ICCM 11v2

ELEVENTH INTERNATIONAL CONFERENCE
ON COMPOSITE MATERIALS
Gold Coast, Queensland, Australia
14th - 18th July 1997
PROCEEDINGS
VOLUME II
FATIGUE, FRACTURE AND
CERAMIC MATRIX COMPOSITES
Editor
Murray L. Scott
AUSTRALIAN COMPOSITE STRUCTURES SOCIETY
WOODHEAD PUBLISHING LIMITED
TABLE OF CONTENTS
Damage Tolerance
Compression After Impact (CAI) Properties of Hat Stiffened CF/PEEK Panels
Fabricated through a Route without Autoclave
Takashi Ishikawa, Masamichi Matsushima
Damage Tolerance of Biomimetic Laminates
Patrick Kim, Toshio Tanimoto
Effect of Loading Parameters on Compression Dominated Fatigue Behavior of
Impact Damaged Composite Laminates
Milan Mitrovic, H. Thomas Hahn, Greg P. Carman
RTM Carbon Composites - Influence of Process Parameters on Compression
Strength After Impact
Florentin Berthet, Pierre Devos, Thierry Ansart
10
17
27
Dynamic Response Behaviour

Free Vibration of Delaminated Composite Plates with Contact Analysis
Meng-Kao Yeh, Ren-Cheng Lai
The Dynamic Response of Inelastic, Delaminated Composite Plates
Todd O. Williams, Frank L. Addessio
Vibration Analysis of Thick Laminated Trapezoidal Plates
C.C. Chen, C. W. Lim, S. Kitipornchai, K.M. Liew
Dynamic Responses of Orthotropic Annular Composite Plates
C.L.D. Huang
The Dynamic Response of a Pultruded Fiberglass Tube Subjected to Lateral
Loading
Louis J. Brunet, James A. Nemes
Vibration Analysis for Composite Cylindrical Shells with a Rectangular Cutout
Young-Shin Lee, Young-Wann Kim
An Analysis of Vibration Damping in Composite Plates Using Complex
Frequency Equations
S. Kalyanasundaram, D.H. Allen, M. Cardew-Hall, A.E. Lowe
Vibrations of a Laminated Composite Beam Elastically Supported at One End
and Carrying a Tip Mass at the Other
T. Yokoyama, Z.M. Yang
36
47
55
64
74
84
94
103
Fatigue
Fatigue Behaviour of Commingled CF/PEEK Composite
Lin Ye, Xiaoxue Diao, Yiu-Wing Mai
Effect of Load Frequency on the Tensile Fatigue Behaviour of Angle-Ply
AS4/PEEK
X.R. Xiao, I. Al-Hmouz
xiii
114
124
Residual Strength of GRP at High Cycle Fatigue

J. Andersons, J. Korsgaard
A Stochastic Cumulative Damage Model for the Fatigue Response of
Laminated Composites
R. Ganesan, S.V. Hoa, S. Zhang, M. El-Karmalawy
Progressive Fatigue Modeling of Laminated Composite Pinned/Bolted
Connections
Mahmood M. Shokrieh, Larry B. Lessard
Prediction of Flexural Fatigue Life for Unidirectional CFRP Laminates
Masayuki Nakada, Takeshi Ishiguro, Yasushi Miyano
Fatigue Behavior of the Carbon/Epoxy and the Carbon/PEEK Composites
Subjected to Low-Energy Impact
Nyan-Hwa Tai, Ming-Chuen Yip, Chih-Ming Tseng, Hui-Chia Yu
Initiation of Free-Edge Delamination in a Composite Laminate Under
Fatigue Loading
Ran Y. Kim, Allan S. Crasto
Fatigue Crack Growth Behavior of a Random Fiber Sheet Molding Compound
Composite
D.R. Atodaria, S.K. Putatunda, P.K. Mallick
Damage Based Fatigue Life Modeling for Brittle Composites
G.M. Newaz, N. Bonora,
Fatigue Damage in Nicalon SiC(f)/Al Composite Wires
Yue Zhuo, Hong Wan, Jin Pan, Deming Yang
135
145
156
167
177
185
194
203
211
Fracture
The Mode-I Fracture Toughness of IM7/LARC-RP46 Composites at High
Temperatures
W.X. Wang, F.G. Yuan, B.D. Potter, R.H. Pater, Y. Takao
Intralaminar Fracture Behaviour and Microstructural Characteristics of
Unidirectional T300/914
Adrian Lowe, Shankar Kalyanasundaram
Rate Effects on Mode I Interlaminar Fracture Toughness in CarbonFibre/Epoxy and Carbon-Fibre/Toughened-Epoxy Composite Laminates
Takayuki Kusaka, Masaki Hojo, Tomoaki Kurokawa, Shojiro Ochiai
Toughness of Marine Composites: Delamination Resistance Under Static
Loading
P. Compston, P.-Y. B. Jar, P.J. Burchill, G.J. Simpson
Improved Interlaminar Fracture Toughness for Vinyl Ester Resin - Fibre Glass
Composites
Peter J. Burchill, Gary J. Simpson
Dynamic Crack Propagation and Perforation of Laminated Composites
Using a Split Hopkinson Pressure Bar
Sylvanus N. Nwosu, Gregory J. Czarnecki
Evaluation of Multiaxial Test Data of UD-Laminate by So-Called "FractureType Strength Criteria" and By Supporting Probabilistic Means
Ralf G. Cuntze
xiv
215
225
234
244
254
263
273
A Singular Hybrid Finite Element Analysis of Interlaminar Crack Problems

Kyohei Kondo, Kazuhiro Yagi
Failure Behavior and Morphology of Hybrid Short-Fibers Reinforced
Composite Under Thermal and Mechanical Loadings
Kh. G. Schmitt-Thomas, Zhen-Guo Yang, R. Malke
Debonding and Fracture Mechanisms of Particulate Filled Polypropylene
I.L. Dubnikova, D.K. Muravin, V.G. Oshmyan
Effects of Pores and Voids on the Interlaminar Delamination Toughness of a
Carbon/Epoxy Composite
Leif E. Asp, Fredrik Brandt
A Statistical Model for Crack Growth in Composite Materials
Biao Wang, Jin Dai, Shanyi Du
Modelling Problems of Damage and Fracture of Notched Fibre Composites
R.A. Dimant, H.R. Shercliff, P.W.R. Beaumont
Improved Fracture Mechanics Predictions for Delamination of Composites
Accounting for Yield Zones
Z. Petrossian, Michael R. Winsom
Damage Mechanics of Reinforcement Cracking in Particle or Short-Fiber
Reinforced Composites
Keiichiro Tohgo, Young-Tae Cho, Hitoshi Ishii
Parallel Computing and Damage in Composite Structures
F. Devries, F. Ln, G. L'Hostis, D. Soulat
A Proper Treatment of Interfacial Cracks in the Presence of Friction
C.T. Sun, W. Qian
An Effective Method for Modeling the Tension Fracture of Composite
Structures
Khanh Trinh, Fu-Kuo Chang
Effect of Strain Rate on Tensile Fracture Behaviour of Fibre Reinforced
Polyamide Composites
M. Todo, K.Takahashi, Ph. Bguelin, H.H. Kausch
Fibre Orientation Effects on Interlaminar Fracture Toughness of Marine
Composites
P. Compston, P.-Y. Jar, P. Davies
Improving of the Fracture Toughness of MG-Alloy Composites Produced by
Powder Metallurgical Technique
P. Abachi, B.L. Mordike, K.U. Kainer
Study on the Mode I Interlaminar Fracture Toughness of Multi-Directional
Laminates
Li Yong, Li Shunling, Xiao Jun, Tao Jie
Experimental Delamination Behavior of Anisotropic Layered Plates with
Different Internal Artificial Delaminations Under Compression Loading
M. Sczepanik-Weinmann, U. Stoll, F.J. Arendts
Effect of Fabric Orientation on Fracture Toughness of Glass Fabric/PC
Composite
J.B. Wang, Y.Q. Sun, S.R. Zheng, X.T. Jing
Interface Modelling in Laminate Composites for Free-Edge Effects Analysis
Mohamed Haboussi, Hlne Dumontet, Jean Louis Billet
xv
290
301
312
322
332
342
353
364
374
387
397
405
415
424
431
438
447
453
Impact
Behavior of Gradient Designed Composite Under Ballistic Impact
Frank K. Ko, Amotz J. Geshury, John W. Song
Hypervelocity Impact Damage to Composite Materials
R.C. Tennyson, G.D. Shortliffe
Results from Testing of the Helicopter Tail Rotor Blade Made of Composite
Laminated Materials After Ballistic Damages
Boko Rauo
Low Velocity Impact Failure Analysis of Sandwich Beam
Ya-Jung Lee, Yau Shyu
Influence of Water Absorption on Delamination Induced by Low-Velocity
Impact and CAI Strength of FRPs
Kenjiro Komai, Kohji Minoshima, Kazuto Tanaka
Low Velocity Impact Damage of Organic Foam Core Sandwich Composites
William J. Craft, Derke R. Hughes, Ajit D. Kelkar
Perforation Characteristics Prediction of Multi-Layered Composite Plates
Subjected to High Velocity Impact
Hideaki Kasano, Kenichi Abe
Finite Element Analysis of the Ballistic Impact Response of Composite Helmet
Materials
J. Van Hoof, M. J. Worswick, P.V. Straznicky, M. Bolduc, S. Tylko
Perforation Characteristics of Fabrics Made of High Strength PE Fibers
K. Takahashi, H. Komatsu
Using a Homogenization Procedure for Prediction of Material Properties and the
Impact Response of Unidirectional Composite
A.D. Resnyansky, E.I. Romensky
Impact Damage and Failure Mechanisms in Structure Relevant
Composite Specimens
J.F.M. Wiggenraad, L.C. Ubels
Impact Damage Growth and Failure of Carbon-Fibre Reinforced Plastic
Skin-Stringer Panels
Emile Greenhalgh, Sunil Singh, Donald Roberts
Simulation of Impact Deformation, Damage & Fracture in
Composite Laminates
Da-Xhi Jiang, Wei Shen, Xing-Ye Wang
464
474
485
495
503
513
522
532
542
552
562
573
583
Ceramic Matrix Composites

Fabrication and Mechanical Properties of Glass Matrix Composites with
Different Fibre Architectures
Heinrich Kern, Volker Winkler, Aldo R. Boccaccini
Thermoelectric Properties of B4C-Based Composite Ceramics Prepared by
Arc-Melting
Takashi Goto, Jianhui Li, Toshio Hirai
Formation and Characterisation of Fibre-Matrix Interfaces in Continuous Fibre
Reinforced Silicon Carbide
Pamela M. Farries, Jean-Bernard Veyret
xvi
593
603
613
Structural Analysis of Ceramic-Matrix Composites

Alain Burr, Francois Hild, Frederick A. Leckie
Failure Mechanism of a Multi-Layer Coating for the Oxidation Protection of
C/C Composites
Laifei Cheng, Litong Zhang, Yongdong Xu
Subsurface Damage Measurement on Carbon Carbon Composite Materials in
Braking Applications
Yves Remond, Christiane Wagner
Microstructural Characterisation of 3D Woven SiC/SiC-based Composites after
Tensile Testing at Room and Elevated Temperature in Different Atmospheres
Ian J. Davies, Takashi Ishikawa, Masaki Shibuya, Tetsuro Hirokawa
The Mechanical Properties of 3D Woven Carbon Fibers Reinforced Ceramic
Matrix Composites
Wenwei Zheng, Xingye Wang, Jiayu Xiao, Fengrong Liu, Dazhi Jiang, Zhaohui Chen
Effects of Temperature and Layup on the Mixed-Mode Delamination Toughness
of a Ceramic Matrix Composite Material
Jonathon J. Polaha, Tor W. Sherwood, Barry D. Davidson
Toughening of Silicon Nitride Matrix Composites by the Addition of Si-C-N
Nanometer Particles
Xingui Zhou, Changrui Zhang, Zhaohui Chen, Kai Xie, Anchen Zhou
Non Linear Behavior of Two Woven Ceramic Matrix Composites Under Cyclic
Loading
Jean-Marie Morvan, Stphane Baste
Fracture in Off-Axis and Angle-Ply Fiber Reinforced Ceramics
Alex S. Selvarathinam, Y.J. Weitsman
The Research of Self-Reinforced HP Silicon Nitride
Li Fengmei
Microstructural Study of the Sintered Composite Composed of TiN and Ni
Shujie Li, Yuping Li, J.H. Maas, J. Boeijsma, B.H. Kolster
Structural Evolution of C/C-SiC Composites Through the Main Production
Steps on the Liquid Silicon Infiltration (LSI) Route
Herbert Mucha, Akira Kamiya, Bernhard Weilage
The Fabrication of Two Dimensional Carbon Fiber Reinforced Alumina
Composites by Pressure Infiltration
Hsien-Kuang Liu, Bor-Horng Lin
Application of Preceramic Polymer For Shape-Forming of Multiphase SiC/Si3N4
Ceramic Parts
Yongqing Li, Zhaohui Chen, Changrui Zhang, Haifeng Cheng, Ancheng Zhou,
Xiaoheng Bao
Key Components of High Temperature Stable High Oxidation Resistant
Ceramic Matrix Composites of SiC/SiC Type Working Out and Technological
Ways of their Making
A.M. Tsirlin, V.G. Gerlivanov, N.A. Popova, Yu E. Pronin, E.K. Florina, S.P. Gubin
Microstructure of the TZP Particulate Composites
Marek Faryna, Zbigniew Pedzich, Lidia Litynska, Krzysztof Haberko
FEM Analysis of Micro-Indentation Test for C/C Composites
H. Serizawa, S. Sato, H. Tsunakawa, A. Kohyama
xvii
621
630
638
646
658
665
675
682
694
703
709
717
725
735
741
751
759
Preparation and Properties of 3D-Carbon Fiber Woven Preform Reinforced SiC

and Al2O3 Hybrid Matrix Composites
Jiayu Xiao, Fengrong Liu, Wenwei Zheng, Xingye Wang, Dazhi Jiang, Zhaohui Chen
Continuous Fibre Reinforced Composites with Mullite Matrix Derived from
Colloidal Precursors
Jaili Wu, Frank R. Jones, Peter F. James
769
775
New and Non-Traditional Materials

Mechanical Properties of Spider Silk
Sueo Kawabata, Frank K. Ko, Mari Inoue, Masako Niwa, John W. Song
Natural-Fibre-Mat-Reinforced Thermoplastic Composites Based on
Flax Fibres and Polypropylene
S. Garkhail, R. Heijenrath, M. van der Oever, H. Bos, T. Peijs
Mechanical Properties of Woven Fabric Carbon/Magnesium-Composites
O. ttinger, W. Schff, C. Hausmann, T. Heyne, R.F. Singer
Modelling Stress Transfer in a Wood Composite
Rowan Paton
Microcellular Graphitic Foam Processing
Kristen M. Kearns, David P. Anderson
Development and Characterisation of a New High Damping Composite:
Mg2Si/Mg
Christine Mayencourt, Robert Schaller
Interface Study of Composite Bands Formed by Powders Plating and Rolling
(PPR)
Sun Yong, Zhang Shuhong, Zhang Daiming, Qian Tiancai, Gao Yun, Li Bin
Mechanical Performance of Compression Molded and Injection Molded Blends
of LaRC Polyimide with Polyetherimide (PEI)
David Olivero, Donald Radford, Sarah Hummel, Charles Smith, Betty Tung
Engineering Composite Electret Materials
L.S. Pinchuk, V.A. Goldade, I.M. Vertyachikh, V.N. Kestelman
Multi-Axial Warp Knitted Layers - a Textile for Reinforcing Concrete
Gerd Franzke, Peter Offermann, Thomas Bischoff, Burkhard Wulfhorst
Processing and Electrical Properties of PTCR Composite Materials
Z.X. Xiong, K.Z. Baba-Kishi
xviii
785
794
804
813
825
833
845
851
859
870
881
Proceedings of ICCM11, Gold Coast, Australia, 14th-18th July 1997
COMPRESSION AFTER IMPACT (CAI) PROPERTIES

OF HAT STIFFENED CF/PEEK PANELS
FABRICATED THROUGH A ROUTE WITHOUT
AUTOCLAVE
Takashi Ishikawa and Masamichi Matsushima
Airframe Division, National Aerospace Laboratory
6-13-1 Ohsawa, Mitaka, Tokyo 181, Japan
SUMMARY: Compression after impact (CAI) properties in composite structure levels were
experimentally obtained and examined. A remarkable feature in the fabrication technology of
the test piece panels was the fusion bonding process without autoclave. This route was
pursued for possible low cost fabrication methodology of carbon fiber (CF)/thermoplastic
composites exhibiting an excellent damage tolerance property. Used test specimens were
stiffened panels with hat shaped stiffeners regarded as optimal for CF/thermoplastic,
CF/PEEK being chosen here. Two kinds of panels in size were tested in compression loading
after drop weight impact for creating delamination. The results are compared with the
previous similar CAI tests for T-stiffened panels of CF/PEEK where the present results
exhibit higher CAI values. A potential of this fusion bonding technology without autoclave is
indicated. Numerical analysis for predicting initial buckling stress was also conducted with a
favorable agreement with experiments.
KEYWORDS: experiments, impact, delamination, thermoplastic composites, fusion
bonding, ultrasonic C-Scan, buckling, finite element analysis
INTRODUCTION
Damage tolerance properties of composites are considered to be important design criticals in
their application to aircraft structures. Particularly, importance of compression after impact
(CAI) values have been recognized in strength design of composite wing or fuselage in recent
years. Therefore, NAL of JAPAN has been conducting a long term research project [1] for
development of tough composite structures to aircraft component. A chosen material system
in this project was CF/PEEK. After basic material properties including compression after
impact (CAI) strengths for NASA type thick plates [2] were obtained, stiffened panels with
four T-shaped stiffeners were tested [3,4] as structural level CAI evaluations. Based on such
achievement, two model wing box structures of CF/PEEK were built and tested [1,5] in
fatigue loading followed by residual static strength tests. The results of these boxes exhibited
high delamination propagation resistance and excellent strength properties. In summary, a
possibility of 35 % weight reduction by CF/PEEK from a baseline metal structure was
verified [1] even in a conservative estimate.
II - 1
Volume II: Fatigue, Fracture and Ceramic Matrix Composites
However, one major disadvantage of CF/PEEK was also

revealed through this work. Since it requires high
temperature and pressure during fabrication procedure, its
autoclave processing may lead to high cost because of
expensive autoclave itself and tooling. This fact could be
an obstacle for rapid spreading of CF/PEEK material into
aerospace composite community. Based on such a
background, a fabrication technique without an autoclave
has been pursued by NAL and Fuji Heavy Industries Inc.
(FHI) in JAPAN for realizing low cost thermoplastic
composite structures. A possible solution might be nontraditional hot-pressing where an appropriate shape of
stiffeners is also a key point of concern. The present
paper describes recent results of this research program. A
detailed description of fabrication process [6] will be
minimized and the focus will be placed on presentation
and discussion of experimental CAI properties. Numerical
analysis of them with locally delaminated hat stiffeners at
skin-stiffener interface was also conducted using a
commercially available finite element code.
DESCRIPTION OF TEST SPECIMENS
As stated before, material employed was CF/PEEK (AS4/
APC-2). Shape and dimensions of a smaller test panel
stiffened by single hat are shown in Fig.1, which is
considered to be a prototype of larger panels. The larger
panel dimensions stiffened by three hats are shown in
Fig.2. A hat shaped stiffener was chosen for its
compatibility with secondary fusion bonding to a flat skin
and anti-buckling properties. A key point of the fusion
bonding technique without autoclave is a development of
a mold-platen containing line heaters.
At first, a
technique with PEEK film was established and two
smaller panels were fabricated with this route, designated
as CPA03 and 04. A problem of non-uniform thickness
distribution in skin-stiffener attachment flanges remained
unsolved in this phase probably due to re-softening
mentioned later. Next, several thermoplastic films with
lower melting point than PEEK and methods for fusion
promotion into CF/PEEK substrate structure were tried in
their combinations.
Some screening interlaminar
toughness and strength tests were conducted and a
combination of PEI and pre-fusion to a substrate was
selected [6]. A benefit of this technique is that resoftening and deformation in the previously formed
region can be minimized during fusion bonding. One
larger panel designated as CPB02 was fabricated through
this method, whereas a larger panel CPB01 was also
II - 2
Fig.1: Single hat panel: shape

and dimensions
Fig. 2: Three hat panel: shape

and dimensions
fabricated through an established PEEK film with autoclave technique for comparison.
Stacking sequences were (45/-45/-45/45/ 0/90)sym.:12 plies and (45/-45/-45/45/0/ 90)sym.:18
plies for skin and stiffener, respectively, regardless of panel sizes. Aluminum end platens
were adhered with a epoxy potting agent for compression tests for both loading edges of the
panel.
OUTLINE OF IMPACT TESTS AND FORMED DELAMINATION
All the panels were subjected to drop weight impact tests on a skin-stiffener bonding region
with the following normalized energy levels (Unit: J/mm) by thickness at a impact point:
CPA03= 2, CPA04= 4, CPB01=6.7 (with Autoclave) , CPB02=6.7 (with PEI film). These
levels were determined so as to create barely visible indentation according to the previous
results [3,4] and preliminary impacts. The reason why the larger panel required greater energy
is supposed that some amount of impact energy was consumed in elastic deformation under a
longer supporting span. Impact points were located on a mid-line parallel to the loading ends.
The panels were supported on a very stiff steel table and hold by C-cramps at impact as
shown in a photo of Fig. 3.
The panels were C-scanned after impact and projection area of delamination was measured
by a ultrasonic C-scanner. A typical delamination image taken through a C-scanner, SDS
3300 by Krautkraemer Japan Co., is shown in Fig. 4 for CPA04 panel. It can be seen that
delamination occurred almost along fusion bonding interface in this case.
Fig. 3: Photo of impact test by drop weight

impactor
Fig.4: Delamination image for CPA04
Although an indication is omitted here, delamination also took place within skin and stiffener
in cases of CPB01 and 02, which is similar to one shot thick plate cases [2]. Relationships
between normalized impact energy and the delamination area in projection are indicated in
Fig. 5 with the previous T-stiffened panel results [3,4]. A general tendency of the present
impact results seems to be reasonable, although larger panels show smaller areas due to the
II - 3
above reason. It can be understood that the present fusion bonding technique without
autoclave worked out well from a standpoint of impact delamination.
OUTLINE OF COMPRESSION TESTS AFTER
IMPACT AND DISCUSSIONS
3.5
3
w/o Autoclave, 1 Hat

by Autoclave, 3 Hats
w/o Autoclave, 3 Hats (PEI)
Ref.3: by Autoclave, 4 T
Compression tests were conducted after 12 strain

2.5
gages were glued onto the smaller panels. For the
2
larger panes, 88 strain gages were used whereas a
1.5
description of results is omitted in this paper because
tests are undergoing now. For the smaller panel, a
1
strain gage map is shown in Fig.6 where even
0.5
number gages were glued back to back of odd ones
0
on a stiffener side with some exceptions. A gage #3
0
2
4
6
8
10
was placed on an impact indentation. A number of
Normalized Impact Energy (J/mm)
Fig.5: Relationships between
gages, 12, was limited by 16 channel capacity of a
normalized impact energy and
data logger, . Thus, these strains, a load and 3 points
delamination area
of deflections were recorded into this data-logger,
DAA 100A by Kyowa Co. JAPAN. Loading edge
conditions were close to clamped due to aluminum end platens. Stainless steel tubes with
sharpened slits acting as knife edges were installed along both side edges so that simply
supported conditions were approximately realized. Compression load was introduced by
Instron 1128 screw driven machine directly through a glued platen. A photo taken during a
compression test is shown in Fig.7.
Fig. 6: Strain gage map for CPA panel
II - 4
Fig. 7: Photo of compression test for CPA
400
400
300
300
200 < Prdc. Bkl. Str.
200
100
Strain at #1
Strain at #2
Strain at #3
Strain at #4
Exp.Bkl.Str.
0 4
4
-1.5 10 -1 10
Strain at #1
Strain at #2
Strain at #3
Strain at #4
100
-5000
5000
-6
Strain (10 )
Fig. 8: Stress-strain behavior for panel

CPA04 (4J/mm)
Stress-strain behavior for the case where impact
delamination affects the final strength, CPA04
with 4 J/mm, is shown in Fig.8.
Note that
delamination of this panel was already shown in
Fig. 4. In Fig.8, a back-to-back pair of strain
gages, #3 and #4 on the delamination, clearly
indicates the behavior of delamination buckling
far prior to the final failure. Such behavior is
basically similar to that of a CF/PEEK Tstiffened panel in Ref. 4. It should be noted that
an inflection point of stress-average strain curve
for this pair provides an experimental initial
buckling stress, being determined as 185MPa for
CPA04. A comparison of this value with a
numerically predicted initial buckling stress will
be discussed later.
For CPA03 panel with 2J/mm impact, a
delamination area measured by C-scan was
1/50of CPA04.
Therefore, its stress-strain
behavior shown in Fig. 9 does not indicate any
symptom of delamination buckling before the
final failure. All curves are almost straight up to
the failure and coincide with each other.
0
4
4
-1.5 10 -1 10
-5000
0
-6
Strain (10 )
5000
Fig. 9: Stress-strain behavior for panel

CPA03 (2J/mm)
500
400
300
200
100
0
0
Single Hat: w/o Autoclave

Ref.4 (4T: by Autoclave FY89)
Ref.4 (4T: by Autoclave FY88)
10
Normalized Impact Energy (J/mm)
Fig. 10: CAI strengths vs.

impact energy
CAI strengths of the smaller hat stiffened panels without autoclave fabrication are shown in
Fig.10 with comparison data of CF/PEEK panels stiffened by four T-shaped stiffeners and
fabricated by autoclave technology in FY 1988 and 1989 [3,4]. Although single stiffener
panels exhibit a disadvantage in terms of CAI strength comparing with multi-stiffener panels
[4], the present hat-stiffened panels show excellent CAI strengths. If we take a note for 4
II - 5
J/mm data of 344MPa, the present fusion bonding quality without autoclave can be regarded
as excellent.
Fig. 11: Final Failure Mode of CPA 04

panel fabricated without autoclave and
impacted with 4J/mm
Fig. 12: Final Failure Mode of CPA 03 panel

fabricated without autoclave and
impacted with 2J/mm
The reason why even a strength at 2 J/mm (CPA03) exceeds no impact strength of 4-T panel
can be ascribed to a good anti-buckling property of hat stiffeners and a small delamination
area by virtue of an excellent fusion bonding quality without autoclave.
Final failure modes of these panels are indicated in photos of Fig.11 and 12 taken from the
stiffener side for panels CPA04 and 03, respectively. In Fig.11 with larger delamination area,
a transverse failure band formation to the introduced loads can be observed which was
observed in flat plate CAI tests [7] without any exception. Although this failure band
formation can be well correlated experimentally with the transverse delamination penetration
in the case of flat plates, ambiguous mechanism remains unexplained in the detail in the
present stiffened panels. At least, the delamination buckling behavior must be a trigger of the
final failure in the present case. In Fig.12 with a very small size of delamination, the final
failure occurred along the loading end platen and it was not influenced by delamination. This
finding is in accordance with the stress-strain behavior indicated in Fig.9. Therefore, this CAI
value of 426 MPa is estimated to be close to the compression strength of the present panel
without delamination. A failure mode similarity to the NASA or SACMA plates with small
delamination radius was verified in Fig.12.
CAI results for three stiffener panels are being obtained now and they would be demonstrated
at the presentation. Thus, a description in the current text is skipped.
PREDICTION OF BUCKLING BEHAVIOR BY FINITE ELEMENT ANALYSIS

Numerical analysis by a finite element method was conducted with simplified delamination
location between skin and stiffener. A commercially supplied code, NISA-II, capable of
handling laminate elements was used here. The used elements were 4 or 8 nodes
isoparametric shell elements composed of assigned stacking sequence of unidirectional plies.
II - 6
Outside of delaminated regions, skin-stiffener attachment flange was merely modeled as a

thicker plate with total number of plies of both constituents where the thickness of fusion
bonding film was neglected. This treatment implies that a slight local bending-stretching
coupling may exist according to an unsymmetrical stacking sequence. In order to idealize a
delaminated panel, a single delamination was assumed to exist between a skin and a hat
stiffener. A projection shape of delamination was simplified into a rectangle nevertheless the
true shape was the one as shown in Fig. 4. An important point of difference from the reality
may be the following: Delamination edges along the fusion bonded region were considered to
be free against opening deflection whereas narrow but tight bonded area remained along these
edges according to a C-scan image. Boundary conditions along the loaded and the side edges
were assumed to be fixed and simply supported, respectively, under consideration of potting
material as discussed in Ref. 3 and tubular knife edges. Buckling under flat compression
loading was simulated in the following manner. A static analysis was done first under the
same specified end shortening at all the nodes along the loading edge and reaction forces at
these nodes were obtained. Linear buckling analysis was done next using corresponding
nodal forces along the loading line.
The employed elastic moduli for unidirectional CF/PEEK lamina are as follows: EL = 117.3
GPa, ET = 10.3 GPa, GLT = 4.62 GPa, L = 0.38, and TT = 0.5 which are compatible with
the experimental values adopted in Ref. 3. Ply thickness values for skin and stiffener regions
were determined as 0.125mm and 0.13mm, respectively, from each measured thickness
divided by ply number. An example of a numerical deformation pattern in the buckled state is
shown in Fig.13 where absolute value of deflection has no meaning according to the nature of
the eigenvalue problem. Because the delaminated area are assumed to be a rectangle as stated
above, skin deformation pattern looks more square than reality. Initial buckling stress of 201
MPa was obtained and plotted by a symbol of < in Fig.8, whereas experimental buckling
stress of 185 MPa determined through back-to-back strain gages was also plotted by on
the left axis in Fig.8. Note that stresses are merely plotted on the vertical axis. A correlation
between these predicted and experimental buckling stress is shown to be very favorable.
However, because the shape and area of the delaminated region, and geometrical structure of
delamination were simplified very much in a finite element modeling, this good agreement is
not considered to be a perfect description of this sort of complicated problem but some
coincidence makes the agreement favorable. A discussion about this point may be
summarized conservatively as follows: The present modeling of delaminated stiffened
structure can predict a trend in deformation pattern of initial buckling and an approximate
value of its initial buckling stress.
Fig. 13: Deformation Pattern in a Buckled Single Hat Panel under Flat Compression
II - 7
CURRENT CONCLUSIONS
A fabrication technology of stiffened panels using secondary fusion bonding without
autoclave was pursued with a considerable success. Two key factors in the technology consist
of a development of a mold with embedded heaters and adaptation of fusion bonding film to a
thermoplastic composite substrate. A goal of this technology is to realize an affordable
CF/thermoplastic structure for future civil transport aircraft. A stiffener shape of a hat was
chosen so as to optimize buckling load and compatibility for fusion bonding. Single and three
hat stiffener panels were fabricated of CF/PEEK (AS4/APC-2) material system. Both panels
exhibited similar or superior delamination resistance under impact loading to panels of CF/
PEEK containing four T-stiffeners which were fabricated and tested previously. CAI
strengths of the present single stiffener panels are greater than those obtained for the 4-T
panels. Thus, the present fusion bonding technology is proven as a potential fabrication
method for an affordable thermoplastic composite component. Some part of the reason why
the present no impact strength is higher than that of the 4-T panels can be attributed to a better
buckling resistance of hat stiffeners than Ts. Numerical predictions by finite element
analysis were conducted for a simplified mechanical model of a delaminated panel. In terms
of an initial buckling stress, correlation between the predicted and experimental values was
favorable. Further effort will be required for complete prediction of the behavior in the
postbuckling region.
Because a baseline technology for low cost thermoplastic composite structures is established,
NAL of JAPAN is seeking to find an application route of heat-resistant thermoplastic
composites to future high speed commercial transport where weight reduction is a crucial
technological barrier. A candidate resin there is not PEEK but PIXA, semi-crystalline
polymer by Mitsui-Toatsu Co. JAPAN. Fortunately, the present fusion bonding technology
for CF/PEEK can be directly transferred to CF/PIXA material system. Structural test results
for this new system will be presented in the future.
ACKNOWLEDGEMENT
The authors owed a great deal of the present work to those who dedicated themselves to
fabrication technology development in Fuji Heavy Industries LTD, e.g., Mr. Shigeru Machida
and Hiroyuki Nakamura. The authors express their sincere gratitude to Mr. Kenji Matsui of
JAMCO Co. LTD, Japan, (Student of Nihon University at moment) for his effort to conduct
most of experiments. They also wish to thank Mr. Hee-Jun Kim of a student of Seoul
National University, and Dr. Ik-Hyeon Choi of a senior researcher of KARI (Korean
Aerospace Research Institute) for their assistance in some parts of the experiments. Their stay
in JAPAN is supported by Japan-Korea Foundation and JISTEC. Fabrication expense of the
panel specimens was funded by STA of Japanese Government for Future Supersonic
Transport Technology Program.
II - 8
REFERENCES
1.
Ishikawa, T., Hayashi, Y., Sugimoto, S., Matsushima, M., and Amaoka, K.,
Development and Test Results of Full CF/PEEK (APC-2) Horizontal Stabilizer Models
- Basis for SST Structure, Proceedings 1st AIAA Aircraft Engineering , Technology, and
Operations Congress, Los Angeles, CA, USA, September 19-21, 1995, also AIAA Paper
95-3931
2.
Ishikawa, T., Matsushima, M., Hayashi, Y., and Noguchi, T., Comparison of Tensile
and Compressive Properties of Some Structural Elements between Carbon/PEEK and
Carbon/Epoxy Composites, Proceedings 5th Japan-US Conference on Composite
Materials, Tama-City, Tokyo, Japan, June 24-27, 1990, Kobayashi, A., Ed., pp.445-452.
3.
Ishikawa, T., Matsushima, M., and Hayashi, Y., Improved Correlation of Predicted and
Experimental Initial Buckling Stresses of Composite Stiffened Panels, Composite
Structures, Vol.26, No.1, 1993. pp.25-38.
4.
Ishikawa, T., Hayashi, Y., and Matsushima, M., Compression after Impact Strengths of
Stiffened Panels Made of CF/PEEK and CF/Epoxy (in Japanese), Journal of the Japan
Society for Aeronautical and Space Sciences, Vol. 42, No.484, 1994, pp. 319-328
5.
Ishikawa, T., Hayashi, Y., Sugimoto, S., and Matsushima, M., Development of Full CF/
PEEK (APC-2) Horizontal Stabilizer Models and Their Strength Test Results,
Proceedings 10th International Conference on Composite Materials, Whistler, British
Columbia, Canada, August 14-18, 1995, Vol.III: Processing and Manufacturing,
Poursartip, A.., Street, K.N., Eds., pp. 693-700.
6.
Machida, S., Nakamura, H., Kawakami, M., Akiba, K., and Ishikawa, T., Development
of Fusion Bonding Technology of Thermoplastic Composites with Focused Heating
Method (in Japanese), Proceedings 37th JSASS/JSME Structures Conference, Fukuokashi, Fukuoka, Japan, July 12-14, 1995, pp.249-252.
7.
Ishikawa, T., Sugimoto, S., Matsushima, M., and Hayashi, Y., Some Experimental
Findings in Compression-after-Impact (CAI) Tests of CF/ PEEK (APC-2) and
Conventional CF/Epoxy Flat Plates, Composite Science and Technology, Vol.55, No.4,
1995. pp.349-363.
8.
Engineering Mechanics Research Corporation, Users Manuals of NISA-II, Version 91,

Troy, MI, April 1991.
II - 9
DAMAGE TOLERANCE OF BIOMIMETIC

LAMINATES
Patrick Kim1 and Toshio Tanimoto2
1
School of Engineering and

Department of Materials Science and Ceramic Technology
Shonan Institute of Technology, 1-1-25 Tsujido Nishikaigan, 251 Fujisawa, Kanagawa, Japan
2
SUMMARY: The strength of a biomimetic CFRP laminate with a stacking sequence similar
to that found in many animal hard tissues is examined. The static strength of the biomimetic
laminate with holes of different diameters is compared to that of a standard quasi-isotropic
and a cross-ply laminate. The first two have similar elastic moduli and strengths without a
hole. The ratio of notched to unnotched strength is also nearly the same, while the cross-ply
laminate has the lowest ratio. This difference is explained in terms of the extent of an
equivalent damage zone, which depends on the laminates architecture. The implications of
the findings for the design of damage-tolerant laminates are discussed.
KEYWORDS: biomimetic composite, damage tolerance, notched sample, strength ratio,

fracture, fiber orientation distribution
INTRODUCTION
Biological materials derive remarkable mechanical properties from their composite nature and
reinforcement architecture. Many biological composites are fiber-reinforced laminates
similar to plywood, with a fiber orientations that varies through their thickness in relatively
constant angle steps [1-3]. As these composites often have a protective or structural function,
such as in insect cuticles, it has been assumed that this fiber architecture may give them a
particularly good strength or damage tolerance. If this should be the case, a biomimetic
microstructure would be a way to make improved composites [3, 4]. Tests have shown that
the quasi-static moduli and unnotched strength of a biomimetic and a conventional quasiisotropic laminate are nearly the same. Gunderson and Whitney however showed that the
impact resistance of the biomimetic laminate, in terms of energy and force to initiate
delamination, was better than that of a standard quasi-isotropic laminate [3]. To examine the
post-impact performance, this paper presents results on the quasi-static behavior of a
biomimetic laminate including some damage. It examines the sensitivity of a biomimetic
laminate to the presence of a damaged area, here modeled as an open hole, and compares the
residual strength and the fracture patterns to those of conventional quasi-isotropic and a crossply laminates.
THEORETICAL BACKGROUND
Various authors, including Neville [1], Giraud et al. [2], and Gunderson and Whitney [3] have
shown examples of biological laminates with a fairly regular angle difference of about 78
II - 10
between successive plies (Fig. 1). This lamination scheme, selected for the biomimetic
laminate in the present study, gives a double helical lay-up with angle steps of 24 for each
helix. For a 16-ply laminate, the outermost plies are oriented at 90 with respect to each
other, and the laminate is nearly quasi-isotropic.
The damage tolerance as related to the ratio of strength of a laminate with and without a hole
was analyzed by the Damage Zone Criterion (DZC) approach developed by Eriksson and
Aronsson [5]. The actual, two-dimensional damage in the laminate is modeled as a fictitious
crack extending from the hole perpendicularly to the applied load. The residual stress acting
on this crack of length d representing the damage zone is assumed to be constant. The
stresses ahead of this crack correspond to a linear elastic stress distribution for an anisotropic
plate:
r = o, x d
a
2 + R 2 + 3R 4 (KT 3)[5R 6 7R8 ] , x d

2
R = D 2x
r =
(1)
where a is the applied stress, r is the stress on the reduced section, o is the strength of the
laminate without hole, and D is the diameter of the hole. The stress concentration factor KT
for an infinitely wide plate depends on the laminates elastic properties:
2
2
A11 A22 A12 + A11 A22 A12
K = 1+
A22
2A66
.5
(2)
where Aij are the laminate stiffnesses. The elastic properties of the laminates were calculated
using classical lamination theory [6]. A correction was made for the finite width of the
samples. This however gives a negligible calculated difference for the sample dimensions
used here and is thus not absolutely necessary. The integral of the stresses on the net section
must equal the applied load:
w/2
a wt
= otd + o + r r
2
D / 2+ d
x = D / 2 +d
)tdx
(3)
Fig. 1: Lay-up of the biomimetic composite laminate

The critical damage zone size d* can be calculated from this equation at the ultimate load N
of the sample with a hole [5]. This damage zone size is taken as an indicator of the damage
II - 11
tolerance of the laminate. A larger damage zone at failure would point to a higher damage
tolerance, as it gives a higher ratio of strength with a hole to strength without a hole.
EXPERIMENTAL
The samples were made from unidirectional (UD) AT400/EP carbon fiber-reinforced epoxy
prepregs (Asahi Kasei Corporation, Japan). The cured material had a longitudinal modulus of
117Gpa and a transverse modulus of about 11GPa. Laminated plates consisting of 16 plies of
prepreg were cured in an autoclave at a pressure of 6bar and 130C for two hours. The
samples were slow cooled under pressure by turning off the heat in the autoclave, in order to
minimize internal stresses. The resulting laminates had a thickness of 1.98mm, 0.06mm.
The cured plates were cut into 20mm wide samples with a water-cooled rotary diamond saw.
The samples were then left to stabilize at room temperature and humidity for at least one
week. The lay-ups of the samples for this study are given in Table 1. Samples were tested
without hole and with a central 4mm or 5mm diameter hole corresponding to a ratio of hole
diameter to sample width of 1/5 and 1/4. The CFRP samples were clamped between two
5mm thick acrylic plates during drilling, in order to keep damage around the hole during
machining to a minimum. The samples without and with a hole had a gauge length of 120mm
and 80mm, respectively. Tapered GFRP tabs were bonded to the sample ends to reduce the
likelihood of damage or slippage in the grips.
The tensile tests were conducted monotonically up to failure at a cross-head velocity of
1mm/min. For the samples without holes the extension was measured by strain gauges
bonded to both sides of the samples. Strength results were accepted only for samples in
which failure damage did not extend into the gripping areas. For the samples with holes, the
load-extension curve and the peak force were recorded. The stress was calculated based on
the full section of the sample.
Table 1: Definition of sample types
Sample Lam. type
Laminate lay-up
CP
Cross-ply
(0/ 90 )4s
QI
Quasi-isotropic
(0 / 45 / 90/ -45 )2s
BM
Biomimetic
(0 /-78 /24 /-54 /48 /-30 /72 /-6 // -84 /18 /-60 /42 /-36 /66 /-12
/90 )
RESULTS AND DISCUSSION

The biomimetic and the standard quasi-isotropic laminates had nearly the same elastic
modulus, while the cross-ply laminate was much stiffer (Table 2). The experimental moduli
of the plain laminates are all within 5% of the predictions. The symmetric quasi-isotropic and
cross-ply laminates had no extension/flexural coupling. The transverse strains on opposing
surfaces of the biomimetic laminate, however, differed significantly, as the axial stress led to
a bending about the longitudinal axis.
II - 12
Table 2: Elastic properties of the unnotched laminates (calculated values in parentheses)

Sample
E1 [GPa]
o [MPa]
CP
66.47 (64.4)
0.065
901
QI
48.84 (50.8)
0.335
676
BM
49.38 (52.0) / 47.98 (52.0)*
0.320 / 0.333*
654 / 619 *
* in two normal directions, parallel to surface fibers
The unnotched laminates had a linear stress strain curve until failure. All samples failed on
the free length, away from the grips. The strength of the biomimetic samples was 3-8% lower
than that of the quasi-isotropic laminate. The cross-ply samples strength was about one third
higher, as a result of the higher proportion of strong 0 plies. The fractured zones had
different appearances in all three samples. The cross-ply samples broke along a straight,
slightly jagged line, perpendicular to the samples axis, and damage was localized in the
vicinity of the fracture. On the other hand, the damaged area in the quasi-isotropic and
biomimetic laminates extended over a length corresponding to more than twice the tensile
samples width, with extensive delamination and splitting. The biomimetic laminate samples
also failed in compression in the longer half of the broken sample, upon recoil at tensile
failure. This happened for none of the two other laminate types, and may be due to a lower
apparent compressive strength of the unsymmetric biomimetic laminate.
The notched strength, the notched-to-unnotched strength ratio, and the equivalent damage
zone size calculated from the strength ratio are reported in Table 3, for laminates with two
different hole diameters. The strengths of the biomimetic and quasi-isotropic laminates with a
4mm hole were about the same. With a 5mm hole, the biomimetic laminate had a clearly
higher strength. Its strength ratio with a 4mm or 5mm hole was slightly over 4% higher
and nearly 15% higher, respectively, than that of the quasi-isotropic laminate. Its strength
was also less variable than that of the latter. Although the cross-ply laminate had a much
higher unnotched strength than the others, its notched strength was just 11% higher than that
of the quasi-isotropic and biomimetic laminates, and its strength ratio was almost 20% lower
for a 4mm hole. From these results, it would seem that the biomimetic laminate is slightly
more tolerant to damage in the form of a hole than the quasi-isotropic laminate, and that this
characteristic is more pronounced for larger hole diameters. Further tests are planned with
other hole sizes, to verify this.
Table 3: Failure loads of the laminates with holes and calculated critical length of the
representative damage zone. Standard deviation given in parentheses
Sample
D [mm]
N [MPa]
N /o
d*
CP
4.0
459.1 (14.1)
0.510
0.45
QI
4.0
5.0
407.3 (49.3)
351.1 (26.4)
0.603
0.519
0.89
0.72
BM
4.0
5.0
399.8 (19.7)
379.4 (18.1)
0.628
0.596
1.01
1.13
The calculated equivalent damage zone d* at failure for the samples with the 4mm hole was
about 13% larger in the biomimetic than in the standard quasi-isotropic laminate. For a
5mm hole, it was over 50% larger. The size of d* for the cross-ply laminate was by far the
II - 13
lowest, at less than half the value for the biomimetic laminate. The higher value of d*
indicates a more extensive lateral propagation of the damage before failure of the sample.
The dependence of the extent of the damage on the laminate lay-up has been modeled by
Chang et al. [6]. The present results concord with the analytical results of that group for the
cross-ply and quasi-isotropic laminates: whereas the cross-ply samples failed suddenly,
without prior sign, the development of damage in the quasi-isotropic samples was signaled by
sharp, high-pitch sounds as failure approached. Although the biomimetic laminate had a
similar calculated extent of damage at impending failure, d*, as the quasi-isotropic laminate,
its different reinforcement geometry is expected to lead to a different spatial distribution of
the damage. The broader distribution of angles between plies and the loading direction is
expected to lead to a wider range of interlaminar shear stresses near the edge and the hole, and
thus to a more progressive evolution of damage. This will be investigated in the future. The
stress distribution in the reduced section calculated by the DZC approach for a hole of 4mm
is shown in Fig. 2 for the three sample types. The distributions for the quasi-isotropic and
biomimetic laminates are nearly the same, with the former having a higher stress in the
damage zone, corresponding to the higher unnotched strength.
Fig. 3 illustrates the fracture observed in the samples with a 4mm hole. The pictures are also
fairly representative of the different fracture patterns observed in the samples with the 5mm
hole. The notched cross-ply laminate, like the unnotched sample, broke along a straight line,
and the damage, involving mainly fiber breakage, was very localized. The locations of the
fracture in the two legs of the reduced section were somewhat staggered. In the quasiisotropic sample, the fracture ran at 45 from the hole, leading to a 90 V-shape. This was
accompanied by delamination exposing mainly 45 layers and some loose debris of pulled-out
delaminated layers. When the surface fibers were parallel to the samples axis, extensive
splitting occurred, while some spalling occurred for 0 surface layers. In the biomimetic
laminate, the angle of the V-shaped fracture was larger, and much less splitting was observed
in the surface layer with longitudinal fibers. Delamination exposed mainly layers with fibers
at angles above 45, leading to the conclusion that the other layers, at angles closer to 0,
failed by fiber breakage. The fact that more fiber breakage could be achieved at fracture for
the biomimetic lay-up may explain the higher relative strength of the notched samples.
II - 14
Keeping in mind its higher resistance to impact damage initiation [3], the higher notched
strength ratio of the biomimetic laminate further contributes to its better overall damage
tolerance compared to the standard quasi-isotropic laminate. The tradeoff for this increase in
performance is the higher complexity of the lay-up, requiring more care during production.
With current automated, computer-controlled manufacturing methods, this should however be
a minor problem when increased damage tolerance is really desired.
CONCLUSIONS
The plain and notched strength of a biomimetic laminate with a reinforcement architecture
similar to that used in biological composites has been examined and compared to the strength
of conventional quasi-isotropic and cross-ply laminates. For plain samples without a hole, the
modulus of the biomimetic laminate is about the same as that of the quasi-isotropic one, and
its strength is only slightly lower. Both have a lower strength than the cross-ply laminate.
When a hole is introduced into the laminate, the strength of the biomimetic laminate is
decreased less than that of the quasi-isotropic, and much less than for the cross-ply laminate.
This higher strength ratio and the larger size of a calculated equivalent damage zone indicates
a somewhat superior damage tolerance of the biomimetic laminate under tension. This
quality is more pronounced for the larger hole size in the present investigation. It is proposed
to be a result of the broader distribution of fiber angles leading to a wider range of
interlaminar stresses near the edges and hole. As the biomimetic laminate is also more impact
resistant, its overall damage tolerance is clearly better than the conventional quasi-isotropic
laminate.
ACKNOWLEDGMENTS
We thank Ms. M. Taguchi for her contribution to the experimental work for this study.
REFERENCES
1.
Neville, A.C., The Biology of Fibrous Composites: Development Beyond the Cell
Membrane, Cambridge University Press, 1993.
II - 15
2.
Giraud, M.M., Castanet, J., Meunier, F.J., Bouligand, Y., The fibrous structure of
coelacanth scales: a twisted plywood, Tissue and Cell, Vol. 10, No. 4, pp. 671-686,
1978.
3.
Gunderson, S.L., Whitney, J.M., A controlled unsymmetric ply orientation for improved
isotropic properties, Advances in Experimental Mechanics, ASME Vol. AMD-146, pp.
99-109, 1992.
4.
Li, S., Zhang, R., Fu, S., Chen, X., Zhou, B., Zeng, Q., A biomimetic model of fiberreinforced composite materials, Journal of Materials Science and Technology, Vol. 10,
No. 1, pp. 34-38, 1994.
5.
Eriksson, I., Aronsson, C.-G., Strength of tensile-loaded graphite/epoxy laminates

containing cracks, open and filled holes, Journal of Composite Materials, Vol. 24, No. 5,
pp. 456-482, 1990.
6. Chang, K.-Y., Liu, S., Chang, F.-K., Journal of Composite Materials, Vol. 25, pp. 274301, 1991.
II - 16
EFFECT OF LOADING PARAMETERS ON

COMPRESSION DOMINATED FATIGUE BEHAVIOR
OF IMPACT DAMAGED COMPOSITE LAMINATES
Milan Mitrovic, H. Thomas Hahn & Greg P. Carman
Mechanical & Aerospace Engineering Department,
University of California, Los Angeles, CA 90095-1597, USA
SUMMARY: In this paper work aimed at characterizing the influence of low-velocity

impact damage and loading parameters on fatigue response of AS4/3501-6 quasi-isotropic
laminates is presented. Constant amplitude compression-compression fatigue tests indicate
that the impact induced delamination does not growth in the absence of local buckling effects
and for the load levels below 50% of the CSAI (Compressive Strength After Impact). In the
presence of overloads maximum compressive strain that these laminates can sustain without
damage growth is reduced substantially, and block loading tests indicate that the high/low
sequence is more damaging than the low/high loading sequence. Constant amplitude tensioncompression fatigue of impacted laminates shows negligible differences in terms of life and
damage propagation modes when compared to the undamaged specimens. Results of this
study suggest that clearly defined damage tolerance criteria for impacted composite laminates
can be derived from a limited number of tests.
KEYWORDS: impact damage, post-impact fatigue, damage tolerance
INTRODUCTION
In aircraft applications composite panels are routinely subjected to both low-velocity impacts
and long term mechanical loading (fatigue). These loading events cause damage to develop in
the form of fiber breakage, matrix micro-cracking, and delaminations which lead to material
property degradation with the most severe being compression strength. This reduction
represents one of the major issues for satisfying the safety requirements of the aircraft
structures. Strength reduction of impacted composite panels has been studied extensively
over the past decades, and it has been shown that it depends both on the type and extent of
damage.
The post-impact fatigue behavior has been studied by a number of researchers on different
materials, lay-ups, and for different loading conditions. Stellbrink [1] studied the fatigue
behavior under the tension-compression loading of T300/69 and T300/914 quasi-isotropic
laminates. From S-N curves, it was concluded that the rate of life reduction is higher for
undamaged than for damaged specimens, and that test data agree better with the sudden death
model than with the degradation model. Influence of impact damage on fatigue behavior of
AS4/3501-6 laminates under tension-tension (T-T), tension-compression (T-C), and
compression-compression (C-C) loading was studied by Ramkumar [2]. Fatigue testing was
performed at stress levels above 60% of the ultimate strength, and while most of the
specimens subjected to T-T loading survived 1 million cycles, all specimens loaded in T-C
II - 17
and C-C failed well before. The author concluded that tensile loading, static or fatigue,
represents the least severe mode to cause additional damage in impacted laminates. In
another experimental study by Ramkumar [3], the effect of imbedded (idealized)
delaminations on the compression behavior of three different stacking sequence of quasiisotropic T300/5208 graphite/epoxy laminates was investigated. From the S-N data, it was
concluded that the threshold value of the maximum compressive stress, at which failure is not
expected to occur, depends on the location and shape of the implanted delamination. Also,
the occurrence of delamination growth and its direction is shown to be dependent of the
laminate stacking sequence and its through-the-thickness location, and the failures were
induced predominantly by the propagation of imbedded delaminations to the tab region.
Blaricum et al., [4], studied the compression dominated loading based on the modified
FALSTAFF flight by flight sequence on a [+45/-45/0/0/]7S XAS-914C impacted specimens.
Fatigue testing is performed under the maximum compressive strain of 0.36%, a value close
to the design strain used in military aircraft. It is reported that for this type of material and
lay-up, the low load levels can be deleted from the testing sequence with no significant effect
on fatigue life, and that the duration effect of high-level loads does not influence fatigue life
significantly. Some researchers have suggested the fatigue damage tolerance criteria for
impacted composite structures that require only static loading [5]. This is based on
observations that the impacted composite panels have very flat compression S-N curves, and
that although compression strength is greatly influenced by the impact damage, any
subsequent reduction under fatigue loading is minimal.
While the real service loads are usually kept much lower than the ultimate failure loads, most
of the research work on fatigue of composite laminates with impact induced damage has
focused on the final failure. The objective of this study is to determine the influence of
loading parameters on impact induced delamination growth during fatigue loading, and the
fatigue design limits (delamination growth thresholds) for composite laminates containing
barely visible low-velocity impact damage. Specifically the effects of load type, load level,
load sequence, and the effect of overloads, their magnitude and place in the load spectrum are
of interest in this study. Method to accelerate fatigue testing is suggested by extending the
results of compression dominated constant amplitude and block loading tests to random
fatigue loading.
EXPERIMENTAL SETUP
Quasi-isotropic [0/+45/-45/90]S4 laminates made from an AS4/3501-6 graphite/epoxy
material system were used in this study. Gage length of the coupons for static and fatigue
tests of 38.1 x 38.1 mm was chosen to prevent the occurrence of buckling effects in the
presence of compressive loads. Impact tests are performed on a Dynatup 8200 drop weight
impact testing machine with a 6.35 mm and 12.7 mm tup diameters and 4.3 kg impactor
weight (cross-head and tup). A modified SACMA SRM 2-88 impact fixture was used to
provide the necessary support for the specimens. Modifications were made to accommodate
specimens dimensions used in this study, and consisted of a smaller 25.4 x 25.4 mm square
cutout instead of the recommended 76.2 x 127 mm cutout. In conjunction with the Dynatup
8200 drop weight impact testing machine, the model 730-I data acquisition and analysis
system was used which can provide complete records of impact energy and force as a function
of time. Static compression tests are performed on a 50 kN capacity Instron test frame. The
specimens were loaded at a displacement rate of 1.27 mm/min, per ASTM D3410-75
standard. An end-loaded compression testing fixture was used, which is similar to NASA
II - 18
short block compression fixture. The test fixture was placed in the test frame upon a spherical
seat which provided alignment of the specimen during compression. Following static tests,
the fatigue was performed on a 10 kN capacity Instron test frame. Monitoring damage
progression during fatigue loading, via X-ray radiography, was done periodically to asses the
change of the damaged area as a function of loading parameters.
IMPACT DAMAGE EVALUATION AND COMPRESSIVE STRENGTH

AFTER IMPACT (CSAI)
Initial impact tests were performed to determine the influence of impact test parameters on
damage development. Impact energies were adjusted to produce barely visible impact
damage for subsequent fatigue loading. The overall impact damage area, in the form of a
delamination, has been determined via X-ray radiography, and the observed damage size is
plotted versus the incident impact energy in Figure 1.a. Caution should be taken when
interpreting results for the maximum damage diameter of 25.4 mm achieved in these tests,
since it coincides with the 25.4 x 25.4 mm square cutout of the clamping fixture. That is, the
damaged area might have been larger if the boundary conditions in these tests were different
(i.e., for the larger unsupported area).
Fig. 1. a) Damage area vs. the incident impact energy

b) CSAI as a function of impact test parameters
As can be seen from Figure 1.a, some specimens that were impacted with the same incident
energy of 1.9 Joules, displayed different damage patterns. That is, some specimens contained
delaminations 12.7 mm in diameter, while others showed no damage at all. Even sectioning
and microscopic examination of these specimens revealed the absence of any damage in the
form of matrix cracking or delamination, suggesting that there exists an impact energy
threshold below which damage cannot be found (~ 2 J for this material system and lay-up).
Above this threshold energy, significant damage occurred, corresponding to 12.7 mm
diameter of damaged area and resulting in a damage diameter to specimen width ratio (d/w) of
0.33. Other researchers have also reported the existence of the minimum energy required to
cause an impact damage. Choi et al. [6] reported that an impact energy threshold exists for
T300/976 graphite epoxy cross-ply composites. The study also showed that stacking sequence
II - 19
significantly influences the impact energy required to initiate impact damage. Reported values
range from 0.52 to 3.1 J, which is similar to the values obtained for our material system. In
another study [7] various lay-ups of the carbon fiber reinforced thermoplastic toughened
epoxy (IM7/977-2) were investigated, including cross-ply and quasi-isotropic. The
normalized energy at the onset of incipient damage was found to be on the average 1.5 J/mm
(or 9.75 J, if expressed in the total impact energy). These values are much higher than
observed in our investigation, which is as expected for 977-2 toughened epoxy system.
Since the objective of this research is to investigate the fatigue growth of barely visible impact
damage, an incident impact energy producing a repeatable damage pattern was chosen (2.1 J).
A summary of static compression tests of these specimens is presented in Figure 1.b,
including all relevant test parameters and associated compressive strengths. As can be seen
from the figure, samples that did not contain visible damage showed negligible degradation in
the compressive strength with respect to the undamaged specimens. That is, the average
CSAI of these specimens (600 MPa) is very close to the value of 621 MPa obtained on
undamaged samples. It can be concluded that the compressive strength of impacted
specimens is only function of the damage present, agreeing well with the results of Cairns and
Lagace [8]. That is, the damage was caused by different impactor mass, velocity and
boundary conditions, but for the equivalent damage an equivalent compressive strength is
measured. Furthermore, while subjected to slightly different impact energies and different
impactors, samples with approximately same delamination of 12.7 mm in diameter
experienced very similar compressive strength reduction (Figure 1.b). Average CSAI of these
samples is 470 MPa (with a small standard deviation of 18 MPa), which corresponds to 25%
reduction when compared to the undamaged specimens. Stellbrink [1] has reported
compressive strength reduction of ~50% for impacted T300/69 and T300/914 quasi-isotropic
laminates with d/w ratios of 0.32 - 0.44, while Ramkumar [2] investigated the behavior of
impacted AS4/3501-6 laminates and observed ~65% reduction in compressive strength.
While the ratio of the damaged area with respect to the overall specimen's dimensions in
Ramkumar study is almost the same as in current study (i.e., d/w of 0.33), the strength
reductions are substantially different. It seems that more accurate qualification of the impact
damage, rather than the overall damaged area, is needed in order to compare different test
results. Furthermore, compression test method, gage length of the specimen (buckling
effects), and definition of the final failure are also parameters which are important in defining
the CSAI. The actual delamination failure of the coupons in this study occurs at ~380 MPa,
while the final collapse of the specimens occurs much later at ~470 MPa, which is defined as
the compressive strength after impact.
POST-IMPACT FATIGUE BEHAVIOR

Summary of fatigue test results including the type of loading, load levels, test frequency and
the number of cycles for each specimen is outlined in Table 1. Two specimens were tested
per each load level which were chosen between 30 and 70% of CSAI. It should be noted that
the variable test frequency (smaller for higher stress levels) was dictated by the limitations of
the test frame capacity.
II - 20
Table 1. Test matrix for constant amplitude and block loading

Specimen
Test frequency Load level
#
[Hz]
[% of CSAI]
Constant Amplitude Compression-Compression
32A6
7.5
40 %
32A7
7.5
40 %
33B1
7.5
50 %
33B4
7.5
50 %
34A8
5
60 %
35A5
5
60 %
33B2
4
70 %
35A6
4
70 %
N
# of cycles
Specimen
Test frequency Load level
#
[Hz]
[% of CSAI]
Compression Block Loading (High/Low)
33B3
4
70 %
10
30 %
33B5
4
70 %
7.5
40 %
Compression Block Loading (Low/High)
31C7
7.5
40%
4
70%
32A6
7.5
40 %
4
70 %
Constant Amplitude Tension-Compression
32C5
7.5
30 %
32C6
7.5
30 %
31C5
5
40 %
33A6
5
40 %
33A5
5
50 %
31C4
5
50 %
* indicates cycles to final failure
N
# of cycles
1,000,000
1,000,000
1,000,000
1,000,000
>500,000
>400,000
141,607*
100,000*
100
1,000,000
100
1,000,000
1,000,000
>100,000
1,000,000
45,296*
1,000,000
1,000,000
>600,000
376,602*
70,500*
110,000*
CONSTANT AMPLITUDE (C.A.) COMPRESSION-COMPRESSION (C-C)

LOADING
The extent and mode of delamination growth under compression loading varied significantly
depending on the applied load level. For specimens cycled at 40 and 50% of CSAI, the
impact induced damage did not grow under cyclic loading up to 106 cycles. At higher load
levels corresponding to 60% of CSAI only minor growth of impact damage was observed.
Moderate delamination growth initiated after 2,000 cycles, and it reached the tab region of the
specimen after 50,000 cycles. Schematic representation of this damage pattern, as observed by
X-ray radiography, is presented in Figure 2. Growth of this thin delaminated region occurred
on the rear impacted side of the specimen, along the lines of impact induced matrix cracks in
the outer 0o ply, and is attributed to local buckling of the outer plies. At higher load levels
(70% of CSAI) different and more severe type of damage growth was observed, as can be
seen from Figure 3. Although there are some variations between specimens, delamination
II - 21
extended throughout the entire gage section at ~10,000 cycles. This was a consequence of
'delamination failure' which occurred between 100 and 1,000 cycles. Microscopic
examination of specimens tested at 70% CSAI indicate more severe damage which spans few
outer plies on the rear impacted side of the specimen. However, catastrophic failure of these
specimens occurred much later, at more than 100,000 cycles. Again, buckling of the outer
plies on the rear impact side was the major factor governing delamination growth at this stress
level.
N=0
N = 2,000
N = 5,000
N = 50,000
N = 500,000
Fig. 2. Growth of impact damage under compression-compression fatigue (60% of CSAI)

N=0
N = 10
N = 100
N = 1,000
N = 10,000
Fig. 3. Growth of impact damage under compression-compression fatigue (70% of CSAI)

S-N curves of impacted and previously [9] tested undamaged specimens are presented in
Figure 4.a. Stress levels in this figure are normalized with respect to the ultimate compressive
strength (UCS) of undamaged specimens, where 100% of CSAI corresponds to 75% of UCS.
It is evident that the fatigue life of impacted specimens is reduced when compared to
undamaged coupons, especially at higher load levels (> 60% of CSAI or 45% of UCS) where
local buckling and damage growth due to delamination failure govern the fatigue response.
At lower load levels for which damage did not grow in fatigue, similar performance in terms
of life is observed.
undamaged specimens
impacted specimens
0.5
0.4
0.4
S / UCS
0.6
0.5
S / UCS
0.6
0.3
0.3
0.2
0.2
0.1
0.1
0
100
0
1000
10000
100000
N (number of cycles)
1000000
100
1000
10000
100000
1000000
N (number of cycles)
Fig. 4. Strength vs. number of cycles (S-N) curve for undamaged and impacted specimens
a) C-C loading, b) T-C loading. Note: stress levels are normalized with respect to the
ultimate compressive strength (UCS) of undamaged specimens
II - 22
CONSTANT AMPLITUDE (C.A.) TENSION-COMPRESSION (T-C) LOADING

The damage growth under the tension-compression loading resembled the fatigue behavior of
undamaged specimens. Specimens cycled at 30 of CSAI revealed only ply cracking, and no
growth of impact damage up to 106 cycles. At higher load levels of 40 and 50 of CSAI
both ply cracking and edge delamination are observed, together with the minor impact
damage growth. Schematic representation of the damage pattern at 50 of CSAI is presented
in Figure 5. While the growth of very thin delaminated region occurs prior to the growth of
delamination from the edges, it is the edge delamination that actually governs the fatigue life
as in the case of non-impacted specimens. Furthermore, comparison of S-N curves of
impacted and previously [9] tested undamaged specimens (Figure 4.b) reveals no differences
between the two. Due to the most damaging nature of T-C loading, failure of the laminates
occurs at lower load levels than in C-C loading. Therefore, the delamination failure, as
observed in constant amplitude C-C loading at 70% of CSAI has no practical importance in
T-C loading, and can not influence the damage growth modes for this type of loading. These
results suggest that long and costly testing of impacted composite laminates is not necessary
for T-C loading as long as the life of non-impacted laminates is known.
N=0
N = 50,000
N = 100,000
N = 110,000
Fig. 5. Growth of impact damage under tension-compression fatigue (50% of CSAI)

COMPRESSION-COMPRESSION BLOCK LOADING - EFFECT OF OVERLOADS
In order to determine the damage tolerance criteria for impacted composite laminates,
constant amplitude tests can provide useful information about the damage growth parameters
at different load levels. Based on the results of C.A. tests the fatigue design limit without
impact damage growth for this material system and type of initial impact damage corresponds
to load levels less than 50% of CSAI (0.5% strain) in the presence of only compressive loads,
and 30% of CSAI (0.32% strain) in tension-compression loading. However, it has been
suggested that delamination in composite materials may grow very rapidly for a small load
range, and that small uncertainties in the applied load may lead to large uncertainties in
delamination growth [10]. This is especially important if the service loads are greater than
anticipated. Therefore, the influence of overloads that might occur during the real service
conditions on damage growth and fatigue life is investigated next. Since only minor damage
growth was observed for specimens tested below 60% of CSAI, emphasis was placed on the
sequence of occurrence of high load levels (i.e. 70%).
Two specimens were subjected to low/high block loading to asses the influence of this type of
load sequence on damage propagation. After one million cycles at 40% of CSAI (with no
visible damage growth) fatigue was continued at 70% until the final failure. Observed
delamination growth for these specimens was very similar to the coupons that were tested
only at 70% load level. Comparison of normalized delamination area (area of damage / gage
section area) for constant amplitude and 40/70 block loading is presented in Figure 6.a. As
II - 23
can be seen from the figure, for the same number of cycles at 70% the same trend in
delamination growth is observed. Also, the number of cycles (at 70%) to final failure of these
specimens is comparable to the coupons fatigued at constant amplitude loading (see Table 1).
For other fatigue damage modes in composite materials (like ply cracking in tension
dominated fatigue) blocked tests have shown that low/high load sequence results in a greater
damage [9]. For that case the low loads by itself may not be damaging, but they may aid the
subsequent high load to produce greater damage than would otherwise occur without it. For
the type of damage in this study the exact opposite behavior is observed.
1.00
1.00
block loading 70/30
C.A. 70% of UCS
0.80
Normalized Delamination Area
Normalized Delamination Area
C.A. 70% of UCS
Block load 40/70

0.60
0.40
0.20
0.00
0.80
block loading 70/40
0.60
0.40
0.20
0.00
1
10
100
1000
10000
100000 1000000
N - number of cycles at 70% of CSAI
10
100
1000
10000
100000 1000000
N - number of cycles at lower load levels
Fig. 6. Delamination growth as a function of number of cycles

a) at 70% of CSAI for constant amplitude (70%) and low/high block loading (40/70%)
b) at 30 and 40% of CSAI for high/low block loading (70/30 and 70/40%)
100 cycles
@ 70%
40% of UCS
N = 10,000
N = 100,000
N = 200,000
N = 500,000
Fig. 7. Growth of impact damage under block compression-compression loading
II - 24
Effect of overloads that occur at the beginning of the load spectrum (high/low test sequence)
was investigated on two coupons which were initially exposed to 100 cycles at 70%, and
subsequently fatigued at lower stress levels (i.e. 30% and 40% of CSAI). Schematic
representation of the damage pattern for the second case, as observed by X-ray radiography, is
presented in Figure 7. As indicated in this figure, following the overloads that cause
'delamination failure', delamination growth occurs even at these low load levels (for which no
growth of impact damage is observed in constant amplitude tests). Normalized delamination
area is plotted versus the number of cycles for these two load sequences in Figure 6.b.
Examination of these data indicate that delamination growth threshold, for laminates
subjected to overloads, is even below 30% of CSAI. Results also suggest that the high/low
test sequence for impact damaged composite laminates is more damaging than the low/high
sequence.
Based on the results of constant amplitude and block loading some conclusions about the
influence of higher load levels on damage progression during random spectrum loading can
be inferred. Since high load levels represent less than 0.05% of the total number of cycles in
the transport wing spectrum test [11], their magnitude and sequence of occurrence is very
important for the appropriate design of standardized certification procedures for impact
damaged composite materials. Since a full flight-by-flight spectrum test is long and costly, a
method of accelerating this test through truncation of either low or high loads is beneficial
[11]. Based on the results of this study, the damage progression modes are expected to be the
same for the baseline and truncated load spectrum as long as the highest load levels are kept
below the levels that cause delamination failure (i.e., <60% of CSAI). On the other hand, if
these high load levels are included in the test spectrum the truncation of either high or low
load levels (following a high load level) might lead to unreasonable estimates of fatigue life.
However, it should be pointed out that these suggestions apply only for the type of initial
damage as used in this study, that is, barely visible low-velocity impact damage.
CONCLUSIONS
Results of impact tests show that a correlation exists between the energy absorbed by the
composite during impact and damage induced in it. Furthermore, static compression tests
demonstrate that the compressive strength is a function of the damage state only (i.e. absorbed
impact energy), and is relatively independent of other impact test parameters. For the
laminates impacted close to the impact damage threshold (resulting in a damage-diameter-towidth ratio of 0.33), a 25% reduction in compressive strength was measured. Subsequent
constant amplitude compression-compression fatigue tests of these specimens indicate that the
delamination does not grow for load levels below 50% of the CSAI. On the other hand,
coupons tested at 70% of CSAI and above, exhibited significant delamination growth even
after a few cycles. The growth was caused by the delamination failure of the plies located on
the rear impact side of specimens. Local buckling of delaminated plies was the reason for
delamination growth even when the load levels were reduced to 30% and 40% of CSAI.
Furthermore, block loading tests indicate that the high/low sequence is more damaging than
the low/high load sequence for compression-compression loading. Fatigue behavior under the
constant amplitude tension-compression loading of impacted laminates shows negligible
differences in terms of life and damage propagation modes when compared to the undamaged
specimens, suggesting that the damage tolerance criteria for this type of loading can be
defined from a limited number of tests.
II - 25
ACKNOWLEDGMENTS
The authors of this paper gratefully acknowledge the support of the Federal Aviation
Administration under grant number 95-G-021. We would also like to thank Peter
Shyprykevich and Joseph Soderquist of FAA for their guidance.
REFERENCES
1.
Stellbrink, K.K.U., Influence of Low-Velocity Impact on the Fatigue Behaviour of

CFRP Laminates, Fatigue and Creep of Composite Materials, Third International
Symposium on Metallurgy and Material Science, M. Lilholt and R. Talreja, Eds., 1982,
pp. 319-327.
2.
Ramkumar, R.L., "Effect of Low-Velocity Impact Damage on the Fatigue Behavior of

Graphite/Epoxy Laminates", Long-Term Behavior of Composites, ASTM STP 813, T.K.
O'Brien, Ed., American Society for Testing and Materials, Philadelphia, 1983, pp. 116135.
3.
Ramkumar, R.L., "Compression Fatigue Behavior of Composites in the Presence of

Delaminations", Damage in Composite Materials, ASTM STP 775, K.L. Reifsnider, Ed.,
American Society for Testing and Materials, 1982, pp. 184-210.
4.
Blaricum, T.J., D.S. Saunders, G. Clark and T.E. Preuss, Damage Tolerance of Impact
Damaged Carbon Fibre Composite Wing Skin Laminates, New Materials and Fatigue
Resistant Aircraft Design, Symposium, D.L. Ed., 14th CAF, 1989, pp. 537-556.
5.
Demuts, E., R.S. Whitehead, and R.B. Deo, "Assessment of Damage Tolerance in
Composites", Composite Structures, Vol. 4, 1985, pp. 45-58.
6.
Choi, H.Y., H.S. Wang and F.-K. Chang, "Effect of Laminate Configuration and
Impactor's Mass on the Initial Impact Damage of Graphite/Epoxy Composite Plates Due
to Line-Loading Impact", Journal of Composite Materials, 26 (6), (1992), pp. 804-827.
7.
Strait, L.H., M.L. Karasek and M.F. Amateau, "Effects of Stacking Sequence on the
Impact Resistance of Carbon Fiber Reinforced Thermoplastic Toughened Epoxy
Laminates", Journal of Composite Materials, 26 (12), (1992), pp. 1725-1740.
8.
Cairns, D.S. and P.A. Lagace, Residual Tensile Strength of Graphite/Epoxy and
Kevlar/Epoxy Laminates with Impact Damage, Composite Materials: Testing and
Design (Ninth Volume), ASTM STP 1059, S.P. Garbo, Ed., ASTM, Philadelphia, 1990,
pp. 48-63.
9.
Hahn, H.T., J. Bartley-Cho, and S.G. Lim, "The Effect of Loading Parameters on
Fatigue of Composite Laminates: Part II", DOT/FAA/AR-95/79 Report, 1996.
10.
OBrien, T.K., "Towards a Damage Tolerance Philosophy for Composite Materials and
Structures", Composite Materials: Testing and Design (Ninth Volume), ASTM STP
1059, S.P. Garbo, Ed., ASTM, Philadelphia, 1990, pp. 7-33.
11.
Phillips, E. P., "Effect of Truncation of a Predominantly Compression Load Spectrum

on the Life of a Notched Graphite/Epoxy Laminate", Fatigue of Fibrous Composite
Materials, ASTM STP 723, 1981, pp. 197-212.
II - 26
RTM CARBON COMPOSITES: INFLUENCE OF

PROCESS PARAMETERS ON COMPRESSION
STRENGTH AFTER IMPACT
Florentin Berthet1, Pierre Devos1, Thierry Ansart2
1
cole des Mines dAlbi Carmaux,Centre de recherche matriaux,

Campus Jarlard-route de Teillet,81013 Albi Ct Cedex 09, France
2
Centre dEssais Aronautique de Toulouse, Division matriaux et structures,
Departement structures et matriaux cellule,
23 avenue Henri Guillaumet, 31056 Toulouse, Cedex, France
SUMMARY: By using the Taguchi method of experimental design, it has been possible to
demonstrate the influence of some processing factors on compression strength after impact of
carbon/epoxy composite materials processed by resin transfer molding. The parameters
studied were injection and holding pressure, mold temperature, fiber bundle size, tackifier
level, injection type and the interaction between injection pressure and mold temperature.
Statistical analysis of results leads to the conclusion that temperature and pressure conditions
modify the interface properties. Moreover, fiber bundle size, pressure, temperature and most
likely tackifier level affect the post impact compressive behavior of the materials.
Nevertheless, RTM can be considered as tolerant to process conditions variations.
KEYWORDS: resin transfer molding, experimental design, processing, quality, compression

strength after impact
INTRODUCTION
Use of advanced composite materials in defense and aeronautics increased dramatically in the
last 30 years because these sectors could bear extra costs of production provided there was a
reduction in the component weight. The challenge which now has to be faced by aerospace
manufacturers is to reduce their costs. For this reason, the Resin Transfer Molding Process
(R.T.M.) currently provides new interest for the aeronautics industry.
Numerous studies have been performed to simulate flow of resin through preforms with
applications in mold design. A new requirement has arisen: the identification of the
relationship between process parameters and mechanical properties of the molded component.
Compression strength after impact is one of the most significant mechanical properties for
aerospace requirements. Resin transfer molding has been identified as a process through
which it is possible to improve impact resistance.
Therefore a study had been conducted experimentally to fill a part of the gap. Compression
strength after impact (C.S.A.I.) has been evaluated.
II - 27
BACKGROUND
Taguchi Method
Experimental design using the Taguchi method is well adapted to evaluate influence of
processing conditions on mechanical properties of RTM CFRP because the potential
parameters are numerous. Experimental design makes it possible to simultaneously minimize
the number of tests necessary and maximize the quality of results.
A test on assumed equality of variance between samples and in samples leads to a separation
of factors in three categories: very significant (VS), significant (S), not significant (NS). The
greater the variance between two levels, the more significant the factor concerned.
An effect is associated with a factor i and a level j. It represents what should theoretically be
added to the average value if the component has been molded at level j of the factor i. That is
to say the effect Eij of having been molded at level j instead of at average level. This effect is
defined as Eqn 1.
Eij=average at level j-general average
(1)
More details can be found in M.Pillet's book (1).

The main difficulty in this kind of problem is to establish a relationship between what
happens microscopically and macroscopically. It is all the more complicated because the
involved mechanisms are multiple. Moreover the expected effects are small at room
temperature.
RTM
RTM is a process in which a dry fibrous preform is impregnated by a thermoset to produce a
composite. In a prior stage called preforming the fibre layup is made to the shape of the
component to produce a preform. The role of the tackifier is to keep the shape of the
component and to make it stiff enough to be handled prior to injection.
It is now widely accepted, that in RTM, resin flow is governed by Darcy's law:Eqn 2.
<
[K ]
v=
grad(p)
(2)
v: front velocity in porous media, : viscosity, : porosity (volume left free by fibers divided
by total volume), p: applied pressure K: permeability tensor.
In RTM porous media for impregnation are heterogeneous. In a component there are two
levels of heterogeneity, microscopic (fiber level) and macroscopic (fabric level). This implies
that there are two types of flow (around bundles and in bundles) and permeability.
II - 28
Available studies of the influence of process parameters on mechanical properties are rare.
Nevertheless Hayward and Harris [2] Dockum and Shell [3], Steenkamer [4], Young [5] and
Michaeli [6] have done some work in this area. But their conclusions are contradictory.
Impact Testing
Impact test procedure
The impact tests were performed using a drop weight tester. The specimens were clamped at
all corners with a rectangular window (75x125 mm), in agreement with Airbus Industrie test
method AITM 1-0010. The drop weight was released at a predetermined height to impact the
center of the specimen with a given energy. Restrike of the drop weight was prevented by
capturing it after first impact.
Non destructive control
Damage areas were measured with C-SCAN.
Compression after impact test procedure
Specimens (150x100 mm) were subjected to a compressive load using an antibuckling C.A.I.
test fixture. The compression tests were conducted at a constant cross head speed of 1mm/mn.
DESCRIPTION OF EQUIPMENT AND PROCESS

Our equipment consisted of:
- A square steel mold allowing fabrication of plates 400*400mm in different
thicknesses.With this mold we can do peripheral injection (using the whole feeder as
indicated on Fig. 1) or directional injection (by filling the side feeders).
- An heated 20T hydraulic press
- An RTM machine (pressurized pot) allowing injection up to 4 bars
- A vaccum pump
- A data recorder (pressures, temperatures, vaccum)
position of evacuation gate
(to vacuum pump)
preform
feeder
injection gate (from injection pot)
Fig. 1: Top view of half mold with preform

The typical resin transfer molding cycle is as follows
1) cutting out of reinforcements
2) assembly of reinforcements
II - 29
3) preforming 5 min at 120C under 2 bars (cohesion of reinforcements is achieved by a

tackifier that is a thermoplastic or a re-actifiable uncatalysed thermoset).
4) mold closure, vacuum application, resin degassing, and injection until mold is filled
5) holding pressure until no more material enters the mold
6) cure 75 min at 160C
7) mold opening
8) post cure in autoclave following a programmed cycle (120 min at 180C)
DESCRIPTION OF EXPERIMENT
Parameters
Parameters chosen for this study are: mold temperature, injection pressure, holding pressure,
fiber bundle size, tackifier level, injection type and interaction between mold temperature and
injection pressure. These parameters are varied between two values according to L8 taguchi
table (Table 1).
Table 1: L8 Taguchi table applied to RTM
specimen
no.
mold injection
temp. pressure
(C) (bars)
1
2
3
4
5
6
7
8
100
100
100
100
140
140
140
140
1
1
2
2
1
1
2
2
Temp.holding tackifier injection

pressure
pressure level
type
Interaction (bars)
(g/m2/
face)
1
2
5
directional
1
3
15
peripheral
2
2
5
peripheral
2
3
15
directional
2
2
15
directional
2
3
5
peripheral
1
2
15
peripheral
1
3
5
directional
fiber
bundle size
(filaments)
3000
6000
6000
3000
6000
3000
3000
6000
We discount the effects of cure and post cure because these two steps are not characteristics
of RTM.
What is specific to RTM is the fiber/matrix interface formation with defect creation during
injection caused by unknown mechanisms. From this one assumes that processing influences
fiber/matrix bonding and thus the mechanical properties of the composite.
Materials
Four fabrics were used, the fabric style was the same (balanced fabric), but fabrics are woven
from fiber bundles of different sizes (3K 400g/m2 or 6K 600g/m2), and contained two possible
levels of tackifier (5g/m2/face or 15g/m2/face). When different fiber bundle size are used,
different layups had to be used to achieve comparable bending stiffness (Table 2).
The injected resin is a special RTM single component epoxy resin .
II - 30
PRESENTATION OF DATA AND RESULTS

For each of the processing conditions, one plate is produced and in each plate six rectangular
samples were cut (150 (in O direction)x100mm ). Specimens were impacted at 5J(x2), 10J,
15J, 25J, 35J.
Table 3: Mass and energy impact conditions
Impact energy (J)
5
10
15
25
35
Drop weight (kg)

2,55
2,55
2,55
6,69
6,69
II - 31
As shown in Fig.2 different fiber bundle sizes do not lead to different maximum deflections.
It proves that working at the same bending stiffness [8] is justifiable and that the change in
properties is a consequence of processing and not a consequence of layup.
The plates produced with different materials can be easily differentiated on Fig.3 by bundle
size: 1,4,6,7 small (3K), 2,3,5,8 large (6K). But as presented in Fig.3 the data are not easy to
interpret and some more work is required .
INTERPRETATION OF DATA
Damage Surface at 5J Impact
For an impact event damage initiation first appears at the fiber/matrix interphase at low levels
of impact energy. There is an energy threshold at which microcracks appear at the
fiber/matrix interface. Then cracks propagate in the matrix and at the interface between two
plies. Thus the damage surface (measured by C.SCAN), for a low impact energy, will reflect
the level of fiber and matrix bonding. Results of variance analysis in agreement with Taguchi
method appear in Table 4.
The calculated value Fexp is more important than the tabulated value Ftheo95 for mold
temperature and Ftheo99 for fiber bundle size. Mold temperature and fiber bundle size have a
significant effect, and very significant respectively on damage area at 5J. For example, a mold
temperature increase from 100C to 140C is associated with a reduction of the damaged area
of 40 mm2 that is to say a variation of twice the effect (19.9 mm2).
II - 32
Table 4: variance analysis table for damage surface at low impact energy
Effect (mm2) Fexp
mold temperature
injection pressure
holding pressure
tackifier level
injection type
fiber bundle size
P-T interaction
-19,9
6,63
-0,38
-11,63
-1,88
46,38
13,63
6,21
0,69
0
2,13
0,06
33,8
2,92
Ftheo95
nu1=1,nu2=8
5,32
5,32
5,32
5,32
5,32
5,32
5,32
Ftheo99
nu1=1,nu2=8
11,26
11,26
11,26
11,26
11,26
11,26
11,26
sign.
S
NS
NS
NS
NS
VS
NS
Increasing mold temperature favours fiber/matrix bonding (damage area reduction with
increasing temperature as indicated by the negative sign of mold temperature effect).
Influence of fiber bundle size is due to a structure effect (although laminates have bending
stiffness as similar as possible) and also a difference of impregnation induced by fiber
mismatch. The bigger the fiber bundle the more difficult to impregnate the bundle and the
more important the damage area.
Kittelson and al. [7] have shown influence of tackifier compatibility with the resin. At higher
temperatures tackifier level influences compression strength. In our study this parameter had
an effect, but was not statistically significant. It is likely that tackifier (level) is also important
(if temperature has to be considered ).
Compression Strength after Impact
A least square fit of experimental data points leads to an equation of CAIS as a function of
impact energy summarized in Table 5. Data are fitted to a power law : Eqn 3.
C . S. A . I . = K . ( E i )
n
(3)
K: coefficient, n:exponent, Ei: incident energy

Analysis of coefficient and exponent by the taguchi method is shown in Table 6 and in Table
7. Pooling method was used to artificially increase the degree of freedom (D.O.F. to 3) of the
residual so as to draw conclusions more easily.
Fiber bundle size, mold temperature and holding pressure have a (very) significant effect on
coefficient (Table 6).
Holding pressure has a positive effect on coefficient of C.S.A.I.. For example, increasing the
holding pressure from 2 to 3 bars is associated with a variation of twice the effect (+8.5 MPa)
that is to say an increase of coefficient of 17 MPa.
II - 33
Table 5: equations for different process conditions

Plate n
equation
y=521x-0,2478 R2=0,967
y=468x-0,2301 R2=0,96
y=438x-0,2082 R2=0,97
y=532x-0,2503 R2=0,987
y=426x-0,2106 R2=0,983
y=501x-0,2379 R2=0,955
y=469x-0,2092 R2=0,97
y=421x-0,2052 R2=0,97
Table 6: Variance analysis for coefficient y:C.S.A.I. x:Ei

Effect
DOF Var.
8,5375
sum of
squares
583,1
-2,99
406,4
-17,8075
Fiber bundle -33,7625

size
residual
Holding
pressure
injection
pressure
Mold
Temp.
Fexp
Ftheo99
1=1,2=3
34,1
sign.
583,1
Ftheo95
1=1,2=3
12,60587 10,1
406,4
8,785889 10,1
34,1
NS
2537
2537
54,84249 10,1
34,1
VS
9119
9119
197,1426 10,1
34,1
VS
138,8
46,26
Table 7: variance analysis for exponent

source
Effect
Fexp
Ftheo95
1=1,2=3
6,629955 10,1
Ftheo99
1=1,2=3
34,1
sign.
0,005963
sum of DOF Var.

squares
0,0003 1
3E-04
Holding
pressure
injection
pressure
Mold
Temp.
Fiber bundle
size
residual
-0,00669
0,0004 1
4E-04
8,340294 10,1
34,1
NS
-0,00919
0,0007 1
7E-04
15,74159 10,1
34,1
-0,01139
0,001
0,001
24,18302 10,1
34,1
0,0001 3
4E-05
NS
Finally mold temperature and fiber bundle size have rather a negative effect on compression
strength after impact. Higher temperatures and fiber bundle size lead to a reduction of
compression strength of the material (Fig.4), and residual property also decreases faster with
impact energy . Different temperature and pressure conditions during injection influence the
mechanical property of the resin. It was proved that fiber bundle size, temperature and
pressure conditions have a statistically significant effect on compression strength after impact.
II - 34
These process parameters are supposed to act through modifying the resins capacity to
prevent fiber microbuckling and energy release rate (GIIc).
CONCLUSION
Experimental design made it possible to demonstrate the effect of some processing factors on
mechanical properties of RTM CFRP. We established the (limited) effect of temperature and
pressure conditions during injection on the final quality of composites. The initiation of
cracks is favoured by large fiber bundles and low injection temperature. The compression and
residual compression resistance are influenced by pressure, temperature and fiber bundle size.
The most significant effect found was fiber bundle size. Nevertheless, RTM can be considered
as tolerant to process condition variations.
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
Pillet, M. Introduction aux plans dexpriences par la mthode taguchi, 2nd edition, les ditions de
lorganisation,1994.
Hayward, J.S. and Harris, B. Effects of process variables on the quality of RTM moldings, SAMPE
journal, Vol. 26, n3, May/June1990, pp.39-46.
Dockum, J.F.Jr. and
Schell, P.L. Fiber directed preform reinforcement: factor that may
influencemechanical properties in liquid composite molding, proceeding of the ASM/ED conference,
1990, pp.393-406.
Steenkamer, D.A. The influence of preform design and manufacturing issues on the processing and
performance of the resin transfer molded composites, Thesis of the University of Delaware, 1994.
Young, W.B.,Tseng, C.W.,Study on the preheated temperatures and injection pressures of the RTM
process, J.of Reinf.Plast.and Comp.,Vol.13-May 1994, pp.467-482.
Michaeli, W.,Dickhoff, J.,Production of fibre reinforced components with high surface quality using a
modified RTM technique, SAMPE 39, April 1994, pp.145-153.
Kittelson, J.L. and Hackett, S.C.,Tackifier/resin compatibility is essential for aerospace grade resin
transfer molding, SAMPE 39, April 1994, pp.82-96.
Bucinell,R.B., Nuismer, R.J.,Koury, J.L.,Response of composite plates to quasistatic impact events,
Composite materials : fatigue and fracture (third volume), ASTM STP 1110, T.K. OBrien, Ed.,
American Society for Testing Materials, Philadelphia, 1991, pp.528-549.
II - 35
FREE VIBRATION OF DELAMINATED COMPOSITE

PLATES WITH CONTACT ANALYSIS
Meng-Kao Yeh and Ren-Cheng Lai
Department of Power Mechanical Engineering
National Tsing Hua University, Hsinchu 30043 Taiwan, ROC
SUMMARY: The free vibration of composite laminated plates with through-the-width

delamination was investigated analytically and experimentally. In analysis, Hamiltons
principle and the finite element method were extended to derive the free vibration equation
for the delaminated composite plates. The transformation method, which reduced the degrees
of freedom of the plate, was applied to deal with the geometric compatibility between the
contact surfaces of upper and lower sublaminates in the delaminated region. The free
vibration of the delaminated composite plate was treated as a two-dimensional plane-strain
eigenvalue problem. In experiment, a shaker system was used to determine the natural
frequencies and the corresponding mode shapes for the delaminated plates. The delamination
becomes significant to affect the natural frequencies as its size exceeds 50% of the laminate
length. The natural frequencies of the delaminated plates decrease with increasing values of
fiber orientation and also decrease as the delaminated region located toward the center of the
composite plates.
KEYWORDS: free vibration, natural frequency, composite plate, delamination
INTRODUCTION
Owing to high strength/weight ratio, composite materials have been widely used in the
aerospace, automobile, defense and civil industries. However, there is occasionally some
defects in our daily used materials. Composite materials have defects as well. Delamination
is one of the defects encountered in most composite laminated plates. It results from low
velocity impact, fatigue loads and manufacturing defects. Therefore, study of influence of
delamination on the composite laminates is rather important.
The dynamic characteristics of composite laminates were investigated by Rayleigh-Ritz
method [1], Galerkin Method [2, 3], and finite element method [4-10]. Among them,
Bhimaraddi [2] studied the effect of initial imperfection on the nonlinear dynamic response of
composite laminates; Bicos and Springer [4] derived the free vibration equation of composite
laminates with damping by Hamiltons principle; Dhanaraj and Palaninathan [5] and Ju et al.
[6] discussed the influence of initial stress on the dynamic characteristics of composite
laminates. Shen and Grady [7], Tracy and Pardoen [8] and Ju et al. [9, 10] assessed the
dynamic characteristics of composite laminates with delamination. Ju et al. [10] analyzed the
contact phenomenon at the delamination; however, they dealt with the case with node pairing
only.
II - 36
Among the methods related to contact mechanics, there are transformation matrix method
[11-13], penalty method [14, 15], Lagrangian multiplier [16, 17] and flexibility matrix method
[18]. The solution accuracy and stability of penalty method is determined by the magnitude
of penalty number and structural stiffness. The Lagrangian multiplier method needs additional
parameters. When using the flexibility matrix method, the node pairing on both sides of
contacting surfaces is required. The transformation matrix method satisfies the geometric
compatibility between the contacting surfaces and reduces the number of degrees of freedom
of the problem. It deals with both sticking and sliding conditions on the contacting surfaces,
which was not discussed before for the composite laminates; therefore, it was used in this
study to obtain the dynamic response of the composite laminates.
In analysis, Hamiltons principle and the finite element method were used to derive the
system equation of the composite laminates. The governing equation omitted the damping
due to its small effects on the composite laminates [19-21]. For the contact points, the
transformation matrix method was used and the contacting conditions were iterated. The free
vibration of composite laminates was treated by a two-dimensional plane strain eigenvalue
problem to find the natural frequencies and mode shapes. The material properties used in the
analysis were obtained from experiment. In experiment, a shaker system was used to obtain
the dynamic characteristics of the composite laminates. The size and the position of
delamination and the fiber orientation were varied and compared with results obtained from
analysis.
HAMILTONS PRINCIPLE
The contact effects on the free vibration response of composite laminates with through-thewidth delamination was investigated in this study. Fig. 1 shows schematic of a cantilever
composite laminated plate with through-the-width delamination. Hamiltons principle is
expressed as [22]
t
t12 (T U)dt + t12Wnc dt = 0
(1)
where T is the total kinetic energy of the laminate system; U is the potential energy of the
system; Wnc is the virtual work done by nonconservative forces; () is a variation symbol;
t 1 , t 2 are the times at which the system configuration is known. Here
1 u
T= V 2 T udv
(2)
1
1
U= V 2 T dv = V 2 T [E]dv
(3)
Wnc = WF + WC
(4)
WF = V u T fdv + S u T FdS
(5)

WC = S uA T FCA dA + S u BT FCBdA
c
c
(6)

where u is the displacement vector; u is the velocity; is the density; is the stress; is
the strain; [E] is the matrix of material constants; WF is the work done by the
nonconservative external force; WC is the work done by the contact force; f is the body force;
F is the surface tractions; FCA, FCB and uA, uB are the contact forces and displacements on the
surfaces above and below the delamination; V = VA + VB and S = S A + S B are the volume
and surface of the composite laminate; S c is the contact area.
II - 37

The contact force FC must be in equilibrium as FC = F CA = F CB . Eqn 6 is expressed as

WC = S ( u A u B ) T F C dA
C
(7)
Transformation Matrix
The displacement a A of an arbitrary contact point p on the surface above delamination is
divided into three parts [11, 12] as: (i)gap displacement a p q , (ii)sticking displacement
a q r , (iii)relative sliding displacement a r s , as shown in Fig. 2.
a A = a ps = a pq + a q r + a r s = [ N ]a B + t a S + h = [G ]a + h
(8)
where h is the gap displacement; [N] is the matrix of shape function; a B is the displacement
vector of point q below delamination; t is the tangent vector of contact point below
delamination; a S is the magnitude of relative displacement. Eqn 8 is the constraint condition
of the contact surface above delamination. [G] includes [N] and t and is called the
transformation matrix. a is a combination of a B and a S . If point p doesnt contact with
point on surface below delamination, [G] is a unity matrix, and a = a A , h = 0 .
Finite Element Method
The two-dimensional isoparametric quadratic element was used in analysis. The laminate is
divided into two parts; part A is above the delamination, part B includes regions below the
delamination and area without delamination (see Fig. 3). In Eqn 7, u A u B is the relative
displacement of points above and below delamination; they can be expressed as
a p s a q r , as shown in Fig. 2. After finite element discretization, the work done by the
contact force WC is
J
j=1
j=1

WC = (a ps a q r ) T ( R C ) = (( t T a + h) T ( R C ))
(9)
where J is the element number above the delamination; R C = Sc [ N ] F C dA is the contact

nodal force to be determined; a is the nodal displacement produced by contact force R C .

Both a and R C are to be determined. The total kinetic energy, potential energy and work
done by nonconservative external force can be expressed in terms of [G] and a as
JA
JB
j=1
j=1
JA
JB
j=1
j=1
1
1 a
T = 2 ([G ] + h) T [ M A ]([G ] + h) + 2 () T [ M B ]()
a
a
a
1
1
U = 2 ([G ]a + h) T [ K A ]([G ]a + h) + 2 (a ) T [ K B ](a )
JA
JB
j=1
(10)
(11)
j=1
WF = ([G ]a + h) T (Q ) + (a ) T (Q )
A
B
(12)
where J A and J B are the element numbers of part A and part B; [ M A ] and [ M B ] are mass
matrices; [ K A ] and [ K B ] are stiffness matrices; Q and Q are external forces. Since
A
B
there is only contact force, no external force, on the contact surface, Q = 0 , Q = 0 , the
A
B
nodal contact force can be expressed by the nodal force equilibrium above the delamination as

R C = [ K A ]a A + [ M A ] A . The frictional force on the contact surface is expressed
a
according to Coulombs law as
II - 38

t T R C = sign d n T R C
(13)
where d is the frictional coefficient; t and n are the tangential and normal vectors on the
contact surface. If a is replaced by global parameter a , after substituting Eqns 9 to 13 into
Hamiltons principle and rearrangement for zero gap displacement at contact point h = 0
and no external forces Q = Q = 0 . The system governing equation can be obtained as
A
[ M ][G ] [ K ][G ]
[[G ]T d sign [ n ]T [I]] A
a
+ A
a =0 (14)
[M B ]
[K B ]
Material Properties
The material property of each layer of the composite laminate is assumed as transversely
isotropic. There are total of five material constants: Longitudinal modulus E 1 , transverse
modulus E 2 , shear modulus G12 and Poisons ratios 12 and 23 , which result in the stressstrain relation of each layer in the fiber principal direction. The material property in the
natural coordinate can be obtained by a transformation matrix between the fiber principal
direction and the natural coordinate. For simplicity, assume that the composite laminate is
under plane strain condition. The stress-strain relation in Cartesian coordinate can be obtained
by a transformation matrix between the natural coordinate and Cartesian coordinate.
The Gaussian quadrature was used in numerical integration. For composite laminate, each
layer may have different fiber orientation; therefore, the material properties are integrated
through layers in the thickness direction. The elemental stiffness matrices [ K A ], [ K B ] can
be obtained through the use of material property matrix in Cartesian coordinate and are
assembled to be the global stiffness matrices [ K ], [ K ] in Eqn 14.
A
B
Free Vibration of Laminated Plates
Assuming that a has the following form.
a = e it
(15)
where is the displacement vector which is not a function of time. After substituting Eqn
15 into Eqn 14, for to have nontrivial solution, let
[ M ][G ] 2 [ K ][G ]
det([[G ] T d sign [ n ] T [I]] A
+ A
)=0 (16)
[M B ]
[ K B ]
Eqn 16 is the standard eigenvalue problem, in which the eigenvalue is the natural
frequency of the system. The composite laminates studied have one end fixed and the other
end free. The following boundary conditions were applied to the system.
u i = 0 , v i = 0 , dv i / dx = 0
(17)
where u i , v i are the displacements in the x- and y-direction at the fixed end, respectively.
From the solution obtained from Eqn 16, the eigenvector can be found to represent the
natural mode shapes of laminated plate.
II - 39
EXPERIMENT
In addition to the finite element analysis, the shaker vibration experiments were performed to
verify the analytical results. The size and location of delamination and the fiber orientation
were varied to assess their effects on the dynamic response of the laminated plate. The
composite specimens were made from graphite/epoxy prepregs (0.12mm thick per ply) with
required size and stacking sequence. They were put in an aluminum mold on a temperaturecontrolled hot press. In the upper region of the aluminum mold air pressure was applied on
top of the specimen; while a vacuum pump was used to empty the gas produced during the
curing process. For the delaminated specimens, a Teflon piece (0.1 mm thick) was imbedded
to form an artificial delamination.
The material property of a single composite layer is assumed as transversely isotropic. Five
material constants E 11 , E22 , G12 , v 12 , v 23 are needed, in which v 23 = v 12 is assumed as
conventional [23, 24]. The rest four material constants were measured on a tensile testing
frame according to ASTM D3039 [25]. Each constant was obtained by the average of three
experiments as E 11 =147.27 6.09 Gpa, E22 =13.89 1.04 Gpa, G 12 = 4.87 0.15 Gpa,
v 12 =0.320 0.032.
Shaker Vibration Experiment
The specimens used in shaker vibration experiment had length 200 mm, width 40 mm and 8
layers. One end of specimen was fixed for 50 mm, the test section had 150 mm long. Before
the experiment, proper position for force sensor and sampling points for output signal were
layout on surface of the laminate specimen. The specimens were then fixed on the test fixture
and shaker, amplifier, signal analyzer, force sensor, accelerometer connected as arranged in
Fig. 4. The force sensor was glued on the specimen. The accelerometer was used to measure
the dynamic response at each sampling point. During the experiment, an excitation was
generated by the signal analyzer, amplified and connected to the shaker. The shaker forced
the specimen to vibrate. The signals from force sensor and the accelerometer on the laminate
specimen were amplified and fed into the signal analyzer to obtain the frequency response
function by fast Fourier transform (FFT). The response function was directed into a personal
computer in which the STAR Modal System was used to calculate the natural frequencies and
mode shapes.

In this study, the free vibration of composite laminates with through-the-width delamination
was analyzed by the finite element method and was verified by shaker vibration experiment.
The position and size of delamination and the fiber orientation were varied to evaluate their
influence on the dynamic response of the laminated plates. Fig. 5 shows three-dimensional
mode shapes obtained from shaker vibration experiment for specimen (w=40 mm, L=150
mm, D=1.1 mm, a/L=2/3, d/D=1/8, [ 0 ]8 ). The first and third mode shapes are bending
types. The second mode shape is a torsional one which cant be obtained from plane strain
analysis. Fig. 6 shows the corresponding two-dimensional bending mode shapes from the
middle section. Fig. 7 shows the corresponding natural modes obtained from the finite
element analysis. The analytical first mode shows a bulge-up separation in the middle
delaminated region, while it is not shown in the experimental results due to counter effect of
II - 40
the weight of accelerometer on the weaker upper delaminated region. The second bending
modes agree well in both analytical and experimental results. Because of high noise and poor
signal acquisition at higher frequency, the third bending mode was unable to found in
experiment; however, the analytical third bending mode showed a contact point in the middle
of delamination.
Variation of Delamination Size
The size of delamination was varied for laminated specimens to find its influence on the
dynamic response of composite laminates. The sizes of delamination were a/L=1/4, 1/2, 2/3
and 5/6. The separate movement of the delaminated region occurs for delamination size
larger than 50% of laminate length. Figs. 8 and 9 show the relation between the bending
natural frequencies and delamination size for laminated specimens (w=40 mm, L=150 mm,
D=1.1 mm, d/D=1/8, [0 ]8 ). In Fig. 8, the natural frequencies of the first bending mode
decrease with the delamination size. Larger reduction can be seen for delamination size larger
than 2/3. The trend of analytical results was consistent with the experimental ones. In Fig. 9,
the natural frequency of the second bending mode for a/L= 1/2 and 2/3 were larger than those
of laminates without delamination, since separating movement occurred for the delaminated
region. The bulge-up of delaminated region increases the bending rigidity of laminate which
results in higher natural frequency for a/L=1/2 and 2/3. As the length of delamination
increases, the upper delaminated region becomes weaker; therefore, the additional rigidity
obtained from the bulge-up movement was smaller and natural frequencies decrease as also
observed in experiment.
Variation of Delamination Position
The position of the delamination in the thickness direction of specimens was varied to
evaluate its effect on the dynamic response of the laminate plates. The positions of
delamination are d/D=1/8, 2/8 and 4/8. Figs. 10 and 11 show the relation between the natural
frequencies and delamination position for laminated specimens (w=40 mm, L=150 mm,
D=1.1 mm, a/L=2/3, [ 0 ]8 ). The natural frequencies of the first bending modes decrease as
the delamination position toward the center of specimen. The experimental natural
frequencies for specimens with d/D=1/8 is lower than those of the specimens with d/D=2/8
due to the difficulty in manufacturing the former specimens. The specimens with d/D=1/8
probably changed their local thickness at the delaminated region. For the second bending
modes, the natural frequencies for specimens with d/D=1/8 and d/D=2/8 are higher than those
of specimens without delamination. This can be explained from separating movement for
mode shapes of specimens at the delaminated region. The bulge-up delamination gave extra
bending rigidity for the composite laminates. For the case with delamination position at the
mid-thickness d/D=4/8, no separating movement found in the bending mode; thus, the
bending rigidity decrease and natural frequencies decrease as well.
Variation of Fiber Orientation

The delaminated plate specimens (w=40 mm, L=150 mm, D=1.1 mm, a/L=2/3, d/D=1/8)
were used with variation of fiber orientation to evaluate its effect for laminates with []2S
stacking sequence. The stacking sequences were [0 ]8 , [ 30 ] 2S , [ 45 ] 2S , [ 60 ] 2S and
II - 41
[ 90 ] 2S . Figs. 12 and 13 show bending natural frequencies for specimens with different
fiber orientation. The bending natural frequencies decrease with fiber orientation for both
first and second bending modes. The natural frequencies decrease more quickly for
between 0 and 45 . The plane strain finite element method can only predict reasonable
results for cross-ply laminated plates which is symmetric about the middle section in the
longitudinal direction.
CONCLUSIONS
In this study, the free vibration of composite laminates with through-the-width delamination
was investigated analytically and experimentally. The two-dimensional finite element
analysis was used to calculate the natural frequencies and mode shapes for the delaminated
composite plates. The shaker vibration experiment was carried out to verify the analytical
results. The following conclusions can be made:
(1)
The natural frequencies of the delaminated plates decrease as delamination size
increases. This effect is more significant for cases with delamination near upper
surface of specimen and size larger than a/L=2/3.
(2)
For laminated plates with small delamination, no local separating movement at
delaminated region occurs and their natural frequencies are close to composite plates
without delamination.
(3)
The natural frequencies of the delaminated plates usually are smaller than those of
laminated plates without delamination; however, for some cases e.g., a/L=2/3,
d/D=1/8, the local separating movement adds bending rigidity which increases the
natural frequencies to a higher value than those of laminates plates without
delamination.
(4)
When the delamination position is at mid-thickness of the laminated plate, no local
separating movement occurs and the natural frequencies are much lower than those
laminated plates without delamination.
(5)
The natural frequencies of the delaminated plates decrease as the fiber orientation
increases for laminates with stacking sequence [ ]2s .
REFERENCES
1
2.
3.
4.
II - 42
Baharlou, B. and Leissa, A. W., "Vibration and Buckling of Generally Laminated

Composite Plates with Arbitrary Dage Conditions," International Journal of
Mechanical Sciences, Vol. 29, No. 8, pp. 545-555, 1987.
Bhimaraddi, A., Non-Linear Free-Vibration Analysis of Composite Plates with
Initial Imperfections and Inplane Loading," International Journal of Solids and
Structures, Vol. 25, No. 1, pp. 33-43, 1989.
Bowlus, J. A., Palazotto, A. N. and Whitney, J. M., "Vibration of Symmetrically
Laminated Rectangular Plate Considering and Rotatory Inertia," AIAA Journal, Vol.
25, No. 11, pp. 1500-1511.
Bicos, A. S. and Springer, G. S., "Analysis of Free Damped Vibration of Laminated
Composite Plates and Shells," International Journal of Solids and Structures, Vol. 25,
No. 2, pp. 129-149, 1989.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
Dhanaraj, R. and Palaninathan, "Free Vibration of Initially Stressed Composite

Laminates," Journal of Sound and Vibration, Vol. 142, pp. 365-378, 1990.
Ju, F., Lee, H. P. and Lee, K. H., "Free-Vibration Analysis of Composite Beams with
Multiple Delaminations," Composites Engineering, Vol. 4, pp. 715-730, 1994.
Shen, M. H. H. and Grady, J. E., "Free Vibration of Delaminated Beams," AIAA
Journal, Vol. 30, No. 5, pp. 1361-1370, 1992.
Tracy, J. J. and Pardoen, G. C., "Effect of Delamination on the Natural Frequencies
of Composite Laminate," Journal of Composite Materials, Vol. 23, pp. 1200-1215,
1988.
Ju, F., Lee, H. P. and Lee, K. H., "Finite Element Analysis of Free Vibration of
Delaminated Composite Plates," Composites Engineering, Vol. 5, pp. 195-209, 1995.
Ju, F., Lee, H. P. and Lee, K. H., "Dynamic Response of Delaminated Composite
Beams with Intermittent Contact in Delaminated Segments," Composites
Engineering, Vol. 4, pp. 1211-1224, 1995.
Chen, W. H. and Yeh, J. T., "Finite Element Analysis of Finite Deformation Contact
Problems with Friction," Computers & Structures, Vol. 29, pp. 423-436, 1988.
Chen, W. H. and Yen, J. T., "Three-dimensional Finite Element Analysis of Static
and Dynamic Contact Problems with Friction," Computers & structures, Vol. 35, No.
5, pp. 541-552, 1990.
Chen, W. H., Chang, C. M. and Yeh, J. T., "An Incremental Relaxation Finite
Element Analysis of Viscoelastic Problems with Contact and Friction," Computer
Methods in Applied Mechanics & engineering, Vol. 109, pp. 315-329, 1993.
Pascoe, S. R. and Mottershead, J. E., "Two New Finite Element Contact
Algorithms," Computers & Structures, Vol. 32, No. 1, pp. 137-144, 1989.
Wanxie, Z. and Suming, Z., "A Finite Element Method for Elasto-plastic Structures
and Contact Problems by PQP," International Journal for Numerical Method in
Engineering, Vol. 26, pp. 2723-2738, 1988.
Gallego, F. J. and Anza, J. J., "A Mixed Finite Element Model for the Elastic Contact
Problem," International Journal for Numerical Method in Engineering, Vol. 28, pp.
1249-1264, 1989.
Pascoe, S. K. and Mottershead, J. E., "Linear Elastic Contact Problems Using Curved
Element and Including Dynamic Friction," International Journal for Numerical
Method in Engineering, Vol. 26, pp. 1631-1643, 1988
Francavilla, A. and Zienkiewicz, O. C., A Note on Numerical Computation of
Elastic Contact Problem," International Journal of Numerical Method in
Engineering, Vol. 9, pp. 913-924, 1975
Zabaras, N. and Pervez, T., "Viscous Damping Approximation of Laminated
Anisotropic Composite Plates Using the Finite Element Method," Computer Method
in Applied Mechanics and Engineering, Vol. 81, pp. 291-316, 1990.
Maaskant, R. and Gibson, R. F., "Analysis of Damping in Composite Laminates,"
AIAA Journal, Vol. 30, No. 7, pp. 1912-1915, 1991.
Hwang, S. J. and Gibson, R. F., "The Use of Strain Energy-Based Finite Element
Techniques in the Analysis of Various Aspects of Damping of Composite Materials
and Structures," Journal of Composite Materials, Vol. 17, pp. 2585-2605, 1992.
Craig, R. R., Jr., Structural Dynamics: An Introduction to Computer Methods, Wiley,
New York, 1981.
II - 43
23.
Kim, K. S. and Hong, C. S., "Delamination Growth in Angle-Ply Laminated

Composites," Journal of Composite Materials, Vol. 20, pp. 423-438, 1986.
Whitcomb, J. D., "Parametric Analytical Study of Instability-Related Delamination
Growth," Composites Science and Technology, Vol. 25, pp. 19-48, 1986.
ASTM, "General Products, Chemical Specialties, and End Use Products," D3039,
Annual Book of ASTM Standards Standard Test Method foe Tensile Properties of
Fiber-Resin Composites, Section 3, Vol. 15.03, pp. 162-165, 1983
24.
25.
Fig. 1: Schematic of a cantilever composite plate with through-the-width delamination.
Fig. 2: Schematic of contact nodal displacement.
Fig. 3: Schematic of finite element mesh for delaminated plate.
II - 44
II - 45
II - 46
THE DYNAMIC RESPONSE OF INELASTIC,

DELAMINATED COMPOSITE PLATES
Todd O. Williams and Frank L. Addessio
T-3 Theoretical Division
Los Alamos National Laboratory
Los Alamos, New Mexico 87545, USA
SUMMARY: The dynamic behavior of metal matrix composite (MMC) plates is considered.
In particular, the influence of inelastic deformations and delamination at the interfaces of the
lamina on the macroscopic and local response of Al2O3/Al plates are studied. The work is
carried out using a recently developed plate theory which models both delamination and
localized history-dependent effects such as inelasticity. A linear debonding model for the
interface is employed for the current work. The theory models both the initiation and growth
of delaminations without imposing any restrictions on the location, size, or direction of
growth of the delamination. In the current work the response of the individual lamina in the
plate are modelled using the Method of Cells (MOC) micromechanical model. The inelastic
behavior in the matrix is modelled using the unified viscoplastic theory of Bodner and
Partom.
The behavior of a Al2O3/Al plate under dynamic cylindrical bending subjected to a ramp and
hold type of loading is examined. For simplicity, the plate is assumed to be composed of a
cross-ply layup. It is shown that both inelastic deformations and delamination have a strong
influence on dynamic plate behavior. The inelastic deformations have stronger effect on the
axial displacement while delamination has greater influence on the deflection.
KEYWORDS: delaminated composite plates, interfacial constitutive laws, dynamic plate
behavior, micromechanics, method of cells, viscoplasticity
INTRODUCTION
II - 47
II - 48
II - 49
II - 50
II - 51
II - 52
II - 53
II - 54
VIBRATION ANALYSIS OF THICK LAMINATED

TRAPEZOIDAL PLATES
C.C. Chen1, C.W. Lim1, S. Kitipornchai1, and K.M. Liew2
1
Department of Civil Engineering, The University of Queensland, Qld 4072, Australia

Division of Engineering Mechanics, School of Mechanical and Production Engineering,
Nanyang Technological University, Singapore 639798
SUMMARY: To account for the effect of transverse shear deformation, the p-Ritz method
incorporating Reddys third-order shear deformation theory has been developed for the free
vibration analysis of thick, laminated, cantilever, trapezoidal plates. In the proposed p-Ritz
method, sets of uniquely defined polynomial functions are used as the admissible trial
displacement and rotation functions. The admissible function consists of the product of a twodimensional function and a basic function which is defined by multiplying the equations of
the plates prescribed boundary. The total energy is derived by using Reddys third-order
shear deformation theory which is able to predict the vibration characteristics of thick
laminates more accurately. A convergence study has been carried out to verify the accuracy of
the results. The results, wherever possible, are compared with published solutions to
demonstrate the versatility and applicability of the proposed method.
KEYWORDS: free vibration, laminate, plate, higher-order shear deformation theory, p-Ritz
method
INTRODUCTION
The characteristics of high tensile strength and light weight has made the laminated plates
gained their popularity in aerospace and marine engineering applications. Therefore, the
ability to accurately determine the structural behavior of these panels is of paramount
importance. The use of shear deformation theories for the analysis of thick laminated
composite plates has emerged because of the inadequacy of the classical laminated plate
theory [1] based on the Kirchhoff-Love hypothesis. The most popular shear deformation
theory is the first-order shear deformation theory [2] which was extended to laminates by
Yang et al.[3] In this theory, a shear correction factor is assumed a priori and it doesnt
account for surfaces free of transverse shear strains. In an effort to circumvent the drawbacks
of the first order shear deformation, various second and higher-order shear deformation
theories have been proposed. The most common higher-order theory was pioneered by
Levinson [4] and Reddy [5,6] which assumes transverse inextensibility. The present method
adopts the simple higher-order theory based on the pioneering work of Reddy [5].
The need of wing-type structures has facilitated the development of trapezoidal laminates. Not
much research in this topic have been done so far. Lee and Lee [7] applied the finite element
method incorporating the first-order theory to investigate the vibration problem of moderately
thick, cantilevered, trapezoidal laminates. Kapania and Lovejoy [8] used Chebychev
II - 55
polynomials as displacement functions in the Rayleigh-Ritz method and utilizied the firstorder theory to study the free vibration of moderately thick, generally laminated, cantilevered,
quadrilateral plates. Liew and Lim [9] integrated the classical theory and p-Ritz method for
the problem of thin trapezoidal laminates. The p-Ritz method which developed by Liew [10]
avoids the difficulty of mesh generation and continuity conditions required in discretization
methods and is applicable to laminates with free, simply supported and clamped edge
conditions. In this study, the p-Ritz method incorporating Reddys third-order shear
deformation theory is proposed for the vibration analysis of cantilevered, thick, laminated,
trapezoidal plates. Convergence of frequency parameters has been carried out to ensure
adequate terms in the displacement and rotation functions. The effects of aspect ratio, taper
ratio, plate geometry and stacking angle on the thick, cantilevered, trapezoidal laminate are
also investigated. The results, wherever possible, are compared with published solutions to
evaluate the validity and to demonstrate the applicability of the proposed method.
MATHEMATICAL FORMULATIONS
Preliminary Definitions
Fig. 1 illustrates a thick, cantilever, laminated, trapezoidal plate composed of N orthotropic
laminae oriented at angle k with respect to the x-axis and with the geometric configuration
of
length-to-thickness ratio a/h,
y
aspect ratio a/b, taper ratio
b/4
c/b, skew angle , and total
thickness h. The reference
Cartesian coordinate system

is located at the mid-plane of
the laminated plate. The
angle which measures

c/2
positive in the clockwise
b
direction, is the degree of
x
skewness of the trapezoidal
c
laminate.
Another
two
definitions of the skew

angles, and , are also
used for the convenience of
a
comparison. The laminae are
cross-plied
or
stacked
symmetrically with respect to
Fig. 1: Laminated trapezoidal plate geometry
the middle surface of the
laminated plate. The interior
angles of laminate geometries are assumed to be no larger than 90 degrees. The vibration
frequencies of the trapezoidal laminate subjected to a variety of skewness angles, aspect
ratios, taper ratios, and stacking angles are to be determined.
II - 56
Energy Expressions
For a thick trapezoidal laminate undergoing small-amplitude vibration, the spatial
displacement at a general point may be resolved into u, v (in-plane motion) and w (out-ofplane motion) components, respectively. By using Reddys higher-order shear deformation
theory, the displacement field along the x-, y-, z-axis can be expressed in the following form:
w( x , y , t )
4z 3
u ( x , y , z , t ) = u 0 ( x , y , t ) + z x ( x , y , t ) 2 x ( x , y , t ) +
x
3h
w( x , y , t )
4z 3
v ( x , y , z , t ) = v 0 ( x , y , t ) + z y ( x , y , t ) 2 y ( x , y , t ) +
y
3h
w( x , y, z , t ) = w0 ( x , y, t )
(1)
where u0 , v 0 , w0 , x , y are the displacement components of the mid-plane along the x-, y-,
and z-directions and rotations of the mid-plane about the y- and x-direction, respectively. For
free vibration problem, the deflection and rotation functions of the laminate mid-plane are
sinusoidal in time, that is
u0 ( x , y , t ) = U ( x , y ) sin t
v 0 ( x , y , t ) = V ( x , y ) sin t
w0 ( x , y , t ) = W ( x , y ) sin t
x ( x , y , t ) = u ( x , y ) sin t
y ( x , y , t ) = v ( x , y ) sin t
(2)
If the laminated plate is made of elastic materials, the strain energy for each ply is given by
Uk =
1
'N k T k dV k
2
(3)
where Uk and Vk are the strain energy and the volume of the kth lamina, respectively.
Neglecting normal stress k , the total strain energy for the entire laminated plate is
U=
hk
1 N
hk 1 ( x x + y y + xz xz + yz yz + xy xy ) k dz dA
2 k =1 A
(4)
Also, the total kinetic energy T associated with the vibration of laminated plate is
u 2 v 2 w 2
h N
T = k + +
dA
2 k =1 A t
t
t
(5)
in which k is the mass density of the kth lamina.

The energy components U and T can be further simplified by introducing the equivalent
modulus for a multidirectional lamina
II - 57
( Aij , Bij , Dij , E ij , Fij , Hij ) =

k =1
hk + 1
hk
(Q ) (1, z, z , z , z , z )dz
2
ij
(6)
(Q ) are obtained by transforming the stacking angle and the stiffness

constants, (Q ) , which are related to the material properties, E , E , , , G , G , G , of
each ply. Eqns (4) and (5) can be further expanded in terms of (U ,V , W , , ) by
in which,
ij
ij
12
21
12
13
23
substituting the strain-displacement relationship, stress-strain relationships and Eqn (2). As

the vibration is periodic in time, we are able to obtain the maximum strain energy Umax and
maximum kinetic energy Tmax in a vibratory cycle. Finally, the total energy functional for
the free vibration of laminated plate is defined as
= Umax - Tmax
(7)
which is to be minimized by the Rayleigh-Ritz method to obtain the frequency.

p-Ritz Method
The displacement and rotation components, U(x,y), V(x,y), W(x,y), u(x,y), and v(x,y), can be
further simplified by using the non-dimensional expressions
x
y
, =
a
b
(8)
In the Rayleigh-Ritz method, the displacement and rotation functions should be expressed in
the form of a series containing unknown coefficients,
m
U ( , ) = ciu iu ( , )
i =1
m
V ( , ) = civ iv ( , )
i =1
m
W ( , ) = ciw iw ( , )
i =1
m
u ( , ) = ci u u ( , )
i
i =1
m
v ( , ) = ci v v ( , )
i
(9)
i =1
where ciu , civ , ciw , ci u , ci v are the unknown coefficients.

Now, using the p-Ritz method, we assumed the p-Ritz shape functions, i u , i v , i w , i u , i v
to be the products of two-dimensional polynomial series and basic functions. The twodimensional polynomial series f i ( , ) are defined as
II - 58
f ( , ) =
i =1
q i
(10)
q =0 i = 0
where p is the highest degree of the set of two-dimensional polynomials.

To ensure automatic satisfaction of geometric boundary conditions, the basic functions are
assumed to be consist of products of boundary expressions of the laminated plate raised to
their associated basic powers. Thus, for the trapezoidal laminated plate shown in Fig. 1, the
basic functions can be structured as shown below
1
b = +
1
1
1 3
c
c
a
+ tan 1 + 1 +
2 b
4 b
2
b
1 c
1
c
a
+ tan + 1 1 +
2 b
4 b
b
(11)
where denotes the associated basic power. If the cantilevered plate clamped at = -1/2,
s
the basic powers are defined as follows,
s
= 0 , for = u, v , w, u , v and s = 2,3,4
s
1 = 1 , for = u, v , w, u , v
w
1 = 2 ,
(12)
The total energy functional derived in equation (7) are minimized with respect to the
unknown coefficients leading to the following governing equation,
K uu
D0
K uv
K vv
0
0
K ww
sym.
0
0
K wu
K uu
M uu
0
K wv
K uv
K v v
sym.
0
0
M ww
M vv
0
0
M w u
M uu
0 c u
0 c v
M wv c w = 0
0 ac u
M vv bc v
(13)
with
D0 =
Q11 h 3
12(1 12 21 )
(14)
Details of the stiffness and mass matrices can be found in [11]. If the laminae are made of the
same material with mass density per unit volume and the same thickness h, the nondimensional eigenvalue becomes
h
(15)
= ab
D0
The vibration frequencies and mode shapes of trapezoidal laminates are obtained by solving
.
II - 59
NUMERICAL STUDIES AND DISCUSSION

Numerical computation has been carried out in double precision on a SGI PowerChallenge
computer. The results have been compared with available solutions of first-order shear
deformation theory. Excellent agreement exists between the present method and previous
solutions. In all examples, laminates have been assumed to be equal thickness and made of
non-dimensional material properties, E1/E2 = 25, G12/E2 = 0.5, G23/E2 = 0.2, G13 = G12, and
= 0.25.
Table 1:
Convergence of frequency parameters of 6-ply trapezoidal laminates with

a/b = 1.0, c/b = 0.3, stacking sequence [302/0] s, a/h = 5, and = 0
Mode sequence number
4
5
6
8
10
11
12
13
14
15
2.3576
2.3575
2.3575
2.3575
2.3575
2.3575
2.3575
4.6118
4.6110
4.6108
4.6108
4.6107
4.6107
4.6107
7.1591
7.1582
7.1581
7.1580
7.1580
7.1580
7.1579
7.5668
7.5668
7.5667
7.5667
7.5667
7.5667
7.5667
9.7511
9.7488
9.7486
9.7484
9.7483
9.7483
9.7483
10.8658
10.8596
10.8584
10.8578
10.8573
10.8571
10.8569
13.4282
13.4223
13.4218
13.4215
13.4213
13.4213
13.4212
15.3492
15.3365
15.3344
15.3337
15.3332
15.3330
15.3328
The trapezoidal laminated plates with aspect ratio 1, taper ratio 0.3, length-to-thickness ratio
5, = 0 , and stacking sequence [302/0]s were solved in the first example. Convergence
studies have been carried out to check the numerical accuracy of this method. The degree of
polynomial p was increased from 8 to 15 to study its effect on the non-dimensional frequency
parameters . As shown in Table 1, the difference between the results for p = 11 and p = 15
are less than 0.02%. Apparently, the degree of polynomial p = 15 is enough to ensure the
convergence of the results and it is used in all subsequent examples.
The effects of aspect ratio, taper ratio and stacking angle on the frequency parameters of
moderately thick symmetric laminates have been investigated in the second example. The
present method is further verified by comparing the results for frequency parameters with the
solutions from Kapania and Lovejoy[8] who applied Chebychev polynomials and first-order
shear deformation theory. As shown in Table 2, the results are in excellent agreement with
previous solutions. It is noted that the frequency parameter decreases with increasing aspect
ratio or taper ratio. The influence of aspect ratio indicates that cantilevered plates with higher
a/b are structurally weaker than those with lower a/b. As for the effect of taper ratio, the
authors believe that the reduction in mass for lower c/b leads to the increase of the frequency.
In the previous example, the increase in stacking angle caused different change in frequency
parameter. Therefore, a cursory examination of the influence of stacking angles and taper
ratios on the vibration of thick trapezoidal laminates is required. In the third example, the
laminates are assumed to be stacked symmetrically with stacking sequence [ / ]s, a
length-to-thickness ratio of 5, and plate geometry = 0. By comparing the lowest four
frequency results , the effects of taper ratio and stacking angle for laminates with different
geometry have been shown in Fig.2. Again, it confirms that a lower taper ratio results in
higher frequency parameter. Also note that the fundamental frequencies decrease as stacking
II - 60
angle increases. For all modes except the fundamental mode, higher frequencies are likely to
occur with the stacking angle between 15 and 45 degrees.
Table 2: Effects of aspect ratio, taper ratio and stacking angle on the four lowest frequency
parameters, 1 = a 2 / h / E 2 , of laminated trapezoidal plates with = 0, stacking
sequence [ / / 0 ] s and a/h = 10
a/b c/b
1
Ref. 8
0
30
45
0.50 0
30
45
1.00 0
30
45
Present
Mode sequence number

2
3
Ref. 8 Present Ref. 8 Present
4
Ref. 8
Present
1 0.25
6.85
5.21
3.7
6.20
4.59
3.20
5.58
4.01
2.75
6.904
5.224
3.704
6.252
4.603
3.206
5.628
4.023
2.759
11.9
16.4
14.9
9.05
12.7
13.2
6.10
8.21
8.45
12.029
16.431
14.826
9.136
12.681
13.160
6.153
8.218
8.444
14.9
19.5
17.3
13.5
18.3
14.3
9.53
14.2
13.1
14.901
19.704
17.402
13.451
18.402
14.227
9.592
14.250
13.116
22.5
32.4
32.8
18.7
27.1
29.2
11.6
19.4
20.1
23.091
32.141
32.319
18.896
26.967
28.816
11.574
19.481
20.026
2 0.25
2.05
1.34
0.885
1.80
1.16
0.755
1.55
0.987
0.640
2.052
1.343
0.884
1.803
1.155
0.754
1.547
0.987
0.639
4.93
5.85
4.04
3.53
5.58
3.86
2.10
3.85
3.72
4.945
5.829
4.021
3.536
5.557
3.849
2.106
3.852
3.712
7.14
8.24
8.65
6.41
6.19
6.50
5.46
5.35
4.08
7.131
8.259
8.634
6.399
6.209
6.489
5.455
5.345
4.072
8.67
12.1
10.0
8.32
10.9
9.87
6.93
9.49
8.73
8.761
12.091
10.886
8.408
10.895
9.852
6.935
9.470
8.709
5 0.25
0.361
0.201
0.128
0.308
0.170
0.108
0.255
0.143
0.0901
0.361
0.201
0.128
0.308
0.170
0.108
0.255
0.143
0.0900
1.46
0.957
0.621
1.05
0.910
0.588
0.614
0.877
0.566
1.461
0.955
0.619
1.050
0.908
0.586
0.614
0.876
0.564
1.75
2.18
1.61
1.63
1.98
1.60
1.54
1.47
1.57
1.753
2.119
1.597
1.633
1.933
1.589
1.544
1.467
1.548
2.24
2.46
1.92
2.03
2.39
1.75
1.73
1.78
1.58
2.176
2.442
1.853
1.969
2.387
1.694
1.713
1.760
1.576
0
30
45
0.50 0
30
45
1.00 0
30
45
0
30
45
0.50 0
30
45
1.00 0
30
45
Consideration is now given to the effect of taper ratios and plate geometry on the 8-ply antisymmetrically cross-plied trapezoids. Numerical results have been illustrated in Fig. 3. As
anticipated, an increase in taper ratio leads to a decrease in the frequency parameter. The
fundamental frequency parameter decreases as the plate geometry increases. However, for
the rest of the modes, the effect of plate geometry on the frequency parameter is significant.
From the numerical results of the third and fourth examples, we may conclude that an
increase in aspect ratio or taper ratio results in an decrease in frequency parameter. As for the
effect of stacking angle on the frequency parameters, no theoretical explanation is available,
to the authors knowledge. However, from the results of this study, we may conclude that: 1)
II - 61
for the bending modes, the frequency parameter decreases when stacking angle increases; 2)
for other modes except torsional modes, the frequency parameter decreases when stacking
angle approaches 0 or 75 and maximum frequency values are most likely to occur when
stacking angle is between 15 and 45; 3) for the torsional modes, the change of frequency
parameter is influenced by the plate geometry.
c/b=0.1,Mode1
c/b=0.1,Mode2
c/b=0.1,Mode3
c/b=0.1,Mode4
c/b=0.3,Mode1
c/b=0.3,Mode2
c/b=0.3,Mode3
c/b=0.3,Mode4
c/b=0.5,Mode1
c/b=0.5,Mode2
c/b=0.5,Mode3
c/b=0.5,Mode4
c/b=0.7,Mode1
c/b=0.7,Mode2
c/b=0.7,Mode3
c/b=0.7,Mode4
c/b=0.9,Mode1
c/b=0.9,Mode2
c/b=0.9,Mode3
c/b=0.9,Mode4
12
frequency parameter
10
8
6
4
2
0
0
15
30
45
stacking angle
60
75
Fig. 2: Effects of taper ratio c/b and stacking angle on the first four non-dimensional
frequency parameters of laminated trapezoidal plates with
stacking sequence[ / ] s , = 0 and a/h = 5
8
Mode1
=0
Frequency parameter
7
6
Mode2
Mode3
Mode4
Mode1
=0
4
3
Mode2
Mode3
Mode4
Mode1
=0
1
0
Mode2
Mode3
Mode4
0.1
0.3
0.5
c/b
0.7
0.9
Fig. 3: Effects of taper ratio c/b on the first four non-dimensional frequency parameters of
laminated trapezoidal plates with stacking sequence [0/90] 4 and a/h = 5
II - 62
CONCLUSIONS
This paper presents an approximate method based on the Rayleigh-Ritz energy approach for
the vibration analysis of thick, cantilevered, stacked symmetrically or cross-plied trapezoidal
laminates. In this method, sets of two-dimensional p-Ritz shape functions are employed as the
admissible displacement and rotation functions. Since the boundary conditions have been
formulated as the intrinsic components of the p-Ritz functions, the boundary conditions are
automatically satisfied at the outset. Several examples have been verified and the results are in
close agreement with available published literature. It can be concluded that an increase in
aspect ratio or taper ratio results in a decrease in frequency parameter. In addition, both plate
geometry and stacking angle should be considered as their influence on the frequency
parameter are significant.
REFERENCES
1. Leissa, A.W., Vibration of Plates, NASA SP-160, Scientific and Technical Information
Division, Office of Technology Utilization, NASA, Washington D. C., 1969.
2. Reissner, E., The Effect of Transverse Shear Deformation on The Bending of Elastic
Plates, Journal of Applied Mechanics, Vol. 12, 1945, pp. A69-A77.
3. Yang, P.C., Norris, C.H., and Stavsky, Y., Elastic Wave Propagation in Heterogeneous
Plates, International Journal of Solids and Structures, Vol. 2, 1966, pp. 665-684.
4. Levinson, M., An Accurate Simple Theory of The Statics and Dynamics of Elastic
Plates, Mechanical Research Communications, Vol. 7, 1980, pp. 343-350.
5. Reddy, J.N., A Simple Higher-Order Theory for Laminated Composite Plates, Journal
of Applied Mechanics, Vol. 51, 1984, pp. 745-752.
6. Reddy, J.N. and Phan, N.D., Stability and Vibration of Isotropic, Orthotropic and
Laminated Plates According to A Higher-Order Shear Deformation Theory, Journal of
Sound and Vibration, Vol. 98, 1985, pp.157-170.
7. Lee, I., and Lee, J.J., Vibration Analysis of Composite Plate Wing, Computers and
Structures, Vol. 37, No. 6, 1990, pp. 1077-1085.
8. Kapania, R.K., and Lovejoy, E.L., Free Vibration of Thick Generally Laminated
Cantilever Quadrilateral Plates, AIAA Journal, Vol. 34, No. 7, 1996, pp.1474-1486.
9. Liew, K.M., and Lim, C.W., Vibratory Characteristics of General Laminates, I:
Symmetric Trapezoids, Journal of Sound and Vibration, Vol. 183, No. 4, 1995, pp. 615642
10. Liew, K.M., The Development of 2-D Orthogonal Polynomials for Vibration of Plates,
Ph.D. Thesis, National University of Singapore, Singapore, 1990.
11. Chen, C.C., Liew, K.M., Lim, C.W., and Kitipornchai, S., Vibration Analysis of
Symmetrically Laminated Thick Rectangular Plates Using the Higher-Order Theory and
p-Ritz Method, Journal of the Acoustical society of America, in press, 1997.
II - 63
DYNAMIC RESPONSES OF ORTHOTROPIC

ANNULAR COMPOSITE PLATES
C.L.D. Huang
Department of Mechanical Engineering, Durland Kansas State University Manhattan,
Kansas 66506-5106 USA
SUMMARY: Using Hamilton's principle, the nonlinear partial differential for large
oscillations and the associated boundary conditions, for thick circular composite plates, have
been derived. A high-order shear deformation thick plate theory rather than a first-order shear
deformation plate theory is implemented in the basic equations. The applications of the finite
element method to the dynamic problems rely on the use of a variational principle to derive
the necessary element properties and motion equations. The assembled equations for an
eigenvalue problem. A numerical technique of using an incremental and iterative scheme is
utilized for calculating frequency responses. The HSG-PRD-HSG and PRD-HSG-PRD
annular composite plate with free-clamped boundary conditions are studied as examples and
their dynamic responses are obtained.
KEYWORDS: non-linear dynamic response, thick annular composite plates
INTRODUCTION
Composite plates are made of laminae, the plates can be layers of isotropic or anisotropic
materials which material properties may or may not depend on orientation. The plates are
heterogeneous in thicknesswise because of difference material layers. Also the thickness of
plates may be thick enough for which the "thin" plate theory might not be sufficient to
describe the physical models. Thus, the equations of motion of a heterogeneous anisotropic
thick plate are needed. Using Hamilton's principle, the nonlinear partial differential equations
for large oscillations and the associated boundary conditions have been derived. A higherorder shear deformation plate theory (the inplane displacements are expressed as a cubic
polynomial functions of thickness coordinate and the transverse deflection is assumed to be a
constant) rather than a first-order shear deformation plate theory is implemented in the basic
equations. The applications of the finite element method to the dynamic problems rely on the
use of a variational principle to derive the necessary element properties and equations. The
assembled finite-element equations for the plate are formed by summing each element
equation obtained in consideration of a single element. Then, the boundary conditions are
imposed on the vector of nodal field variables, so that the appropriate boundary conditions are
satisfied. The assembled equations form an eigenvalue problem. The developed model is
then applied to study an axisymmetric large vibration of heterogeneous orthotropic annular
composite plate. A numerical technique of using an incremental and iterative scheme is
utilized for calculating frequency responses. The HSG-PRD-HSG (High-Strength-Graphite
epoxy sandwiched with two filaments of PRD-49-III developed by E.I. DuPont de Nemours
and Company, Inc.) and PRD-HSG-PRD annular composite plates with free-clamped
boundary conditions, are used as examples, and their dynamic responses are obtained.
The effects of thickness of plate which are due to the shear deformation, rotatory inertia,
inplane deformation and inplane inertia, are examined and discussed.
II - 64
EQUATIONS OF MOTION IN FINITE ELEMENT FORM
II - 65
as:
II - 66
II - 67
II - 68
With the boundary conditions, for linear vibrations, eqn (20) can be solved directly by
numerical computation to obtain the frequencies by using a standard algorithm. For
nonlinear vibrations, the nonlinear part of element stiffness matrix [K](e)NL contains the
unknown function f in eqn (17), and an iterative procedure (Raju and Rao, 1982) [3] must be
used to find their frequency responses. A computer code is written for the purpose of
obtaining numerical results.
NUMERICAL RESULT AND DISCUSSION

In the numerical calculations, the total thickness of the plate h is chosen as 6.35 mm, and the
ratio of outer radius-thickness (r0/h) is taken as 10 and 25, as the thick and thin annular
sandwich plates, (see Figure 1) respectively. Two interior hole sizes are used, i.e. R = r0/r0 =
0.1 and 0.3. The material properties of HSG and PRD-49 III are listed below:
density
kg/m3x103
E11
N/m2x1010
E22
N/m2x1010
E55
N/m2x1010
12
21
High-strength
graphite
epoxy (HSG)
1.558
12.40
1.03
0.55
0.27
0.0224
PRD-490 III
epoxy (PRD)
1.391
7.93
0.41
0.21
0.30
0.0155
Material
With these material properties, the element matrices in equations (18) and (19) and the basic
equation (20), the frequency responses are obtained and plotted in Figures.
Figures 2 and 3 show the linear frequency, affected by the variation of sandwiched thickness
ratio of annular sandwich plates, comprised of PRD 49-III and HSG with r0/h=25 (thin plate)
and 10 (thick plate), respectively. Both show that, for a given sandwiched thickness ratio
value of 2 hf/h, the linear frequency response of a HSG-PRD-HSG annular sandwich plate is
higher than that of a PRD-HSG-PRD annular sandwich plate. The homogeneous HSG
annular plate (2hf(PRD)/h=0 or 2 hf(HSG)/h=1) has a higher frequency response than the
homogeneous PRD annular plate (2hf(HSG)/h=0 or 2hf(PRD)/h=1). The frequency response of a
annular sandwich plate comprised of PRD core with HSG facings is higher than that of a
homogeneous PRD plate. On the contrary, the frequency response of an annular sandwich
plate with HSG core and PRD facing is lower than that of a homogeneous HSG plate. Figure
2 shows the frequency of HSG-PRD-HSG annular sandwich plates increases with an increase
of the sandwiched thickness ratio. The maximum frequency occurs approximately at
2hf/h=0.8 for both hole sizes R=0.1 and 0.3. The frequency decreases slightly when the ratio
2hf/h is greater than 0.8. The frequency at 2hf/h=0 is the frequency of a homogeneous HSG
plate. In Figure 3, the frequency of HSG-PRD-HSG annular sandwich plate increases
monotonically with an increase of the sandwiched thickness ratio, and reaches the maximum
II - 69
value, which is the frequency of the homogeneous plate made of HSG. In comparison of
Figures 2 and 3, the characterics of frequency curves with respect to sandwiched thickness
ratio are similar with some variations. This phenomena is due to the effects of shear
deformation and rotatory inertia which must be included in the thick plate theory.
Figure 4 shows the frequency ratio NL/L versus the sandwiched thickness ratio 2hf/h at the
finite amplitude of wmax/h=1.0. The PRD-HSG-PRD annular sandwich plate has higher
frequency ratio than that of HSG-PRD-HSG annular sandwich plate for 0<2hf/h<0.7. The
maximum frequency ratio NL/L occurs approximately at 2hf/h=0.4 for the annular sandwich
plate comprised of HSG core with PRD facings. At the highest frequency ratio, the
frequencies of nonlinear vibrations are 5.2% and 1.81% higher than those of linear vibrations,
respectively.
REFERENCES
1.
Reddy,J.N,"A Refined Nonlinear Theory of Plate with Transverse Shear Deformation",

J. Non-Linear Mech., Vol.20 (9/11),1984,pp.881-896.
2.
Huang, C.L.D. and Lu, H.Y., "Nonlinear Vibrations of Orthotropic Thick Annular
Plates Using Higher Order Shear Deformation Theory", Composites Engineering, Vol.
1, No. 10, pp. 985-994.
3.
Raju,K.K.and Rao,G.V., "Large Amplitude Asymmetric Vibration of Some Thin Shells

Revolution." Journal of Sound and Vibration. Vol.44,1982,pp.327-333
II - 70
Figure 2: Linear Frequency Responses for Various Thickness Ratios of Thin Plates
II - 71
Figure 3. Linear Frequency Response for Vaious Sandwiched Thickness Ratio of Thick Plates
II - 72
Figure 4. Nonlinear Frequency Response for Various Thickness Ratio of Thin Plates
II - 73
THE DYNAMIC RESPONSE OF A PULTRUDED

FIBERGLASS TUBE SUBJECTED TO LATERAL
LOADING
Louis J. Brunet and James A. Nemes
Mechanical Engineering Department,
McGill University 817, rue Sherbrooke ouest, Montral QC Canada, H3A 2K6
KEYWORDS:
deflection
II - 74
pultruded tubing, impact, quasi-static modeling, lateral loading, large
II - 75
II - 76
II - 77
II - 78
II - 79
II - 80
II - 81
II - 82
II - 83
VIBRATION ANALYSIS FOR COMPOSITE

CYLINDRICAL SHELLS WITH A RECTANGULAR
CUTOUT
Young-Shin Lee and Young-Wann Kim
Department of Mechanical Design Engineering,
Chungnam National University, 200 Kung-dong, Yousung-ku, Tae-jon, 305-764, Korea
KEYWORDS: rectangular cutout, ring stiffened composite cylindrical shell, discrete

stiffener theory, plain weave
II - 84
II - 85
II - 86
II - 87
II - 88
II - 89
II - 90
II - 91
II - 92
II - 93
AN ANALYSIS OF VIBRATION DAMPING IN

COMPOSITE PLATES USING COMPLEX
FREQUENCY EQUATIONS
S. Kalysanasundaram*, D. H. Allen**, M. Cardew-Hall*, A.E. Lowe*
*Department of Engineering, Australian National University, Australia
**Department of Aerospace Engineering, Texas A & University, U.S.A.
KEYWORDS: vibration damping, viscoelasticity, complex frequency equations, specific

damping capacity, cylindrical bending
II - 94
II - 95
II - 96
II - 97
II - 98
II - 99
II - 100
II - 101
II - 102
VIBRATIONS OF A LAMINATED COMPOSITE BEAM

ELASTICALLY SUPPORTED AT ONE END AND
CARRYING A TIP MASS AT THE OTHER
T. Yokoyama and Z. M. Yang
Department of Mechanical Engineering
Okayama University of Science, 1-1 Ridai-Cho, Okayama 700, Japan
SUMMARY: The present work deals with the free vibration analysis of a symmetrically
laminated composite beam with one end elastically supported and having a tip mass at the
other free end. The governing equations for free vibrations of the composite beam are derived
from Hamilton's principle accounting for the effects of shear deformation and rotatory inertia.
The individual influences of ply orientation, beam end constraint, tip mass and distance
between the the tip mass centroid and the point of attachment on the natural frequencies of the
composite beam are examined and presented in three-dimensional graphs. It is shown that the
increase in the (tip mass)/(beam mass) ratio, the (radius of gyration of the tip mass)/(beam
length) ratio and the (distance between the tip mass centroid and the point of
attachment)/(beam length) ratio yields the decrease in the fundamental frequencies. The
validity of the analysis is verified by comparison with the frequency results available in the
literature.
KEYWORDS: laminated composite beam, natural frequency, beam end constraint, tip mass,
ply orientation, coupling effect, timoshenko beam theory
INTRODUCTION
The increasing use of laminated composite beams as robot arms or rotating blades requires a
deeper understanding of vibration characteristics of these beam structures. A beam theory
based on the classical lamination theory [1] is inadequate for accurate prediction of the
natural frequencies of laminated composite beams, because the effects of shear deformation
are pronounced for composites due to the high ratio of extensional modulus (in-plane
modulus) to transverse shear modulus. The bending-torsion coupled vibrations of generally
orthotropic beams have been studied by several authors [2-5] including the effects of shear
deformation and rotatory inertia. The same vibration problem of symmetrically laminated
composite beams was solved by Chen and Yang [6] using finite element approach. A similar
study of orthotropic beams was subsequently performed by Suresh et al. [7] including crosssectional warping effects, neglecting rotatory inertia and shear deformation effects. The free
bending vibrations of symmetrically laminated composite beams were discussed by
Chandrashekhara et al. [8] and Abramovich [9]. Later on, the influence of tip mass on the
natural frequencies of laminated composite beams was considered by Chandrashekhara and
Bangera [10]. Recently, the free vibrations of unsymmetrically laminated composite beams
were analyzed by Krishnaswamy et al. [11], Abramovich and Livshits [12] and Abramovich
et al. [13].
II - 103
The present paper is concerned with the free vibrations of a laminated composite beam
elastically supported at one end and carrying a tip mass at the other free end as a basic model
for a single-joint flexible robot arm or a rotor blade. The governing equations for free
vibrations of the composite beam are derived from Hamilton's principle accounting for the
effects of shear deformation and rotatory inertia. A parametric study is carried out to
elucidate the individual effects of ply orientation, beam end constraint, tip mass and distance
between the tip mass centroid and the point of attachment on the natural frequencies of the
composite beam.
FORMULATION OF PROBLEM
Laminate Analysis
A laminated composite beam is considered, referred to a system of Cartesian co-ordinates (
see Fig.1 ) with the origin on the midplane of the beam and the x-axis being coincident with
the beam axis. The length, width and height of the beam are denoted by L, b and h,
respectively. In an effort to include the coupling effects among bending stiffness, twisting
stiffness and shearing stiffness of the composite beam, the linear elastic plate theory is applied
in this study. Neglecting the in-plane displacements compared with the flexure-induced
displacements, the displacement field for the composite beam can then be assumed as:
(1)
where u, v and w are the displacement components in the x, y and z directions,
respectively,
is the lateral displacement of a point on the midplane, and and are the
rotations of the normals to the cross-sections. The strain-displacement relations are given by
(2)
The curvatures (
) can be written as
(3)
The laminate beam constitutive relations are derived based on the laminate plate theory as
(4)
(5)
II - 104
where (
) are the moments, (
) are the shear force resultants, and
(
) are the transverse shear strains. The laminate stiffness coefficients in Eqs (4) and (5)
are defined by
,
(6,7)
where k is the shear correction factor dependent on the shape of the cross-section [14], and
are the transformed material coefficients from the principal material coordinate system to
the x-y-z coordinate system. For the laminated composite beam under study, the couples
(
) and the resultant
can be assumed to be zero. However, the curvatures (
)
and the shear strain
are assumed to be non-zero. Thus, Eqs (4) and (5) can be rewritten as
,
(8,9)
where
(10)
(11)
Note that the D12 term denotes bending-bending stiffness, the D16 and D26 terms refer to
bending-twisting stiffness, and the A45 term represents shearing-shearing stiffness of the
laminated composite beam.
Governing Equations
The kinetic energy T of the laminated composite beam can be written as
(12)
where and are the translational and rotary inertia of the cross section of the composite
beam, defined by
(13)
and
being the mass density of the composite beam material. The strain energy
is given by
II - 105
(14)
The governing equations for free vibrations of the composite beam can be derived from
Hamilton's principle, i.e.
(15)
where t1 and t2 are two arbitrary time variables and denotes the first variation. Substituting
Eqs (12) and (14) into Eqn (15) and integrating by parts leads to
(16)
Solution Procedure
The harmonic solutions to Eqn (16) can be assumed in the form
(17)
where W (x) is the lateral vibration mode, (x) is the rotational vibration mode and is the
circular frequency. Substitution of Eqs (17) into Eqs (16) yields
(18)
where
(19)
By eliminating W () and () , Eqs (18) can be transformed into two uncoupled differential
equations as
(20)
The solutions to Eqs (20) can be expressed as
II - 106
Case (I)
(21)
where
(22)
Case (II)
(23)
where
. The constants are related by the coupled Eqs (18) as
(24)
Boundary conditions and frequency equation

The associated boundary conditions for the composite beam under consideration may be
written as
at
(25a)
(25b)
(25c)
at
(25d)
where
and are the rotational and translational spring stiffnesses, respectively. J is the
rotational moment of inertia of the tip mass, M is the tip mass, and d is the distance between
the center of gravity of the tip mass and the point of attachment. Using Eqs (17) and the
non-dimensional variable , Eqs (25) may be rewritten as
at
(26a)
(26b)
II - 107
at
(26c)
(26d)
With the help of relations of integration constants in Eqs (24), inserting the solutions Eqs (21)
into the boundary conditions Eqs (26) gives four homogeneous linear algebraic equations. In
order that the solution other than zero exists, the coefficients (i = 1,2 ...4 ) must be equal to
zero. This leads to the determinantal equation as
(27)
in Eqn (27) are omitted because of space limitations. For a given beam,
The coefficients
the frequency coefficients
can be found by solving the frequency equation
(27), and the corresponding natural frequencies i (i = 1, 2, 3...) can then be calculated from
Eqs (19). The roots of Eqn (27) are obtained by applying the false position method [15] with a
tolerance of
.
NUMERICAL RESULTS AND DISCUSSION
In order to verify the validity of the present formulation, we analyze the free vibrations of a
laminated composite beam made of AS/3501-6 graphite-epoxy with the following material
properties [8] :
The short-thick beam ( L/h = 15 or r = 0.019 ) is considered for the purpose of comparison
with the available exact solutions. The beam width b is taken as unity for all the problems.
The shear correction factor k is taken as 5/6 for the rectangular cross-section. The frequency
results are given in the non-dimensional form
(28)
For computational convenience, the following non-dimensional parameters are introduced :
(29)
where I is the second moment of area of the beam cross-section, m is the mass of the beam
and R is the radius of gyration, which is directly related to J = MR 2.
Table 1 shows the first five natural frequencies of a cross-ply laminated composite beam with
different degrees of constraint at one end. It is seen that the natural frequencies of the
composite beam increase with increasing
and
and approach those of a cantilever
composite beam from Ref.[8]. Figure 3 displays the effect of ply orientation on the
fundamental frequency of the composite beam with a rotational and translational constraint at
one end. It is important to note that the fundamental frequency of the composite beam
becomes smallest at the ply orientation [49.4/-40.6]s. Although not shown here, the same
II - 108
tendency is observed for the higher mode frequencies. This may be due to the fact that the
shear deformation parameter s in Eqs (19) for the composite beam has a minimum at this ply
orientation in the region 0o << 90o.
The influences of the various parameters are presented in three-dimensional graphs of the
non-dimensional natural frequency. Figure 4 shows that the tip mass M reduces
significantly the fundamental frequency of the composite beam more than the radius of
gyration of the tip mass R. Figure 5 indicates the effects of the tip mass M and the distance d
between the tip mass centroid and the point of attachment on the fundamental frequency. As
with Fig.4, the tip mass M has a greater effect on the fundamental frequency than the distance
d. Figure 6 represents the influences of R and d on the fundamental frequency of the
composite beam with a constant tip mass at the free end. Obviously the effect of d on the
fundamental frequency is higher than that of R.
CONCLUSIONS
The natural frequencies of a symmetrically laminated composite beam with one end
elastically supported and carrying a tip mass at the other free end have been calculated and
plotted. A parametric study was conducted to examine the individual effects of ply
orientation, beam end constraint, tip mass and distance between the center of gravity of the
tip mass and the point of attachment on the natural frequencies of the composite beam. The
accuracy of the numerical results was confirmed by comparison with the available frequency
results. The present results show that the increase in the (tip mass)/(beam mass) ratio, the
(radius of gyration of the tip mass)/(beam length) ratio and the (distance between the center of
gravity of the tip mass and the point of attachment )/(beam length) ratio yields the decrease in
the fundamental frequencies.
REFERENCES
1.
Vinson, J.R. and Sierakowski, R.L., The Behavior of Structures Composed of

Composite Materials, Kluwer Academic Publishers, Dordrecht, 1987, p.119-147.
2.
Abarcar, R.B. and Cunniff, P.F., "The Vibration of Cantilever Beams of Fibre
Reinforced Material", Journal of Composite Materials, Vol.6, 1972, pp.504-517.
3.
Teoh, L.S. and Huang, C.C., "The Vibration of Beams of Fibre Reinforced Material",
Journal of Sound and Vibration, Vol.51, No.4, 1977, pp.467-473.
4.
Teh, K.K. and Huang C.C., "The Vibrations of Generally Orthotropic Beams, A Finite
Element Approach", Journal of Sound and Vibration, Vol.62, No.2, 1979, pp.195-206.
5.
Teh, K.K. and Huang, C.C., "The Effects of Fibre Orientation on Free Vibrations of
Composite Beams", Journal of Sound and Vibration, Vol.69, No.2, 1980, pp.327-337.
6.
Chen, A.T. and Yang, T.Y., "Static and Dynamic Formulation of a Symmetrically
Laminated Beam Finite Element for a Microcomputer", Journal of Composite
Materials, Vol.19, 1985, pp.459-475.
II - 109
7.
Suresh, J. K., Venkatesan, C. and Ramamurti, V., "Structural Dynamic Analysis of

Composite Beams", Journal of Sound and Vibration, Vol.143, No.3, 1990, pp.503519.
8.
Chandrashekhara, K., Krishnamurthy, K. and Roy, S., "Free Vibration of Composite

Beams Including Rotary Inertia and Shear Deformation", Composite Structures, Vol.14,
1990, pp.269-279.
9.
Abramovich, H., "Shear Deformation and Rotary Inertia Effects of Vibrating Composite
Beams", Composite Structures, Vol.20, 1992, pp.165-173.
10.
Chandrashekhara, K. and Bangera, K. M., "Vibration of Symmetrically Laminated

Clamped-Free Beam with a Mass at the Free End", Journal of Sound and Vibration,
Vol.160, No.1, 1993, pp.93-101.
11.
Krishnaswamy, K., Chandrashekhara, K. and Wu, W.Z.B., "Analytical Solutions to

Vibration of Generally Layered Composite Beams", Journal of Sound and Vibration,
Vol.159, No.1, 1992, pp.85-99.
12.
Abramovich, H. and Livshits, A., "Free Vibrations of Non-Symmetric Cross-Ply

Laminated Composite Beams", Journal of Sound and Vibration, Vol.176, No.5, 1994,
pp.597-612.
13.
Abramovich, H., Eisenberger, M. and Shulepov, O., "Vibrations and Buckling of CrossPly Nonsymmetric Laminated Composite Beams", AIAA Journal, Vol.34, No.5, 1996,
pp.1064-1069.
14.
Cowper, G.R., "The Shear Coefficient in Timoshenko's Beam Theory", Journal of

Applied Mechanics, Vol.33, 1966, pp.335-340.
15.
Conte, S.D. and de Boor, C., Elementary Numerical Analysis: An Algorithmic

Approach, 3rd Ed., McGraw-Hill, 1981, p.72.
II - 110
z
y
h
x
b
L
Fig.1. Coordinate system and dimensions of a laminated composite beam.
kr
J
m
kt
Fig. 1 A composite beam with root flexibilites and tip mass
Fig. 2. A laminated composite beam with a rotational and translational

constraint at one end and carrying a tip mass at the other free end.
Table 1. Non-dimensional natural frequencies of a cross-ply laminated

composite beam with different degrees of constraint at one end ( L/h = 15 )
(M= d=R= 0)
kr = kt
Mode No.
10
[0/90/90/0]
10
10
10
cantilever
(Ref. 8)
0.2492
0.8800
0.9226
0.9231
0.9241
0.9568
3.6710
4.8739
4.8888
4.8925
5.9228
8.1904
11.3731
11.4345
11.4400
13.0597
14.8569
18.5699
18.6920
18.6972
20.9092
22.3594
26.0309
26.2082
26.2118
II - 111
1 .0
6
( kr = kt = 10 , M = 0 )
0 .8
0.6
0.4
B
B
[60/-30]s
[49.4/-40.6] s
[90/0 ]s
[75/-15] s
[4 5/-4 5]s
[15/-75]s
[0/90] s
0.0
[30/-60] s
0 .2
Fig.3. Effect of ply orientation on fundamental frequency of a laminated

composite beam with a rotational and translational constraint at one end.
( k r = k t = 106 , d = 0 )
[0/90/90/0]
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0.0
0.0
0.0
0.0
0.2
0.1
0.2
R
0.3
0.4
0.5 1.0
0.8
0.6
0.4
M
Fig.4. Non-dimensional fundamental frequency versus R and M for a

cross-ply laminated composite beam with a rotational and translational
constraint at one end.
II - 112
( k r = k t = 106 , R = 0 )
[0/90/90/0]
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0.0
0.0
0.0
0.0
0.1
0.2
d
0.4
0.3
0.6
0.4
0.5 1.0
0.8
0.2
Fig.5. Non-dimensional fundamental frequency versus d and M for a

cross-ply laminated composite beam with a rotational and translational
constraint at one end.
( k r = kt = 106 , M = 0.4 )
[0/90/90/0]
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0.0
0.0
0.0
0.0
0.1
0.1
0.2
0.2
0.3
0.3
0.4
0.4
0.5 0.5
Fig.6. Non-dimensional fundamental frequency versus R and d for a cross-ply

laminated composite beam with a rotational and translational constraint at one
end and having a constant tip mass at the other free end.
II - 113
FATIGUE BEHAVIOUR OF COMMINGLED CF/PEEK

COMPOSITES
Lin Ye, Xiaoxue Diao and Yiu-Wing Mai
Centre for Advanced Materials Technology, Department of Mechanical & Mechatronic
Engineering, The University of Sydney, NSW 2006, Australia
SUMMARY: The static and fatigue properties of continuous AS4 carbon fibre (CF)
reinforced polyetheretherketone (PEEK) composite laminates made from commingled
unifabric prepreg were investigated under static tension and tension-tension fatigue loading.
Three different lay-ups were studied, namely [0]16, [90]16 and [02/902]2S. The results for static
stiffness and strength as well as fatigue life and residual strength under different fatigue stress
amplitudes were evaluated, and the damage and failure mechanisms of the composite
laminates were discussed. The wear-out model was applied to fit the S-N curves and estimate
residual strength of composite laminates, and good correlation was obtained. Compared to
composite laminates made from pre-impregnated tape prepreg (e.g. APC-2), the commingled
CF/PEEK composite laminates are more sensitive to fatigue loading because of imperfect
fibre bundle alignment.
KEYWORDS: fatigue behaviour, CF/PEEK composites, commingled prepreg, tensiontension fatigue, damage mechanisms
INTRODUCTION
Fatigue behaviour of a composite structure is an important property for engineering
applications. For many years the primary matrix materials used in polymer-based composites
have been thermosets. High performance thermoplastics have been introduced as matrix
materials in recent years, which have many advantages over their thermosetting counterparts.
In particular, thermoplastic matrices offer the potential for simpler, less time-consuming
manufacturing cycles [1,2]. Continuous carbon fibre (CF) reinforced polyether-ether-ketone
(PEEK) composites constitute one group of the most competitive materials in the aerospace
and defence industries. CF/PEEK prepreg is usually supplied in a form of pre-impregnated
tapes (e.g. APC-2). However, more recently CF/PEEK prepreg has been available in form of a
woven fabric, consisting of unidirectional and continuous carbon fibres mingled with
continuous PEEK fibres, which are interwoven with continuous PEEK fibres in the transverse
direction. This type of prepreg is particularly suitable for manufacturing components of
complex geometry. The fatigue behaviour of CF/PEEK composites produced from this type
of prepreg has not been studied comprehensively. The main objectives of this work are to
investigate the static and fatigue properties of commingled unifabric CF/PEEK composite
laminates under static tension and tension-tension fatigue loading and to compare them with
those of continuous CF/PEEK composite laminates made from the pre-impregnated tape
prepreg (e.g. APC-1, APC-2). The damage and failure mechanisms in the composite
laminates are discussed and the fatigue performance of the laminates is evaluated using the
wear-out model.
II - 114
MATERIAL PREPARATION AND EXPERIMENTAL PROCEDURE

The material used in this study was commingled continuous carbon fibre reinforced PEEK
unifabric (UD AS4 3K/PEEK) supplied by the Polymer Research Division, BASF. A diagram
of the commingled unifabric system is shown in Fig. 1. The diameter of PEEK fibres is
approximately 130 m in the cross-weave and 30 m in the longitudinal direction, where
PEEK fibres were mingled with carbon fibres. The global weight percentage of PEEK
polymer in the preform is 32%. With a density of 1.265 g/cm3 for non-crystallised PEEK [3],
the fibre volume fraction amounts to 55% in a fully consolidated composite.
Carbon Fibre
PEEK Fibre
Commingled
Warp
PEEK Weft
Fig. 1 Schematics of commingled CF/PEEK composite unifabric prepreg.
Temperature [ C]
t=60min
400
o
10 C/min
2 oC/min
Pa = 1MPa
t (Holding Time)
Time [min]
Fig. 2 Processing cycle of CF/PEEK laminates from commingled unifabric prepreg.

Unidirectional prepreg layers were produced through edge-welding the unifabric with a
soldering iron into 199199 mm sheets, with subsequent cutting. The prepreg layers were
then placed in a steel mould of a 200200 mm square cavity in specified stacking sequences
and consolidated using a laboratory hot press. Two different lay-ups were chosen in
manufacturing the panels, i.e. unidirectional [0]16 and [02/902]2S. The 16 layers of prepreg
gave a subsequent panel thickness of approximate 2.2 mm after consolidation. Upilex
polyimide films (50 m in thickness) coated with Frekote-FRP release agent were used
between the prepreg stack and the mould surface to ensure easy removal of the panels after
consolidation. The processing cycles of the composite laminates are outlined in Fig. 2.
Two different geometries of specimens were chosen, i.e. straight-side for static tests and
dog-bone shape for fatigue tests, shown in Fig. 3. The specimens were end-tabbed for
II - 115
testing, bonding aluminium tabs of 40 mm in length and 2 mm in thickness with a CIBAGEIGY Araldite superstrength adhesive.
L
B
W
R
38mm
2.2mm
2mm
Lay-up
[0]16
[0]16
[90]16
[90]16
[02,902]2S
[02,902]2S
38mm
120mm
200mm
W[mm]
14.0
9.5
22.0
22.0
22.0
22.0
B[mm]
14.0
6.8
22.0
18.0
22.0
18.0
L[mm]
120.0
80.0
12.0
80.0
120.0
80.0
R[mm]
100.0
100.0
100.0
Fig. 3 Geometrical dimensions of coupon specimens.

The static tensile tests were performed using Instrons 1195 and 5567 respectively at a
constant cross-head rate of 1.0 mm/min. A clip gauge with a gauge length of 25 mm was
attached to the middle part of specimens during testing to measure the strain response and
consequently determine the stiffness. The fatigue tests were conducted on an Instron 8501
hydraulic serve machine under load control at a frequency of 8 Hz with a stress ratio of 0.1.
Some tests were continued until the specimen failed in order to establish S-N curves, the rest
tests were terminated at specified stages of fatigue life for measurements of residual-tensile
stiffness and strength as well as investigations of damage mechanisms.

Static Properties
From the stress-strain response, the static strength and stiffness of laminates with three
different lay-ups are determined, listed in Table 1, which are compared with the
corresponding properties of XAS/PEEK (APC-1, Vf=55%) [4], AS4/PEEK (APC-2, Vf=60%)
[1] and AS4/Epoxy (Vf=68%) [5] laminates as well as those of a pure PEEK resin [3] in Fig.
4. The longitudinal tensile strength of the CF/PEEK commingled unidirectional composite
laminate is 17% higher than that of APC-1 composite which is of the same fibre volume
fraction as the former, but is clearly lower than those of APC-2 and AS4/Epoxy composites
because these are of higher fibre volume fractions. In principle, the longitudinal tensile
strength of the CF/PEEK commingled composite laminates should be lower than that of
composites of the same fibre volume fraction made from the pre-impregnated tape prepreg,
because in the former the fibres are not stretched out as in the latter so that the fibre
reinforcement is not as effective as in the latter. However, the reason why the CF/PEEK
commingled composite is of higher static tensile strength than APC-1 is probably due to the
fact that the strength of AS4 fibres is about 16% higher than that of XAS fibres [6]. The
longitudinal stiffness of the CF/PEEK commingled composite is about 10% lower than those
of APC-2 and AS4/Epoxy because of higher fibre volume fractions in both composites.
II - 116
The average transverse tensile strength of the CF/PEEK commingled composite is almost the
same to that of APC-2 [1] but is 28% higher than that of AS4/Epoxy composites, indicating
that the failure strength of unidirectional composites under transverse tension is controlled by
the matrix strength. The transverse strengths of all PEEK matrix composites are lower than
that of the pure PEEK resin as shown in Fig. 6. Unlike the longitudinal tensile strength which
is determined almost entirely by the fibre strength, the transverse tensile strength is governed
by many factors including properties of the fibre and matrix, the interface bond strength, the
presence and distribution of voids, the residual stress and interactions between these
mechanisms. The transverse stiffnesses of CF/PEEK, APC-2 and AS4/Epoxy are quite close
to each other, and they are all much higher than that of the pure PEEK resin.
2,500
140
120
Eo [GPa]
o [MPa]
2,000
APC-1
CF/PEEK
APC-2
AS4/Epoxy
PEEK
1,500
1,000
100
APC-1
CF/PEEK
APC-2
AS4/Epoxy
PEEK
80
60
40
500
20
0/90
90
Lay-up
0/90
90
Lay-up
(a)
(b)
Fig. 4. Static strength (a) and stiffness (b) of CF/PEEK commingled

laminates in comparison with those of other composite materials.
Table 1 Properties of commingled CF/PEEK composite laminates
Strength o
Lay-up
Stiffness Eo
Ultimate Strain
Mean [MPa]
CV[%]
Mean [MPa]
CV[%]
Mean [MPa]
CV[%]
[0]16
1507.05
7.20
124.57
3.43
1.17
17.20
[90]16
88.61
9.58
9.03
1.56
1.31
7.57
[02/902]2S
744.17
8.87
68..27
4.73
1.18
6.15
The static uniaxial strength of the CF/PEEK [02/902]2S cross-ply laminate is very close to that
of APC-1 [7] laminate with 11 plies (six in the axial direction and five in the transverse
direction) because of the higher strength of AS4 fibres although a [02/902]2S laminate contains
a less percentage of 0o plies. The uniaxial stiffness of CF/PEEK [02/902]2S laminate is 7%
lower than APC-2 and 13% lower than AS4/Epoxy, but about 10% higher than APC-1.
Fatigue Properties
The experimental S-N data for the CF/PEEK composites with three different lay-ups are
shown in Figs. 5-7 with comparison with those of unidirectional and cross-ply APC-1 [4] as
well as the best fit curve of the pure PEEK resin [3], respectively. Compared with APC-1,
II - 117
the S-N data of the unidirectional CF/PEEK commingled composite exhibit quicker
degradation in Fig. 4, indicating that this material is more fatigue sensitive.
2,000
a [MPa]
1,500
95%
1,000
95%
500
0
Log N f
Fig. 5 S-N curve of unidirectional [0] 16 CF/PEEK laminate (: static data, : fatigue data) in
comparison with APC-1 (1).
120
95%
a [MPa]
100
PEEK
80
60
95%
40
20
0
Log N f
Fig. 6 S-N curve of unidirectional [90] 16 CF/PEEK laminate (: static data, : fatigue data)
in comparison with PEEK resin.
a [MPa]
800
95%
600
95%
400
200
0
Log N f
Fig. 7 S-N curve of cross-ply [02/902] 2S CF/PEEK laminate (: static data, : fatigue data) in
comparison with APC-1 (1).
II - 118
The data of cross-ply laminates shown in Fig. 5 exhibit the same trend, and the experimental
results of APC-1 cross-ply laminates are almost always above those of CF/PEEK laminates
because of a higher percentage of 0o plies. The data of the CF/PEEK [90]16 laminates are
much more scattered. The best fit curve of S-N data for the pure PEEK resin is above all
the data of the CF/PEEK laminates shown in Fig. 6.
Performance Degradation
A series of tests for measurements of residual-tensile stiffness and strength were conducted
on the unidirectional [0]16 and cross-ply [02,902]2S CF/PEEK composite laminates which had
been fatigued for some specified cycles. The residual strength degradation with cycle number
for the CF/PEEK [0]16 and [02,902]2S composite laminates subjected to different fatigue stress
levels is plotted in Fig. 8, which exhibit large scatter in experimental data. There is globally
little degradation in residual strength during most part of fatigue life for both laminates until
the final stage of fatigue damage which causes dramatic degradation in residual strength
followed by catastrophic failure of the laminates.
1,200
1,200
r [MPa]
1,600
r [MPa]
1,600
800
400
400
0
800
3 4 5
Log N
(a)
(b)
800
600
600
r [MPa]
800
r [MPa]
3
4
Log N
400
200
200
0
400
Log N
(c)
Log N
(d)
Fig. 8 Residual strength for [0] 16 unidirectional laminate (a) a=50%o, (b) a=60%o
and [02/902] 2S cross-ply laminate (c) a=60%o, (d) a=70%o, wear-out model,
o: experimental data.
II - 119
Failure Mechanisms
Under both static and fatigue loading, fibre breakage is the main micro-damage mode for the
unidirectional laminate, which occurs either individually or in the form of bundles as shown
in Fig. 9. In the areas where fibres are stretched out, fibres usually break or fragment
randomly, while in the areas where fibre bundles are misaligned, fibre bundles fracture occurs
because of local stress concentration (Fig. 9b). With the increase of fatigue cycle the fibre
bundle fracture occurs continuously. This probably is the reason why the CF/PEEK
commingled laminates is much more fatigue sensitive than APC-1. The fibre/matrix
interfacial adhesion appeared to be quite good because the isolated and broken fibres are
almost never striped but always coated with the PEEK matrix, shown by the scanning electron
micrography (Fig. 9c).
The [02/902]2S cross-ply laminate exhibits multiple damage modes including transverse matrix
cracking in the 90o plies in the first stage of damage development and delamination growth
along the interfaces between 0o and 90o plies originated from the tips of transverse matrix
cracks at the late stage of damage development, and final fibre breakage in 0o plies. Fig. 10
shows the polished edge of a failed specimen under static loading, illustrating transverse
matrix cracks and extensive delamination. When the laminate is subjected to fatigue loading,
damage usually starts in 90o plies in the form of matrix cracking, which multiplies in number
and extends in length with the increase of fatigue cycle. Delamination then initiates at the tips
of transverse matrix cracks and propagates through the 0/90 interface (Fig. 11), causing the
continuous load transfer from 90o plies to 0 o plies until the massive fibre breakage occurs in
0o plies, which results in the catastrophic failure of the specimen. However, although
numerous transverse matrix cracks and large delaminations are observed in the post-failed
specimens, the damage in the CF/PEEK composite laminates appears fairly late at the present
fatigue stress levels. This phenomenon is different from that of carbon/epoxy laminates [5], in
which transverse matrix cracks appear quickly because of brittle characteristics of epoxy
resins. In the CF/PEEK composite laminates the tough characteristics of PEEK matrix delay
the damage initiation to a late stage of the fatigue life.
Fiber Bundle
Matrix Pool
Bundle Fracture
(a)
(b)
(c)
Fig. 9 Schematic of misaligned fibre bundle (a), SEM photography
of fracture of misaligned bundles (b) and fibre/matrix adhesion (c).
II - 120
Fig. 10 Photography of the polished edge of

a failed [02/902] 2S specimen under static
loading.
Fig. 11 Photography illustrating microfracture modes in [02/902] 2S laminate

subjected to cyclic stress a=60%o for
N=684,000cycles
WEAR-OUT MODEL ANALYSIS

To evaluate the fatigue behaviour of the CF/PEEK laminates of different lay-ups, the wearout model of residual strength proposed by Halpin et al. [8] is applied. The degradation of
residual strength of composites under fatigue loading is assumed to be in the form of
= 0 Aa ( N 1)
r
(1)
where r, o, a are the residual strength, the static strength and the maximum applied cyclic
stress, respectively, and A are material constants to be determined. This equation means that
the strength of a composite laminate decreases from the initial value o to r after N cycles of
fatigue loading with the maximum cyclic stress a. When the residual strength equals to the
maximum applied stress, i.e. r=a, Eq. (1) is reduced to the S-N relation of the laminate,
e = a [1 + A( N 1)]S , S =
(2)
where e is the equivalent static strength. The formulation of the wear-out model is completed
assuming that the equivalent static strength obeys a two-parameter Weibull distribution, and
the probability of the static strength higher than e, or the reliability, is given by

P(e ) = exp e

(3)
where and are the Weibull shape and scale parameters, respectively. To evaluate fatigue
behaviour of composite laminates, these four parameters , A, and can be determined,
fitting the wear-out model to the experimental data.
The model parameters, calculated with the maximum likelihood method [9, 10], are obtained
for the CF/PEEK laminates of three different lay-ups, listed in Table 2. The S-N curve of each
composite laminate calculated from Eq. (2), with the equivalent static strength determined by
II - 121
e=(1+1/), is shown as a solid line in Figs. 5-7, respectively. The dashed lines denote the
95% confidence level. It can be seen that the wear-out model provides a good estimation to
the experimental data.
Table 2. Parameters of fatigue model for CF/PEEK commingled composite laminates.
Lay-up
[MPa]
[0]16
0.0630
0.1117
21.80
1549.25
[90]16
0.0237
0.0629
8.78
88.19
[02/902]2S
0.0604
0.0409
17.69
769.66
The residual strength degradation behaviour is simulated for both unidirectional [0] 16 and
cross-ply laminates [02/902]2S at different fatigue stress levels using the model parameters in
Table 2, which is shown as a solid line in Fig. 8. The difference between the wear-out model
and experimental data for the [0]16 laminate at a fatigue stress level of 50% static strength is
clearly visible. While for other stress levels, the wear-out model is correlated with the
experimental data very well for both [0]16 and [02/902]2S laminates.
CONCLUSIONS
Based on the present study, the following points can be highlighted:
1. The static properties of the CF/PEEK composite laminates made from commingled
unifabric prepreg are higher than those of APC-1 because of the higher tensile strength of
AS4 fibres, but are clearly lower than those of APC-2 due to a lower fibre volume
fraction.
2. The composite laminates exhibit higher fatigue sensitivity under tension-tension fatigue
loading than APC-1 because of fibre bundle breakage at the locations where fibre bundles
are misaligned.
3. The matrix cracking in transverse plies and extensive delaminations between 0o and 90o
plies initiated from the tips of transverse matrix cracks are two main damage mechanisms
in the cross-ply laminate. Compared with carbon/epoxy composite laminates, the damage
in the CF/PEEK laminates appears relatively late, because the tough characteristics of the
matrix delay the initiation of damage.
4. The wear-out model shows good correlation with the experimental data for both S-N
curve and residual strength.
ACKNOWLEDGMENT
L. Ye wishes to thank the Polymer Research Division, BASF, Germany, for supplying the
testing materials. X. X. Diao is supported by an Overseas Postgraduate Research Scholarship
and a Sydney University Postgraduate Research Award. The partial support from the
Australian Research Council (ARC) on this project is appreciated
II - 122
REFERENCES
1. Thermoplastic Composite Materials Handbook, ICI Thermoplastic Composites, 1992,
pp.1
2. Henaff-Gardin, C. and Lafie-Frenot, M. C., Fatigue Behaviour of Thermoset and
Thermoplastic Cross-Ply Laminates, Composites, 23(1992)358-303
3. Rigby, R. B., Polyetheretherketone, in Engineering Thermoplastics, Properties and
Applications, Margolis, J. M., Ed., Marcel Dekker, Inc. (1995), pp. 299-314
4. Maekawa, Z., Hamada, H., Lee, K. and Kitagawa, T., Reliability Evaluation of
Mechanical Properties of AS4/PEEK Composites, Composites, 25(1994)37-45
5. Charewicz, A. and Daniel, I. M., Damage Mechanisms and Accumulation in
Graphite/Epoxy Laminates, in Composite Materials: Fatigue and Fracture, ASTM STP
907, Hahn, H. T., Ed., American Society for Testing and Materials, Philadelphia, 1986,
pp.274-297
6. Engineering Materials Handbook, Vol.1, Composites, ASM International, 1987, pp.112113
7. Dickson, R. F., Jones, C. J., Harris, B., Leach, K. C. and Moore, D. R., The
Environmental Fatigue Behaviour of Carbon Fibre Reinforced Polyether Ether Ketone,
Journal of Materials Science, 20(1985)60-70
8. Halpin, J. C., Jerina, K. L. and Johnson, T. S., Characterisation of Composites for the
Purpose of Reliability Evaluation, in Analysis of the Test Methods for High Modulus Fibres
and Composites, ASTM STP 521, Whitney, J. M., Ed., American Society for Testing and
Materials, Philadelphia, 1973, pp.5-64
9. Sendeckyj, G. P., Fitting Models to Composite Materials Fatigue Data, in Test Methods
and Design Allowable for Fibrous Composites, ASTM STP 734, Chamis, C. C., American
Society for Testing and Materials, Philadelphia, pp.245-260
10. Diao, X. X., Ye L. and Mai, Y. W., Fatigue Behaviour of CF/PEEK Composite
Laminates Made from Commingled Prepreg Part I - An Experimental Study, Composites,
in revision
II - 123
EFFECT OF LOAD FREQUENCY ON THE TENSILE

FATIGUE BEHAVIOUR OF ANGLE-PLY AS4/PEEK
X. R. Xiao, I. Al-Hmouz
Concordia Center For Composites, Concordia University, Dept. of Mechanical Engineering
1455 de Maisonneuve Blvd. W., Montreal, Quebec H3G 1M8
SUMMARY: This paper presents the results of a study aimed at understanding the load
frequency effect on fatigue life of thermoplastic composites. The tensile fatigue behaviour
(R=0.13) of straight sided angle-ply [45]4s AS4/PEEK was investigated at three frequencies.
The cyclic stress-strain data and temperature variation during fatigue tests were recorded. A
model correlating the thermal effect to the load frequency on the fatigue life of the
thermoplastic composite was proposed.
KEYWORDS: fatigue, load frequency, thermoplastic, hysteresis loop, thermal effect
INTRODUCTION
The load frequency is known to have a substantial effect on fatigue performance of polymer
composites, especially on laminates dominated by the matrix properties [1-8]. Literature data
indicate that the load frequency has a two-fold effect on fatigue of polymer composites: the
creep effect and the thermal effect. Due to the viscoelastic nature of polymeric matrices,
matrix dominated laminates tend to creep during low frequency fatigue testing and, hence,
decreasing load frequency tends to reduce the fatigue strength of the material. On the other
hand, the viscoelastic dissipation under cyclic loading leads to autoheating of the matrix
material. The friction at damaged sites also generates heat. Since the heat generation is
proportional to load frequency, increasing frequency would increase the temperature rise and,
thus, result in softening of the material. Depending on the material system, the level and wave
form of cyclic stress, increasing load frequency may either increase or decrease the fatigue
life determined by the balance between the creep and thermal effects.
Fatigue failures associated with large-scale hysteretic heating have been observed in polymers
[9,10] and in glass fiber reinforced epoxy [1] where fatigue life decreased as load frequency
increased. The effect of hysteretic heating was less substantial in graphite/epoxy and
consequently fatigue life increased as load frequency increased [3]. The same trend was
observed in Boron/epoxy [6].
Several studies on fatigue of thermoplastic composite AS4/PEEK have shown that hysteretic
heating in AS4/PEEK is much more pronounced as compared to that in thermoset composites
[4,5,11] and the fatigue life decreases remarkably with increasing load frequency [4,5]. DanJumbo et al [5] have developed a model to account for the load frequency effect on fatigue
life. Their study, however, was conducted with notched specimens where the thermal effect is
restricted as compared to that in smooth specimens because the heat generated at notch areas
can dissipate into a much larger surrounding materials at lower stresses.
II - 124
The present work is aimed at understanding the thermal effect associated with load frequency
on fatigue life of thermoplastic composite AS4/PEEK. To emphasize the thermal effect,
straight sided angle-ply [45]4s specimens were used. Tensile-tensile fatigue tests were
conducted at three frequencies. The cyclic stress-strain data and temperature variation during
fatigue tests were recorded. A model correlating the thermal effect to the load frequency on
the fatigue life of the thermoplastic composite was proposed.
EXPERIMENTAL
Specimens
AS4/PEEK [0/90]4s panels of 12x12 inches were made from prepreg using a WABASH hot
press following ICI's procedure. The straight sided [45]4s specimens were then cut from the
laminates. The nominal dimensions of the specimens were 200x18x2 mm with 38 mm long
aluminum end tabs, which follows the specification of ASTM D3479. The ultimate tensile
strength of the [45]4s laminates was measured to be 338 MPa, which was very close to the
value reported in literature. Fig.1 shows the typical stress-strain curve obtained by static
tensile tests.
Fatigue Test
The tensile-tensile fatigue tests were carried out using a MTS machine under load controlled
mode with sinusoidal load wave form. The stress ratio in fatigue was R=0.13. Three load
frequencies were investigated: 1 Hz, 5 Hz and 10 Hz. At each frequency, fatigue tests were
conducted at three stress levels: 60%, 70% and 80% of the ultimate tensile strength. To
complete the S-N curve, few more fatigue tests were conducted at other stress levels.
During fatigue tests, the load level was monitored from the signal of the load cell while the
axial strain was measured using an extensormeter. The cyclic stress-strain data were collected
at selected cycles. The temperatures were measured during the test using thermocouples
attached to the center region of the specimen surface and recorded by a computer data
acquisition system.

S-N Data
The data on fatigue strength - number of cycle to failure (S-N) obtained are presented in Table
1. The values of the equilibrium or the maximum temperature recorded during fatigue tests
are also listed in Table 1. The S-N curves for three frequencies are plotted in Fig.2, where
symbols are the test data. It is evident that increasing frequency reduced the fatigue life and
this life reduction effect appeared to be larger at lower stress levels.
II - 125
II - 126
II - 127
II - 128
II - 129
Hysteresis Loop
Fig.3 shows the typical hysteresis loops recorded during fatigue tests at 60% stress level for
the three load frequencies. As seen the shape of the hysteresis loops changed gradually during
fatigue and the way of its change depended upon load frequency. Initially the hysteresis loop
had a shape of slim ellipse and was symmetric to its center, which may be approximated by
two triangles placed base to base as observed by other researchers [1]. With increasing the
number of cycle, the loop nearly collapsed in Fig. 3a and Fig. 3b (1 Hz and 5 Hz,
respectively) but expanded in Fig 3c (10 Hz).
At other stress levels, the evolution of hysteresis loop also appeared to depend upon
frequency. To compare the effect of frequency and stress level on the evolution of hysteresis
loop, the area of hysteresis loop is plotted versus the normalized number of cycle N/Nf in
Fig.4, where Nf is the fatigue life. At 1 Hz (Fig.4a), the area of hysteresis loop decreased as
the number of cycles increased whereas at 5 Hz and 10 Hz (Fig.4b and Fig.4c) the trend
reversed and this trend became stronger at higher stress levels.
The evolution of hysteresis loop correlated well with the damage extent observed on specimen
surfaces. The main type of damage observed on fatigue failed specimens was matrix cracking.
It was found that, with increasing load frequency and stress level, the intensity of the matrix
cracks increased. The additional energy dissipation as reflected by the expanded hysteresis
loops is likely due to friction at these damaged sites.
Another noticeable point in Fig.4 is the accumulation of creep strain. Regardless of the load
frequency and stress level used, all specimens failed by fatigue exhibited a total maximum
strain in the range of 18-20%, which is in the range of failure strain obtained by static tests.
Temperature Rise
The typical temperature variations during fatigue tests are shown in Fig.5 for the three load
frequencies at three stress levels. At 1 Hz, temperature rise reached the equilibrium for fatigue
tests conducted at all stress levels and the equilibrium temperature increased with stress
levels. At 5 Hz, a peak in temperature rise curve, which is similar to that reported by [4], was
observed at tests conducted at 60% level. At higher stress levels, monotonic temperature rises
were recorded. At 10 Hz, continuously rising temperatures were measured at all stress levels.
The values of the equilibrium or the maximum temperature recorded during fatigue tests have
been given in Table 1.
It should be noticed that the temperature measured at specimen surfaces for samples tested at
10 Hz sometimes exhibited a lower value than those tested at 5 Hz at the same stress level.
This is believed due to the delay in thermal conduction. During fatigue tests at higher load
frequencies and higher stress levels, the temperature differences between the center and the
surfaces of the specimens are much higher and the specimens may fail before the temperature
rises are reflected at their surfaces. Indeed, continuous temperature rises were recorded on
specimens tested at 10 Hz even after the fatigue test had been halted upon failure .
Taking this factor into consideration, we may conclude that the decrease in fatigue life of
AS4/PEEK [45]4s laminates with increasing load frequency is directly related to the
temperature rise associated with hysteretic heating. This is to say that the effect of load
frequency on fatigue life is mainly thermal nature.
II - 130
Table 1 Results of fatigue tests for [45] 4s APC2/AS-4 samples

Sample#
Frequency
(Hz)
2-B-3
4-B-93
2-C-4
1-A-3
1-B-1
1-B-8
1-B-6
3-C-4
3-C-22
-C-32
1-L-6
2-C-2
2-B-2
1-A-4
3-B3
3-B-1
1-K-2
1-S-5
1-A-9
1-K-3
4-B-8
4-B-7
1-B-4
2-D-5
1-S-1
1-K-7
1-S-2
1-S-8
1-S-9
1-K-1
1-L-2
2-D-4
2-D-2
2-D-3
1-S-6
1-S-7
1-A-7
1
1
1
1
1
1
1
1
1
1
1
1
5
5
5
5
5
5
5
5
5
5
5
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Load level
%ult
60
60
65
70
70
70
70
80
80
80
80
93
60
60
70
70
70
70
70
75
80
80
80
60
60
60
60
70
70
70
70
75
80
80
80
80
80
Number of
cycles
1,400,000*
1,200,000*
50700
20650
16000
21000
18800
2900
3000
3850
3570
1500
100,000
186,000
2200
1500
1100
2600
2620
350
200
220
200
540
575
1040
580
450
260
280
470
450
200
180
315
120
100
Temperature
(C)
34 E
32 E
43 E
40 E
45 E
46 E
45 E
45 E
45 E
84 F
84 F
127 F
135 F
152 F
127 F
100 F
110 F
97 F
108 F
152 F
140 F
136 F
95 F
119 F
80 F
110 F
102 F
135 F
90 F
Sample did not fail after this number of cycles.

E : equilibrium temperature.
F : failure temperature.
II - 131
A MODEL FOR THERMAL EFFECT ON FATIGUE LIFE

Literature data and our experimental results have shown that thermal degradation associated
with hysteretic heating is the dominant frequency effect on fatigue life of thermoplastic
composite AS4/PEEK.
For viscoelastic materials, the rate of hysteretic heating is determined by the hysteretic energy
dissipated under fatigue loading [9].
w = fD"(f,T)2
(1)
where w =energy dissipated, f = frequency, D" = the loss compliance, = the peak stress.
The hysteretic heating can also be measured directly from the area of hysteretic loop. Hanh
and Kim [2] have shown that the hysteretic energy can be approximated by
w = C(1-R)2(-t)
(2)
where C is a parameter related to the mid-width of the hysteretic loop, R is the stress ratio of
the fatigue test and t is the threshold stress below which no hysteretic heating is measured.
The hysteretic heating rate is given by
q = C(1-R)2(-t)f
(3)
From heat transfer analysis, the equilibrium temperature is found to be proportional to q and
hence may be predicted by
T = T0+ '[(1-R)2(-t)f]
(4)
where ' and are parameters, T0 is the environment temperature.

An observation of S-N curves suggests that these curves may be described by the following
function
Log(N)(-0')=C
(5)
where 0' is the fatigue limit, C is a parameter. The strength of viscoelastic materials is
sensitive to temperature, so is the fatigue limit. Assuming that the effect of temperature on
fatigue limit 0' follows an exponential function
0' = 0exp[-"(T-T0)]
(6)
With Eq.4, Eq.6 becomes

0' = 0exp{-[(1-R)2(-t)f]}
II - 132
(7)
Substitute Eq.7 into Eq.5 and introduce stress parameters p=/ult, p0=0/ult and pt=t/ult
we obtain the mathematics model for thermal effect on fatigue life of materials
Log(N){p-p0exp[-((1-R)2p(p-pt)f)]}=C
(8)
Eq.8 is actually a variation of the model presented by Dan-Jumbo et al [5] but, instead of
employing the temperature rise data, here the thermal effect is directly related to the stress
level and load frequency. In addition, the model in [5] is based on crack growth in viscoelastic
media, whereas Eq.8 is based on the assumption of a general strength degradation of the
material and hence it is suitable for smooth composite samples.
By curve fitting the data given in Table 1 and with p0=0.55, pt=0 and R=0.13, the values of
the parameters in Eq.8 were determined as follows: C=1.7, =1 and =0.5. The S-N curves
predicted by Eq.8 are presented in Fig.2 by dashed lines to compare with the test data. In
general, the model predicted the trend of load frequency effect except for 5 Hz at 60% stress
level.
CONCLUSIONS
Tensile fatigue tests of [45]4s AS4/PEEK laminates were carried out at three frequencies: 1
Hz, 5 Hz and 10 Hz. Substantial temperature rises were recorded during fatigue tests,
especially at higher frequencies and stress levels. Fatigue life reduced significantly as load
frequency increased. The results clearly indicate that the effect of load frequency on fatigue
life was dominated by thermal effect. A model correlating thermal effect with load frequency
and stress level was proposed. The prediction by the model agreed reasonably well with the
test data.
ACKNOWLEDGMENT
This work is supported by National Science and Engineering Research Council of Canada.
We also like to thank Mr.J.Dufour of IMI, National Research Council of Canada for the
assistance in fabrication of composite plates, Mr.P.Oullette and Mr.M.El-Karmalawy of
Concordia Center for Composites for their assistance in fatigue tests.
REFERENCES
1.
J.W.Dally and L.J.Broutman, "Frequency effects on the fatigue of glass reinforced

plastics", J. Composite Materials, Vol.1, 1967, 424-442.
2.
H.T.Hahn, R.Y.Kim, "Fatigue behaviour of composite laminate", J. Composite

Materials, Vol.10, 1976, 156-180.
3.
C.T.Sun, W.S.Chan, "Frequency effect on the fatigue life of a laminated composite",

Composite Materials: Testing and Design, ASTM STP 674, Ed. S.W.Tsai, 1979, 418430.
II - 133
4.
D.C.Curtis, D.R.Moore, B.Slater, N.Zahlan, "Fatigue testing of multi-angle laminates of

CF/PEEK", Composites, Vol.19, No.6, 1988, 446-452.
5.
E.Dan-jumbo, S.G.Zhou, C.T.Sun, "Load-frequency effect on fatigue life of

IMP6/APC-2 thermoplastic composite laminates", ASTM STP 1044, G.M.Newaz, Ed.,
1989, 113-132.
6.
K.L.Reifsnider, W.W.Stinchcomb, T.K.O'Brien, "Frequency effects on a stiffness based

fatigue criterion in flawed composite specimens", Fatigue of Filamentary Composite
Materials. ASTM STP 636, K.L.Reifsnider and K.N.Lauraitis, Eds., ASTM, 1977, 171184.
7.
D.Kujawski, F.Ellyin, "Rate/frequency-dependent behaviour of fibreglass/epoxy

laminates in tensile and cyclic loading", Composite, Vol.26, 1995, 719-723.
8.
J.F.Mandell, U.Meier, "Effect of stress ratio, frequency, and loading time on the tensile
fatigue of glass-reinforced epoxy", Long-Term Behaviour of Composites, ASTM STP
813, ASTM, 1983, 55-77.
9.
R.W.Hertzberg, J.A.Manson, Fatigue of Engineering Plastics, Academic Press, New

York, 1980.
10.
S.Suresh, Fatigue of Materials, Cambridge University Press, New York, 1991.
11.
C.C.M.Ma, S.H.Lin, N.H.Tai, J.F.Wu, G.Y.Wu, C.L.Ong, Y.C.Chang, M.F.Sheu,

Y.R.Yang, "Fatigue behaviour of carbon fiber reinforced thermoplastic and thermoset
polymer composites (I) [45]4s laminates under tension-tension loading, Proceedings of
ICCM10, Vol.I, 1995, 569-576.
II - 134
RESIDUAL STRENGTH OF GRP AT HIGH CYCLE

FATIGUE
J. Andersons1, J. Korsgaard2
1
Institute of Polymer Mechanics, 23 Aizkraukles iela, Riga LV-1006, Latvia

2
LM Glasfiber A/S, Denmark
SUMMARY: Residual strength degradation of glass-fiber reinforced polyester composite of

a lay-up typical for wind rotor blade material is studied at low-cycle fatigue. Gradual
reduction of remaining strength is observed as expected for GRP, accompanied by increasing
scatter. Residual strength model based on strength-life equal rank assumption yields accurate
approximation of experimental data. Strength reduction at stress level corresponding to highcycle fatigue (N>106 cycles) appears to correlate well with higher stress level test results
indicating that strength degradation at design stress level can be evaluated using low-cycle
tests. Assuming that strength degradation model parameters do not depend on applied stress
level, residual strength data obtained at low stress level tests of comparatively short duration
can be used to estimate the average fatigue life at the same stress thus considerably reducing
total test time.
KEYWORDS: fatigue, residual strength, glass/polyester composite
INTRODUCTION
Fail-safe strength design of engineering composite structures has to ensure that appropriate
residual static strength level is maintained during design life. The latter is typically too long to
allow conventional fatigue and residual strength tests to be performed at the desired low
design stress level. Hence fatigue test data are usually extrapolated to low stress levels of
interest. In order to improve the accuracy of remaining strength and life estimates, a model of
composite fatigue behavior during high-cycle fatigue verified by experimental data is
necessary.
Residual strength models have been developed for prediction of residual strength and life
distribution under constant amplitude, block [1], and spectrum loading [2]. Modifications
taking into account mean stress effect [3], competing failure modes [4], degradation
mechanism change [5] have been proposed. In [6] residual strength model is used to
extrapolate experimentally derived S-N curves. The model predictions satisfactorily agree
with experimental data in the stress range covered, which in most investigations is limited by
that corresponding to fatigue life of about 106 cycles. Less research is dedicated to strength
degradation at lower cyclic stress levels which are of primary significance for long-term
applications.
II - 135
With this aim residual strength degradation at a range of stress levels has been studied
experimentally, and a method of fatigue life estimation based on limited amount of residual
strength tests at the stress level of interest is proposed.
MATERIAL AND TESTS

Glass fiber reinforced polyester used as a blade material for wind turbines was tested. The
material was produced by hand lay-up, and comprised chopped strand mat, unidirectionally
reinforced, and fabric layers in the following sequence: /CSM, fabric,(CSM, UD)2/ S. The
resin used was an orthophthalic polyester (Norpol 410M910). The average fiber volume
fraction was 41 %. Longitudinal modulus and tensile strength were correspondingly E= 30
2 GPa and S= 672 21 MPa.
Dog-bone shape specimens were used for tensile static strength and fatigue tests. Static
strength distribution is shown in Fig. 1, and fatigue diagram - in Fig. 2. Fatigue tests were run
at ambient conditions. Loading frequency was 17 Hz, stress ratio r = 0.1. Residual tensile
strength degradation was studied at a number of stress levels. The results are shown in Table
1.
Table 1: Residual strength
Maximum
stress, MPa
Duration of
loading,
cycles
284
2.5 104
5 104
1 105
1.24 105
2 105
4 105
4 105
4 105
8 105
1.6 106
235
221
206
186
167
Duration of
loading,
fraction of
nominal life*
.17
.33
.66
.82
.15
.30
.16
.08
.07
.06
Number of
specimens
tested
7
9
7
8
6
7
5
5
7
5
Average
residual
strength,
MPa
604
555
507
493
628
598
628
634
635
617
Standard
deviation of
residual
strength, MPa
27
35
53
82
27
30
31
28
35
3
life fraction reported in the Table is calculated with respect to linear approximation of lgN vs.
diagram
MODELING OF RESIDUAL STRENGTH DEGRADATION
Majority of residual strength R degradation models can be derived from the following rate
equation:
dR
= f ( , r, S )g( R)h( n)
dn
II - 136
(1)
supplemented by initial condition R( 0) = S and failure criterion R( N ) = , where R residual strength, S- static strength, N- fatigue life at cyclic stress with maximum value , n current number of loading cycles. Non-linearity of strength reduction is accounted for by
appropriate choice of functions g and h. Models derived assuming that strength degradation
rate does not explicitly depend on either n (i.e. h=1) or R are commonly used. Since there is
little physical background for favoring any of these approaches, both of them are considered
below.
Statistical Model
Taking g =
S 1
1 +1
R
, h 1 and f = c
, residual strength model [7] is recovered,
S c N
and residual strength R of a specimen with initial static strength value S after fatigue loading
of n cycles at a constant maximum stress level is:
S n
R =S c
S c N
(2)
Residual strength distribution is derived in [7] assuming Weibull distribution function for
static strength S:
FS (S ) = 1 exp[ S ]
(3)
and fatigue life (at constant maximum stress level ):

c n /c
FN (n) = 1 exp +
N
(4)
All variables with stress dimension are normalized by Weibull scale parameter of static
strength distribution; N - characteristic fatigue life at stress level . It follows from (4) that it
is related to the mean life N :
c
N = N 1 + + c

(5)
Residual strength distribution is given by Eqn 6:

FR ( R ) = 1 exp[ S0 S ( R) ]
(6)
where by S(R) is denoted the static strength value corresponding to residual strength R and
obtained solving (2) for S. Parameter S0 accounts for the fact that residual strength can be
determined only for the specimens that have survived fatigue loading, i.e. for which
R(n) > . Therefore S0 is calculated from (2) replacing R by .
II - 137
The model described comprises five material parameters. They were estimated by maximum
likelihood method as corresponding distribution function parameters in the following order.
First, strength distribution scale and shape parameters and were evaluated fitting (3) to
static strength test results (see Fig. 1), then parameter c characterizing fatigue life scatter and
characteristic S-N curve lg N = A B parameters A and B were determined
approximating constant amplitude fatigue data (Fig. 2) by (4), and the remaining residual
strength degradation rate parameter - fitting (6) to residual strength data at = 284 MPa
(Fig. 3). Parameter values are shown in Table 2.
Table 2: Model parameter values
35.1
, MPa
683
c
10.8
A
10.7
B, MPa-1
.019
2.45
Deterministic Model
If h( n) = n 1 and g( R) 1 are taken in Eqn 1, it follows that residual strength degrades
under constant amplitude cyclic loading according to relation (7):
n
R = S (S )
N
(7)
Although the model allows statistical treatment, strength and fatigue life in Eqn 7 are
normally substituted by their average values therefore making model deterministic (see e.g.
[8]).
Deterministic model, unlike the statistical one, cannot account for apparent slowing down of
strength reduction at n approaching mean fatigue life N (and caused by failure of weaker
specimens during cycling). Therefore average residual strength values at = 284 MPa
obtained at loading duration up to .66N only were used for parameter determination,
resulting in = 1.33 (see Fig. 4).
Residual strength data apparently form a mastercurve (Fig. 5), when plotted as a function of
nominal life fraction spent under cyclic load. It makes it possible to roughly estimate residual
strength reduction at low cyclic stress level using high stress level data. Variation of
remaining strength for small to medium loading duration in terms of life fraction (n/N<.35) is
very close to that of static strength with the exception of the lowest stress level considered =
167 MPa. Standard deviation of residual strength for the latter was almost an order of
magnitude less then that of static strength.
In order to check whether strength degradation model parameters are stress dependent, they
were determined from two residual strength data sets obtained at = 284 MPa (corresponding
fatigue life N=1.9 105) and = 235 MPa (N=1.7 106). Stress levels were chosen so that they
lay within the stress range at which fatigue tests were performed, so that reliable estimate of
II - 138
corresponding fatigue lives was available. Model parameters for data generated at = 284
MPa are listed above, while for = 235 MPa they are = 2.36 (statistical model) and = 1.28
(deterministic model). The difference in parameter values determined at different stress levels
is negligible for both models, therefore they can be considered material parameters. Prediction
of residual strength distribution at = 235 MPa by statistical model (Fig. 6, 7) and mean
strength degradation by deterministic model (Fig. 4) using model parameters determined at
= 284 MPa show very good agreement with test data.
Assuming that further reduction of cyclic stress level will not influence model parameters,
low stress residual strength data can be used for estimation of the corresponding fatigue life.
For deterministic model, relation for mean life estimate at stress follows from Eqn 7:
S
N = n
S R
1/
(8)
For statistical model, maximum likelihood method was used for determining the value of N*
in Eqn 2 that yields the best fit of distribution function (6) to residual strength data obtained at
the corresponding maximum cyclic stress . Theoretical residual strength distributions are
plotted in Fig. 8-10. This method allows to calculate also standard deviation of N, thus
allowing to evaluate the accuracy of prediction. For the lowest cyclic stress value considered
estimated variation of N was of the same order as N clearly indicating that statistical model is
not applicable in this case.
Predicted fatigue life at low stress is shown in Fig. 2. Fatigue life estimates by both residual
strength models are very close to each other, and slightly below extrapolation line of
experimental S-N data.
CONCLUSIONS
Residual strength data vs. life fraction of GRP under load form a mastercurve, from which a
rough estimate of strength reduction at the cyclic stress level of interest can be obtained.
Strength degradation model parameters appear not to depend on cyclic stress level. It permits
to estimate fatigue life of composite at a given stress level from a limited amount of residual
strength test results at the same test level.
ACKNOWLEDGMENTS
This work was partially supported by CEC contract JOU2-CT93-0277 - Advanced Flexible
Composite Rotor for Large Wind Turbines.
II - 139
REFERENCES
1. Yang, J. N. and Jones, D.L., The effect of load sequence on statistical fatigue of
composites, AIAA Journal, 1980, Vol. 18, N 12, pp. 1525-1531.
2. Yang, J. N. and Shanyi Du, An exploratory study for fatigue of composite under
spectrum loading, Journal of Composite Materials, 1983, Vol. 17, Nov., pp. 511-526.
3. Kim, R.Y., Effect of mean stresses on fatigue behavior of composite laminates, Proc.
ICCM 7, 1982, pp. 621-626.
4. Whitney, J.M., A residual strength degradation model for competing failure modes,
Long-Term Behavior of Composites, ASTM STP 813, 1983, pp. 225-245.
5. Talreja, R., Fatigue of Composite Materials, 1985, 125 p.
6. Donaldson, S.L., and Kim, R.Y., Life prediction of glass/vinylester and glass/polyester
composites under fatigue loading, Proc. of ICCM-10, Vol. I: Fatigue and Fracture, 1995,
pp. 577-584.
7. Yang, J.N., and Jones, D.L., Load sequence effects on the fatigue of unnotched composite
materials, Fatigue of Fibrous Composite Materials, ASTM STP 723, 1981, pp. 213-232.
8. Reifsnider, K.L., and Stinchcomb, W.W., A critical element model of the residual strength
and life of fatigue-loaded composite coupons, Composite Materials: Fatigue and Fracture,
ASTM STP 907, 1986, pp.298-313.
0.8
F(S)
0.6
0.4
0.2
0
600
620
640
660
680
700
720
S , MPa
Fig. 1: Static tensile strength distribution of GRP
II - 140
400
350
300
, MPa
250
200
150
100
50
0
4
4.5
5.5
6.5
7.5
lg N
Fig. 2: Fatigue diagram

(x) - S-N data; mean life predicted by statistical () and deterministic () models
800
700
600
R, MPa
500
400
300
200
100
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
n/N
Fig. 3: Residual strength reduction at = 284 MPa

Statistical residual strength model prediction for failure probabilities .5 ( ) ;
.1 and .9 (- - -)
II - 141
700
600
R, MPa
500
400
300
200
100
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
n/N
Fig. 4: Mean strength degradation at = 284 MPa () and = 235 MPa ()

Lines -deterministic model prediction
800
R, MPa
700
600
500
400
sigma=284 MPa
235 MPa
221 MPa
205 MPa
186 MPa
167 MPa
300
0
0.1
0.2
0.3
0.4
n/N
Fig. 5: Residual strength vs.nominal life fraction for different cyclic stress levels
II - 142
0.8
0.6
0.6
F(R)
0.8
F(R)
0.4
0.4
0.2
0.2
0
450
550
0
450
650
550
650
R, MPa
R, MPa
Fig. 6: Residual strength distribution after Fig. 7: Residual strength distribution after
200000 cycles at = 235 MPa
0.8
0.8
0.6
0.6
F(R)
F(R)
0.4
0.4
0.2
0.2
0
550
600
R, MPa
650
700
0
550
600
650
700
R, MPa
Fig. 8: Residual strength distribution after Fig. 9: Residual strength distribution after
II - 143
0.8
F(R)
0.6
0.4
0.2
0
550
600
650
700
R, MPa
Fig. 10: Residual strength distribution after 800000 cycles at = 186 MPa
II - 144
A STOCHASTIC CUMULATIVE DAMAGE MODEL

FOR THE FATIGUE RESPONSE OF LAMINATED
COMPOSITES
R. Ganesan, S.V. Hoa, S. Zhang and M. El-Karmalawy
Concordia Centre for Composites, Department of Mechanical Engineering
Concordia University, Montreal, Quebec, Canada.
SUMMARY: Polymer-matrix fibre-reinforced laminated composites display considerable

variability in their fatigue response. Existing works on the quantification of fatigue response of
composite laminates have deployed statistical and probabilistic approaches. These approaches
are however inadequate for the prediction of the fatigue response as well as its probabilistic
characteristics, and for the determination of reliability for use in design. In the present paper, a
new cumulative damage model is developed based on a stochastic approach. The damage
evolution in the composite laminate is quantified through the compliance increase in the laminate
based on the embedded Markov process theory. The prediction of increase in compliance and the
associated probabilistic characteristics is carried out based on this model. The fatigue response of
a graphite epoxy composite laminate is then analyzed based on the new model using the test
data. The composite material considered is AS 4/3501-6 graphite epoxy cross ply lay up [0/90]4s.
KEYWORDS: fibre-reinforced polymer-matrix laminated composites, fatigue behaviour,
stochastic process modelling and analysis, markov chains
INTRODUCTION
Laminated composites are known for the high degree of variability in their mechanical
properties. In particular, the behaviour and engineering properties of polymer-matrix laminated
composite materials are significantly affected by the technological process defects, design errors
at both material and structural level, and environmental effects which are random in nature.
However, for quantifying the fatigue behaviour of composite laminates, so far only a
deterministic approach has in most cases been deployed. This deterministic approach has
resulted in disagreement between theory and practice.
Weibull [1] introduced into fatigue failure studies a probability distribution function in order to
account for the variations in the life data of mechanical components made of metallic materials.
Reliability analyses have been developed for metallic structures based on this distribution
function. The Weibull distribution function has also been used to model the fatigue life data of
composites [2-4].
The Weibull distribution function frequently provides a reasonable fit to the probability
distributions of life times that have been obtained from fatigue tests or service operation. Some
aspects of variability in life time can be assessed and accounted for. However, (i) the probability
distribution function of life times has not been related to the fatigue damage process, (ii) it is not
possible to separate and account for the major sources of variability in material response, (iii) the
II - 145
correlation characteristics of test data are not accounted for, and (iv) the nature of physical
process is not reflected. Further, there are only two limiting states of the material in the Weibull
distribution based fatigue model, that are (i) satisfactory (no damage) and (ii) failure (completely
damaged). This imposes severe limitations on the validity and applications of this model. For
instance, one can not incorporate the initial damage distribution (for a composite, this refers to
the fluctuations in the initial elastic modulus) nor it is possible to trace the damage accumulation
during service using the Non-Destructive Testing (NDT) techniques. A rigorous stochastic
approach can take into account all sources of variability in material response and lead to a
reliable prediction of material response.
Fundamental to the design of composite structures is a model that can be deployed for the
prediction of fatigue damage accumulation. The development of a cumulative damage model
based on a rigorous stochastic approach and the associated prediction scheme is the objective of
the present work.
MARKOV PROCESS MODELLING OF COMPLIANCE

The model developed here considers the "Cumulative Damage (CD)" due to fatigue, which is the
irreversible accumulation of fatigue damage in the composite material. A "Stochastic
Phenomenological Model" is developed, since the experimental data display considerable
randomness. The central idea is that in a suitably defined "Duty Cycle (DC)" a random amount
of fatigue damage is accumulated in the composite laminate. This fatigue damage in turn results
in the reduction of the elastic modulus of the laminate.
It is quite natural and reasonable to assume that the increment of damage at the end of a DC
depends in a probabilistic manner on the amount of damage present at the start of the DC, on that
DC itself, and is independent, in the probabilistic sense, of how (i.e. the damage development
path) the damage was accumulated up to the start of that DC. This corresponds to the Markov
assumption in the stochastic process theory.
It may be noted here that the Markov assumption is consistent with the system theory concepts
that are well established. The laminate specimen can be viewed as a dynamic system. At any
instant of time (or equivalently the load cycles), the compliance of the laminate is analogous to a
variable system parameter. The input to the system, i.e. the excitation, is the loading during a
DC. The increase in the compliance of the sample (i.e. damage increment) can be viewed as an
output (i.e. response). The Markov assumption then simply implies that the response of the
system (which is the increase in compliance at the end of a DC) is a function of the system
parameter (i..e. the fatigue damage that has been present in the laminate before the DC was
applied which is indicated by the laminate compliance at the start of the DC), and the excitation,
and is independent of the way with which the value of the system parameter (i.e. the fatigue
damage in the laminate at the start of DC) has been arrived at.
It is well known that (i) unlike the metals in a composite laminate various damage mechanisms
become simultaneously active, and (ii) the viscoelastic response of the laminate is pathdependent. However, the fatigue damage in a laminate is not quantified in terms of the length of
the fatigue crack or any other physically-observable variables (for which the path dependency is
an important consideration) as in the case of metals. Rather, a material parameter such as
compliance that is sensitive to, reflective of and representative of whatever forms of fatigue
damage is deployed as a "system descriptor" or "damage development indicator". Also, this
II - 146
parameter is determined from the instantaneous response of the laminate and not from the
transient response.
Test data regarding the elastic modulus are collected using nominally-identical composite
specimens that are subjected to fatigue loading. The compliance, which is the reciprocal of
elastic modulus, increases as the fatigue damage sets in. The compliance is denoted here by C
and the number of load cycles is denoted by n. The values of the stress ratio and the stress
amplitude of the fatigue loading are denoted by R and S respectively.
For each test specimen, the C - n curve can be plotted. Due to the inherent variability associated
with the accumulated damage, this curve varies from specimen to specimen. Typical sample
realizations are depicted in Figure 1. The stochastic process modelling the compliance is denoted
as {C(n, S, R), n N, S A, R B}, where N, A and B are the respective sample spaces of the
number of load cycles, the stress amplitude and the stress ratio. The sets N, A and B are the
index sets of the stochastic process {C(n, S, R)}.
The discrete set of damage states is determined based on the following procedure. The entire
range of compliance values is calculated from the sample test data. This range is then divided
into m number of equal or non-equal intervals, (C2 - C1), (C3 - C2), ........., (Ci+1 - Ci), ......, (Cm+1 Cm), where Cm+1 is the maximum value of the compliance.
Finite Markov Chain
The compliance process is viewed as a Markov process. The compliance process is nonhomogeneous because the compliance is a continuous function of the number of load cycles. The
index set N of this non-homogenous Markov process is (n0 , n1 , n2 ,......, nj-1 , nj, .........nmax ) and
the state space is (1, 2, 3, ......, j-1, j, .......m). Both these index sets are discrete. A Markov
process that has discrete index sets is called as Markov Chain (MC). Since the index sets are
finite, the compliance becomes a Finite Markov Chain.
Transition Probability Function (TPF)
As a result of the Markovian property and non-homogeneity of the stochastic process,
pij (nk-1 , nk ) = P [C(nj) = j | C(nj-1) = i ]; nk nk-1
(1)
where pij (nk-1 , nk ) is the Transition Probability Function (TPF), which is the probability of the
composite specimen entering a new damage state j at loading cycle nk, given that it is currently in
damage state i at loading cycle nk-1 . Based on the definition of the conditional probability
density function, joint probability density function and Bayes' theorem [5] this equation can be
recast into
pij (nk-1 , nk ) = P{[Cj C(nk ) Cj+1 ] [Ci C(nk ) Ci+1 ]}/P[Ci C(nk ) Ci+1 ]
(2)
The individual probability density functions of random variables Ck-1 and Ck are denoted by fCk-1
(ck-1) and fCk (ck ) respectively. The joint probability density function corresponding to these two
II - 147
correlated random variables is denoted by fCk-1 Ck ( ck-1 , ck ). The TPF between damage states i
and j, can be determined according to
pij (nk-1, nk ) = ,CjCj+1 ,CiCi+1 fCk-1 Ck (ck-1 , ck) dck-1 dck /{,CiCi+1 fCk-1 (ck-1)dck-1 }
(3)
Transition Probability Matrix (TPM)

Generalizing Eq. (2) for all possible damage states included in the state space, the Transition
Probability Matrix (TPM) of the composite specimen, denoted by [-(nk-1 , nk)], when the load
cycle is increased from nk-1 to nk, can be obtained. Since all damage states are mutually exclusive
and collectively exhaustive, the sum of the probabilities in any row of the TPM that has non-zero
probabilities is equal to unity. When the load cycle is nk-1 , the compliance is C(nk-1) = i. Now, if
the load cycle is increased to nk , the compliance is C(nk) = j. It is also possible that C(nk) = C(nk1) so that j = i. However, the system can not make a transition form a higher state to a lower state,
i.e. C(nk) C(nk-1) so that j i. The final state m is the "absorbing state" and so pmm (nk-1 , nk) = 1.
M-Step Transition Probability Matrix
The transition probability matrix of the composite specimen corresponding to any initial load
cycle n0 and any final load cycle nm can be determined in terms of the transition probability
matrices, based on the stochastic process theory. The resulting matrix of order m x m is called as
the "M-Step Transition Probability Matrix (M-Step TPM)" and is denoted by [- (n0 , nm)]. In
order to determine this matrix, use is made of the Chapman-Kolmogorov Theorem [5]. The
product equation for the M-Step TPM is obtained as
[- (n0 , nm)] = [- (n0 , n1)] x [- (n1 , n2)] x [- (n2 , n3)] x......x [- (nm-1 , nm)]
(4)
In order to determine the unconditional probabilities of the damage states of the specimen, given
the initial damage state distribution, when the load cycle is increased from n0 to nm, the following
equation is derived.
{-U(n0 , nm)} = {- (0)} x [- (n0 , n1)] x [- (n1 , n2)] x......x [- (nm-1 , nm)]
(5)
In this equation, {- (0)} denotes the initial damage state (row) vector of the specimen which
consists of the probabilities of the specimen in different damage states at load cycle n0.
Prediction of Damage States at any Load Cycle
The initial and final values of load cycles that have been included in the test program are first
considered. This duty cycle is expressed in terms of m number of smaller duty cycles. For each
smaller duty cycle, the transition probability matrix is determined according the procedure given
in the foregoing. This way, m number of transition matrices will be available. For each entry in
the transition probability matrix, m number of values will be available. For instance, consider
any transition probability pij. If 10 steps (smaller duty cycles) are considered, ten values of this
transition probability viz., 1pij , 2pij , 3pij , ..........., 10pij will be available. These ten values are
considered as constituting a random variable. Further, it is assumed here that this random
variable, denoted by pij has a Gaussian distribution.
II - 148
The ten sample values are used as a basis for predicting the value of transition probability pij that
corresponds to the duty cycle defined by the final load cycle value of the 10-the (smaller) duty
cycle and the value of the load cycle at which the prediction is to be made. The relevant rationale
and procedure are outlined in the following.
(i) For the random variable pij, a Gaussian probability density function is fitted using the ten
sample values available. The parameters of this Gaussian density function, ij and ij , can be
determined following the standard parameter estimation techniques [6]. A suitable statistical
analysis program is available in the MATLAB software, and this program was used in the
present work.
(ii) A standardized random variable zij is defined according to the relationship
zij = (pij-ij)/ij. This standardized random variable will also have Gaussian distribution because
the Gaussian nature of probability distribution does not change when a linear transformation of
normal random variable is performed.
(iii) Now, the concepts of reliability theory and failure probabilities [7] are deployed. It can be
proved using the probability theory and system reliability theory that a value of pij that
corresponds to a reliability R (say 90%) can be obtained by first determining the value of z90
such that the area bounded by the density function curve of zij between the range - to z90 is 0.9,
and then calculating the value of pij according to the relationship (z90 ij ) + ij . For instance,
when the reliability is 90%, the value of z90 is approximately equal to 1.3. Now consider a
transition probability, say p33. By fitting a Gaussian distribution, the parameters of Gaussian
density function for p33 can be obtained and say, they are equal to, 0.4819 and 0.2843 (it is
worthwhile to note the magnitude of randomness in the values of p33, which is approximately
60% in the above example). The value of p33 which will have a reliability of 90% (i.e. the
probability that this value will be exceeded in actual test results is 10%) is {[1.3 (0.2843)] +
0.4819} which is equal to 0.8515. This is the predicted value of p33. This procedure is followed
to determine the predicted values of all the transition probabilities. This way, the complete
transition probability matrix is determined.
Now, the (10+1) - step duty cycle defined by the initial value of the load cycle in the test
program and the value of the duty cycle at which the prediction is to be made, is considered. Ten
transition probability matrices can be determined from test data and the eleventh transition
probability matrix is predicted. The unconditional probabilities of damage states corresponding
to the load cycle at which prediction is to be made can be determined using Eqs. (4) and (5). The
value of m is equal to 11. These unconditional probabilities define the histogram of damage
states corresponding to the load cycle where prediction is to be made.
Analytical Probability Density Functions
The probability density function of the compliance at load cycle nk-1 , which is denoted as fCk-1
(ck-1) is determined based on the Maximum Entropy Method [7]. Based on this method, the
analytical form of the probability density function fCk-1 (ck-1) is taken to be
fCk-1 (ck-1)
p
exp [0 + i ck-1i ]
i=1
(6)
II - 149
The values of the parameters i , i = 0, 1, 2,..., p are determined such that the entropy of the
random variable ck-1 given by
S = - , fCk-1 (ck-1) ln [ fCk-1 (ck-1) ] dck-1
R
(7)
where R denotes the range of the random variable ck-1 , is maximum, subject to the following
constraints: (i) the total probability is unity, i.e.
=
1
(8)
, fCk-1 (ck-1) dck-1
R
and (ii) the probabilistic moments of the random variable ck-1 defined in terms of the probability
density function fCk-1 (ck-1) are equal to the corresponding statistical moments obtained from the
compliance data, i.e.
Mi , i = 1, 2, ..., q
(9)
, ck-1i fCk-1 (ck-1) dck-1 =
R
where Mi is the i-th statistical moment about the origin of the compliance values ck-1 .
Further details regarding the Maximum Entropy Method can be found in Ref. 7. The above
problem should be solved based on the techniques of optimization. The optimization technique
that is best suited to this problem is the Nonlinear Programming Technique proposed by
Jacobson and Oksman [8].
Joint Probability Density Functions
The joint probability density function is determined based on the relative frequency approach [6],
in the form of a three-dimensional histogram.
EXPERIMENTAL INVESTIGATION
The composite material considered is AS 4/3501-6 Graphite Epoxy cross ply lay up [0/90]4s.
Plates made up of this material and of dimensions 280 mm x 250 mm x 1 mm were
manufactured. The manufacturing conditions are as follows: Each plate is first heated for 1 hour
at a temperature of 116o C and at a pressure of 586 kPa (full vacuum), then it is heated for 2
hours at a temperature of 176o C and at a pressure of 655 kPa (full vacuum). Each plate was then
used to produce 10 specimens each of dimensions 280 mm x 25 mm x 1 mm. Aluminium plates
30 mm x 30 mm x 0.2 mm were attached to the gripping regions of the specimens, using the
ASTM D 3039/D 3039-93 test procedure. The effects of relaxation stresses induced during
manufacturing usually disappear about six months after fabrication, as described in [9]. In order
to minimize the effects of stress relaxation, all the tests for a batch of specimens were carried out
at a fixed time, that is one month after conditioning, after which 90% of the recovery has taken
place.
The environmental conditions can seriously affect the damage properties of composites. In
particular, the internal stress caused by machining and the moisture induced during machining
can change the value of the damage measure, and so the specimens must be conditioned before
testing. The method previously used in Ref. 9 has been employed in the present work, in this
regard. It consists of (i) heating the specimens at 100o C in an oven for 2 hours, (ii) raising the
II - 150
temperature to 170o C and holding it for two hours, and (iii) leaving the specimen at room
temperature for at least one week before testing. All the plates were tested for defects using Cscanning technique and using microscopic inspection for voids, and only those plates that met
the required quality were chosen for static and fatigue tests.
Static Testing
In the experimental investigation, considerable attention has been paid to ensure the quality of
the composite specimens and test procedures. Two specimens were chosen from each plate to be
tested for static properties and strength, and those specimens were chosen from a random
position in the plate. The specimens were tested for their tensile properties in accordance with the
ASTM D 3039/D 3039-93 standard test method for tensile properties of polymer matrix
composite materials. Only those plates with specimens having tensile strength and modulus
within 10% variation of the average tensile strength and average modulus of all tested
specimens, were accepted. All other specimens were rejected. The (statistical) average tensile
strength of the tested specimens was 614 MPa with a standard deviation of 49.5 MPa. The
(statistical) average Young's modulus of the tested specimens has been 61.151 GPa with a
standard deviation of 2.27 GPa.
Fatigue Testing
Tests were conducted at room temperature on an MTS hydraulic machine. Specimens were
clamped to the machine grips using Al tabs as outlined in the ASTM D 3039/D 3039-93 test
procedure. An on-axis T-T fatigue load at 5 Hz frequency and with a load ratio R of 0.1 was
applied. The maximum load is chosen to be that required to induce gradual failure to the
specimens. The value chosen corresponds to 70% of the average specimen strength. This value
represents a realistic value used in the design of composite structures. The fatigue/static strength
ratio for uniaxial CFRP is widely quoted [10] as being about 70% for endurance of about 107
cycles. The stress level selected is at a middle level between the higher applied stress levels
resulting in higher stiffness at failure and shorter fatigue lives, and the lower stress levels which
induce more total damage than higher stress levels (as evident by the greater amount of stiffness
reduction at failure). The fatigue modulus was obtained by dividing the maximum applied stress
by the corresponding strain. The values were monitored continuously during load cycling. The
values of fatigue compliance at different number of load cycles were plotted in Figure 1.
NUMERICAL RESULTS FOR THE MARKOV CHAIN

Step 1:Determination of the damage states: The largest value in the state space i.e. the largest
value of the compliances of all test specimens, is 1.7831 x 10-11 Pa-1 and the lowest value is
1.6114 x 10-11 Pa-1, as can be seen from Figure 1. This range of compliance values is then
divided into seventeen damage states of same size.
Step 2:Generate Probability Density Functions for Each Compliance Value Set C(nk-1): The first
four statistical moments, i.e. M1, M2, M3 and M4, of each compliance value set C(nk-1) are
determined and are used as constraints in the determination of fC k-1 . The maximum entropy
method yields the values of the five parameters viz., i, i=0, 1, 2, ..., 4, for each
fC k-1. The values of these parameters are listed in Table 1.
II - 151
Table 1: List of the values of i at five load cycles

0
0 cycle
101 cycles
151 cycles
201cycles
251 cycles
-59.75927
-52.95307
-10.20866
11.63882
-1.021048
1
12.11768
4.758957
0.032665
-1.913443
-0.437083
2
-0.976104
-0.163753
0.0168172
0.0765457
0.0163917
3
0.0350279
0.0024078
-0.000467
-0.001223
-0.000210
4
-0.000468
-0.000013
0.0000034
0.0000067
0.0000008
Step 3:Determination of Joint Probability Density Functions: Five three-dimensional histograms

that correspond to five smaller duty cycles (0, 101), (101, 151), (151, 201), (201, 251) and (251,
301) are determined.
Step 4:Evaluation of Transition Probability Matrices: The transition probability matrices can
now be evaluated using Eq. (3). The single-variable integration is performed using a subroutine
available in MATLAB software. At this stage, the sum of the non-zero transition probabilities in
each row of the transition probability matrix is calculated. Theoretically, the sum must be equal
to unity. However, for practical purposes, a tolerance margin of 10% is used in the present
work. If the sum of the transition probabilities is more than 1.1 or less than 0.9, suitable
modifications have to be made as to the bounding compliance values of the damage states Ci , i =
1, 2, ..., 18 and/or the bounds used for the single-variable distributions. An iterative procedure is
hence called for. The five transition probability matrices each of order 17x17 corresponding to
the five duty cycles (0, 101), (101, 151), (151, 201), (201, 251) and (251, 301) are obtained using
the above procedure. The matrices [(0, 101)] and [(101, 151)] are considered for illustrative
purposes and those parts of the rows and columns of these matrices that have non-zero entries are
given below:
[(n0, n1)] = [(0, 101)]:
Row 1 0.0 0.4369 0.2496 0.1872 0.1248 0.0 0.0
Row 2 0.0 0.2003 0.5342 0.1335 0.0668 0.0668 0.0
[(n1, n2)] = [(101, 151)] :
Row 2 0.0 0.3064 0.7149 0.0
0.0
0.0 0.0
Row 3 0.0 0.0
0.5243 0.5243 0.0
0.0 0.0
Row 4 0.0 0.0
0.0
0.3935 0.5902 0.0 0.0
Row 5 0.0 0.0
0.0
0.0
0.2792 0.5583 0.0
Row 6 0.0 0.0
0.0
0.0
0.0
0.9089 0.0
Step 5:Calculation of m-step Transition Probability Matrices: The 5-step transition probability
matrix [(n0, n5 )], which is [(0, 301)] can be determined using Eq. (4). The entries in the first
two rows and first six columns of the transition probability matrices have non-zero values and
they are given below:
[(n0, n5)] = [(0, 301)] =
Row 1 0.0 0.4369 0.2496 0.1872 0.1248 0.0
Row 2 0.0 0.2003 0.5342 0.1335 0.0668 0.0668
II - 152
Step 6:Calculation of the Unconditional Probabilities of Compliances at Load Cycle 301: This
step is carried out using Eq. (5) to determine the probability density function of compliances at
the end of duty cycle (0, 301). The row vector {(0)} consists of the probability distribution of
initial damage states of the specimens and is determined from the distribution of C(0). For the
present case the row vector {(0)} is given by: {(0)} = {0.516129, 0.483871, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0}. The unconditional probabilities of the compliances at n5=301 is
given below:
{U(0, 301)} = {0.0 0.0967 0.0967 0.3870 0.2580 0.1290 0.0322 0........0}
Prediction of Damage State Probabilities at Load Cycle 351
Now, the duty cycle (301, 351) is considered. Using the transition probability matrices [(0,
101)], [(101, 151)], [(151,201)], [(201, 251)], and [(251, 301)], the transition probability
matrix [(301, 351)] is predicted with a reliability of 90%. All the non-zero transition
probabilities are located within the first six rows and six columns of predicted matrix [(251,
301)]:
[(301, 351)] :
Row 1 0.0 0.3414
Row 2 0.0 1.1415
Row 3 0.0 0.0
Row 4 0.0 0.0
Row 5 0.0 0.0
Row 6 0.0 0.0
0.1950 0.1463 0.0975 0.0

0.7022 0.1043 0.0522 0.0522
0.8515 0.6060 0.0
0.0
0.0
1.0259 0.5545 0.0
0.0
0.0
1.2590 0.4752
0.0
0.0
0.0
1.2924
The unconditional probability matrix is given by {U(0, 351)}: {0.0 0.0709 0.0965 0.2989
0.3494 0.0899 0.0944 0 0........ 0}, and the corresponding histogram is plotted in Figure 2.
From the test data, the histogram of compliances at load cycle 351 is determined (Figure 3).
They are: {0.0 0.0968 0.0968 0.3548 0.2581 0.0968 0.0968 0 0......... 0 }. A reasonable
agreement can be observed.
CONCLUSIONS
The fatigue behaviour of laminated composites is considered and a stochastic approach to model
and analyze this behaviour is developed. The following major sources of variability in
cumulative damage process are incorporated in the stochastic model: (i) Initial damage state, (ii)
Severity and order of both the load cycles as well as the material damage, (iii) State of damage at
the failure of the test specimen. The severity of load cycle and the material damage (including
the changes in the material properties) is specified in the model in terms of the probability
transition matrix. The variations in the severity of repetitive load cycles and the material damage
are incorporated through the non-homogeneous nature of the damage evolution process. The
randomness in the initial damage due to manufacturing errors, technological process defects,
design errors etc., is specified by the initial damage state vector. The stochastic model developed
in the present work possesses the capability of prediction.
II - 153
REFERENCES
1. Weibull, W. 1951. A Statistical Distribution Function of Wide Applicability. Transactions of
the ASME: Journal of Applied Mechanics, 1951, Vol. 18, pp. 293-297.
2. Hahn, H. T. and Kim, R. Y. 1975. Proof Testing of Composite Materials. Journal of
Composite Materials, Vol. 9, pp. 297-311.
3. Ramani, S. V. and Williams, D. P. 1977. Notched and Unnotched Fatigue Behaviour of
Angle-Ply Graphite/Epoxy Composites. In: Fatigue of Filamentary Composite Materials,
ASTM STP 636. Eds.: K. L. Reifsnider and K. N. Lauraitis, pp. 27-46.
4. Talreja, R. 1987. Fatigue of Composite Materials. Technomic publishing Company,
Lancaster.
5. Cox, D. R. and Miller, H. D. 1965. The Theory of Stochastic Processes, Metheun and Co,
London.
6. Bendat, J. S. and Piersol, A. G. 1986. Random Data: Analysis and Measurement Procedures,
Wiley Interscience, New York.
7. Siddall, J. N. 1983. Probabilistic Engineering Design: Principles and Applications. Marcel
Dekker, New York.
8. Siddall, J. N. 1982. Optimal Engineering Design. Marcel Dekker, New York.
9. Ni, R. G. and Adams, R. D. 1984. The Damping and Dynamic Moduli of Symmetric
Laminated Composite Beams - Theoretical and Experimental Results. Journal of Composite
Materials, Vol. 18, pp. 104-121.
10. Cardrick, A. W. and Smith, M. A. 1974. An Approach to the Development of Meaningful
Design Rules for Fatigue-Loaded CFRP Components, Composites, Vol. 5, pp. 96-100.
II - 154
II - 155
PROGRESSIVE FATIGUE MODELING OF

LAMINATED COMPOSITE PINNED/BOLTED
CONNECTIONS
Mahmood M. Shokrieh1 and Larry B. Lessard2
1
Department of Mechanical Engineering, Iran University of Science and Technology

Tehran, Iran
2 Department of Mechanical Engineering, McGill University, 817 Sherbrooke St. West
Montral, Qubec, H3A 2K6, Canada
SUMMARY: In this research, a deterministic model for predicting the behaviour of

laminated composite pinned/bolted connections under fatigue loading is described. A
phenomenological modeling technique, called progressive fatigue damage modeling, is
established which is capable of predicting the residual strength, residual stiffness, and fatigue
life of composite laminates. Stress analysis, failure analysis, and material property
degradation rules are the three major components of the model. A three dimensional,
nonlinear, finite element technique is proposed for the stress analysis. Based on the state of
stress, different failure modes are detected by a set of fatigue failure criteria. To remove the
requirement of the large experimental data base for the failure analysis a normalization
technique is established. By using this technique the restriction of the application of the
failure criteria to limited states of stresses is overcome. Material properties of each element
are degraded by using sudden and gradual material property degradation rules. For this
purpose an analytical technique is established to predict the degradation of material properties
of a unidirectional ply under multiaxial fatigue loading. Based on the model, a computer code
is developed that simulates cycle by cycle behaviour of composite laminates under fatigue
loading.
KEYWORDS: fatigue, pin, bolt, progressive modeling, failure criteria
INTRODUCTION
One of the earliest papers in fatigue of composite laminates with stress concentrations was
published by Owen and Bishop [1]. Today, there are several experimental [2-5] and
analytical [6-8] research papers in this field. However, the present state of knowledge is still
in the stage of development and improvement. Among the existing models, the semiempirical model of Kulkarni et al. [6], the critical element model of Reifsnider et al. [7], and
the damage growth model of Spearing et al. [8] are all systematic and methodological,
therefore they are worthwhile for discussion. The semi-empirical model of Kulkarni et al. [6]
was developed based on a mechanistic wearout [9] framework. The wearout philosophy
treats fatigue damage as the growth of pre-existing flaws or discontinuities in a material. By
growing the flaws, the strength of the material decreases and reaches to the level of the state
of stress and finally, catastrophic fatigue failure occurs. The semi-empirical model was
evaluated by experimental techniques, but little correlation between the theory and
II - 156
experiments was found. Although the semi-empirical model of Kulkarni et al. [6] never
showed a correlation with experimental results, there are however many interesting points and
valuable information in their work. The critical element model of Reifsnider et al. [7] was
developed based on a mechanistic approach. Their mechanistic approach is based on
micromechanical representations of strength. Sendeckyj [10] listed some of the weaknesses
of this model. In the critical model, it is assumed that failure occurs suddenly and everywhere
in the lamina, which is not consistent with the experimental observations. Also in the model,
the dependence of lamina failure stresses on the lamina thickness in laminates is not properly
considered. Moreover, the model does not properly account for delamination between the
laminae in the laminate. The damage growth model of Spearing et al. [8] is based on
quantification of notch tip damage by the extent of the individual failure processes, such as
splitting in the 0o plies and delamination between the 0o ply and off-axis plies. There are
some limitations in the damage growth model which are listed here. The model is laminate
geometry dependent, i.e., by changing the lay-up of the laminate, the model must be modified.
Moreover, the existence of transverse ply cracks, which is an important failure mode, is
ignored in their model. Furthermore, they assume that the damage pattern is similar in all
plies and delamination shape is the same at all interfaces which is not a general assumption.
Also, the shape of the notch is pre-defined. In addition, their model is only capable of
considering cyclic tensile loading. Thus all existing models for fatigue analysis of composites
have limitations which make them unsuitable for general use.
PROBLEM STATEMENT
Consider a composite plate with a hole under fatigue pin/bolt loading. The coordinate axes,
dimensions, and variables are shown in Fig. 1. At the start of cyclic loading, the strength of
the material is greater than the stress state, therefore there is no static mode of failure
anywhere on the plate. By increasing the number of cycles from zero, based on the stress
states at each point of the plate, material properties at those points are degraded as a function
of number of cycles. By increasing number of cycles with more degradation of the material
properties and redistribution of stresses, failure begins in regions where the strength of the
material is less than the stress state at that point. After failure initiation, stresses are
redistributed around failed regions. By further increasing the number of cycles, failure
propagates in different directions. Finally after a certain number of cycles the laminate cannot
tolerate additional cycles. At this point the maximum number of cycles is reached and the
laminate has failed completely.
II - 157
Fig. 1: Composite plate under pin/bolt load
COMPONENTS OF THE MODEL

Progressive fatigue damage model is an integration of three major components. These are
stress analysis, failure analysis, and material property degradation rules. There are some
similarities between static and fatigue progressive damage modeling. However, there are
many features in fatigue progressive damage modeling that make this technique quite
different from static progressive damage modeling. Details of these differences will be more
clear in the next sections. In the following, after a brief explanation of each component, the
model is established by the integration of the three components.
Stress Analysis
By considering edge effects, delamination, out-of-plane buckling, and stacking sequence
effects that are fully three dimensional phenomena, and because of material nonlinearity, a
three dimensional nonlinear finite element technique is used for the stress analysis. Assuming
symmetric geometry, loading, and ply lay-up, suitable boundary conditions are used (Fig. 2).
Radial fixed displacement boundary conditions have been applied on nodes around the hole at
the load region to simulate the pin load. To simulate a fully tightened bolt, a uniform pressure
is applied on the washer region. Moreover, to simulate friction between the washer and the
plate, fixed displacement boundary conditions in radial and tangential directions are applied
on nodes on the surface of the composite plate and under the washer region. The material
properties of a unidirectional ply are assumed to be transversely isotropic. By this
assumption, in three dimensions, the number of material parameters (strengths, stiffnesses,
and Poisson's ratios) is reduced from eighteen to eleven. To achieve high accuracy, the
isoparametric quadratic 20-node brick element was used. Also by using a large number of
elements near the edge of the hole and at layer interfaces, stress singularities have been
considered.
II - 158
Fig. 2: Finite element model

Failure Analysis
Despite the existence of an extensive amount of research on biaxial/multiaxial fatigue of
metals [11], research in the same field on composite materials [12,13] is less complete.
Literature reviews on multiaxial and biaxial fatigue loading of composite materials presented
by Found [12], and Chen and Matthews [13] state that further research in this field is needed.
In this research, seven different failure modes for the unidirectional ply under multiaxial state
of stress are considered which are: fiber tension, fiber compression, fiber-matrix shearing,
matrix tension, matrix compression, normal tension and normal compression failure modes.
Suitable stress-based failure criteria for detecting these modes of failure under static and
fatigue loading conditions are derived. The effect of material nonlinearity on the
mathematical form of the failure criteria is also considered. Difficulties and limitations of
application of such failure criteria, in traditional forms, for unidirectional plies under
multiaxial fatigue loading conditions for general stress state and stress ratios are discussed.
Under fatigue loading conditions, the material is loaded by a stress state which is less than the
maximum strength of the material, therefore there is no static mode of failure. However, by
increasing the number of cycles, the material properties degrade and eventually lower to the
level of the stress state and, at this point, catastrophic failure occurs. The idea of using
polynomial failure criteria to predict the life of a composite ply under multiaxial fatigue
loading has been utilized by many investigators [14-16]. They used the fatigue strength, as a
function of number of cycles, in the denominators of failure criteria instead of the static
strength of the material. This strategy is potentially beneficial, however in practice, the
application of their models are restricted to very specific conditions.
To show the restriction of application of fatigue failure criteria in traditional forms, consider
the following fatigue failure criterion introduced by Hashin [15] for fiber tension fatigue
failure mode of a unidirectional ply under a two-dimensional state of stress (biaxial fatigue),
2
xy
xx
= g2 +
+
f
X t (n,, )
Sxy (n,,)
(if gf+ > 1, then failure)
(1)
where X t (n,,) is the residual longitudinal tensile strength of a unidirectional ply under
uniaxial fatigue loading, and S xy (n, , ) is the residual in-plane shear strength of a
unidirectional ply under uniaxial shear fatigue loading conditions. Both Xt and Sxy are
functions of n, and , which are number of cycles, stress state and stress ratio, respectively.
In practice, designers must deal with a wide range of states of stress, varying from low to
high. Therefore, in order to apply Eq. 1, the residual longitudinal tensile fatigue strength and
II - 159
residual in-plane shear fatigue strength of a unidirectional ply ( X t (n,,) and Sxy (n, , ) ) must be
fully characterized under different stress levels and stress ratios. This requires a large
quantity of experiments just to predict the fiber in tension fatigue failure mode of a
unidirectional ply under simple biaxial fatigue loading conditions. By considering the other
modes of failure and the multiaxial states of stress which are encountered in the real fatigue
design of composite structures, the proposed method is faced with severe difficulties.
To overcome the difficulties arising from the large quantity of experiments required at
different stress states and stress ratios to characterize the material, many investigators [14-16]
restricted their models to specific stress ratios. This assumption is too restrictive for general
cases. For example, in the analysis of a pin/bolt fatigue loaded composite laminate, using a
constant stress ratio leads to incorrect results. Clearly, for this problem there are different
states of stress at different points in the material. Also, after applying fatigue load on a
notched composite laminate, failure initiates near the stress concentrations and the material
property degrades, therefore the stress ratio and the state of stress are not constant at different
points. This means that in practice, stresses redistribute during the fatigue loading. By
considering the different behaviour of a unidirectional ply for each combination of the stress
state and stress ratio, an infinite number of experiments would be required in order to fully
characterize the residual properties of a unidirectional ply under arbitrary state of stress and
stress ratio. To eliminate the aforementioned obstacle of using the quadratic polynomial
failure criteria for a wide range of stress state and stress ratio, a novel technique (generalized
residual material property degradation model) which has been explained in detail in other
articles [17,18] published by authors. The fatigue failure criteria for different modes of
failure are similar to static failure criteria, except that the material properties are not constants
but functions of number of cycles, stress state, and stress ratios. It should be added that the
effects of material nonlinearity on the fatigue failure criteria are also considered. A list of
different fatigue failure criteria suitable for detecting various modes of failure is explained
elsewhere [19] and for brevity is not repeated in this paper.
Material Property Degradation
As failure occurs in a ply of a laminate, material properties of that failed ply are changed by
the material property degradation rules. For each mode of failure, there exists an appropriate
sudden material property degradation rule. Some of the failure criteria are catastrophic and
some of them are not. Thus for each failure mode a suitable rule exists. For instance, after
detecting fiber tension failure mode, which is a catastrophic mode, all of the stiffnesses and
Poisson's ratios of the failed ply are reduced abruptly to zero. As another example, for matrix
tension failure mode, only the transverse stiffness and Poisson's ratios are changed to zero.
Moreover in fatigue loading, after each number of cycles, material properties of each ply
(strengths, stiffnesses, and Poisson's ratios) are degraded, even though failure is not detected
by any of the above mentioned failure criteria. For this purpose an analytical technique is
established to predict the gradual degradation of material properties of a unidirectional ply
under multiaxial fatigue loading.
The residual strength of a unidirectional ply under arbitrary uniaxial state of stress and stress
ratio is calculated using the following equation:
II - 160
log(n ) log(.25)
(R s ) +
R (n, , ) = 1
log(N f ) log(.25)
(2)
where, R(n,,) = residual strength, Rs = static strength, n = number of applied cycles, =

magnitude of applied maximum stress, Nf = fatigue life at , = stress ratio, and and =
experimental curve fitting parameters. The residual stiffness of a unidirectional ply under
arbitrary uniaxial state of stress and stress ratio is calculated using the following equation:
1
log( n) log(.25)

Es +
E (n,, ) = 1
log N log(.25)
f f
( f)
(3)
where, E(n,,) = residual stiffness, Es = static stiffness, f = average strain to failure, and
and = experimental curve fitting parameters. The normalized fatigue life of a unidirectional
ply under arbitrary uniaxial state of stress and stress ratio is calculated using the following
equation:
u=
ln(a f )
= A + Blog N f
ln[(1 q )(c + q )]
(4)
where, f and u = curve fitting parameters, a = (max - min)/2 = alternating stress, m =

(max + min)/2 = mean stress, q = m/t, a = a/t, c = c/t, t = tensile strength, c =
compressive strength. For the shear loading condition Eq. 4 must be modified. For a
unidirectional ply under shear, the definitions of tensile strength and compression strength are
meaningless, i.e., there is no difference between positive and negative shear. Therefore, the
word strength is used instead. By considering this, the parameter c in Eq. 4 is equal to one
(c=1) for shear fatigue conditions. Also, the experimental results for unidirectional plies
under in-plane shear loading conditions show that a better curve fitting is achieved by adding
a log10 to the left-hand side of Eq. 4. Therefore, Eq. 4 is changed to the following form for
the simulation of fatigue life of the unidirectional ply under shear loading conditions. The
left-hand side of Eq. 5 is still denoted by u for brevity.
u = log10 (
ln(a f )
) = A + Blog Nf
ln[(1 q)(1 + q)]
(5)
PROGRESSIVE FATIGUE DAMAGE MODEL

The model is explained by means of a flowchart shown in Fig. 3. As shown in this figure, the
finite element model must first be prepared. In this step, material properties, geometry,
boundary conditions, maximum and minimum fatigue load, maximum number of cycles,
incremental number of cycles, etc., are defined. Then the stress analysis, based on the
maximum and minimum fatigue load, is performed. Consequently, the maximum and
minimum induced on-axis stresses of all elements are calculated. It should be emphasized
that on-axis stresses for each ply of each laminate of each element at Gauss points are
II - 161
calculated and averaged. Therefore, the stress ratio ( = min/max) for each element is
determined. In the next step, failure analysis is performed and the maximum stresses are
examined by the set of fatigue failure criteria. If there is a sudden mode of failure, then the
material properties of the failed plies are changed according to appropriate sudden material
property degradation rules. The stiffness matrix of the finite element model is rebuilt and the
stress analysis is performed again. New stresses are examined by the set of fatigue failure
criteria. In this step, if there is no sudden mode of failure, an incremental number of cycles
are applied (e.g., n = 100). If the number of cycles is greater than a preset total number of
cycles, then the computer program stops. Otherwise, material properties of all plies of all
elements are changed according to gradual material property degradation rules using the
generalized material property degradation technique. Then stress analysis is performed again
and the above loop is repeated until catastrophic failure occurs, or the maximum number of
cycles (pre-defined by the user) is reached. If catastrophic failure is reached, then fatigue life
and the mechanisms of failure due to fatigue loading have been achieved. It should be noted
that if the maximum number of cycles is selected as a large number, such as 109, then the
fatigue life of the problem can be obtained by the algorithm. If the computer program stops
because the user-chosen maximum number of cycles is reached, then the mechanisms of
failure due to fatigue loading are found by examining the final state of damage. Furthermore,
in the latter case, residual strength of the composite laminate is obtainable by performing a
progressive static damage modeling on the final results of progressive fatigue damage model.
Fig. 3: Flowchart of progressive fatigue damage model

In order to verify the fatigue progressive damage model, a complete knowledge of material
properties (stiffness, strength, and fatigue life) of a unidirectional ply under static and fatigue
loading conditions is required. Hence, to run the computer program developed for
progressive fatigue modeling of laminated composites, the mechanical behaviour of a
unidirectional ply under uniaxial static and fatigue loading conditions must be fully
characterized. For this purpose, the material properties of a unidirectional composite laminate
in static and fatigue loading conditions; in the fiber and matrix directions; under tensile,
compressive, in-plane-shear and out-of-plane-shear loading are required.
II - 162
RESULTS
To verify the fatigue simulation capability of the model, fatigue life and residual strength tests
are performed for different ply configurations and load levels. To show the generality of the
model, simulated results for fatigue life of the pin/bolt-loaded composite laminates are also
compared with independent experimental results available in the literature. Also, the residual
fatigue life of the pin-loaded composite laminates under two consecutive load levels (low and
high) is measured experimentally and the results are compared with simulated results by the
model. In the following some of the typical results obtained are presented. Additional results
are published by the authors elsewhere [17-19]. The results of the fatigue life of pin-loaded
[04/904]s and [904/04]s cross-ply laminates made of AS4/3501-6, are shown in Fig. 4. As
shown, similar to the static strength behaviour, the pin-loaded [904/04]s laminate shows a
higher fatigue life than the [04/904]scross-ply laminate for the same fatigue loading
conditions. The main reason for this behaviour is explained by the lower singular stress state
for the [904/04]s cross-ply laminate near the edge of the hole.
To evaluate the progressive fatigue damage model by independent experimental data, the
fatigue life of a pin-loaded quasi-isotropic [0/90/45]s laminate made of AS4/3501-6 is
simulated and compared with the experimental results of Herrington and Sabbaghian [20].
The fatigue life of the pin-loaded quasi-isotropic laminate is simulated by the progressive
fatigue damage model and compared with the experimental results (Fig. 5).
Fig. 4: Fatigue life curves of pin-loaded [04/904]s and [904/04]s laminates
II - 163
Fig. 5: Fatigue life of pin-loaded quasi-isotropic [0/45/90]s laminates (simulation by the

model and experiments by Herrington and Sabbaghian [20])
The progression of damage in a pin-loaded angle-ply [+454/-454]s laminate at different

number of cycles is shown in Fig. 6. The simulated number of cycles to failure is 25000
cycles, while the experiments show a range of 15000 to 26900 cycles. As shown, an excellent
agreement between the simulated and experimental results is achieved for mode of failure.
simulated by the model after 0, 18000 and 25000 fatigue loading cycles
experiment at final failure
Fig. 6: Progression of damage in a pin-loaded angle-ply [+454/-454]s laminate
II - 164
SUMMARY
The present research establishes a new modeling approach to simulate the behaviour of
composite laminates under general fatigue loading conditions. Using the idea of the
traditional progressive damage model which is capable of simulating of static behaviour of
composite laminates, the progressive fatigue damage model is developed. The model consists
of three major components: stress analysis, failure analysis and material property degradation
rules. The model determines the state of damage at any load level and number of cycles, from
failure initiation and propagation to catastrophic failure. The model is able to predict the
residual strength, residual life, final failure mechanisms and final fatigue life of composite
laminates under general fatigue loading conditions. Based on the model, a computer code is
developed which simulates the cycle-by-cycle behaviour of composite laminates under fatigue
loading conditions. The capabilities of the model are examined by simulation of a pin/boltloaded composite laminates as a complicated example.
REFERENCES
1.
Owen, M. J. and Bishop, P. T., Fatigue Properties of Glass-Reinforced Plastics

Containing a Stress Concentrator, Journal of Physics - D: Applied Physics, Vol. 6,
(1973), pp. 2057-2069.
2.
Simonds, R. A. and Stinchcomb, W. W., Response of Notched AS4/PEEK Laminates

to Tension/Compression Loading, Advances in Thermoplastic Matrix Composite
Materials, ASTM STP 1044, G. M. Newaz, Ed., (1989), pp. 133-145.
3.
Bakis, C. E., Simonds, R. A. and Stinchcomb, W. W., A Test Method to Measure the
Response of Composite Materials Under Reversed Cyclic Loads, Test Methods for
Design Allowables for Fibrous Composites, ASTM STP 1003, C. C. Chamis and K. L.
Reifsnider, Eds., (1989), pp. 93-110.
4.
Xiao, J. and Bathias, C., Fatigue Behaviour of Unnotched and Notched Woven
Glass/Epoxy Laminates, Composites Science and Technology, 50, (1994), pp. 141148.
5.
Xiao, J. and Bathias, C., Fatigue Damage and Fracture Mechanism of Notched Woven
Laminates, Journal of composite Materials, Vol. 28, No. 12, (1994), pp. 1127-1139.
6.
Kulkarni, S. V., McLaughlin Jr., P. V., Pipes, R. B. and Rosen, B. W., Fatigue of
Notched Fiber Composite Laminates: Analytical and Experimental Evaluation,
Composite Materials: Testing and Design (Fourth Conference), ASTM STP 617,
(1977), pp. 70-92.
7.
Reifsnider, K. L., The Critical Element Model: A Modeling Philosophy, Engineering

Fracture Mechanics, Vol. 25, Nos. 5/6, (1986), pp. 739-749.
8.
Spearing, S. M. and Beaumont, P. W. R. and Kortschot, M. T., The Fatigue Damage

Mechanics of Notched Carbon Fibre/PEEK Laminates, Composites, Vol. 23, No. 5,
(Sept., 1992), pp. 305-311.
9.
Halpin, J. C., Jerina, K. L. and Johnson, T. A., Characterization of Composites for the
Purpose of Reliability Evaluation, Analysis of the Test Methods for High Modulus
Fibers and Composites, ASTM STP 521, (1973), pp. 5-64.
II - 165
10.
Sendeckyj, G. P., Life Prediction for Resin-Matrix Composite Materials, Fatigue of

Composite Materials, Ed. by K. L. Reifsnider, (1990), pp. 431-483.
11.
Garud, Y. S., Multiaxial Fatigue: A Survey of the State of the Art, Journal of Testing
and Evaluation, JTEVA, Vol. 9, No. 3, (May 1981), pp. 165-178.
12.
Found, M. S., A Review of the Multiaxial Fatigue Testing of Fiber Reinforced

Plastics, Multiaxial Fatigue, ASTM STP 853, K. J. Miller and M. W. Brown, Eds.,
(1985), pp. 381-395.
13.
Chen, A. S. and Matthews, F. L., A Review of Multiaxial Biaxial Loading Tests for
Composite Materials, Composites, Vol. 24, No. 5, (1993), pp. 395-406.
14.
Sims, D. F. and Brogdon, V. H., Fatigue Behavior of Composites under Different

Loading Modes, Fatigue of Filamentary Materials, ASTM STP 636, K. L. Reifsnider
and K. N. Lauraitis, (1977), pp. 185-205.
15.
Hashin, Z., Fatigue Failure Criteria for Unidirectional Fiber Composites, Journal of
Applied Mechanics, Vol. 48, (Dec. 1981), pp. 846-852.
16.
Ryder, J. T. and Crossman, F. W., A Study of Stiffness, Residual Strength and Fatigue
Life Relationships for Composite Laminates, NASA Contract Report CR-172211,
1983.
17.
Shokrieh, M. M., and Lessard, L. B., Multiaxial Fatigue Behaviour of Unidirectional

Plies Based on Uniaxial Fatigue Experiments: Part I. Modeling, International Journal
of Fatigue, accepted, 1996.
18.
Shokrieh, M. M., and Lessard, L. B., Multiaxial Fatigue Behaviour of Unidirectional

Plies Based on Uniaxial Fatigue Experiments: Part II. Experimental Evaluation, Int.
Journal of Fatigue, accepted, 1996.
19.
Shokrieh, M. M., Progressive Fatigue Damage Modeling of Composite Materials, Ph.

D. Thesis, McGill University, Department of Mechanical Engineering, Montreal,
Quebec, Canada, 1996.
20.
Herrington, P. D. and Sabbaghian, M., Fatigue Failure of Composite Bolted Joints,

Journal of Composite Materials, Vol. 27, No. 5, (1993), pp. 491-512.
II - 166
PREDICTION OF FLEXURAL FATIGUE LIFE FOR

UNIDIRECTIONAL CFRP LAMINATES
Masayuki Nakada1, Takeshi Ishiguro2 and Yasushi Miyano1
1
Materials System Research Laboratory, Kanazawa Institute of Technology,

Ohgigaoka Nonoichi, Ishikawa 921, Japan
2
Graduate School, Kanazawa Institute of Technology,
Ohgigaoka Nonoichi, Ishikawa 921, Japan
KEYWORDS: fiber reinforced plastics, static strength, creep strength, fatigue strength,
time-temperature superposition principle
II - 167
II - 168
II - 169
II - 170
II - 171
II - 172
II - 173
II - 174
II - 175
II - 176
FATIGUE BEHAVIOR OF THE CABON/EPOXY AND

THE CARBON/PEEK COMPOSITES SUBJECTED TO
LOW-ENERGY IMPACT
Nyan-Hwa Tai1, Ming-Chuen Yip1,2, Chih-Ming Tseng2 and Hui-Chia Yu2
1
Materials Science Center and 2 Department of Power Mechanical Engineering

National Tsing- Hua University, Taiwan, R.O.C.
SUMMARY: Fatigue behaviors of carbon fiber reinforced thermosetting and thermoplastic

composites with and without subjecting low-energy impact were investigated. The [0/45/90/45]2S T300/976 carbon/epoxy and AS-4/PEEK laminates were used in this study. The impact
response of these two laminates shows a linear force-deformation relationship in the initial
penetration impact loading, which implies an elastic behavior of these two composites; the
following zigzag peaks indicate fracture initiation and arrest. The fatigue tests were performed
at various stress levels under tension-tension loadings with a fatigue frequency of 3 and stress
ratio of 0.1. The median rank method and Weibull distribution were applied to predict the
fatigue probabilities of the specimen under static and fatigue tests. The S-N curves for the
undamaged and low-energy impact damaged laminates are compared.
KEYWORDS: low-energy impact, fatigue behavior, T300/976 carbon/epoxy, AS-4/PEEK,

Weibull distribution, S-N curve
INTRODUCTION
Although fiber reinforced composite shows many remarkable features such as lower specific
strength and modulus, better corrosion resistance, and precise tailoring, the inherent
anisotropic property of fiber reinforced composites results in much more complicated fracture
mode than that of the conventional materials. During the past two decades, carbon/epoxy
composites have been adopted in many applications, and numerous studies on the fatigue
properties of the laminates have been reported[1,5]. On the other hand, carbon fiber
reinforced thermoplastic composite such as Gr/PEEK laminate is a relatively new material;
only a few studies were performed on the topic of fatigue behavior for Gr/PEEK laminates[68]. Besides fatigue behavior, impact behavior is also a very important property for
composites in engineering applications[9]. On the applications of composites in transportation
vehicles, the damages can be introduced during maintenance such as dropped tool and during
service such as debris rebound; all these damages result in reduction of performance.
However, very few articles were published to investigate the effects of the combined loading,
fatigue and impact, on the performance of the Gr/epoxy and Gr/PEEK composites. The
purposes of this work are to study the fatigue behavior of the low-energy impacted
composites and to compare the performance of the quasi-isotropic thermosetting (Gr/epoxy)
and thermoplastic (Gr/PEEK) composites.
II - 177
EXPERIMENTAL
P3051F Gr/epoxy and AS-4/PEEK prepregs were supplied by Toray in Japan and Imperial
Chemical Industry (I.C.I.) in U.K., respectively. The Gr/epoxy and Gr/PEEK laminates were
fabricated in autoclave and hot press, respectively. Specimens were cut from fabricated panels
through a water-cooled diamond table saw. The dimensions of the test coupons are 227mm x
19.1 mm x 2mm with a gage length of 127 mm. Non-tapered glass/epoxy end tabs with
dimensions of 50 mm x 19.1 mm x 1.6 mm were used.
The cross-section area of the test specimens for penetration impact was 100 mm x 100 mm.
The impact procedure was conducted using a steel impactor of cylindrical shape with a
hemispherical nose 19.0 mm in diameter and 17.01 kg in weight. The diameter of the support
ring is 76 mm.
Low-velocity impact tests were conducted using a Dynatup/GRC drop weight impact tester.
The specimens were firmly fixed between flat circular rings with a support ring 18 mm in
diameter. The impact machine is equipped with a hemispherical nose 12.7 mm in diameter
and a rebound apparatus to prevent multiple impacts. Various energies for low-energy impact
tests were employed by changing the hammer height.
THEORETICAL ANALYSIS
The two-parameter Weibull distribution function, which is shown in Eq. (1), was adopted to
analyze strength, residual strength after impact, and fatigue life.
X s
i

R i (X i )= 1 exp
Xo
(1)
The median rank is employed in Eq. (2) for failure probability, R(Xi) at Xi, 1<i<m,
R (X i )=
i 0.
3
m + 0.
4
(2)
where Xi denotes the test strength of fatigue life of the ith specimen.
Based upon the data of fatigue cycles to failure under various stress levels, the S-N curves
were plotted using linear regression method.

Figures 1a and 1b show the relationship among force, deformation, and energy during
penetration impact loading for both Gr/epoxy and Gr/PEEK laminates, respectively. Although
the differences in energies required to penetrate these two laminates are not obvious, the
responses during impact loading of these two composites are different. A linear relation
between force and deformation for these two laminates can be observed during the initial
penetration process, which implies an elastic behavior of the composites. After the linear
region, a force relaxation which indicates the formation of a fracture surface is detected for
II - 178
Gr/epoxy laminates only. On the other hand, no obvious peak can be detected before reaching
the peak load for Gr/PEEK laminates. As compared the deformation at which the peak load is
reached of these two laminates, Gr/PEEK composites show greater deformation than
Gr/epoxy composites, which implies better damage tolerance of the Gr/PEEK laminates.
Further verification can be made by comparing the response of the composites after peak
load; a sharp drop in force is detected, which indicates the formation of a large fracture
surface in Gr/epoxy composites. However, only slightly reduction in force can be observed
for Gr/PEEK laminates, which implies less force relaxation and fracture surface formation.
The total energy to penetrate the laminate is denoted as Ei.
Figure 2 shows the strengths of the unimpacted specimens and the impact-damaged
specimens loaded under various impact energies. The strength is reduced with increasing
impact energy for both composites. Moreover, the rate of decrease in tensile strength for
Gr/PEEK is higher than that of the Gr/epoxy laminates. After impact loading, a wider
distribution in tensile strength is also detected.
The failure probabilities of the unimpacted and the 25% Ei impacted Gr/epoxy laminates were
compared in Figure 3a; a slight difference in failure probability can be observed; and the
impact damage results in a broader distribution in strength is also detected. However, nearly
no difference in failure probability predicted by Weibull function can be observed for
Gr/PEEK laminates (Figure 3b), although some experimental data have a significant deviation
from theoretical prediction.
Figure 4a shows the S-N curves of unimpacted and impacted Gr/epoxy composites. A linear
stress-fatigue life relationship is obtained. The shift of the curve implies that low-energy
impact results in lower fatigue life, however, no significant variation in slope is observed,
which indicates both composites (unimpacted and impacted) having similar fatigue
sensitivity. Figure 4b shows the S-N curves of the unimpacted and 25%Ei impacted Gr/PEEK
composites, and it reveals that low-energy impact loading results in the reduction of fatigue
life for Gr/PEEK composites. A linear S-N curves were obtained using linear regression
analysis. In our previous study, the Gr/PEEK composite shows two-segment S-N curves[6];
the difference between these two results may be due to different testing conditions and
analytical methods. Furthermore, it is also observed that fatigue life distribution becomes less
spread at lower stress level when the Gr/PEEK composites were subjected to a low-energy
impact.
CONCLUSIONS
Penetration impact loadings were applied to Gr/epoxy and Gr/PEEK laminates; Gr/epoxy
shows a brittle fracture behavior, whereas Gr/PEEK shows a plastic fracture mode. The
tensile strength reduces with increasing impact energy for both Gr/epoxy and Gr/PEEK
laminates. A good correlation between experimental data and theoretical analysis for failure
prediction in tensile strength is achieved. The data of fatigue life to failure indicate that the
fatigue life is reduced when the specimens were subjected to a low-energy impact, and the
distribution becomes less spread for both Gr/epoxy and Gr/PEEK composites.
II - 179
ACKNOWLEGEMENT
This work was supported by the National Science Council of the Republic of China, under
Contract No. NSC 86-2212-E-007-037.
REFERENCES
1. Yang, J.N., Jones, D.L.,, and, S.H. and Meskini, A., A Stiffness Degradation Model for
Graphite/Epoxy Laminates, J. Composite Materials, Vol. 24,1990, pp. 753-769.
2. Hahn, H.T. and Hwang, D.G., Failure Characterization of a Graphite/Epoxy Laminate
Through Proof Testing, Composite Materials: Testing and Design (Sixth Conference),
ASTM STP 787, I.M. Daniel, Ed., American Society for Testing and Materials, 1982, pp.
247-273.
3. Whitney, J.M., Fatigue Characterization of Composite Materials, Fatigue of Fiber
Composite Materials, ASTM STP 723, American Society for Testing and Materials,
Philadelphia, 1981, pp. 133-151.
4. Rotem, A. and Nelson, H.G., Residual Strength of Composite Laminates Subjected to
Tensile-Compressive Fatigue Loading, J. Composites Technology and Research, Vol. 12,
No. 2, 1990, pp.76-84.
5. Spearing, S.M. and Beautmont, P.W. R., Fatigue Damage Mechanisms of Composite
Materials Part III: Prediction of Post-Fatigue Strength, Composites Science and
Technology, Vol. 44, 1992, pp. 299-307.
6. Tai, N.H. Ma, C.C.M., and Wu, S.H., Fatigue Behavior of Carbon Fiber/PEEK Laminates
Composites, Composites, Vol. 28,No. 8, 1995, pp.551-559.
7. Curtis, D.C., Moore, D.R., Slater, B., and Zahlan, N., Fatigue Testing of Multi-angle
Laminates of CF/PEEK , Composites, Vol.19, No. 6, 1988, pp. 446-452.
8. OBrien, T.K., Fatigue Delamination Behavior of PEEK Thermoplastic Composite
Laminates, J. Reinf. Plast. Compos. Vol. 4, No. 4, 1988, pp.341-359.
9. Cantwell, W.J. and Morton, J., The Impact Resistance of Composite Materials - A
Review, Composites, Vol. 22, No. 5, 1991, pp. 347-362.
II - 180
3 0 .0
4 0 0 0 .0
2 0 .0
2 0 0 0 .0
1 0 .0
0 .0
(HJ XH
QU \MR O V
) U HQZ Q
RF H WR V
6 0 0 0 .0
0 .0
0 .0
5 .0
1 0 .0
1 5 .0
2 0 .0
2 5 .0
'
HIRUP
DWLRQP
P
1(a)
3 0 .0
4 0 0 0 .0
2 0 .0
2 0 0 0 .0
1 0 .0
0 .0
(HJ XH
QU \MR O V
) U HQZ Q
RF H WR V
6 0 0 0 .0
0 .0
0 .0
5 .0
1 0 .0
1 5 .0
2 0 .0
2 5 .0
'
HIRUP
DWLRQP
P
1(b)
Fig. 1: Relationship among deformation, force, and energy during penetration impact
loadings for (a) Gr/epoxy and (b) Gr/PEEK laminates
II - 181
5HVLGXDO 6WUHQJWK03D

,PSDFW (QHUJ\(L

2(a)
5HVLGXDO 6WUHQJWK03D

,PSDFW (QHUJ\(L

2(b)
Fig. 2: Residual strength of the specimens after being subjected to various impact energies
for (a) Gr/epoxy and (b) Gr/PEEEK laminates
II - 182
'LVWULEXWLRQ )XQFWLRQ

XQLPSDFW GDWD
(L LPSDFW GDWD
:HLEXOO GLVWULEXWLRQ

5HVLGXDO 6WUHQJWK LQ ;
3(a)
'LVWULEXWLRQ )XQFWLRQ

XQLPSDFW GDWD
(L LPSDFW GDWD
:HLEXOO GLVWULEXWLRQ

5HVLGXDO 6WUHQJWK LQ ;
3(b)
Fig. 3: Weibull distribution of the composites without and with subjected to 25% Ei impact
loading for (a) Gr/epoxy and (b) Gr/PEEK laminates
II - 183

6WUHVV /HYHO
(L LPSDFW
XQLPSDFW

)DWLJXH &\FOHV WR )DLOXUHORJ 1

4(a)
(L LPSDFW
XQLPSDFW
6WUHVV /HYHO

)DWLJXH &\FOHV WR )DLOXUHORJ 1

4(b)
Fig. 4: S-N curves of the composites without and with subjected to 25% Ei impact loading for
(a) Gr/epoxy and (b) Gr/PEEK laminates
II - 184
INITIATION OF FREE-EDGE DELAMINATION IN A

COMPOSITE LAMINATE UNDER FATIGUE
LOADING
Ran Y. Kim and Allan S. Crasto
University of Dayton Research Institute, Dayton, OH 45469, USA
SUMMARY: This paper is concerned with the onset of free-edge delamination in composite
laminates subjected to fatigue loading. The material system investigated was a carbon fiberreinforced toughened bismaleimide (IM7/5250-4). A quasi-isotropic laminate with a
[902/30]s lay-up was employed and the onset of free-edge delamination determined under
static loading and compression-compression fatigue loading. The interlaminar stresses in the
free-edge region were calculated using an analytical model, taking into consideration residual
curing stresses. The stress for the onset of delamination under static compression and the
cycles for the onset of delamination under fatigue loading at various fatigue stress levels were
experimentally determined. Delamination occurred at the 30/-30 interface for all specimens
under both static and fatigue loadings. The critical component of the interlaminar stresses
was found to be the interlaminar shear stress xz at the 30/-30 interface due to mechanical
loading and curing residual stresses. This stress xz was therefore used to predict the onset of
delamination in conjunction with the S-N relationship for in-plane shear, which was obtained
from the fatigue testing of [45]2s specimens. The experimental data for both static and
fatigue loadings compare well with the analytical prediction.
KEYWORDS: delamination onset, interlaminar shear stress, fatigue, bismaleimide.
INTRODUCTION
Free-edge delamination of composite laminates under in-plane static loading has been
extensively studied since the early 1970s [1-5]. From the results of these studies, for
example [2, 4, 5], the nature of the interlaminar stress components is well understood and the
onset of delamination under static loading is reasonably well predicted. However, little work
has been reported on the onset of delamination under fatigue loading. The objective of this
study was to develop the capability to predict the onset of free-edge delamination in
composite laminates subjected to fatigue loading. The material system investigated was a
carbon fiber-reinforced toughened bismaleimide (IM7/5250-4). A quasi-isotropic laminate
with a [902/30]s lay-up was employed and the onset of delamination determined under
compression-compression fatigue loading. Applied compressive loading was chosen, to
eliminate the interaction of transverse cracks which usually occurs prior to delamination
under applied tension. A laminate was designed to produce an interlaminar normal tensile
stress under applied uniaxial compression. When this laminate is subjected to an axial
compressive stress, delamination occurs due to interlaminar stresses in the free-edge region
prior to the formation of transverse cracks. The interlaminar stresses at the free-edge region
were calculated using an analytical model [4], taking into consideration curing residual
stresses. The maximum-stress failure theory was applied to predict the initiation of
II - 185
delamination. Because of the presence of a severe stress gradient along the free edge of the
specimen, the concept of an effective stress, the average stress over a distance from the free
edge equivalent to the thickness of one ply, was utilized. The thermomechanical properties
required for the analytical calculations were determined experimentally in a complementary
study. The stress-free temperature was also determined from the deflection of an
unsymmetric laminate strip as a function of temperature to calculate the residual interlaminar
stresses due to curing. The stress for the onset of delamination under static compression and
the cycles for the onset of delamination under fatigue loading at various fatigue stress levels
were experimentally determined. Acoustic emission was employed to detect the onset of
delamination, and the result was confirmed visually using a microscope. Delamination
occurred at the 30/-30 interface for all specimens under both static and fatigue loadings. The
critical component of the interlaminar stresses was found to be the interlaminar shear stress
xz at the 30/-30 interface due to mechanical loading and curing residual stresses. This stress,
xz, was therefore used to predict the onset of delamination in conjunction with the S-N
relation for in-plane shear which was obtained from fatigue testing of [45]2s specimens. The
experimental data for both static and fatigue loadings compare well with analytical
predictions.
EXPERIMENTAL
The material system selected for this study was a graphite/bismaleimide, IM7/5250-4,
obtained in the form of unidirectional prepreg tape. A laminate with a stacking sequence of
[902/30]S was fabricated according to the manufacturer's recommended cure cycle and
postcured at 227C for five hours. The panel was cut into rectangular specimens 15 cm long
and 2 cm wide using a diamond-impregnated saw blade. Glass/epoxy end-tabs, 5 cm long,
were bonded at each end leaving a gage section of 5 cm, and the specimen edges were
polished to enhance microscopic detection of delamination. Unidirectional laminates were
also fabricated to determine the necessary thermoelastic properties for calculation of the
interlaminar stresses in the free-edge region and to measure the appropriate strengths required
for the prediction of the onset of delamination. The average fiber content was determined to
be 64 percent by volume. The Poisson's ratio yz was determined from the tensile loading of
a [90]16T specimen with strains measured in y- and z-directions and utilized to calculate Gyz.
A device was designed and built to provide side support to the specimen and thereby prevent
premature failure due to specimen buckling under compressive load. This side support has a
central opening to accommodate an acoustic emission transducer to detect the onset of
delamination under static as well as fatigue loading. An acoustic emission transducer was
mounted on the flat face of the specimen using high-temperature vacuum grease to monitor
the onset of delamination. A brittle white coating was applied to one of the free edges for the
visual observation of the onset of delamination during loading. At the first indication of
delamination, the specimen was unloaded and the delamination confirmed by microscopic
examination of the specimen edge.
Of all the interlaminar stress components, the interlaminar shear stress xz, was found to be
responsible for the free-edge delamination under applied compressive loading. Therefore, to
predict the onset of delamination under fatigue loading, an S-N curve must be generated for
interlaminar shear loading. Fatigue strength for in-plane shear loading was assumed to be
equivalent to fatigue strength for interlaminar shear loading, and fatigue life data were
generated from the uniaxial tension-tension fatigue of the [45]2s specimens. Five stress
levels were chosen with a minimum of five replicates at each stress level. All specimens were
II - 186
fatigued to final failure on an MTS test machine in the displacement-control mode. The
fatigue tests were conducted at a frequency of 10 Hz and a stress ratio of R= 0.1.
Fatigue life data were reduced by employing a statistical procedure which allows the
generation of S-N curves with some statistical value without resorting to an extremely large
data base. The details of this procedure can be found in [6]. The outline of this procedure is
as follows:
(a) The fatigue life data at each stress level are fitted to a two-parameter Weibull distribution
X(n) = exp[-(n/N)a]
(1)
where a is the shape parameter and N is the scale parameter (characteristic fatigue life).
These parameters are estimated using a maximum-likelihood method.
(b) All the fatigue life data for each stress level are normalized with respect to the
characteristic life estimated in step (a). Then the common shape parameter is determined
by pooling the normalized data for all stress levels, and the characteristic life for each
stress level is re-estimated using the common shape parameter.
(c) The estimated characteristic lives are fitted to a classical power law of the S-N
relationship
S(n) = aNb
(2)
where a and b are material constants which are determined by least-square regression
analysis. N is the characteristic fatigue life estimated in step (b).
The fatigue cycles at the onset of delamination were determined under compressioncompression fatigue with a frequency of 5 Hz and R= 1, using a [902/30]S specimen with the
same dimensions as those used in the static tests. Specimens were fatigued until the onset of
delamination and unloaded to confirm the delamination under a microscope. A total of 12
specimens were tested at six different stress levels with 1 to 3 replicates at each stress level.
The location of the interface where delamination occurred was identified for each specimen
under a microscope.
The thermomechanical properties of the material system determined for calculation of
interlaminar stresses and prediction of the onset of free-edge delamination are listed in Table
1.
Interlaminar Stresses in the Free-Edge Region
Interlaminar stresses at each interface in the free-edge region were calculated using the
variational model developed by Pagano [4]. The distribution of interlaminar stresses, z, yz,
and xz, at the 30/-30 interface and the midplane due to an applied compression of 1 MPa and
due to thermal loading alone (from curing residual stresses) are shown in Figures 1-4. There
is only an interlaminar normal tensile stress present at the midplane under either mechanical
loading or thermal loading due to curing residual stresses. All three interlaminar stress
components exist at the 30/-30 interface. Interlaminar stresses at the other ply interfaces were
II - 187
found to be insignificant compared to those at the midplane and the 30/-30 interface. There is
a large gradient in the interlaminar stress components in the vicinity of the free edge. At both
the midplane and the 30/-30 interface, the interlaminar normal stress is tensile due to
mechanical loading, whereas it is compressive due to residual curing stresses. This
compressive interlaminar normal stress increases the applied stress for the onset of
delamination at these interfaces. A large interlaminar shear stress xz exists at the 30/-30
interface and is negative for both applied mechanical loading and curing residual stresses.
Table 1: Thermomechanical properties of IM7/5250-4 under ambient conditions
Longitudinal modulus, Ex, GPa
Transverse modulus, Ey=Ez, GPa
Shear modulus, Gxy=Gxz, GPa
Transverse shear modulus, Gyz, GPa
Major Poissons ratio, xy= xz
Minor Poissons ratio, yz
Coefficient of thermal expansion
Longitudinal, x
Transverse, y=z
Stress-free temperature, C
Fiber volume,Vf, %
Cured ply thickness, mm
Shear strength, MPa
Transverse strength, MPa
165
10.3
5.8
3.3
0.31
0.56
10-6/C
0.45
24.7
177
64
0.127
122
61.3
The subscript x refers to the fiber direction and y and z refer to directions perpendicular to the
fiber direction.
Interlaminar stress, MPa
0.25
x =-1 MPa
0.2
0.15
0.1
0.05
0
-0.05
0
0.1
0.2
0.3
0.4
0.5
Distance from free edge, mm
Fig. 1:Interlaminar normal stress distribution at the midplane due to applied compression
II - 188
0.2
=-1 MPa
x
yz
-0.2
xz
-0.4
-0.6
0
0.1
0.2
0.3
0.4
0.5

Fig. 2:Interlaminar stress distributions at the 30/-30 interface due to applied compression
5
T=-153C
0
-5
-10
-15
-20
-25
-30
0
0.1
0.2
0.3
0.4
0.5
Fig. 3: Interlaminar normal stress distribution at the midplane due to curing residual stress
II - 189
T=-153C
0
-5
yz
z
-10
xz
-15
-20
-25
0
0.1
0.2
0.3
0.4
Distance from free edge, mm,
0.5
Fig. 4: Interlaminar stress distributions at the 30/-30 interface due to curing residual stress
Prediction of the Onset of Delamination Under Static Loading

The results of the interlaminar stress analysis were applied to the prediction of the onset of
delamination in conjunction with a failure theory [7]. The average value of each stress
component at the free edge, instead of the maximum value, was assumed to be the effective
stress level, i.e. that stress which is responsible for failure. The effective stress is obtained by
averaging the interlaminar stress components over a fixed distance ho from the free edge
along the width of the laminate at the interface in question and is given by
ho
I(z) = 1
ho
I(y,z)dy
0
(3)
The fixed distance ho was taken as a thickness of one ply in all cases.
For prediction of the stress level at the onset of delamination and its location (interface), the
maximum stress criterion was applied. Failure via delamination was assumed to occur when
any of the effective stresses, I(z), at each interface first reached the corresponding strength.
By comparison of all interlaminar stresses with the corresponding strengths at the 30/-30
interface and the midplane for the laminate used in this study, the critical effective stress
component was found to be the interlaminar shear stress, xz, at the 30/-30 interface. Thus,
the onset of delamination was predicted when this effective stress reached the interlaminar
shear strength. The interlaminar shear strength of the laminate was assumed to be equal to
the in-plane shear strength of the material system. Results from [8] demonstrate that the
difference between the interlaminar shear strength and in-plane shear strength is less than two
percent for the graphite/epoxy AS4/3501-6 composite system.
The predicted stress for the onset of delamination compares fairly well with the experimental
value as shown in Table 2. The experimental value is an average of four specimens with a
coefficient of variation, (Cv), of 4.5 percent.
II - 190
Table 2: Comparison of prediction and experiment for onset of delamination under static
loading
[902/30]s
Cv,%
Onset stress, MPa

Prediction
Experiment
346
312
4.5
Delaminated interface
Prediction
Experiment
30/-30
30/-30
-
Delamination occurred at the 30/-30 interfaces for all specimens, as shown in Figure 5. The
exact location of the delamination with respect to the ply interface could also be identified
from these micrographs. No transverse crack was found to occur prior to delamination.
Fig. 5: Micrograph showing the onset of delamination for the [902/30] s specimen
This is consistent with the analytical prediction of the stress to first-ply failure being greater
than the stress for the onset of delamination under these conditions.
Onset of Delamination Under Fatigue Loading
Since xz is responsible for the free-edge delamination, the inplane shear S-N relationship and
the calculated xz under a given load (including curing residual stress) were utilized to predict
the onset of delamination under fatigue loading. Fatigue failure was assumed to occur when
the fatigue interlaminar shear stress, xz, reached the fatigue shear strength of the material
system. Figure 6 shows the shear S-N relationship determined under tension-tension fatigue
from the [45]2s laminate.
II - 191
Shear stress, MPa
100
90
80
70
60
50
100
1000
10
10
Cycle to failure
10
10
Fig. 6: S-N relationship for shear and fatigue data of [902/30] S laminates in terms of the
calculated effective stress xz at the 30/-30 interface. Solid line: shear S-N relation from
[45]2s laminate; open circle: fatigue data from [45]2s laminate; and solid circle: fatigue
data from [902/30] s laminate
The open circles are the fatigue data obtained from experiment, and the solid line represents
the regression line which was fitted to Eqn 2 using the data reduction scheme described
previously. The material constants a and b in Eqn 2 determined from the regression analysis
are 135.7 and -0.065 , respectively. For a given interlaminar shear stress the number of
cycles for the onset of delamination can then be predicted from Eqn 2. The interlaminar
shear stress generated in the fatigue loading of any laminate can be calculated using the
global-local model. The fatigue data at the onset of delamination for the [902/30]s specimen
under compression-compression fatigue are plotted (solid circle) in Figure 6. The shear
stresses in the ordinate are obtained from the applied fatigue stress on the [902/30]s laminate
using the global-local model. The experimental data appear to fit the predicted S-N relation
reasonably well with a little higher trend. This higher trend may be partially attributed to the
difficulty in determining the exact number of cycles for the onset of delamination under cyclic
loading. It may also be due to the inability of the acoustic emission system to capture a very
small signal from the initiation of delamination. Further work is needed to refine the
technique.
CONCLUSIONS
An investigation was conducted to determine the initiation of free-edge delamination in
IM7/5250-4 graphite/epoxy laminates under static and fatigue loading. Predictions regarding
stress level and location (interface) for the onset of delamination agree very well with
experimental results under static loading. The approach used in this work appears to be
promising for prediction of onset of delamination under fatigue loading. In both static and
fatigue loading no transverse ply failure was observed prior to free-edge delamination for this
laminate. The absence of any damage prior to delamination might be helpful for prediction of
II - 192
the delamination. The onset of delamination in this laminate can be accurately determined
experimentally by monitoring strains in the axial and thickness directions and the use of
acoustic emission. Further study is required to obtain confidence in this approach using a
variety of laminates and different loading conditions.
REFERENCES
1. R.B. Pipes and N.J. Pagano, Interlaminar Stresses in Composite Laminates under Uniform
Axial Extension, J. Composite Materials (1970), v4, 538-548.
2. N. J. Pagano and R.B. Pipes, The Influence of Stacking Sequence on Laminate Strength,
J. Composite Materials (1971), v5, 50-57.
3. J.M. Whitney and C.E. Browning, Free-Edge Delamination of Tensile Coupons, J.
Composite Materials (1973), v6, 300-303.
4. N.J. Pagano, Stress Fields in Composite Laminates, Int. J. Solids & Struct. (1978), v14,
385-400.
5. N.J. Pagano and S.R. Soni, Global-Local Laminate Variation Model, Int. J. Solids &
Struct. (1983), v19, 207.
6. W. J. Park and R. Y. Kim, Statistical Analysis of Composite Fatigue Life, Progress in
Science and Engineering of Composites, T. Hayashi, et al., ICCM-IV, Tokyo, Japan,
October, (1982),v1, 709-716.
7. R.Y. Kim and S.R. Soni, Experimental and Analytical Studies on the Onset of
Delamination in Laminated Composites, J. Composite Materials (1984), v18, 70-79.
8. R. Y. Kim, F. Abrams, and M. Knight, Mechanical,Characterization of a Thick
Composite Laminate, Proc. of American Society for Composites, Technomic Publishing
(1988), 711-718.
ACKNOWLEDGMENT
This work was performed under U.S. Air Force Contract No. F33615-95-D-5029. The
authors wish to express their appreciation to Mr. Ron Esterline of the University of Dayton
Research Institute for his help in the preparation and testing of specimens.
II - 193
FATIGUE CRACK GROWTH BEHAVIOR OF A

RANDOM FIBER SHEET MOLDING COMPOUND
COMPOSITE
D. R. Atodaria1, S. K. Putatunda1 and P. K. Mallick2
1
Department of Materials Science and Engineering, Wayne State University,

Detroit, Michigan 48202, USA
2
Department of Mechanical Engineering, University of Michigan-Dearborn,
Dearborn, Michigan 48128, USA
SUMMARY: The purpose of this study is to investigate the fatigue crack growth behavior of
a random fiber sheet molding compound (SMC-R) composite and to examine the applicability
of a new fatigue crack growth rate model to this material. In this study, the fatigue crack
growth rate of SMC-R was measured using a compliance approach. The new fatigue crack
growth rate model is based on a power law equation and takes into account not only the effect
of the stress intensity factor range, but also the effect of crack growth at various stages of
loading using a weight average stress intensity factor. It was observed that this new model
can represent the fatigue crack growth rate of this SMC-R at three different load ratios in a
single unifying curve.
KEYWORDS: random fiber composites, sheet molding compounds, fatigue crack growth,
compliance, power law equation
INTRODUCTION
Randomly oriented short fiber reinforced sheet molding compound (SMC-R) composites are
finding increasing use in many structural applications, including automotive structures and
bridge structures. Many of these structures experience repeated loading in service and fatigue
failure must be considered as one of the critical failure modes in their design. Use of SMC in
these structures requires that a better understanding of the fatigue failure process and a
suitable fatigue life prediction technique be developed for this class of composite materials.
There is a now a significant number of publications on the fatigue behavior of continuous
fiber reinforced laminated composites; however, publications on the fatigue behavior of
random fiber composites are extremely limited. In an earlier work [1], we proposed a new
empirical model to characterize the fatigue crack growth behavior of random fiber
composites. In the present work, we have studied the fatigue crack growth rate of a random
fiber SMC composite at three different load ratios and extended the application of our model
to the fatigue life prediction of this material.
II - 194
PROPOSED MODEL
An understanding of fatigue behavior is important for any material that experiences repeated
loading during service. There are basically two approaches for generating information and
knowledge of the fatigue behavior of materials: (1) the S-N curve or the -N curve approach,
and (2) the fracture mechanics approach. The fracture mechanics approach assumes the
presence of inherent cracks or flaws in the material and attempts to predict its life using
empirical models of fatigue crack growth rate. The most commonly used Mode I fatigue
crack growth rate model is based on the Paris power law equation [2]
da
dN
= A( K )
(1)
where, da/dN = fatigue crack growth rate

= difference in the maximum and minimum Mode I stress intensity factors
K
A, m = curve-fitted power law constants for the material
Eqn. (1) has been used extensively for metals and to some extent, with composite materials as
well. Wang et. al. [3] employed Eqn. (1) to establish fatigue crack growth rate in an SMC.
They found the value of A and m to be 7.58x10-8 and 9.65, respectively. Similar attempts
have also been made to fit Eqn. (1) to fatigue crack growth rate of other random fiber
composites, such as injection molded thermoplastic composites [4].
The Paris model, given by Eqn. (1), assumes that the fatigue crack growth rate depends
only on the stress intensity factor range, _K. For a fatigue test with zero load ratio, K = Kmax ,
and thus, according to Eqn. (1), the fatigue crack growth rate depends only on Kmax. regardless
of the mean load imposed during cycling. Furthermore, for any given value of stress intensity
factor range , the fatigue crack growth rate is independent of other values of K within this
range. For many composites, it is observed that crack growth occurs at stress levels that are
lower than the maximum stress imposed in the fatigue cycle and, therefore, stress intensity
factors lower than the Kmax will also have some effect on the fatigue crack growth rate of such
composites. In each cycle, the crack may grow in a very slow and progressive manner as
the stress increases from the minimum to the maximum value; however, the extent of damage
growth may not be equal at each stress level. It is expected that the crack growth rate will be
slower at low stress levels and will be faster as the stress increases toward the maximum
value. Based on this assumption, we have proposed a modified power law model which takes
into account this progressive crack propagation process by including a weight average value
of K in addition to the stress intensity range that is already present in the Paris model.
According to this modified model, the fatigue crack growth rate can be expressed as
II - 195
In the present work, we have used the proposed model for a random fiber sheet molding
compound composite. However, unlike the carbon fiber reinforced PEEK, the fatigue
failure does not occur by the enlargement of a single visible crack. In the first few fatigue
tests performed with SMC specimens, it was observed that the fatigue crack in this material
does not follow a well defined path. As the fatigue loading is continued, a number of small
cracks appear in the vicinity of the original notch tip. These cracks are oriented at different
angles with the notch plane. With a traveling microscope focused on the specimen surfaces,
the opening and closing of these cracks can be clearly seen; however it was difficult to judge
whether these cracks are only at the surfaces or also extend deep into the thickness of the
specimen. Eventually one of these cracks extends in length, but stops after a slight extension.
With continued cycling, multiple small cracks appear in the vicinity of the extended crack and
the whole process of stop-and-go crack growth process is repeated. Another point to note is
that the crack extension occurs in SMC in a zigzag manner, often deviating from the plane of
the original notch. This created a problem in that the crack length could not be accurately
measured during the fatigue test. This has led us to use the increase in specimen compliance
as an indirect measure of the fatigue crack growth in SMC. The compliance measurement
technique is described in the next section.
II - 196
EXPERIMENTAL PROCEDURE
The material used in this study is a sheet molding compound (SMC) composite containing 25
mm long chopped E-glass fibers randomly oriented in a thermosetting polyester resin. The
fiber weight fraction is estimated to be 30 percent. In addition to the randomly oriented fibers
and the matrix, the other major ingredient in the material is a calcium carbonate filler which is
randomly dispersed in the matrix. The nominal thickness of compression molded SMC plate
was 3.6 mm.
The fatigue experiments were conducted in an MTS servohydraulic testing machine using a
18 kN load cell cartridge. Compact tension specimens with the dimensions shown in Fig. 2
were used in all fatigue tests. The crack opening displacement during the fatigue tests were
measured using a clip gauge, which was inserted at the notch mouth in the beginning of each
test. Fatigue cycling was performed in a load-controlled mode at room temperature using a
frequency of 2 Hz. Three different load ratios, Pmin/ Pmax. , were used, namely 0.1, 0.3 and 0.5.
Fig. 2: Compact tension specimen
In order to use the compliance measurement as an indirect measure of fatigue crack growth,
we had to first establish a compliance calibration equation which relates the specimen
compliance to an equivalent crack length. The calibration curve was obtained using 10
specimens with different initial notch lengths, varying in length from 12 to 20 mm. Each of
these specimens was cycled once between 0.67 kN and 1.33 kN at 0.1 Hz. The specimen
compliance, defined as the ratio of crack opening displacement (COD) to load, is determined
from the loading portion of the load vs. COD diagram obtained during this test.
The specimen compliance during the fatigue test was measured at regular intervals. Instead of
applying the full load in the first cycle, the load was increased in steps. After reaching the end
of each load step, the specimen was allowed to fatigue for a few cycles at 2 Hz. before
recording the hysteresis loop depicting the load vs. the crack opening displacement at 0.1 Hz.
The load was then increased to the next step. This process was continued until the full load
II - 197
was reached and the specimen was subsequently allowed to cycle between the maximum and
minimum loads for the duration of its life. The specimen compliance was calculated from the
digital data of load and crack opening displacement using the Excel software. A function
called linest in the Excel software was used in to calculating the slope of the loading portion
of the hysteresis loop.
RESULTS
The compliance calibration diagram for the SMC material considered is shown in Fig. 3.
Following equation was obtained relating the compliance to the crack length :
C = 5.903 x 10-7 a 1.276
(3)
The stress intensity factor for the compact tension specimen was calculated using the
following equation [5]:
K =
P
B
a
(W )
where, P = applied load on the specimen, B = specimen thickness,

W = width of the specimen from the center of the loading point,
a = crack length measured from the center of the loading point, and
for a compact tension specimen,
II - 198
(4)
Fig. 4 shows the fatigue crack growth rate vs. K for three different load ratios considered in
this study. This figure shows that the fatigue crack growth rate of the SMC increases with
increasing load ratio. If the Paris equation is used to model the fatigue crack growth rate,
there will be different constants for different load ratios. Similar observation can be made if
the fatigue crack growth rate is plotted as a function of Kaverage (Fig. 5); however, the
distinction between the load ratios is much narrower in this figure. Next, the effects of K and
Kaverage are combined using different values of and for the powers of these two parameters,
and as shown in Fig. 6, the fatigue crack growth data for all three load ratios fall in a very
narrow band for = 0.15 and = 0.85. This indicates that Eqn. (2) can be used to unify the
load ratio effect in the SMC. The best fit line drawn through the data points for all three load
ratios has the equation given by
da
dN
= 2. 06x10
20
( K
11.2
0.85
Kaverage )
0.15
(5)
The experimental fatigue crack growth rate for each load ratio was compared with the fatigue
crack growth rate predicted by Eqn. (5). One such comparison is shown in Fig. 7 for load
ratio R=0.1. As can be seen in Fig. 7, there is a close agreement between the experimentally
II - 199
determined values and predicted values. Similar agreement was also observed for the other
two load ratios.
II - 200
CONCLUSIONS
1.
The compliance measurement approach was shown to be an effective method for

monitoring the fatigue crack growth in random fiber SMC composites.
2.
The proposed model is capable of unifying the fatigue crack growth rate at three load
ratios considered into a single power law equation:
da
dN
3.
= 2 . 06 x 10
20
( K
0 . 15
0. 85
K average )
11. 2
Good agreement was found between predicted fatigue crack growth rate using the
proposed model and the experimentally measured fatigue crack growth rate in a random
fiber SMC composite.
REFERENCES
1. Atodaria, D.R., Putatunda, S.K. and Mallick, P.K., A Fatigue Crack Growth Model for
Random Fiber Composites, Journal of Composite Materials, Vol. 31, No.8, 1997 (in
print).
2
Paris, P.C. and Erdogan, F., A Critical Analysis of Crack Propagation Laws, Trans.
ASME Journal of Basic Engineering, Vol. D85, 1963, pp. 528-534.
II - 201
3. Wang, S.S., Chim, E.S.-M., and Zahlan, N.M., Fatigue Crack Propagation in Random
Short-Fiber Composite, Journal of Composite Materials, Vol. 17, 1983, pp. 251-266.
4. Karger-Kocsis, J. and Friedrich. K., Fatigue Crack Propagation in Short and Long FibreReinforced Injection Moulded PA 6.6 Composites, Composites, Vol. 19, No. 2, 1988,
pp. 105-114.
5. Hertzberg, R.W., Deformation and Fracture Mechanics of Engineering Materials, 4th
Edition, John Wiley and Sons, Inc., New York, 1996, pp. 757.
II - 202
DAMAGE BASED FATIGUE LIFE MODELING FOR

BRITTLE COMPOSITES
G. M. Newaz1 and N. Bonora2
1
Mechanical Engineering Department,Wayne State University,

505 Anthony Wayne Drive, Detroit, MI 48202, USA
2
Industrial Engineering Department,University of Cassino,
Via G. De Biasio 43, 03043 Cassino (FR), Italy
SUMMARY: Fiber reinforced composites are usually notch insensitive. Damage in

composite, under static loading is usually due to serveral concurrent fracture modes as fiber
failure, matrix cracking, fiber-matrix debonding, fibre pull-out and delamination. These
mechanisms reduce the material load carrying capability. In fatigue this situation can be even
more complicated. In brittle composite systems where all constituents have brittle behavior
with no dissipation processes other than the creation of new fracture surfaces - they exhibit a
strong connection between the first damaging fatigue cycle and the subsequent damage
evolution. In addition, damage produced during the first cycle can be directly related to the
one produced due to monotonic loading. In the present paper a fatigue damage model for
brittle composites is presented. The effect of complex damage state was examined in terms of
stiffness loss as a function of the accumulated strain during cycling. The model, developed on
phenomenological basis, has been verified on several composite systems such as
unidirectional ceramic matrix and cross-ply polymeric composites.
KEYWORDS: fatigue life model, brittle composite, stiffness loss, damage
INTRODUCTION
Prediction of fatigue life in brittle composites where only microcracks dominate the overall
dissipation characteristics remains an important challenge as growth laws based on fracture
mechanics for single cracks are inadequate. Due to the complex damage modes that occur in
a composite material under static or cycling loading, it is appropriate to address the complex
damage states where several basic fracture types such as matrix cracks, fibre breaks,
delamination, and fibre-matrix disbonds and their mutual interactions are accounted for,
particularly, in developing life models. The best way to investigate fatigue effects on
composite materials is to assess the degradation of the material properties such as elastic
moduli as a function of the number of cycles [1-3]. A direct consequence of distributed
microcrack-type damage is tensorial stiffness property change in composites.
There is an important connection between the first cycle damage and subsequent damage
evolution in fatigue in laminated composites. The damage that occurs in the first cycle is no
different than that which occurs during monotonic response. In order to clearly understand
the implication of this initial damage, it is important to develop the link between monotonic
damage and fatigue damage to develop proper fatigue life models. Monotonic response of
unidirectional CMCs have been studied extensively by a number of researchers [4-12].
II - 203
Proportional limit of unidirectional CMCs have been associated with matrix cracking and
subsequent nonlinear behavior has been identified to be due to progressive fiber failure [9,10].
Matrix cracking in CMCs has been the subject of in-depth investigation by researchers as well
[12,13]. Continuum damage mechanics models can be used to predict the stiffness loss in
these composites due to the presence of damage which are typically microcrack entities such
as matrix cracks, fiber matrix disbonds and fiber cracks [3]. Monotonic damage evolution and
subsequent modeling of the stiffness changes using continuum approaches were investigated
for the hybrid composite [14]. Cyclic damage has been characterized by a number of
researchers for glass matrix composites which confirm that a combination of damage such as
matrix cracking, fiber-matrix debonding and fiber fracture depending on the lay-up of the
composite [15]. Polymeric composites (PMCs) brittle matrices such as epoxy also can be
categorized as brittle composites because of the nature of dissipation characteristics due to
distributed matrix cracks. Other PMCs also show brittle characteristics as plastic deformation
is suppressed at the microstructural level as much of the load is carried by carbon fibers.
Microcracking in toughened epoxies and in thermoplastic matrices are also dominant
dissipation mechanisms precluding plastic deformation that occurs readily in bulk matrices.
As far as distributed microcracking is concerned in PMCs during fatigue, they can be
categorized as brittle where energy dissipation is primarily controlled by free surface creation.
MODELING ASPECTS
Fig. 1 shows the data scatter for several fatigue tests performed with ceramic matrix
composites [16] at room temperature with the same R-ratio and frequency and max . In order
to arrange experimental data in a more meaningful way, the number of cycles has been
normalized by Nf. This representation allows one to separate the scatter relative to Nf and to
underline the main feature of the fatigue damage evolution law. In this case, it is possible to
identify a lower bound for the degradation of the moduli.
The fact that brittle composites are very sensitive, in terms of accumulated damage, to the
very first cycle is indicative of a high sensitivity to the applied strain. It has been seen that it is
more appropriate to approach composites in terms of strain instead of stress, since due to the
non-homogeneity of the micro-structure, strains have to be the same while stresses are
different in matrix and in the fiber.
If the strains applied in the first few cycles are high enough to introduce damage in the
microstructure, this will lead to stiffness loss. The subsequent cycles will see a progressive
evolution of the initiated damage state that will develop more gradually up to the moment of
final fracture. It is very difficult to develop an accurate model, defined on micro-mechanical
basis that is able to predict the physical evolution of complex damage states. The main
features can be obtained from the fatigue tests in terms of the shape of the stiffness reduction
trend. Many authors pointed out that brittle materials such as glass or glass-ceramic materials
show practically no range of fatigue behavior that results in a crack of the order of an existing
flaw size growing instantaneously. A number of authors [15,17] suggested that fatigue cycling
does not lead to new damage modes in the composites.
If fatigue cycling does not introduce additional damage modes other than the growth of the
already existing complex damage state, it follows that the evolution features related to the
fatigue process are all contained in the shape of the distribution of the fatigue data.
Furthermore, cyclic damage measurements taken at different cycles should fall on the same
II - 204
static damage curve shown in Fig.2. The last observation has been verified by plotting
damage measurements taken at different number of cycles versus the total strain applied
during the cycling. When a specimen is cycled in stress control, due to creation of new
surfaces, the true applied strain increases with the number of cycles as shown in Fig. 3. The
progressive material damage is illustrated in Fig. 4. The applied strain is partially recovered
elastically in the cycle but there is a contribution that accumulates as ratchetting, due to
incomplete closure of cracked surfaces during unloading.
The relation that describes the damage evolution via modulus reduction in the applied loading
direction, ED /E0, as a function of the number of cycles, have to be bounded between 0 and 1
for a number of cycles comprised between 1 and Nf. The following relation :
1
ln( N )
ED
= B ( ) 1
E0
ln( N f )
(1)
seem to be appropriate to describe the experimental data trend. When the number of cycle
reaches Nf, i.e. the number of cylces at fracture, the ratio E D /E0 goes to zero. E with
subscripts D and 0 represent current and initial stiffnesses, respectively. Note that D= 1ED/E0 . When N=1, then the amount of damage at the first cycle is determined by the imposed
strain field in the material. In fact only if the strain level is high enough to produce damage in
the first place, then further damage will develop during cycling.
To determine the amplitude constant B, it is sufficient to observe that the first cycle, if the
strain rate is not extremely high, is similar to a static load ramp up to the imposed maximum
strain. For this reason, the amount of damage produced in the first cycle have to be related to
the static damage curve as suggested by the authors [14]:
1
ln( / th ) m
B ( ) = (1 D0 ) ( Dcr D0 )1
ln( f / th )
(2)
that substituted in Eqn. 1 leads to:

ED
E0
((1 - D0 ) - ( Dcr - D0 )
ln e / eth
ln e f . / eth
1
m
) [ 1-
ln N
ln N f
(3)
In addition to CMCs, the model can be applied to PMCs as the fatigue damage process in
these materials is dominated by matrix cracking which manifests as stiffness loss as a function
of fatigue cycles. As in the case of CMCs, the damage in PMCs in terms of matrix cracks
and/or fiber cracks are the energy dissipation events that are completely controlled by creation
of new surfaces only. Consequently, based on damge processes at work in fatigue, both types
of composites exhibit brittle behavior. Working with data from literature [18-19] and
analyzing them using the approach discussed earlier, the proposed fatigue damage model can
predict the trend in damage evolution as shown in Figures 5-6.
II - 205
CONCLUSIONS
Damage in CMCs as well as PMCs are strain controlled. More particularly, magnitude of
damage as manifested by stiffness loss is dependent on the strain experienced by the
composite as described by Eqn. 3. The fatigue life model presented requires the evaluation of
a number of parameters such as th , cr and m which can be obtained through a static test.
Only two parameters, exponent and the number of cycle at fracture Nf, have to be
determined using fatigue data. Anyway, only few tests performed at two or three different
stress levels will be sufficient to identify the parameters in a proper way. The nature of
damage evolution as dictated by stiffness loss during fatigue can be cast in a functional form
that captures the essence of strain controlled damage. Fatigue life for a given reduction in
stiffness can be estimated from the functional form if nominal final life is known from limited
experimental data. This approach seems to work well for both CMCs and PMCs. Hence, this
basic life prediction model for fatigue that captures the essence of the trend in fatigue data in
brittle composites shows significant promise.
ACKNOWLEDGEMENT
The authors acknowledge the funding for this research work from the IHPTET program
support from the Propulsion and Materials Laboratory of Wright Laboratories. Mr. Larry
Zawada of the Materials Directorate was the program monitor.
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
Highsmith, A. L. and Reifsnider, K. L., "Stiffness-Reduction Mechanisms in Composite

Laminates," in Damage in Composite Materials, ASTM STP 775, American Society for
Testing and Materials, Philadelphia, PA, 1982, pp.103.
Laws, N., Dvorak, G.J., and Hejazi, M., "Stiffness Changes in Unidirectional
Composites Caused by Crack Systems, " J. Mechanics of Materials, vol.2, 1983, p.123
Talreja, R. , Fatigue of Composite Materials, Technical University of Denmark, 1985.
Prewo, K.M., "Tension and Flexural Strength of Silicon Carbide Fiber-reinforced Glass
Ceramics", Journal of Materials Science, Vol. 21, 1986, p. 3590.
Prewo, K. M., "Advanced Characterization of SiC Fiber reinforced Glass Matrix
Composites", Office of Naval Research Contract N00014-81-C-0571, Interim report,
June, 1983.
Zawada, L.P., Butkus, L.M. and Hartman, G.A., "Room Temperature Tensile and
Fatigue properties of SiC Fiber-Reinforced Aluminosilicate Glass", Ceramic
Engineering Science and Proceedings, Vol. 11, 1990, p. 1592.
Nardone, V.C. and Prewo, K.M., "Tensile Performance of Carbon Fiber-Reinforced
Glass", Journal of Materials Science, Vol. 23, !988, p. 168.
Newaz, G.M., Brust, F.W., Jarmon, D.C. and Cairo, R.R., "Deformation, Failure and
Mechnical Response Modeling in Dual Fiber CMC", Presented at the Annual ACS
Meeting, Cocoa beach, Florida, January, 1994.
Rousseau, C.Q., "Monotonic and Cyclic Behavior of a SiC/CAS Composite", ASTM
STP 1080, J.M. Kennedy, H.M. Mouller, and W.S. Johnson Eds., 1990, p. 136.
II - 206
10.
11.
12.
13.
14
15.
16.
17.
18.
19.
Prewo, K.M., Johnson,B.and Starett, S., "Silicon Carbide Fiber-Reinforced GlassCeramics", Journal of Materials Science, Vol.24, 1989, p.1373.
Sbaizero, O. and Evans, A.G., "Tensile and Shear properties of Laminated Ceramic
Matrix Composites", Journal of American Ceramic Society, Vol. 69, # 6, 1986, p. 481.
Aveston, and J.Kelly, "Theory of Multiple Fracture of Fibrous Composites", Journal of
Materials Science, 1973, Vol.8, p. 352.
Budiansky, B., Hutchinson, J.W., and Evans, A.G., "Matrix Fracture in FiberReinforced Ceramics", Journal of Mechanics and Physics of Solids, Vol. 34. p. 35.
Bonora, N. and Newaz, G., "Modeling Damage Evolution in a Hybrid Ceramic Matrix
Composite Under Static Tensile Loading", accepted for publication in J. Engineering
Materials and Technology, ASME, Nov., 1996.
Karandikar, P.G. and Chou, T.W., "Microcracking and Elastic Moduli Reductions in
Unidirectional Nicalon-CAS Composites Under Cyclic Fatigue Loading", Proceeding of
the 16th Annual Conference on Composites and Advanced Ceramics, Cocoa Beach,
Florida, Jan., 1992.
Newaz, G.M. and Bonora, N., Fatigue Life Modeling in Hybrid CMCs, in Thermal
and Mechanical Test methods and Behavior of Continuous Fiber Ceramic Composites,
ASTM STP 1309, M.G.Jenkins, S.T. Gonczy, E. Lara-Curizio, N.E Ashbaughand L.P.
Zawada, Eds., American Society for Testing and Materials, 1996.
Evans, A.G. and Fuller, E.R., "Crack Propagation in Ceramic Materials under Cyclic
Loading Conditions", Metall. Trans. 5, (1974), p.27.
Charewicz A. and I.M. Daniel, Damage Mechanisms and Accumulation in
Graphite/Epoxy Laminates, in Composite Material:Fatigue and Fracture, ASTM STP
907, H.T.Hahn Ed., 1986, pp. 274-297.
Bonora, N., Marchetti, M. and Millela, P.P., Theoretical Forecasting and Experimental
Validation of Damage Tolerance and Accumulation in Glass/Epoxy laminates, J. of
reinforced Plastics and Composites, Vol II, 1992, pp. 56-81.
II - 207
II - 208
II - 209
II - 210
FATIGUE DAMAGE IN NICALON SIC(F)/AL

COMPOSITE WIRES
Yue Zhuo, Hong Wan, Jin Pan and Deming Yang
State Key Laboratory for Fatigue and Fracture of Materials
Department. of Materials Engineering and Applied Chemistry
National University of Defense Technology
Changsha, Hunan, 410073, China .
SUMMARY: To understand the fatigue damage mechanism in Nicalon SiC(f)/Al composite,

the wires of SiC(f)/Al were fabricated by ultrasonic liquid infiltration method. Tensile -tensile fatigue test on the composite wires was carried out at stress ratio R=0.1 at room
temperature and at 573 K . Tests show that the dominant damage model in the wire is
progressive fiber fracture and the progressive fiber fracture is due to the deorientation of fiber
to the stress axis. Test at 573 K shows no evident differences in the fatigue property between
room temperature and 573 K . The wires after exposure at 873 K for 30 minutes, through the
interface bonding force has been increased by interface raction, are of lower fatigue property
than those unexposured.
KEY WORDS: metal-matrix composite, silicon carbide fiber, fatigue damage
INTRODUCTION
Despite its evident importance in numerous engineering application, few published
investigation exist on the fatigue properties and damage of SiC multifilaments reinforced
Aluminum composite. The specimens used in this paper are composite wires with the fibers
uniaxially oriented in the loading direction. Its fatigue failure mode is different from the
fatigue crack growth mode in short fibers reinforced aluminum composite[1]. The fatigue
failure mode in continuous fiber reinforced metal matrix composites (MMC) is controlled by
the following three constituents of the system: fiber, matrix, and fiber/matrix interface.
W.S.Jonson basing on the relative strain to fatigue failure grouped possible failure modes of
MMC into three categories[2]--- matrix dominated, fiber dominated and self-similar damage
growth. The purpose of this paper is to find the failure mode and damage process in this
composite wires. The experiment results also provide a method to improve the fatigue
properties of SiC fiber/Al composite.
EXPERIMENTAL MATERIALS & METHODS

The composite wire used here is reinforced by a bind of 500 Nicalon SiC fibers produced by
NIPPON Carbon Co. Ltd. The matrix is industrial pure Al 1100 infiltrated by ultrasonic liquid
infiltration method. Total wire length of a batch is 200 to 500 meters and the fibers are
oriented in the length direction. Specimen cut from the same batch of wire are 50 mm and
have the same mechanical properties. The main properties of the fiber and composite wire are
listed in table 1. Tension-tension fatigue tests were conducted in load control on a 1 x 103N
II - 211
EHF FB01-4LA material test machine with a stress ratio R=0.1. A sinusoidal waveform with
a frequency of 20 Hz was used. The fatigue fractograph is observed by a S-360 scanning
electron microscope.
Table 1: The main properties of the fiber and composite wire
materials diameter density (g/cm3)
SiCf
2.55
10-15 m
wire
0.52 mm
2.60
UTS (MPa)
2500-3000
1100-1200
E (GPa)
180-200
110-120
Vf %
40-45
Load/elongation curve was detected by a punctuation extensometer and recorded by means of

an X-Y recorder. If the specimens fractured near the grips its date would not be count in.
Specimens tested at 573 K were heated by a tube stove and the gips were not in the stove.
Fig. 1: S-N curve of NICALON SiCf /Al composite wire

---- at room temperature
---- at 573 K

Figure 1 is the S--N curve of the composite wires. From the curve it can be known that the
fatigue stress to failure is decreased quickly with the increase of cycle numbers. When the
cycle index is 106 its fatigue strength is 440 MPa, less than one half of its static tensile
strength. For fibers in the wire is ceramic material and its fatigue stress to failure generally
has no relationship with the numbers of stress cycle. The fatigue strength of the Al matrix is
more than 90% of its static tension stress while the fatigue cycle number is 106 [3]. Why does
the fatigue stress of the composite wire decrease so rapidly?
To find the reason, a fatigue test at 573 K was taken on the wires. At this temperature the
mechanical properties of fiber can not be influenced, but the properties of Al matrix can be
changed. Test result is shown in fig. 1. The composite wire at 573 K is of the same fatigue
property with its room temperature. Treated the wire at 873 K for 30 min. than test its fatigue
property at Smax=700 MPa at room temperature. The cycle index to failue is 5 x 1025 x 103
and much lower than that of untreated specimen. Nicalon SiC fibers can react with Al at 873-913 K and form Al2O3 and Al4C3. This interface reaction can increase the bonding force
between fibers and matrix and can decrease the mechanical properties of the fibers in the
II - 212
specimens. So the fatigue damage in this composite wire must be mainly dominated by the
fibers.
Fig. 2: The relation of static tensile strength with fatigue cycle number
x---- Smax=700 MPa
o ---- Smax= 450 MPa
Fig. 3: The photomicrograph of the composite wire x 100
Fig 4: The SEM fatigue fractograph of the composite wire

To find the fatigue damage process in the wire, a specimen after 106 cycles on Smax=450 MPa
and a specimen fractured by static tensile test have been extracted by NaOH solution. The two
specimens are of the same length (50 mm) before extracted. The fiber in the fatigue specimen
become short about 5 mm, but the fiber in static test specimen is about 50 mm same with the
length of fibers in untested specimen. To confirm this fiber frcture process during fatigue test,
several specimens stressed on Smax=700MPa and Smax=450MPa were tested to different cycle
II - 213
numbers and then tested their static tensile strength. The relation between cycle numbers and
static tensile strength are shown in Fig. 2. The tensile strength is decline with the increase of
fatigue cycle index. So the fibers fracture must be progressive as the fatigue test going. This
fiber length after fatigue test at certain conditions can also give us some information about the
bonding force between fiber and matrix.
What made the fibers in fatigue specimen fractured one by one? For the composite wire was
fabricated by ultrasonic liquid infiltration method and exists the deorientation in the length
direction, as shown in photomicrograph (Fig. 3). The deorientation resulted in the stress
difference among the fibers during fatigue test. The Al matrix displayed a good special fatigue
ductility and cooperated with the rotation of fibers while one fiber of hard orientation
fractured after another. A SEM fatigue fractograph as Fig. 4 shows that the Al matrix in a
specimen after 106 cycles is still appeared as a ductility fractograph characteristic. Electric
conductivity of the fatigue specimen before and after fatigue test is unchanged too. The
specimen elongation during the fatigue process was detected and the result is shown on Fig. 5.
This specimen elongation may be caused by improving its fiber deorientation and that is the
reason of the fiber fractured one by one.
x10-4
Fig. 5: The specimen elongation during the fatigue test

CONCLUSION
1. The progressive fiber fracture is the dominant damage model in this composite wire.
2. The deorientation of the fibers to the stress axis resulted in the fibers fracture one by
one and this fatigue process is with the cooperation of matrix fatigue ductility.
ACKNOWLEDGMENTS
The authors would like to thank to Professor Wang Zhong Guang and Mr. Gong Wen Ku ,
Yao Ge for useful discussions and assistance with the fatigue test.
REFERENCES
1.
Xi Cong Liu and C. Bathias, Fatigue damage development in Al2O3/Al

composite, Composites, Vol.24, No. 3(1993)pp282-287
2.
W.S.Johnson, Fatigue damage accumulation in various metal matrix composites,

NASA Technical Memorandum 89116, N87-18416, 1987
3.
G.J.Dvorak and W.S.Johnson, International Journal of Fracture 16(1980)585-607
II - 214
THE MODE-I FRACTURE TOUGHNESS OF

IM7/LARC-RP46 COMPOSITES AT HIGH
TEMPERATURES
W.X. Wang1. F. G. Yuan2, B. D. Potter2,
R. H. Pater3 and Y. Takao1
1
Research Institute for Applied Mechanics,

Kyushu University, Kasuga, Fukuoka 816, Japan
2
Department of Mechanical and Aerospace Engineering,
North Carolina State University, Raleigh, NC 27695, USA
3
NASA Langley Research Center, Hampton, VA 23681, USA
KEYWORDS: fracture toughness, high temperature, mode I, energy release rate,

viscoelasticity and viscoplasticity, loading rate, crack growth rate
II - 215
II - 216
II - 217
II - 218
II - 219
II - 220
II - 221
II - 222
II - 223
II - 224
INTRALAMINAR FRACTURE BEHAVIOUR AND

MICROSTRUCTURAL CHARACTERISTICS OF
UNIDIRECTIONAL T300/914
Adrian Lowe and Shankar Kalyanasundaram
Department of Engineering, Australian National University, Australia
KEYWORDS: shear, failure mechanisms, carbon/epoxy, fractography, phase separation
II - 225
II - 226
II - 227
II - 228
II - 229
II - 230
II - 231
II - 232
II - 233
RATE EFFECTS ON MODE I INTERLAMINAR

FRACTURE TOUGHNESS IN CARBON-FIBRE/EPOXY
AND CARBON-FIBRE/TOUGHENED-EPOXY
COMPOSITE LAMINATES
Takayuki Kusaka 1 , Masaki Hojo 2 , Tomoaki Kurokawa 3 and Shojiro Ochiai 2
1
Hyogo Prefectural Institute of Industrial Research,

3-1-12, Yukihira-cho, Suma-ku, Kobe 654, Japan
2
Mesoscopic Material Research Centre, Kyoto University,
Yoshidahonmachi, Sakyo-ku, Kyoto 606-01, Japan
3
Department of Mechanical Engineering, Setsunan University,
17-8, Ikedanakamachi, Neyagawa 572, Japan
KEYWORDS: delamination, mode I, fracture touchness, rate effect, impact strength, crack
propagation, unstable fracture, stick-slip phenomenon
II - 234
II - 235
II - 236
II - 237
II - 238
II - 239
II - 240
II - 241
II - 242
II - 243
TOUGHNESS OF MARINE COMPOSITES:

DELAMINATION RESISTANCE UNDER STATIC
LOADING
P. Compston 1 , P.Y.B. Jar 1 , P.J. Burchill 2 and G. Simpson 2
1
Department of Engineering, The Australian National University,

Canberra, ACT 0200, Australia
2
Defence Science and Technology Organisation, Aeronautical and Maritime Research
Laboratory, G.P.O. Box 4331, Melbourne, Victoria 3001, Australia
KEYWORDS: delamination resistance, marine, glass fibre, woven roving, unidirectional,

vinyl ester.
II - 244
II - 245
II - 246
II - 247
II - 248
II - 249
II - 250
II - 251
II - 252
II - 253
IMPROVED INTERLAMINAR FRACTURE

TOUGHNESS FOR VINYL ESTER RESIN - FIBRE
GLASS COMPOSITES
Peter J. Burchill and Gary J. Simpson
Ship Structures and Materials Division,
Aeronautical and Maritime Research Laboratory, DSTO,
PO Box 4331, Melbourne, Victoria 3001, Australia.
SUMMARY: Toughening the matrix using modified acrylonitrile - butadiene copolymers

improves the delamination resistance of vinyl ester resin - glass woven roving composites.
Analysis of the largest (or smallest) measurements of toughness by extreme value theory
shows that the improvement to the matrix resin is fully to transferred to the composite. The
glass content of the composite does not appear to determine the delamination resistance for
contents of 38 - 70 vol%. For the composites used in this study the contribution to
interlaminar fracture toughness by the reinforcement ranges from 0 to 1400 Jm-2.
KEYWORDS: delamination resistance, extreme value theory, vinyl ester resins, acrylonitrile
/ butadiene copolymers, woven glass roving reinforcement
INTRODUCTION
Fracture in cloth - resin composites occurs most easily parallel to the reinforcement between
the layers of cloth resulting in delamination of the composite and a large reduction in
compressive strength. Generally tensile loading produces this delamination, and the cracks
grow at the fibre-matrix interface or through failure of the matrix resin, or a combination of
both. Improvement of the fracture toughness of composite structures will require the
suppression of this crack growth. Most studies of fracture in composites have looked at
unidirectional fibre - epoxy resin based materials [1], but vinyl ester resin composites are just
as prone to delamination which is governed by the brittle nature of the matrix and the fibre matrix bond. From previous studies, two energy absorbing mechanisms can be identified
contributing to the delamination resistance; energy is absorbed by deformation of the resin or
by the effects of fibres bridging the crack [2-6]. Bridging fibres absorb energy through their
being pull out of the matrix and through fracture. Another mechanism by which the fracture
energy may be increased is by crack deflection which may arise due to imperfections in the
composite or through local changes in fibre orientation, that will be experienced with cloth based composites.
Vinyl ester resins (VE) are typically a mixture of methacrylate esters produced from epoxy
resins diluted with other monomers such as styrene and are generally cured at room
temperature to give materials with a much higher resistance than polyester resins to moisture
and corrosive chemicals. A commonplace VE resin is made from the diglycidylether of
bisphenol A and is the work horse of the vinyl ester resin market. A feature which gives these
II - 254
resins a strong advantage over epoxy resins apart from price is that they can be formulated to
provide almost any desired pot-life. Studies of the properties of bulk castings show them to
be brittle, however recent work on toughening has shown that large improvements in
toughness can be gained by the addition of suitable reactive liquid polymers (RLP) which
yield a separated elastomeric phase upon cure [7,8]. This study will examine how well the
improvement in resin bulk properties transfer to VE resin / glass fibre composites.
The hand lay-up and vacuum bagging method of composite fabrication was suspected of not
being very reproducible in terms of glass to resin ratio, in addition, toughened resins have a
higher viscosity which will affect resin wet-out and flow in the preparation of the composite.
Hence, no attempt was made to control the glass content, but rather allow deliberate variation
in the process to give a glass to resin ratio which was as wide as practicable between
composites. In this way, a more confident comparison was expected of the influence different
resins have on composite toughness. This approach was hoped would also provide data on
the fibre bridging and weave effects of the reinforcement which were suspected of being
influenced by resin content.
EXPERIMENTAL
Materials
Two vinyl ester resins used were: Derakane 411 (Der 411) a standard bisphenol A epoxy
based vinyl ester resin, and Derakane 8084 (Der 8084) a toughened version of Der 411
internally modified with a reactive acrylonitrile - butadiene copolymer during manufacture.
The additives were: a vinyl terminated acrylonitrile - butadiene copolymer (BF Goodrich
VTBN X33) and a BF Goodrich carboxy terminated acrylonitrile - butadiene copolymer prereacted with an aliphatic oligomeric epoxide (S22) [9]. The reinforcement was a woven
roving - AR106, Colan Products Pty. Ltd., nominal areal density 630 g/m2 - with a vinyl ester
resin compatible finish. The structure of the cloth was warp 29.5 yarns per 10cm, 1200 tex.,
and weft 15.8 yarns per 10cm, 1800 tex. All other materials were commercial chemicals used
as received.
Resin Cure and Composite Fabrication
The resins and their blends with the additives were cured at room temperature using the
catalyst system cobalt octoate (0.3 pph, of a 41% wt/vol solution in white spirit) and
methylethylketone peroxide (1.5 pph, of a 40% dimethyl phthalate solution), and the gel time
was adjusted with dimethylaniline (accelerator) and 2,4-pentanedione (retarder). Curing of
all resin systems was completed overnight at room temperature then for 90 minutes at 90oC.
The test methods used for evaluating bulk properties were, Heat Distortion Temperature
(ASTM D648), T-Peel (ASTM D1876) and fracture toughness from double torsion geometry
[10].
Various techniques were used to alter the glass : resin ratio in the hand lay-up/vacuum
bagging method, involving timed application of vacuum, and pot formulations to change gel
times. The composite panels were assembled with 16 plies of cloth (300mm x 500mm) cut
with the warp direction measuring 300mm. A thin Teflon film was placed as a de-bond zone
between the 8th and 9th plies to give an initial crack of about 50mm in the test specimens.
II - 255
Fracture Resistance
Double Cantilever Beam (DCB) specimens (25 x 190mm) with the long axis parallel to the
warp direction were cut from the composite panels with the de-bond layer at one end. To load
the specimen, tabs were attached to the de-bond end; and one side of the specimen was
painted with whitener to aid observation of the crack and marking its length. The specimen
was carefully put under a tensile load to open the de-bonded zone and pre-cracked to about 52
mm from the beginning. The specimen was then unloaded to its original position, and the
Instron Testing Machine (Model 4502) set at a crosshead speed of 2 mm/min. When a predetermined crosshead displacement was reached, the specimen was unloaded by about 5% of
the displacement before marking the crack length. The first measurement was generally
between 60 to 75 mm from the open end of the beam. The corresponding load and
displacement before the unloading were recorded automatically by the computer. Further
measurements were taken until 12 measurements (excluding the initial crack length) were
recorded. After the test, the specimen was unloaded and crack lengths were measured using a
travelling microscope with all measurements taken from the tab-end.
Load (P), displacement (d) and crack length (a) data were analysed by two methods, one due
to Hashemi, Kinloch and Williams [11] and another due to Rosensaft [12]. Both methods of
data reduction gave essentially the same values for fracture toughness and the values reported
here are from analysis using equations given by Hashemi et al.
Compliance (C = P/d) is related to specimen dimensions and material properties by:
C = 8N .
( a + h )
Bh 3 E11
(1)
N is a correction factor which accounts for the stiffening effects of the attachments required to
load the specimen and corrections due to large displacements and attachment tilting, other
symbols are: thickness 2h, width B, modulus E11. The term h, adjusts the measured crack
length for compliance of the un-cracked part of the specimen [11,13,14].
Strain energy release rate (GI) is calculated from:

G I = ( F N ).3 P 2 B(a + h)
(2)
F is a factor which corrects for the shortening of the crack length due to large displacements.
Equations for F and N can be found in references [11,14].
II - 256

Matrix Resins
The ability of the rubber additives to improve adhesion and crack resistance of the vinyl ester
resin Der 8084 is detailed in Table 1. A marked improvement in peel strength is achieved by
the addition of either rubber, and similar improvements are seen in the fracture energy of the
castings. The lowering of HDT due to the rubber should also be noted, and is an unfortunate
consequence of rubber toughening of any type of resin
Table 1: Properties of the resin systems
Resin
Additive
411
8084
8084
8084
NONE
NONE
6.86% X33
6.86% S22
Fracture
Energy
kJ/m2
0.13
0.4
1.40
2.40
Peel Strength
N/mm
HDT
0
C
0.5
0.9
2.31
3.01
101
80
73
70
Composite Properties
A minimum of five specimens was tested from each composite panel, and from each
specimen generally twelve measurements of toughness were obtained as the crack propagated
between the plies. Load deflection curves for these composites showed that. crack growth
generally progressed by stick / slip with some continuous slow crack growth regions which
were greater in extent with toughened matrix resins.
Toughness kJm
-2
2 .5
1 .5
1
0
10
20
30
40
50
60
O rd e r
Fig. 1: Toughness measurements on Der 8084 + X33 composite specimens
II - 257
Despite the uniformity with which a particular composite was made, it was apparent early in
the project that the specimens were not the same, and there was a wide variation in calculated
toughness values in one specimen associated with crack initiation and arrest. Fig. 1 gives the
toughness values from the five specimens of a Der 8084 + X33 composite. The values for
each specimen have been placed in serial order and is not the order in which they were
determined, toughness measurements did not depend on position along the specimen.
An analysis of variance showed that the specimens were not identical, and a similar result was
found for all the composite panels. Nevertheless, if the compliance measurements for each
specimen are normalised for variation in thickness and width, and plotted as the cube root
against crack length, all the values fall on a common line as seen in figure 2, and as expected
if equation 1 was obeyed. The slope of this line is (8/E11)1/3.
-1 1/3
Normalized Compliance ((mm N ) )
15
12
6
3
0
-50.00
0.00
50.00
100.00
150.00
200.00
-3
Crack Length (mm)
Fig. 2: Compliance data for all specimens tested from a Der 8084 + S22 composite
This agreement in modulus for the specimens from one panel indicates that the variation in
toughness is not due to poor fabrication of the panel. The average toughness determined from
a composite specimen was always greater than the toughness of the neat resin.
Figure 3 gives a plot for each resin type of the observed average specimen toughness against
specimen thickness. Thickness is a function of the glass content which for these composites
varied between 38 and 70 volume %, and each line is constructed from measurements on at
least four composite panels.
II - 258
3.5
Toughness kJ m
-2
3
2.5
Der
Der
Der
Der
2
1.5
8084 + S22
8084 + X 33
411
8084
1
0.5
0
6
9
Thic kne ss m m
10
11
12
Fig. 3: Variation of toughness with specimen thickness

These results show that there is an increase in toughness of a composite with increase in resin
toughness, however, the scatter makes the difference difficult to quantify. The slopes of these
lines suggests that there is little or no dependence on glass content. Thus the variation in
interlaminar toughness of these composites seems to be determined wholly by the conditions
along the plane which the crack is forced to follow. These conditions appear not to be
determined by a global glass to resin ratio. Previous work ascribes the toughness of a
composite as being determined by the toughness of the resin and by fibre bridging effects [26]. Improvements to resin toughness have been said not to be fully transferred to a composite,
and that there was a limit above which the benefits of further toughening of the matrix were
marginal [3,4].
Extreme Value Analysis
An alternate analysis of the results has been carried out applying extreme value theory
[15,16]. Firstly, because of the very weak dependence on glass content the results from all the
composites made from a particular resin were pooled, and the values from each specimen
ordered from smallest to largest. Thus from the four composites made with Derakane 411,
twenty smallest (or largest) values of toughness were obtained which were again ranked in
increasing order. These values have been presented as a Gumbel plot in figure 4, in which the
ordinate scale is a function of the cumulative probability (F(x)) defined as i/(n+1) where i is
the rank of a particular value, and n is the total number of values.
This figure represents the cumulative probability of observing a toughness less than or equal
to a given value, and the linearity of the plot indicates a reasonable fit to the Gumbel
distribution - a Type 1 asymptotic distribution of smallest or largest extremes. Data from Der
8084 composites are also displayed, and the differences between these two curves probably
reflects different interactions of the resin with the reinforcement
II - 259
-Ln(-Ln(F(x)))
3.5
3
2.5
2
1.5
1
0.5
0
-0.5 0
-1
-1.5
-2
95
Der 411
90
Der 8084
Cumulative
Probability
%
70
50
0.5
1.5
25
2.5
10
-2
Toughness (kJ m )
Fig 4: Probability plot of maximum toughness recorded for each specimen

Fig. 5 shows the Gumbel plot for the maximum specimen toughness values for Derakane
8084 blended with S22, and compared with the values obtained from composites with the
untoughened resin. Here the advantages of toughening the resin is clearly seen, the offset
between the curves is similar to the change in toughness of the neat resin when blended with
the additive. Likewise, a plot for Der 8084 + X33 composites gave the improvement to the
toughness as an offset of about 1.1kJm-2. These plots also indicate that the probability of
having a toughness less than a specified value is much reduced for composites made with the
modified resin.
3.5
3
2.5
Toughened
Un-toughened
95
90
-Ln(-Ln(F(x)))
2
1.5
1
70
0.5
Cumulative
Probability
%
50
0
-0.5 0
25
10
-1
-1.5
-2
1
-2
Toughness (kJ m )
Fig. 5: Probability plot of effect of resin toughening on composite delamination
II - 260
Because the crack growth behaviour in these composites was seen to be a stick - slip process,
then the maximum values observed for each specimen are most likely to be for crack growth
initiation. The wide range of these values would then seem to be associated with the varying
effects of crack pinning by the glass filaments and from crack deflection as a result of
following the undulating surface of the cloth. The smallest of these maximum values could
be as a result of initiation from resin rich areas within the composite. Comparison with the
largest values then gives a measure of the contribution due to glass fibres and the structure of
the cloth.
Table 2 lists these smallest and largest values for the four resin systems , and from the above
assumptions the difference gives how much this particular glass cloth might contribute to
toughness in a composite. This contribution is thus about 0 to 1.4kJ m-2 in addition to the
toughness of the resin for Der 8084 based composites. Except for the Der 8084 + S22 matrix,
the smallest values are much larger than the toughness of the unreinforced resins especially in
the case of Der 411, and shows that these values also have a contribution from the
reinforcement.
Table 2: Smallest and largest values of the observed maximum toughness
Composite Matrix
Smallest Maximum
(kJ m-2)
Largest Maximum
(kJ m-2)
Der 411
Der 8084
Der 8084 + X33
Der 8084 + S22
0.76
0.69
1.70
2.47
1.60
2.05
3.19
3.80
Contribution by
Glass Cloth
(kJ m-2)
0.84
1.36
1.49
1.33
CONCLUSIONS
The delamination resistance of vinyl ester resin - glass woven roving composites can be
greatly improved through toughening the resin. Scatter in the results and the use of average
values hides the true advantage of using a toughened resin in a composite to improve
interlaminar fracture resistance. However by applying extreme theory to the largest (or
smallest ) toughness values recorded from each test specimen, the benefits of increased resin
toughness becomes much more apparent. For the resin systems used, the improvement to the
resin is clearly transferred to the composite. The reinforcement also contributes a variable
amount to the delamination resistance which appears to be between 0 and 1400 Jm-2. Global
variation in glass content does not appear to be a factor in determining this reinforcement
contribution.
REFERENCES
1.
Martin, R.H., Interlaminar Fracture Characterization: A Current Review, NASA -CR-187573,

1991.
2.
Slepetz, J.M. and Carlson, L., in Fracture Mechanics of Composites, ASTM Special Technical
Publication 593, 1975, p.143.
3.
Bascom, W.D., Bitner, J.L., Moulton, R.J. and Siebert, A.R., The interlaminar fracture of
organic-matrix woven reinforcement composites, Composites 1980, 9
II - 261
4.
Hunston, D.L., Composite Interlaminar Fracture: Effect of Matrix Fracture Energy, Compos.
Technol. Rev., Vol. 6, 1984, p. 176.
5.
Huang, X.N. and Hull, D., Effects of fibre bridging on GIc of a unidirectional glass / epoxy
composite, Compos. Sci. Technol., Vol 35, 1989, p. 283.
6.
Briscoe, B.J. and Williams, D.R., Interlaminar Fracture Toughness of Aramid/Epoxy

Laminates, Compos. Sci. Technol., Vol. 46, 1993, p. 277.
7.
Pham,S. and Burchill, P.J., Toughening of Vinyl Ester Resins with Modified Polybutadienes,
Polymer, Vol. 36, 1995, p. 3279.
8.
Burchill, P.J. and Pearce, P.J., Epoxy Acrylate - Based Resins, Polymeric Materials
Encyclopedia, Vol. 3, Salamone, J.C., Ed., CRC Press Inc.,1996, 2204.
9.
Material synthesised by P.J. Pearce according to information provided by BF Goodrich.
10. Kies, J.A., and Clark, B.J., Fracture propagation rates and times to fail following proof stress in
bulk glass, Fracture 1969, Ed., Pratt, P.L., (Chapman and Hall. London, 1969), p. 483.
11. Hashemi, S., Kinloch, A.J. and Williams, J.G., The analysis of interlaminar fracture in uniaxial
fibre-polymer composites, Proc. Royal Soc., Vol. A427, 1990, p. 173.
12. Rosensaft, M. and Marom, G., A one Specimen Mode I Delamination Test for Composite
Materials, J. Comp. Tech. Res., Vol. 10, 1988, p. 114.
13. Kanninen, M.F., An Augmented Double Cantilever Beam Model for Studying Crack
Propagation, Int. J. Fracture, Vol. 9, 1973, p. 83.
14. Williams, J.G., The Fracture Mechanics of Delamination Tests, J. Strain Anal., Vol. 24, 1989,
p. 207.
15. Gumbel, E.J., Statistics of Extremes, Columbia University Press, New York, 1966.
16. Mann, N.R., Schafer, R.E. and Singpurwalla, N.D., Methods for Statistical Analysis of
Reliability and Life Data, John Wiley & Sons Inc., New York, 1974.
II - 262
DYNAMIC CRACK PROPAGATION AND

PERFORATION OF LAMINATED COMPOSITES
USING A SPLIT HOPKINSON PRESURE BAR
Sylvanus N. Nwosu 1 and Gregory J. Czarnecki 2
1
Department of Physics, Dillard University, New Orleans, LA 70122, USA

2
Wright Laboratory, Wright Patterson AFB, Ohio 45433, USA
KEYWORDS:
Hoskinson bar
perforation, dynamic impact, energy, crack propagation, laminates,
II - 263
II - 264
II - 265
II - 266
II - 267
II - 268
II - 269
II - 270
II - 271
II - 272
EVALUATION OF MULTIAXIAL TEST DATA OF

UD-LAMINAE BY SO-CALLED "FRACTURE-TYPE
STRENGTH CRITERIA" AND BY SUPPORTING
PROBABILISTIC MEANS
Ralf G. Cuntze
Division of Spacecraft, MAN Technologie AG, Liebigstr. 5a, 85757 Karlsfeld/Munich.
SUMMARY: Overall objective of this elaboration is to calibrate fracture criteria resp. to

define size and shape of the fracture body as cheaply as possible by executing the
characteristic tests only.
Based on knowledge, achieved by investigating v. Mises, Mohr
- Coulomb and the new physically based Hashin - Puck Strength Criteria for inter fibre
fracture (IFF) of brittle unidirectional laminae, a new and general concept for the derivation of
fracture criteria will be proposed and applied to fibre reinforced plastics (FRP).
The fracture body derived here is basically piecewise smooth, each piece representing a single
failure mode. As interpolation functions the invariants associated to the material's symmetries
are utilised. Physical basis is the reference to the 2 fracture-types in a material: normal
fracture (NF) and shear fracture (SF). For the subsets of failure modes "fibre fracture (FF)"
and "IFF" two conditions for FF and three for IFF are derived. These five conditions describe
the five failure modes or mechanisms occurring, and five failure modes are the maximum
number a transversally-isotropic material, modelled a crystal, can possess.
In the transition zone of two failure modes or domains of mixed fracture, respectively, a
probabilistic modelling has to be applied. This finally leads again to a smooth surface of the
complete fracture body.
As the most remarkable results of the elaboration have to be pointed out: Consider
micromechanics resp. real material stresses in the constituents fibre and matrix (incl.
interface) which only can fail, however, formulate and visualise in lamina stresses at
composite resp. macromechanical level and think in Mohr's fracture stresses.
The application of the criteria to test results is very promising. Erroneous results, possible if
applying the traditional global (stress) interaction criteria, should not be achieved.
KEYWORDS: fibre reinforced composites, strength criteria, test data evaluation
INTRODUCTION
"Hot spots" of load carrying isotropic structures have to be designed versus various (Fig. 1)
strength failure modes including onset of yielding and fracture resp. onset of cracking.
In case of laminates built from brittle UD-laminae the stress man has to dimension each
II - 273
lamina (ply-by-ply) versus IFF and fibre-fracture (FF). Whereas IFF indicates the onset of
fibre-parallel cracking in one lamina of the laminate FF will indicate the onset of fibre
cracking. First FF of a lamina usually marks the final fracture of the laminate. A reliable
prediction of IFF under 3D-states of stress is mandatory for the calculation of progressive
failure.
As well in the stress analysis as in the failure criterion of the strength analysis there are
scattering design parameters. The uncertainty of these parameters (loads, strengths,
geometrical quantities, young's module, ...) is of physical nature or of statistical nature
(shortage of information due to a too small sample size of a certain design parameter
measured). Besides this there is always some uncertainty in the calculation model.
This scatter is usually considered in the design by the use of fixed, deterministic factors of
safety, better called design factors of safety (DFOS), which are based on experience with
structural tests. In aerospace industry structures are dimensioned by using DFOS j which
increase the so-called design limit load (DLL) to the design ultimate load (DUL) according to
jult DLL. This procedure includes the idea of proportional loading and that the reserve factor
fRes is related to the external load only. Above "load factors" are discriminating the onset of
yield (e.g. jpo,2 = 1.1) for isotropic material and ultimate fracture (e.g. jult = 1.5).
The design allowables to be used as strength values are statistics-based minimum values
[Cun96/2]. This might be a so-called Mil-Hdbk 17 "B-value" with 90 % reliability and 95 %
confidence probability, the latter number regards the confidence in the transfer of the finite
number of sample test data to the parent distribution with its infinite number of data.
Proofs of Design or design verifications done by analysing and/or testing a structure have to
be given for failure modes such as stability, deformation, vibration and strength. Target of the
designer is to achieve a margin of safety MS 0 for each single failure mode and - in this
paper - for fracture modes as a subset of them. In order to achieve a reliable margin of safety
(MS = fRes -1) in the Strength Proof of Design the engineer should accurately know the
normally non-linear 3D-state of stress in the critical material element and its separation into
load-induced and residual stresses!
Fracture criteria as a subset of strength criteria should enable the engineer to perform the
Strength Proof of Design for any 3-dimensional state of stress resp. combination of stresses
with a minimum amount of test data.
Verified fracture criteria as a subset of strength criteria (Fig. 1) enable to do this. They can be
calibrated after execution of the basic strength tests, only.
The traditional criteria such as Tsai/Wu have some shortcomings which forces research to
correct or replace them [VDI97]. However, in the last four years heavy progress has been
made in Germany with the derivation of more physically-based fracture criteria for brittle
unidirectional isolated laminae. On the basis of [Has80] A. Puck developed the "Hashin-Puck
(Action-Plane) Strength Criteria" for IFF [Puc92] assuming 2 modes. Additionally, tests were
performed under a German grant to verify the criteria [VDI97].
Parallel, the author tried to generate strength criteria, each of which should be related to one
single failure mechanism resp. fracture mode (not type) and formulated by invariants, thus
taking the material's symmetries into account. Basis are the types of fracture in a material
II - 274
which are normal stress induced fracture (NF) and shear stress induced fracture (SF). This
fracture-type concept proposed may be applied to any brittle or ductile behaving material
[Cun96/2].
Fig. 1: Strength Failure Modes (modi) of Composites. (NF = normal fracture, SF = shear
fracture. F||, ...,Fcr are the UD-lamina's failure functions representing fracture modes)
In this paper brittle UD-laminae are described, only. The fracture criteria derived are stress
criteria and have as foundations: The material is homogeneous (smeared or averaged material)
and can be modelled an ideal crystal with its symmetries and mutual independent basic
strengths from a meso/macromechanical point of view. However, for the in reality
heterogeneous composite (is a structure not a material) the micromechanical behaviour of the
constituents (real materials) fibre, matrix and interphase will be considered too.
FRACTURE MECHANISMS IN UD-LAMINAE

Fracture is understood in this arcticle as a separation of material, which is free of damage such
as cracks but not free of defects and flaws prior to loading. In this context crushing of a
porous material has to be seen a separate fracture mechanism resp. (fracture) failure mode
and type.
Idealising a material a crystal one can draw from Table 1 the information: The number of
symmetries determines the number of strengths. The higher the structural level of the material
the more basic strengths have to be measured.
Due to the fact 5 strengths being usually measured in case of a transversally-isotropic lamina
the application of the crystal model is still general practice.
Fig. 2 shall give answer to the question: How many types of fracture are recognised in case of
isotropic and transversally-isotropic ideal materials? These are just two: the shear stress
induced fracture SF (Mohr's stress n is acting at the "plane" of fracture) and normal stress n
II - 275
induced fracture NF - as we know - for isotropic materials, but also two for the UD-material.
However, the letter case comprises several SF and NF which belong to FF and IFF.
Table 1 : Number of Strengths and Fracture Types for Various Ideal Materials (t = tension, c
= com-pression; ||, = parallel, transversal to the fibre; W = warp, F = fill. Superscript c =
critical)
material has
allocation to
crystal type
symmetries
material example
elasticity
quantities
increasing structural level

1
2
transversally
rombically
isotropic
anisotropic
5
9
UD-lamina,
fabrics
mat
E||, E, G||
EW, EF, GWF, FW, E3,
||,
3W, 3F, GF3, GW3
no structure
0
isotropic
2
resin matrix,
mono. ceramics
E,
(2)
I1, I2, or J2 ...
(5)
R||t, R||c, Rt,
Rc, R||
(5)
I1, ..., I5, ...
t
c
K Ic , K IIc
t
c
t
K || c , K || c , K c
strength quantities Rt, Rc or

(resistances)
(R , R)
invariants
fracture
toughnesses
(2)
(2)
K cc , K || c
RWt, RWc, RFt, RFc, R3t,

R3W, RFW, R3F, R3c (9)
I1, ..., I7, ...
(5)
Also the physical fracture "planes" are pointed out in the figure.
A comparison of Fig. 2 with Table 1 leads to an essential conclusion: One strength governs
one distinct fracture type.
The characterisation of the strength of transversally-isotropic composites requires the
measurement of five independent basic strengths: R||t, R||c (fibre parallel tensile and
compressive strength) as well Rt, Rc (tensile, compressive strength perpendicular to the
fibre direction) and R|| (fibre parallel shear strength). R||t is determined by the strength of the
constituent fibre, R||c is generally governed by shear stability. This includes different
microfailure mechanisms: The matrix may shear under loading and does not stabilise the
generally somewhat misaligned fibres embedded. Hence it comes to bending and "kinking"
(structure behaviour). But, the load "grasping" fibre as the well-stabilised, stiffer constituent
may shear (this is material behaviour) under ||c and ||, too. Rt is determined as well by the
relatively low strength properties of the matrix (cohesive failure) and the interphase material
in the interface fibre-matrix (adhesive failure caused by a weak fibre-matrix bond), as by the
fibres acting as embedded stress risers.
II - 276
x
fp
R II
45
x
n
brittle
fp
x
n
X1
SF II
II
X3
45
- 21
IFF
X2
X2
SF
X1
X1
NF
II
R II
X3
SF II macromechanically
ductile
fp
t tensile strength
R
(ductile tearing)
t
R cohesion
X2
NFII
X3
c
R compressive strength
c
R cohesion
n
X2
X1
Shear Fracture (SF)

n
R II
X3
FF
R t tensile strength
R = R t separating strength
x
n
X3
Normal Fracture (NF)
fp
45
NFII micromechanically
X2
R
SF
X1
Fig. 2: Fracture Types ( modes) in case of Brittle and Ductile behaving Isotropic Material
and Fracture Modes in case of Transversally-isotropic Material (nA ( = fp) = n).
Well known from literature is the shift of the fracture limit of IFF strain resp. stress to higher
values the absolutely thinner the lamina is and the stiffer in comparison the embedding
laminae and the laminate are. Whereas the free isolated lamina (UD-specimen) if transversally
tensile loaded is underlying a weakest-link failure (resulting in a Weibull distribution for Rt)
the embedded lamina will be strain controlled by its neighbours. Inherent defects or
microcracks as often reported can grow together to form mesocracks only in case they have a
chance to open. However, these microcracks are arrested fracture-mechanically by the
embedding laminae.
In order to consider the constraint effect on a "thin" embedded lamina following [Fla 82] and
some own investigations the author recommends e.g. for CFRP (tthr 0,35 mm) as correction
formula for the design allowable Rt due to
Rt (t < tthr) = Rt t thr / t .
(2.1)
GENERAL FORMULATIONS OF STRENGTH CRITERIA AND CONDITIONS

>
Strength criteria F = 1 are generally formulated as (isotropic case shown for simplicity)
<
>
F ({}; {R}) = F(x, y, z, zy, zx, yx; Rt, Rc) = 1

<
... criterion.
(3.1)
In the failure (here fracture) function F the vector {} contains the six stresses in the coordinate system of the structure - as above - or the six lamina stresses. The vector {R}
contains all basic strengths to be applied such as Rt, Rc in isotropic case.
II - 277
Failure surfaces are determined by the strength condition F = 1

F (fRes {}; {R} = 1
... condition
(3.2)
To achieve this, the stresses in equation 3.1 have to be multiplied by the reserve factor f Res,
which is usually dropped for reasons of a simpler writing.
Composites with a polymeric matrix exhibit brittle behaviour from a macroscopic point of
view. Thus a formulation of strength criteria can be based on the fracture hypotheses of Mohr
and Coulomb, adapted to UD-laminae. This was suggested by Hashin [Has80]. Hence, using
Mohrs fracture stresses, which act together at the same plane, the isotropic equations read:
= F({Mohr}, R) = 1
... NF
(3.3a)
= n/R = 1
(normal stress hypothesis)
F
= F({Mohr}, R) = 1...
... SF
(3.3b)
= n/(R - n) = 1 e.g.(shear stress hypothesis, Mohr-Coulomb)
With R: = separating strength (resistance) R: = cohesion (resist.), : = Coulomb friction.
F
This outlines: one condition is to be used for each fracture mechanism which is dominated by
one fracture resistance R or R (s. also [Cun96/2] for isotropic materials. A resistance (of
the fracture plane) corresponds with the basic strength associated if the action plane of a
distinct stress comes to be the potential (load must be high enough, too) fracture plane: e.g. in
brittle case max = R = Rt (isotropic) or Rt (transversally-isotropic). In case of brittle
behaviour only R can be measured and vice versa R in case of ductile behaviour. Therefore,
R principally remains undeterminable until the fracture angle fp is determined. This is also
valid for the fracture plane stresses {Mohr}.
The "curve fitting" of the course of test data can be performed much easier by not taking
Mohr's stresses and the 2 resistances R, R but {} and the 2 basic strengths. However,
equation (3.1) will now be used in Mohrs sense meaning one basic strength governs one
fracture mechanism NF or SF
F = F ({}, RNF) = F({}, Rt) = 1 and F = F ({}, RSF) = 1.
(3.4)
Both, the physical basis and the numeral handling are improved by using invariants as
interpolation functions, a well known and classical approach.
DERIVATION OF "FRACTURE-TYPE STRENGTH CRITERIA"

Description of 3D-States of Stress
Three-dimensional states of stress in a lamina may be described by lamina stresses {}. A
further possibility is given by using Mohr's fracture stresses mentioned above. They are
acting in the potentially physical fracture "plane" (Fig. 3) and are decisive for fracture. In case
of
II - 278
II - 279
The prediction of failure - made by whichever condition - is satisfied first. Terms in table 3
enveloped by dots are local correction terms to approximate (fit) the mixed fracture domains.
They are not global interaction terms as e.g. in the global fracture criteria of Tsai-Wu.
The stresses 21 and 1t do not have a common action plane, but marginally act
micromechanically together. However, 21 and 1c almost have the same action plane in case
of brittle constituents fibre and matrix.
How the stress efforts are adding, linearly or quadratically or even cubically is determined by
the course of test data.
The fracture type Crushing Fracture is added in Table 3 for reasons of completeness. It
belongs to a material which due to its porosity will fail by internal voluminous degradation
under compressive stresses and not in "planes" any more.
Determination of IFF Calibration Conditions besides the Basic Strengths
Putting the compressive strength
R c
into
F delivers
a = b - 1.
(4.1)
A second equation, if using the measurable fracture angle fpc as the other obtainable

information of the lateral compression test will be Mohr-Coulomb's condition (c: = cos fp c )

F F
c s

(4.2)
/
= = + cot an 2 c = 0,5 ( ) yielding
b = 1/(4 c 2-1).
fp

n n
s c
Trials to estimate the tensile fracture stress tt in an appropriate test had no success
[VDI97]. However, a good estimation can be achieved by using the information Rc being
Weibull distributed. Following [Awa78] easily can be derived
tt = Rt/ k 2
with
k = Weibull's module.
(4.3)
With respect to Fig. 6 a value for a point on the "bulge" ( 2 D , 2 D ) can be estimated or
||
2D
2D
directly a value for the linearized slope
m|| ( || - R||)/
(4.4)
Analytical Derivation of Mohr's Envelope Curve
Resolving the fracture condition F = 1 for 2
delivers in case of 23 = 0
2 . 2b= -R c a + 23b + a 2 R c 2 8a R c b 3 + 4 b R c 2 = f ( 3 )

~
c
with
And
II - 280
nt
(a R c 2 b 3 + 2 b 2 ) / ( 4 b c / ( 2 3 )), 3 = var iable
= | ~~ (-2 + 3) |,
sc
c
s
s
n = ~ 22 + ~ 23, ~ =
1 ~2 .
c
(4.5)
(4.6)
(4.7)
Then the cohesion can be calculated from
R = nt (n = 0).
This quantity corresponds to
A
Puck's R [Puc96].
PROBABILISTIC MODELLING OF MULTIPLE FAILURE DOMAINS

The application of probabilistics will connect adjacent modes.
The influence of the FF modes on IFF can be taken into account very simply by practical
correction terms (see [Cun96/1]) in the IFF conditions or by probabilistics as shown now.
To estimate the joint probability of failure of the different IFF modes as logical model of the
failure system the series system or union U of " = 3 IFF failures will be taken. This delivers a
failure probability pf on the safe side (index S for sum, see [Cun96/2]), of
pf p f, = 1 - (1-fS)(1-f||S)(1-fS)
(5.1)
since small p f, are expected. Correlation among the (component) failures leads to a
smaller (system) failure probability pf, above expression therefore is conservative. In case of a
Weibull-distributed strengths
Rt, Rc, R|| as "sum" function
r
f S = 1 exp[ ( ) k ]
(5.2)
w
is valid (it represents the distribution of the extreme defects) with r = stress variable, k =
Weibull's module (characterising the scatter of a strength) and w = shape parameter. A further
parameter characterising the position of the distribution is put zero. This is in any case
permitted for very brittle materials (e.g. monolithic ceramics, k 2. An estimation of
Weibull's parameter is possible [Pli95] on basis of the mean value m and the coefficient of
variation of the parent distribution of each strength measured (bounds: cov < 25 %, k > 4,8)
by
k 1.2 /cov;
w = m /(1+1/k);
m= R
with
= Gamma function.
(5.3)
As estimations for m, v the values of a well sized sample are taken without considering the
confidence interval.
Combining (5.1) and (5.2) yields the following equation for the IFF body
r

w

r
+ ||
w||
k||
r
+
w

= "n (1 pf ) .
(5.4)
Instead of a uniaxial stress the equivalent stress as representation of the multiaxial state of
stress will be put in (5.4): r = ,
r|| = ||,
r = .
It can be calculated in
II - 281
case of F (shorter than for the IFF cases having cubic invariants)
F = f Re s / Rc = 1
f Re s = [-a I2 +
or
= Rc/ f Re s
(5.5)
a 2 I 2 2 + 4 b I 4 + 4c R c 2 / R || 2 / 2 [ b I 4 / R c + c I 3 R c / R 2 || ].
For the IFF Proof of Design an engineering approach estimates fRes in the mixed failure
domain
(1 / f Re s ) k = (1 / f Re s ) k + (1 / f Re s|| ) k + (1 / f Re s ) k = f(fRes(modes))
(MFD)
with
(5.6)
k min k (safe side) causing a maximum out-smoothing.

APPLICATION TO GLASS/EPOXIDE-FRP
Determination of Curve Parameters (applying MATHCAD)

For the application of the criteria achieved at first all informations on the basic strengths: The
type of distribution, the mean value R and the coefficient of variation cov have to be
provided. Table 4 represents an example set of data. The sources are various investigations
[Kna72], [ZTL], [VDI97], due to the fact that a real complete test series of all interesting
combinations of stresses including 1 for FRP is not existing.
Table 4: Strength Properties (f = 0.60) and data of Calibration Points (mean values)
c
fp = 49 ; = 73 MPa; = 95 MPa
tr = 86 MPa, 1.05 R || ( 2 = 0)
( 2 , 21 ) = ( 40, 0),(28, 40 ),(0, 58),( 90, 72)

t
R || : Weibull, 1500 MPa, cov = 6 %
c
R || : Weibull, 1200 MPa, cov = 7 %
t
R : Weibull,
40 MPa, cov = 12 %
c
R : Weibull, 144 MPa,
R || : Weibull, 61 MPa,
cov = 7%
cov = 10 %
Fracture Curves: Deterministic Approach incl. Effect of Correction Terms

Fig. 4 comprises four different cross-sections of the five-dimensional IFF-body (2, 3, 23,
31, 21): The graph (21, 2) represents stresses in the plane of the lamina which are of
highest interest in the justification of the stress-output delivered from Classical Laminate
Theory (CLT). Three curves F|| (transversal SF), F|| (transversal-parallel SF) and F||
(transversal-parallel NF) envelope the "50 % safe domain". The graph (31, 2) outlines by
t
"Puck's corner" that in contrary to (21, 2 ) the shear stress 31 has not the same action plane
II - 282
as 2t. The remaining two graphs (2, 3) and (23, 2) are well-known from isotropic
materials. In these isotropic cross-sections fracture can be described excellently by the
homogenised lamina stresses (2, 3, 23) or artificially smeared stresses, respectively. In the
hydrostatic domain 2c = 3c -10 Rc the IFF cureves will be closed by F||z, due to
Poisson's effect.
Fig. 4: IFF-curves (Glass FRP).(MFD = mixed

failure domain = fracture due to 2 modes, MPF
= multiplane fracture of the same mode NF)
II - 283
Fig. 5: FF-and IFF-curves (Glass FRP).

(with correction terms, different scales)
Fig. 5 outlines the limited applicability of the homogenized lamina stresses. The lamina
(composite) stress 1 is not the fracture active stress, however, the stress 1f of the constituent
(material) fibre. In order to nevertheless remain in the graphical display on composite level
the fibre stress has to be multiplied by the fibre volume fraction f. This should be an
acceptable approach due to the extreme difference in the stiffness and thereby in the load
carrying capacity of fibre and matrix. Tests performed, see [VDI97], show the effect of
Poisson: Under biaxial compression the load can be only increased until tensile fracture stress
or strain of the fibre will be met.
Simple 2D-Approximation of the (2, 21) Fracture Curve
For an improved but nevertheless simple approximation of the (2, 21)-test results, which are
mandatory for the evaluation of CLT stresses, the utilisation of the formulae
3
F|| : a 2 / R t + c 2 / R t 2 + d 212 / R || 2 + e 2 212 / R || = 1

(6.1a)
F:
II - 284
a 2 / R c + b 2 / R c 2 + c 212 / R || 2 = 1
(6.1b)
is recommended with a , b from the

equations (4.1, 4.2) and the calibration points
indicated in Fig. 6 by a square. Due to
micromechanical interaction caused by the
matrix the author micromechanically combines
NF|| and NF in his approach (Puck: SF and
SF|| macrome-chanically).
50
nt
60
c
2 fp
50
50
t
R
MFD
100
50
40
100
t
R
2
50
150
c
-R
80 R
2D
Fig. 6: 2D-Approximation (50%, Mean Curve.

{ Test Data from [Kna72]. calibration points
The probabilistic out-smoothing in the MFD or

transition zone, resp., of the piecewise smooth
fracture-type related domains of the IFF-body
shall be visualised by Fig. 7.
c
-R
150
Fracture Curves: Probabilistic Approach
21
100
3
2
0
-50
20
-160
-120
-80
-40
-R
40
t
R
Fig. 8: An analytically derived Mohr-Coulomb

"Envelope" Curve (bold) with MFD-circles and
Transition Circle (dashed).
( R = friction cohesion, = tan ,

m (C )
m
i (C )
X =(mk) t
4h i
2h t
N
iO
I
N
iO
N X2
i
I
O I
N
iO
N X2
i
I
A-curve
-100
B-curve
mean-curve
-150
-200
Fig. 7: "Global Fit" versus "Modus Fit".

Design curves showing Probabilistic Outsmoothing in the Mixed Failure Domain (MFD)
Further the traditional "global fit" by polynomial

interaction failure models (applied by most of
the traditional strength criteria) is compared with
the "modus fit" proposed in this article. The application of the probabilistic out-smoothing
leads again back to one equation for the description of the global course of test data. Based on
the numerical investigations for Fig. 7 one further conclusion can be drawn that multiple
fracture in a mode, e.g. NF caused by biaxial tensile stresses, will be not covered by the
simple modus fit ( c = 0 ). This correction term is necessary to approximate the test data in
the MPF regime. Beside this, the term also has an effect in the MFD regime (SF, NF),
which is not adverse because such brittle materials with a high contents of defects show both
sorts of out-smoothing.
Essential for the dimensioning of FRP-structures is the "B-value Design Curve", most often.
Fig. 8 eventually shows (probably for the first time) an analytically derived Envelope Curve.
To remind us of the former enveloping procedure the compression circle and the transition
(the fracture mechanism jumps from NF to SF) circle are included. Also the relationship of
II - 285
50
fracture angle, friction angle and the local Coulomb friction coefficient c derived from a
Rc-loading are given.
In the MFD no realistic fracture angle will be determined. Consequently the nt (n) curve
again has to be an evelope of the circles (2d, 3) or (2, 3d) the values of which are the
curve points of the probabilistic function (5.4) describing the transition from SF to NF.
And, looking at the FEM codes the fracture conditions offered for a so-called "linear or
parabolic Mohr-Coulomb Material" is nothing but a continuous mathematical description of
an envelop of Mohr's circles in the MFD. It just has relation to Mohr's fracture angle outside
the MFD where a "pure" SF or NF mode is dominating.
CONCLUSIONS, OUTLOOK AND REMARKS
Test Data Fit and Failure Modes
The author has the same point of view as L.J. Hart-Smith: It is scientifically incorrect to
employ polynomial interaction failure models whenever the (micromechanical) failure
mechanism of the critical constituent of the composite changes with the state of stress.
And - valid for any material - the main shortcoming of global criteria is: A change in one
strength (which belongs to a distinct fracture mode) has an effect on the whole failure
surface (surface of fracture body) which represents all modes ([Har93, Cun94]).
Interactions between different failure modes are incorrect. Interactions between stresses
affecting the same failure mode in the same constituent of the composite are permitted.
Separate characterisations are needed for each failure mechanism in each constituent of a
composite of materials. The fit has to be performed for each mode separately.
It does not matter whether a state of stress comes from a one-, a two- or a threedimensional stressing as long it belongs to one fracture mechanism or mode.
Consequently all these states of stress can be described by the same fracture condition.
In the isotropic planes test data excellently can be described by the homogenized stresses
2, 3, 23 (composite = smeared or averaged material). However, if the UD-lamina is
e.g. subjected to (1, 2, 3) the constituent (real isotropic or anisotropic material) stress
1f will be responsible for fracture and not the composite stress (1).
Strength Criteria:
There are 3 categories of failure criteria: stress-based, strain-based and energy-based. The
criteria can be grouped into independent, partly (stress) interactive and globally
interactive ones. Their development may be based on micromechanical (constituent level)
and/or (meso/) macromechanical investigations (lamina resp. composite level).
The author prefers stress criteria in the proof of design because: stresses fracture the
material, residual stresses are taken into account and strain history is considered in the
strength values.
II - 286
The concept of strength criteria presented exhibit a physical basis, a good fit of test data
course, invariance with respect to coordinate transformations, a relatively easy
application involving numerical stability and applicability to ductile behaving composites
as well.
From the several applications a preliminary assessment of the concept is possible:

- the fracture conditions F = 1 determine the shape of the failure surface with its five
piecewise smooth areas, representing the five modes
- a separation of mechanical and probabilistic modelling would be advantageous for an
accurate interpretation of test results but is not completely possible. The probabilistic
tool has to be taken to smooth-out in the mixed failure domain (MFD) resp. transition
domain of two or more modes the adjacent mode-related pieces of the "failure
surface"
- the out-smoothing approximately can be done by applying correction terms instead of
performing real probabilistic calculations
The application of the concept to various isotropic test results was very successful and is
very promising in case of transversally-isotropic UD-laminae, too. A big challenge will
be an application to the various woven fabric laminae.
Because "high tech" composite parts normally have notches and often have to be
designed to damage tolerance they were qualified more or less by tests. So, the physical
drawbacks of the traditional strength criteria applied to unnotched (and notched)
structures have not been revealed.
Design (dimensioning) and Proof of Design:
There are 5 basic strengths (or 5 resistances, if Mohr would be strictly applied) and as
much modes or fracture mechanisms, respectively.
To establish 3D-design curves the concept just needs for the "fitting" of the 5 pure modes
the 5 basic strength allowables, the fracture angle fpc (pertaining Coulomb's friction) and
a calibration point at the (2, 21)-bulge. If micromechanical interaction will be taken

into account 3 more calibration points have to be provided or can be assumed from
experience with a FRP family in order to determine the remaining curve parameters
belonging to the correction terms. The curve parameters of each failure function in this
paper are computed- beside the basic strength points - from further distinct calibration
points, only. Generally of course, a root mean square-fit is foreseen to achieve their
values.
For the dimensioning the designer needs the coloured contour plots or profiles of the
stresses as output of the FEM stress analysis and for the Proof of Design he should get
offered by the FEM codes coloured profiles of fResmode too. This is indirectly still
practiced when plotting the equivalent stress of v. Mises (3 J2/yield2 = 1
fResmode =
(yield/equiv)0,5 = yield/3J2)0,5 ) plots for the mode "onset of yielding" in case of

isotropic material or for NF with the principal tensile stress (I/Rt = 1 fResmode =
Rt/I). The lowest fResmode computed for IFF and FF drives the design.
II - 287
In the mixed failure domain equ. (5.6) delivers an estimation for fRes = f(fRes(modes),
therewith bypassing together with the correction terms a real probabilistic out-smoothing.
The determination of fRes in case of cubic invariants is a little more laborious. If residual
stresses have to be taken into account, due of
[VDI97].
fRes . {}(load) + {}(residual), see
ACKNOWLEDGEMENTS
The author highly appreciates the excellent support of his colleague B. Szelinski, the fruitful
discussions with Dr. "John" Hart-Smith and the comments of members co-operating under the
German R&D contract number 03N8002 of the BMBF [VDI97].
REFERENCES
[1]
Awaji, H. and Sato, S.: A Statistical Theory for the Fracture of Brittle Solids under
Multiaxial Stresses. International Journal of Fracture 14 (1978), R13-16
[2]
Boehler, J.P.: Failure criteria for glass-fiber reinforced composites under confining
pressure. J. Struct. Mechanics 13, 371-393
[3]
Cuntze, R.G.: "Fracture-type Strength Criteria" formulated by Invariants which consider

the Materials Symmetries of the Isotropic/Anisotropic Material used. Conf. on
Spacecraft Structures, Materials and Mechanical Testing. ESA-CNES-DARA:
Noordwijk, March 1996
[4]
Cuntze, R.G.: Bruchtypbezogene Auswertung mehrachsiger Bruchtestdaten und

Anwendung im Festigkeitsnachweis sowie daraus ableitbare Schwingfestigkeits- und
Bruchmechanikaspekte. DGLR-Jahrestagung, Dresden, Sept. 96, Tagungsband 3
[5]
Flaggs, D.L., Kural, M.H.: "Experimental Determination of the In Situ Transverse

Lamina Strength in Graphite Epoxy Laminates". J. Comp. Mat. Vol 16 (1982), S. 103116
[6]
Hart-Smith, L.J.: An Inherent Fallacy in Composite Interaction Failure Curves.

Designers Corner, Composites 24 (1993), 523-524
[7]
Hashin, Z.: Failure Criteria for Unidirectional Fibre Composites. J. of Appl. Mech. 47
(1980), 329-334
[8]
Hufenbach, W. and Kroll, L.: A New Failure Criterion Based on the Mechanics of 3Dimensional Composite Materials. ICCM-10, Whistler, Canada, 1995
[9]
Jeltsch-Fricker, R.: Bruchbedingungen vom Mohrschen Typ fr transversal-isotrope

Werkstoffe am Beispiel der Faser-Kunststoff-Verbunde. ZAMM 76 (1996), 505-520
II - 288
[10] Knappe, W. und Schneider, W.: "Bruchkriterien fr unidirektionalen Glasfaser/Kunststoff unter ebener Kurzzeit- und Langzeitbeanspruchung". Kunststoffe, Bd.
62 (1972), 864
[11] Kopp J. and Michaeli, W.: Dimensioning of Thick Laminates using New IFF Strength
Criteria and some Experiments for their Verification. Proceedings "Conf. on Spacecraft
Structures Materials and Mechanical Testing", ESA, 27-29 March 1996
[12] Mohr, O.: Welche Umstnde bedingen die Elastizittsgrenze und den Bruch eines
Materials? Civilingenieur 44 (1900), 1524-1530, 1572-1577
[13] Paul, B.: A Modification of the Coulomb Mohr Theory of Fracture. J. of Appl. Mech.
(1961), 259-268
[14] Puck, A.: Ein Bruchkriterium gibt die Richtung an. Kunststoffe 82 (1992), S. 607-610
(A failure criterion shows the Direction - Further Thoughts on the Design of Laminates
- Kunststoffe German Plastics 82 (1992), 29-32)
[15] Puck, A.: Festigkeitsanalyse von Faser-Matrix-Laminaten. (Modelle fr die Praxis).
Mnchen: Carl Hanser Verlag, 1996
[16] Rowlands, R.E.: Strength (Failure) Theories and their Experimental Correlation. In Sih,
G.C. and Skudra, A.M.: editors, Handbook of Composites, Vol. III, chapter 2, Elsevier
Science Publisher B.V., Madison, WI, U.S.A., 1985, 71-125
[17] Shn, S. und Gldner, H.: Bruch- und Beurteilungskriterien in der Festigkeitslehre.
Fachbuchverlag Leipzig-Kln, 2. Ausgabe, 1993
[18] Tsai, S.W. and Wu, E.M.: A General Theory of Strength for Anisotropic Materials.
Journal Comp. Mater, Vol. 5 (1971), 58-80
[19] Cuntze, R.G., et.al.: Zum verbesserten Festigkeitsnachweis von Bauteilen aus Faserkunststoffverbund. - Modellbildung und Experimente zu einer physikalisch begrndeten
Festigkeitshypothese fr die unidirektionale Schicht. VDI-Fortschrittbericht, Reihe 5.
BMBF-Vorhaben 03N8002 (A. Puck, adviser until 1995)
[20] Dornier, Fokker, MBB, DLR: Investigations of Fracture Criteria for Laminae (in
German). 1975-1980, Grant from Ministry of Defence, BMVg, Koblenz
II - 289
A SINGULAR HYBRID FINITE ELEMENT ANALYSIS

OF INTERLAMINAR CRACK PROBLEMS
Kyohei Kondo1 and Kazuhiro Yagi2
1
Department of Aeronautics and Astronautics, University of Tokyo

7-3-1 Hongo, Bunkyo-ku, Tokyo 113, Japan.
2
Space Systems Department, Aerospace Division, Nissan Motor Co., Ltd.
3-5-1 Momoi, Suginami-ku, Tokyo 167, Japan.
SUMMARY: Singular stress distributions near interlaminar crack in composite laminate

under various loadings are analyzed by a singular hybrid element method which is developed
based on the elasticity solutions and the variational principle of a hybrid functional.
Asymptotic solutions for displacements and stresses near the interfacial crack between
dissimilar anisotropic materials are utilized as the displacement and stress expressions within
the crack element. The singular hybrid element is used in conjunction with the conventional
displacement-based, quasi three-dimensional elements to study the detailed nature of the
displacements and stresses in vicinity of the interlaminar crack tip.
Analyzing a small interfacial crack in a large composite under mechanical loading, we
demonstrate that the present numerical solutions are in good agreement with the analytical
ones for a interfacial crack in infinite composite. Then, we apply the present method for
analysis of delamination cracks originating from transverse cracking in composite under
various loadings. Comparisons of the present solutions with the results by the displacement
finite element method show the accuracy and efficiency of the present finite element
procedure.
KEYWORDS:, delamination, stress singularity, potential energy release rate, finite element
method, singular element, hybrid element, variational principle
INTRODUCTION
Delamination is the most commonly observed failure mode in composite laminates under
mechanical and hygrothermal loadings which have been extensively utilized for the
lightweight structures. Onset and growth of the delamination can be characterized by the
fracture mechanics based on the singular stress distributions at the crack tip. The stress field
near the interfacial crack can be analyzed by the conventional displacement finite element
method [1, 2] or the singular hybrid finite element method [3, 4].
In this paper, a singular hybrid element method for studying the stresses near the interfacial
crack is presented. The fundamental solution for the displacements and stresses in vicinity of
the crack tip is reviewed first. The singular crack element is formulated based on the
elasticity solutions and a hybrid variational principle. Comparisons of the present numerical
II - 290
solutions with the analytical ones or the numerical ones by the displacement finite element
method demonstrate the accuracy and efficiency of the present finite element method.
ASYMPTOTIC SOLUTION FOR DISPLACEMENTS AND

STRESSES NEAR INTERFACIAL CRACK TIP
II - 291
II - 292
HYBRID FINITE ELEMENT FORMULATION
II - 293
Stiffness Matrix Formulation for Singular Crack Element

If we truncate the series expansion in Eqns 14 and 15 at n =N, we have the displacements and
tractions at the interelement boundary of Am as
In the present finite element approach, a 17-node, hybrid crack element shown in Fig.2 with
51 degrees of freedom is constructed. The total number N of terms in the truncated series in
Eqn 23 is takes as 14. Standard quadratic interpolation functions are used for the L in Eqns
25 to ensure matching of the boundary displacements of the singular hybrid element with
those of adjacent 8-node, 24 degrees of freedom, isoparametric regular elements as shown in
Fig.3.
II - 294
NUMERICAL RESULTS
To confirm the singular hybrid finite element solutions, we analyze a large carbon fiber
reinforced plastics (CFRP) [0o/90v] laminate with a small central crack of length 2a subjected
to the uniform far field stress 22 as shown in Fig.4. The total and singular stress
distributions along the interface are shown in Fig.5. The singular stress distributions are
shown in Fig.6 together with those obtained by the analytical solution for an infinite plate [7].
Comparison of the numerical solution for the interlaminar singular stresses with the analytical
one is also showre in Table 1. The potential energy release rates from the two approaches are
A large CFRP [o/-o] laminate with a small central crack is also analyzed. The numerical
solution for the potential energy release rates is shown in Fig.7 together with the analytical
one [7].
Then, we analyze CFRP laminate with delamination originating from transverse cracking as
illustrated in Fig.8. The stress distributions along the interface of CFRP laminate under
tensile strain 11 are shown in Fig.9 together with those obtained by the displacement finite
element method [2] with fine discretization as illustrated in Fig.10.
CONCLUSIONS
A singular hybrid interfacial crack element has been developed to study the stresses near
interfacial crack combining with the conventional regular elements. The present numerical
solutions for a small interfacial crack in a large composite are in good agreement with the
analytical ones for a crack in an infinite composite. And comparison of the present solutions
for delamination originating from transverse cracking with those by the conventional
displacement-based finite element method show the accuracy and efficiency of the present
finite element procedure.
REFERENCES
1.
Kondo, K. and Aoki, T., "An Energy Release Rate Approach for Free-edge Delamination
Problem in Composite Laminates", Composite Structures 4, Marshall, I.H., Ed.,
Proceedings Fourth International Conference on Composite Structures, Paisley, Scotland,
Vol. 2, 1987, pp.241-257.
2.
Kondo, K. and Yasue, Y., "Transverse Crack and Delamination in Composite Laminates
Subjected to Mechanical and Hygrothermal Loading", Composites '95 : Recent Advanced
in Japan and the United States, Kimpara, I., Miyairi, H. and Takeda, N., Eds.,
Proceedings of Seventh Japan-U.S. Conference on Composite Materials, Kyoto, Japan,
1995, pp.91-98
3.
Wang, S.S. and Yuan, F.G., "A Singular Hybrid Finite Element Analysis of Boundary Layer Stresses in Composite Laminates", Int. J. Solids Structures, Vol.19, No.9, 1983,
pp.825-837.
II - 295
4.
Kim, Y. and Im, S., "Delamination Crack Originating from Transverse Cracking in
Gross-ply Laminates under Various Loadings", Int. J. Solids Structures, Vol.30, No.16,
1993, pp.2143-2161.
5.
Ting, T.C., "Explicit Solution andf Invariance of the Singularities at an Interface

Crack in Anisotropic Composite", Int. J. Solids Structures, Vol.22, No.9, 1986, pp.965983.
6.
Washizu, K., Variational Principle in Elasticity and Plasticity, Second Edition, Pergamon
Press, 1974.
Qu, J. and Bassani, J. L., "Interfacial Fracture Mechanics for Anisotropic Bimaterials",
Journal of Applied Mechanics, Vol.60, No.2, 1993, pp.422-431
II - 296
II - 297
II - 298
II - 299
II - 300
FAILURE BEHAVIOR AND MORPHOLOGY OF

HYBRID SHORT-FIBERS REINFORCED COMPOSITE
UNDER THERMAL AND MECHANICAL LOADINGS
Kh. G. Schmitt-Thomas, Zhen-Guo Yang, and R. Malke
Institute of Applied Materials Engineering
Technical University of Munich, Arcisstrae 21, 80333 Munich, Germany
SUMMARY: The mechanical property, microstructure morphology, and failure behavior of

hybrid fiber reinforced thermoplastic composite containing defects were studied under
coupled thermal and mechanical loading. The system investigated was hybrid short Al/C
fibers reinforced polyaryletherketon composite, in which aluminum-fiber and carbon-fiber
were 30 wt.-% and 10 wt.-%, respectively. First fracture testing of specimens with different
artificial defects including notch and crack-like slot subjected to thermal and mechanical
loading was carried out using three-point short-beam bending test; the effect of changing
temperature on mechanical strength of the composite was then discussed; next failure mode
and corresponding morphology of broken samples were systematically characterized by SEM;
and finally, the macroscopic and microscopic failure behavior of the composite under coupled
thermal and applied loads were addressed based on experimental data. Results show that
high temperature is able to significantly affect macroscopic fracture mode, microscopic
failure behavior, and ruptured morphology of the composite along with its mechanical
properties.
KEYWORDS: Hybrid short-fiber, PAEK resin, high temperature, failure behavior, notch,
slot, aluminum-fiber, carbon-fiber.
INTRODUCTION
Hybrid composite consisting of two different fibers and single matrix is often advantageous
over a regular composite. With alternative choice of fiber combination, it offers improved
material properties such as thermal conductivity property, fracture toughness, and mechanical
strength due to mixture effect of fibers. Therefore hybrid composite has been received
considerably attention in the literature. A number of investigations have been made for
performance analysis and characterization of hybrid continuous-fiber reinforced composite,
for instance, selected papers like static problem [1], dynamic problem [2], fracture mechanics
[3], and engineering application [4]. Meanwhile people are also increasingly interested in
hybrid short-fiber reinforced composite as it can be used as a secondary structure to substitute
steel materials due to its high strength to weight ratio and convenience of molding
geometrical complex component. But most of works concentrated on the improvement of its
physical properties including thermal and electrical conductivity [5-8], and surface treatment
of fibers [9-10].
Like laminated composite for structural materials, however, hybrid short-fiber reinforced
composite is sometimes serviced in more complicated environment such as high temperature
and coupled thermal-mechanical load. Unfortunately, an available data about the influence
II - 301
of changing temperature on mechanical property and failure behavior of a given material is

very seldom. Furthermore the works done were only limited to consider the effect of single
short-fiber on material property regarding high temperature [11-13]. It is thus necessary to
perform laboratory tests to simulate actual operation environment of the hybrid composite and
further to evaluate its structural integrity under practical service condition.
This study aims to investigate the effect of changing temperature on strength performance and
failure behavior of hybrid short-fiber reinforced polymeric composite. The matrix material
chosen is Polyaryletherketon (PAEK); it is a kind of new high performance semi-crystalline
thermoplastic resin with glass transition temperature of around 170 C, which combines
excellent high-temperature property with retaining high elastic modules and high impact
strength [14]. Reinforcements used are aluminum-fiber and carbon-fiber, respectively; the
former is used for improving thermal conductivity property of the composite and the latter for
enhancing mechanical performance. Since this kind of hybrid composite is a typical
advanced composite material, it is able to be used as secondary component in some places
where local service temperature is sometimes higher than 160 C, such as aerospace motorcasing, heating-removal fan, and power pump. Hence the present study will provide
important basis for the practical use of PAEK composite as lightweight component
replacement. For this purpose, three-point short beam bending test were carried out for
specimens containing different artificial defects under varying temperature; then
microstructure and fracture mode of the specimens were viewed by optical microscopy, and
failure behavior and morphology of the broken samples were observed in detail with scanning
electron microscopy.
EXPERIMENTAL DETAILS
Materials
The matrix used is a commercial grade PAEK resin, called Ultrapek A1000, which is
provided by BASF; aluminum-fiber and carbon-fiber are supplied by Japanese company.
Carbon-fiber is a 3 mm long with a diameter of 7 m, and aluminum-fiber, called as AlMg2.5
metallic-alloy fiber, is a length of 3 mm and its cross-section looks like a quasi triangularshape with each side length of about 90 m. Based on the previous work on its thermal and
electrical property [5], the hybrid composite to be considered in this study will consist of 30
wt.-% of aluminum-fiber and 10 wt.-% of carbon-fiber. According to designation of ASTM
D790, rectangular specimens for three-point short-beam bending test, in which the
geometrical dimensions are 80 mm long, 10 mm wide and 4 mm thick, respectively, were
manufactured directly by injection molding. The fundamental properties of the relevant
materials are listed in Table 1.
Testing
Optical microscopy was utilized for the visualization of fiber alignment and orientation in the
specimen and polarized light was used to view the crystallization of matrix in the composite.
Fracture surface and morphology of broken samples were observed in a scanning electron
microscope (SEM) and samples were previously gold coated before SEM observation.
Differential scanning calorimetry (DCS) was performed for measuring glass transition
temperature of the composite, and the results relevant to thermal properties of the composite
with different contents of fiber measured by DSC are given in Table 2.
II - 302
Table 1: Fundamental properties of constituents consisting of hybrid composite

Materials
Tensile strength
(MPa)
Matrix A1000
104
AlMg2.5 fiber
210
Carbon fiber
3950
PAEK composite
106
Tensile modulus
(MPa)
4100
70000
2380000
10280
Density
( g / cm 3 )
1.3
2.68
1.77
1.60
Ultimate
elongation (%)
4
7.5
1.55
2.35
Table 2: Thermal properties for the hybrid composite with different fiber content
Sample No.
Matrix
Fiber content
(wt. %)
1
2
3
PAEK
PAEK
PAEK
0
19.13
40
Tg
Tc
Tm
169
169
170
339
336
327
393
393
380
Note: Tg denotes glass transition temperature, Tc initial melting temperature, and Tm melting
temperature; sample 2 only contains aluminum-fiber and sample 3 contains 30 wt.-%
aluminum-fiber and 10 wt.-% carbon-fiber.
Three different kinds of specimens containing different artificial defects were prepared for
bending test. The artificial defects with semi-notch and crack-like slot were machined into
the center on the edge of each specimen. The geometrical size and configuration of defects are
described in the following: the slot is approximately 0.4mm wide and the depth from the
edge is 2 mm long; the notch has a radius of R=2 mm and depth from the edge is also 2mm
long.
With temperature ranging from room temperature to 200 C, short beam three-point bending
test were conducted in Instron 4515 universal test machine with a temperature chamber, and
loading rate was chosen to be 5 mm/min. After testing, some samples were cut from section
of the fractured specimens for observing their fracture behavior and morphology.
RESULTS AND DISCUSSIONS

Microstructure details
In general, injection-molded hybrid short-fiber reinforced composite is fairly inhomogeneous
because most of the fibers are aligned in the direction of mold-filling. Fig.1, which illustrates
the fiber distribution and orientation along the transverse and longitudinal direction, shows
the fibers are preferentially oriented parallel to mold-filling direction. Aluminum-fibers are
tendency to the longitudinal direction, in which fiber length points to the flow direction of
mold-filling. However, in the direction of transverse section, there is very seldom for fiber
length to locate along this direction. As a result, the microstructure of the composite made
II - 303
through injection molding displays anisotropic property, which will also influence fracture
behavior of the material.
a) along transverse direction (200x)

b) along longitudinal direction (200x)
Figure 1: Profile of fiber distribution and orientation
Effect of changing temperature on composite strength
The relationship between bending stress and strain of the specimen without artificial defect
subjected to different temperature is depicted in Fig. 2. From this figure, we can see that at
room temperature, once the applied stress reaches yielding stress of the material, the
composite breaks immediately, with no visible plastic deformation. When temperature is
raised to 100 C, the relationship of material 's stress-strain still keeps linear basically without
the effect of temperature and the load-bearing capability of the composite reduces a little.
The value of maximum stress at this case is only smaller than that of its counterpart at room
temperature by a quarter and the magnitude of Young's modulus is almost same each other.
Figure 2: The effect of changing temperature on material strength (without defect)

For the case of 200 C, however, the composite softens greatly and displays nonlinear plastic
deformation feature for the high temperature; the whole stress-strain curve looks smooth, and
the amount of plastic deformation is considerably large. Consequently, the load-bearing
capability of the composite considerably reduces once temperature is above a certain value.
Higher temperature makes inherent energy of the material be raised so movement of
II - 304
molecular becomes more easier. When temperature is risen up to 200 C, namely higher than
glass transition temperature of the resin, the reduction in strength of the composite is
extremely large, irrespective of how much of the content of reinforcement fiber; the
maximum stress at this time is only half a value at room temperature. Furthermore, the
Young's modulus also decreases dramatically, which is only 20% of Young's modulus at
room temperature. Comparing to strength performance of the material, the influence of
temperature on the composite's stiffness is significantly larger than that of its influence on
composite's strength. Similar feature can also been seen from specimens with artificial
defect, which will further be discussed in an another separated paper [15]. Therefore, high
temperature not only influences strength of the composite but also changes the deformation
behavior of the composite. Moreover, when temperature is well above glass transition
temperature of resin, the effect of fiber orientation becomes less significant due to easy
movement of resin molecular. So high temperature will cause resin damage and shorten
residual life of the composite.
Effect of defect type on material property
Figure 3 shows the effect of different defect shape on the strength of the composite at room
temperature and the associated mechanical properties measured are listed in Table 3.
Table 3: The effect of defect type on composite strength at room temperature
Defect type
non-defect
notch
slot
Young's
modulus
(MPa)
5343
4617
4485
Maximum
load
(KN)
1.164
0.500
0.488
Maximum
Stress
(MPa)
173.20
75.58
71.94
Displacement
at break
(mm)
1.209
0.495
0.481
Strain at
break
(%)
4.48
1.83
1.79
Figure 3: The effect of defect shape on the strength of the composite (T=24C)
Because of stress concentration, like metal material, load-bearing capability of the composite
containing defect substantially reduces. In this figure, it can be seen that the maximum stress
specimen containing artificial defect can undertake is much smaller than that of specimen
II - 305
without defect. Usually, their stress values are less than a half stress value of specimen
without defect (see Table 3). In addition, the deformation curves of specimens with and
without defect are also quite different. Regarding specimen without defect, its stress increases
with the increase of strain until final fracture of specimen; whereas for the specimens with
defect, their stress first increases with increasing of strain and then the stress in the specimen
rapidly decreases due to resin yielding. This implies once resin material enters yielding,
specimen will immediately fail without any softening-strengthening phenomena. This is also
the same trend for both specimens with defect and specimen without defect. However, for the
specimens with notch and slot, their stress-strain curve and tendency look very similarly. The
reason is as follow: due to stress concentration, stress at the tip of the notch and slot will
reach yielding stress of resin and surrounding materials are thus blunted; but stresses a little
far away from the tip zone are still less than yielding stress of the resin, and the materials
nearby the tip zone are surrounded by a large amount of linear-elastic material. In this
instance, stress at the tip of notch and slot will be relaxed and redistributes by means of
plastic deformation of the resin to let tip displacement open largely than before. Moreover,
under the condition of high temperature, the material possesses larger ductility and permits
bigger displacement value at the tip of defect in comparison with that at room temperature.
As a result of these, local stress concentration is not critical to some extent to the magnitude
of width at the tip of defect for the tough resin like PAEK. Instead the elastic-plastic behavior
of resin material becomes important. Therefore if the resin is a tough material, the critical
magnitude of the width at defect tip is a little dependent on the extend of stress concentration
caused by defect type but be more dependent on their material's elastic-plastic behavior. This
maybe the main reason why the stress-strain curves of the specimens with notch and slot
resemble one another.
Macroscopic fracture mode
For three kinds of specimens, the macro fracture mode of the composite is dependent largely
upon temperature. Visual observation of the broken samples indicates that the hybrid
composite dissipates fracture energy primarily by a combination of saw-tooth deflection of
fracture surface, plastic deformation of specimen, as well as fiber's pull-out, debonding, and
rupture (see Fig. 4, and the detailed analysis will be addressed in the following section on
microscopic failure behavior).
Figure 4: The macroscopic fractography of specimens containing notch at room temperature

Fig. 5 shows fracture profile of three kinds of specimens with varying temperature. At room
temperature, the fractured specimen keeps fundamentally original shape and the fractured
section of three specimens almost separates in the way of brittle. It means that for brittle
fracture mode there is very little energy dissipation consumed by plastic deformation prior to
II - 306
final failure. With the increase of temperature, failure manner is changed from quasi-brittle
mode at room temperature to ductile mode at 100 C. At the time when temperature is equal
to 100 C, the composite possesses more ductile behavior due to increase of inner energy and
the whole specimen occurred a little plastic bending.
a) T=24C
b) T=100C
c) T=200C
Figure 5: Optical micrography of the fracture profiles for different temperature
Compared to its original shape, the geometrical shape of the fractured specimen happen
variation; most obviously, fracture surfaces for specimens with notch and slot still connect
each other after failure. That implies ductile fracture will consume a certain amount of
fracture energy before breaking through plastic deformation as specimens have undergone a
larger bend plastic deformation. As temperature is risen to 200 C, failed specimen displays
large plastic deflection. In this circumstance, fracture surface of all three specimens do not
break apart; furthermore, deformation process exhibits non-linear plastic flow due to viscousplastic stretching of matrix. This means the movement and slide of polymer molecular
become more easier for high temperature, and resin material tends to be viscous-plasticity
behavior. Thus the characteristic of resin damage at high temperature will be viscous-plastic
stretching instead of only yielding. For this reason, fracture surfaces of specimens at 200 C
keep together. All of these cases demonstrate the macro fracture feature of the composite has
a great good relation with temperature.
II - 307
Microscopic failure behavior

SEM photography of the fracture surface for hybrid composite with changing temperature are
shown in Figs 6-8.
In a sense of micro-level, the fundamental failure behavior of the hybrid composite comprises
fiber pull-out (Fig. 6(a)), fiber debonding (Fig. 7), fiber rupture (Fig. 8), and plastic
deformation of matrix (Fig. 6 (b) and (c). The most obvious characteristic for the failure of
the composite is that morphology of the matrix is extremely quite different with changing
temperature (Fig. 6). At room temperature, matrix morphology is basically flat with cusp and
exists a little shear band with brittle behavior; matrix morphology at 100 C gets rough and
appears many ductile dimples, illustrating deformation is in the manner of elastic-plasticity;
while T=200 C, fractography of matrix looks like a shape of mop filling with fibrils,
exhibiting viscous-plastic flow deformation is caused by excessive high temperature.
a) T=24C
b) T = 100C
Fig. 6: The effect of changing temperature on

morphology of matrix
c) T=200C
Moreover, the interface of matrix and fiber becomes severe due to difference of their thermal
expansion coefficients. However, there is no obvious feature to well characterize the effect
of defect type on the morphology owing to insensitivity of tough matrix to defect type. The
interface between carbon-fiber and matrix is dependent to great extent on their position and
place. Two cases for bonding surface of carbon fibers at room temperature are depicted in
Fig. 6 (a) and Fig. 7(a). Fig. 6(a) demonstrates excellent interface of carbon-fibers and Fig.
7(a) poor situation for the interface of carbon-fibers. Why is the interfacial surface of carbonfiber fully different for the same process condition? The crystallization and extend of resin
II - 308
PAEK on the surface of carbon-fiber account for this result. With reference to Fig. 6(a), there
exist a very thin layer of resin film adhere to the surface of the carbon-fiber. This thin layer is
resulted directly from crystallization of PAEK resin shown in Fig. 9, in which the interface
between resin and carbon-fiber bonds strongly. By contrast, there is very clean on the surface
of carbon-fiber shown in Fig. 7, which indicates there is no any crystallization thin layer
adhesive to the surface of carbon-fiber. This event is, in fact, closely related to cooling way
of the extruded reinforced pellets and molding process of the specimen when molded by
injection machinery. At a time when hybrid-fiber reinforced PAEK pellets are extruded, they
need to be cooled fast to fix their shape. If the resins are located inside the pellet, they are
gradually cooled and thus have relatively more time to crystallize on the surface of the fibers;
but if the resins situate on or near the surface of the pellet, they are immediately cooled and
thus create brittle amorphous structure without any crystallization.
a) T = 24C
b) T=200C
Figure 7: Carbon-fiber pull-out and debonding
50 m
Fig.
8: The morphology
aluminum-fiber
of
ruptured Fig. 9: The crystallization photo of matrix on

the surface of aluminum- and carbonfiber
Similarly, when specimen is manufactured from the pellets by injection machine, cooling of
molding tool also affects the crystallization of resin. Resins lying on the surface of specimen
are usually not easy to crystallize for the fast cooling of tool, and resins near the center of the
specimen are easy to be crystallized due to longer period of cooling so that they form
crystallization morphology. It is these reasons that result in completely different interface
II - 309
between carbon-fiber and matrix. Another interesting thing about interface of the carbonfiber is worthwhile to noting there is no fracture phenomena of the fiber due to its too short.
Comparing with carbon-fiber, however, the interfacial bonding between aluminum-fiber and
matrix is always relatively poor. As indicated in Table 1, elongation rate of aluminum-fiber is
much larger than that of carbon-fiber, but larger than that of resin. When applied stress is
above yielding stress of resin, the tensile stress will automatically transfer to aluminum-fiber,
making it be elongated longitudinally and contracted transversely. With the increase of tensile
stress, the cross-section of the aluminum-fiber is shrunken greatly until its final rupture ( see
Fig. 8), resulting in the surface of aluminum-fiber to be separated from matrix. While this
pattern is equivalent to pull-out of carbon-fiber, their ruptured section is not a smooth flat,
but looks like steep rock. Besides, for the reason of high temperature, the number of pull-out
of carbon-fiber at 200 C is more than that at room temperature (see Fig. 7); furthermore,
carbon-fibers seem to be moved obviously due to large viscous-plasticity stretching of matrix.
These further illustrate temperature is a crucial factor to influencing both deformation
morphology of matrix and fiber interface.
CONCLUSIONS
Based on the experimental results of PAEK composite with varying temperature and failure
analysis for broken samples, the following conclusions may be drawn:
(1) Micro failure behavior of the composite is closely related to temperature. Matrix
morphology at a room temperature is cusp flat with a small brittle shear bands, at 100 C
matrix surface gets rough and exists a lot of ductile dimples, and in the case of 200 C matrix
deformation looks like a mop consisting of fibrils with a characteristic of viscous-plastic flow.
(2) The macro failure behavior of the composite is also affected by high temperature. With
the increase of temperature, failure behavior is changed from quasi-brittle mode at room
temperature to ductile mode at 100 C; when temperature equals 200 C, fracture mode
becomes large plastic deflection.
(3) The interface situation between fiber and matrix is dependent mainly upon crystalline
degree of resin and service temperature. With high crystallization, the interface bonds
strongly, and vice versa. Furthermore, the interface of resin and fibers becomes poor with the
temperature; high temperature tends to weaken the fiber/matrix interfacial bond because of
the difference in their thermal expansion coefficients.
(4) The temperature effect on strength performance of the composite is very significant. With
the increase of temperature, the strength reduction of the composite is extremely large and the
strength at 200 C is only half a value at room temperature. However, as PAEK resin is a
kind of tough material, the type of defect only influences magnitude of local stress
concentration, but not influences failure behavior of the material.
REFERENCES
1.
Gruber, M. B. and Chou T., Polymer Composites, 4, 265-269, 1983
2.
Jang, B Z. et al, Comp. Sci. & Tech., 34, 305-335, 1989
Aronhime, J. et al, Comp. Sci. & Tech., 43, 105-116, 1992
II - 310
Johnson, W. S., Li E. and Miller J. L., in Proceedings of ICCM-10, Vol.V, 227-234, 1995
5.
Schmitt-Thomas, Kh.G. and Malke R., in Proceedings of ICCM-10,Vol. IV, 449-456,

1995
6.
Dunn, M. L. et al, J. Comp. Mater., 27 (15), 1493-1512, 1993
7.
O'Rourke, J. M., Bushby R. S. and Scott V. D., Comp. Sci. & Tech., 56, 957-965, 1996
8.
Takei, T. H., Hatta H. and Taya M., Mater. Sci. & Eng. A, 131, 145-152,1993
9.
Maldas, D. and Kokta B. V., J. Reinforced Plastics and Composites, 11, 1093-1103,
1992
10. Maldas, D. and Kokta B. V. , J. Comp. Mater., 25, 375-390, 1991

11. Schultz, J. M. and Friedrich K. , J. Mater. Sci., 19, 2246-2258, 1984
12. Yau, S. S. and Chou T. W., in ASTM STP 1003, Vol.2, pp45-54, 1989
13. Sarasua, J, Remiro P. M. and Pouyet J., Polymer Composite, 17 (3), 468-477,1996
14. BASF, User's manual for Ultrapek, Germany, 1978
15. Influence of temperature and defect on failure mechanism and properties of PAEK/hybrid
short-fiber composite (in preparation, to be submitted to Composite Science &
Technology)
II - 311
DEBONDING AND FRACTURE MECHANISMS OF

PARTICULATE FILLED POLYPROPYLENE
I.L.Dubnikova, D.K.Muravin and V.G.Oshmyan
Polymer and Composite Department, Semenov Institute of Chemical Physics,
Russian Academy of Sciences,
4 Kosygin street, 117977 Moscow, Russia
SUMMARY: Regularities of plastic deformation and fracture of particulate-filled

polypropylene are analyzed. Experimental data on elongation at break b of PP, filled by
particles of different fraction and size are presented. Debonding and plastic flow
micromechanisms are studied. The effect of micromechanisms on composite macroscopic
behavior is analyzed. Two concentration regions of different debonding and flow
micromechanisms were found. At low filler fraction (< 15 - 20 vol. %) particles debond
independently at the initial stages of drawing and a microhomogeneous flow mechanism takes
place. In this region deformation occurs with necking. Above a critical filler content,
debonding proceeds in a correlated way with the formation of craze-like deformation zones
transverse to the stretching axes. Concentration of craze-like zones is determined by filler
fraction, particle size and the level of interfacial interaction. Whether microhomogeneous or
craze-like mechanisms occur essentially determines the regularities of further deformation
and fracture: localized or macrohomogeneous flow, ductile or quasi-brittle fracture, the drop
or the rise of ultimate strains with increase of particle diameters.
KEYWORDS: particulate filled polypropylene, particle size, mechanical properties,
debonding, plastic flow, craze, fracture mechanism.
INTRODUCTION
It is well established both by measurements and theoretical approaches that a level of
interfacial interaction plays an important role in particulate filled composites deformation [15]. It is shown, in particular, that weakening of adhesion causes a diminution of composite
stiffness [1,2,4] and yield stress [1,3,5]. Yield strain drops sharply with filler fraction in a case
of well-bonded particles and slightly diminishes in an opposite case [3,5].
A simplified model of composite fracture was proposed in [5] in framework of deformation
theory of plasticity. It is based on the numerical analysis of stress-strain distribution (SSD)
and critical value criterion of maximum shear strain. The model predicts a sharp drop of
ultimate elongation with an increase in filler fraction in a case of a perfect interfacial bond
and a gradual decrease of the last in a case of poor adhesion. The predictions are in qualitative
agreement with experimental data, however the model does not account two important
factors.
An effect of the particle size is the first one. Debonding in the particulate filled polymers
leads to a formation of the pores on initial stages of deformation. Further growth and
confluence of the pores seems to be an adequate mechanism of ductile materials (PP, PE)
II - 312
rupture [6,7]. In a case of the brittle material a dramatic drop in the strength with an increase
in a defect size forms a basic point of a linear fracture mechanics. Whenever pores are the
defects responsible for the fracture of high plastic materials, their size likely affects on
fracture parameters in analogous manner. From the other side not only absolute size, but also
concentration of the defects should determine fracture parameters. This concentration in turn
is determined by a level of interfacial interaction. As was shown in [8,9], the particle
diameter, d , is responsible for this level, namely, the debonding stress, d, decreases with an
increase in d . In accordance with a model of [5], the less is a portion of debonded particles,
the less should be ultimate strain.
A second simplification of a model [5] is in a consideration of uniform SSD. High plastic
polymers and composite on their basis mostly deform in heterogeneous manner with
formation and growth of the neck. In this case a sharp reduction in composite deformability
(ductile-brittle transition) clearly should coincide with a disappearance of a strain hardening,
i.e. equality between a strength b and a neck stress n [10].
In the present study an effect of particle content and size on macroscopic behavior of
particulate-filled PP is analyzed. An establishment of the correlation between debonding
processes at the initial stage of deformation, plastic flow mechanisms and composite fracture
regularities is the goal of this investigation.
EXPERIMENTAL
Isotactic PP was used as a matrix. Polymer molecular weight characteristics, measured by a
GPC method, were: Mw=6.3x105, Mw/Mn=3.5. Aluminum hydroxide Al(OH)3 was used as
the filler. Four fractions of mean, d , minimum, d min , and maximum, d max , diameters were
used: 1 (0.5-1.5), 2.5 (1.5-5), 8 (4-16) and 25 (5-50) m.
Composites were prepared in a Brabender mixing chamber at T=190C, duration of mixing
was 10 min. Calcium stearate (2 wt.% of the filler fraction) was added to avoid filler
aggregation. Polymer degradation was prevented by Topanol (0.3 wt.%) and
dilauriltiodipropionate (0.5 wt.%). Filler content was in the range 0-70 wt.% (0-47 vol.%). A
filler surface treatment by octametilcyclotetrasyloxane was used into the series of high filler
content to reduce adhesive interaction.
The specimens for mechanical testing were prepared by the molding at T=190C and the
pressure p=10 MPa with subsequent cooling under the pressure. The drawing of the blade
shaped samples was performed on the Instron-1122 at the room temperature and strain rate
0.67 min--1.
The structure of the deformed composite samples was observed by scanning electron
microscopy (SEM) with the use of JSM-35C.
II - 313

The effect of filler fraction and size on ultimate elongations of particulate-filled PP.
The dependencies of the elongations at break b on the filler volume fraction and particle size
are shown in Fig. 1. A series of regularities is seen from these data.
1. Ultimate strains decrease with the rise of filler content. The increase of b in the case of
large particles ( d = 25 m ) in the range of filler fraction 15%30 vol.% is an interesting
exclusion from this regularity. It will be discussed below. Here we note that only two
publications (Fu, Wang and Shen [11,12]) are known to us as other experimental evidence
on the rise of polyolefines fracture properties with an increase of rigid filler content.
Authors found brittle-ductile impact strength transition for the HDPE-CaCO3 composites
in the case of filler surface modification by phosphate.
2. The ductile-brittle transition is rather sharp in the case of small particles ( d = 1, 2.5 m )
and occurs at br 25 vol.%. The increase of d leads to the more smooth concentration
dependencies and to the shift to the right of br (br36 vol.% for d = 8 m ).
3. An increase of the particle size leads to the drop of the elongation at break at small filler
fraction (<15 vol.%) and to the rise of the b above 30 vol.%. Intermediate region is
characterized by the extreme-type size dependencies (Fig. 2).
Three regions of volume fraction are also characterized by different modes of macroscopic
deformation.
b,%
1 m
600
2.5 m
400
8 m
25 m
200
0
0
10
20
30
40
50
,vol.%
Fig.1. The elongation at break versus filler volume fraction

for different particle size
Below =15 vol.% the drawing is accompanied by formation and propagation of a neck.
Strain-hardening is briefly recognized in the case of small particles. This circumstance
II - 314
provides propagation of the neck all through the sample. However, the increase of the particle
size causes the drop in strain-hardening. It results in the fracture of samples, filled by 14
vol.% of large inclusions ( d = 25 m ), at the initial or intermediate stage of the neck
propagation.
In the intermediate region (15 vol.%30 vol.%) a tendency to unstable neck propagation is
observed. Formation of two or three necks is seen in the case of small particles. The increase
of d leads to the disappearance of the neck and the tendency to homogeneous flow.
Above 30 vol.%, filled by small particles ( d = 1, 2.5 m ) composites break in quasi-brittle
manner with formation of a narrow neck. The increase of d provides macrohomogeneous
flow, but up to the certain threshold, dependent of particle size. For the composites of
d = 8 m a tendency to the formation of a narrow neck and quasi-brittle fracture is revealed at
36 vol.%.
The surface treatment of small particles by octametilcyclotetrasyloxane rises elongation at
break of composites filled by 28 vol.% and 36 vol.% (Fig. 2, dashed lines) [13]. This
procedure causes the transition from localized flow and quasi-brittle failure to
macrohomogeneous deformation and ductile fracture in the case of d = 2.5 m .
Experimental data on filler fraction and size dependencies of ultimate strains and
microscopical observations of samples deformed demonstrate the change of debonding and
flow micromechanisms at cz between 15 and 20 vol.%.
b, %
600
500
=15%
400
300
=28%
200
100
=36%
0
0
2 5
10
15
20
25
d ,m
F ig .2 . T h e e lo n g a tio n s a t b re a k ve rsu s p a rticle
d ia m e te r. D a sh e d lin e s p o in t to su rfa c e tre atm e n t.
II - 315
The transition of debonding and plastic flow micromechanisms with the increase of filler
fraction.
Exfoliation along the surface of a single inclusion obviously results in the change of SSD in
its close vicinity. So, it is natural to expect the occurrence of different debonding
micromechanisms depending on particle content.
Uncorrelated debonding and microhomogeneous flow mechanism.
The change of SSD is not noticeable far from the source of this change. So, in the case of the
large interparticle distance, or, equivalently, at a small filler fraction debondings of different
particles should proceed independently. Uncorrelated debonding mechanism (Fig.3a) causes
microhomogeneous flow (see Fig.3b, which corresponds to the region of the developed neck).
A picture of this type was observed by SEM for the composite samples with <15
vol.% [14]. We underline the following features.
1.
2.
Exfoliation occurs along the surfaces of all the particles.

Longitudinal elongation of the every pore formed is equal, with a good precision, to the
draw ratio mean elongation of the neck region (microhomogeneous deformation).
(a)
(b)
(c)
(d)
Fig.3: Scheme representation of uncorrelated (a), correlated (c) debonding micromechanisms

and sequent microhomogeneous (b), and craze-like (d) plastic flow
Correlated debonding and craze-like flow micromechanism.

Quite another situation takes place in the case >1520 vol.%. Debonding of the single
inclusion considerably changes SSD of a neighbor because of the small interparticle distance.
The maximum stress intensity in the uniaxially stretched composites is reached in the area of
II - 316
a pore equator. It should result in correlated debondings of neighbor particles and the
formation of the craze-like microporous zones, extended in the direction transverse to the
stretching axes. Correlated debonding (Fig.3c) causes the craze-like flow mechanism (Fig.3d).
Such picture was observed by SEM for the composite samples with >20 vol.% [14].
Contrary to the previous case, debonding is characterized by the following features. A
noticeable portion of the particles outside the craze-like zones remain bonded.
Because of the sharp difference in the compliances, further deformation mainly concentrates
inside the craze-like microporous zones, whenever the other regions remain weakly stretched.
The first regularity may be clarified by the following estimation of the screen effect
efficiency, which is in the drop of the stress in the area of craze banks, caused by formation of
micropores inside. Let us denote the debonding stress (the stress value sufficient for the
exfoliation along the surface of the given particle) by d and the stress of polymer bulkhead
inside the crack by bh. It is natural to suggest that the pore fraction inside the craze-like
zones is equal to the filler volume content and to use the minimum cross section area
estimation for the calculation of the stress, cz, tightening the craze-like zone banks:
(1)
cz = bh 1 2 3 .
Correlation between d and cz
cz d
(2)
determines the conditions for the further debonding of the particles in the vicinities of the
craze banks. It is seen that even in the case of the weak adhesive strength (low d value,
which is analyzed in the present paper) new debondings formation in the process of further
stretching will be forbidden at sufficiently high filler content.
As noted, the increase of the particle size results in a drop of the debonding stress d [8,9] and
thereby in the rise of pore concentration. The surface treatment by the antiadhesive
component is another route for a debonding stress decrease [9,13].
The stage of secondary debondings depends upon the state of bulkheads in the primary crazes.
bh and, hence, cz can be raised in the process of further drawing, if the bulkheads deform in
the elastic or strain hardening region. bh is a fixed value, equal to the polymer matrix yield
stress Y, if the bulkheads are in the stage of plastic flow. So, if the craze-like flow
mechanism is realized, debonding may occur, not only at the beginning of drawing, but at the
developed or even final stages. In the case of the sample, filled by 36 vol.% of small
( d = 2.5 m ) particles only few zones are seen outside the narrow neck. Inside the neck their
concentration is much higher because of the strain-hardening stage of the bulkheads in
deformation zones.
Macroscopic consequences of microhomogeneous or craze-like flow.
Localized or macrohomogeneous flow.
Unfilled PP is characterized by a well-emphasized extreme-type engineering diagram and
localized plastic flow, but the true stress increases with draw ratio. An extreme type of the
engineering diagram is caused by a conservation of volume and, therefore, a drop of the crosssection area on draw.
The increase of pore concentration in the filled PP leads to a more noticeable growth of
II - 317
material volume on deformation and, hence, to a less drop of the cross-section area. In the
framework of a microhomogeneous mechanism (15 vol.%), debonding occurs for every
particle at the initial stage of deformation. However, because of the small filler fraction, the
volume of the pores formed is not sufficient for the qualitative change of engineering
diagrams and the samples deform with necking for the every particle size studied.
The transition to macrohomogeneous flow would take place at the further increase of filler
fraction if the microhomogeneous mechanism remains valid. But the change of
microhomogeneous flow mechanism by the craze-like process complicates the situation. If d
is low, debonding in accordance with (1), (2) is not forbidden and the volume of the pores
formed becomes sufficient for the change of the engineering diagram from the extreme to
increasing ones. This is the reason for macrohomogeneous deformation of high filled
composites (> 20 vol.%) in the cases of large ( d = 8, 25 m ) inclusions or small ones
( d = 2.5 m ), treated by anti-adhesive component. In the case of untreated small particles the
criterion (2) of debonding is not valid, because of the increase of the debonding stress. So,
micropore concentration remains low and the samples deform with the necking.
Ductile and quasi-brittle fracture.
The fracture properties of particulate-filled ductile polymers are determined by the growth
and confluence of micropores, formed in the process of debonding. So, debonding and the
consequent flow mechanism are extremely important.
In the framework of microhomogeneous flow mechanism (< 15 vol.%) a filler fraction can
be sufficiently low for the complete adhesive failure at the initial stages of deformation
(composite yield stress Y decreases with , but can remain higher than d). The composites
characterized by ductile fracture. The drop of b with the increase of particle size (Fig. 1, the
initial portions of the curves, Fig. 2, =15 vol.%) is governed by the geometric factor of
rupture, which is in the aggravation of fracture parameters with the growth of the size of a
defect, responsible for fracture.
In the framework of the craze-like mechanism, complete debonding is not obtained because of
the screen effect discussed. In this case the main part of deformation is concentrated in the
narrow microporous zones, so ultimate strains depend upon the specific amount of the last by
means of at least two reasons. The first is the increase of the elongation due to the increase in
the concentration of the deformed regions. The second reason is the rise of the strain at break
of polymer bulkheads inside the craze, as a result of the decrease of the local strain rate, at
fixed value of the last, with an increase of the crazes density. Therefore, geometric factor of
fracture contradicts with the debonding one, which is in the change of degree of debonding
processes completion. The influence of this contradiction on b is illustrated by the
intermediate and final portions of the curves in Fig. 1 and by the plots of Fig.2, corresponded
to = 28 and 36 vol.%.
dVd
d d is a volume fraction of the particles, debonded in

the stress interval between and +d), are schematically represented at Fig.4. The increase
of filler fraction does not noticeably change the debonding stress [9], i.e., the abscissa position
of the curves 1-3 in Fig. 4a, but gives a proportional increase of the ordinate. The porosity of
the sample should increase due to this factor. However, the stress cz in the vicinities of the
Debonding stress distributions, (
II - 318
craze banks drops with in accordance with correlation (1). Competition between these two
factors admits the rise or drop of craze-like zones density and, therefore, the existence of
increasing or decreasing portions of ultimate strains dependencies upon filler fraction. This
phenomenon is the reason of the rise of b with in the region 15 vol.%30 vol.% in the
case d = 25 m and the drop of the last in the case of small particles (Fig. 1). Dramatic
diminution of the microporous zones density provides the transition from ductile to quasibrittle fracture. The corresponding filler fraction, br, depends upon the debonding stress
(which is the decreasing function of d ) and should be shifted to the higher with the
increase of particle size. This regularity is experimentally confirmed by the increase of br
from 25 vol.% for d = 1 and 2. m to br36 vol.% for d = 8 m (Fig. 1).
,
5
Portion of debonded
particles
(a)
1
cz
3
2
1<2<3
0
0
Stress
Portion of debonded
particles
d1>d2>d3
(b)
0
0
Stress
Fig.4: Scheme representation of the debonding stress distributions: (a) - for different
filler fractions (1<2<3) and fixed particle size; (b) - for the various particle diameters
( d1 > d 2 > d 3 ) and fixed filler fraction. Vertical lines denote the stresses cz in the vicinities of
the craze banks (see formula (1))
II - 319
Contrary, at the fixed filler fraction the cz value remains unchanged, whenever the change of
the particle size results in the horizontal shift of the debonding stress distribution (Fig4b). The
surface treatment by the anti-adhesive component facilitates debonding [9] and is an
alternative way to achieve such a shift. So, at the high filler content, debonding is most
important and causes an increase of b with the increase of inclusion size (Fig. 2, = 36
vol.%, solid line). Competition between adhesive and geometric factors of fracture (the last
remains valid in framework of the craze-like mechanism) is the reason of the extreme size
dependence of b at =28 vol.% (Fig. 2, solid line) and at high filler content in the case of
anti-adhesive coating of the particles (Fig. 2, = 28 and 36 vol.%, dashed lines).
CONCLUSIONS
1. Fracture properties of particulate-filled PP have been studied in the wide range of filler
fraction and size.
2. Two debonding micromechanisms and the transition between them at cz 15% were
discovered. The first is uncorrelated debonding at the initial stages of deformation and
occurs at low (< 15 vol.%). The second micromechanisms is correlated debonding of
neighbor particles inside the narrow zones, transverse to the loading axis.
3. Two different material flow mechanism - microhomogeneous and the craze-like one ( >
15 vol.%) - have been found. In the case of uncorrelated debonding the
microhomogeneous deformations form in the neighborhood of the pores. The correlated
debondings of neighbor particles results in the craze-like flow mechanism. In this case
plastic deformation is mainly concentrated inside craze-like regions.
4. An increase of micropore fraction is the reason for the transition from localized drawing
(with necking) to macrohomogeneous draw.
5. Two factors of composite failure are formulated: a geometric one, which is in the drop of
fracture properties with an increase of particle (pore) size; and a debonding factor, which
is in the formation of the sources of fracture in the processes of phase exfoliation and in
the change of degree of debonding processes completion dependent of particle size.
6. In the framework of microhomogeneous flow mechanism, geometric factor governs the
fracture, and ultimate elongations decrease with an increase of filler content and particle
size.
7. In the framework of craze-like mechanism, the debonding factor of fracture plays an
important role. The increase of debonding stress with the decrease of the particle size
results in the drop of the craze-like zone concentration and in a decrease of br of the
ductile-brittle transition. Competition between geometric and debonding factors causes
extreme dependence of the elongation at break upon d in the intermediate region 15
vol.%<<30 vol.%. At >30 vol.% ultimate strains increase with the rise of d , due the
increase of deformation zone concentration.
REFERENCES
1.
Nielsen, L.E., Mechanical Properties of Polymer Blends and Composites, Marcel

Dekker, Inc., New York, 1974.
2.
Manson, J.A. and Sperling, L.H., Polymer Blends and Composites, Plenum Press, New
York, 1976.
II - 320
3.
Pukanszky B., Fekete E. and Tudos F., Surface Tension and Mechanical Properties in
Polyolefine Composites, Makromolecular Chemie, Macromolecule Symposium, Vol.
28, 1989, pp. 165-186.
4.
Lyapunova, M.A., Knunyatz, N.N., Manevitch, L.I., Oshmyan, V.G. and Shaulov
A.Yu., Simulation of the Effect of Interfacial Bond on Elastic Properties of
Particulate Composite, Mekhanika komposithich materialov, No. 2, 1986, pp. 231234.
5.
Gorbunova, N.V., Knunyantz, N.N., Oshmyan,. Manevitch, L.I., Oshmyan, V.G. and
Topolkaraev, V.A., On the Effect of Interfacial Bond on Elastic-Plastic Properties of
Particulate Filled Composite, Mekhanika komposithich materialov, No. 2, 1990, pp.
336-339.
6.
Topolkaraev, V.A., Tovmasyan, Yu.M., Dubnikova, I.L, Petrosyan, A.I. Meschkova,

I.N., Berlin, Al.Al., Gomza, Yu.P., Schilov, V.V, The Effect of Particle Size on the
Structure and Deformation Behaviour of Particulate Filled HDPE with Different
Molecular Mass, Mekhanika komposithich materialov, No.4, 1987, pp.616-622.
7.
Dubnikova, I.L, Topolkaraev, V.A. Paramzina, T.V., Gorokhova, E.V., Dyachkovskii,

F.S., Plastic Properties of Particulate Filled PP, Vysokomolekularnye Soedineniya,
Ser. A, Vol. 32, No 4, 1990, pp.841-847.
8.
Zhuk, A.V., Knunjantz, N.N., Oshmjan, V.G., Topolkaraev, V.A. and Berlin, A.A.,
Regularities of Adhesive Failure in Particulate Filled Polymers, Makromolecular
Chemie, Macromolecule Symposium, Vol. 44, 1991, pp. 295-299.
9.
Zhuk, A.V., Knunyants, N.N., Oshmyan, V.G., V.A.Topolkaraev and A.A.Berlin,

Debonding microprocesses and interface strength in particle filled polymer materials
Journal of Material Science, Vol. 28, 1993, pp. 4595-4606.
10.
Bazhenov, S.A., The Effect of Particles on FailureModes of Filled Polymers,

Polymer Engineering and Science, Vol. 35, No. 10, 1995, pp. 813-822.
11.
]Fu, Q., Wang, G. and Shen, J., Polyethylene Toughened by CaCO3 Particles: BrittleDuctile Transition of CaCO3 Toughened HDPE, Journal of Applied Polymer Science,
Vol. 49, No. 4, 1993, pp. 673-682.
12.
Fu, Q., Wang, G. and Shen, J., Effect of Morphology on Brittle-Ductile Transition on
HDPE / CaCO3 Blends, Journal of Applied Polymer Science, Vol. 49, No. 11, 1993,
pp. 1995-2002.
13.
Dubnikova, I.L, Gorokhova, E.V., Gorenberg, A.Ya., Topolkaraev, V.A.,The Effect

of Small Additivites of Octametylcyclotetrasiloxane on Deformation Behaviour of the
Particulate Filled Polyolefines, Vysokomolekularnye Soedineniya, Ser.A, Vol. 37, No
9, 1995, pp.1535-1544.
14.
Dubnikova, I.L., Oshmian, V.G., Gorenberg, A.Ya, Mechanisms of Particulate Filled

Polypropylene Finite Plastic Deformation and Fracture, Journal of Material Science,
Vol.31, 1996.
II - 321
EFFECTS OF PORES AND VOIDS ON THE

INTERLAMINAR DELAMINATION TOUGHNESS OF
A CARBON/EPOXY COMPOSITE
Leif E. Asp and Fredrik Brandt
The Aeronautical Research Institute of Sweden
P.O. Box 11021, SE-161 11 Bromma, Sweden
SUMMARY: Presence of pores and voids is known to impair mechanical properties such as
strengths and moduli of composite materials. At least, for aeronautical applications it is
important to understand how defects such as voids affect the interlaminar delamination
toughness of a composite laminate. Therefore an investigation on the influence of voids on the
interlaminar delamination toughness of a carbon fiber/epoxy composite in pure modes I and II
as well as mixed mode has been performed. Voids are found to have a deleterious or no effect
on the critical strain energy release rate at crack growth initiation. However, at propagation in
pure mode I and mixed mode load-cases presence of voids causes a R-curve behavior. The
increase in Gc, which is strong in mode I and moderate in mixed mode is inherent to changes
in failure mechanisms. In pure mode I, ply splits are observed to bridge the crack as the crack
locally leaps between plies at propagation. In mixed mode (GII/G0.5), the crack propagates
in one interface only, multilayer cracking in the crack tip region results in increased GC at
crack propagation. From the observed R-curve behavior it is concluded that pores and voids
do not impair the delamination toughness of the HTA/6376C composite in static loading.
KEYWORDS: Voids, pores, manufacturing defects, interlaminar delamination toughness,

delamination tests.
INTRODUCTION
Poor manufacturing of laminates may cause appearance of voids and pores in and in-between
plies. To date, laminates containing detectable voids are discarded. Cost will be significantly
reduced if laminates containing voids can be accepted in some applications. For this reason it
is important to understand the influence of voids on the mechanical properties of composite
laminates. In particular, for aeronautical applications, the effect of voids on interlaminar
delamination toughness is of interest. This is motivated by interlaminar delamination growth
being one of the major failure modes in composite structures, such as aircraft wing skins.
Interlaminar delamination or fracture may be defined as debonding of plies due to the
interlaminar stresses present in the laminate.
It is well known that voids degrade mechanical properties, e.g. tensile and shear strengths, of
composites. Over the years a number of studies have been presented on this matter.
Comprehensive reviews on the work in this area have been given by Judd and Wright [1], and
more recently by Cantwell and Morton [2]. To date, work has been focused on the effects of
void content on strengths and moduli of composites. Especially, the interlaminar shear
strength (ILSS) has been of concern [1-4]. Results presented on the effects of voids and pores
II - 322
on mechanical properties are found to differ substantially between studies. For instance, for
carbon fiber/epoxy composites, the ILSS has been reported to decrease between 1 and 21%
for the first one percent voids [1]. This inconsistency is coherent with the difficulty to
measure true void content reported by Ghiorse [5]. In spite of the efforts made, to the authors
knowledge, no results on the influence of voids on the interlaminar delamination toughness of
CF/EP composites have been reported in the literature.
The objective of this study is to investigate the influence of voids and pores on the
interlaminar delamination toughness of a HTA/6376C composite laminate in pure modes I
and II as well as mixed mode loading.
EXPERIMENTAL METHODS
Test methods
dI
dc
c
L
Fig. 1: FFAs mixed mode bending rig

To study the effects of pores and voids on delamination toughness of HTA/6376C composite
laminates, several test methods were utilized. Tests in modes I and II as well as mixed mode
were performed. For Mode I, the common DCB (Double Cantilever Beam) test was used. In
mode II, specimens were tested in three-point bending in the ENF (End Notch Flexural) test.
Finally, in the mixed mode tests, the specimens were tested under combined loads in the
MMB (Mixed Mode Bending) test. Seven specimens were tested in each test method. All tests
were performed in a modified version of the mixed mode rig originally suggested by Reeder
and Crews [6,7]. In this study only one mixed mode ratio was investigated together with the
pure mode tests. According to the analysis by Reeder and Crews [6], an expected mode ratio,
GII /G, of 0.5 was chosen by setting c=41 mm, where L=50 mm, see Fig. 1.
Three-point bending tests were performed to determine the flexural modulus of the uncracked
laminates at room temperature. In these experiments the distance between support rollers was
65 mm. All experiments were performed at ambient conditions.
II - 323
Specimens
All specimens were fabricated from HTA/6376C carbon/epoxy prepreg supplied by Ciba
Geigy. The laminates were cured according to the suppliers recommendations (at 6 bars
pressure). However, to introduce voids in one of the laminates the applied vacuum was
tampered with during cure. This was done in the following manner: General consolidation of
the laminate was achieved as the vacuum-bag was evacuated prior to curing. After evacuation
the vacuum pump was turned off. The laminate was heated at a rate of 2C/min to a
temperature of 175C. As the laminate was heated the pressure inside the bag rose to 6 bar.
When the laminate had been heated for one hour the bag was ventilated at 1 atmosphere
pressure. Ventilation was repeated as the temperature reached the plateau value of 175C.
After one hour at 175C the laminate was left at a pressure of 1 atmosphere.
The laminates included a 35 mm long artificial delamination as a starter crack. The artificial
delamination consisted of a 7.5 m thick Upilex 7.5S polyimide film from UBE. The thin film
used reduces the resin pocket created at the film edge and hence its influence on the critical
strain energy release rate [8]. After curing, the laminates were C-scanned for flaws, such as
delaminations, and cut into specimens in a diamond wheel cutter. Prior to testing all
specimens were dried in an oven at 70C for two weeks. The dimensions of the rectangularly
shaped specimens were; width 20 mm, length 150 mm and nominal thickness 3.3 mm. The
nominal thickness of the void-free laminate was 3.1 mm. The specimen lay-up was
[012//(5/04)s], where the sign // refers to the interface plane, i.e. the plane of the artificial
delamination. The off-axis angle of 5 was applied to reduce fiber bridging at delamination
growth. In a study by Olsson et al. [8] it was concluded that an off-axis layer at the interface
suppresses fiber bridging in DCB and MMB tests. Also, Olsson et al. [8] showed that the offaxis layers in the composite have a negligible effect on the local G distribution through the
specimen width. To release any adhesion between the polyimide film and the composite the
specimens were pre-loaded prior to testing.
Three-point bend specimens were cut to the dimensions; width 5 mm and length 75 mm with
a nominal thickness of 3.3 mm.
Measurements
The delamination beam DCB, ENF and MMB tests were performed in a screw-driven Zwick
tension load machine at a constant displacement rate of 0.5 mm/min. In the DCB tests a 250
N load cell was used, while in the ENF and MMB tests a 5 kN load cell was utilized. The
specimens were connected to the test rig by hinges with automatic couplings, for details see
ref. 7.
Crack growth in the DCB, MMB and ENF specimens was measured manually with a
traveling microscope. The microscope was equipped with an instrumented dial gauge
providing continuos reading of the observed crack tip position. The dial gauge had a
maximum stroke length of 20 mm. In the DCB tests, use of a 20 mm extension made it
possible to measure a total crack growth length of 40 mm. Mode I opening displacement in
the DCB test was measured with a string/wheel potentiometer drawn by a wire attached to the
load point of the specimen. In the ENF and three point bend tests the vertical load point
displacement was measured with an instrumented dial gauge. In the mixed mode MMB test,
the vertical fulcrum displacement, dc, as well as the opening displacement, dI, were measured
with instrumented dial gauges, see Fig. 1. The vertical fulcrum displacement was measured
II - 324
directly on the specimen, whereas the opening displacement was measured on an extension of
the upper hinge in the direction of the applied load. In the experiments loads, displacements
and crack lengths were acquired with an IBM PC equipped with Labtech Notebook software.
The void content was determined by image analysis, using the OPTILAB image analysis
software of Graftek. The void content of the composite was averaged over 36 cross-sections
covering the entire area (20*3.3 mm2) of totally 20 specimens. Interlaminar void contents in
0/+5 and +5/-5 interlayers were determined separately in one dimension on the respective
interface. The void content of the laminate was determined on specimens cut parallel as well
as perpendicular to the 0-direction, whereas interlaminar void contents were determined on
specimens cut parallel to the 0-direction, only.
DATA REDUCTION
Here, a short description of the methods used for the data reduction is given. A more complete
description was presented in ref. 7. The methods used for evaluation of the DCB and ENF are
based on load, displacement and crack length measurements. The results for the strain energy
release rates for MMB load cases are achieved by superposition of the pure mode I and mode
II load cases, see Figs. 2 and 3.
The mode I strain energy release rate was determined using an expression by Hashemi et al.
[9],
GI =
3 PI d I
2 b(a + )
(1)
In Eq. (1) b is the specimen width, a is the measured crack length and is a crack length
correction. is determined analytically using an expression by Olsson [10] (the shear factor is
set to K=5/6),
1
EF h
E
EF
+4 F +
=
+
6E T a
6E T 2
h 12KG T
EF
GT
(2)
The strain energy release rate in mode II was determined using the following expression by
Carlsson and Gillespie [11]
3
1 C PII
6 L + 3a L2
SH
9a PII d II
d II
a
,where CSH =
G II =
2b
20 bhG T
(2L3 + 3a3 )
2
(3).
In the equations presented, h is half the thickness of the specimen, EF is the measured
longitudinal flexural modulus. ET and GT are the transverse Youngs and shear moduli,
respectively. Here, data from the void-free laminate are used [7], ET=10.5 GPa and GT=5.25
GPa. Eqs. (1) and (3) have low sensitivity to the accuracy in estimations of these mechanical
properties.
II - 325
3c-L
4L P
DCB
+
c+L
ENF
c+L
2L
= c+L
MMB
c+L
3c-L
P
4L
c+L
4L P
c+L
4L P
c
L P
2L
c-L
P
2L
Fig. 2: Separation of modal loads in the MMB test
The results for the strain energy release rates for MMB load cases are achieved by
superposition of the pure mode I and mode II load cases, see Figs. 2 and 3. The respective
mode I, PI , and mode II, PII , load components, which have been worked out by Reeder and
Crews [6], are proportional to the applied load, P, as:
3c L
PI =
P
4L
c + L
PII =
P
L
(4)
where P, c and L are defined in Fig. 1. Further, geometry determines the displacements
associated with mode I opening, dI, and mode II flexure, dII. The mode I opening
displacement, dI, is measured. The mode II displacement, dII , is calculated from dI and the
fulcrum displacement, dc, as
d II = dc + c , where c = d I L
1
16 L2 d 2
I
(5)
In the analysis of the three-point bend tests, determining the flexural modulus, influence of
shear deformation was taken into account.
II - 326
d1 /2
dI
d1 /2
c
dc
dII
Fig. 3: Separation of modal displacements in the MMB test

The laminate void content was determined to 0.9%. The interlaminar void contents between
the 0/+5 (the position of the starter crack) and +5/-5 plies were 1.4% and 9.2%,
respectively. The scatter in data is large as the voids were irregularly distributed in the
laminate. For this reason some of the cross-sections contained a large amount of voids as
others were void-free. The largest values for the void content of the laminate and interlaminar
void content measured in one specimen were 2.2% and 25.8%, respectively. All the observed
voids were needle shaped, meaning that they were long in the direction parallel to the fibers
compared to their width. The length of the individual voids varied, whereas the void diameter
was typically that of 1-2 fiber diameters.
The flexural modulus of the laminate containing voids was determined to 111 GPa, which is
7.5% lower than that previously measured flexural modulus of the void-free laminate [7]. This
is in good agreement with results by Ghiorse [5], who reported a decrease in the flexural
modulus of 5.3% for a carbon fiber/epoxy cross-ply laminate for the first one percent voids.
Delamination tests
Table 1: The critical strain energy release rates of the tests on the void-free and voided
laminates. Standard deviations are presented within brackets. All units in J/m2.
Void-free laminate
voided laminate
Test method
GC (initiation)
GC (max)
GC (initiation)
GC (max)
DCB
(mode I)
229.0 (17.8)
253.4 (6.6)
239.8 (9.7)
441.4 (50.8)
ENF
(mode II)
883.1 (117.5)
883.1 (117.5)
811.3 (57.0)
811.3 (57.0)
MMB
(mixed mode)
478.8 (43.7)
487.1 (45.5)
381.4 (47.6)
454.5 (70.3)
The results of the delamination tests are summarized in Table 1. In the table critical strain
energy release rates, GC, for void-free specimens and specimens containing voids, referred to
as voided laminate, are presented. Table 1 lists data for GC at crack growth initiation as well
as maximum GC at crack propagation. The maximum GC provides information about the R-
II - 327
curve behavior of the specimens, i.e. the dependency of the critical strain energy release rate
on crack length. All data for void-free specimens were reported in a previous study [7]. In ref.
7 a more comprehensive description of the delamination crack growth behavior of the voidfree specimens is given.
DCB tests
The presence of voids did not affect the critical strain energy release rate at crack growth
initiation in pure mode I, see Table 1. However, their presence affected the crack growth
behavior. Although stable crack growth was observed in both void-free specimens and
specimens containing voids, their over all behavior were considerably different. For void-free
specimens no R-curve behavior was found. In contrast, for specimens containing voids a very
strong R-curve behavior was observed. Typical appearances of the critical strain energy
release rates as a function of crack lengths are depicted in Fig. 4. For these specimens the
critical strain energy release rate increased as the crack propagated. The maximum value at
crack propagation was 84% higher than that at crack growth initiation, see Fig. 4. This shift in
crack growth behavior is explained by a change in failure mechanism. In the void-free
specimens an insignificant amount of crack bridging was established at the crack tip. In the
specimens containing voids, however, ply splits (approximately 1-2 mm wide) were observed
to bridge the crack. The bridged zone was measured to be 25-30 mm long. The observed
crack bridging appeared as the crack tip locally leaped between plies as the crack propagated.
The scatter in data for the laminate containing voids is significantly larger than for the voidfree laminate, see Table 1. Thus, the irregularity in the distribution of voids affects the
interlaminar delamination toughness of the composite loaded in pure mode I.
500
Gc (J/m2)
400
300
200
Voided laminate
100
Void-free laminate
0
30
40
50
60
70
a (mm)
Fig.4: Schematic of the mode I critical strain energy release rates dependence on crack
length for the void-free and void containing (voided) laminates
In a recent study, Hou et al. [12] reported results on the influence of voids on the interlaminar
delamination toughness for CF/PEI woven fabric composites. In accordance to the results
presented here, Hou et al. reported an increase in the value of the critical strain energy release
rate at crack propagation with an increase in void content. However they found this increase
in the energy release rate to originate in an enlarged crack tip region rather than in ply-
II - 328
splitting. Due to the fiber architecture, no ply splits were reported to appear in the woven
composites.
ENF tests
For both types of laminates the crack growth was unstable when loaded in pure mode II. In
the ENF tests, presence of voids was found to have a slightly deleterious effect on the critical
strain energy release rate. GIIC at crack growth initiation was 8% lower for the laminate
containing voids compared to the void-free one. To assure that the resin pocket created at the
starter film tip did not affect the measured energy release rate, tests of four pre-cracked ENF
specimens were performed. The specimens were pre-cracked in mode I until the starter crack
length exceeded 37 mm. These tests confirmed the results, and hence any effect of the resin
pocket was excluded. However, in these additional tests it was found that stable crack growth
can be achieved as the initial crack length exceeds 37 mm. According to the analysis by
Reeder and Crews [6] stable crack growth is expected for a crack length larger than 35 mm.
Observations of the crack tip indicated that the crack propagated in one plane, only. As a
consequence, no crack bridging was detected. This is coherent with observations reported for
mode II tests on CF/PEI woven composites by Hou et al. [12].
MMB tests
Stable crack propagation was achieved in both laminates. In mixed mode loading the presence
of voids was found to decrease the critical strain energy release rate at crack growth initiation
by 20%. However, as the crack propagated the critical strain energy release rate increased and
a R-curve behavior was recorded for the void containing specimens. Due to this R-curve
behavior the recorded maximum strain energy release rate at propagation was at the same
level for the two laminates. As in the DCB test results, the scatter in data, at propagation, for
the laminate containing voids is larger than for the void-free laminate, see Table I. Crack
bridging was not observed in either of the two laminates. As the crack propagated in the 0/5
interface only, the recorded R-curve behavior of the laminate containing voids is explained by
multiple crack planes at the crack tip. Multiple crack planes were observed on the edge of the
specimens at crack propagation and in post-mortem optical microscopy.
The mode mixture in the MMB test deviated from the expected GII/G=0.5. The achieved
mode mixture was found to be 0.54 for the laminate containing voids as well as for the voidfree laminate. These deviations from the expected mode mixture are due to the evaluation
method used. Based on linear elasticity, methods based on load and load-displacement
measurements should give the expected mode mixity of 0.5 [6]. In fact, load based methods
give a mode mixture of 0.5. However, a previous investigation [7] reveled non-linear loadcompliance relationships and indicated methods based on load-displacement measurements to
be more accurate than methods based on load measurements only. For this reason, results
based on load and displacement measurements only are reported here.
Dependency of delamination toughness on mode mixture
In Fig. 5 the critical strain energy release rates at crack growth initiation are presented as a
function of mode mixture. To show trends, data points for void-free laminate and the laminate
containing voids, marked voided laminate, have been curve fitted by interpolation.
II - 329
The graph shows that presence of voids reduced the critical strain energy release rate at crack
growth initiation in pure mode II and mixed mode (GII/G0.5). In pure mode I, however, GC
was unaffected by the present voids. In contrast, at crack propagation, presence of voids most
strongly affected the critical strain energy release rate in pure mode I.
Voided laminate
Gc (J/m2)
800
Void-free laminate
600
400
200
0
0
0.25
0.5
0.75
GII/G
Fig. 5: Critical strain energy release rate at crack growth initiation as a function of mode
mixture. Interpolations between data points are included to show trends for the two
laminates.
CONCLUSIONS
The influence of voids on the interlaminar delamination toughness of a carbon fiber/epoxy
composite in pure modes I and II as well as mixed mode has been investigated. Voids were
found to have a deleterious or no effect on the critical strain energy release rate at crack
growth initiation. However, at propagation in pure mode I and mixed mode load-cases
presence of voids was found increase GC. This increase, which was found to be very strong in
mode I and moderate in mixed mode, is inherent to changes in failure mechanisms. In pure
mode I, ply splits were observed to bridge the crack as the crack locally leaped between plies
at propagation. In mixed mode (GII/G0.5), the crack propagated in one interface. However,
multiple cracks observed at the crack tip and in front of the crack, resulted in an increased GC
at crack propagation in mixed mode loading. From the observed R-curve behavior it is
concluded that pores and voids do not impair the delamination toughness of the HTA/6376C
composite in static loading.
II - 330
REFERENCES
1.
Judd, N.C.W. and Wright, W.W., Voids and their effects on the mechanical properties
of composites- An appraisal, SAMPE Journal, 14, 1978, pp. 10-14.
2.
Cantwell, W.J. and Morton, J., The significance of damage and defects and their
detection in composite materials: A review, J. strain analysis, 27, 1992, pp. 29-42.
3.
Bowles, K.J. and Frimpong, S., Void effects on the interlaminar shear strength of
unidirectional graphite-fiber-reinforced composites, J. Comp. Mat., 26, 1992, pp.
1487-1509.
4.
Wisnom, M.R., Reynolds, T. and Gwilliam, N., Reduction in interlaminar shear

strength by discrete and distributed voids, Comp. Sci. Techn., 56, 1996, pp. 93-101.
5.
Ghiorse, S.R., Effect of void content on the mechanical properties of carbon/epoxy

laminates, SAMPE Quart., 1993, pp. 54-59.
6.
Reeder, J.R. and Crews, Jr, J.H., Mixed-mode bending method for delamination
testing, AIAA Journal, 28, 1990, pp. 1270-1276.
7.
Asp, L.E., Moisture and temperature effects on the interlaminar delamination

toughness of a carbon/epoxy composite, FFA TN 1996-52, The Aeronautical Research
Institute of Sweden, 1996.
8.
Olsson, R., Thesken, J.C., Brandt, F., Jnsson, N. and Nilsson, S., Investigations of
delamination criticality and the transferability of growth criteria, FFA TN 1996-31,
The Aeronautical Research Institute of Sweden, 1996.
9.
Hashemi, S., Kinloch, A.J. and Williams, J.G., Corrections needed in double cantilever
beam tests for assessing the interlaminar failure of fibre composites, J. Mat. Sci. Lett.,
8, 1989, pp. 125-129.
10.
Olsson, R., A simplified improved beam analysis of the DCB specimen, Comp. Sci.
Techn., 43, 1992, pp. 329-338.
11.
Carlsson, L.A. and Gillespie, Jr., J.W., Mode-II interlaminar fracture of composites,
In Application of fracture mechanics to composites, K. Friedrich, Ed., Elsevier,
Amsterdam, 1989, pp. 113-158.
12.
Hou, M., Ye, L. and Mai, Y.-W., Effect of moulding temperature on flexure, impact
strength and interlaminar fracture toughness of CF/PEI composite, J. Reinf. Plast.
Compos., 15, 1996, pp. 1117-1130.
II - 331
A STATISTICAL MODEL FOR CRACK GROWTH IN

COMPOSITE MATERIALS
Biao Wang, Jin Dai and Shanyi Du
Research Center for Composite Materials, Harbin Institute of Technology, Harbin,China
SUMMARY: In this paper, the crack propagation problem in composite with random
microstructures was considered in detail. Along the crack path, the material is assumed to
contain randomly distributed crack-arresting points which have stochastic fracture strength.
The model can be used to describe a large variety of composite materials, such as continuous
fiber composites and short-fiber composites, etc. The aim of this study is to calculate the
probability of fracture for a given material, crack length and external load. As an example,
the unidirectional fiber reinforced composite was considered in detail. The fiber strength was
assumed to be a random variable with the Weibull distribution, and both the strong and weak
interface between the fiber and matrix were considered. It was found the fracture strength was
sensitive to the interface behavior.
KEYWORDS: heterogeneous materials, micro-mechanics, statistical model, crack growth
INTRODUCTION
The fracture toughness of heterogeneous materials, such as composites is quite different from
the fracture toughness of their constituents. For example, composite materials can be obtained
with much higher fracture toughness than the matrix materials. It is a well-known fact that
such enhancements of fracture strength are due to the matrix, the reinforcement and the
interface interaction with each other. Whereas, it is a extreme difficult task to predict the
overall toughness and the probability of fracture from data on the microstructure.
Many theories have been proposed to understand the fracture mechanism of heterogeneous
solids, such as fiber-reinforced composite [6,7]. The most significant works are those on fiber
reinforced ceramic and the other brittle matrix composites, for example Ref.[1-5]. When the
tensile load which is much lower than the utmost load of the material is applied on fiberreinforced brittle matrix composite, closely spaced matrix cracks are induced perpendicular
with the direction of the tensile load, and each crack almost completely traverses the material,
but the applied load continues to be supported by unbroken fibers that bridge each matrix
crack between its faces. In this paper, different fracture mechanism was considered. Along the
crack path, the material is assumed to contain randomly distributed crack-arresting points
which have stochastic fracture strength. For the crack to propagate over a distance x , it must
cut off all the obstacles, such as fibers, in this area. The model can be used to describe a large
variety of composite materials, such as continuous fiber composites and short-fiber
composites, etc. Through the microstructural analysis, the hyperbolic differential equation
was derived for determining the probability density function of crack length for given external
load, the average fracture toughness of the material was also obtained. As an example, the
unidirectional fiber reinforced composite was considered in detail. The fiber strength was
II - 332
assumed to be a random variable following the Weibull distribution, and both the strong and
weak interface between the fiber and matrix were considered. It is found that the fracture
strength is sensitive to the interface behavior.
THE STATISTICAL MODEL FOR CRACK GROWTH IN COMPOSITES

Consider a general heterogeneous materials containing a straight crack as shown in Fig.1.
inclusion
crack
Fig. 1: Heterogeneous material containing a crack

When the applied load increases, the crack should propagate forward. Along the crack path,
the inclusions, such as fiber, play the role of arresting point for the crack propagation. Since
the inclusion usually has a stochastic strength value, and is distributed randomly in space, the
following assumption are introduced in this model.
1. The occurrence probability of a crack arresting-point in region (x, x+ x ) is given by
( x ) x ;
2. The probability for no crack arresting-point in region (x, x+ x ) is given as 1 ( x ) x ;
3. The occurrence probability for two and more crack arresting-points in this region is the
higher-order smaller quantity as o ( x ) .
If the strength distribution of the crack arresting-point is given by F, ( x ) x can be derived
as following
( x ) x =Prob[there is an inclusion in the region (x, x+ x )]

Prob[the inclusion will not break at the tip of the
crack under the action of external load]
= nx[1 F ]
(1)
where n is the average number of the inclusions in unit length. In deriving Eqn (1), We
assumed that distribution of the inclusions in space is statistically uniform.
If P0 ( x ) is denoted as the probability that there is no crack arresting-point in region ( x0 , x ),
one can obtain
P0 ( x + x ) = P0 ( x )[1 ( x ) x ] = P0 ( x ) ( x ) P0 ( x ) x
(2)
II - 333
Therefore, one can derive

dP0 ( x )
= ( x ) P0 ( x )
dx
(3)
The solution of equation (3) is given by

x
P0 ( x ) = exp ( x )dx
x0
(4)
If the crack starts to propagate from x0 under the action of the external load, the probability
for it to be arrested in the region ( z , z + z ) is given by
z
< z| x 0 > z = exp ( x )dx ( z ) z

x0
(5)
therefore the probability density function of arresting the crack at z is given by

z
< z| x0 >= ( z ) exp ( x )dx

x0
(6)
Once the crack starts to propagate at x0 , the probability for it to be arrested without increasing
the load is given by

Ps = < z| x 0 > dz = 1 exp ( z )dz

x0
x0
(7)
Therefore, once the crack starts to propagate at x0 , the probability that it will propagate to
infinity and induces the material to fracture is given by

Pf = 1 Ps = exp ( z )dz
x0
(8)
As defined in Kunin[6], a load-specimen configuration is stable if Pf = 0 , i.e., the crack will

be arrested with probability 1, otherwise it is unstable.
The mean crack arrest length for cracks initiated at x0 is

z
m1 ( x0 ) = < z| x0 > dz = x0 + exp ( x )dx dz
x0
x0
x0
II - 334
(9)
The above analysis is suitable for crack propagation spontaneously in a single excursion
without increasing the load. In what following, we try to incorporate the effect of increasing
load to derive the differential equation for determining the crack length distribution p( x , ) .
Let us consider a loaded solid described in Fig.1. Denote by ( ) the random position of the
crack tip at an external load , and by P( x , t )x the probability of finding the crack tip at a
length x at the external load , assuming that at = 0 the tip was at x0
p( x , ) x = Pr ob[ x ( ) x + x| (0) = x0 ]
(10)
To relate p( x , + ) to the distribution p(, ) of crack locations at , notice first that the
crack tip ends up at the length x at the load + in one of the following mutually
exclusion ways: during , it either propagated to x from one of the length z between the
notch tip and x, x0 z < x , or it was at x at the load and remained there. By the formula of
total probability, one obtains
x
p( x , + ) x = < x| z > dx
x0
+ (1
dF
P( z , ) z
d
dF
) p( x , ) x
d
= < x| z > p( z , ) f ( z , )dzx

x0
+[1 f ( x , ) ] p( x , ) x
where f ( z , )( =
(11)
dF
) is the probability density function of strength for a crack arrestingd
point.
Rearranging equation (11), one can derive
p ( x , )
= < x| z > p( z , ) f ( z , )dz f ( x , ) p( x , )
x0
x
(12)
From equation (6), one obtains

< x| z >= ( z )
< x| x0 >
< z| x 0 >
(13)
Substitution of equation (13) into equation (12) yields
p ( x , )
=< x| x0 > ( z ) p( z , ) f ( z , )/ < z| x0 > dz f ( x , ) p( x , )
x0
x
(14)
II - 335
Divide equation (14) by < x| x0 > and denote

q( x, ) =
p( x , )
< x| x0 >
(15)
Then equation (14) becomes
q ( x , )
= ( x ) f ( z , )q ( z, ) dz f ( x , )q ( x , )
x0
x
(16)
Differentiating equation (16) in x, one obtains the following equation for q:
2q ( x, )
f ( x , )
q ( x , )
= ( x ) f ( x , )q ( x , )
q ( x, ) f ( x, )
x
x
x
(17)
Substitution of equation (15) into equation (17) yields
2 p p
< x | x 0 > p
< x | x0 >
1
1
+ f ( x, )

x x
x
< x| x0 >
< x| x0 >
f
1
+ p( x , ) ( x ) f ( x , )
f ( x , ) < x| x0 >
x
< x | x0 >
x
2
2
1
+
< x| x 0 >
< x | x0 >
< x| x0 > = 0
2
< x | x 0 > x
< x| x0 > x
(18)
Equation (18) can be expressed in the form as
2p
p
p
+ a ( x , ) + b( x , )
+ c( x , ) p = 0
x
x
(18)
where
a( x, ) = f ( x, )
b( x , ) =
c( x , ) =
+
II - 336
2
x x0
x x0
1
x x0
x x0
1
x
x x0
x x0
f
1
( x) f ( x, )
f ( x, )
x
x
x x0
x x0 x x0
2
1
x x0
x
x x 0 x
(19)
From equation (6),one finds that if we know the microstructural characteristics of the
material, represented by the average number of inclusions in unit length n ,the strength
distribution F( ) of the crack arresting-points and the applied load, the parameters
a ( x , ) , b( x , ) , c( x , ) can be determined easily. If the initial crack tip is at x 0 when
= 0 ,the probability density function p( x , ) of the crack tip location at can be
determined by equation (18) and the initial condition
p( x , ) = ( x x 0 )
p( x , ) = 0, x < x 0 , 0
(20)
The derivation of equation (18) is quite similar with the work of Kunin[6],whereas in Kunins
work, the mode of crack growth was modeled by a Markovian stochastic pattern of a
microscopic random jump, followed by a random waiting time, a random jump, and so on. In
our model ,the applied load plays the role of the time t. In the following part, as an example,
the unidirectional fiber reinforced composite was considered in detail.
CRACK GROWTH IN UNIDIRECTIONAL FIBER COMPOSITES
In most composites of current interest ,the extensional stiffness of the fibers is much greater
than that of the matrix. As a result, the fibers support most of the load in case of uniaxial
tensile loading. Considering a 90 0 -slit notch in a unidirectional composite(Fig.2), for it to
extend in an unstable manner at an applied load, it is necessary that the intact fiber at the root
of the notch breaks. This requires that the stress in this fiber exceeds its strength .Generally
speaking, the fiber strength can be reasonably assumed to follow the Weibull distibution,
therefore
k
(21)
F ( ) = 1 exp ( x )
0
where kx is the stress concentration factor for the intact fiber at the tip x of the notch.
Fig.2: A 90 0 -Slit notch in unidirectional composite
To calculate the stress concentration factor for the intact fiber, following Zweben[1], the
modified shear-lag model was used as shown in Fig.3.
II - 337
Fig.3: Model for the analysis of stress concentrations

including nonelastic matrix effect
It consists of a central core of n broken fibers separated from two adjacent intact fibers by a
matrix region whose effective width is d. The intact fibers are representative of regions of
fibers which are subjected to a stress concentration resulting from the notch. The intact fibers,
in turn, are flanked by regions of matrix of width d, beyond which lies material which is
assumed to have the behavior of overall composites, so as the part of central core.
Because of the large elastic shear stress concentration in the matrix at the root of the notch
and the relatively weak interfacial strength and strength of most composite materials, it can be
anticipated that there will be nonelastic effects such as debonding and plasticity in the matrix
between the core and intact fibers at the root of the notch. We therefore include in the model a
region of matrix in which the behavior is nonelastic , such as the interfacial slip. The axial
dimension of this region is 2a. In this nonelastic region , the shear stress is assumed to be
constant and equal to the interfacial failure stress.
The boundary conditions at x=0 are zero stress in the core and zero displacement in the intact
fibers. Because of the mixed boundary conditions ,we choose as unknowns the axial
displacements of the core and intact fibers,U 0 and U 1 , respectively. By using the
nondimensional quantities, the equation of equilibrium are
d 2 u0
n
2 m = 0
d 2
Ef
d 2 u1
+m = 0
u1 +
Ec
d 2
(22)
d 2 u0
2(u1 u0 ) = 0
d 2
Ef
d 2 u1
=0
+ 2u1 u0
2
Ec
d
(23)
where
II - 338
ui =
Gm A f 1/ 2 1
1 A f d 1/ 2
(
) Ui , m = (
)
m
E f Gm h
E f dh
=(
E f Af d
Gm h
) 1/ 2 x
, = (
E f Af d
Gm h
) 1/ 2 a
(24)
and
A f = fiber cross-sectional area
E f = fiber extensional modules
E c = composites extensional modules
Gm = matrix shear modules
= nondimensional length
d = fiber spacing
h = material thickness
u= nondimensional axial displacement
= nondimensional axial coordinate
m = matrix shear stress
m = nondimensional shear stress

= applied stress at infinity
The boundary conditions are
du0 (0)
= 0,
d
u1 (0) = 0 ,
du0 du1
=
=1
d
d
(25)
In addition, the displacements and forces should be continuous at = ,

and also
u1 ( ) u0 ( ) = m
(26)
By solving equations (22) and (23) under the given boundary condition, the stress
concentration factor for the intact fiber at the tip of the notch which contains n broken fibers is
given by
k x (0) =
du1 ( )
d = 0
(27)
The material constants are:
II - 339
E f = 360GN / m 2 , f = 0.4 , = 6.98 , 0 = 355 Mpa , d = 8 10 6 m , L = 25.4 10 3 m

The results are shown in Fig.4
kx
Fig. 4 The stress concentration factor for the intact fiber

Substituting the stress concentration factor into equation (21),then into equation(8),we obtain
the failure probability of the specimen under the given applied load and the notch size. The
results are shown in Fig.5 as the function of the interfacial strength m between the fiber and
matrix.
< >
Pf
0.06
0.04
0.02
300
Number of broken fibers n
Fig.5 The failure probability of unidirectional

composites with a notch
600
1000
Number of broken fiber
Fig.6. The average fracture stress

of the notched specimen
To calculate the average fracture strength of the unidirectional composite with a notch, one
can write,
dp f
max
max
=
(28)
d = max
p f d
0
0
d
where max is the utmost strength for all the specimens of the material, the result given by
equation (28) is shown in Fig.6.
II - 340
CONCLUDING REMARKS
In this paper, a statistical model for crack growth in general heterogeneous solid was proposed
based on the material microstructure and the strength distribution of the inclusion. As an
example, the unidirectional fiber composite with a 90 0 -slit notch was considered in detail.
We believe that the effect of the interfacial strength on the fracture stress was magnified due
to the assumption that the fiber will break at the point with the maximum stress. In fact, the
breaking point along the intact fiber should be widely dispersed due to the stochastic nature of
the fiber strength.
REFERENCES
1.
Zweben, C., An Approximate Method of Analysis for Notched Unidirectional

Composites, Engineering Fracture Mechanics, Vol.6, No.1,1974
2.
Cox, B.N., Marshall, D.B. and Thouless, M.D., Influence of Statistical Fiber Strength
Distribution on Matrix Cracking in Fiber Composite, Acta Metall, Vol.37, p1933, 1988.
3.
Rubinstein, A.A. and Kang Xu, Micromechanical of Crack Growth in Fiber-reinforced

Ceramics J.Mech.Phys.Solids,V.40, p105, 1992.
4.
Gao, Y.C.,Mai, Y.W. and Cotterall, B., Fracture of Fiber-Reinforced Materials, Journal
of Applied Mathematics and Physics (ZAMP), V.39,p.550, 1988
Chang, Y.C.,Wang, A.S.D. and Chou, T.W., On matrix Cracking in Fiber Reinforced
Ceramics, J.Mech.Phys.Solids, V.41, p.1137, 1993.
6.
Kunin, B.I., A Stochastic Model For Slow Crack Growth in Brittle Materials, ASME
Appl.Mech.Rev., V.47,No1,173, 1994.
7.
Jeulin, D., Fracture Statistics Models and Crack Propagation in Random Media, ASME
Appl.Mech.Rev,V.47, No.1,p.141, 1994.
II - 341
MODELLING PROBLEMS OF DAMAGE AND

FRACTURE OF NOTCHED FIBRE COMPOSITES
R. A. Dimant, H. R. Shercliff and P. W. R. Beaumont
Cambridge University Engineering Department
Trumpington Street, Cambridge CB2 1PZ, England
SUMMARY: The Kortschot, Spearing, Beaumont model of notch-tip damage in carbon

fibre/epoxy laminates has had some success at predicting the early stages of damage growth
in glass fibre/epoxy laminates under monotonic and cyclic loading. However, the model is not
universally applicable in its present form to the propagation of damage in glass/epoxy,
particularly in high cycle fatigue. This difficiency in the model can be attributed to the effects
of the transverse (90) ply crack on the breakage of glass fibre within the 0 ply neighbour.
Although the K-S-B model takes into account the effect of transverse ply cracking on changes
in laminate compliance, it does not consider the stress concentration effect of the transverse
ply crack on the localised tensile stress in the fibres of the adjacent 0 ply.
KEYWORDS: modelling, damage, fracture, cracks, carbon fibre, glass fibre, epoxy
INTRODUCTION
An important difference between carbon/epoxy and glass/epoxy is in the degree of anisotropy.
Whilst the ratio of longitudinal modulus to transverse modulus of a unidirectional
carbon/epoxy laminate is between 15 and 20, that of glass/epoxy is closer to 3 or 4. This
means for a cross-ply glass/epoxy laminate, the 90 ply carries a greater share of the applied
tensile load than would carbon fibre. Thus, the relative magnitude of the stress increment
transferred into a longitudinal 0 ply adjacent to a transverse ply crack (tpc) in the 90 ply
neighbour is greater for the glass/epoxy.
The Kortschot, Spearing and Beaumont model is of damage accumulation in the vicinity of a
notch-tip. This physical model is based on the identification by direct observation of matrixdominated cracking processes in cross-ply and multi-directional carbon/epoxy laminates
under monotonic [1,2] and cyclic loading [3,4]. This paper assesses the applicability of the KS-B model to a wider range of composite material systems, notably glass/epoxy laminates, in
terms of:
(1)
(2)
(3)
(4)
II - 342
monotonic and cyclic tensile loading

damage, applied stress, and load cycles
transverse ply thickness
terminal damage-state, fracture stress, and post-fatigue strength.
K-S-B MODEL
The K-S-B model is based on the direct observation and assessment of matrix-dominated
cracking in the vicinity of a notch-tip: a split in the 0 ply; a delamination at the 0/90
interface, roughly triangular in shape; and a crack in the transverse or 90 ply, (see Kortschot
and Beaumont [1,2] and Spearing and Beaumont [3,4]). Under monotonic or cyclic loading,
the damage zone doesn't change it's shape, it simply grows in size. The extent of damage can
be characterised by a straightfoward measurement of split length and angle at the tip of the
split and delamination crack which forms the triangle, known as the delamination angle. This
is physically sound for cross-ply carbon/epoxy laminates.
The fracture stress of the damaged laminate depends on competition between these
mechanisms which, on the one hand, reduce the stress concentration of the notch by notch-tip
blunting, and on the other by reducing the tensile strength of the load bearing 0 ply close to
the notch-tip. In carbon/epoxy, the notch-tip blunting mechanisms dominate and the postfatigue strength of the laminate actually goes up. This may not be the case in other fibrematrix systems.
COMPOSITE SYSTEMS, SPECIMEN DESIGN

Composite Materials
Glass/epoxy laminates (913G-E) were prepared by Polymeric Composites, Bristol (UK) and
carbon/epoxy laminates (924C-T300) were made by Ciba-Geigy, Duxford (UK). The
laminates were moulded into sheets measuring 600mm by 750mm and cured in accordance
with the manufacturers specifications. Tensile tests in monotonic and cyclic loading were
performed on (0/90/0), (0/90)s, (0/902)s and (0/904)s laminate configurations.
Specimen Geometry
Parallel-sided tensile specimens were cut to a width of 20mm and a gauge length of 160mm
using a guillotine. Two 2mm diameter holes, centres spaced 6mm apart, were drilled at the
mid-section of the specimen. These holes were connected by an 8mm long central notch, cut
with a diamond-coated hacksaw blade. The notch/specimen width ratio (2a/w = 0.4), is
similar to that used by Kortschot and Beaumont [1,2] and Spearing and Beaumont [3,4]. For
comparative purposes, some tests were carried out on unnotched samples.
Quasi-static Tensile Strength and Fatigue Strength
Five duplicate tensile tests were performed to determine the quasi-static fracture strength. A
computer-controlled Instron 6025 screw-driven mechanical testing machine was used for this
purpose at a displacement-rate of 1mm/min. Similar notched specimens were cyclicallyloaded in tension using an Instron 1271 servo-hydraulic mechanical testing machine in load
control with a cyclic stress ratio R = 0.1 at a frequency of 5Hz. Each specimen was fatigued
at 80% of it's quasi-static fracture strength.
II - 343
Damage Observation and Assessment

Damage growth in carbon/epoxy laminates was observed using a zinc iodide penetrant dye
solution and X-ray radiography. The optical properties of glass/epoxy laminates allow the
direct observation of damage growth. Split length in glass/epoxy was determined by placing a
micrometer against the actual surface of the specimen and from measurements taken from
photographs of cracks at regular intervals. Where possible to do so, the delamination angle
was calculated by assuming a triangular-shaped delamination crack, after K-S-B.
QUASI-STATIC FRACTURE STRESS MEASUREMENTS

Components of Damage
In notched glass/epoxy laminates, as with carbon/epoxy, the mechanisms of cracking are:
(1)
(2)
(3)
(4)
splitting within the 0 ply

transverse ply (90 ) matrix cracking
delamination of the 0/90 interface
fibre fracture within the 0 ply.
Damage Growth
Damage growth in glass/epoxy with increasing load and ply thickness is shown in Fig. 1.
(0/90/0) glass/epoxy
A split initiates at the notch-tip and propagates under increasing load. A fine dispersion of
closely-spaced transverse ply cracks also develops in the vicinity of the notch-tip. At stresses
approaching the fracture stress of the laminate, fibres in the longitudinal 0 ply snap
sequentially at the notch-tip. By these processes, the notch-tip damage zone grows by a series
of localised fibre breaks until the laminate finally separates into two pieces.
(0/90)s glass/epoxy
The appearance and development of damage are similar to that observed of the (0/90/0)
laminate. Having either a single or double transverse ply within the laminate, fibre fracture is
not dominant but part of the overall failure process. The relationship between applied tensile
stress and normalised split length (l/a) is one where split length scales with the square of
applied stress, as found and modelled by Korschot and Beaumont for carbon/epoxy.
(0/902)s glass/epoxy
A split initiates at an applied stress close to the fracture stress. Transverse cracks in the
quadruple-layered 90 ply are prominent and widely-spaced. At final fracture, damage
propagates from the tip of the notch to the edge of the specimen in a series of sub-failures
with fibres snapping in the 0 ply. This effectively increases stably the length of the notch. A
new split initiates in the 0 ply at the edges of the notch-tip crack. The notch extends by stable
crack growth until the laminate fractures completely. Figure 1 indicates that fibre breaks in
II - 344
the 0 ply are coincident with the plane of transverse ply cracks in the vicinity of the notch-tip.
Delamination at the 0 /90 interface is apparent.
In this 8-layer transverse ply laminate, failure is characterised by the growth at the notch-tip
of a single transverse crack only. Under increasing stress, the tpc extends towards the edge of
the specimen. Close to specimen fracture, fibre breakage in the 0 ply adjacent to the notchtip is apparent along the line of the tpc. Ultimate failure of the laminate is accompanied by
delamination of the 0 /90 interface which initiates at the single transverse ply crack.
We observe that damage which develops in the (0/902)s laminate is a combination of the two
modes of failure. The transverse ply thickness of the (0/902)s laminate represents a transition
between dominating failure mechanisms.
Fracture Stress and Transverse Ply Thickness

The effect of transverse ply thickness on laminate fracture stress, where fracture stress is
expressed as the average failure stress carried by the longitudinal 0 plies, is shown in Fig. 2.
Fig. 2: The effect of transverse ply thickness on notched glass/epoxy fracture stress, where
fracture stress is expressed as the average stress carried by the longitudinal 0 plies
FATIGUE DAMAGE AND POST-FATIGUE STRENGTH

Components of Fatigue Damage in Glass/Epoxy
The components of damage are similar to but more extensive than observed in monotonic
loading. Even in the early stages of fatigue damage, it is possible to identify and determine
the delamination angle (Table 1).
II - 345
Fatigue Damage Growth

The extent of fatigue damage in glass/epoxy with load cycling at 80% of laminate strength
depends on ply thickness (Fig. 3). The stress indicated for each specimen is based on the
stress carried by the 0 plies of the laminate.
(0/90/0) glass/epoxy
Initially, fatigue damage accumulates with the characteristic pattern identified by Kortschot,
Spearing and Beaumont (Fig. 3). Furthermore, an additional damage mode develops between
10K and 20K load cycles: fibres in the 0 ply at the tip of a split fracture through the thickness
of that ply. This creates effectively a secondary notch at the tip of the split from which a
second series of splits form. As the cross-sectional area reduces, the stress in the remaining
ligament increases, which subsequently affects the split growth-rate. With additional fatigue,
further splitting occurs at broken fibre bundles. In this way, random damage develops in the
composite. Damage progresses from the notch-tip towards the edge of the specimen, as it
does so reducing the load-carrying capacity of the laminate. At 70K load cycles, the damage
has spread throughout the laminate. In one ligament of specimen, four or more groups of
splits can be seen. Damage continues to propagate until catastrophic failure, in this example,
at 71.3K load cycles. The development of a non-geometrical damage zone occurred between
10K and 20K load cycles.
(0/90)s glass/epoxy
Initially, a similar pattern of damage to that reported by Kortschot, Spearing and Beaumont is
observed in specimens fatigued at 340 MPa (Fig. 3). However, by 10K cycles, it is apparent
that each split is made up of a multiple array of closely spaced cracks. While these cracks
initiated from or near to the notch-tip, the behaviour of the specimen resembles essentially
single crack growth as observed in isotropic materials. At 40K load cycles, a new split has
nucleated at the tip of a bundle of fractured fibres lying in the plane of the notch. At 60K load
cycles, secondary damage has grown beyond the original splits resulting in additional fibre
fracture. This process continues until the weakened specimen is no longer able to support the
applied load. It fails at 78.8K load cycles. Most fatigue failures occurred at approximately
II - 346
II - 347
II - 348
120K load cycles. In this material, damage could no longer be described in simple
geometrical terms by 20 or 30K load cycles.
Overall, damage is different to that observed in laminates having fewer transverse plies (Fig.
3). It is dominated by the development of splits and delaminations, rather than by fibre
fracture. At 40K load cycles, some damage has extended across the specimen by fibre
fracture but the extent of this damage is considerably less than in laminates having only one
or two transverse plies. The form of damage remains predominantly one of splitting and
delamination. Finally, the material fractured at 73.5K load cycles.
The most significant feature of this damage is the absence of fibre fracture in the 0 ply prior
to fatigue failure (Fig. 3). Damage develops simply as splitting at the notch-tip with
delamination at the 0 /90 interface. Interaction between the delamination and transverse ply
cracking plays an important part in the damage process. Failure of the specimen occurs at
9.4K load cycles and is characterised by large delaminated regions extending across the width
of the specimen.
0
UNNOTCHED POST-FATIGUE STRENGTH

Aluminium end-tabs were bonded to 140mm wide by 300mm length plates of glass/epoxy
and carbon/epoxy laminates with a heat-curing resin. Each plate was cyclically-loaded in a
20-tonne Losenhauserwerk servo-hydraulic mechanical test machine.
Glass/epoxy laminates were cyclically-loaded to 50% of their ultimate strength of 600MPa.
Laminates were fatigued in load control up to 10K load cycles at 5Hz and a stress ratio R =
0.2. To begin with, the carbon/epoxy laminates were also fatigued at 50% of their fracture
strength but due to a large number of laminate failures below 10K cycles, the maximum
cyclic stress was reduced to 720MPa or 35% of ultimate.
Parallel-sided specimens for determining post-fatigue tensile strength were cut and ground
from these fatigue-damaged plates to a size of roughly 6mm wide by 100mm length. To
ensure failure within the gauge length, these specimens were further waisted to a width of
2mm at the mid-plane using a 125mm diameter grinding wheel. Thirty specimens were
prepared for each fatigue loading history and laminate. The post-fatigue tensile tests were
performed using an Instron 6025 screw-driven mechanical testing machine, as described in
section 4.1.
Unnotched Post-Fatigue Tensile Strength
For comparative purposes, the post-fatigue strength was calculated by assuming the applied
load is carried essentially by the longitudinal plies only. This method of normalisation is
based on the premise that the laminate close to failure has a multiple array of matrix cracks in
the transverse ply. Thus, the applied load is carried principally by the 0 ply.
II - 349
II - 350
Glass/Epoxy laminates
The monotonic tensile strength of the longitudinal ply decreases continuously as the thickness
of the transverse ply increases (Fig. 4). The error bars correspond to one standard
deviation.The fracture stress of the (0/90/0) laminate is approximately 20% greater than that
of the (0/904)s laminate. This variation in strength is sensitive to the density or spacing of the
transverse ply crack which in turn depends on thickness of the 90 layer. By 10K cycles,
fatigue has little effect on the residual strength of (0/902)s and (0/904)s laminates, while the
residual strength of (0/90/0) and of (0/90)s laminates has decreased by 18% and 8%,
respectively, (Fig. 5). The implication is that for laminates containing a single or double
transverse ply, fatigue results in rapid degradation of the laminate strength. The transition
occurs for laminates having three transverse plies, or a thickness of 0.45mm .
Recall that in the notched laminate having a single or double central 90 ply, fatigue loading
resulted in massive, progressive fibre-fracture prior to ultimate failure. In monotonic loading,
fibre breakage was evident only at loads close to ultimate failure. In laminates containing
four and eight central transverse plies, the dominant mode of failure was by delamination. It is
the absence of delamination and notch-tip blunting at the 0 /90 interface of the single or
double transverse ply which is responsible for the strength degradation.
Carbon/Epoxy laminates
The effect of transverse ply thickness on the monotonic tensile strength of carbon/epoxy is the
opposite to that observed for glass/epoxy, where the fracture stress of (0/904)s is
approximately 20% greater than the (0/90/0) laminate. The general effect of load cycling is
shown in Fig. 6. This is the mirror-image of the pattern observed for glass/epoxy. Cracks form
in the transverse ply. Fatigue damage results in an increase in laminate compliance as the
density of transverse cracks increases. The post-fatigue strength of the carbon/epoxy is
relatively unaffected since the strength of the longitudinal ply is essentially retained.
FINAL REMARKS
The most significant difference between the carbon/epoxy and glass/epoxy laminate
containing a notch is that the latter fails in fatigue whereas the post-fatigue strength of the
former actually goes up. In fatigue of unnotched glass/epoxy, there are significant reductions
in the residual strength of the laminate containing a single or double transverse ply.
Delamination at the 0/90 interface of the 4 and 8-transverse ply laminate, however, results
in crack-tip blunting effects. The result is that strength of the thicker laminate becomes
independent of transverse ply thickness because of these blunting effects. The relation
between unnotched post-fatigue strength and transverse ply thickness of carbon/epoxy is the
mirror-image due to the difference in the degree of anisotropy.
The Kortschot, Spearing, Beaumont model which has considerable success with carbon/epoxy
has limited application to glass/epoxy due to the development of complex damage modes in
the vicinity of a notch. The evidence presented here indicates that the breakage of glass fibre
which results from transverse ply cracking is not confined to the interlaminar region but
develops through the thickness of the ply. This substantially compromises the load-carrying
capability of the laminate. Emphasis must shift , therefore, from modelling the geometry of
II - 351
the damage zone to include in the model a component for the strength degradation of the 0
ply and by taking into account the thickness of the transverse ply, also.
ACKNOWLEDGEMENTS
RAD acknowledges the financial support of the EPSRC in the form of a CASE Studentship.
We acknowledge the valuable discussions with Professor M. T. Kortschot, Professor S. M.
Spearing, Dr P. A. Smith and Professor M. F. Ashby throughout this and earlier work.
REFERENCES
1.
Kortschot, M.T., Beaumont, P.W.R., "Damage Mechanics of Composite Materials - Parts

1-4", Composites Science and Technology 39 (1990) 289-301; 303-32; 40 (1991) 147165; 167-179.
2.
Kortschot, M.T., Beaumont, P.W.R., "Damage-Based Notched Strength Modelling",

ASTM STP 1110 Composite Materials: Fatigue and Fracture (3rd Volume) Editor:
O'Brien, pp 596-616 (1991).
3.
Spearing, S.M., Beaumont, P.W.R., Ashby, M.F., "Fatigue Damage Mechanics of

Notched Graphite-Epoxy Laminates", ibid pp 617-637 (1991).
4.
Spearing, S.M., Beaumont, P.W.R., "Fatigue Damage Mechanics of Composite Materials

- Parts 1-4", Composites Science and Technology 44 (1992) 159-168; 169-177; 299-307;
309-317.
II - 352
IMPROVED FRACTURE MECHANICS PREDICTIONS

FOR DELAMINATION OF COMPOSITES
ACCOUNTING FOR YIELD ZONES
Z. Petrossian and Michael R. Wisnom
Department of Aerospace Engineering, University of Bristol, Bristol, BS8 1TR, U.K.
SUMMARY: Analysis was carried out on a composite specimen containing discontinuous

plies, previously tested in three-point bending. The locations of delaminations from the block
of discontinuous plies were determined using the strain energy release rates from a finite
element analysis. A mixed-mode failure criterion, based on the strain energy release rate,
grossly overestimated the failure loads of the specimen under two different ratios of bending
to out-of-plane shear loading. An improved approach, also based on the mixed-mode failure
criterion but which took account of yielding in the resin at the onset of delamination, was able
to predict the failure loads to within 3% of the experimental values.
KEYWORDS: delamination, fracture mechanics, strain energy release rate, finite element
analysis, discontinuous plies, through thickness stress, plastic zone, initial defect
INTRODUCTION
Fibre-reinforced composite materials have lower transverse and interlaminar strengths
compared to fibre-direction strengths. This means interlaminar failure can occur when the
interlaminar stresses are much lower than the in-plane stresses. Significant interlaminar
stresses can arise, for instance, at the ends of discontinuous plies. Such discontinuities are
found in composite structures when plies are dropped off within a laminate to taper the
thickness. Interlaminar shear stresses can arise when out-of-plane loading is applied to
laminates. Helicopter rotor blades, aircraft wing boxes and floor beams are examples of
structures susceptible to both out-of-plane loading and bending moments, which contain
dropped plies. Bending moments cause local interlaminar stresses at discontinuous plies, and
out-of-plane loading causes overall shear stresses. The combination of these two stress
distributions can lead to cracks initiating at the ends of the discontinuous plies and
propagating along the resin rich layers between the continuous and discontinuous plies. These
cracks or delaminations can greatly reduce the strength and stiffness of the laminate, which in
turn can lead to catastrophic failure of the whole structure. It is therefore very desirable to be
able to predict the loading conditions at which delamination occurs.
Delamination is often predicted using the strain energy release rate, G. This approach was
used to predict delamination from the free edges of laminates in [1]. Furthermore, G is usually
calculated from finite element analysis via the virtual crack closure technique. This was done
to investigate delamination from the ends of discontinuous plies in [2,3]. In [2] the laminates
were in four-point bending and in [3] pure axial tension was applied. In [4] a stress based
criterion was used to predict the initiation and location of damage in a composite specimen in
II - 353
pure bending. Fracture mechanics was then used to predict the growth of delamination from
the initial damage. Finally, in [5] fracture process zones at the free edge of laminates were
taken into account. This enabled the initiation and growth of delamination to be predicted
without any assumed initial defect.
In this paper, delamination from the ends of discontinuous plies is predicted with fracture
mechanics taking into account the material yielding which occurs before crack propagation.
The results of this improved approach are then compared to predictions which do not take
account of the yielding, and to experimental data.
In [6] an account is given of the testing of straight unidirectional laminates in three-point
bending, shown schematically in Fig. 1. The laminates consisted of 32 plies of E glass/ epoxy
913 and were nominally 10mm wide. Four of the plies near the tension surface were cut
across the complete width. The bending moment caused local interlaminar stresses to arise
around the cut. The shear force caused overall interlaminar shear stresses to arise in the
section of the specimen between the loading rollers. The combination of these two stress
distributions caused cracks to initiate at the cut and propagate along the interfaces of the
continuous and cut plies, as seen in Fig. 1. Thus, specimen failure was due to delamination at
the top left and bottom right of the cut. The ratio between the overall and local stresses was
varied by changing the axial distance, d, between the cut and the central loading roller. The
test results indicated a strong interaction between the two types of stress distributions. It was
thought that this might be because the overall interlaminar shear stresses enlarged the zones of
yielding in the resin rich layers between plies where the delaminations initiated.
When yielding occurs in the resin, the plies on either side are able to move relative to each
other. Hence, strain energy can be released even when no physical crack is present. Thus, the
larger the yield zone, the longer the effective crack length [7]. This is similar to the Irwin
approach where the crack length is assumed to equal its physical length plus the length of the
yield zone ahead of the crack tip. The yield zone length is estimated assuming linear-elastic
material behaviour. His approach was developed for predicting crack propagation in metal
plates with an initial crack. In this paper it is adapted for predicting delamination in composite
laminates with no initial delamination.
Cracks Propagating from Cut
Cut
Continuous
Plies (22)
a
Cut
Plies (4)
Continuous
Plies (6)
20mm
20mm
Fig. 1: Schematic diagram of specimen with cut plies in three-point bending

The specimens described above are analysed. Two positions of the cut are studied; d = 5mm
and d = 15mm. A finite element model is created and the strain energy release rates of various
II - 354
positions and combinations of cracks is evaluated. From the curves of G with respect to crack
length, the locations of the cracks are predicted. The failure load is then predicted both with
and without taking the yield zones into account. Both predictions are then compared to the
experimental results.
FINITE ELEMENT MODEL

A two-dimensional finite element model was created using CPS4 elements in ABAQUS [8].
These are four-noded plane-stress elements. A plot of the mesh can be seen in Fig. 2, together
with the boundary conditions corresponding to d = 5mm. For d = 15mm the same model was
used but the boundary conditions were moved left along the mesh by 10mm. Hence, the
central loading point was slightly to the right of the centre of the model when d = 5mm, and
slightly to the left when d = 15mm. This is exactly what was done during the tests. Since there
were stress singularities present at the top and bottom of the cut a fine mesh was created
around these two points. The mesh at the cut can be seen more clearly in Fig. 3. The cut is
shown by the bold vertical line and four possible delamination paths are indicated with dashed
horizontal lines. At the cut the elements are one ply thick and about 0.031mm long, and the
length gradually increases moving away from the cut. There are dual coincident nodes along
the possible crack interfaces and ends of the block of cut plies. During the testing of the
specimens [6], the resin between the ends of the cut plies was seen to crack well before
delamination occurred. Thus, the nodes on either side of the cut are not held together. Each
pair of nodes along the crack interfaces is initially held together by one stiff spring acting in
the horizontal direction and another acting in the vertical direction. This is to enable the
virtual crack closure technique to be used. There are also gap elements along the interfaces to
stop the two surfaces from crossing once the springs are removed.
Fig. 2: Finite Element Model with Boundary Conditions Corresponding to d = 5mm
II - 355
A downward vertical concentrated load per unit width, P, is applied to the node at the point of
contact of the central roller. The node at the point of contact of the left-hand roller is
constrained to have zero x and z translation. The node at the point of contact of the roller on
the right is constrained to have zero z translation. P is equal to the average experimental
failure load per unit width. The value of P is 432.4 N/mm for d = 15mm, and 284.0 N/mm for
d = 5mm. Refer to Table 1 for the material properties and dimensions of the model. The
nominal value of the specimen thickness was initially assumed in the finite element analysis.
From the corresponding results, the crack locations were determined. For the failure load
predictions, it was deemed more accurate to use the average measured thickness of the
specimens and to correct the axial Youngs modulus for the lower fibre volume fraction.
However, the change in the values of G were small enough not to affect the predicted crack
locations.
Fig. 3: Close-Up of Mesh showing Cut and Possible Locations of Delaminations
II - 356
Table 1: Material Properties and Dimensions.

Axial Youngs Modulus, Ex, for Nominal Thickness
43.9 GPa
Axial Youngs Modulus, Ex, for Measured Thickness
41.2 GPa
Through-Thickness Youngs Modulus, Ez
15.4 GPa
Shear Modulus, Gxz
4.34 GPa
Poissons Ratio, xz
0.3
Length of Specimen
80 mm
Nominal Thickness
4 mm
Measured Thickness
4.33 mm
Width
10 mm
Distance between Outer Rollers
40 mm
DETERMINATION OF CRACK LOCATIONS BY TOTAL ENERGY APPROACH

The total strain energy release rate, GT, was firstly found for individual cracks at each of the
four possible crack locations, seen in Fig. 3. GT was then found for cracks propagating
simultaneously along all four crack paths, and finally for cracks propagating simultaneously
along the top left and bottom right interfaces.
GT was calculated from the change in total strain energy in the model as the crack was
allowed to propagate by gradually removing the horizontal and vertical springs along the
interfaces. This was done at each increment of crack length, a. The schematic definition of a is
given in Fig. 1.
Graphs of GT against a were plotted for all the cases and are given together in Fig. 4 for d =
15mm, and Fig. 5 for d = 5mm. It can be seen that in every case the value of GT decreases
with crack length until it reaches a minimum value, Gmin. After this, the value of GT generally
grows with crack length. This means, as the load is increased, the cracks will propagate in a
stable manner up to a short crack length (corresponding to the minimum value of Gmin) and
then crack propagation will be unstable. This behaviour was observed in the tests in [6].
Therefore, catastrophic failure occurs when Gmin reaches the critical value, Gc.
II - 357
Total Strain Energy Release Rate (N/mm)
2
Top Left & Bottom Right
Bottom Right
Top Left
All Locations
Bottom Left
Top Right
1.5
0.5
0
0
0.5
1.5
Crack Length (mm)
Fig. 4: Strain Energy Release Rate with Respect to Crack Length for
Different Positions and Combinations of Cracks (Cut 15mm from Central Roller)
Total Strain Energy Release Rate (N/mm)
1.5
1
Top Left & Bottom Right
Bottom Right
Top Left
All Locations
Bottom Left
Top Right
0.5
0
0
0.5
1.5
Crack Length (mm)
Fig. 5: Strain Energy Release Rate with Respect to Crack Length for
Different Positions and Combinations of Cracks (Cut 5mm from Central Roller)
II - 358
Since catastrophic failure depends on the value of Gmin, the locations of the delaminations at
failure can be determined by comparing the values of Gmin corresponding to the different
cases studied. The values of Gmin are given in Table 2. Referring to the table, it is clear that
for d = 15mm the most likely type of failure is by delaminations propagating simultaneously
from the top left and bottom right of the cut, since the value of Gmin is the highest for this
situation. For the d = 5mm case, Gmin is highest when a crack propagates from the bottom
right corner only. However, cracks from both the bottom right and top left of the cut were
observed in the tests.
Although cracking is primarily mode II due to the type of loading, a small amount of mode I
is generally present. Furthermore, the difference between the values of Gmin for cracking at
the bottom right only, and simultaneous cracking at the bottom right and top left is not very
great. Thus, the fact that cracking occurred simultaneously at the bottom right and top left,
even when the value of GT was lower in the d = 5mm case, can be explained by more mode I
being present. It was therefore concluded that the mode ratio would have to be taken into
account for accurately predicting the failure loads.
Table 2: Minimum Values of GT for Various Cracks (From Figures 4 and 5)

d = 15mm
d = 5mm
Crack at Top Left Only
0.34 N/mm
0.67 N/mm
Crack at Top Right Only
0.00 N/mm
0.00 N/mm
Crack at Bottom Right Only
0.39 N/mm
1.03 N/mm
Crack at Bottom Left Only
0.00 N/mm
0.00 N/mm
Cracks at All Locations
0.23 N/mm
0.63 N/mm
Cracks at Bottom Right & Top Left
0.45 N/mm
0.93 N/mm
AILURE LOAD PREDICTION USING STANDARD MIXED MODE APPROACH

Cracks were assumed to propagate simultaneously (and at the same rate) along the top left and
bottom right interfaces, for both cut positions. The following mixed-mode criterion was used
to predict the failure loads:
GI / GIc + GII / GIIc = 1
(1)
where GI and GII are the mode I and mode II components of the strain energy release rate,
respectively, and GIc and GIIc are their critical values. For E glass/ epoxy 913, GIc = 0.25
N/mm and GIIc = 1.08 N/mm, [9].
GI and GII, were found with respect to the crack length, a, using the virtual crack closure
technique. This assumes the strain energy released when a crack propagates is equal to the
energy required to close the crack. GI was calculated from the spring forces and relative nodal
displacements in the vertical (z) direction, and GII was calculated from the spring forces and
relative nodal displacements in the horizontal (x) direction. GI and GII were then each
normalised by their corresponding critical values.
II - 359
The graphs of normalised G against a for the two cut positions (d = 15mm and d = 5mm) can
be seen in Figures 6 and 7. One can see the sum of the normalised strain energy release rates,
(GI/GIc + GII/GIIc), decreases to a minimum value as the cracks start propagating. Then, after a
crack length between 0.2mm to 0.3mm, the value starts to increase again. As explained in the
previous section, this means the load at catastrophic failure is governed by the minimum value
of (GI/GIc + GII/GIIc). At the average experimental failure loads, the minimum values of
(GI/GIc + GII/GIIc) are 0.61 for the d = 15mm case, and 0.83 for the d = 5mm case. Thus,
according to Eqn 1, both the predicted failure loads are higher than the experimental failure
loads. The strain energy release rate is proportional to the square of the applied load, so the
predicted failure loads per unit width are:
P = 553 N/mm
for d = 15mm
P = 312 N/mm
for d = 5mm.
and
Normalised Strain Energy Release Rate
GI / GIc
GII / GIc
1.5
GI/GIc+GII/GIIc
1
0.5
0
0
0.5
1.5
Crack Length (mm)
Fig. 6: Normalised Strain Energy Release Rate at

Experimental Failure Load (Cut 15mm from Central Roller)
II - 360
Normalised Strain Energy Release Rate
2.5
GI / GIc
2
GII / GIc
GI/GIc+GII/GIIc
1.5
0.5
0
0
0.5
1.5
Crack Length (mm)
Fig. 7: Normalised Strain Energy Release Rate at

Experimental Failure Load (Cut 5mm from Central Roller)
IMPROVED FAILURE LOAD PREDICTION

The axial distance from the cut to which the zones of yielding material extended was
estimated from the finite element analysis. This was done at the top left and bottom right of
the cut. The resin along the interfaces of the cut and continuous plies was assumed to yield
when the interlaminar shear stress exceeded 70 MPa. This value was estimated from the
measured in-plane shear stress-strain response of glass/ epoxy 45 specimens in tension [10],
and assuming transverse isotropy. The yield zone length at the top left was averaged with the
length at the bottom right. This was the effective initial crack length. The corresponding value
of (GI/GIc + GII/GIIc) was found by interpolation. If the value of (GI/GIc + GII/GIIc) was less
than one at the effective initial crack length, it meant the predicted failure load was higher
than the experimental failure load, and vice versa. If this was the case, the value of (GI/GIc +
GII/GIIc) and the yield zone lengths were recalculated at higher (or lower) values of P. No
further finite element calculations were necessary since the shear stress was proportional to P,
and the value of (GI/GIc + GII/GIIc) was proportional to P2. After a small number of iterations,
the predicted failure load converged. The values for d = 5mm and d = 15mm are given in
Table 3, alongside the previous predictions and the average experimental values. The values
of the yield zone lengths in the d = 5mm case were 0.70 mm at the top left and 0.48 mm at the
bottom right of the cut. The values for d = 15mm were 1.57 mm at the top left and 0.38 mm at
the bottom right. Although there was a big difference between the latter values, averaging
them together was considered to be the best solution for the purposes of keeping the approach
manageable.
II - 361
Note that both the strain energy release rate and the yield zone lengths were calculated
assuming linear-elastic material properties. The estimated yield zone lengths would be longer
if non-linear stress-strain behaviour was taken into account. However, the strain energy
release rate would be lower since yielding limits the stresses at the crack tip. Thus, to some
degree these two effects cancel each other out. Also, to accurately predict the yield zone
length and corresponding strain energy release rate, the resin along the interfaces would have
to be modelled using non-linear material properties. This would require much more
computational time, and for the purpose of engineering design, the approach used above is
thought to be reasonable.
Table 3: Failure loads (per unit width)
d = 15mm
d = 5mm
Experimental Average
432.4 N/mm
284.0 N/mm
Prediction (Standard Mixed-Mode)
553 N/mm
312 N/mm
Prediction (Including Yield Zones)
444 N/mm
280 N/mm
DISCUSSION
The analysis showed that when no account is taken of yielding, linear-elastic fracture
mechanics is not able to accurately predict delamination in laminates with significant overall
interlaminar shear stresses. Referring to Table 3, it can be seen that the failure loads predicted
without accounting for yielding at the onset of delamination are both very unconservative.
The prediction for the d = 15mm case is 28% higher than the experimental value, and the
prediction for the d = 5mm case is 10% higher. The prediction corresponding to the d = 15mm
case is worse because the ratio between the shear force and the bending moment is higher
when the cut is further from the central roller. Hence, the relative importance of the overall
interlaminar shear stress is greater, and the yield zones are generally longer when d = 15mm.
When the improved approach is applied to the two cases, the predictions are within 3% of the
experimental results. Therefore, the improved approach was able to accurately predict the
failure loads of laminates with two different levels of yielding.
Previously it was found that a linear interaction equation between the effects of overall
interlaminar shear stresses and local stresses at the cut gave a good prediction of failure [6].
The results presented here explain the interaction in terms of the increased effective crack
length as the overall interlaminar shear stresses become high compared with the yield stress.
CONCLUSIONS
A composite specimen containing cut plies under three-point bending was analysed. The ratio
between the bending moment and the shear force at the cut was varied by changing the
position of the cut relative to the applied loads. The strain energy release rate was calculated
for delaminations spreading from the cut using finite element analysis. A mixed-mode failure
criterion, based on the strain energy release rate, grossly overestimated the failure loads for
both cut positions. An improved approach, also based on the mixed-mode failure criterion but
which took account of yielding in the resin at the onset of delamination, was able to predict
II - 362
the failure loads to within 3% of the experimental values. This shows the importance of
considering the yielding that occurs before crack formation when predicting delamination in
laminates under high overall interlaminar shear stresses.
ACKNOWLEDGEMENTS
The support of Westland Helicopters and the UK Engineering and Physical Sciences Research
Council is gratefully acknowledged.
REFERENCES
1.
OBrien, T. K., Characterization of Delamination Onset and Growth in a Composite

Laminate, Damage in Composite Materials, ASTM STP 775, K. L. Reifsnider, Ed.,
American Society for Testing and Materials, 1982, pp. 140 - 167.
2.
Wisnom, M. R. and Jones, M. I., Delamination of Unidirectional Glass Fibre-Epoxy

with Cut Plies Loaded in Four Point Bending, Journal of Reinforced Plastics and
Composites, Vol. 14, 1995, pp. 45 - 59.
3.
Harbert, S. J. and Hogan, H. A., An Analysis of Curvature and Layup Effects on

Delamination in Notched Composite Beams, Journal of Reinforced Plastics and
Composites, Vol. 11, 1992, pp. 443 - 457.
4.
Minguet, P. J. and OBrien, T. K., Analysis of Composite Skin/Stringer Bond Failure

Using a Strain Energy Release Rate Approach, Proc. Of ICCM-10, Whistler, B.C.,
Canada, August, 1995, pp. 245 - 252.
5.
Amrutharaj, G. S., Lam, K. Y. and Cotterell, B., Delaminations at the Free Edge of a
Composite Laminate, Composites, 27B, 1996, pp. 475 - 483.
6.
Wisnom, M. R. and Jones, M. I., Delamination due to Interaction between Overall

Interlaminar Shear and Stresses at Terminating Plies, Composite Structures, Vol. 31,
1995, pp. 39 - 47.
7.
Wisnom, M. R., Shear Fracture of Unidirectional Composites without Initial Cracks,

Composites Science and Technology, Vol. 52, 1994, pp. 9 - 17.
8.
ABAQUS, Hibbitt, Karlsson and Sorensen Inc., 100 Medway Street, Providence, R.I.,
U.S.A.
9.
Partridge, I., Cranfield University. Private communication.
10.
Wisnom, M. R., The Effect of Fibre Rotation in 45 Tension Tests on Measured

Shear Properties, Composites, Vol. 26, No. 1, 1995, pp. 25 - 32.
II - 363
DAMAGE MECHANICS OF REINFORCEMENT

CRACKING IN PARTICLE OR SHORT-FIBER
REINFORCED COMPOSITES
Keiichiro Tohgo, Young-Tae Cho and Hitoshi Ishii
Department of Mechanical Engineering,
Shizuoka University, 3-5-1, Johoku, Hamamatsu 432, Japan
KEYWORDS: damage mechanics, reinforcement cracking, particle or short-fiber reinforced

composites, broken ellipsoidal inhomogeneity, load carrying capacity, equivalent inclusion
method, mean field theory, micromechanics
II - 364
II - 365
II - 366
II - 367
II - 368
II - 369
II - 370
II - 371
II - 372
II - 373
PARALLEL COMPUTING AND DAMAGE IN

COMPOSITE STRUCTURES
F. Devries, F. Ln, G. L'Hostis, D. Soulat
Laboratoire de Modlisation et Mcanique des Structures, UPMC / ENSAM / ENS Cachan U.R.A. 1776 C.N.R.S., Universit Pierre et Marie Curie - Tour 66, 5eme tage, 4,
place Jussieu - 75252 Paris cedex 05, France
SUMMARY: A model for the damage's evolution in composite materials is here proposed. It
consists in a microscopic/macroscopic approach requiring the solution of two uncoupled
steps. The first one consists in the computation of the composite's homogeneous equivalent
behaviors, damaged by defects postulated at the microscopic composite's level. The second,
consists then in solving a nonlinear, time-dependent, macroscopic problem.
These steps require unfortunately a lot of computations whose numerical costs may be
prohibitive. It is here proposed to reduce them, thanks to the development of a strategy
allowing to take benefit from the parallel computers performances.
Starting from two three-dimensional, industrially-realistic examples, one shows that this
methodology may be fruitfull if one knows how to choose a subdomain decomposition, taking
in account the damages and the heterogenities of the composite's representative volume
element.
KEYWORDS: damage, thermoelasticity, homogenization, subdomain decomposition

techniques, parallelism
INTRODUCTION
This paper deals with the simulation of damage's propagation in composite structures. To this
way we use an homogenization technique which leads to a microscopic/macroscopic
approach. To this purpose we first postulate damage criteria in the composite representative
volume element. These criteria introduce parameters which are chosen as microscopic
damage parameters. The use of the homogenization technique, allows then to construct a
damaged macroscopic medium model, based on a microscopic approach of a body weakened
by microdefects. The damage's evolution may be then modelled and computed inside an
adjusted thermodynamic framework.
Unfortunately this methodology requires a lot of computations, required for the construction
of composite homogenized behaviors connected to each of its microscopic damaged
configuration. In addition to the three-dimensional geometry, a good description of the
damage process needs the choice of a lot of microscopic damage parameters and a lot of
values for these parameters. Thus, this methodology requires large computer CPU times and
large memory spaces, and comes up against the limitations prescribed by the actual computers
technology.
II - 374
A way to get round these overcosts consists in using multiprocessor computers. We propose
thus in this communication to show how one can use homogenization techniques and
substructuring techniques to take advantage of the parallel computers performances.
In order to assess the benefit of this methodology, we consider two kinds of three-dimensional
composites, used at the industrial level as covering materials for aeronautical engines. The
first one is made up of glass microballoons coated in a resin (syntactic foams). The observed
damage process of this composite consists of the implosion of the balloons. The second
example concerns sewed layers of unidirectional composites. Due to the different expansion
coefficients of these materials, when the structure undergoes a high temperature increase,
damage occurs under the form of delaminations between constituents.
A MODEL FOR DAMAGE'S EVOLUTION SIMULATION IN COMPOSITE

STRUCTURES
In order to simulate the damage's evolution in composite structures, we use hereafter a
microscopic/macroscopic approach of a macroscopic medium weakened by microdefects. To
this purpose we classically consider two scales: the microscopic scale is attached to the
representative volume element of the composite and is thus representative of microscopic
effects such damages; its basis is denoted by (0, y1, y2, y3); at the macroscopic level any point
of the structure is marked with respect of a system of axis denoted by (0, x1, x2, x3).
This approach consists first in postulating damages at the microscopic level. The connected
damaged homogeneous behavior laws are then computed thanks to the homogenization
principle. In order to simulate the damage's evolution in the composite structures under
consideration, it remains lastly to solve, at the macroscopic level, a nonlinear, time-dependent
problem .
Computation of Homogenized Damaged Behaviors
We consider in what follows that the composite under consideration admits a geometric and
mechanic periodicity, such that this material may be defined by the knowledge of its period or
basic cell, denoted by Y. In order to describe the damage's evolution in this material, let us
consider its basic cell as a real structure containing defects, as for example stress-free holes
(denoted by C). Let us associate to these defects, parameters, denoted in vectorial notations by
D. Thanks to the homogenization process for periodic media [1], [2], one can associate to this
microscopic description of the damaged composite material a homogeneous equivalent
behavior. To this way and when a thermo-elastic behavior for the constituents is chosen, we
have to find the microscopic deviation in temperature and the microscopic displacement
field u, solutions in the elastic part of Y (denoted by Y*) to the following problems, also
known as cellular problems [3], [4]:
II - 375
Elasticity cellular problem:
(1)
divy = 0 , = a [ e(u) - (T - T0 ) ] in Y*
u - E y Y-periodic , (n) Y-antiperiodic
(n) = 0 on 0C , << e(u) >>Y = E
Thermal conduction cellular problem:
(2)
*
divy q = 0 , q = - y
in Y
- F y Y-periodic , q . n Y-antiperiodic
q . n = 0 on 0C , << y >> Y = F
In these problems divy and y respectively denote the divergence and gradient operators with
respect to the microscopic coordinates y. We have also used the following notations:
1
1
e(u) = ----- ( y u + t y u) , << y f >>Y = ----- f n dS
-
2
|Y|
(3)
Y-periodic admits the same value at points belonging to opposite faces of Y

and respectively denoted by , a, , , c the conduction, stiffness, expansion tensors, the mass
density and the specific heat, and introduced the fields E, F, T which represent the
macroscopic strain tensor, the macroscopic temperature gradient tensor and the macroscopic
temperature (T0 is the reference macroscopic temperature).
Once these problems have been solved, one may then construct the damaged homogenized
behavior sought, throught the computation of the stiffness, conduction and expansion
homogenized tensors Qh(D), h(D), h(D) which connect the microscopic fields to the
macroscopic ones by the relations:
(4)
< > = = Qh (D) E - Qh (D) h (D) (T - T )
Y
0
h
< q > = S = - (D) F
with
1
< f >Y = -------- f(y) dy
|Y|
*
Y
and the homogenized specific heat denoted by Ch(D).

Let us lastly specify that thanks to the linearity of these problems with respect to the
macroscopic data (strain E, temperature deviation (T-T0), temperature gradient F) the
solutions to problems (1) and (2) are usefully splitted in the following forms:
II - 376
(5)
u = u k h Ekh + w (T - T0 ) (k, h {1, 2, 3})
k
= F
(k {1, 2, 3})
(with uk h = u h k thanks to Ekh = Ehk )
allowing to solve once and for all these problems, undependently from the macroscopic fields
which have to be now computed.
Damage's Evolution
Once previous step has been carried out, one knows each of the homogenized equivalent
behaviors of the composite material under consideration, corresponding to given values of the
damage parameters considered at the microscopic level.
For a given thermo-mechanical loading prescribed to the composite structure (which occupies
the volume ), the damage's evolution is then modelled by introducing the macroscopic free
energy density defined by:
h
h
1
1 < > Y C (D )
h
h
(6) < >Y (E, T, D) = ----- E Q (D) E - Q (D) (D) E (T - T0 ) - ----- -------------------------------- (T - T0 )2
2
2
T0
where the vector D (n entries assumed) is constituted by the damage parameters.
The volume dissipation is then given by:
n
(7)
dDi S
Fi ------- - ----- x T
dt T
i =1
where we have denoted by Fi the thermodynamical damage force associated to the damage
parameter Di, defined by:
(8)
0{< >Y (E, T, D) }

Fi = - -----------------------------------------------------0Di
(i = 1, 2, , n)
A useful means for satisfying the positivity of the dissipation (7) consists in using the
thermodynamic framework of standard generalized materials [5]. Assuming the classical
decoupling between the mechanical and thermal dissipations and starting from the existence
of pseudo potentials of dissipation denoted by th(x T) and mec(F1, , Fn) which are
chosen as positive or zero functions, convex and zero in 0, this framework allows one to
obtain the complementary laws under the forms:
th
(9)
mec
0 ( x T)
dDi 0 (F1 , , Fn )
S
----- = - ------------------------- , ------- = ------------------------------------------0Fi
T
dt
0( T)
(i = 1, , n)
The choice of these potentials has to be a posteriori justified by collation of results with
experimental data. In what follows we choose the thermal potential in order to obtain the
II - 377
classical Fourier's law for the temperature. Concerning the mechanic potential, it may be for
instance chosen as follows:
n
i
mec
1 + i
(10) (F1 , , Fn ) = ------------- | F i |
( i > 0, i > 0 i {1, , n} )
i =1 1 +
i
This particular choice leads to a damage's evolution law of Norton's kind [6].
Once these choices have been made the evolution of damages will be obtained by the solution
of the following macroscopic problem:
for which it is required to specify boundary and initial conditions.

The damage's evolution may be thus computed in the two steps given above which consists
in:
1.
constructing the homogeneous equivalent damaged behaviors connected to chosen
values of microscopic defects;
2.
solving the macroscopic time-dependent problem describing the damages evolution.
THE SCHUR COMPLEMENT METHOD FOR THE SOLUTION OF CELLULAR

PROBLEMS
The method explained above for the computation of damage's evolution in composite
structures has been already successfully used for the study of damage by fibers breaks in
unidirectional composites [7]. It requires however the solution of the cellular problems (1), (2)
for each set of values of the damage parameters postulated at the composite's microscopic
level. These computations may induce very large CPU times in the case where, like here, the
geometry is three-dimensional. This is the reason why we have developed a parallel way for
these solutions; it consists in using a subdomain decomposition technique (Schur complement
method) and allows to take benefit from the parallel computers performances.
In order to present the application of this technique to the solution of cellular problems, let us
consider the thermal cellular problems (2). Thanks to the decomposition (5) it may be easily
shown that the fields k are the solutions to the following problems:
II - 378
Let us then introduce a subdomain decomposition of Y into 2 subdomains (for reader's

convenience), denoted by (1)Y and (2)Y. We denote by the common boundaries of these
subdomains and the external boundary of Y (on which periodicity conditions for tk are
prescribed).
Let us next introduce the fields (i)k, (i)qk solutions in the subdomain (i)Y to the following
problem:
where we have denoted by (i)n the unit outward normal to (i)Y and introduce the (not yet
known) temperature field which has to be characterized.
These fields will coincide with the restrictions of k and qk in (i)Y if and only if they satisfy
the continuity of normal heat flux on :
(14)
(1)
(1)
n=
(2)
(2)
on
In order to deduce the problem whose solution is , let us use the linearity of problem (13)
and thus split its solution as follows:
(15)
(i) k
(i)
(i)
= + ,
(i) k
(i)
q = +
(i)
where (i)k and (i)k are the solutions to:
II - 379
Problem (16) may be solved once for all. Let us associate to its solution the normal heat flux
on , defined by: (i)bk = (i)k . (i)n .
Next, let us introduce the linear operator (i)S, defined on the subdomain (i)Y, which associates
to the prescribed (but unknown) temperature of problem (17) its connected normal heat flux
on :
(18)
(i)
S : (i)S() = [ (i) .
(i)
n ] |
Thanks to these quantities the normal heat flux connected to the solution of problem (13)
satisfies:
(19)
[(i)qk .
(i)
n] | =
(i)
S() + (i)bk
The continuity condition (14) leads then to the interface problem to be solved on which
writes:
(20)
Find such that:
(1)
S() + (2)S() = - ((1)bk + (2)bk )
The symmetric, definite positive operator S = (1)S + (2)S is called the Schur complement
operator [8]. It determines the problem to be solved at the interface between the
subdomains. This problem is classically solved by use of the iterative preconditioned
conjugate gradient algorithm, whereas the problems to be solved in each subdomain
(problems 16 and 17) are fruitfully solved by a direct method, such the Cholesky method.
As mentioned above the use of this method for the "in parallel" solution of the cellular
problems requires the choice of a subdomain decomposition for the composite's basic cell.
Many subdomain decomposition procedures are available [9], [10]. However, in order to
obtain an efficient speed-up, this choice has to almost satisfy some principles, such the
minimization of the interface problem size and the load-balancing of the tasks on each of the
processors used. We shall see hereafter that mechanical phenomena complicate this choice.
II - 380
APPLICATION OF PARALLELISM
In order to illustrate the means given above for the parallel simulation of damage's
propagation in composite structures, we consider two kinds of three-dimensional composites,
used at the industrial level as covering materials for aeronautical engines. These materials and
the damages which are experimentally observed are now briefly presented.
Syntactic Foam
The first material we consider is a syntactic foam. It consists of an isotropic resin into which
nine glass balloons are arranged in a cubic centered network (see fig. 1). Their internal
boundaries are assumed to be stress free.
B3
B1
B2
B
B6
Fig. 1: Geometric arrangement of the syntactic foam and subdomain decomposition
The experimentally-observed damage process in this composite consists of the implosion of

the microballoons, and thus their disappearances. In order to simulate this kind of damage we
attach to each (say Bi) of these balloons a scalar denoted by Di which take the value 1 or 0
depending on whether the balloon is broken or not.
The parallel solution of the cellular problems has been carried out thanks to a decomposition
in 8 subdomains shown in figure 1. This choice has been directly induced by the geometry of
the basic cell. Indeed, it insures the load-balancing of tasks on each of the 8 processors which
are allocated to each of the 8 subdomains, and minimize the size of the interface problem to
be solved.
The characteristics of this decomposition are presented in figure 2, which points out the
number of degrees of freedom (Finite Element Method) for the global and interface problems
corresponding to each of the damaged configurations of the composite's basic cell, in the case
of the elasticity cellular problems (1).
II - 381
8
7
Interface problem - Central balloon broken
Interface problem - Central balloon not broken
Global problem - Central balloon broken
5
4
3
2
1
0
0
Number of broken corner-balloons
Global problem - Central balloon not broken
5000
10000
15000
20000
25000
30000
Number of degrees of freedom
Fig. 2: Size of the damaged elastic problems

Thanks to this decomposition, the parallel solution of problems (1) on a MIMD machine
(KSR1), leads to times savings which appear very good, as figure 3 attests (the C.P.U. times
given in this figure correspond to the solution of one cellular problem (1)) .
8
Times in parallel
Times in sequential
Number of broken corner balloons
6
5
4
3
2
1
0
0
50 100 150 200 250 300 350 400 450 500 550 600
C.P.U. Times
Fig. 3: C.P.U. times for the solutions of one elastic cellular problem
Indeed, from these results, it appears that the times speed-up is around 7 and thus very close
to the optimal and maximal one which is equal to 8.
Sewed Layers of Unidirectional Composites
The second composite we have considered, consists in two layers of unidirectional
composites, whose orientations are respectively equal to 0 and 90 degrees, held in contact
(one hopes so) by a third constituent, perpendicular to them, called in the sequel the "picot".
Due to the different expansion coefficients of the materials of this basic cell, when the
structure undergoes a thermo-elastic loading, the damage occurs under the form of
debondings between the constituents. We postulate thus three kinds of defects which are
modelled by cavities (denoted by Ci in figure 4) with zero thickness and whose lengths are
denoted by Li . These cavities respectively model the debondings appearing at the interface
between the layers and the picot and between the layers themselve.
II - 382
picot
y
debonding between
3
the layer and the picot
C2
picot
y3
debonding between
the layers:
C
3
layer n 2 ; orientation: 90
h
layer n 1 ; orientation: 0
y2
y
debonding between the layer and the picot:
2
1
C1
Fig. 4: Basic cell of the sewed composite - Location and geometric modelization of damages
We associate to these defects three damage parameters defined as follows:
L1
L2
L3
(21) D1 = ----D2 = ----D3 = ----h1
h2
R
where h1, h2 and R respectively denote the thicknesses of the layer number 1, number 2 and
the radius of the picot.
Unlike the syntactic foam considered before, the choice of a subdomain decomposition for the
basic cell of this composite is not straightforward, because of its geometry and the presence of
damages. This is the reason why we have considered two subdomain decompositions. The
first one consists of 20 subdomains whose boundaries coincide with the debondings between
the composite's constituents (the degrees of freedom located at the debondings belong thus to
the problem to be solved at the interface ). The second one consists of 8 subdomains whose
boundaries, a priori, do not coincide with the damages localizations. The main disadvantage
of the first decomposition stems from the bad load-balancing it induces, whereas the second
insures a good load-balancing. Let us lastly note that both decompositions lead to some
"floating"subdomains (the local problems such (17) do not admit an unique solution); this fact
requires the use of a particular preconditioner technique. We used to this purpose the "Coarse
Neumann" [11] and "Neumann-Neumann" preconditioners [12].
Fig. 5: Decompositions in 8 and 20 subdomains

The characteristics of both decompositions are presented on the following figure, for two
particular states of damage of the composite's basic cell:
II - 383
25000
D =D =0
1
Number of degrees of freedom
20000
8 subdomains - D 3= 0
8 subdomains - D = 1
3
15000
10000
5000
Global problem
Interface problem
Configurations
Fig. 6: Size of the elastic problems

We report in figure 7 the C.P.U. times consumed for the solution of one elastic cellular
problem by use of these two decompositions and the two preconditioning techniques.
Starting from these results it first appear that, from the point of view of parallelism, the
"Coarse Neumann" preconditioner lead to better times savings than the "NeumannNeumann".
From the times savings point of view, the decomposition in 8 subdomains leads to a speed-up
around 6, whereas the decomposition in 20 conducts to a speed-up which is only around 4
(the best are respectively 8 and 20). These results clearly assert that load-balancing is an
essential "pre-requisite" for the obtention of good times-savings in parallel, even if it is very
difficult to realize this task in case of heterogenous and damaged structures.
Once this choice has been made, the results presented in the case of these two applications,
clearly show that parallelism and the methodology used, are powerful.
CONCLUSION
Thanks to the use of the homogenization technique and a subdomain decomposition
technique, a computational strategy and a parallel numerical tool have been made available
for the simulation of damage's propagation in composite materials. They enable to obtain
appreciate times-savings in most cases. However, the results presented in this paper, clearly
show that the problem of the choice of a well-suited subdomain decomposition for the
composite's representative volume element, remains to be solved, even if the times-savings
we have obtained are very encouraging.
In order to allow for the proposed modelization of damage's evolution in composite materials,
to fully reveal its potentialities, the work, which has began consists in developing a parallel
strategy for the solution of the nonlinear, time-dependent, macroscopic problem.
II - 384
8 subdomains
20 subdomains
1400
1300
In Parallel steps
In Sequential steps (assembling)
1200
1100
(1) Neumann-Neumann sequential

(2) Neumann-Neumann parallel
(3) Coarse Neumann sequential
(4) Coarse Neumann parallel
C.P.U.Times (s)
1000
900
800
700
600
500
400
300
200
100
0
(1)
(2)
(3)
(4)
(1)
(2)
(3)
(4)
Fig. 7: C.P.U. times for the solutions of one elastic cellular problem versus subdomain
decomposition and preconditioning technique
ACKNOWLEDGMENTS
This work has been made available thanks to the collaboration of the I.N.R.I.A. and more
particularly the Modulef department, which has given us the opportunity to use the KSR1
parallel machine.
REFERENCES
1.
Duvaut G. "Analyse Fonctionnelle et mcanique des milieux continus. Application

ltude des matriaux composites lastiques structure priodique - Homognisation".,
Theoritical and Applied Mechanics, WT. Koiter editor, North-Holland, 1976, pp.119132.
2.
Sanchez E. "Non homogeneous media and vibration theory", Lecture Notes in Physics,
127, Springer Verlag, Berlin, 1980.
3.
Francfort G. "Homogenization and linear thermoelasticity", SIAM Journal of

Mathematical Analysis, Vol. 14, 1983, pp. 696-708.
4.
Hashin Z."Thermoelastic properties of fiber composites with imperfect interface",

Mechanics of Materials, Vol. 8, 1990, pp. 333-348.
5.
Halphen B., Nguyen Quoc Son. "Sur les matriaux standards gnraliss", Journal de
Mcanique, Vol. 14, 1985, pp. 39-63.
6.
Friaa A. "Le matriau de Norton-Hoff gnralis et ses applications en analyse limite",

Comptes Rendus de l'Acadmie des Sciences I, Vol. 286, 1978, pp. 953-956.
II - 385
7.
Devries F.. "A fatigue damage model by fibres breaks in unidirectional composites",
Proceedings of the Fourteenth annual Energy-Sources Technology, ASME, Composites
Materials Techonoly, Houston, USA, 1991, pp. 137-144.
8.
Bjordstad P., Wildlund O. "Iterative method for the resolution of elliptic problems on
regions partitioned into substructures", SIAM Journal of Numerical Analysis, Vol. 23,
1986, pp. 1097-1120.
9.
Farhat C. "A simple and efficient automatic F.E.M domain decomposer", Computers
and Structures, Vol. 28, 1988, pp. 579-602.
10.
Malone JG. Automated mesh decomposition and concurrent finite element analysis for
hypercube multiprocessor computer. Computational Methods Applied Mechanical in
Engineering, Vol. 70, 1988, pp. 27-58.
11.
Mandel J. "Balancing domain decomposition", Numerical Methods in Engineering, Vol.

19, 1993, pp. 233-241.
12.
Agoshkov V.I. Poincare's-Stecklov operators and domain decomposition methods.

Proceedings of the Fourth International Symposium on Domain Decomposition
Methods for Partial Equations, Glowinski, Golub, Meurant, Periaux eds., SIAM,
Philadelphie, USA, 1987, pp. 73-112.
II - 386
A PROPER TREATMENT OF INTERFACIAL

CRACKS IN THE PRESENCE OF FRICTION
C. T. Sun and W. Qian
School of Aeronautics and Astronautics, Purdue University,
West Lafayette, Indiana 47907-1282, USA
SUMMARY: Frictional sliding on interface crack surfaces results in weak crack tip stress
singularity and zero strain energy release rate. A new fracture criterion based on finite
extension strain energy release rate is proposed to capture the intrinsic fracture toughness. The
finite extension strain energy release rate is shown to represent the magnitude of the singular
stress field. Numerical simulation of a center crack in a bimaterial infinite media under remote
shear as well as fiber pull-out and push-out are presented to illustrate the frictional effect in
both small and large scale contacts near the crack tip.
KEYWORDS: friction, fracture, interfacial crack, crack tip stress singularity, strain energy
release rate, toughness, fiber pull-out and push-out
INTRODUCTION
The development of classical fracture mechanics is based on the open crack assumption, i.e.,
the crack surfaces are not in contact. That assumption leads to, in linear elastic solids, stress
singularity of inverse square root that enables one to employ either strain energy release rate
or stress intensity factor to characterize fracture toughness and predict crack movement.
When crack surfaces are under combined compression and shear, the effect of sliding
frictional stresses on the crack surfaces cannot always be ignored. Only in the case in which
the neartip contact zone is extremely small, the K-dominated field basically agrees with that
of the open crack solution. This was discussed in detail by Sun and Qian [1] for bimaterial
interfacial cracks under Mode I loading. On the other hand, if the contact zone is finite,
frictional contact of the crack surfaces can no longer be neglected. Most importantly, the
frictional contact alters the stress singularity to weak singularities [2]. Consequently, the
strain energy release rate as conventionally defined vanishes [3] and cannot be used as a
fracture parameter.
Many researchers have attempted to extend the fracture mechanics approach to treat various
engineering problems involving interfacial cracks with frictional contact. Fiber pull-out and
push-out problems have attracted the most attention (e.g., [4-7]). Delamination in thin
film/substrate micro-electronics systems [8] and composite laminates [9] also have been
investigated. Various methods were proposed by these authors to derive the strain energy
release rate or to employ different parameters for characterizing the interfacial crack
toughness. Many of these approaches suffer deficiencies mainly due to the weak singularity
and the energy dissipation associated with the frictional sliding of crack surfaces.
II - 387
In this study, the concept of strain energy release rate for interfacial cracks in the presence of
friction is reexamined. A finite element based numerical procedure is introduced to calculate
the strain energy release rate and energy dissipation due to friction for a finite crack
extension. An explicit relation between the generalized stress intensity factor and the finite
extension strain energy release rate is derived based on the neartip field. Thus, the finite
extension strain energy release rate with a fixed crack extension can be used to represent the
magnitude of the singular stress field and, therefore, to quantitatively characterize the
intrinsic fracture toughness. For numerical examples, a center crack in a plane strain infinite
bimaterial medium under shear loading (see Fig. 1) and fiber pull-out and push-out tests (see
Fig. 2) are simulated.
, u 0
R f 4R f 6 R f
stiffer
1 , 1
h= 40 R f
fiber
2 , 2 left tip
right tip
Matrix
composites
softer
Fig. 1 A bimaterial centered crack panel,

20m 20m, under remote loading
= .12MPa
Fig. 2 Geometry and boundary condition of

fiber pullout and pushout specimen.
Rf = 0.64, Ef = 100, Em = 1, f = m = 0.3,
35 % fiber in composites.
NEAR TIP FIELD AND STRAIN ENERGY RELEASE RATE

For a crack in homogenous media with friction or an interface crack without friction, there is
always square root singularity associated with the neartip field. However, it was originally
shown by Comninou [2] for an interface crack between two dissimilar isotropic media with
friction and later confirmed and extended by Deng [3] for an interface crack between two
dissimilar anisotropic media that the stress singularity r is always weak ( < 0.5 ) for a
static crack under monotonic loading. The singularity index is related to one of Dundurs
parameters and coefficient of friction as
cot =
(1)
It is easy to see that is 0.5 if or is zero which corresponds to homogenous media or

frictionless condition, respectively. The neartip stress field ahead of the crack tip along the
bimaterial interface and the relative crack surface sliding displacement in plane strain
condition are
II - 388
xy ( r ,0) = K II (2 r )
xy ( r , ) = K II cos (2 r )
(2)
yy ( r , ) = K II sin (2 r )
u x ( r ) = u x ( r , ) u x ( r , ) = K II sin / 2(1 )(2 ) r 1

where the generalized stress intensity factor K II is defined as
K II = lim ( 2r ) xy ( r,0 )
r0
] [
= ((3 4 1 )(1 ) + (1 + )) / 2 1 + ((3 4 2 )(1 + ) + (1 )) / 2 2
(3)
(4)
In Eqn (4), i and i are the shear modulus and Poissons ratio, respectively, for material i
(i = 1,2).
It has been pointed out by Dundurs and Comninou [10] that the problem of interface crack
with friction is a linear process in the context of monotonic loading. In this context, relation of
the strain energy release rate and stress intensity factor can be obtained through a finite crack
extension a as
1 a

G ( a ) =
xy ( r ,0) xy ( a r , ) u x ( a r ) dr
2 a 0
(5)
K 2 sin
II
4(1 )( 2 )
(2 ) (1 ) cos
a 1 2
2(1 )
( 3 2 )
In Eqn (5), is the gamma function.

The conventional strain energy release rate is defined as G = lim G( a ) . It is seen from Eqn
a 0
(5) that strain energy release rate G vanishes as a 0 due to the fact that < 0.5 in the
presence of friction.
FINITE ELEMENT PROCEDURE FOR ENERGY CALCULATION

The finite element method is used to perform the energy calculation during crack extension. A
two-step release technique is recommended so that self-similar mesh along the entire crack
surface is not needed in calculating strain energy release rate as well as dissipation energy rate
during crack extension. In calculating the strain energy release rate at the crack tip of finite
contact, special care should be taken to separate the elastic nodal force from the frictional
nodal force due to crack surface frictional traction near the crack tip.
In view of the foregoing, a procedure to calculate the strain energy release rate and dissipation
energy rate based on crack tip nodal forces and displacements is presented. A two-step crack
release method is used to allow non-self-similar mesh along the crack surface. Contact
element is used behind and ahead of crack tip. For four-noded elements, the strain energy
II - 389

release rate G( a ) and dissipation energy rate G d ( a ) can be calculated as
)(
1
a
a

G ( a ) =
Fx Fy u ( 2) u (1)
x
x
2 a
N
G d ( a ) = G d ( a ) + G e ( a )
d
N
G d ( a ) =
(6)
1
a
Fy u ( 2) u (1)
x
x
2 a
where G e ( a ) is the portion of the dissipation energy rate from existing crack surface, and
d
N
G d ( a ) is the portion of the dissipation energy rate from the new crack surface. Fxa and Fya
are the nodal forces at crack tip node a (see Fig. 3). Superscripts 1 and 2 in Eqn (6) stand
for the states before and after crack release, respectively.
y
r
1
2
a
b
Fig. 3: Bimaterial crack and its coordinate system

A center crack lying between two dissimilar infinite isotropic media under remote shear
loading (see Fig. 1) is used to illustrate the validity of the aforementioned technique. The
commercial code ABAQUS is used to perform the analysis. The size of the panel is 20m
20m with a finite crack of 2m. The Dundurs parameters and coefficient of friction are
chosen to be 0.5. A similar problem of infinite plate was also investigated by Dundurs and
Comninou [10] using elastic dislocation approach. The stress intensity factor obtained by
calculating the strain energy release rate and using the relation in Eqn (5) for closed crack tip
(right tip in Fig. 1) is given together with those from Dundurs and Comninou [10] in Table 1.
There is excellent agreement between these two methods, and the difference is believed to be
due to the finite size effect of the panel in the present study compared with the infinite
medium in Dundurs and Comninous study.
Table 1: Comparison of stress intensity factor K II from present analysis and K d obtained
II
by Comninou, et al. [10], K II = (1 + 2 2 )K II / ( 2a ) , = 0.5, = 0.5
a / a
3.18 10 2
1.59 10 2
7.9610 3
3.9810 3
II - 390
K II (present)
0.9723
0.9720
0.9724
0.9737
(K
II
K d / K d 100%
II
II
1.28
1.25
1.29
1.42
A FRACTURE CRITERION
For the center cracked problem as discussed in the previous section, it is recognized that there
is always a sizable contact zone at one crack tip (denoted as closed crack tip), and an
extremely small contact zone at the other tip (which can be virtually regarded as open crack
tip) under pure shear loading. This problem is studied using the finite element procedure to
understand the frictional effect on the interfacial crack under small scale as well as large scale
contacts. In the finite element analysis, the applied shear stress is assumed to be 0.12 MPa,
1 = 35 MPa, 2 = 0.35 MPa, and 1 = 2 = 0.0. These material properties yield = 0.49.

The strain energy release rate G( a ) , dissipation energy rate G d ( a ) as well as total energy

release rate G( a ) + G d ( a ) for the problem of Fig. 1 with = 0.49 and = 0.5
corresponding to various crack extensions are shown in Fig. 4 and Fig. 5 for open and closed
crack tips, respectively.
Energy Release Rates
30000
GI
G II
G
Gd
G+Gd
15000
0
0.00
0.02
a/a
0.04
Fig. 4: Finite energy release rates at the open crack tip for the panel shown in Fig. 1
Energy Release Rates
30000
15000
G
Gd
G+G
d
0
0.00
0.02
a/a
0.04
Fig. 5: Finite extension energy release rates at the closed crack tip
for the panel shown in Fig. 1
At the open crack tip, the contact zone rc / a is usually around 10 5 to 10 7 and can be
virtually treated as an open crack tip. The modified crack closure technique can therefore be
applied to calculate the strain energy release rates. It is seen in Fig. 4 that individual finite
extension strain energy release rates are a dependent, while the total strain energy release
rate as well as the dissipation energy rate remain constant for different crack extensions. It is
II - 391
noted that dissipation energy rates are fairly small compared with strain energy release rate.
This indicates that the frictional effect on the open crack tip is fairly small. It can also be
shown that the stress distribution ahead of the open crack tip is very close to those of the
oscillatory solution obtained based on traction free crack surface condition. It is concluded
that frictional effect at the open crack tip is negligible. All existing approaches for the
calculation of fracture parameters for bimaterial interfacial cracks are thus applicable (e.g.,
see Ref. 1).
As for the fracture behavior at the closed crack tip, it is clearly seen in Fig. 5 that the finite
extension strain energy release rate decreases, and dissipation energy rate increases, while
total energy release rate remains constant when crack extension a decreases. Recall the
relationship between strain energy release rate and crack extension a in Eqn (5) that strain
energy release rate indeed decreases and eventually vanishes when a approaches zero due
to the weak stress singularity. The strain energy release rate is therefore a -dependent, while
the increase of dissipation energy rate make up the decrease of strain energy release rate for
total energy balance when a decreases. The implication of decreasing strain energy release
rate is that the classical strain energy release rate concept is no longer valid because of its zero
value. Consequently, a new finite extension strain energy release rate associated with a finite
crack extension is needed to quantify the weak singular stress field ahead of the crack tip.
The decreasing behavior of strain energy release rate was also noticed by Stringfellow and
Freund [8] who computed the J-integral for the frictional sliding fracture in a thin film on a
substrate. It was pointed out later by Deng [3] that the path independence of J-integral no
longer exists due to the crack surface traction resulting from friction, and the J-integral for a
vanishingly small contour becomes zero indicating strain energy release rate also vanishes.

In view of the foregoing, a fracture criterion using the strain energy release rate G of a critical
finite crack extension a 0 as a fracture toughness parameter is proposed, i.e.,
We
U

G( a 0 ) =
G d ( a 0 )
= G c ( a 0 )
a
a
(7)
where We , U and G d ( a 0 ) are the external work done, strain energy change and the
dissipation energy rate for a crack extension of a 0 , respectively. From Eqn (5), it is seen

that, for a fixed a , the finite extension strain energy release rate G has a unique relation
0
with the generalized stress intensity factor K II . The fracture criterion given by Eqn (7) states
that the interfacial crack would grow if the neartip stress field reaches a certain critical state.
Note that the simple relations below
dU 1 dWe
,
=
da 2 da
G=
dU
da
(8)
between strain energy release rate, strain energy and work done in linear elastic fracture
mechanics are no longer valid in the presence of friction. Table 2 shows the result obtained
from the finite element analysis for the center crack problem (Fig. 1) with = 0.49 and
different coefficients of friction. It is evident that the above relations are no more valid in the
presence of friction.
II - 392
Table 2: Energy ratios for = 0.49 and different coefficients of friction
U / We
0.02
0.2
0.5
0.4917
0.4883
0.4705
(W

+ G( a ) a / We
0.5084
0.5117
0.5295
FIBER PULL-OUT AND FIBER PUSH-OUT

The mechanical properties of the fiber/matrix interface are important in determining the
mechanical behavior of fibrous composites. The local response of the fiber-matrix interface
during fracture has pronounced effects on the mechanical performance and structural integrity
of the composite. Fiber pull-out as well as fiber push-out are the two major methods to
determine the fiber/matrix interfacial properties as well as to understand the toughening effect
of fiber composites due to frictional sliding after debonding.
There have been numerous theoretical models based on fracture mechanics to analyze the
fiber-matrix interface toughness and toughening effect due to fiber sliding. In these studies,
classical strain energy release rate was the key fracture parameter in determining the
debonding initiation and debonding progression. However, most researchers did not account
for the effect of the weak interfacial stress singularity due to friction. Some researchers used
approximate interfacial shear stress distributions to calculate the strain energy release rate
assuming G = dU / da (e.g. [4,6]). Consequently, none of these models can accurately
characterize the intrinsic interfacial toughness. In this section we investigate the debonding
process in the fiber pull-out and fiber push-out problem using the new finite extension strain

energy release rate G( a ) .
It is noted that fiber pull-out and fiber push-out models adopted here are axisymmetric.
However, It can be shown that the neartip stress field ahead of the crack tip and the
characteristic equation for determining the singularity are exactly same as the plane strain
problem. These features were noticed by Atkinson [11] who compared the elastic equilibrium
equations for both axisymmetric and plane strain conditions near the crack tip. Therefore, the
result of 2D plane problem is valid for fiber pull-out and push-out problem.
An axisymmetric section of the fiber-matrix-composite specimen shown in Fig. 2 is used for
the present study. The fiber, matrix and the composite regions are assumed to be isotropic.
The elastic moduli of the composite region are obtained using the rule of mixtures with fiber
volume fraction of 35%. The Youngs moduli for the fiber and matrix are 100 units and 1
unit, respectively, and the Poissons ratios are assumed to be 0.3 for both. The characteristic
crack extension length a 0 is taken to be 0.01 throughout the debonding process. Two types
of loading are used for various debonding lengths; one is constant load for both fiber pull-out
and push-out, the other is constant displacement loading for fiber push-out only.
An open debonding crack tip is associated with the fiber pull-out under constant load (see
Fig. 6). The total strain energy release rate increases at initial debonding stage and become
steady state after debonding length reaches two times the fiber diameter. For the open crack,
II - 393
Energy Release Rates *

Ef/2Rf
the critical strain energy release rate G c with a mode mixity must be determined.
0.5
GI
0.4
GII
G
0.3
0.2
0.1
0
0
10
a/2Rf
Fig. 6: Finite extension energy release rates for fiber pull-out under constant load
0.5
0.5
G
0.4
0.4
Gd
U/We
Energy Release Rates *

Ef/2Rf
During fiber push-out under constant load , the debonding crack surface is completely
closed and the frictional effect is significant. Figure 7 shows the finite extension strain energy
release rate, dissipation energy rate and total energy rate for various debonding lengths. It is

seen that G and G d increase until debonding reaches two times the fiber diameter. There is a
slight decrease for both energy rate afterwards. Figure 8 gives the ratios of strain energy
change in the elastic body and total external work done during crack release for two different
coefficient of friction . It is seen that those ratios are no longer 0.5 as these in the
frictionless case , especially at initial and final debonding stages. However, Eqn (8) may still
be a good approximate when is very small.
G+Gd
0.3
0.2
0.3
u=0.5
0.2
u=0.2
0.1
0.1
0
0
10
a/2Rf
Fig. 7: Finite extension energy release rates for

fiber change and push-out under constant
load
10
a/2Rf
Fig. 8: Ratio of strain energy

external work done during crack
growth in fiber push-out for
= 0.5 and 0.2.
Results for fiber push-out under constant displacement loading u 0 = 0.2 are shown in Fig. 9
and Fig. 10 for two coefficient of friction. It can be easily seen that the initial crack growth is
unstable. Stable crack growth prevails after debonding reaches two times the fiber diameter.

Besides, It is seen that there are virtually no change in total energy rate G + G d and external

applied force for different friction coefficient. However, value of G for = 0.5 is much
smaller than that for = 0.2 . This indicates that a larger applied force is needed for larger
II - 394

coefficients of friction to raise G to the level of the critical intrinsic fracture toughness G c .
1.2
Energy Release Rate

*h2/ (Efu02Rf)
Energy Release Rate

*h2 / ( Ef u02Rf)
1.2
G
1
Gd
0.8
G+Gd
0.6
0.4
0.2
Gd
G+Gd
0.8
0.6
0.4
0.2
0
0
0
10
a/2Rf
10
a/2Rf
Fig. 9: Finite extension energy release rates

for fiber push-out under constant
displacement u 0 = 0.2 with = 0.5 .
Fig. 10: Finite extension energy release

rates for fiber push-out under
constant displacement u 0 = 0.2
with = 0.2.
EFFECT OF COMPRESSIVE LOADING

An open crack may be forced to close if compressive loads normal to the crack surfaces are
applied in addition to the shear load. This condition may exist in fiber composite due to the
mismatch of coefficients of thermal expansion between the fiber and matrix. For such a
closed interfacial crack, the near tip stress field has a stronger singularity, i.e., > 0.5 [12],
and the classical strain energy release rate becomes unbounded. Consequently, the intrinsic
interfacial fracture toughness must be measured in terms of the finite extension strain energy
release rate for a critical distance ao. However, the toughnesses of the two closed cracks with
stronger and weaker singularities, respectively, should not be interchanged.
CONCLUSIONS
The presence of friction on interfacial cracks lying between two dissimilar media may cause
the crack surfaces to close and crack tip stress singularity to become less singular than the
inverse square root singularity. Depending on the loading direction, the contact size may be
extremely small or finite. For the case of small contact zones, the crack tip can be treated
effectively as an open crack and the fracture criterion in terms of total strain energy release
rate and mode mixity can be employed. For the case of finite contact zones, the classical
strain energy release rate vanishes and thus cannot be used as a fracture parameter as in linear
fracture mechanics. A strain energy release rate based on a finite crack extension is shown to
be uniquely related to the neartip stress field and is capable of capturing the intrinsic fracture
toughness of the interfacial crack.
ACKNOWLEDGMENT
This work was supported by Office of Naval Research grant No. N00014-90-J-1666 to Purdue
University. Dr. Y.D.S. Rajapakse was the grant monitor.
II - 395
REFERENCES
1. Sun, C.T. and Qian, W., The Use of Finite Extension Strain Energy Release Rates for
Fracture in Interface Cracks, to appear in International Journal of Solids and Structures.
2 Comninou, M., Interface Crack with Friction in the Contact Zone, Journal of Applied
Mechanics, Vol. 44, 1977, pp. 780-781.
3. Deng, X., An Asymptotic Analysis of Stationary and Moving Cracks with Frictional
Contact along Bimaterial Interfaces and in Homogeneous Solids, International Journal
of Solids and Structures, Vol. 31, No. 17, 1994, pp. 2407-2429.
4. Gao, Y. C, Mai, Y.W. and Cotterell, B., Fracture of Fiber Reinforced Materials,
Journal of Applied Mathematics and Physics. (ZAMP), Vol. 39, 1988, pp. 550-572.
5. Sigl, L.S. and Evans, A.G., Effects of Residual Stress and Frictional Sliding on Cracking
and Pullout in Brittle Matrix Composites, Mechanics of Materials, Vol. 8, 1989, pp.1-12.
6. Kerans, R.J. and Parthasarathy, T.A., Theoretical Analysis of the Fiber Pullout and
Pushout Tests, Journal of the American Ceramics Society, Vol. 74, 1991, pp. 1585-1596.
7.
Hutchinson, J.W. and Jensen, H.M., Models of Fiber Debonding and Pullout in Brittle
Composites with Friction,, Mechanics of Materials, Vol. 9, 1990, pp. 139-163.
8.
Stringfellow, R.G. and Freund, L.B., The Effect of Interfacial Friction on the BuckleDriven Spontaneous Delamination of a Compressed Thin Film, , International Journal
of Solids and Structures, Vol. 30, No. 10, 1993, pp. 1379-1395.
9. Buchholz, F.G., Wang, H., Ding, S. and Rikards, R., Delamination Analysis For CrossPly Laminates under Bending with Consideration of Crack Face Contact and Friction,
Proceedings of the 10th International Conference on Composite Materials (ICCM-10),
Vol. 1, Whistler, BC, Canada, August 14-18, 1995, pp. 141-148.
10. Comninou, M. and Dundurs, J., Effect of Friction on the Interface Crack Loaded in
Shear, Journal of Elasticity, Vol. 10, No. 2, 1980, pp. 203-212.
11. Atkinson, C., Avila, J., Betz, E. and Smelser, R.E., The Rod Pull Out Problem, Theory
and Experiment, Journal of Mechanical and Physical Solids, Vol. 30, No. 3, 1982, pp.
97-120.
12. Qian, W. and Sun, C.T., Frictional Interfacial Fracture under Combined Compressive
and Shear Loading, to be published.
II - 396
AN EFFECTIVE METHOD FOR MODELING THE

TENSION FRACTURE OF COMPOSITE
STRUCTURES
Khanh Trinh1, Fu-Kuo Chang2
1
Structural & Thermalmechanical Modeling Department, Sandia National Laboratories

PO Box 969, Livermore, California 94551, USA
2
Department of Aeronautics and Astronautics, Stanford University
Stanford, California 94305-4035, USA
SUMMARY: A modeling method that uses finite element analyses together with a
progressive damage model was developed to predict the fracture strength of composite shells
with cut-outs subjected to in-plane tensile loads. The use of a progressive damage model for
composites involves degrading elastic stiffnesses when material failures are detected. This
material degradation can cause severe non-convergence difficulties in the solution process of
the finite element analyses. This problem can prematurely terminate the finite element
analyses and prevent accurate fracture strength predictions, especially for larger panels. In the
developed modeling method, artificial viscosity was added to an existing material damage
model to overcome the convergence problems. Fracture strength predictions using this
modeling method compared well with experimental data for both small and large notched
panels. Furthermore, the developed method was able to overcome convergence problems
using the quasi-static solution procedure. This characteristic makes the developed modeling
method much more computationally effective than modeling methods that require dynamic
solution procedures, which are often used to partially treat the non-convergence problems.
KEYWORDS: damage tolerance, composite progressive failure, finite element modeling,

artificial viscosity.
INTRODUCTION
Predicting notch strengths of composite laminates with cut-outs is an important and difficult
modeling problem that has been studied extensively. This type of modeling is required to
study the damage tolerance of composite structures such as airplane fuselages and other
composite pressure vessels. Numerous analytical techniques ranging from semi-empirical
fracture models to comprehensive numerical methods have been proposed [1]. Most existing
modeling methods work well for small coupons, but are not effective for larger panels [2].
Large panels here indicate panels that are wider than ten inches. Comprehensive numerical
methods are required to study the damage tolerance of structures with multiple failure modes.
For example, there are three failure modes in a damage tolerance study of an airplane
fuselage: skin fracture, stiffener failures, and skin/stiffener joint failures.
In this study, a modeling method that uses finite element analyses together with a progressive
failure model was developed to simulate failures of large composite panels with cut-outs.
Figure 1 shows the type of geometry and loads to be studied using this modeling method. In
II - 397
this paper, notch strength predictions for flat notched panels under uniaxial tension are
presented. The materials under consideration here are continuous fiber reinforced composite
systems such as graphite/epoxy.
cut-out
cut-out
cut-out
Figure 1: Geometry and loads to be studied

Modeling failures in composites is difficult for at least two reasons. First, the failure
mechanism is complex. It involves many failure modes including fiber breakage, matrix
cracking, and shear failures. These different failure modes cause different types of material
degradation. In addition, these failure modes interact. To accurately predict notch strengths,
the material model must capture the relevant and important failure modes and the
corresponding material degradations. Second, the finite element analyses must overcome
numerical problems caused by the material degradations. This material nonlinearity in
composites is severe because the fibers, the main load carrying constituent, fail in a brittle
manner. The brittle fiber failure introduces a singularity in the constitutive relations. This
singularity can cause serious convergence problems in the nonlinear solution process of the
finite element analyses. Furthermore, brittle failure is a form of strain softening, which may
cause the results to be mesh sensitive [3].
The first problem, the failure mechanism, has been extensively studied and many composite
progressive failure models have been proposed. In this study, a composite failure model
recently developed by Shahid and Chang [4] was adopted. The focus of this study was on the
second problem, for which there is very little reported work. The goal was to overcome
numerical problems caused by the brittle fiber failure, so that accurate notch strength
predictions could be made.
Although not studied extensively, the non-convergence problem is important to the notch
strength modeling problem, especially for large panels. In small coupons, accumulated
damage including fiber failure is normally confined to a small area at the notch before final
failure occurs. However, for large panels, experiments have shown that damage can
accumulate to a substantially larger area before final failure occurs [5]. The extensive damage
could significantly reduce material properties in this area which could potentially cause the
convergence problems. In the small coupon analyses, even if numerical problems prematurely
terminated analyses, notch strength predictions based on maximum analysis loads can often
be made without great loss of accuracy. This practice will often underpredict notch strengths
of larger panels. For larger panels, numerical problems due to brittle fiber failures must be
overcome to obtain accurate notch strength predictions.
II - 398
NUMERICAL PROBLEMS
The numerical problems that prematurely terminated the finite element analyses include the
non-convergence problem in the Newton-Raphson nonlinear solution scheme in most cases
and floating point problems in a few other cases. Floating point problems can be cured by
simply degrading stiffnesses to small, but non-zero numbers. However, the non-convergence
problem is not trivial. A non-convergence problem is the condition where the solver can not
find an equilibrium solution after several Newton-Raphson iterations inside a load increment,
even after the load increment had been cut back several times. Note that Arc-length methods
are not solutions to these problems. All analyses in this study were displacement controlled,
and the displacement boundary condition was defined a-priori to increase monotonically.
The brittle material degradation caused by fiber failures in composite is the main cause of the
convergence problem. Figure 2 shows this troublesome characteristic in the on-axis stress
and strain behavior of a unidirectional ply subjected to a uniaxial tension load. This abrupt
and total stiffness degradation is a singularity in the constitutive relations. The brittle
degradation often causes two conditions that can lead to the non-convergence problem:
shocks in the displacement field, and an ill-conditioned consistent tangent stiffness.
xx, xx
xx
xx
Figure 2: Abrupt stiffness degradation caused by fiber failure
A displacement field shock is a large and sudden jump in the structure deformation. This
occurs when the loaded composite laminate adjusts itself to the directional stiffness loss. This
directional stiffness loss is unique to brittle failures in laminate composites. For a strain
driven finite element analysis, large deformation moves the solution (equilibrium
displacement field) far away from the current displacement field, making it difficult for the
nonlinear solution scheme to converge.
The consistent tangent stiffness is the coefficient matrix in the linear matrix equation used to
advance the solution in the nonlinear solution scheme [6]. When this matrix becomes illconditioned, the solution scheme has more difficulty converging.
In the developed modeling method, an existing composite progressive damage model was
adopted to model the failure response of composite structures, and artificial viscosity was
added to this damage model to overcome the convergence problem. The following sections
describe the modeling method, a brief description of the composite failure model, and the
artificial viscosity material model. Notch strength comparisons between experimental data
and model predictions are presented, together with the effects of the artificial viscosity on
fracture strength predictions.
II - 399
MODELING METHOD
The commercial finite element code ABAQUS/Standard [7], together with a user defined
material model was used to predict notch strengths of notched composite panels. ABAQUS
reduced integration layered shell elements S4R and S8R were used. The loads were uniaxial
tension and were perpendicular to the cut-outs (first problem in figure 1). The load was
applied quasi-statically by fixing the panels bottom side and prescribing a displacement
boundary condition to the top side (displacement control analyses). The full NewtonRaphson method (not modified Newton) was used in the nonlinear solution process. Finite
deformation theory was used since large deformations were anticipated at the notch.
The quasi-static solution procedure makes this modeling method very effective
computationally compares to modeling methods that require dynamic solution procedures.
Although, the notch strength problem is a static problem, dynamic solution procedures are
often used because it can partially improve the convergence characteristic of the analyses.
The composite progressive failure model used in this study was based on a model recently
developed by Shahid and Chang [4], which describes the failure criteria, and degradation
rules. This material model is a nonlinear elastic model (elastic with damage). Four failure
modes are considered: fiber breakage, fiber compression failure, matrix cracking and shear
failure. Modified Hashins failure criteria were used to predict the failure modes [8]. Failures
and damages are tracked in each ply, and ply stiffnesses are reduced based on the cumulative
damage state of the ply. The constitutive relations are as follows:
xx
xx
x
Qxx (d) Qxy (d)

0
Qyx (d) Qyy (d)
0
0
0
Qss (d)
xx
yy
xy
(1)
where d is the material damage state, and Qij are stiffnesses.

The important degradation characteristic is that stiffness degradations are gradual and small
for matrix cracking and shear failure modes, and the degradation is abrupt and total for the
fiber failure modes (see figure 2). The degradations due to matrix cracking and shear failures
cause no convergence problem in the analyses. However, the brittle degradation due to fiber
failure can cause severe convergence problem.
To overcome the convergence problem caused by the fiber failure mode, artificial viscosity
was added to the material model. The added artificial viscosity acts as a shock absorber and
damps out displacement jumps and discontinuities in the stress and strain fields. The
viscosity term also adds viscous stiffness to the material, which helps to keep the consistent
tangent matrix from becoming ill-conditioned. The added viscosity will also help to reduce
the mesh sensitivity problem [3].
The constitutive relations of the progressive damage model with the added artificial viscosity
is as follow:
b(d)Exx xx
xx
Qxx (d) Qxy (d)
0
xx
yy
0
+ b(d)Eyy yy
(2)
xx = Qyx (d) Qyy (d)
xy
0
0
Qss (d)
xy
b(d)Gxy xy
II - 400
tf > t > (tf+t0)
b(d) = 0
(3)
(t-t )
b(d) = b0 [1- f ]
t0
tf < t < (tf+t0)
where ij are the strain rates, b0 is the artificial viscosity parameter, tf is the fiber failure time,
and t0 is the viscosity ramp down period. b0 and t0 are parameters required for the viscous
model. However, a constant t0 was used in this study so that the viscous model is essentially a
simple one parameter model. Since the artificial viscosity may add undesirable effects to the
analysis result, this viscosity should be added only when it is needed. In the current model,
the artificial viscosity is added after the fiber failure mode is detected. The viscosity is also
removed slowly over a predetermined period after this fiber failure occurs.
For quasi-static analyses, time in the strain rate equation is directly related to the load
progression in the analysis. For example, if the total displacement boundary condition is
applied at time 1.0, then half of this displacement boundary condition is applied at time 0.5.
Figure 3 shows the effect of the artificial viscosity to the material degradation of a
unidirectional ply subjected to uniaxial load.
higher viscosity
xx
xx, xx
xx
Figure 3: Effects of artificial viscosity on the fiber failure mode
The artificial viscosity can change the results of an analysis; too much viscosity will increase
a structures stiffness, and erroneously increase the notch strength prediction. Thus, it is
important to use enough viscosity to overcome convergence problem, but not too much. For
this modeling method to work, the problem must have a proper range of artificial viscosity.
When using an artificial value in the proper range, assuming it exists, the artificial viscosity
will be enough to overcome convergence problems, and will not cause error to the strength
prediction. Figure 4 shows the idea of the proper viscosity range in an ideal problem.
The proper range of artificial viscosity is problem dependent, since the susceptibility of an
analysis to convergence problems and the notch strength prediction depends on the loading,
the geometry, the material, the discretizations, and the solution path. This means there is no a
priori knowledge of the proper amount of viscosity, and analysis iterations may be required to
obtain a valid notch strength prediction. Limiting the viscous model to one parameter (b0)
will simplify any required parametric studies.
II - 401
Failure load (Pmax)
proper range
artificial viscosity
Figure 4: Proper range of artificial viscosity in an ideal problem
MODEL VERIFICATION
Finite element analyses were performed to validate the developed modeling method for 11
test cases where notch strength experimental data was available. These test cases included 7
small coupons and 4 large panels. The small coupon test data was obtained from Sandia
National Laboratories [9]. The large panel test data was obtained from the Boeing Company
[5]. The load in all test cases was uniaxial tension. Table 1 shows the geometry, material, and
lay-up for the test cases.
Table 1. Geometry, material and lay-up of test cases.
2a
Case
70B
54B
76B
70C
70E
54E
76E
P8
P9
P12
P12
B
II - 402
w (mm)
38.1
38.1
38.1
38.1
25.4
25.4
25.4
914.
914
1524.
1524.
a (mm)
6.35
6.35
6.35
6.35
9.5
9.5
9.5
102..
114.
152.
152.
R (mm)
6.35
6.35
6.35
1.59
1.59
1.59
1.59
0.89
0.89
0.89
0.89
Material
T800/3900-2
T800/3900-2
T800/3900-2
T800/3900-2
T800/3900-2
T800/3900-2
T800/3900-2
AS4/938
AS4/938
AS4/938
AS4/938
Lay-up
[(45/90/-45/0)2]s
[-45/0/45/902/-45/902/45/0]s
[45/90/-45/0/45/90/45/0/45]s
[(45/90/-45/0)2]s
[(45/90/-45/0)2]s
[-45/0/45/902/-45/902/45/0]s
[45/90/-45/0/45/90/45/0/45]s
[45/90/0/60/90]s
[45/0/90/30/0]s
[45/0/90/30/0]s
[45/90/0/60/15/90/60/0/90/45]
For comparison, the same analyses were performed using the progressive damage model
without the added artificial viscosity. Material properties and parameters required for the
composite progressive damage model can be found in reference 4.
Table 2 summarizes notch strength results. These results show that analyses with and without
artificial viscosity predicted the notch strength of small coupons well. Analyses without the
viscosity severely underpredicted the notch strength of the large panels. Notch strength
prediction for large panels using analyses with the viscosity compared well with experimental
data.
Table 2: Failure load comparison
Test
70B
54B
76B
70C
70E
54E
76E
P8
P9
P12
P12B
Experimental data (N)

45,400
31,200
50,300
44,100
16,900
13,800
25,800
347,000
316,000
601,000
668,000
Model with viscosity (N)

47,200
45,000
57,000
47,200
21,500
16,500
27,100
356,000
312,000
650,000
601,000
Model w/o viscosity (N)

47,200
45,000
57,000
46,700
21,400
15,600
25,400
191,000
169,000
312,000
245,000
Table 3 shows the effects of artificial viscosity on failure load predictions for seven of the
cases shown above. These results show that a low level of viscosity was required to
overcome the non-convergence problem, and that failure load predictions were not very
sensitive to the viscosity level.
Table 3: Effects of artificial viscosity on failure load prediction
Case
analysis 1
b0
70B
54B
70C
70E
76E
P8
P9
1.E-5
1.E-6
1.E-5
1.E-6
1.E-6
1.E-4
1.E-4
Failure
load (N)
47,200
45,000
47,200
21,500
27,100
356,000
312,000
analysis 2
b0
1.E-4
1.E-5
1.E-4
1.E-5
1.E-5
1.E-3
1.E-3
Failure
load (N)
47,200
45,000
48,000
21,500
27,100
356,000
334,000
analysis 3
b0
1.E-3
1.E-4
1.E-3
1.E-4
1.E-4
3.E-3
3.E-3
Failure
load (N)
50,700
45,000
51,200
21,500
32,000
401,000
343,000
II - 403
CONCLUSION
It has been demonstrated that artificial viscosity can be used to overcome non-convergence
problems caused by brittle fiber failures when modeling composites. A simple artificial
viscosity model along with a modeling method on using this viscous model to predict notch
strengths of composites has been presented. Together with the progressive damage model
developed by Shahid and Chang [4], the developed artificial viscosity modeling method was
successfully used to predict the notch strength of both small and large composites panels
subjected to uniaxial tension load. The artificial viscosity modeling method allowed accurate
notch strength predictions for large panels, which could not be effectively done previously.
ACKNOWLEDGEMENTS
This work was supported by the U.S. Department of Energy under Contract #DE-AC0494AL85000.
REFERENCES
1.
J. Awerbuch, and M. S. Madhukar, Notched Strength of Composite Laminates:

Predictions and Experiments-- A review, Journal of Reinforced
Plastics and
Composites, Vol. 4, pp. 1-159, 1985.
2.
L. B. Ilcewicz, et. al., Tension Fracture of Laminates for Transport Fuselage Part 4:
Damage Tolerance Analysis, Fourth NASA/DoD Advanced Technology Conference,
NASA CP3229, pp. 254-298, 1993.
3. A. Needleman, Material rate dependence and mesh sensitivity in localization problems,

Compu. Methods Appl. Mech. Eng., 67, 68-85, 1987.
4. I. Shahid, and F.K. Chang, An Accumulative Damage Model for Tensile and Shear
Failures of Laminated Composite Plates, Journal of Composite Materials, Vol. 29, pp.
926-981, 1995.
5. T.H. Walker, et. al., Tension Fracture of Laminates for Transport Fuselage - Part 2:
Large Notches, Third NASA Advanced Technology Conference NASA CP 3178, pp.
727-758, 1992.
6. T.J.R. Hughes, Nonlinear Finite Element Methods, Stanford University custom course
materials, Stanford Bookstore, 1996.
7. ABAQUS is a commercial general purpose finite element code, Hibbit, Karlsson &
Sorensen Inc., Pawtucket, Rhode Island, USA, 1994.
8. Z. Hashin, Failure Criteria for Unidirectional Fiber Composites, Journal of Applied
Mechanics, 47, pp. 329-334, 1980.
9. W. Lu, and K. Trinh, An Experimental and Computational Study of Progressive Failure
of Laminated Composites with Cut-outs, 10th International Conference on Composite
Materials, Whistler, British Columbia, Canada, August 14-18, 1995.
II - 404
EFFECT OF STRAIN RATE ON TENSILE FRACTURE

BEHAVIOUR OF FIBRE REINFORCED POLYAMIDE
COMPOSITES
M. Todo1, K. Takahashi1, Ph. Bguelin2 and H.H. Kausch2
1
2
Research Institute for Applied Mechanics, Kyushu University, Kasuga, Fukuoka 816, Japan
Polymer Laboratory, Swiss Federal Institute of Technology, CH-1015 Lausanne, Switzerland
SUMMARY: Tensile testing of four different woven fabric polyamide composites was
performed at various loading rates ranging from 1x10-3 to 3m/s using a servohydraulic testing
apparatus. Two kinds of reinforcements, carbon fibre (CF) and glass fibre (GF), and two
kinds of matrices, polyamide-6 (PA6) and modified polyamide-6 (mPA6), were the
components of the four composites. The results showed that both tensile strength and failure
strain of these composites tend to increase with increase of strain rate except for GF/mPA6
whose tensile strength and failure strain were stabilized at high rates (>1x100 1/s). Fracture
regions of the tested specimens were also observed to study micromechanisms of tensile
failure using a polarized optical microscope. The results showed that in the composites with
mPA6 matrix, extensive microcracking occurred in the matrix regions prior to the final
failure. Rate effects on the tensile fracture behaviour of the polyamide composites are
discussed on the basis of these experimental results.
KEYWORDS: polyamide composites, tensile behaviour, strain rate dependence, high speed
testing
INTRODUCTION
Dynamic tensile testing of continuous fibre reinforced polymer composites has been
performed to characterize the tensile mechanical behaviour of the composites [1-6]. Dynamic
mechanical properties such as elastic modulus, tensile strength and failure strain were
obtained in these studies by using specific high speed tensile testing systems such as a one bar
method [1], split-Hopkinson pressure bar techniques [2-5] and a falling weight impact tester
[6]. Dynamic testing results are sometimes compared with static data obtained by different
tests to study rate effects on the mechanical properties. On the other hand, as one attempt to
extend static test to a higher rate test, Bguelin and Kausch have developed a high speed
testing system using a modified servohydraulic testing machine and an optical extensometer
[7]. They have applied this system to study rate effects on the tensile mechanical behaviour
of polymers in a wide range of loading rates.
In the above studies, most of the attention were paid to thermoset matrix composites. With
increasing interests in thermoplastics as matrices for polymer composites, many works have
been carried out to characterize the mechanical behaviour of thermoplastic matrix composites;
however, few attempts have been made to study the dynamic tensile behaviour of the
composites. In the present study, dynamic tensile testing of continuous fibre reinforced
II - 405
polyamide composites was carried out by means of the high speed testing system developed
by Bguelin and Kausch, and effects of strain rate on the tensile fracture properties were
investigated. In this paper, the experimental results are presented, and the tensile failure
mechanisms of this type of thermoplastic matrix composites are discussed.
Material and Specimen
Four different polyamide composites were studied in this work. Plain woven fibre cloth made
of carbon fibre (CF) or E-glass fibre (GF) was used as the reinforcement in these composites.
The matrix was polyamide-6 (PA6) or modified polyamide-6 (mPA6). It is known that mPA6
has higher stiffness and strength than PA6. The four composites are denoted hereafter by
CF/PA6, CF/mPA6, GF/PA6 and GF/mPA6. Thin laminates of 1mm thickness containing six
plies of reinforcements were fabricated for each material. The fibre volume fractions were
56% for CF/PA6 and CF/mPA6, 51% for GF/PA6 and 54 % for GF/mPA6. Tensile specimens
were cut from the laminates, and the direction of the warp threads corresponded with the
tensile loading direction. Specimen geometry is shown in Fig.1.
Testing Apparatus
Tensile testing of the four composites was carried out in a range of loading rate from 1x10-3 to
3 m/s using a servohydraulic testing machine. This apparatus has been modified so as to
reduce dynamic effects at high rates of loading by using a viscoelastic damper in the loading
unit [8]. A piezoelectric load cell was used for the measurement of load in high sensitivity.
Strain was measured by means of an optical extensometer that was designed for the purpose
of dynamic testing [8]. Two optical fibres were placed on specimen surface at an interval of
35 mm as shown in Fig.1, and the tensile deformation of the specimen was evaluated from the
positions of two light points emitted through the optical fibres.
Microscopic Observation
Polarized optical microscopy (POM) was carried out to investigate fracture micromechanisms
of the composites. Longitudinal cross sectional area near failure region was examined by
II - 406
POM. For each of the four composites, a thin sample (about 100 m thickness) was prepared
from the failure region of a tested specimen by using cutting and polishing technique. The
prepared samples were observed on a polarized optical microscope under transmitted and/or
reflected light.

Tensile Testing
Stress-strain relations of four different polyamide composites obtained at the lowest loading
rate (1x10-3 m/s) are shown in Fig.2. A knee phenomenon was observed for all four
composites. It has been postulated that the nonlinearlity in woven fabric composites is caused
by micromechanical deformations such as shear deformation of the longitudinal threads,
extensional deformation of the matrix regions and transverse cracking of the transverse
threads [9,10]. It is clearly seen from Fig.2 that the nonlinear stress-strain behaviour after
the knee
II - 407
point was much larger in the glass fibre composites than that in the carbon fibre composites.
Dependence of the initial tensile modulus on strain rate is shown in Fig.3. The tensile moduli
of CF/PA6, CF/mPA6 and GF/PA6 tended to slightly increase as strain rate increased, while
the modulus of GF/mPA6 appeared to be insensitive to strain rate.
Dependence of the tensile strength on strain rate is shown in Fig.4. The tensile strength of the
carbon fibre composites increased with increase of strain rate. CF/mPA6 exhibited the highest
tensile strength of all four composites at all strain rates as a result of the combination of high
strength carbon fibre and mPA6 matrix. Thus, the better tensile properties of mPA6 than PA6
was well transferred to this type of composite system. The tensile strength of GF/PA6 also
increased with increase of strain rate. The increase was 54% at the highest rate (4x101 1/s),
while that of CF/PA6 was 28%. It should be noted that the tensile strength of GF/PA6 at the
highest rate was almost equivalent to that of CF/mPA6, which possessed the highest strength
at all rates. On the other hand, the tensile strength of GF/mPA6 increased as strain rate
increased up to 1x100 1/s and then slightly decreased with increase of strain rate.
II - 408
II - 409
As a result, the strength of GF/mPA6 was equivalent to that of CF/PA6 and higher than that
of GF/PA6 at the lowest rate (1x10-2 1/s) but the least of all four composites at the highest rate
(3x101 1/s).
Dependence of the failure strain on strain rate is shown in Fig.5. The failure strain of the glass
fibre composites was larger than that of the carbon fibre composites as expected because glass
fibre generally shows larger elongation than carbon fibre. The failure strain of the carbon
fibre composites slightly increased with increase of strain rate. The failure strain of GF/PA6,
on the other hand, increased as strain rate increased, whereas the failure strain of GF/mPA6
increased up to 1x100 1/s and then slightly decreased at higher rates. These strange odd rate
dependencies on the tensile strength and the failure strain of GF/mPA6 appear to be related to
the tensile failure mechanism of this composite discussed later.
Strain-rate dependence of the absorbed fracture energy is shown in Fig.6. The absorbed
energy was directly estimated by numerical integration of the stress-strain curves. The
absorbed energy for the glass fibre composites was higher than that of the carbon fibre
composites at all strain rates tested, which is ascribed to the larger failure strain of the glass
fibre composites as seen in Fig.5. Values of the absorbed energy for the carbon fibre
composites and GF/PA6 tended to increase with increase of strain rate. However, the rate
dependence for GF/mPA6 was very similar to those of the tensile strength (Fig.4) and the
failure strain (Fig.5). Thus, it appears that the modified polyamide resin mPA6 fulfills its
superior tensile performance as a matrix for carbon fibre composite at all strain rates tested.
On the other hand, mPA6 loses its good function as a matrix for glass fibre composite under
high strain rates (>1x100 1/s).
Fracture Mechanism
Failure regions of the carbon fibre composites at the highest loading rate (3 m/s) are shown in
Fig. 7. For CF/PA6, a relatively straight fracture line perpendicular to the tensile direction
was observed, whereas pull-outs of fibre bundles were significant in CF/mPA6. This tendency
was seen at all the testing rates. Failure regions of the glass fibre composites at the highest
loading rate (3 m/s) observed under transmitted light are shown in Fig.8. Dark regions in
these photographs are thought to correspond to the damages such as matrix cracking,
debonding, interfacial failure and delamination. More extensive damage zone was clearly
observed in
II - 410
II - 411
II - 412
GF/mPA6 than that in GF/PA6. The damage development prior to the final tensile failure of
GF/mPA6 was larger than that of GF/PA6 at all testing rates.
POM photographs of the failure regions of the carbon fibre composites at the highest loading
rate (3 m/s) are shown in Fig.9. For CF/mPA6, microcracks were formed perpendicular to
the tensile direction and they were observed in the matrix regions and the transverse threads.
It is also seen that these cracks propagated toward the interface between the longitudinal
threads and the transverse threads. As a result, interfacial debonding must be generated. On
the other hand, such kinds of failure modes were not observed in CF/PA6. POM photographs
of the failure regions of the glass fibre composites generated at the highest loading rate (3
m/s) are shown in Fig.10. For GF/mPA6, extensive microcracks were observed in both the
matrix regions and the transverse threads. Interfacial debonding and delamination were also
seen in GF/mPA6. For GF/PA6, only a few cracks were initiated at the fracture point. At all
the loading rates, microdamage formation was observed in both the carbon and the glass
fibre composites with mPA6 matrix. Thus, the modified polyamide composites tended to
generate microcracking in the matrix regions and also in the transverse threads under tensile
loading condition. This may be related to the brittleness of mPA6. As seen in Figs.4, 5 and 6,
the tensile failure properties of the carbon fibre composites could be improved by using
mPA6 as a matrix at any testing rates. It is therefore considered that the microdamages
observed in CF/mPA6 (Fig.9 (b)) did not affect the tensile fracture properties. On the other
hand, in the case of GF/mPA6 (Fig.10 (b)), extensive microcracking occurred prior to the
final failure and it prevented the increase of the tensile strength under high strain rates
(>1x100 1/s), although GF/mPA6 showed better tensile performance at low loading rates
(<1x100 1/s).
CONCLUSIONS
Tensile fracture properties of four different polyamide composites were studied at loading
rates from 1x10-3 to 3 m/s using a high speed tensile testing apparatus. Carbon fibre
composite with modified polyamide-6 matrix showed better tensile performance at all testing
rates than carbon fibre composite with polyamide-6 matrix. Tensile mechanical properties
such as tensile strength, failure strain and absorbed failure energy for both the carbon fibre
composites tended to increase with increase of strain rate. Glass fibre composite with
polyamide-6 matrix showed superior tensile performance especially at higher strain rates. The
tensile mechanical properties of this composite dramatically increased as strain rate increased.
On the other hand, the tensile mechanical properties of glass fibre composite with modified
polyamide-6 matrix increased as strain rate increased up to 1x100 1/s and then slightly
decreased at higher strain rates (>1x100 1/s). As a result, the mechanical properties of the
modified composite were higher than those of the unmodified composite at low rates (<1x10 0
1/s). However, the relation was reversed at higher rates (>1x100 1/s). Extensive microdamage
formation observed in the modified composite was related to this kind of rate dependence.
ACKNOWLEDGEMENTS
This work was supported by Grant-in-Aid for International Scientific Research (Joint
Research No. 07044160) sponsored by Monbusho. The authors also wish to thank Nittobo Co.
for providing the polyamide composites.
II - 413
REFERENCES
1.
K.Kawata, S. Hashimoto and N. Takeda, In Proc. 4th Int. Conf. on Composite

Materials, ICCM IV, Eds T. Hayashi et al., Tokyo, 1982, 829-836.
2.
J. Harding and L.M. Welsh, J. Mater. Sci., 18, 1983, 1810-1826.
3.
C.A. Ross, W.H. Cook and L.L. Wilson, Exp. Tech., 8, 1984, 30-33.
4.
Z.G. Liu and C.Y. Chiem, ibid, 12, 1988, 20-21.
5.
G.H. Staab and A. Gilat, J. Comp. Mater., 29, 1995, 1308-1320.
6.
S. Barre, T. Chotard and M.L. Benzeggagh, Composites, Part A, 27A, 1996, 1169-1181.
7.
Ph. Bguelin, M. Barbezat and H.H. Kausch, J. Phys. III France, 1, 1991, 1867-1880.
8.
Ph. Bguelin, PhD Thesis No1572, EPFL CH Lausanne, 1996.
9.
T. Ishikawa and T.-W. Chou, J. Mater. Sci., 17, 1982, 3211-3220.
10.
T. Ishikawa and T.-W. Chou, J. Comp. Mater., 17, 1983, 399-413.
II - 414
FIBRE ORIENTATION EFFECTS ON

INTERLAMINAR FRACTURE TOUGHNESS OF
MARINE COMPOSITES
P. Compston 1 , P.-Y. Jar 1 and P. Davies 2
1
Department of Engineering, The Australian National University,

Canberra, ACT 0220, Australia
2
Marine Materials Laboratory, IFREMER,
Centre de Brest BP 70 - 29280, Plouzan, France
KEYWORDS: delamination, woven glass fibres, fibre orientation, marine, thermoset,

toughness,
II - 415
II - 416
II - 417
II - 418
II - 419
II - 420
II - 421
II - 422
II - 423
IMPROVING OF THE FRACTURE TOUGHNESS OF

MG-ALLOY COMPOSITES PRODUCED BY POWDER
METALLURGICAL TECHNIQUE
P. Abachi, B. L. Mordike and K. U. Kainer
Institute of Werkstoffkunde and Werkstofftechnik
T.U.Clausthal, Agricolastr.6, 38678 Clausthal-Z., Germany
SUMMARY: Metal matrix composites of QE22 magnesium alloy reinforced with SiC
particles of different shapes were produced using powder metallurgical technique. Particle
content of up to 25 volume per cent was chosen. The fracture toughness tests were carried out
at room temperature and at 125C in the as extruded as well as in the T6 heat treated
condition. The reinforcement decreases in general the fracture toughness of the alloy QE22.
Sharp irregular SiC particles reduce the fracture toughness of composites more than that of
the blocky type. This was observed for composites tested under all experimental conditions.
KEYWORDS: fracture toughness, magnesium alloy, SiC particles, powder metallurgical

technique
INTRODUCTION
Light metal matrix composites offer higher specific strength, higher elastic moduli,wider
temperature range for application and lower thermal expansion than conventional light metals
and alloys. These can be exploited by designers seeking high temperature strength,
dimentional stability and light weight. Although aluminium and its alloys have been
frequently used as the light metal for the matrix of MMCs, magnesium alloys with a much
lower density have recently been considered as an alternative to Al for the matrix [1-3].
Incorporation of ceramic reinforcements into a light metals matrix enhances markedly the
physical and mechanical properties of these metals, but the composites exhibit low fracturerelated properties, which limit their widerspread application [4-6]. Therefore, the fracture
toughness behaviour of metal matrix composites has been the subject of the recent studies
[7,8]. The examination of the effect of the particles shape and content on the fracture
toughness of a magnesium alloy, QE22, which is a heat treatable and relatively temperature
resistant magnesium alloy was the main objective of the present work.
TEST PROCEDURE
Powder metallurgical technique was used to produce metal matrix composites. For the
matrix QE22 magnesium alloy powder of 39 m average diameter was prepared by gas
atomizing. The alloy powder was mixed with SiC particles of 9 m average diameter. The
consolidation of the cold prepressed alloy powder-particle mixture was carried out by
extrusion. Two types of particle shapes, blocky and sharp, were used to examine the effect of
the particle shape on the mechanical properties of composites. The content of SiC particles
II - 424
reinforcement was in all cases 25 volume per cent. Unreinforced QE22 magnesium alloy
powder was consolidated under the same conditions for use as a reference material to
compare the mechanical properties. Examination of the unreinforced alloy and composites
were carried out on samples in the as extruded as well as in the T6 condition. For the T6 heat
treatment the specimens were solution treated at 530C for 6 h followed by air cooling and
ageing for 8h at 204C. The microstructures of alloy and composites were examined by light
microscopy. Fracture toughness tests were carried out at room and elevated temperature using
the short rod specimen method of measuring fracture toughness [ 8-10 ]. The fracture
surfaces of the tested specimens were examined in the scanning electron microscope.

The micrographs of the figures 1 and 2 show the microstructures of the materials tested.
Figure 1 shows the particle distribution for composites of 20 volume per cent of blocky and
sharp particles respectively. The SiC particles are aligned in the extrusion direction. The
microstructure of the matrix alloy of the composite contains much finer precipitates than the
unreinforced QE22, ( Fig. 2a ). A comparison of the microstructure of the composite in the
extruded and in the T6 condition indicates that the particles facilitated the precipitation of the
second phase, which may be Mg12Nd or Mg12 Nd2Ag, ( Fig. 2b ) [ 11-14].
a.
b.
Fig. 1: Microstructure of extruded QE22 + 20 vol. % SiC composites.

a) Blocky shape particles, b) Sharp shape particles.
II - 425
a.
b.
Fig. 2: Microstructure of QE22 + 20 vol. % SiC composites

a) as extruded, b) as T6 heat treated condition.
The fracture toughness tests of materials in extruded condition were carried out at room
temperature and at 125C, as well as in the T6 heat treated condition. Figure 3 shows the
results of fracture toughness of unreinforced and composites reinforced by SiC particles of
blocky type. The fracture toughness values of composites decrease with increasing content of
the reinforcement independently of test conditions. However, the T6 heat treated composites
reveal lower fracture toughness compared with those extruded. The results of fracture
toughness values of specimens carried out at 125C were higher than those tested at room
temperature.
Fig. 3: Showing the results of fracture toughness tests on QE22 alloy and composites with
blocky type SiC particles
Figure 4 shows the results of fracture toughness of composites reinforced by SiC particles of
the sharp type. The fracture toughness values of composites were lower compared with the
unreinforced alloy. The fracture toughness values of composites reinforced by the SiC
particles of the sharp type show little change with increasing content of particles. These
unexpected results may be due to proper distribution of the particles in the matrix alloy and
the production procedure as a whole.The differences between the fracture toughness values of
these composites in the as extruded as well as T6 condition were not as much observed as by
II - 426
composites with SiC particles of blocky type. Nevertheless, composites with SiC Particles of
the blocky type show higher fracture toughnesses over the whole particle content range
than those of sharp type, ( Fig. 5 ).
Sic Particle Content (Vol.), Sharp Type
Fig. 4: Showing the results of fracture toughness tests on QE22 allow and composites with
sharp type SiC particles
Sic Particle Content (Vol. %)
Fig.5: The effect of particle shape on fracture toughness values of composites at RT .

The fractograph of the fracture surfaces of composites showed a prefentially oriented fracture
plane in the extrution direction ( Fig. 6 ). In addition the fracture surface of composites
revealed mainly a brittle fracture appearance. Occationally some dimples like ductile fracture
around the particles and frequently in the form of the elongated dimples in the extrusion
direction on the plane perpendicular to the fracture surface were observed ( Figs. 7, 8 ). The
micrographs of figure 1 indicate on proper distribution of the reinforcement but sometimes
clustering of the particles due to poor mixture of alloy powder and particles was
observed ( Fig. 8 ). This may be one reason for the scatter of the results.
II - 427
Fig. 6: A fractograph of the short rod specimen of the composites with 20 vol. % SiC
particles ( blocky type )
Fig. 7: Showing the fracture surface of a composite tested with some elongated dimples
II - 428
Fig. 8: Showing an example of SiC particles clustering and dimple like fracture around
some particles
CONCLUSION
1.
The QE22 magnesium alloy may be reinforced using powder metallurgical technique
by SiC particles of blocky and sharp shapes.
2.
As a result of production route the composites show

microstructures in the extrusion direction.
3.
The SiC particles are responsible for the fine microstructure in composites and
facilitate the precipitation of the second phase on T6 heat treatment.
4.
The fracture toughness of the QE22 alloy decreases with the SiC particle
reinforcement. However, the decrease in fracture toughness depends on the particle
shape and content. The fracture toughness of the composites as improved more in
the case of blocky type particles as reinforcement than for the sharp particles.
5.
T6 heat treatment of the composites causes additional embrittlement of the composites.
6.
The particles facilitate crack formation on loading. This may be confirmed by the
effect of particle shape on fracture toughness of tested composites.
directionally oriented
II - 429
REFERENCES
1.
Rack H.J.,Light - weight, High - Performance Metal Matrix Composites, in

Processing and Properties for PM Composites, Kumar P. and Ritter A., Eds, The
Met. Soc.,AIME, Feb. 1988, pp.155-168.
2.
Arsenault R.J., The Strengthening of Aluminium Alloy 6061 Fiber and Platelet
Silicon Carbide, Mat. Sci. Eng., 64, 1984, pp. 171-181.
3.
Kainer K.U. and

Mordike B.L., Herstellung
und Eigenschaften
kurzfaserverstrkten Magnesiumlegierung, Metall, 44, 1990, pp.438-443.
4.
Divecha A.P.,Fishmann S.G.and Karmarker S.D.,Silicon Carbide Reinforced

Aluminium-a Formable Composite, J. of Metals 33, Sept. 1981, pp. 12-17.
5.
Logsdon W.A. and Liaw P.K., Tensile, Fracture Toughness and Fatigue Crack
Growth Rate Properties of SiC Whisker and Particulate Reinforced Al-Metal Matrix
Composites, 83 -ID3- NODEM - PI (Westinghouse Scientific Paper), December 1983.
6.
Crowe C.R., Gray R.A. and Hasson

D.F., Microstructural controlled
FractureToughness of SiC / Al Metal Matrix Composites, in Proc. ICCM-V,
Harrigan W.C. ,Jr., Strife .J. and Dhingra A.K.,Eds, The Metallurgical Society Inc.,
Warrendale, P.A., USA, 1985, pp. 843- 866.
7.
Evans J.T., Fracture and Subcritical Crack Growth in Alumina Fiber Magnesium Composites , Acta Metall. 34, No.10, 1986, pp. 2075 - 2083.
8.
Purazrang K., Abachi P. and Kainer K.U , Mechanical Behaviour of Magnesium

Alloy MMCs Produced by Squeeze Casting and Powder Metallurgy Techniques,
Composites Eng., 3, 1993, pp. 489 - 505.
9.
Barker L.M., A Simplified Method for Measuring Plane

Toughness, Eng. Fracture Mech., 9, 1977, pp. 361 - 369.
10.
Standard Test Method for Plane - Strain ( Chevron - Notch ) Fracture Toughness
of Metallic Materials, in Annual Book of ASTM Standard, E1304-89, Vol. 03.01,
1990, pp. 927-937.
11.
Evdokimenko V.I. and Kripyakevich P.I., The Crystal Structures of Magnesium

Rich Compounds in the Systems La-Mg,Ce-Me and Nd-Mg, Soviet PhysicsCrystallography, 8, 1963, pp.135-141.
12.
Payne R.J.M.and BaileyN.Improvement of the Age Hardening Properties of

Magnesium-Rare Earth Alloys by Addition of Silver, J. of the Inst. of Met., 88,
1960, pp. 417-427.
13.
Lorimer G.W., Structure-Property Relationship in Cast Magnesium Alloys,

Magnesium Technology, The Institute of Metals, 1987, pp.47-53.
14.
Borradaile J.B., Threadgill P.L. and Mordike B.L., The Structure of Complex MgAg-RE-Th Alloys, in Proc.4th Int. Conf. on Strength of Metals and Alloys,
Laboratorie de Physique de Solide, E.N.S.M.I.M. I.N.P.L., Nancy, 1976, pp.1385-1388.
II - 430
Strain
von
Fracture
STUDY ON THE MODE I INTERLAMINAR

FRACTURE TOUGHNESS OF MULTI-DIRECTIONAL
LAMINATES
Li Yong, Li Shunling, Xiao Jun, Tao Jie
Department of Material Science and Engineering, Nanjing University of Aeronautics and
Astronautics, 29 Yu Dao Street, 210016, Nanjing P.R. China
KEYWORDS: laminate, delamination, crack, fracture toughness
II - 431
II - 432
STACKING SEQUENCES
II - 433
II - 434
II - 435
II - 436
II - 437
EXPERIMENTAL DELAMINATION BEHAVIOR OF

ANISOTROPIC LAYERED PLATES WITH
DIFFERENT INTERNAL ARTIFICIAL
DELAMINATIONS UNDER COMPRESSION
LOADING
M. Sczepanik-Weinmann, U. Stoll, F.J. Arendts
Institute for Aircraft Design,

University of Stuttgart, Pfaffenwaldring 31, D-70550 Stuttgart, Germany
KEYWORDS:
delamination, buckling, local, global, compression loading, Moir,
interferometry, artificial inserts
II - 438
II - 439
II - 440
II - 441
II - 442
II - 443
II - 444
II - 445
II - 446
EFFECT OF FABRIC ORIENTATION ON FRACTURE

TOUGHNESS OF GLASS FABRIC/PC COMPOSITE
J.B. Wang 1 , Y.Q. Sun 1 , S.R. Zheng 2 and X.T. Jing 3
1
Mechanical Engineering Department, Northwest Institute of Textile Science and

Technology, Xi'an, 710048, China.
2
Chemical Engineering Department, Northwesten Polytechnical University,
Xi'an, 710072, China
3
Material College, Xian University of Science and Technology,
Xi'an, 710048, China
KEYWORDS: woven fabric composite, fabric orientation, fracture toughness, glass fabric,
polycarbonate
II - 447
II - 448
II - 449
II - 450
II - 451
II - 452
INTERFACE MODELLING IN LAMINATE

COMPOSITES FOR FREE-EDGE EFFECTS ANALYSIS
Mohamed Haboussi, Hlne Dumontet, Jean Louis Billet
Laboratoire de Modlisation et de Mcanique des Structures (URA CNRS 1776)
ENSAM/UPMC/ENS Cachan, 151 Bd. de l'Hpital 75013 Paris
SUMMARY: Two interface models based on physical considerations are suggested and used
to analyse the free-edge effects in unidirectional multi-layered composites. These models
question the pertinence of the classical perfect interface model to represent the interface
physical reality. The first model is transition spatial evolution behavior law defined for an
interlayer according to the laying direction. It is based on a microscopic analysis of the fiber
distribution in the interface vicinity between two layers. The numerical simulation of this
model gives exact stress tensor in all the laminate, especially the values of the interlaminar
stresses which verify exactly the free-edge conditions. The second model is an interface law
defined on material surface, resulting from the resolution of a problem to the limits on the
interlayer, simulating a very thin flexible layer. This model also gives no singular free-edge
interlaminar stresses.
KEYWORDS: unidirectional laminates, free-edge effects, transition behavior law, wall

effect, asymptotic expansion method, material surface law, singular interlaminar stresses
INTRODUCTION
The idea that the quality of a composite does not depend only on the nature and properties of
its constituents but also of their binding region is now known [1]. Many samples exist where
the interface modelling wake has been considered responsible for real or supposed
miscalculations [2,3]. Enhancing the interfacial region analysis is not a simple task, as this
zone does not exist itself, it is formed during the material manufacture [4]. Thus, some
unidirectional thermodurcissable laminate processes like RTM, induce microscopic scale
changes, that manifest by resin focusing at the interface. This focusing is due notably to
specific fiber distribution at the interface vicinity, generated by "the wall effect", introduced
by Billet for laminate composites [5], Fig. 1.
The aim of this work is to investigate the influence of a refined interfacial regional description
on free edge effect and to compare this approach with classical methods of analysis.
Usually, at the free-edge vicinity of a laminate plate, a complex 3D state stresses is observed,
presenting an important concentration effect, able to provoke delamination failures. The
classical theories or the piece wise approach do not reveal this state of stresses, many
structural free-edge effect developments have been then proposed these last years, [6]. In
these models, the interlaminar interface is commonly assimilated to a perfect surface, or
sometimes to a very thin layer of an adhesive material, [7]. These approaches, after
simulation, generate singular free-edge interlaminar stresses. These singularities create
II - 453
difficulties in the application of onset delamination criteria, as they overestimate the physical
interlaminar stresses, [8].
In this work, we start by setting a transition behavior model for the interface region or the
interlayer. This model constitutes the essence of our approach. The evolutive law has
therefore been served in a 3D finite element tensile test simulation of a (0,90)s laminate.
Comparisons with classical results show that our approach allows the best perception of the
stresses in the totality of the composite laminate, especially at the free edge.
This approach presents the inconvenient of being numerically expensive and badly adapted to
the study of any multi-layer containing several interfaces. We have therefore decided to
develop an interface law defined on a material surface, translating the evolutive continue
behavior of the interface region to render our approach more profitable. In order to achieve
this, we have used the asymptotic expansion method where the interface thickness constitutes
the small parameter of the problem.
We obtain a first interface law, asymptotically equivalent, to transmission conditions through
a thin interlayer, relating by stiffness the stresses interface vector to the displacement jumps.
Note that, we have mathematically obtained the supposed imperfect interface model proposed
in [3].
Several basic numerical tests have permitted the theoretical validation of the interface law.
The interlayer was successively discretized for different values of its thickness. The numerical
stiffnesses values converge when the thickness decreases towards the theoretical values given
by the interface law.
Furthermore, the interface law is implemented in the industrial finite element software
ABAQUS. Different tests validate the interface card implementation and show its capacity to
restore the stress and the displacement vector at the interface correctly. Finally, this law has
been used to analyse the free-edge effects in (0/90)s laminate in extension, presenting a very
thin flexible interface, representing an adhesive material.
INTERLAYER TRANSITION NOTION

The classical piecewise approach that averages the behavior in the thickness of each layer and
assimilates the interfacial region as a perfect adherence surface is only an approximation of
the reality. Indeed, the interface, which is the area separating two layers, is a layer of very
weak thickness, Fig. 1.
This layer is constituted in part of resin in the case of organic matrix laminates, with a specific
fiber distribution.
So, to describe more clearly the interlaminar unidirectional laminate interface, we propose a
model that takes into account the three dimensional and heterogeneous nature of the interface.
This model is based on the wall effect notion, a classical notion in granular mechanics [10].
II - 454
Fig.1: Laminate micrographic section (S. I. Anderson and K. Nielsen, ECCM 5 1992)
Wall Effect Comments
The wall effect allows us to demonstrate the fiber specific spatial distribution in the vicinity of
a contact surface. In the case of a laminate, the contact surface can be a layer whose
reinforcements are oriented differently to the studied layer or the composite fabric mould
wall.
Let ' be a longitudinal surface of a layer, parallel to the wall , Fig. 2. In full mass, by unit
area of this surface ', the fiber section area Sf and the resin one Sr are in the same proportion
as the volume rates (Sf + Sr = Vf + Vr = 1).
If ' and are too close in comparison with the fiber diameter, the surface fiber fraction Sf
becomes insignificant compared to the resin fraction Sr (Sf<<Sr).
stochastic distribution
of fibers
x
wall effect
wall
h = F ( f , Vf)
V
f
Fig. 2: Definition of the wall effect and the transition area

In contact with the wall, by a geometrical effect, the layer enriches in resin and sufficiently far
from the wall by a distance equal to the fiber diameter f, we find the full mass composition,
the fiber rate is then equal to its average value Vf . This description explains for example the
external appearance of the moulding manufactured composite.
II - 455
Transition Behavior Law
x3
Vf
V
f
Fig. 3: Fiber rate distribution along the thickness

( V f = 50 % , f = 20 m )
Fig. 4: Theoretical evolution of Vf
The study of volume fraction of the resin is based on the definition of a representative volume
(or slice) of a periodic fiber arrangement whose width corresponds to the periodicity.
Therefore, we quantify the volume fraction of the fibers in the region situated between the
plan ' and the wall . We show that the distribution according to x3 depends, for a given
arrangement, on the diameter and the average volume rate of the fibers in the layer.
The curve, Fig. 3, represents this distribution in the case of an hexagonal fiber arrangement. In
numerical simulations, we have retained an evolutive fiber distribution in a reduced zone
whose thickness is equal to the fiber diameter, Fig. 4.
C
The elasticity tensor of the transition region, ijkl , is then obtained by a classical
homogenisation technique [9]. It appears as a function of x3 due to the x3 evolution fiber rate,
linking up with the behaviors of the adjacent layers , Fig. 5.
x3
C+
ijkl
I+
transition
a rea
Interface
I-
ijkl
Fig. 5: Evolutive behavior law
II - 456
ijkl
NUMERICAL STUDY
The laminate studied is made of 4 elementary unidirectional layers, (0,90)s, with the same
thickness, h, Fig. 6. The layers are transversely isotropic with the following characteristics :
E11 = 38250 MPa ; E22 = E33 = 9200 MPa
G23 = 3135 MPa ; G13 = G12 = 3770 MPa ; 12 = 13 = 0.325
x3
U imposed
displacement along x
x3
0
90
x2
0
90
90
0
x
x
2b=16h
Fig.6 a: Laminate plate (0,90)s
Fig. 6 b: F. E. modelized part
A refined meshing of the plate has been implemented, Fig. 7, using quadratic 20 nodes solid
finite elements listed C3D20 on the industrial software ABAQUS.
Fig. 7: F. E. meshing
It contains 1200 elements and 26259 degrees of freedom. This specific element permits the
introduction of an evolutive law on the thickness.
Two numerical simulation types have been carried out :
- the first corresponds to the classical perfect piecewise approach of the interface, without a
transition law, referenced "SLT".
- the second corresponds to the proposed model and takes into account the evolutive
transition law identified previously, referenced "LTE".
II - 457
Fig. 8: 33 along the interface with an expansion at the vicinity of the free edge
Fig. 9: 23 along the interface with an expansion at the vicinity of the free edge
The results in figures 8, 9, 10 and the expansions associated, show the concentration effect
expected of the stresses at the intersection of the free edge and the interface. On the other
hand, successive meshing refinements at the interface and the free edge vicinity show the non
singular and singular character of the free-edge interlaminar stresses respectively, when using
a continuous evolutive behavior interface law and conditions of a classical perfect interface.
So, the taking into account of the fiber specific distribution in the vicinity of the interface
leads to more physical stresses [12], which are more appropriate for delamination onset
criteria applications.
In addition, this approach allows us to consider a resin interlayer met in some injection
manufacturing processes (RTM). Similarly, the existence of transition film at the composite
periphery during their manufacturing with kit under pressure can be simulated.
These numerical calculations, undertaken in ABAQUS, have needed a very thin discretization
of the interface region. Nevertheless, it is necessary to note that the cost of this approach
render it badly adapted to the study of any industrial multilayer; this is why, we have decided
to develop a simplified transition element, so as to integrate them in a more industrial
meshing, making the approach numerically profitable.
II - 458
Thus, a first interface model based on asymptotic calculations is established. It permits us to

substitute this interlayer discretization for an interfacial law defined on a material surface
without thickness. Its numerical implementation in an industrial software code resulted to a
representative interface card of a transition layer composed by an adhesive material.
Fig. 10: 33 along the free edge (x2=0)

ASYMPTOTIC INTERFACE LAWS
Using asymptotic development, we seek to replace the very thin interlayer behavior law given
in Fig. 5. To make matters simpler, we assume, that it concerns a linear elastic behavior per
part that connects with the adjacent layers behavior :
C ijkl
represents the elasticity tensor of the resin. e is a dimensionaless number less than 1. l is a
length equal thereafter to 1, which will not affect the general character of the statement. More
flexible than the adjacent layers, the interfacial region behavior and particularly the resin one
is a function of its dimensionaless thickness :
II - 459
The typical studied tensor is then an orthotropic rigidity tensor, function of the expanded
variable y3 = x 3 / e, the reduced thickness e and a real power value m .
We use the asymptotic expansion that consists to research displacements and stresses in the
interlayer function of the small parameter e ' under the form :
e x ,x ,x
U ( 1 2 3 ) = U 0 ( x1 ,x2 , y3 ) + eU 1 ( x1 ,x 2 , y3 ) + o( e2 )
e x ,x ,x
( 1 2 3) =
n0
en 0 n 0 ( x1 ,x2 , y3 ) + en0 + 1 n 0 + 1 ( x1 ,x 2 , y3 ) + o( en 0 + 2 )
(5)
depending on m value.
e
Displacement and stresses

are solutions of a classical problem with perfect
adhesion conditions on the interfaces I+ et I- between the interlayer and the adjacent layers.
To this problem, we add a "non-interpenetrability" physical condition :
( U,
U ( x 3 e) U ( x 3 ) 3 e
e
2l
x3 0
0 1
(6)
By replacing (4) in these equations, we obtain succession balance equations corresponding to

the development orders. At the first order, the resolution of the problem provides the
following interface law asymptotically equivalent to transmission conditions through a very
thin layer, whose behavior is given by (3).
Mechanical interpretation of the interface law :

When m<1, a case corresponding either to quite flexible or rigid layer, we find perfect
transmission conditions.
II - 460
When m 3 1,a case corresponding to a flexible or a very flexible layer, we obtain interface
conditions that permit sliding and decohesion, the stiffnesses (k1,k2,k3) being respectively the
coefficients of elastic sliding and decohesion.
-
when these coefficients take infinite values, we find again the transmission law of
perfect interface, m<1.
when they take zero values, the interface does not ensure its binding function
0
(stresses transmission at the interface, = 0 , m 1 ).
when k1, k2 and k3 take intermediate values, the interface is elastic.
when the normal stress 33 is negative (crushing), the relation (8) imposes the
normal displacement continuity, according to the laying direction.
The interface region plays a role in the global composite system resistance through stiffnesses
k1, k2 and k3, which control the transfer level of stresses.
NUMERICAL LAW INTERFACE TREATMENT

In this paragraph, we are going to show that previously established interface conditions can be
implemented in an industrial software ABAQUS.
Elementary tests are first undertaken to validate both the interface law on the
theoretical aspect and its implementation.
The interface card has then been used to analyse the influence of a thin flexible
interlayer on laminate free-edge effects.
Plane strain elementary tensile test :
The test structure is an isotropic plate with an isotropic flexible behavior interlayer, (m=1),
submitting to a tension displacement loading, Fig. 11.
U=1 m
Ec= 2eE
x
2
x1
E = 69 GPa
= 0.33
U=1 m
Fig. 11: Elementary test
Fig. 12: Normal stiffness (MPa/mm)
A first interface law validation can be made by comparison between explicit stiffnesses
obtained analytically by formula (7) to stiffnesses obtained numerically equal to the ratio
between the interlayer stresses and the jump displacements, for different decreasing small
thickness values. We note that for sufficiently thin thickness, the numerical normal stiffness
converges towards the explicit one, Fig. 12. For this simple tensile test sample, it is not
possible to obtain numerically the tangential stiffness by discretizing the interlayer, because
the sliding and the shear stress vanish in this case.
II - 461
To see if the interface law correctly restore the stresses and displacements at the interlayer, we
have replaced the interlayer by a material surface where we have traduced the interface law
using an interface card, elaborating by associating two finite elements in ABAQUS ; a spring
finite element, and a contact finite element insuring the "non-interpenetrability" condition and
authorising the weak sliding.
We can thus verify in figures 13 and 14, for example, that the interface card, whose results are
referenced as "CI", restore the interlayer lateral displacement and the shear stress well that
naturally tends to zero in this case.
Fig. 13: Interlayer lateral displacement
Fig. 14: Interlayer shear stress
Since the law and its implementation are validated, we have then used it to simulate the freeedge effects in a laminate, presenting a thin flexible interlayer.
Tensile test of a (0/90)s laminate composite
Fig. 15: Interlaminar stresses at the free-edge

Vicinity
Fig. 16: 33 distribution along the free edge
For this study, we use the same geometrical and mechanical characteristics of the (0,90)s
laminate used previously.
II - 462
The first important test result undertaken in this part of the work shows that the stresses
obtained by the interface card, with a certain degree of approximation representative of a
flexible thin layer, are not singular, Fig. 15. These stresses have the same magnitude as the
stresses obtained before using a transition evolutive behavior law.
Nevertheless, in Fig. 16, a loss of the free-edge stress gradient intensity can be seen. This can
be explained by the flexible character of the interface, whose stiffnesses here correspond to
those of pure resin. This is not the case for the evolutive behavior.
In conclusion, the established interface law is representative of a very thin flexible layer. It is
defined on material surface and allows us to undertake numerical simulations of thin and
flexible layer cheaply. Now, we are working on the global evolutive behavior tensor to
establish a representative surface material interface law.
REFERENCES
1.
J. K. Kim and Y. W. Mai, Interfaces in Composites, Materials Science and Technology,

Volume 13, 1993.
2.
F. Erdogan and G. D. Gupta, "Layered Composites with an Interface Flaw", Int. J. Solids
Structures, 1971, Vol. 7, pp. 1089 - 1107.
3.
J. D. Achenbach, H. Zhu, " Effect of Interfacial Zone on Mechanical Behavior and

Failure of Fiber Reinforced Composites", Journal of the Mech. and Phy. of Solids, vol.
37, 1989, pp. 381-393.
4.
N. L. Hancox, "High Performance Composites with Resin Matrices", Handbook of

Composites, Volume Editors : A. Kelly and S. T. Mileiko, North-Holland, 1983
5.
J. L. Billet, T. Ben Zineb, B. Ben Lazreg, "Surface Transition Effect And Analysis of
Free Edge Stresses for Laminated Composite Plates", Journes Nationales des
Composites 9, France 1994.
6.
N. J. Pagano, S. R. Soni, "Models of Studying Free-Edge Effects", Composite Materials

Series, Interlaminar Response of Composite Materials, Volume 5, Elsevier, 1989.
7.
S. S. Wang, "Elasticity Solutions for a Class of Composite Laminate Problems with

Stress Singularities", in Herakovich Hashin Mechanics of composite materials recent
advances, 1983.
8.
R. Y. Kim, S. R. Soni, "Experimental and Analytical Studies on the Onset of

Delamination in Laminated Composites", Journal of Composite Materials, Vol. 18,
January 1984.
9.
J. Sanchez-Hubert, E. Sanchez-Palencia, "Introduction aux mthodes asymptotiques et

l'homognisation", Masson Editeur, 1992.
10. M. Mooney, "The viscosity of Concentrated Suspension of Spherical Particles", Journal

of Collods and Interface Science, Vol. 6, p.162, 1950.
11. Z. Hashin, B. W. Rosen, The Elastic Moduli of Fiber Reinforced Materials, Journal of
Applied Mechanics. June 1964 - Vol 31 - pp. 223-232..
12. D. W. Oplinger, B. S. Parker, F. P. Chiang, "Edge-Effect Studies in Fiber-Reinforced
Laminates", Experimental Mechanics, September 1974.
II - 463
BEHAVIOR OF GRADIENT DESIGNED COMPOSITE

UNDER BALLISTIC IMPACT
Frank K. Ko1, Amotz J. Geshury1, and John W. Song2
1Fibrous
Materials Research Center, Department of Materials Engineering

Drexel University, Philadelphia, PA 19104, USA
2US Army Natick Research ,Development and Engineering Center
Natick, MA 01760, USA
SUMMARY: A ceramic plate is often added to the surface of the polymer composite backing
in order to create a mechanism to destroy the projectile and spread the impact energy. In order
to lessen the weight of the hard ceramic face plate and introduce additional energy dissipation
and defeating mechanisms, a Gradient Design Composite (GDC) was developed; Ceramic
spheres are embedded in a matrix to form a face component while the backing plate is
composed of Spectra Shield composite.
The ceramic spheres have the capability of defeating the projectile similar to ceramic plate.
The ballistic limit of a GDC composite is a function of sphere size. The bigger the sphere, the
bigger the penetration hole, thus improve the ballistic performance by having more backing
materials to participate in absorbing the kinetic energy. Visual examination of the
experimental panel shows that the backing material have gone through severe delamination
and out-of-plane deformation.
KEYWORDS: composite armor, gradient design concept, small arms body armor
INTRODUCTION
Composite materials have been used successfully as a light weight material for many
protective applications under low and high velocity impacts. By proper selection of material
systems and design of fiber architecture, it has been shown that a high level of damage
containment can be achieved as demonstrated in a compression after impact test [1]. As the
velocity of impact and the nature of the impactor changes, as in the case of a very high
velocity projectile( ~ 3000 ft/s.), a ceramic plate is often added to the surface of the polymer
composite backing in order to create a mechanism to destroy the projectile and spread the
impact energy over a wide area on the backing plate. Current small arms protective body
armor systems used in the U.S. Army frequently consist of a ceramic face backed by fiber
reinforced composite. These body armors is used to protect effectively against threats such as
.30 caliber type bullets which have a higher and more concentrated energy impact than
fragments. In order to lessen the weight of the hard ceramic face plate and introduce
additional energy dissipation and defeating mechanisms against the projectile, a Gradient
Design Composite (GDC) was developed by a combination of material and geometric
hybrid [2].
II - 464
The Gradient Design Composite - GDC (Fig. 1) is a design concept wherein the material
system consists of a harden component, an energy dissipation component and a damage
containment component which are organized according to a gradient of varying functions.
Ceramic spheres of various sizes are embedded in a matrix of appropriate rigidity to form a
face component. The backing plate is composed of fiber reinforced composite with various
levels of structural integration.
Multiaxial linear systems

strength and energy absorption
maximizing
Hardened layer with spheres for destroying the

projectile tip and creating a greater surface area
to contact the plate - rapidly dispersing the
energy
Fig. 1: Gradient design concept for multifunctional ballistic composites

The goal of a recent study [3] was to verify the ballistic protection of Gradient Design
Composite (GDC) consist of aluminum oxide ceramic spheres as a hard/facing layer and
Spectra Shield as a backing layer. The target zone for armor protection was: areal density
between 6.0 to 10.0 lb/ft2 and Ballistic protection based on Military Standard V50 Ballistic
Test for Armor, MIL-STD-662E of 3000-4000 ft/s.
EXPERIMENTAL WORK
Experimental Design
The experimental design of the research program was aimed to verify the ballistic protection
performance of the GDC with areal density between 6.0 to 10.0 lb/ft2. A two layer composite
armor of a ceramic sphere layer bonded by epoxy and a Spectra Shield backing could be
designed with various sphere sizes and thickness of ceramic layer and backing plate offering a
range of protection levels. An analytical model developed by Florence [4], was modified and
employed for predicting ballistic limit velocities in the experimental design stage.
V50 = [ Hb R2 (1 + ) / 0.91mp]
1/2
where, the values used in the computations were the following

V50: the ballistic limit of the armor
: the ultimate tensile strength of the backing plate, 3.0 x 109 Pa (Spectra 1000 fiber tensile
strength was used in calculation for estimation of an ultimate strength of Spectra
Shield backing plate.)
: the rupture strain of the backing plate, 3.4% (Spectra 1000 fiber breaking elongation was
used in the calculation)
Hc, Hb: the thickness of the ceramic tile and back-up plate, respectively.
II - 465
c, b: densities of ceramics and backing composite materials, respectively; 2610 kg/m3 for
aluminum oxide sphere layer, 3400 kg/m3 for ceramic tile and 1000 kg/m3 for Spectra
backing layer
mp: the projectile mass.
mt: the mass of the target involved in the penetration, varies depending on the thickness of the
two layers process, mt=R2 (cHc +bHb)
: the ratio of mt/mp
R: the radius of the zone of the target affected by the impact

r: the projectile radius, 3.81 mm.
A constant impact target zone of R=17.5 mm was used for all the experimental panels. This
value is based on previous test information [2]. Fig. 1 shows predicted V50 as function of the
weight of the experimental panels. The panels with areal density below 6.0 lb/ft2 were added
to evaluate the backing (Spectra Shield) only. The experimental panels with areal density
above 10.0 lb/ft2 were added to allow investigation of 3 layer sphere structures. Table 1
provides detailed description and predicted performance of the experimental panels. Spectra
Shield layers varied from 105 to 145. One, two, and there layers of ceramic spheres were
used while two sizes of spheres: 7/32 or 1/2. The V50 performance of the panels were
predicted to be from around 3000 to 4000 ft/s based on Florences model [4] in which the
value of R, the radius of the target zone affected by the projectile, used was 17.5 mm for all
the experimental panels.
Materials and Panel Fabrication
Spectra Shield was provided in roll form [0/90], 54 wide, weighing 3.8 oz/yd2 or 0.027
lb/ft2 per ply. The Spectra Shield prepregs were prepared with Kraton thermoplastic resin,
which is a copolymer of styrene and butadiene produced by Shell.
5000
Target Zone
4000
3000
2000
1000
0
0
10
12
Areal density (lb/ft^2)
Fig. 2: Experimental panels - predicted V50 vs. areal density
II - 466
14
Table 1: Experimental panels - predicted performance

Panel No.
1- 3
4-6
7-9
10
13
16-17
22 - 23
27
No. of
Spectra Shield
145
105
125
145
105
125
105
105
Layers
Spheres (7/32)
0
1
1
1
2
2
1 (1/2)
3
Predicted
Areal Density (lb/ft2)
3.991
5.855
6.405
6.956
8.82
9.37
9.667
11.785
Performance
V50 (ft/s)
2923
2859
3229
3593
3367
3768
3499
3808
A specially designed steel mold was used to make 12x12 ballistic testing samples. Dake
Hot Press was used for hot pressing the prepregs. Thermocouples were used to monitor the
temperature of the upper and lower platens and Frekote 700-NC releasing agent by The
Dexter Corporation was used for mold release.
The Spectra Shield prepregs were stacked to desired number of layers/total areal density
using the following empirical equations
Thickness of the panel(in) = N (# of plies) x 0.0053 (in, thickness/ply)
(2)
Areal Density (lb/ft2)= 4.9 x thickness of the panel (in)
(3)
The layers were placed into the hot mold (125C ) for 20 minutes under constant press of 125
Psi, and observe the temperature of the filled mold maintained at 120C. Then the panel was
let to cool down to 100C under pressure and removed from the mold.
Processing of the Sphere/Hard Facing Layer

As mentioned before, in order to see the effect of aluminum oxide spheres size as hard facing
plate, two sizes were selected: 7/32 and 1/2, these spheres were fabricated into 1,2 or 3
layers. Fig 3. shows the arrangement of 7/32 ceramic spheres. In the case of 1/2 spheres
(panel No. 22 and 23), in order to fill the gaps between the spheres a smaller size (1/8)
spheres were added as filler as seen in Fig. 4. The spheres were bonded to each other as well
as to the backing material by epoxy (Shell Epon Resin 828 and Shell Epi-Cure 3282). The
spheres and epoxy were placed on the bottom of the mold, and then the spheres were evenly
arranged and distributed to maximize the packaging density inside the mold. At this stage, a
Spectra Shield Panel made by hot press was placed into the mold on top of the sphere
mixture. Then the mold was closed, and placed on the press for curing under pressure of 500
Psi. To get the desired mechanical properties, all the assembled armors were post cured under
100C for 2 hours.
II - 467
Fig. 3: Arrangement of 7/32 ceramic

spheres
Fig. 4: 1/2 spheres and 1/8 spheres as

filler
Panel Properties
Table 2 list the structure and physical properties of the experimental samples that were made
for the study.
Table 2: Experimental panels - structure and physical properties
Panel
No.
Spectra
Layer
Spheres
layer/size
1
4
7
10
13
16
23
27
145
105
125
145
105
125
105
105
0
1/ 7/32
1/ 7/32
1/ 7/32
2/ 7/32
2/ 7/32
1/ 1/2
3/ 7/32
Spectra
Spheres
Total
Thickness
Thickness
Thickness
(in)
0.833
0.602
0.722
0.838
0.603
0.720
0.607
0.603
(in)
0.228
0.227
0.226
0.439
0.452
0.541
0.755
(in)
0.833
0.831
0.949
1.064
1.042
1.171
1.147
1.358
Spectra
Weight
(lb)
3.950
2.838
3.418
3.975
2.860
3.419
2.885
2.877
Spheres
Weight
(lb)
3.084
3.088
3.125
5.556
5.446
7.085
8.509
Total
Weight
(lb)
3.950
5.92
6.51
7.10
8.416
8.87
9.97
11.386

Ballistic Testing
Ballistic testing were carried out at H. P. White Laboratory, Inc. in accordance with the
Military Standard V50 Ballistic Test for Armor, MIL-STD-662E. The projectile used for
the testing was 7.62 mm M60 Ball type with a lead core encased in a brass jacket. The
projectile weighs about 7 to 9 grams. The exact type and weight of the projectile are
proprietary information.
A machine gun was used for the firing. The samples were rigidly mounted by heavy duty
clamps with the area of impact normal to the line of fire. A witness plate made of 0.002 thick
aluminum sheet was located 6 inch behind and parallel to the test sample. Four electronic
counter type chronograph measuring devices were used to measure the velocity of the
projectile.
One to four shots were fired on each sample. The location of the next shot was selected upon
inspecting the target after the previous shot. The normal up-and-down firing procedure was
II - 468
used. In this study, the ballistic limit V50 is defined as the average of the highest partial
penetration velocity and the lowest complete penetration velocity. Table 3 shows the results
obtained in the ballistic testing.
In the case of Panel No. 10, and 27, the V50 values were incalculated due to a missing data:
either partial penetration or complete penetration information was unavailable. In these cases
the high partial or the low complete projectile velocity were chosen to represent the ballistic
performance of the experimental panel. Fig. 5 shows the ballistic performance of the GDC
panels.
Table 3: Summary of ballistic testing results
Panel No.
Spectra
Layer
1,2,3
4,5,6
7,8,9
10
13
16,17
22,23
27
145
105
125
145
105
125
105
105
Spheres
Average
layer/size Thickness
(in)
0
1/ 7/32
1/ 7/32
1/ 7/32
2/ 7/32
2/ 7/32
1/ 1/2
3/ 7/32
0.839
0.835
0.952
1.070
1.036
1.176
1.156
1.302
Average
Weight
(lb)
Number
of Shots
V50
High
Partial
Low
Complete
3.95
5.91
6.51
7.10
8.40
8.85
9.97
11.37
6
4
2
2
2
2
2
2
2456
2943
3493
2372
2900
3428
NA
3472
3549
3674
3794
2551
3000
3557
3560
3582
3617
3732
NA
Incalculable
3527
3583
3703
Incalculable
Fig 5: V50 results of the experimental panels
Figures 6-A,B,C,D, and E shows the penetration hole created on the facing (sphere) layer.
II - 469
Fig.6A: Penetration hole panel No. 1 (no

spheres)
Fig. 6B: Penetration hole panel No. 5 (1

layer)
Fig. 6C: Penetration hole panel No. 17 (2

sphere layers)
Fig. 6D: Penetration hole panel No. 27 (3

layers)
Fig. 6E: Penetration hole panel No. 23 (1/2)
Table 4 contains information regarding the size (diameter) of the penetration hole from the
front and the thickness of the Spectra Shield layers that did not showed evidence of
delamination (pre-delamination thickness). These results shows the direct relationship
between the sphere size and sphere layers on the front penetration hole. Fig. 6 illustrated the
relation between the number of sphere layers and the size of the penetration hole. There are no
evidence that the Spectra Shield layer - the backing layer influence the size of the
penetration hole.
Fig. 7 - A and B shows typical delamination and Out-of plane deformation of the Spectra
Shield layer as observed from the back side of the experimental panels after the ballistic
testing. Taking into the consideration the difficulties in measuring and averaging the intact
layer (as reported in Table 5) it is clearly evident that the bigger the penetration hole size, the
delimitation region start earlier due to decrease in the projectile velocity.
II - 470
Table 4 Penetration hole and pre-delamination thickness

Panel No.
Spectra
Shild
Spehers
layer/size
1/ 7/32
Penetration
Hole Diameter
Size (mm)
4.66
15.58
Pre-Delamination
Thickness
inch/number of
layers
0.551/104
0.353/67
1,2,3
4,5,6
145
105
7,8,9
125
1/ 7/32
14.99
0.384/72
10
145
1/ 7/32
14.36
0.532/100
13
105
2/ 7/32
23.14
0.260/49
16,17
125
2/ 7/32
20.92
0.369/70
22,23
105
1/ 1/2
35.18
0.230/43
27
105
3/ 7/32
27.71
0.468/88
Fig.7: Typical Delamination and Out-of plane deformation of the Spectra Shield
A - Panel No. 2 and B - Panel No. 16
COMPARISON OF RESULTS WITH PREDICTION

Experimental V50 as well as the predicted V50 based on Florences model of the experimental
panels are listed in Table 5. Only the experimental panels with both facing/sphere layer and
backing/Spectra Shield layer are listed. The last column in Table 5 shows the difference
between the experimental V50 and the predicted V50 (% Difference = (V50 Experimental V50 Predicted)/V50 Experimental). It is clear that the modified Florence model with several
simplifying assumptions provide a good prediction regarding the ballistic velocities at the
experimental design stage. However, when measuring R - the radius of the zone of the target
affected by the impact and recalculate V50 values based on the penetration hole size as listed
in Table 5 a totally different values for V50 are obtained. Table 6, lists the predicted V50 from
Florences model based on:
- cone radius from previous study (R=17.5mm)
- actual measured value of the cone radius.
II - 471
Table 5: V50 -Experimental vs. Prediction

Panel
No.
4,5,6
Spectra
Layer
105
Spheres
layer/size
V50
V50
Experimental
Predicted
Difference
1/ 7/32
2943
2859
+2.8
7,8,9
125
1/ 7/32
3493
3229
+7.6
10
145
1/ 7/32
Incalculable
3593
-0.9
3367
+4.5
13
105
2/ 7/32
<3560
3527
16,17
125
2/ 7/32
3583
3768
-5.2
22,23
105
1/ 1/2
3703
3499
+5.5
27
105
3/ 7/32
Incalculable
3808
-0.3
>3794
In order to overcome the above limitation, it is suggested to try and introduce a geometric
factor into Florences model. The geometric factor could provide a link between the material
and geometric parameters of the ceramic spheres and the ballistic limits. more experimental
results are needed to explore the values of this suggested factor and relay them with the
spheres size, shape and packaging.
Table 6: V50 - 2 Predictions
Panel No.
4,5,6
7,8,9
10
13
16,17
22,23
27
Spectra
Layer
Spheres
layer/size
V50
Predicted
R=17.5 mm
Actual R
V50
Predicted
R=actual
105
125
145
105
125
105
105
1/ 7/32
1/ 7/32
1/ 7/32
2/ 7/32
2/ 7/32
1/ 1/2
3/ 7/32
2859
3229
3593
3367
3768
3499
3808
7.79
7.50
7.18
11.57
10.46
17.59
13.86
792
834
862
1632
1539
3536
2494
CONCLUSIONS
The ceramic spheres have the capability of defeating the projectile similar to ceramic tiles as
evident by the flattening of the projectile. The fracture cone radius varied according to the
sphere radius and energy dissipation was through the breaking up of the ceramic spheres and
failure of the backing composite by shearing, fiber cutting and extensive delamination. The
ballistic limit of a GDC composite is a function of sphere size. The bigger the sphere, the
bigger the penetration hole, thus improve the ballistic performance by having more backing
materials to participate in absorbing the kinetic energy. Visual examination of the
experimental panel after the ballistic testing shows that the backing material (Spectra Shield
composite) have gone through severe delamination and out-of-plane deformation.
ACKNOWLEDGMENTS
The authors thank US Army Soldier systems Command, Natick RD & E Center for
sponsoring the study.
II - 472
REFERENCES
1.
Ko, F. K., Three dimensional Fabrics for Composites, in Textile Structural Composites,
edited by Chou, T.W. and Ko, F.K., pp. 129-171, Elsevier Science Publishers
B.V.,
Amsterdam 1989.
2.
Ko, F.K., Yu, J.Z., and Song, J.W., Characterization of Multifunctional Composite
Armor, Proceedings of the American Society for Composites, Atlanta, GA. October 8,
1996.
3.
Ko, F.K., Geshury, A., and Song, J.W., Multifunctional Composite Armor - Phase II,
Final Report to be submitted to U.S. Army Natick.
4.
Florence, A.L., Interaction of Projectile and Composite Armor, Part II, Stanford
Research Institute Report, AMMRG-CR-69-15, 1969.
II - 473
HYPERVELOCITY IMPACT DAMAGE TO

COMPOSITE MATERIALS
R. C. Tennyson and G. D. Shortliffe
University of Toronto Institute for Aerospace Studies
4925 Dufferin Street, North York, Ontario, Canada, M3H 5T6
SUMMARY: This paper describes an experimental program undertaken to study the impact
damage of laminated graphite/PEEK plates and cylinders subjected to particle impacts at
velocities ranging from 3 to 7 km/s. This velocity range corresponds to that found in space
due to impacting micrometeoroids and space debris (MOD). To assess this problem, impact
tests were conducted at three facilities in the U.S., with most of the tests undertaken at the
NASA Johnson Space Center in Houston, using their two-stage light gas guns. Aluminum
spheres were fired at composite material targets using various diameters and velocities to
determine the damage correlation with energy, laminate thickness and material properties.
Flat plate targets were first investigated to establish damage thresholds, followed by tests on
laminated cylinders characteristic of the robot arm structures being designed for the Space
Station. This paper presents the results of these studies showing the correlations found in
terms of impact crater size, front face total damage, rear face spallation damage and
secondary damage resulting from the ejecta plume impacting the rear wall of the cylinder.
KEYWORDS: hypervelocity impact damage, composite materials, orbital debris, design.
INTRODUCTION
Spacecraft in low Earth orbit (LEO) are vulnerable to impact damage resulting from collisions
with micrometeoroids and orbital debris (MOD). Micrometeoroids originate naturally from
planetary or asteroidal collisions and cometary ejecta. Although there is no particular
direction in which micrometeoroids approach the Earth, spacecraft experience a bias in the
direction of travel (known as the RAM direction). This can readily be seen from the impact
distributions measured on the NASA Long Duration Exposure Facility (LDEF) which spent
almost six years in space. Figure 1 presents this data for impacts arising from the MOD
environment and it is quite apparent that the RAM direction recorded significantly more
impacts, although it is noteworthy that even the back surface was hit as well. Artificial space
debris consists of everything from spent satellites and rockets to aluminum oxide fuel
particles, paint chips and fragmentation objects from collisions of these bodies in orbit. A
plot of the MOD environment as a function of altitude and particle size is shown in Fig. 2.
Although micrometeoroids have been a design issue since the 1960s, concern about the
rowing amount of orbital debris did not receive serious attention until the mid-1980s. The
problem has grown to such an extent that the probability of impact by space debris exceeds
that due to micrometeoroids in some cases. There is currently an astounding 2.5x106 kg of
debris in LEO. Since first entering space in 1957, almost 20,000 objects have been launched
with approximately 7,000 still in orbit.
II - 474
Polymer matrix composites are used extensively in spacecraft structures and components,
such as antenna struts, panels and low distortion frames. The largest composite structure in
space is the Canadian robot arm on all of the Space Shuttles. A much more complex robot
system consisting of graphite/PEEK tubes is currently being assembled for the Space Station
Remote Manipulator System. Although these materials provide significant advantages in
terms of high specific stiffness and strength, as well as low thermal expansion properties, the
problem of MOD impact damage needs to be addressed.
Fig. 1: Circumferential distribution of micrometeoroid/debris impacts on LDEF

(NASA M&D/SIG Rept; Aug. 1990).
Fig. 2: Micrometeoroid and orbital debris flux vs. particle size as a function of altitude
(NASA CR#BB000883A, Jan. 1991).
Very few studies have been conducted on MOD hypervelocity impact (HVI) damage to
polymer matrix composites. Some of the earlier work on graphite/epoxy plates includes that
done by Yew et al [1] (identified in graphs as Austin data) at the NASA Johnson Space
Center in 1986 using two of their light gas guns to achieve hypervelocity impact conditions.
A wide range of plate thicknesses was investigated using quasi-isotropic lay-ups.
II - 475
A more detailed series of tests was reported by Christiansen in 1988 [2] (identified in graphs
as Eagle data) in which both low and high modulus graphite/epoxy plates and tubes were
studied. Again, various target thicknesses were used and different impact angles were
employed. It was observed that fiber modulus has a greater effect than lay-up. Christiansen
developed two impact damage models based on this data for energies above and below 150 J.
Lower energy impact data on both graphite/epoxy and PEEK plates were reported in [3]
(identified as Auburn data) but the range of test parameters was very restricted.
The following report presents the results of a series of tests on PEEK/IM7 and PEEK/AS-4
flat plate and cylinder laminates of varying thickness and lay-ups conducted at the NASA
Johnson Space Center (JSC data), the Southwest Research Institute (SwRI data), and the
NASA Marshall Space Flight Center (MSFC data) using their two-stage light gas guns.
EXPERIMENTS
Test Set-Up
A series of flat plates (primary target) were tested in these facilities with a second plate
(secondary target) located about 33 cm behind the primary impact plate. The purpose of the
second plate was to determine the ejecta damage emanating from the first wall impact and
assess the extent of rear wall penetration and overall structural degradation (i.e., loss in
stiffness and strength). All tests were conducted with aluminum impactor spheres of varying
diameter and at different velocities. Energies ranging from 2 to 2000 J were investigated. In
the following graphs, results from other test programs are also included for comparison
purposes. A summary of our test data is presented below in Table 1.
Table 1: HVI shot matrix performed on graphite/PEEK composite
II - 476
A subsequent series of tests on composite cylinders were performed using the 0.50 Cal twostage light gas gun at the Johnson Space Center. The gun is equipped with a Cordin highspeed camera system, capable of snapping 2.25 million frames per second of the projectile
impact and the debris cloud formation processes. The cylinder targets were 30 cm long, 33
cm in diameter and 2.7 mm thick. Aluminum spherical projectiles, 3.18 mm to 9.13 mm in
diameter, travelling at velocities ranging from 6.4 km/s to 6.91 km/s, were used in all of the
experiments. Only normal impacts were performed. The test results for this series are
summarized below in Table 2.
Table 2: HVI database for the composite cylinder targets.
Target
Shot Id
#1
#2
#3
#4
#5
#6
Dp
[mm]
3.18
5.16
5.95
9.13
9.13
9.13
Projectile
V
[km/s]
6.4
6.8
6.91
6.56
6.56
6.55
E
[J]
961
4646
7396
24007
24009
23937
Damage
Dc *
[mm]
8.5
13.2
14.3
18.3
18.4
18.2
*The equivalent circular diameter (Dc) represents the diameter of a circle which encompasses
the same area as the irregularly-shaped crater hole or damage area.
HYPERVELOCITY IMPACT (HVI) TEST RESULTS

Entry Crater Damage
The entry crater model derived from the HVI data is presented in Fig. 3. It is based on a
*
functional relationship between the observed entry crater diameter, Dc (mm) , and an energy
t
, where E is the projectiles kinetic energy (Joules), t is the target thickness
Dp
(mm), and Dp is the projectile diameter (mm). This parameter was first introduced by
Christiansen [2] for use with composites, and is reused in this report with similar success.
parameter
Although other test data for graphite/epoxy materials are included, the model is fitted with a
regression line, forced to pass through zero, which only considers those data points involving
PEEK matrix targets, impacted with aluminum or glass spherical projectiles.
Each entry in the legend of the graph provides four distinct pieces of information. A legend
entry consists of a data point marker, the test series alias, the type of target material (carbon
fibre/PEEK or GRE), and the type of projectile material (aluminum, glass, or nylon). For
example, a legend entry such as : JSC PEEK (Al) implies that solid squares on the graph
represent data points for the JSC test series using PEEK based composite targets. The
projectiles used in the experiments are aluminium spheres.
II - 477
Fig. 3: Entry crater model.

The functional relationship is given by,
t
(1)
Dp
The model is generally applicable to PEEK and epoxy matrix based composites. The carbon
fibres used all have a modulus in between 135 and 235 GPa. Applicable laminate thicknesses
range from 0.5 mm up to 6.7 mm. The model is independent of laminate lay-up. It is
consistent for a broad range of projectile diameters, extending from 0.4 mm to 9.13 mm,
travelling at velocities ranging from 4 to 7.5 km/s. Aluminum, glass, and nylon projectiles are
compatible with the model.
Dc = .93
Total Entry Damage

From a design point of view, the entry damage on the front face is more critical since it
includes the crater hole and associated damage to the laminate. Entry damage, as defined by
the equivalent diameter method, is plotted in Fig. 4 as a function of the same energy
parameter. Included in this plot is the regression curve based on C-scan measurements of the
sub-surface damage zones. It can be seen that even though the delamination lies below the
surface, the enhanced area is only about 20% greater than the surface area. The entry damage
area can be estimated from measuring the crater area, based on the equivalent diameter
method, using the formula: Entry Area = 20 x Crater Area.
II - 478
Fig. 4: Entry damage diameter vs.
t
.
Dp
Rear face spallation damage on the primary impact plate was also measured, the results of
which are plotted in Fig. 5. The regression analysis indicates that the damage area is
comparable to that observed on the front face for a given energy parameter.
Fig. 5: Spall damage diameter vs.
t
.
Dp
II - 479
Fig. 6: Crater vs. projectile diameter.

A plot of the crater diameter as a function of the projectile diameter is shown in Fig. 6. The
regression curve was constructed only through the PEEK data, excluding the nylon cylinder
results. There appears to be no energy dependence for the range of parameters studied.
Interestingly, one can see that the nylon results correlate with the curve quite well, although
the sub-millimeter SiC particle impact data do not agree at all. Generally, one can assume
that for hypervelocity impacts for particles greater than 1 mm in diameter, the crater hole size
produced will be about 2.7 x the projectile diameter. Below this threshold, it would appear
that the hole diameter can be taken as equal to the particle diameter.
SECONDARY DEBRIS PLUME DAMAGE

Considerable effort was made to investigate the debris plume size and corresponding damage
to the rear plate. The debris plume is composed of impactor fracture particles and delaminated
spallation particles. Based on detailed mapping of the debris impact sites (see Fig. 7), the
primary debris plume cone angle was measured for flat plates. These results are plotted in
Fig. 8 and demonstrate a consistent angle of about 11.6.
The next series of experiments using the composite cylinders produced some valuable insight
into the debris and ejecta cloud processes. The Cordin high-speed camera is capable of
snapping approximately 80 frames of the target over the entire impact process, for an elapsed
time of about 100 s.
They depict the progression of the intact projectile, the initial impact, the formation of the
debris cloud, and finally, the rejection of ejecta out the front surface of the cylinder. An
excerpt of the footage produced in the #4 experiment is shown in Fig. 9. In the first frame,
the incoming projectile can be seen just prior to impacting the surface of the cylinder. In the
subsequent frames, the formation of the secondary debris and ejecta clouds is clearly
apparent. Analysis of the debris cloud velocity reveals that the cloud tip travels at
approximately the same velocity as the projectile.
II - 480
Fig. 7: 87 distinct craters catalogued.
Fig. 8: Debris cone angle vs.
t
.
Dp
The energetic debris clouds generated in shots #4 and #6 completely destroyed the rear wall
of the cylinder, as can be seen in Fig. 10. The photographic analysis, combined with post
impact visual observations of the cylinders, yields a debris cloud cone angle of,
dc 17
(2)
where dc represents the 100 mm cutout radius of holes from the debris cloud. For the
cylinders tested, the damage zone was about 200 mm in diameter. This is more than the 11.6
II - 481
previously found during the flat plate tests. The vastly higher impact energies involved may
contribute to the wider debris cloud spread.
Fig. 9: High-speed photography of the impact process in model #4.
STRUCTURAL FAILURE ASSESSMENT

In the design of a composite structure for spacecraft applications, there are two apparent
failure modes which must be considered:
II - 482
1. An energetic MOD fragment creates a large crater and surrounding damage area, causing
structural failure.
2. An energetic secondary debris cloud destroys a larger area of the rear wall of an enclosed
composite structure (such as a cylinder).
The first scenario, structural failure of the cylinders due to a large entry crater, is highly
unlikely, due to the improbability of encountering a large enough MOD particle. However,
the fracture damage from the debris plume is sufficiently large in area to cause a major
reduction in structural stiffness and buckling strength. For example, the effect of a 200 mm
diameter cutout on the buckling strength of a cylinder is shown in Fig. 11. Strength
reductions of the order of 52% and 61% were found for compression (Fx), torsion (Mx) and
bending (My) loading, respectively.
Fig. 11: Effect of cutouts on reduction of buckling strengths.
CONCLUSIONS
As part of this study, two dozen HVI experiments were conducted using flat plate and
cylindrical sections of PEEK matrix based composites. Many parallel studies were uncovered
which had investigated HVI response in similar composites. All of these experiments were
amalgamated to form an extensive HVI database specific to carbon fibre PEEK and epoxy
based composites.
From the database, an entry crater damage model was produced. The entry crater diameter
t
was found to be functionally dependent on a defined energy parameter, 3 E
, over a wide
Dp
range of target thickness, lay-up, fibre and matrix material, as well as projectile diameter,
II - 483
velocity, energy, geometry, and composition. In general, the crater diameter can be modelled
t
by he relationship Dc =.93 3 E
.
Dp
The initial projectile impact produces a secondary debris cloud of high energy projectile and
target fragments. The debris cloud cone angle, which spans the radius of influence
attributable to debris cloud damage, was estimated to be 17, as observed during tests on the
cylinders. The secondary debris cloud is capable of producing severe damage to any
spacecraft composite component lying in its path.
ACKNOWLEDGMENTS
The authors wish to acknowledge the financial support of the Institute for Space and
Terrestrial Science, an Ontario Centre of Excellence, the Canadian Space Agency, and Spar
Aerospace Ltd., for their in-kind support of this project. Special thanks to J. L. Crews, E.
Christiansen, J. Kerr of NASA JSC, and A. Nolan of NASA MSFC for their assistance and
use of their hypervelocity impact test facilities. Without their willing assistance, this project
would not have been possible.
REFERENCES
1. Yew, C. H., Yang, C. Y., and Crews, J., A Phenomenological Study of the Effect of
Hypervelocity Impacts on Graphite-Epoxy Plates, Engineering Mechanics, University of
Texas at Austin and NASA JSC, 1986.
2. Christiansen, E. L., Investigation of Hypervelocity Impact Damage to Space Station
Truss Tubes, Journal of Impact Engineering, Vol. 10, 1990.
3. Tennyson, R. C., Debris and Micrometeoroid Impact Damage on Spacecraft Materials
and Structures, Final Report, Canadian Space Agency, Contract No. 9F011-3-1372/01XSD, March 1994.
II - 484
RESULTS FROM TESTING OF THE HELICOPTER

TAIL ROTOR BLADE MADE OF COMPOSITE
LAMINATED MATERIALS AFTER BALLISTIC
DAMAGES
Boko Rauo
Aeronautical Department, Faculty of Mechanical Engineering,
University of Belgrade, 27 Marta 80, 11000 Belgrade, Yugoslavia
KEYWORDS: vibratory and fatigue characteristics, ballistic damages, survivability
II - 485
II - 486
II - 487
II - 488
II - 489
II - 490
II - 491
II - 492
II - 493
II - 494
LOW VELOCITY IMPACT FAILURE ANALYSIS OF

SANDWICH BEAM
Ya-Jung Lee, Yau Shyu
Department of Naval Architecture and Ocean Engineering, National Taiwan University, 73
Chou-shan Street, Taipei, Taiwan, R.O.C.
SUMMARY: As concluded experimentally on low-velocity impact behavior of sandwich

beam subjected to the cylindrical impactor, the core material and face sheets dominate the
failure modes. The main failure pattern of sandwich beam with lower density core material is
shear cracks in foam (Mode I), whereas crushed in core and damage in the top face sheets
right underneath the impactor (Mode II) when core material density is higher. Finite element
software ABAQUS is used, and maximum stress failure criterion and a modified stiffness
degradation method are adopted to investigate the initiation of failure, the failure modes and
post-failure behavior of sandwich beam. The numerical results are in good agreement with
experimental results. Parametric studies, including stiffness ratio of face sheets to core and
ratio of impactor diameter to simple support span, are also conducted
KEYWORDS: Impact, sandwich beam, post-failure behavior, stiffness degradation,

maximum stress failure criterion.
INTRODUCTION
Owing to the advantages of lightweight and high bending strength, the composite sandwich
materials play an important role in today's industry. In marine engineering, more and more
ships of small sizes such as fishing boats, and yachts, use sandwich panels in ship hull
construction. However, inadvertent collision during the fabrication process, wave impact
during sailing, and so on, may induce damage of sandwich structures and cause significant
reduction of the stiffness and strength of materials. Thus, further understanding of impact
loading, impact response and impact damage of composite sandwich is necessary for the sake
of safety.
Much work has been done on impact resistance and impact tolerance of composite sandwich
panels as in Ref. [1]-[3]. However, because of the complication of contact behavior between
the projectile and sandwich panels, it is very difficult to predict the impact behavior by
establishing mathematical model. An well-known approach, proposed by Tan and Sun in Ref.
[4], considered for determining the impact response is, firstly, to measure the local contact
behavior of the panel by static indentation test; secondly, to use this experimental results in
conjunction with an impact analytical model. This approach, however, can not be regarded as
true prediction since it requires the fabrication of the entire sandwich and indentation tests to
be made for various impactors and structures under consideration. The true prediction should
possess the ability to predict the impact response from a knowledge of the behavior of the
individual components of the sandwich panels as in Ref. [3]. Therefore, many of researches
use finite element method to analyze the impact responses and predict the fracture initiation of
II - 495
materials, however, simulation on post-failure behavior by FEM is rarely seen.. The

objectives of this study is to develop a failure analysis procedure by FEM, which can
investigate the failure modes and post-failure behavior of sandwich beam subjected to the
cylindrical impactor. Maximum stress failure criterion and a modified stiffness degradation
method are adopted to analyze the failure responses, and compared with the experimental
results. After verification of this failure analysis procedure, parametric studies, including the
stiffness ratio of face sheets to core and ratio of impactor diameter to simply-supported span
are conducted as well.
LOW VELOCITY IMPACT TESTS

Sandwich panels made of laminated face sheets MAM (M : REM 300-65 glass mat; A : T-900
Aramid) and two DIVINYCELL core material with different density 0.1, 0.2 g/cm3 are taken
into consideration. These sandwich panels are fabricated by IHI craft in Japan. The dimension
of specimens is 9cm4.5cm and specimen are simply-supported at two edges with a span of 6
cm. The detailed description of test system and test procedure can be referred to Ref. [5]. As
stated before by authors in Ref. [5], the main failure pattern of sandwich beam with foam core
density of 0.1 is shear cracks in core material, located at some distance away from the impact
point and with an inclined direction of 40~50 degrees. As the impact energy increases, the
cracks will extend to the face-core interface, and then induce the delamination. The
characteristic of impact force history when shear cracks occur is an sharp load drop. In
contrast, when foam core density changed to 0.2, no shear cracks are found in core material,
but crush in core and damage in top face sheets beneath the impact point are found, however,
no abrupt load drop is shown in the impact force history. As mentioned above, it is evident
that the core material has a great influence on failure modes, and for the sake of simplicity,
the failure modes of sandwich beams with core density of 0.1 and 0.2 are named as Mode I
and Mode II, respectively.
NUMERICAL SIMULATION AND DISCUSSION

The commercial finite element software ABAQUS and user-defined subroutine UMAT are
used to simulate the failure behavior of sandwich beams. Because of the line load condition,
the sandwich panels subjected to a cylindrical impactor can be modeled as 2-D plain strain
problem. Since the core material dominates the failure behavior, an precise understanding of
the response of core material to stress is required, especially the yielding or failure behavior.
In Ref. [6], Shaw and Sata suggest that failure of foam is governed by the maximum principal
stress, independent of the minor principal stress and that statement is strongly supported by
Patel and Finnie in Ref. [7] and Zaslawsky in Ref. [8]. The micro-structure of foam
considered in this study is similar to those used by Ref. [6]-[8], and due to the foam behavior
which yields plastically under compression, but cracks under tension, the maximum principal
stress failure criterion is chosen to predict the onset of cracks and yielding in foam core. The
failure prediction and the algorithm as well as an illustration to modify the stiffness when core
is degraded due to cracks or yielding are as follows;
When 1 > Xt , cracks occur, then E1, G12, 12 are reduced to zero
When 2 > Xc , yielding occur, then E2, G12 are modified depending on
the plastic behavior of foam material
II - 496
where, Xt and Xc are the tensile and compressive stresses of foam core , respectively.
As for the face sheets, because the delamination and matrix crackings on laminated face
sheets have little effects on failure mechanism, it can be reasonably ignored and only fibers
damage is considered. When fibers damage occurs in tension or in compression, a degradation
factor of 0.14 is used to modify the E1 as suggested by Ref. [9].
STATIC TESTS SIMULATION

According to the phenomenon that the damage patterns of sandwich beams under dynamic
condition are similar to those produced under static condition, thus, for saving the computing
time, the check on performance of numerical failure analysis is conducted under static case.
Mode I
The typical static tests curve of sandwich beams indented by a cylindrical impactor is shown
in Fig. 1. For convenience, several checkpoints are marked along the curve. Initially, the
specimen is loaded up from checkpoint O to A, without any observable damage. Over the
checkpoint A, minor matrix crackings on top face sheets and yielding of the core in the region
underneath the indentor result in the reduction of bending stiffness of specimen. When the
load path reaches B, the loading immediately drops from B to C due to the onset of shear
cracks in core and the extension of cracks along the face-core interfaces. During the path CD,
the specimen continuously deforms without further extreme failure. The comparison between
numerical and experimental F-d curves is shown in Fig.2. It can be seen that those curves fit
each other well. On the numerical F-d curve, there is a slight drop resulted from fibers
breakage right over the level of A. This discrepancy is owing to the drawbacks of that method
which simulates the fibers damage by multiplying a degradation factor right after the failed
elements detected.
The progressive illustrations of simulation for shear cracks and yielding in core are shown in
Fig.3, Fig.4, respectively. It can be concluded from Fig.3 that the progressive crack failure
pattern is in consistence with the one observed during experiments.
Fig.1: The basic pattern of static indentation

of sandwich beam with core density 0.1
Fig.2: The comparison on numerical and

experimental results
II - 497
Fig. 3: The progressive shear crack failure Fig 4. The progressive yielding in core
Mode II
When the core density is changed to 0.2, under the same configuration and loading condition,
the failure patterns of sandwich beams are different from those of Mode I. From Fig. 5, the
specimen is loaded from O to A without any failure, however, over the level of A, the core
exhibits plastic behavior and face sheets are damaged in the region right under the indentor.
As going further along the path AB, the degree of plasticity in core and damage in face sheets
grows continuously. When checkpoint B is reached, fibers of top face sheets beneath the
indentor are almost broken totally and a vertical crack located in the center line of the core is
found. As the loading path moves, the crack propagates downward and the loading-carrying
capacity of specimen loses gradually as seen in path BC. The comparison on computational
and experimental F-d curves is shown in Fig.6. Although these curves are generally in
consistent with each other, however, as mentioned before, the modeling on fibers damage
needs to be further improved.
II - 498
Fig. 6: The basic pattern of static indentation of sandwich beam with core density=0.2
Fig. 7: The comparison of numerical and experimental results
Dynamic impact analysis

After verification on static failure analysis of sandwich beam, the same failure analysis
procedure is applied to the impact study. Due to the low velocity impact, effect of strain rate
is not considered here. The mass of the impactor is 1.52Kg and the incident impact velocity is
2.174 m/s, making the impact energy 3.59 Joules. As expected, the failure patterns of
sandwich beams with core density 0.1 and 0.2 are almost the same to those produced in static
tests. Fig.7 shows the comparison on numerical and experimental impact force histories of
sandwich beam with core density 0.1. It is evident that they fit each other very well.
COMPARISON ON STATIC AND DYNAMIC BEHAVIOR
As concluded above, under the low impact velocity considered in this study, the failure modes
and failure patterns of dynamic cases can seem to be simulated from static cases. This
conclusion can be further supported from the viewpoints of energy. In static case, the work
done to specimens by the indentor can be obtained from integrating the force-displacement
curve of the indentor. The amount of this work can also be regarded as the energy absorbed
by specimens under the assumption of no energy loss. In dynamic case, there are two ways to
calculate the energy absorption of specimens. (a) computing the loss of kinetic energy of the
II - 499
projectile during impact (b) calculating the work done by the projectile in use of integrating
the dynamic force-displacement curve of the projectile. These methods to evaluate the energy
absorbed by sandwich beam are compared in Fig. 8. From Fig. 8, it is surprisingly shown that
three curves fit each other very well. Further, at the same energy level, the energy absorbed
curves are shown to be quite the same under various combinations; large mass impactor with
low velocity and small mass one with high velocity. From the discussion above, it can be
reasonably concluded that the dynamic behavior can be approximately simulated from the
static one. For example, if the energy needed to move along the loading path from O to B as
shown in Fig.2 is known, then the estimation whether the specimen will fail or not, for given
impact velocity and mass of the impactor, can be done by comparing the static energy value
with that of 1/2 mv2.
Fig. 7: The comparison on impact force

histories
Fig. 8: The comparison on energy absorbed

curves
PARAMETRIC STUDY
Core Material
Under the same face sheets MAM and geometric configuration, several sandwich beam with
different core material densities are investigated. Core densities of 0.08, 0.1, 0.13, 0.16, 0.2
are taken into account and denoted by H80, H100, H130, H160, and H200, respectively. The
elastic properties, yielding strength, and failure strength of each are listed in Table 1. The
effective in-plane elastic property of face sheets is about 14.25 Gpa. The results of parametric
study are tabulated as shown in Table2.
Table 1: Material properties of DIVINYCELL foam core
yielding
failure
Core E (MPa)
strength
strength
(MPa)
(MPa)
H80
80
1.2
2.2
H100
115
2.0
2.75
H130
140
2.5
4.2
H160
170
3.4
5.1
H200
230
3.8
5.8
II - 500
Table 2: Parametric study on core materials

failure
energy
failure
Core strength of needed to
mode
sandwich(N)
fail
H80
1454
4.79
Mode I
H100
1874
4.51
Mode I
H130
1888
8.18
Mode II
H160
2066
6.86
Mode II
H200
2287
6.03
Mode II
From Table 2, it can be seen that the failure modes of sandwich beams with core H80, H100
are Mode I, the others are Mode II. Typically, no matter failure modes, the higher the core
density is, the higher the failure strength of sandwich beam is. However, for the
corresponding failure modes, the higher the core density is, the smaller the energy needed to
fail is. These results show that sandwich with core material of higher density has better static
fracture resistance, however, worse impact resistance from the viewpoint of energy.
Ratio of indentor diameter to span
The diameter (R) of the indentor is fixed to 1.2 cm, and the span (S)of simple support
boundary condition are changed to 2 cm, 3 cm, 4 cm and 5 cm, thus, the corresponding ratio
of R/S is 0.12, 0.15, 0.2 and 0.3. The results of parametric study are shown in Table3. Under
the ratio of R/S considered, failure modes of all the sandwich beams are Mode I. As the ratio
of R/S increases, the failure strength of specimen also increases, however, the opposite
tendency is obtained from the energy viewpoint. Therefore, as the increase of ratio of R/S,
sandwich beams have better static fracture resistance, but worse impact resistance.
CONCLUSION
To simulate the failure mechanism of sandwich beams subjected to a cylindrical impactor, the
maximum principal stress criterion and a stiffness degradation method are coded into the userdefined subroutine UMAT provided by finite element software ABAQUS. This failure
analysis procedure can precisely model two of different failure modes; Mode I: shear cracks
in core and Mode II: yielding in core and damage in top face sheets right under the indentor.
In the range of impact velocity considered in this study, it can be concluded that the dynamic
behavior can be simulated from static results. The parametric studies on core density, and
ratio of R/S are also conducted. The failure modes of sandwich beams with lower core density
trend toward Mode I, whereas toward Mode II with higher core density, however, all the
failure modes of sandwich beams are Mode I in the range of ratio of R/S considered in this
study.
Table 3: Parametric study on ratio of R/S
R/S
failure
energy
failure
strength of needed to
mode
sandwich(N)
fail
.12
2219
6.048
Mode I
0.15
2389
4.51
Mode I
0.2
2639
4.173
Mode I
0.3
3222
4.085
Mode I
II - 501
REFERENCES
1. Michael L. Bernard and Paul A. Lagace, Impact Resistance of Composite Sandwich
Plates, J. Reinforced Plastics and Composites, Vol. 8, 1989, pp.432-445
2. W. K. Shin and B. Z. Jang, Instrumented Impact Testing of Composite Sandwich Panels,
J. Reinforced Plastics and Composites, Vol. 8, 1989, pp270-289
3. J. A. Nemes and K. E. Simmonds, Low-Velocity Impact Response of Foam core
Sandwich Composites, J Com. Mat., Vol. 26, No. 4, 1992, pp. 500-519
4. T. M. Tan and C. T. Sun, Use of Statical Indentation laws in the Impact Analysis of
Laminated Composites Plates , J. App. Mat., Vol. 52, 1985, pp. 6-12
5. Y. J. Lee, Y. Shyu, Low Velocity Impact Behavior of Composite Sandwich Used in Ship
Structures , Proceedings Tenth International Conference of Composite Materials, 1995
6. M. C. Shaw and t. Sata, Int. J. Mech. Sci. 8, 1996, pp. 496
7. M. R. Patel and I. Finnie, Lawrence Livermore Laboratory Report UCRL-13420, 1969
8. M. Zaslawsky, Exper. Mech. No. 2 February, 1973, pp.70
9. Youngchan Kim, Julio F. Davalos and Ever J. Barbero, Progressive Failure Analysis of
Laminated Composite Beam, J. Composite Materials, Vol. 30, No. 5, 1996
II - 502
INFLUENCE OF WATER ABSORPTION ON

DELAMINATION INDUCED BY LOW-VELOCITY
IMPACT AND CAI STRENGTH OF FRPS
Kenjiro Komai, Kohji Minoshima and Kazuto Tanaka
Department of Mechanical Engineering, Kyoto University
Yoshida-Honmachi, Sakyo-ku, Kyoto 606-01, Japan
SUMMARY: An investigation of water absorption effects on the impact fracture mechanism

and compression after impact (CAI) behavior was conducted using carbon fiber reinforced
thermoplastic matrix composite (AS-4/PEEK) and aramid fiber reinforced epoxy matrix
composite (K-49/828), which have a quasi-isotropic stacking sequence of [0/45/90]2S. The
impact tests were carried out using a falling weight tester. Impact-induced internal damage
was observed with a scanning acoustic microscope (SAM). The delamination area of AS4/PEEK was much smaller than that of a CF/epoxy composite (MM-1/982X), and no
influence of water absorption was observed. In contrast, that of K-49/828 was the largest, and
the impact-load-induced delamination area was increased by water absorption. In the case of
AS-4/PEEK, the compressive strength of the impacted specimen at 3J was the same as that of
a virgin specimen, and no influence of water absorption was observed. Impact energy affected
the CAI strength of K-49/828 as much as that of MM-1/982X. The fracture surface was
closely examined using a scanning electron microscope, and the fracture mechanisms were
discussed.
KEYWORDS: delamination, CF/PEEK composite, aramid fiber reinforced epoxy matrix

composite, impact loading, compression after impact, water absorption, nondestructive
inspection, scanning acoustic microscope
INTRODUCTION
Advanced fiber-reinforced polymer matrix composites (FRP) that have high strength and high
stiffness-to-weight ratios are now widely used in high technology engineering applications.
When these are used in a service condition, impact events, such as tool drops, occur, and
environmental exposure also affects the mechanical properties of FRPs. Hence, influences of
impact and environment on mechanical properties of FRPs must be clarified. It is well known
that carbon/epoxy laminates are susceptible to low velocity impact [1,2]. To improve impact
resistance, ductile matrix such as PEEK [3] and aramid fibers [4-6] are reported to be
alternative candidate materials. In this study, effects of water absorption on impact fracture
and compression after impact (CAI) behavior were investigated using carbon fiber reinforced
thermoplastic matrix composite (AS-4/PEEK) and aramid fiber (Kevlar 49) reinforced epoxy
matrix composite (K-49/828), having a quasi-isotropic stacking sequence of [0/45/90]2S.
We compared the results with those of carbon fiber reinforced epoxy matrix composite (MM1/982X) [1].
II - 503
EXPERIMENTAL PROCEDURES
The composites used in this study were carbon fiber reinforced thermoplastic matrix (PEEK)
composite (AS-4/PEEK) and aramid fiber (Kevlar 49) reinforced epoxy matrix composite (K49/828), having a quasi-isotropic stacking sequence of [0/45/90]2S. Average thickness,
fiber-volume fraction, tensile strength and initial elastic modulus are listed in Table 1. The
data for MM-1/982X [1] is also listed. Rectangular specimens of 200mm in length and 40mm
in width were used.
Specimens were immersed in distilled water at 80C for 2 months. The weight gain of the
specimens in time is shown in Fig.1, with the data of angle-ply AS-4/PEEK ([45]3S) and
unidirectionally reinforced AS-4/PEEK ([0]14)[7]. In the case of AS-4/PEEK, the quasiisotropic laminates absorbed more water than angle-ply and unidirectionally reinforced
laminates. The angle-ply and the unidirectionally reinforced laminates were well fabricated,
and no void was observed. However, the quasi-isotropic laminates contained many voids.
Figure 2 shows the surface and the side surface of the quasi-isotropic laminate of AS-4/PEEK.
Table 1: Materials used in this study

Name
Construction of laminate
AS-4/PEEK
Average thickness mm
Fiber volume fraction %
Tensile strength MPa
Initial modulus GPa
2.0
61
720
48
K-49/828
[0/45/90]2S
2.4
66
500
31
MM-1/982X
2.5
59
840
63
Fig. 1: Relationship between weight gain and (immersional duration)1/2

This is the reason why the quasi-isotropic laminate of AS-4/PEEK absorbed more water than
other laminates ([45]3S and [0]14). The weight gain of K-49/828 leveled off at 2.0%. The
weight gain of MM-1/982X was about 1.0%[1]. Specimens immersed in water will be
referred to as wet specimens, while the specimens held in room air referred to as dry
specimens.
II - 504
The impact tests were carried out by a falling weight tester. The steel ball (1/2 inch in
diameter) was attached to the falling weight. By this steel ball, the center of specimen whose
both ends (75mm) were fixed was impacted. Compression tests after impact (CAI) were
performed on impacted dry and wet specimens, as well as non-impacted dry and wet
specimens as a reference. Aluminum tabs were glued to both ends of a specimen, giving a
gage length of 63mm. Compression tests were carried out at a constant crosshead speed of
1mm/min, using a electro-hydraulic servo controlled fatigue testing machine (Shimadzu,
EHF-FB20, Load Capacity: 200kN).
The internal damage after impact was observed using a scanning acoustic microscope (SAM,
Olympus Optical Co., UH3) with a pulse wave of 50MHz for 1600ns gate time. The fracture
surfaces of the specimens were observed using a scanning electron microscope (SEM, JEOL,
JSM-5400LV).
EXPERIMENTAL RESULTS AND DISCUSSION

Impact Test
Figures 3 and 4 respectively show the front and rear surfaces of impacted dry specimens of
AS-4/PEEK and K-49/828. Front surface of AS-4/PEEK has a crease and rear surface has
cracks. For K-49/828, there is no crease and crack. This means that the visible surface damage
of K-49/828 is less than that of AS-4/PEEK.
Figures 5(a) and (b) show SAM photographs of a AS-4/PEEK specimen before and after
impact test. In the case of AS-4/PEEK, even a virgin specimen had internal defects introduced
by insufficient matrix infiltration. Therefore, the subtraction of SAM images between virgin
and impacted specimens was adopted to evaluate impact-induced delamination. The result is
shown in Fig. 5(c). We use the same method for K-49/828, which had no internal defect such
as observed in AS-4/PEEK.
(a)surface
(b) cross section

Fig.2: Virgin specimen (AS-4/PEEK)
Figure 6 shows the impact-induced delamination area as a function of impact energy.

Delamination area of AS-4/PEEK was smaller than that of MM-1/982X, and no influence of
water absorption was observed. For K-49/828, impact induced-delamination area was the
largest, and it was increased by water absorption. The threshold impact energy below which
no-delamination occurred was 3J for both dry and wet AS-4/PEEK specimens. For K-49/828,
it was assumed to be almost 0J.
II - 505
On the fracture surfaces of an impacted AS-4/PEEK specimen, many defects introduced

during manufacturing could be observed (Fig.7). Those defects were imaged by SAM as
white areas, shown in Figure 5(a). Figure 8 shows a cross section of an impacted point. A
crack is initiated at a defect. Improved manufacturing process will further increase the impact
resistance of the composite.
(a) Front surface
(b) Rear surface
Fig. 3: Impacted specimen (AS-4/PEEK)
The characteristic fracture feature of AS-4/PEEK is that the fiber surface was covered by
matrix material, irrespective of water absorption (Fig.9). This indicates that the composite had
superior fiber/matrix interfacial strength. There is little difference in fracture morphology
between dry and wet specimen. For K-49/828, the fracture was caused by interfacial
debonding (Fig.10), indicating low fiber/matrix interfacial strength of the composite. Figure
11 shows the magnified fiber surface in K-49/828. Although water absorption was reported to
promote fiber splitting [8-10], the number of fiber splitting of the wet specimens were less
than that of dry ones. This means that water absorption degraded the fiber/matrix interfacial
strength, thereby promoting the impact-induced delamination by water absorption.
(a) Front surface
(b) Rear surface
Fig. 4: Impacted specimens (K-49/828)
II - 506
Fig. 5 (a): SAM photograph of virgin specimen (AS-4/PEEK)
Fig. 5 (b): SAM photograph of impacted specimen (AS-4/PEEK)
Fig. 5 (c): Extracted delamination (AS-4/PEEK)
II - 507
Fig. 6: Relationship between impact energy and delamination area
Fig. 7: Fracture surface (AS-4/PEEK,

Fig. 8 Cross section of an impacted point
Dry specimen, Impact energy: 5J)
(AS-4/PEEK), Dry specimen, Impact energy: 5J)
(a)Dry specimen
(b) Wet specimen
Fig. 9 Fracture surface of AS-4/PEEK (Impact energy: 5J)

CAI Test
The relationship between impact energy and residual compressive strength after impact is
illustrated in Fig. 12. In the case of AS-4/PEEK, the compressive strength of the impacted
II - 508
specimen at 3J was the same as that of a virgin specimen, because no impact-induced

delamination occurred. Above 3J, however, residual compressive strength decreased with an
increase in impact energy. Also we must note that no influence of water absorption was
observed. The compressive strength of K-49/828 was the lowest in all FRPs tested, as the
aramid fiber has low compressive strength [11].
To examine the ratio of the residual compressive strength to the compressive strength of nonimpacted specimens, the relationship between impact energy and normalized residual
compressive strength was analyzed (Fig.13). The normalized strength was defined by residual
compressive strength divided by compressive strength of non-impacted specimen in a
respective environment. The normalized strength of AS-4/PEEK was the highest, indicating
superior resistance against impact. For K-49/828, it was almost the same as that of MM1/982X.
(a) Dry specimen
(b) Wet specimen
Fig. 10: Fracture surface of K-49/828

(Impact energy: 5J)
(a) Dry specimen
(b) Wet specimen

Fig.11: Magnified fiber surface
(K-49/828, Impact energy:5J)
The relationship between impact-induced delamination area and normalized residual

compressive strength is illustrated in Fig. 14. At a certain delamination area, the normalized
residual compressive strength of K-49/828 was higher than that of MM-1/982X. However,
impact-induced delamination area of K-49/828 was larger than that of MM-1/982X (Fig.6).
II - 509
Therefore, as is shown in Fig. 13, the normalized residual compressive strength of K-49/828
and MM-1/982X was similar when compared at a certain impact energy.
II - 510
CONCLUSIONS
1. Even though composites have initial defects introduced by a defective manufacturing
process, the delamination area can be accurately evaluated by the subtraction of SAM
images between virgin and impacted specimens.
2. The surface damage tolerance of K-49/828 is superior to that of AS-4/PEEK.
3. Delamination area of AS-4/PEEK was the smallest, and no influence of water absorption
was observed. For K-49/828, impact induced-delamination was the largest, and it was
increased by water absorption. The impact-induced delamination tolerance of PEEK
matrix composite is superior to that of epoxy matrix composite.
4. The compressive strength of K-49/828 was the lowest in all FRPs tested. However, the
normalized residual compressive strength of K-49/828 was the same as that of MM1/982X. In the case of AS-4/PEEK, the normalized CAI strength was the highest,
indicating superior resistance against impact. Also there is no influence of water
absorption in AS-4/PEEK, showing superior resistance against water absorption.
5. At a certain delamination area, the normalized residual compressive strength of K-49/828
was higher than that of MM-1/982X. However, the delamination area of K-49/828 was
larger than that of MM-1/982X. Therefore, the normalized residual compressive strength
of K-49/828 became almost the same as that of MM-1/982X, when compared at a certain
impact energy.
6. The fracture surface examined indicates that AS-4/PEEK has superior fiber/matrix
interfacial strength. However, K-49/828 has low fiber/matrix interfacial strength. In the
case of K-49/828, the number of fiber splitting of the wet specimens were less than that of
dry ones, bcause water absorption degraded the fiber/matrix interfacial strength.
II - 511
REFERENCES
1.
Komai, K., Minoshima, K. and Yamasaki, H. "Evaluation of Low-Velocity Impact

Induced Delamination by Scanning Acoustic Microscope and Influence of Water
Absorption on Delamination and Compression after Impact of CFRP", Proceedings of
the First International Conference on Mechanics of Time Dependent Materials, Emri, I.
and Knauss, W.G. Eds., 1995, pp. 287-292.
2.
Morita, H. and Matsumoto, H., Characterization on Impact Damage Resistance of

CFRP Laminates under Falling Weight Test, Transactions of The Japan Society of
Mechanical Engineers, Vol.61, No.581-A, 1995, pp.127-132(in Japanese).
3.
Morton, J. and Godwin, E. W., Impact Response of Tough Carbon Fiber Composites,
Composite Structures, Vol.13, 1989, pp.1-19.
4.
Dorey, G., Sidey, G. R. and Hutchings, Impact properties of carbon fiber/Kevlar 49

fiber hybrid composites, Journal of Composites, January, 1978, pp.25-32.
5.
Sakai, H. and Kuhino, S., Aramid Fiber Kevlar as reinforcing Material of Plastics,
Reinforced Plastics, Vol.31, No.12, 1985, pp.591-596(in Japanese).
6.
Kamiyoshi, M., Aramid Fiber, Reinforced Plastics, Vol.33, No.1, 1987, pp.28-32(in
Japanese).
7.
Komai, K., Minoshima, K. and Nitta, Y., Influence of Water Environment on Fracture
and Fatigue Strength of CF/PEEK Composite, Proceeding of the 1994 Annual Meeting
of JSME/MMD, No.940-37, Vol.B, 1994, pp.382-383(in Japanese).
8.
Allred, R.E. and Roylance, D. K., Transverse moisture sensitivity of aramid/epoxy

composites, Journal of Materials Science, Vol.18, 1983, pp.652-656.
9.
Komai, K. and Shiroshita, S., The Influence of Water and Test Frequency on the
Fatigue Strength of Aramid/Epoxy Composites, Proceeding of the Thirty-fifth Japan
Congress on Materials Research, 1992, pp.153-159.
10.
Komai, K., Minoshima, K. and Turu, K., The effect of Environment on the Tensile
Fracture Behavior of Aramid Fibers, Proceeding of the 72nd JSME Spring Annual
Meeting, Vol., 1995, pp.196-197(in Japanese).
11.
Kato, S., Aramid Fiber, Journal of the Japan Society for Composites Materials,
Vol.17, No.51991pp.198-204(in Japanese)
II - 512
LOW VELOCITY IMPACT DAMAGE OF ORGANIC

FOAM CORE SANDWICH COMPOSITES
William J. Craft, Derke R. Hughes, and Ajit D. Kelkar
Department of Mechanical Engineering, North Carolina A&T State University,
Greensboro, NC 27411, USA
SUMMARY: The purpose of this study was to determine the amount of damage resulting in
a sandwich shell composite due to impact by a mass with different impact energies. In
addition, information on the failure characteristics versus the energy of the impact was
desired. While there has been substantial work on impact damage of a number of laminated
composites, little work exists on sandwich composites. The authors intended to investigate
low velocity impact of sandwich panels using a drop weight system. The dynamic response
of sandwich shells to low velocity impact is complex, and our ability to detect threshold
damage is important, and the influence of core density on impact dynamics has not been
investigated. We just concluded tests on 23 sandwich panels of three core densities at the
Wright Patterson Flight Dynamics Laboratory Dynatup drop weight facility. Impact energies
ranged from 0.28 to 36.9 Newton Meters.
While the number of samples is low, we have been able to determine the threshold energy
producing incipient front-face damage as well as the energy producing maximum impact
force after which catastrophic front face failure results. In addition, we have compared our
results with those of laminated composite panels with a comparable number of plies.
Sandwich panels are surprisingly robust, particularly resisting lower energy damage, and the
density of the core has minimal affect on the panel stiffness. However, core density has a
dramatic affect on the contact forces at impact. In addition, peak forces during impact are
equivalent to peak forces achieved in quasi-static tests indicating that the inertia of a sandwich
system is of no significant influence.
KEYWORDS: low velocity impact, sandwich plates, damage, damage tolerance
INTRODUCTION
Low velocity impact is important in sandwich shell composites for a number of reasons.
Perhaps most importantly, low-velocity impact can lead to damage which is frequently not
detected by visual means. Secondly, it can lead to damage which substantially reduces the
ability of the structure to survive the loads it was designed to resist. Since modern aircraft
and aerospace vehicles are designed to minimize added weight, there is less margin of safety
than with fixed structures making an accurate knowledge of failure more important.
Since composite materials are typically strong and stiff, they are prime candidates for aircraft
structures, and since many applications are stiffness rather than strength driven, sandwich
shell composites are ideal candidates for efficient aerospace structures. Some properties
shared by many fiber-reinforced composites laminates include a low ultimate strain and
II - 513
strength through the thickness [1]. Sandwich shell composite core materials also have a low
shear and compression strength in comparison to the face sheets.
By comparison to
laminates, sandwich shell structures exhibit excellent stiffness and strength in bending and a
low specific weight.
While this sounds like an ideal structural material for aeronautical and aerospace applications,
impact resistance restricts the use of such structures where public safety is an issue. Low
velocity impact testing, where velocities are under 12.2 m/s, is particularly important because
1) it can weaken the material without producing visible damage, 2) there are so many
phenomena including improper manufacturing and assembly of sandwich shell components
which can lead to this damage, and 3) the need to create more cost-effective, stiff, lightweight structures is forcing the industry to reconsider sandwich shells. Other factors limiting
the usefulness of such materials include their tendency to absorb moisture. In fact foam cores
made of polymethacrylimide (PMI) absorb high amounts of water as compared to honeycomb
cores [2]. The objectives of the research were to investigate the low velocity impact damage
at different energies, to determine the characteristics three different core densities contribute
to the impact resistance of sandwich composites, and to compare the nature of damage
inflicted by static and dynamic events.
PREPARATION OF PANELS
Only sandwich plates formed into layered flat plates of dimension 178 mm by 254 mm are
used for this study. The thickness of the plates is standardized at 19 mm although a planned
extended study will include other core thicknesses. The external (face) sheets are made of
nine laminae made of single plain weave carbon graphite fibers with an epoxy resin
(Diaminodiphenly Sulfone). This produces a final sheet of from 1.65 to 1.73 mm in
thickness. The resulting plane weave laminate is similar to the [0/90] orientation in
unidirectional laminates. The [0/90] stacking sequence is a common one which parallels that
of the Cristescu [3] study. The McQuillen & Gause study, where a stacking sequence of [+
_
45/02/+45]s was used, gave very similar results for low velocity impact damage [4].
_
The external sheets of graphite epoxy laminates were fabricated and tested in the Center for
Composite Materials Research (CCMR) within the Department of Mechanical Engineering at
North Carolina A&T State University. The outer and inner face sheets were laminated in a
hydraulic press. The Hydraulic Press was initially heated to 121C and allowed to reach a
pressure 689.5 kPa and temperature of 177C. throughout all layers of
The coefficient of expansion for the resin and fibers are substantially different, and they can
cause parasitic internal stresses unless processing is done carefully and consistently. After the
laminate has been maintained at the prescribed temperature and pressure, the laminate was
allowed to cool at 2 degrees per minute. This rate allows enough time for an annealing action
reducing the parasitic stresses due to the thermal expansion mismatch. Material Property tests
were conducted on the face sheets via an MTS-810 machine; the in-plane properties along the
fiber directions are: 68.2 GPa (tensile modulus), 633 MPa (tensile strength), and 412 MPa
(compressive strength). The external sheets strength properties were taken from a nine ply
laminate specimens cut from sheets originally of dimension 0.42 Meter by 0.51 Meter. The
tensile tests followed the American Standard for Test and Materials (ASTM) D 638-94b.
Compression tests followed ASTM D 1621-94.
II - 514
Rohacell is a polymethacrylimide (PMI) foam. The core material for our sandwich panels
was fabricated from one layer of Rohacell. The construction was chosen to approximate an
efficient structure maintaining both high stiffness and strength, in resisting bending loads.
Rohacell foam can be acquired in a variety of densities and shear and compressive strengths.
One important consideration in using it for this work is that we could experiment with
combinations of core properties to obtain an optimized structure.
The Rohacell core is easily machinable and the cost is less than standard honeycomb
(Nomex); however, it is hygroscopic and it has both a low shear strength and stiffness. Mass
densities of 32, 52, and 75 Kg/M3 have been selected for the core since they represent a range
of properties of sandwich cores, and the effect of core density on impact damage is an
important part of the study. They are referred to as R-31, R-51, and R-71. The Table below
indicates some important stiffness and strength parameters for Rohacell.
Table 1: Rohacell stiffness and strength properties
Density
(Kg/M3)
Young's
Modulus
(N/mm2)
Shear
Modulus
(N/mm2)
Tensile
Strength
N/mm2
Compressive
Strength
N/mm2
Shear
Strength
N/mm2
Flexural
Strength
N/mm2
32
36
13
1.0
0.4
0.4
0.8
52
70
19
1.9
0.9
0.8
1.6
75
92
29
2.8
1.5
1.3
2.5
Once the face sheet is formed, the internal sheet is bonded to each of the external sheet in an
oven with AF-163 film adhesive. A prescribed heating and cooling rate as part of the curing
process helps to control parasitic stresses. Once the sandwich shell laminated composites
structures are assembled, they are checked for warping and dimension. They are edge cut to
their final shape required for low velocity impact tests, measured for dimensional tolerances,
and weighed.
There was concern that conventional cutting equipment could cause delamination at the facecore interfaces, and an Ingersoll-Rand HS-3000 water jet system was used to final cut all
specimens. The closed cell geometry is destroyed near machining cuts, and since a water jet
was used, each specimen absorbed some water and had to be dried. To do this, specimens
were placed into an autoclave for 6 hours at 71.1C. After cooling, Dow Corning RTV # 737
100% silicone rubber was applied to specimens for water proofing and left intact there
through C-SCAN evaluations.
All impact studies were performed at the Wright-Patterson Flight Dynamic Laboratory
Dynatup machine drop weight system. In our configuration, the system employs a freefalling drop weight and precision recording system The drop weight mass was 3.37
Kilograms, and contact with the specimen was made through a 12.7 mm radius indentor
normal to the specimen centers. The Boeing test fixture simulating simple support was used.
The fixture is important since it controls fixity; fixity affects specimen compliance which
changes the force-time impact curve.
II - 515
EXPERIMENTAL RESULTS
All impact specimens were identical except for the density of the Rohacell foam core. The
specimens of Table 2. are sorted by density and drop height. A total of twenty-three
specimens were tested in the Dynatup machine. For completeness, specimen ID, impact
velocity (m/s), maximum contact force (Newtons), and impact duration (msec.) are displayed.
The drop height was measured using a precision gage block set. The quantities including
impact velocity, impact force, and impact event duration were all determined experimentally
through drop-weight accelerometers which had an output sampling rate of 40 samples per
millisecond. All other data in this study can be derived from these values.
Table 2: Tested impact specimens sorted by core density and drop weight height
Specimen
ID
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Drop
Height
mm
25.40
39.42
51.77
75.74
101.60
127.00
139.70
25.67
39.22
51.36
75.77
101.50
127.00
139.70
25.40
38.86
51.13
75.44
101.20
101.60
127.00
127.00
138.90
Density
designation
Impact vel.
(m/s)
R-31
R-31
R-31
R-31
R-31
R-31
R-31
R-51
R-51
R-51
R-51
R-51
R-51
R-51
R-71
R-71
R-71
R-71
R-71
R-71
R-71
R-71
R-71
0.7071
0.8778
0.9997
1.2009
1.3929
1.5575
1.6368
0.7102
0.8717
0.9967
1.2009
1.3929
1.5575
1.6368
0.6706
0.8656
0.9967
1.1979
1.3868
1.3929
1.5606
1.5636
1.6368
Contact
force
(N)
1458.90
1569.30
1721.40
1887.70
2176.90
2264.00
2303.60
1672.90
1949.10
2216.00
2393.50
2471.80
2698.20
2537.60
1790.30
2162.20
2471.30
2985.10
2852.90
2890.30
3253.70
2944.60
3047.30
Impact
duration
(msec.)
10.97
11.65
12.15
12.90
12.57
13.25
13.43
9.05
9.70
9.88
10.55
11.23
10.95
11.75
8.02
7.93
8.43
8.25
9.72
9.40
8.85
9.60
9.92
Before dynamic tests were conducted, a static test was conducted on a sample of each density.
Static tests were run on a universal testing machine with the same indentor and Boeing fixture
which would be used in all drop weight impact tests. All static tests was conducted at
II - 516
crosshead velocities of 0.02 millimeters per second. Samples required 5 to 10 minutes for the
maximum contact force to occur. In each case, the time force curve was smooth until it was
reached. Afterwards, the force decreased somewhat erratically and was accompanied by
popping noises indicating the integration of additional damage.
An initial drop height for each density was selected by trial-and-error. The initial heights
were chosen so that a contact force of about 50% of the static test maximum contact force
would be developed. From this initial setting, the drop weight height was increased until
additional drop heights produced no higher contact force maximum. After this, the drop
weight heights were decreased until a height was reached which produced no detectable
impact damage.
Two techniques were used to determine the force producing incipient damage in these
sandwich plates. Incipient damage was determined by: 1) decreasing the drop weight height
until there was no visible asymmetry in the force-time curves and 2) noting at what level the
impact force-time plot first deviated from a smooth curve. This point was characterized by
substantial "ringing" in the accelerometer output. As it turns out, both methods produced
results which were very similar. This is to be expected if the inertia of the plates is negligible
in the impact event and if there are no significant time-dependent mechanics at work. A postimpact study of foam core structures has been reported by Caprino and Teti [5]; new theories
of non-linear behavior mechanics have been proposed by Kuhhorn and Schoop [6], but our
study did not extend past the energy producing maximum contact force since this constitutes
failure of the outer face sheet.
The force-time histories for all twenty-three specimens of all three density core sandwich
panels were numerically smoothed and displayed in an overlay as Figure 1. All impact tests
on all twenty-three sandwich panels are included. These graphs are separated into three
sections as a function of core densities and drop heights. For comparison, one overlayed
graph of force-time histories for R-51 for the lowest, mid-level, and highest impact energies is
also reproduced in as recorded format, Fig. 2.
II - 517
The fact that higher impact forces are developed in stiffer panels for a given drop height is
easily seen in Fig. 3 where three specimens (numbers 4, 11, and 18) were subjected to [almost
identical] drop heights of about 75 millimeters. The force change due to the foam cores is
substantial even though all face sheets are identical. Note also the increased contact time and
the equal impulse to all three specimens. The largest impact force developed in the R-71
highest density core panel specimen is 58% more than the force developed during impact with
the R-31 specimen.
II - 518
Figs. 4 and 5 are overlays of tests on all twenty-three specimens. The maximum impact force
is shown both as a function of drop height and core density. Least square generated curves
are generated for the portion of the curves indicating a leveling-off at the high drop heights as
a result of the inability of the specimen to sustain additional impact load.
For low drop heights, the R-51 and R-71 curves in Fig. 4 are linear as indicated by the dark
lines. The R-31 curve is less linear in this range indicating the potential that some damage
has already occurred. Impact duration in milliseconds is plotted vs. drop height for each of the
three core density specimens. Fig. 5 shows data and a linear curve fit for each density foam.
Even though there is scatter, the figure demonstrates that lower density cores contribute
dramatically to the ability of the panels to decelerate the drop weight more slowly, thereby
extending the drop height at which catastrophic failure occurs. As a practical measure, we
attempted to compare the performance of sandwich panels to that of laminated panels of
equivalent numbers of layers, panel dimensions, and fixture.
Prior impact tests were conducted by the authors on 16, 32, and 48 ply laminated panels [7].
While these panels were made of quasi-isotropic layers of uniaxial laminates, the results
attained on 16 ply panels parallels closely the current study. Fig. 6 shows a comparison of the
threshold damage energy levels in panels in each study. Note that the energy required to
produce incipient damage in the R-31 sandwich and the 16 ply laminates are virtually
identical. What is remarkable is the energy required to produce threshold damage in
sandwich panels. R-51 and R-71 density core sandwich composites with 9 ply plain weave
graphite laminate face sheets are about as resistant to incipient damage as are laminated
panels of 32 ply and 48 ply, respectively.
While sandwich panels are robust in absorbing energy without suffering incipient damage,
they lack resistance to damage at higher energies. This is clearly shown in Fig. 7. In this
case, the energy producing maximum force and damage of the front face sheet is plotted for
each of the three density foam sandwich panels. Beside these data are plotted the equivalent
energies producing backface spalling in 16, 32, and 48 ply laminated panels. Here there is
substantial improvement in the laminated thicker panels.
II - 519
Calculations were also made comparing the coefficients of D [bending stiffness] Matrix for
the sandwich specimens with those of the laminated specimens of [7]. While bending stiffness
coefficients for the laminated panels varied greatly, those of the sandwich panels were almost
identical no matter which density core is used; they varied by less than 1%. In ongoing
studies, deflection measurements were made on sandwich beams made of the same panels of
this study. Such materials deflected up to four times the amount predicted by bending theory.
This indicated that there is substantial shear deformation in low-density sandwich panels.
This may also help explain the resistance of these panels to low energy impact.
CONCLUSIONS
This work shows that damage in sandwich shells and laminated composite structural panels is
substantially different. A major difference is the localization of damage in sandwich panels.
In this case, the face sheets become more loaded in membrane rather than in bending.
Another difference is the compliance of the core in sandwich panels as a factor in their
damage resistance. While we did not present extended data about NDE techniques, we C-
II - 520
SCANNED all tested panels. That information is currently being analyzed, however it has
been difficult to determine the nature and degree of damage to sandwich panels through CSCAN techniques.
Another interesting result is the substantial effect of core density on impact force. Higher
impact forces were developed at a given drop height in panels made of stiffer cores even the
classical laminate theory bending stiffness coefficients hardly changed. However, impact
forces developed only until the material strength limit was reached no matter how much the
impact energy was increased. At threshold damage levels, there was little difference in contact
forces developed.
There was also no indication that the inertial of sandwich panels had an affect on the
maximum force sustained at peak energy impact. Perhaps the most interesting finding was the
ability of sandwich panels to perform exceedingly well in resisting initial damage. In fact, the
higher density sandwich panels performed better than laminated panels of over 2.5 times their
specific weight. On the other hand, if failure of a sandwich panel is defined as failure of the
outer skin, then thin laminated panels perform about the same as sandwich panels while thick
laminated panels perform substantially better.
ACKNOWLEDGEMENTS
In closing, we wish to thank the NASA Center for Research Excellence for the support of
Mr. Derke Hughes and for assisting with materials and travel expenses. We also wish to
thank the Center for Composite Materials Research for the use of fabrication facilities
and Wright Patterson Air Force Base for the use of structural testing and NDE facilities.
REFERENCES
1.
Kelkar, A., "Analysis of Composite Laminates Subjected to Low Velocity Impact

Loading", Wright-Patterson Air Force Base, AFB 45433-6553, June 1992, pp. 4-5.
2.
Granville, D. M., "Moisture Effects of Polymethacrylimide Foam and Honeycomb

Comb Core in Sandwich / Skin Structures", 43rd Annual National Forum and
Technology Display of the Helicopter Society, St. Louis, MO, May 1987.
3.
Cristescu, N., Malvern, L. E., and Sierakowski, R. L., "Failure Mechanisms in

Composites Plates Impacted by Blunt-Ended Penetrators", Foreign Object Impact
Damage to Composites, ASTM STP 568, American Society for Test and Materials,
1975, pp. 159-172.
4.
McQuillen, E. J. and Gause, L. W., "Low Velocity Transverse Normal Impact of

Graphite Epoxy Composite laminates," Journal of Composite Materials, Vol. 10,
January, 1976.
5.
Caprino, G., and Teti, R., "Impact and post-impact behavior of foam core sandwich
structures", Composite Structures, 1994, pp. 47-55.
6.
Kuhhorn, A., and Schoop, H., "A nonlinear theory for sandwich shells including the
wrinkling phenomenon," Archives of Applied Mechanics, 1992, pp. 413-427.
7.
Kelkar, A. D., Craft, W. J., and Sandhu, R. S. "Study of Progressive Damage in Thin
and Thick Composite Laminates Subjected to Low Velocity Impact Loading." PVPVol. 13, Recent Advances in Structural Mechanics, ASME 1993.
II - 521
PERFORATION CHARACTERISTICS PREDICTION

OF MULTI-LAYERED COMPOSITE PLATES
SUBJECTED TO HIGH VELOCITY IMPACT
Hideaki Kasano and Kenichi Abe

Department of Mechanical Systems Engineering, Takushoku University
815-1 Tate-machi, Hachioji City, Tokyo 193, Japan
KEYWORDS: high velocity impact, impact perforation, composite layered plate analytical
modeling, residual velocity, ballistic limit velocity, conservation laws
II - 522
II - 523
II - 524
II - 525
II - 526
II - 527
II - 528
II - 529
II - 530
CONCLUSION
II - 531
FINITE ELEMENT ANALYSIS OF THE BALLISTIC

IMPACT RESPONSE OF COMPOSITE HELMET
MATERIALS
J. van Hoof 1, M.J. Worswick 1, P.V. Straznicky 1, M. Bolduc 2, S. Tylko 3
1
Department of Mechanical and Aerospace Engineering, Carleton University,

1125 Colonel By Drive, Ottawa, Ontario, K1S 5B6, Canada
2
Defence Research Establishment Valcartier, 2459 Pie XI Blvd. North,
PO Box 8800, Courcelette, Quebec, G0A 1R0, Canada
3
Biokinetics and Associates Ltd., 2470 Don Reid Drive,
Ottawa, Ontario, K1H 8P5, Canada
SUMMARY: The response of aramid fibre-reinforced composites to ballistic impact is

investigated using the finite element method. A user material routine incorporating fibre
breakage, delamination, and matrix crushing has been developed and implemented within the
LS-DYNA3D code. Post failure response has been implemented for the various failure
modes using the MLT approach. Penetration and backplane delamination due to a 1.1g
fragment simulating projectile (FSP) have been predicted for flat plates. The effect of post
failure modelling on the extent of damage and on the backplane response is shown to be
significant.
KEYWORDS: ballistic impact, post failure, delamination, fibre breakage
INTRODUCTION
Fibre-reinforced laminate composite materials offer a variety of desirable characteristics for
ballistic helmet applications such as light-weight, high strength, good damage tolerance, and
tailorability. However, a major disadvantage of composite materials is the existence of the
so-called backplane deformation. This deformation mode occurs through delamination of the
composite after partial penetration by the projectile. In such cases, complete penetration of
the helmet may be prevented, but the motion of the delaminated backplane can be sufficient to
cause serious head injury to the helmet wearer. In order to support further optimisation of the
ballistic performance of composite materials, a more thorough understanding of the
mechanisms causing backplane deformation is required. This paper presents a numerical
study into the parameters determining the extent of backplane deformation.
For a given impact velocity, the extent of penetration and backplane deformation will be
determined by the amount of energy transferred from the projectile to the composite material.
Langlie and Cheng [1] indicated that punching of the material, fibre breakage, and
delamination are the major mechanisms of energy absorption in ballistic composite materials.
While a variety of models exist for the prediction of failure initiation in composite materials,
recent work [2,3,4] has shown that post failure models can significantly improve damage
predictions and reduce mesh sensitivity in models of low velocity transverse impacts on
graphite fibre-toughened epoxy resin systems. In the current research, the effects of post
II - 532
failure behaviour on the predicted delamination and backplane response under ballistic impact
conditions are studied.
The impact of a fragment simulating projectile on woven Kevlar laminates is simulated using
the LS-DYNA3D [5] finite element code modified to incorporate user material subroutines
developed as part of this research. Models of 9 mm thick plates were considered under impact
velocities of 500 m/s.
CONSTITUTIVE MODEL
Failure Criteria
Failure criteria for fibre breakage and matrix fracture or crushing were based on a modified
version of the Chang and Chang model [6,7] implemented within a user material subroutine
written for this work. Delamination was modelled outside of this constitutive routine as
described below.
Since the current research considers woven composite materials, the mechanical properties
are assumed equal in both in-plane directions. Therefore, fibre breakage can be caused by
tensile and shear stresses acting in the in-plane directions, leading to the following failure
criteria
2
ij
fi = ri 0 : failure

j =1 Sij
3
(1)
< 0 : safe
where i=1,2, Sij are the in-plane strength values (either tensile or shear), and ri are the failure
thresholds which are equal to 1. When fibre breakage occurs in the i-direction, ij are set to
zero.
During impact, the composite material in the impacted area will be compressed by the FSP.
This will lead to high compressive stresses in the impacted area and high shear stresses in the
surrounding area. When these stresses exceed their related strength values, the composite will
fail and the FSP will punch a hole in the material. This penetration failure mechanism was
modelled using the following criterion
2
3j
r3 0 : failure
f3 =
j =1 S3 j
3
(2)
< 0 : safe
where S3j are the through-thickness strength values (either compressive or shear), and r3 is the
failure threshold and is equal to 1. When penetration failure occurs, all stresses are set to
zero.
II - 533
Post-Failure Model
The post-failure model used in this study, known as the MLT model, was developed by
Matzenmiller et al. [8] and is summarised below. This constitutive model was specifically
developed to describe anisotropic damage in elastic-brittle fibre-reinforced composites. The
elastic-brittle behaviour of these materials is characterised by the formation and evolution of
microcracks (surface discontinuities) and cavities (volume discontinuities). These defects are
considered irreversible and cause primarily stiffness degradation. A set of damage variables,
wi, were introduced to indicate the state of anisotropic damage and to reduce the stiffness in
the damage related directions. The model assumes linear elasticity if the damage state (state
of defects) does not change. This implies linear elastic unloading and reloading until the
previous damage state is reached again. All non-linear effects of the constitutive behaviour
are attributed to damage. Further, the orthotropic nature of the material is maintained
throughout the damaging process. The stress-strain response according to the MLT-model,
indicating the stiffness reduction due to damage, is depicted in Fig. 1.
Following Matzenmiller, et al. [8], the stiffness reducing damage variables wi are introduced
in the compliance matrix C
1
(1 w ) E
1
11
12
E1
13
E1
[C] =
0
21
E2
1
(1 w 22 ) E2
23
E2
31
E3
32
E3
1
(1 w33 ) E 3
1
(1 w12 )G12
1
(1 w23 )G23
(1 w13 )G31
(3)
and the stiffness matrix is obtained by inverting the compliance matrix: S = C-1
The MLT model is based on classical continuum damage mechanics which assumes that only
the undamaged part of the cross-section (net-area) carries loading, i.e. transmits stresses [9].
Consequently, the failure criteria should be defined in terms of the effective stresses, referred
to the net area. In the MLT damage model, the effective stresses and the nominal (true)
stresses are related by
{ } = [M] {}
where,
II - 534
(4)
1
1 w
11
0
[M ] =
0
1
1 w 22
1
1 w 33
1
1 w12
1
1 w 23
1
1 w13
If we substitute Eqn. 4 into the previously defined failure criteria and write the resulting
criteria in terms of strains, we find the following failure criteria for the MLT-model for
example for longitudinal tensile fibre failure:
2
c66 13
c + c + c13 33
c 44 12
g1 = 11 11 12 22
r1
+
+
(1 w11 )S11
(1 w12 )S12
(1 w13 )S13
(5)
The symbol gi is used to indicate that the failure criteria are expressed in terms of strains
instead of stresses. The damage variables are represented by Weibull distribution functions,
e.g., for longitudinal tension
w 11 = 1 e
1 11
me f1
(8)
with,
f1 =
Xt
E1
(9)
which is the nominal failure strain for a linear elastic material (in the absence of damage
growth). Similar relations hold for the other damage parameters. The exponent m in Eqn. 8
determines the shape of the Weibull function. Fig. 2 shows the effect m has on the postfailure strength. High values of m approach instantaneous failure.
The damage thresholds ri indicate the boundary of the elastic range. The material behaves
linear elastically within these boundaries. The role of the damage threshold is equal to that of
the yield stress in plasticity. The damage thresholds are initially set to 1, but will
monotonically increase with increasing damage once failure has occurred. When at a certain
time the strain state exceeds the boundary of the elastic range, the damage thresholds and
damage variables need to be updated. Matzenmiller et al. [8] derived a method to
accommodate coupling of growth for the individual damage variables in the various damage
modes. For reasons of simplicity, damage growth coupling was not implemented in these
preliminary calculations and the damage variables only grew when the corresponding strains
increased.
II - 535
The above model was implemented as a user material subroutine for brick elements in LSDYNA3D. A numerical advantage of the MLT-model is that the damage level and, therefore,
the failure mechanism, is based on the strain state. Thus, the duration of unloading after
failure directly depends on the strain rate, which is significant in high velocity impacts. The
elements that failed in penetration are deleted when w33 is equal to 0.999.
Delamination Failure
Delamination is implemented in the current model by meshing the bottom layer of the
composite separately and connecting this layer to the rest of the plate with so-called tiebreak
interfaces (slide line type 9 in LS-DYNA3D). The tiebreak interface rigidly connects the
nodes of the slave segment to those of the master segment until the failure criterion is
exceeded. The tiebreak failure criterion is given by
2
Fdelam

= n + s
Sn
Ss
(10)
where n and s are the normal and shear interface stresses, respectively, and Sn and Ss are the
corresponding strengths. Delamination occurs when Fdelam > 1. At this point, the interface is
allowed to separate, simulating inter-ply cracking. A similar approach was undertaken by
Hung et al. [9], who also modelled delamination in LS-DYNA3D with tiebreak interfaces and
their numerical results were in good agreement with experimental data.
MATERIALS
Elastic properties used for the woven aramid material studied are listed in Table 1. The
strength values for the composite are given in Table 2. The 4340 steel FSP material was
modelled using a yield strength of 1,034 MPa and a hardening modulus of 685 MPa. Young's
Modulus was taken as 207 GPa and Poisson's Ratio was 0.3.
Table 1: Elastic properties for woven aramid from [1]

Young's Moduli
(MPa)
Poisson's Ratio
E11
E33
21
31
28,269
5,516
0.05
0.07
Shear Moduli
(MPa)
Density
(g/cm2)
G12
G31
2,579
2,579
1.439
Table 2: Strength properties for woven aramid, partly from [1]

Tensile Strength
(MPa)
S11
S22
Sn
482.6 482.6 100.0
II - 536
Compressive
Strength (MPa)
S33
350
S12
34.5
Shear Strength
(MPa)
S23
S31
350
350
Ss
100
The strength values in Table 2 were obtained from [1], except for Sn, Ss, S33, S23, and S31
which were chosen to give reasonable results. These values are physically unrealistic, but
planned impact tests on plates and helmets will provide more realistic values for these
strengths.
MODEL
A finite element model of a woven aramid composite plate was built using I-DEAS Masters
Series, Version 2.1. The planar dimensions of the plate were 100 mm by 100 mm and the
thickness was 9 mm. The plate was impacted by a .22 calibre chisel-nose Fragment
Simulating Projectile (FSP, MIL-P46593). The FSP was made of 4340 steel and weighed 1.1
grams. The impact speed of the FSP was 500 m/s.
A typical mesh of the flat plate model with the FSP is shown in Fig. 3. Both the plate and the
FSP were modelled with 8-node brick elements. To reduce the computational time, only onehalf of the problem was modelled through symmetry. Fig. 4 shows a close-up of the impact
site and FSP.
An automatic eroding contact algorithm was used to simulate contact and penetration between
the FSP and the impacted area of the plate. This algorithm uses a penalty function approach
to prevent penetration of contacting elements. In the event of failure of the composite through
the erosion criterion given by equation Eqn. 3, the failed elements are deleted and the contact
surfaces are automatically updated to consider the next layer of material. A delamination
plane (tiebreak interface) was inserted adjacent to the inner-most layer of elements.
RESULTS
Figure 5 compares the effects of post failure modelling on the FSP velocity time histories. In
the instantaneous failure model, the stresses are immediately set to zero once the material fails
in a certain mode. As already shown in Fig. 2, the stiffness degradation in the MLT model
strongly depends on the value of m. Figure 5 shows that a higher amount of post failure
strength leads to a higher energy transfer from the FSP to the composite material, which
results in greater deceleration of the FSP. When the energy absorption is sufficient, complete
penetration of the composite is avoided and backplane deformation occurs. This is the case
for m=2, where the FSP is completely stopped by the backplane and rebounds. In all other
cases the backplane is completely penetrated by the FSP. The predicted final velocity of the
FSP after complete penetration depends heavily upon the post failure treatment. Figure 5 also
shows that the MLT predictions approach those using instantaneous failure for high values of
m.
The backplane displacement time histories for a node located at the centre of the delaminating
backplane are shown Fig. 6 as function of the post failure exponent m. The only case for
which the backplane was not penetrated by the FSP was for m=2. In all other post failure
cases the backplane was initially pushed down by the FSP, but eventually failed in crushing.
Therefore, the nodal displacements for these cases are not representative for the total
backplane displacement. The instantaneous failure case was omitted because the backplane
area under the FSP was completely eroded.
II - 537
Figure 7 shows a series of deformed mesh plots for m=2. Penetration of the upper layers
occurs through crushing and transverse shear failure as seen in Figs 7 (a) and (b). The nature
of the eroding contact algorithm can be observed as a progressive removal of elements ahead
of the penetrator. As the FSP nears the backplane its velocity decreases, resulting in lower
normal compressive strain rates and through-thickness shear strain rates. This enables the inplane stresses to build up before the element is eroded. Therefore, the dominant failure mode
will gradually change from crushing to fibre breakage. When the FSP reaches the backplane
its velocity will be so low that crushing of the underlying composite material will no longer
occur and the composite will fail in fibre breakage and delamination. This agrees with
Langlie and Chengs findings [1], who observed that the failure process in a composite plate
under ballistic impact can be visualised by three subsequent layers dominated by throughthickness failure, in-plane tensile failure, and delamination failure, respectively. Figure 8
shows the case for m=10, where the backplane is initially pushed down, but eventually fails in
crushing.
DISCUSSION AND CONCLUSIONS

The foregoing shows that the damage exponent m has a considerable effect on the impact
response of the composite material. Unfortunately, m has no physical meaning and cannot be
directly determined from experiments. Williams and Vaziri [4] suggested that m can be
numerically calibrated to give the best representation for a certain impact test. Impact tests on
flat plates and helmets are underway and these tests will hopefully yield a reasonable estimate
for m. In addition to m, the numerical damage predictions are also strongly affected by the
strength values. Material testing on the woven composite material and the aforementioned
impact tests will yield better estimates for these strength parameters.
Despite its limitations, the implemented post failure model yields more realistic results than
could have been obtained with the instantaneous failure models currently implemented in LSDYNA3D. The MLT model is a promising and versatile numerical tool for simulating
progressive damage in fibre reinforced composite materials.
In future, mesh refinement and delamination layers between each ply layer or between groups
of plies at regular intervals through thickness will further improve the current model. Post
failure modelling in the delamination criterion and linking the constitutive model with the tiebreak interface model will also yield a more realistic response.
ACKNOWLEDGEMENTS
Financial support for this research was provided by the Defence Research Establishment
Valcartier and the Natural Sciences and Engineering Research Council of Canada.
REFERENCES
1.
Langlie, S. and Cheng, W., A High Velocity Impact Penetration Model for Thick
Fiber-Reinforced Composites, ASME, Pressure Vessels and Piping Division
(Publication), 174, New York, 1989, pp. 151-158.
II - 538
2.
Majeed, O., Worswick, M.J., Straznicky, P.V., Poon, C., Numerical modelling of
Transverse Impact on Composite Coupons, Canadian Aeronautics and Space Journal,
Vol. 40, No. 3, 1994, pp. 99-106.
3.
Straznicky, P.V., Poon, C., Worswick, M.J., Fuoss, E., Majeed, O., Vietinghoff, H.,
Damage Resistance in Composite Materials, Proceedings of ICCM-10, Whistler,
British Columbia, Canada, August 14-18, Vol. V: Structures, pp. 607-614.
4.
Williams, K. and Vaziri, R., Finite Element Analysis of the Impact Response of CFRP
Composite Plates, Proceedings of ICCM-10, Whistler, British Columbia, Canada,
August 14-18, Vol. V: Structures, pp. 647-654.
5.
Hallquist, J.O., LS-DYNA3D Users Manual (Nonlinear Dynamic Analysis of

Structures in Three Dimensions), Version 936, Livermore Software Technology
Corporation, Livermore, CA, May 1995.
6.
Chang, F.K. and Chang, K.Y., Post-Failure analysis of Bolted Composite Joints in
Tension or Shear-Out Mode Failure, Journal of Composite Materials, Vol. 21, 1987,
pp. 809-833.
7.
Chang, F.K. and Chang, K.Y., A Progressive Damage Model for Laminated
Composites Containing Stress Concentration, Journal of Composite Materials, Vol.
21, 1987, pp. 834-855.
8.
Matzenmiller, A., Lubliner, J. and Taylor, R.L., A Constitutive Model for Anisotropic
Damage in Fiber-Composites, Mechanics of Materials, Vol. 20, 1995, pp. 125-152.
9.
Kachanov, L.M., Introduction to Continuum Damage Mechanics, Martinus Nijhoff,

Dordrecht, 1986
10.
Hung, K.-S., Nilsson, L. and Zhong, Z.-H., Numerical Studies on the Delamination
Mechanism in Laminated Composites under Impact Loading, Proceedings of ICCM10, Whistler, British Columbia, Canada, August 14-18, Vol. V: Structures, pp. 623-6
II - 539
II - 540
II - 541
PERFORATION CHARACTERISTICS OF FABRICS

MADE OF HIGH STRENGTH PE FIBERS
K. Takahashi and H. Komatsu
Research Institute for Applied Mechanics, Kyushu University, Kasuga, Fukuoka 816, Japan
SUMMARY: Perforation study was performed for textile fabrics made of high strength PE
fibers to study effects of textile structure, strand twisting and fiber surface treatment on
perforation energy, i.e., fabrics resistance at perforation. An air gun facility was used to
accelerate a steel bullet in a velocity range 50-200 m/s. High speed photography was
performed to observe high speed deformation of specimen cloth until breakage. The results
shows that two kinds of perforation modes exist depending on the textile structure, and the
perforation energy depends on the bullet velocity.
KEYWORDS: polyethylene, fabrics, ballistic impact, perforation
INTRODUCTION
Textile fabrics made of high strength fibers have been widely used for industrial and leisure
purposes. As the textile fabrics are regarded as high impact materials, they have been utilized
to protect mankind under the conditions in which people may be injured by projectiles. For
example, specially designed clothes such as space suits and bulletproof jackets are made from
high impact textile fabrics. Therefore, it is important to understand the impact resistance of
fabrics to know the limit of such clothes and to improve their properties. Also, fabric
composites have been developed by impregnating fabrics into resins. Therefore, it is useful to
characterize the impact properties of fabrics to know the impact resistance of the composites.
Aramid fibers such as Kevlar 29 and 49 have been widely used for high impact clothes, while
high strength polyethylene (PE) fibers have been also developed. Since some of PE fibers
have lower density and higher strength than the aramid fibers, the PE fibers may have an
advantage over the aramid fibers as fabrics for high impact clothes.
A few attempts have been made to study impact performance of textile fabrics. Montgomery
studied the effects of the shape of projectiles on impact energy absorbed by fabrics [1,2].
Wilde et al. studied impact performance of fabrics based on high speed photography [3,4].
Recently, Morrison et al. investigated the perforation phenomena of aramid laminates [5]. On
the other hand, Leech et al. studied impact behavior of cloths and nets by projectile
theoretically, and proposed a model to explain the behavior [6].
In the above studies, the velocities of the projectiles were in the range of several hundred
meter per second. On the other hand, in the present study, impact testing was performed at
relatively low bullet velocities (<200 m/s). Dependence of the bullet velocity on the
perforation energy absorbed by PE cloths was mainly investigated.
II - 542
EXPERIMENTAL
Material
High strength polyethylene cloths tested in the present study are shown in Table 1. In Table 1,
surface treatment means that the oil used during the process of fiber forming is removed.
Photographs of typical cloth surfaces and schematic drawings of the corresponding textile
structures are shown in Fig.1. The mechanical properties of the row fiber which is used for the
fabrication of cloth, are as follows:
density: 0.98 g/cm2, tensile strength: 2.92 GPa, tensile modulus: 137 GPa.
Air Gun
A multipurpose air gun was used for the impact testing and the testing facility is shown in
Fig.2. The air gun consists of a high pressure vessel, a launcher barrel and a low pressure
vessel in which cloth specimen is held. The length of the whole apparatus is 4,000 mm. High
pressurized air of 5 liter up to 150 atm can be filled up in the high pressure vessel using air
compressor or high pressure gas cylinder. The inner diameter and the length of the launcher
barrel are 25 mm and 2,000 mm, respectively. The diameter and the length of the low
pressure vessel are 600 mm and 1,100 mm, respectively, and the inside pressure is reduced to
about 10-2 torr. This vessel has two pairs of window for observation. The projectile (sabot)
used for the present study is made of polyethylene, and the geometry is shown in Fig.3.
Measurement of Bullet Velocity and High Speed Photography
The whole system for the measurement of bullet velocity and high speed photography is
shown in Fig.4. A chrome steel bullet (9.5 mm diameter and 3.55 g mass) is placed in the
front of the polyethylene sabot, and the accelerated sabot stops at the sabot stopper placed at
the end of the launcher barrel. As a result, only the steel bullet flies and collides against cloth
specimen. The bullet velocity Vi before the collision is measured from the time difference
when the bullet crosses over the two laser lights as shown in Fig.4. The deformation process
of the cloth specimen after the collision and the bullet velocity Vo after the perforation are
obtained using multiplex photography with transmitted light pulse emanated from the multiflash.
Absorbed Perforation Energy
If a steel bullet collides against cloth specimen at velocity Vi and the velocity reduces to Vo
after perforation, then the perforation energy absorbed by the cloth can be estimated from the
difference of the kinetic energy of the bullet before and after the perforation, i.e.
E i Eo =
1
2
m ( Vi 2 V o2 )
(1)
where Ei and Eo are the kinetic energies before and after the perforation, respectively, and m
the mass of the steel bullet.
II - 543
Fixation of Cloth Specimen

Cloth specimen is fixed perpendicular to the direction of bullet flight and the bullet collides at
the center of the cloth. Detail of the specimen fixture is shown in Fig.5 (1). An example of
cloth specimen after impact test is also shown in Fig.5 (2). Figs.5(3) and (4) show the twodimensional distribution of elongation for unperforated cloth A obtained at Vi =59.9 m/s, and
Fig.5(5) show the side view of the same specimen just after the impact. The side view of
perforated cloth A obtained at Vi =76.8 m/s is shown in Fig.5 (6).

High Speed Photography
Typical examples of high speed photography are shown in Fig.6(a) for cloth A and Fig.6(b)
for cloth B. In the caption of Fig.6, PI expresses the interval of light pulses. For both cloth A
and B, the figure (1) shows an unperforated result photographed at a velocity very close to the
critical velocity at which perforation occurs. The figures (2), (3) and (4) show perforated
results obtained at higher impact velocities. It is clearly seen from Fig.6 that for cloth A, the
yarns are cut off completely and there is no pull-out of yarn after the perforation; on the other
hand, the steel bullet perforates through the cloth B pulling out several yarns. It is also noted
that for cloth B, different pull-out modes can be observed depending on the velocity Vi. In the
present study, cloths A, A and D show the same tendency of perforation with cutting of yarns
as shown in Fig.6(a). On the other hand, for cloths B, B, C and E, steel bullet perforates
through each cloth accompanied with pull-out of yarns as shown in Fig.6(b).
Critical Perforation Velocity and Critical Perforation Energy
When a steel bullet collides a cloth specimen, the bullet cannot perforate through the cloth at
the velocities lower than a critical value of impact velocity. The bullet perforates though the
cloth when the impact velocity reaches a higher value than the critical velocity. This critical
velocity is called the critical perforation velocity Vic and the corresponding energy absorbed
by cloth is called the critical perforation energy Ec. For each cloth, Vic and Ec are shown in
Fig.7 and Fig.8, respectively. The results shown in Fig.7 and 8 are very useful to estimate the
perforation resistance of each cloth. However, these results may not present enough
information to compare the perforation energies absorbed by the cloths which have different
conditions such as the size of yarn, strand twisting, textile structure and raw material. In this
study, the specific perforation energy is defined by
E R = E c WR
(2)
where WR is the areal weight density, which is the weight of cloth per 1 m2. For each cloth,
ER value is shown in Fig.9. Although ER does not reflect the effect of each of the above
conditions on the perforation energy, it is a convenient method to evaluate the perforation
energy from comprehensive point of view.
The followings are obtained from the data for the specific perforation energy ER.
II - 544
(1) Effect of surface treatment

Treated cloth absorbs larger energy than untreated one even if the yarn size, the number of
twists, the textile structure and the areal weight density are the same in both cloths. It appears
that this is mainly because the frictional force between yarns in the treated cloth becomes
larger than in the untreated one owing to the removal of oil used during fiber forming process.
(2) Relation between the areal weight density and perforation energy
Regarding plain woven untreated polyethylene fiber cloth, the specific perforation energy
absorbed by cloth A is much lower than those of cloths C, D and E as shown in Fig.9. The
difference of the mean values is approximately 10 J/kg/m2. The critical perforation energy Ec
is shown as a function of the areal weight density WR in Fig.10. A linear relation between Ec
and WR is observed in Fig.10. It is also observed that Ec of cloth A is lower than that of cloth
D, although both cloths have the almost same WR value. It may be understood that this is due
to the effect of the number of twists considering that the other conditions are the same in both
cloths. Thus, it is expected that the number of twists, 160 times/m, for cloth D gives higher Ec
value than 80 times/m for cloth A.
On the other hand, the Ec value of basket woven untreated polyethylene cloth B is lower by
20 % than the extrapolated value of the plain woven untreated cloths. It is expected that a
group of basket woven cloths also has a linear relation between Ec and WR , and a straight
line situated at a lower position than the line for the plain woven cloths may be drawn in
Fig.10. This is further discussed in the following.
(3) Relation between the difference of textile structure and perforation energy
As described above, the perforation energy absorbed by plain woven cloth is larger than that
absorbed by basket woven cloth. This relation may originate in the difference of textile
structure. By comparing plain woven and basket woven cloths, in which the size of yarn and
the yarn density are the same, it can be said that the frictional force between yarns in plain
woven cloth is larger than in basket woven cloth, because the number of the crossing points of
warp and fill for plain woven cloth are four times that for basket woven cloth. This results in
the larger perforation energy absorbed by plain woven cloth.
Dependence of Perforation Energy on Vi
It is interesting to know the Vi-dependence of the perforation energy absorbed by cloth when
steel bullet collides the cloth at relatively low velocities ( 200 m/s) exceeding the critical
perforation velocity. Dependence of the perforation energy, Ei-Eo, on Vi is shown in Fig.11.
Fig.11 (a) shows the classification of marks used in the figures (b)-(h). The figures (b), (c) and
(g) show the examples of the case in which steel bullet cuts off the yarns. On the other hand,
the figures ( d), (e), (f) and (h) show the examples of the case in which steel bullet perforates
through the cloth accompanied with the pull-out of yarns. From these experimental results,
the following can be shown with respect to Vi-dependence of perforation energy.
(1) Relation between the difference of pull-out modes and perforation energy
Perforation energy absorbed by cloth depends on the pull-out mode, i.e. the number of yarns
pulled out. In general, larger number of pull-out results in larger absorbed perforation energy.
II - 545
(2) Vi-dependence of perforation energy

Perforation energy of polyethylene cloth tends to increase with increase of Vi in a certain
range of velocity greater than the critical velocity and decrease at higher velocities. However,
for cloth B (Fig.11(d)) and cloth E (Fig.11(h)), wide scatters are observed in the data because
the perforation energy varies with the pull-out mode. In general, the perforation energy tends
to increase with increasing Vi. In Figs.11(d) and (h), Vi-dependence of the perforation energy
is not clear for the velocities greater than 200 m/s. Considering the Vi-dependency of the other
cloths, it is expected that for cloth B and E, the perforation energy also decreases with
increase of Vi for Vi>200 m/s, and then reaches a constant value. This Vi-dependency appears
to depend on the shape and size of projectiles and the range of the impact velocity [2].
Generally, absorbed perforation energy by cloth due to the collision of steel bullet depends on
the way of the specimen fixation and the mechanical properties of the row fibers. It is also
considered that the pull-out resistance of yarn, the change of textile density generated by the
collision and the microdamage morphology of yarn and filament including heat effects have
influence on the perforation energy. A further study is needed for understanding the detailed
mechanisms.
CONCLUSIONS
Perforation study of high strength polyethylene fiber cloths was performed in a velocity range
50-200 m/s using an air gun facility. The followings have been shown by the study.
1. The perforation energy initially increases linearly with increase of the bullet velocity.
However, it tends to decrease over a limit which is around at 150 m/s.
2. The critical perforation energy, which is defined as the minimum energy required to cause
perforation, can be shown as a linear function of areal weight density of fabrics.
3. Two kinds of perforation modes, that is, cutting of yarns and pull-out of yarns are observed
depending on the textile structure.
4. Effects of fiber surface treatment as well as of strand twisting are shown to exist.
REFERENCES
1.
T.G. Montgomery, PhD Thesis, 1980, North Carolina State University.
2.
T.G. Montgomery, P.L. Grady and C. Tomasino, Textile Research Journal, 52 (1982),
442.
3.
A.F. Wilde, Textile Res. J., 44 (1974) 772.
4.
A.F. Wilde, J.J. Ricca, J.M. Rogers and L.M. Cole, Polym. Eng. Sci., 12 (1972) 41.
5.
C.E. Morrison and W. H. Bowyer, Proc. ICCM 2, (1980) 233.
6.
C. Leech, J.W.S. Hearle and J. Mansell, J. Text. Inst., 70 (1979) 469.
II - 546
II - 547
II - 548
II - 549
II - 550
II - 551
USING A HOMOGENIZATION PROCEDURE FOR

PREDICTION OF MATERIAL PROPERTIES AND
THE IMPACT RESPONSE OF UNIDIRECTIONAL
COMPOSITE
A. D. Resnyansky 1 and E. I. Romensky 2
1
School of Mechanical and Manufacturing Engineering, UNSW, Sydney 2052, Australia

2
Sobolev Institute of Mathematics Siberian Division of Russian Academy of Sciences,
Novosibirsk 630090, Russia
SUMMARY: The aim of the work is to design hierarchical model for description of the
dynamic behaviour of unidirectional composite. On the basis of the Maxwell-type viscoelastic
model [1] a procedure of derivation of dynamic equations for the composites is proposed. The
space arrangement of constituents in the direction transversally to the symmetry axis is out of
consideration. In the present paper the bonding between the constituents is assumed to be
perfect. Comparing data of the known theories and experiments with the results obtained, the
effective elastic characteristics are in good agreement for fibrous composites, and are
coincident with the known theory [2] for laminated composites. Assuming finite elastic
deformations, the procedure gives the equations containing the well-known plastic spin. The
model obtained is applied to the problem of the shock-wave propagation in an obliquely
loaded laminated composite. The shock-wave splitting is observed that is in agreement with
the experiment [3]. The phenomenon is explained by analysis of peculiarities of the loaddeformation curve. The cause is seen in the variation of the strength properties of the
composite due to the reinforcement rotation during the loading.
KEYWORDS: hierarchical model, homogenization approach, shock-wave propagation, offaxis loading, laminated composite
INTRODUCTION
Practical needs in the consideration of composite as a homogeneous anisotropic medium gave
a rise in popularity of homogenization procedures. This consideration provides quite correct
general description of the process of loading. Simple homogenization rules work quite well in
numerous applications and they are very convenient for the finite-element computations of
structures containing composite members. However, in some cases the micromechanical
peculiarities are of interest and description of residual stresses and internal microstresses is
required.
Micromechanical approaches give a useful information but they are restricted to the
consideration of individual fibre [2, 4-6]. Therefore, the homogenization approach is best
suited for an analysis the composite as a whole. Most of the known procedures aim to derive
II - 552
elastic moduli of composite in order to use the anisotropic theory of elasticity in finite element
computations.
Our objective is to design a procedure for derivation of evolution equations. In doing so the
effective elastic characteristics are a by-product. Three-level hierarchy is established during
the application of the procedure. The low level is associated with the microstructural process
in the composite's constituents (irreversible processes in the lattice caused by dislocations and
microdefects). In the paper this level is realized by the description of the constituents by the
Maxwell-type viscoelastic model [1]. The second level is associated with the micromechanics
of the composite and results from the description of the fibre-matrix interaction. Finally, the
upper level is presented by the description of the composite as a whole (the phenomenological
description) within the framework of the conservation laws and constitutive equations; these
latter are derived from the basic model (the low level) and the homogenization procedure (the
second level).
The resultant equations can also be used in the finite element computations. An application of
the model to the three-point bending problem is presented in the proceedings [7]. The
Onzager principle, expressing the symmetry of inelastic reactions to the load, is valid for the
model. This principle is of fundamental significance for the design of finite-element algorithm
because it enables one to derive the symmetrical stiffness matrix.
THE HOMOGENIZATION APPROACH: BRIEF DESCRIPTION

Homogenization procedure has to take into consideration a great number of factors. The
composite is a very complex structure and the interaction between components is hard to
analyse. Therefore, a certain degree of idealisation is necessary to describe both the composite
as a whole and at least the main features of the constituents' interaction.
The main problem in the design of the phenomenological theories is to take into consideration
as many factors as possible and to make the derivation as simple and utilitarian as possible.
As a basis, we take quite complex model for the description of the constituents. This gives us
some reserve in manipulation by assumptions of the second level. Because the utility of
derivation and thermodynamical correctness of the model are among our aims, we select a
simple set of assumptions for the description of the constituents interaction.
Due to the presupposed perfect bonding of components the equality of velocities in the fibres
and matrix is the case. The load transfer between the components is quite complex and
depends on the length and quality of fibres. Distribution of radial and circumferential stresses
around fibre and the interaction of the stress fields among individual fibres and matrix are also
nontrivial issues in the micromechanics of composites. Taking as an example the fibrous
unidirectional composite, we consider that the following assumptions are absolutely
necessary: i) the difference between the constituents stresses in direction of the fibre line can
not be neglected; ii) this difference induces internal microstresses acting in the longitudinal
direction. For simplifying assumptions we take: i) stresses in the transversal direction are
distributed uniformly in both composites components; ii) the mixture rule for stresses in the
longitudinal direction is valid. Mathematical representation of the assumptions stated above
is:
(
(
,
ij1) = ij2) = ij , ( i, j ) (11)
(1)
II - 553
Here the x1 - direction is assumed to be coincident with the fibre direction, the stresses inside
(
(
fibres are ij1) , inside matrix - ij 2) , `averaged' stresses of the composite as a whole - ij .
Hereafter, the upper indices in parentheses are associated with the number of a composite's
(1
constituent (1 - fibre, 2 - matrix). It is assumed that only the difference between 11 ) and
(
112) can not be neglected. This difference generates internal microstresses determined by
(1
(
variable 11 that presents the difference between elastic microstrains 11 ) and 112 ) :
(1
(
11 = 11 ) 112) .
(2)
(
Connection with the internal microstresses is obvious because microstresses ij k ) are related
(
to elastic microstrains ij k ) by the Hooke's law. An interesting peculiarity of this structural
parameter is that it is irreducible at elastic deformations of the composite as a whole including
the stress release. Therefore, 11 is connected directly with the residual stresses in the
composite.
(
On the microlevel the longitudinal stress 11 is a sum of the longitudinal microstresses ij k )
proportionally to the fibre volume fraction c :
(1
(
11 = c 11 ) + (1 c ) 112)
(3)
Thus, this procedure avoids the fibre arrangement, the component's bonding and local
distributions of the transverse stresses but enables the longitudinal microstresses to be taken
into account. Additional hypotheses for elastic microstrains and thermophysical variables
(temperature T and specific entropy s ) are:
T = T (1) = T ( 2 ) , s = c s (1) + (1 c ) s ( 2 ) ,
(1)
(1
11 = 11 ) 11 ,
(1)
(4)
(5)
( 2)
(
(
,
ij = c ( ij1) ij ) + (1 c )( ij2) ij ) , ( i, j ) (11),
(6)
(k)
here ij - microstrains induced in the constituents by the irreversible processes after the
stress release has occurred.
The origin of the least obvious hypothesis (5) results from the micromechanical consideration
of the loading-unloading cycle of unidirectional composite in the longitudinal direction. A
detailed analysis and derivation of (5) have been carried out in [8]. Absolutely similar
hypotheses are constructed for the unidirectional laminated composite.
Introduction of the assumptions for microstrains and temperature gives an opportunity to
connect all microstresses, microstrains and corresponding thermophysical characteristics to
the macrovariabes. If the equations of dynamic behaviour for each the composite's
components are given, then, using the hypotheses stated above, we can obtain: i) relationships
between stresses and strains (including the temperature and entropy); ii) equations of the
dynamics of composite.
II - 554
Detailed derivation for the case of small elastic deformations is stated for the unidirectional
fibrous and laminated composites in [8,9]. Generalisation for the case of finite elastic
deformations is based on the use of a transformation of a rectangular system associated with
fibre into the Cartesian coordinate system. In fact, the transformation is a matrix of the fibre
rotation and it determines fibre location in the stress release state.
The transformation of the tensor variables (strains, stresses, the structural parameter )
results in the appearance of the plastic spin in the right-hand sides, which are responsible for
the inelastic behaviour of composite. In general form the system of equations of the model is:
da ij
dt
+ a ik
uk
= a ik kj ,
xj
d ij
dt
dbij
dt
= ik bkj ,
= ij + ik kj ik kj ,
(7)
where
~
~
~
~
~
11 = 11 11 + 12 22 + 12 33 + 1 q ,
~
~
~
~
~
22 = 12 11 + 22 22 + 23 33 + 2 q ,
(8)
~
~
~
~
~
33 = 12 11 + 23 22 + 22 33 + 2 q ,
~
~
~
~
~
~
~
ij = 66 ij , 11 = 1 11 + 2 22 + 2 33 + q , i j
here ij , i, - the coefficients derived by use of the homogenization procedure from the
functions of the Maxwell-type model for the composites components [9], they depend on the
relaxation functions and the stress state in the constituents. The symmetry of coefficients in
(8) expresses the fundamental fact - the Onzager symmetry principle. In the system (7) ij are
components of a tensor which is responsible for irreversible change in the fibre orientation, it
~
is a given function when the functions ij are given. The tensor with components bij is a
~
~
tensor of irreversible rotation for fibre, q = E ~ , a tilde-variable f is a tensor in the
ij
rectangular system associated with either the fibre or the normal to lamina, it is expressed
~
through that in the Cartesian system by the standard transformation rule: f ij = bik f km bmj . The
system (7) is completed by the conservation laws for energy and momentum, and a given
dependence of the internal energy E on the strain tensor, structural parameter and entropy.
EFFECTIVE ELASTIC CHARACTERISTICS

To illustrate the influence of the fibre content on elastic muduli, the characteristics in this
section will be analysed for the case of isothermal deformation. Invoking the procedure of
homogenization, the Hooke's laws for constituents yield the stress-strain relations which are
typical for the transversely isotropic material:
II - 555
11 = C11 11 + C12 22 + C12 33 ,

22 = C12 11 + C22 22 + C23 33 ,
(9)
33 = C12 11 + C23 22 + C22 33 , ij = 2C 66 ij , i j

where
C11 = + 2 +
+
1
C22 =
+
1
+
1
+
, C12 =
+
+
1
1
, C23 =
+
1
1
C66 = ( C22 C23 ) =
2
1
+
,.
(10)
here , are the Lame coefficients. Expressions in the angle parentheses present the mixture
rule K = cK 2 + (1 c )K 2 . K = cK 2 + (1 c )K 2 . The longitudinal and transverse Young's
moduli E1 , E 2 , the shear modulus G2 and the longitudinal and transverse Poisson's
coefficients 1 , 2 can be easy linked with C ij:
2
2C12
E1 =
,
C22 + C23
E2 =
(C
22
2
C23 ) C11 ( C22 + C23 ) 2C12
(C
11
2
C22 C12 )
2
C12
C11 C23 C12
G2 = C66 , 1 =
, 2 =
2 .
C22 + C23
C11 C22 C12
],
(11)
It is interesting to note that the procedure applied to the laminated structure gives the elastic
moduli which coincide identically with those stated in the popular monography by
Christensen [2]. Therefore, we shall analyse the dependencies of the elastic constants on the
fibre content only for the fibrous composite.
The well-known estimates for an effective modulus of the composite K c (K f - the respective
elastic modulus of fibres, K m - the elastic modulus of matrix) are the Voigt estimate (the
mixture rule) and the Reuss estimate, respectively:
K c = cK f + (1 c )K m , 1 K c = c K f + (1 c ) K m ,
where
c - the fibre volume concentration.
The longitudinal Young's modulus E1 for the composite in most of theories is in very good
agreement with the Voigt estimate (the rule of mixture), experiments are in agreement with
this rule, too. Our formulas are not an exclusion and they are practically coincident with the
linear dependence of the mixture rule.
II - 556
Fig. 1: The transversal Youngs modulus (a) and the rigidity modulus (b) for the glass/epoxy
fibrous composite
Theoretical and experimental dependencies are not so simple for the transversal Young's
modulus E 2 and the rigidity modulus G2 . The well-known approximation for the moduli is the
Reuss estimate. It is interesting to note that our dependence for the rigidity modulus is exactly
the Reuss estimate. Experiments stated in [4] give the moduli (points in Fig. 1) for the
glass/epoxy composite (elastic constants for fibre and matrix are E f = 731G Pa ,
.
G f = 302G Pa , f = 022, E m = 345 G Pa , G m = 18G Pa , m = 035). Our results
.
.
.
.
.
(curves 1 according to equations (10), (11)) are compared in Fig. 1 with the theory of [4] and
the Reuss estimate of the effective moduli. It is seen that the discrepancy between all the
theories and experiment is of the same order.
Experimental data for the graphite/epoxy-resin composite by Goggin [10] are stated by points
in Fig. 2. Elastcic data for the constituents are: the type 1 fibre E f = 395G Pa , f = 028; the
.
type 2 fibre
E f = 230 G Pa , f = 028; the fibre in the transversal direction:
.
E f = 276G Pa, f = 028; the epoxy resin system: E m = 39 G Pa , m = 033. Fig. 2, (a)
.
.
.
.
corresponds to the transversal Young's modulus E1 , Fig. 2, (b) - the rigidity modulus G2 , Fig.
2, (c,d) - the longitudinal and transversal Poisson's ratios 1 and 2 , respectively. Curves 1,
2, 3 in Fig. 2, (a) correspond to the three types of fibres mentioned above according to the
formulas (10), (11), curve 1 in Fig. 2, (b-d) - to the fibre type 1 calculated by the same
formuli.
The data by Goggin have quite large scatter, especially for the Poisson's ratio. The Young's
modulus of the fibre type 2 is less than that of the fibre type 1. However, the transverse
Young's modulus of the composite with the fibre type 2 is higher than that with the fibre type
1 (Fig. 2, (2)). Appearently, this feature is associated with the strong anisotropic properties of
fibres. Our model of the constituents is an isotropic material. Therefore, we calculated curve 3
in Fig. 2 (a) for a fibre with the Youngss modulus corresponding to the experimental
transversal Youngs modulus of the graphite fibre.
Among theoretical data the well-known formuli are those by R. Hill stated in [5]. For the
constants from (10) they take the form:
C11 = + 2 c (1 c )
( )
1

+ + M
, C12 = c (1 c )
( )( +
1
2 2 )

+ + M
II - 557
( 1 2 )
1
,
C 66 = ( C22 C23 ) = c (1 c )

2
+M
2
(12)
( + )
1
( C22 + C23 ) = + c (1 c ) 1 1 2 2 ,
2
++M
2

here f = (1 c ) f 1 + cf 2 . It is easy to check that at M = 0 these formuli reduce to
dependencies (10). Because M has the meaning of the shear modulus we take as an example
M = . Calculations according to (12) give the curve 4 in Fig.2, (a) and curves 3 in Fig. 2,
(c-d). The semi-empirical dependencies by C.Chamis stated in [6] are the curve 5 in Fig. 2, (a)
and curves 2 in Fig. 2, (b-d). Our formuli describe experiments on the Poissons ratio in the
best way among theoretical data mentioned above. For the transversal Youngs modulus and
the rigidity modulus the discrepancy between all the theories and experiment is again of the
same order.
Fig. 2: The effective elastic characteristics for the graphite/epoxy composite. The transversal
Youngs modulus (a) (points are the experiments [10]); the rigidity modulus (b); the
longitudinal Poissons ratio (c); the transversal Poissons ratio (d)
The comparison gives us a confidence in use of the homogenization approach for the
derivation of dynamic equations.
THE SHOCK WAVE RESPONSE OF COMPOSITE

To illustrate the potential of the model a computation of the shock wave propagation in
obliquely oriented laminated composite has been conducted. The method of calculation of the
system (7) is the finite-difference Godunov scheme. The composite structure has been chosen
in accordance with an experimental sample by Sve and Okubo [3], where the Al/Lexan
composite has been oriented symmetrically (Fig. 3,a) and obliquely (Fig. 3, b) to the shock
wave.
II - 558
Fig. 3: Sketches for the straightward and obliquie shock-wave loading of the laminated
composite
In order to understand the process of the off-axis loading let us consider the loading of the
representative volume of the composite by uniaxial stress. The sketch of the process is drawn
in Fig. 4. The corresponding load-deformation curve computed by the model shows that in
certain range of orientations the stress response is complex due to rearrangement of the
composite structure during the loading. At the symmetrical loadings (the cases 1 and 3) the
composite strengths correspond to those of the fibre or matrix. For the off-axis loading (the
case 2) the composite strength changes during the deformation, depending on the
reinforcement direction.
Fig. 4: Schemes of the off-axis loading and calculation of the load-deformation curve for the
case of rotation of the reinforcement
Coefficients , , of the functions ij in (8) which are responsible for the inelastic
behaviour of the constituents are constructed from experimental dependencies of the yield
limit on the strain rate. The method for the construction of the functions can be found in [11].
II - 559
Fig. 5: Calculation of the shock-wave response of the symmetry-oriented and obliquely

oriented laminated composite; 1 - the velocity profile, 2 - the microstress profile, 3 - the
change in the fibre orientation
The shock-wave propagation calculated for the samples in Fig. 3 has a number of
peculiarities. At symmetrical loading (velocity versus coordinate in the sample at a moment of
time - the curve 1, Fig. 5, a), the fibre orientation does not change during the process (curve
~
3), (curve 2) associated with the internal microstress (the difference in elastic strains
between the constituents) changes only after the elastic wave has passed and the plastic wave
has induced the microstress in the sample.
However, in the case of off-axis loading (the orientation angle 0 = 47 ) the fibre orientation
becomes an essential factor due to the rearrangement stated in Fig. 4. Initially, the elastic
wave acts as in the preceding case (velocity - curve 1 in Fig. 5, b), then the successive load
changes the layer orienation and modifies the strength properties of the sample. It is seen from
Fig. 5,(b) that in the second wave the angle orientation has been changed (curve 3), a rise in
~
(curve 2) indicates that the composite in these points is out of elastic range and the change
is associated with the structural rearrangement. The third wave performs the final loading up
to the limit determined by the striker.
Fig. 6: Comparison between the calculated (curve 1) and the experimental [3] (curve 2)
velocity profiles
The comparison of the calculation with the experiments [3] shows good agreement between
the velocity profiles (Fig. 6).
II - 560
CONCLUSION
A homogenization procedure for unidirectional composite is proposed in the paper. The aim
of the approach is to design a three-level hierarchical model for description of the dynamic
behaviour of the composite materials. Effective elastic characteristics obtained by the
procedure for the fibrous and laminated composite are within the range of the well-known
theoretical and experimental data. The model is suitable for description of the shock-wave
data for composites. The calculation of the obliquely loaded composite is in good agreement
with the known experiment [3].
REFERENCES
1. Godunov, S.K. and Romensky, E.I., "Nonstationary Equations of Theory of Nonlinear
Elasticity in the Eulerian Coordinates", Prikl. Mekh. i Tekh. Fiz. [Engl. transl. Appl.
Mech. Techn. Phys.], Vol. 13, No. 6, 1972, pp. 124-144.
2. Christensen, R.M., Mechanics of Composite Materials, John Wiley, New York, 1979.
3. Sve, C. and Okubo S., "Experiments on Pulse Propagation in an Obliquely Laminated
Composite", Journal of Applied Mechanics, Vol. 39, 1974, pp. 1052-1056.
4. Tsai, S.W. and Hahn, H.T., Introduction to composite materials, Technomic Pub.,
Westport, 1980.
5. Tewary, V.K., Mechanics of Fibre Composites, John Wiley, New York, 1978.
6. Chawla, K.K., Composite Materials, Springer-Verlag, New York, 1987.
7. Resnyansky, A.D., Simulation of Flexure in Off-axis Three-Point Bending Test by a
Finite-Element Code, The Present Proceedings.
8. Resnyansky, A.D. and Romensky, E.I., "Model of Viscoelastic Unidirectional Composite
with Coupled Thermal and Elastic processes", Proceedings 19th International Symposium
on Shock Waves, Marseille, France, July 26-30, 1993, Vol. III: Shock Waves in
Condensed Matter and Heterogeneous Media, Brun, R., Dumitrescu, L.Z., Eds, pp. 273278.
9. Resnyansky, A.D. and Romensky, E.I., "Model of Dynamic Deformation of a Laminated
Thermoviscoelastic Composite", Combustion, Explosion, and Shock Waves (Fizika
Gorenia i Vzryva), Vol. 29, No. 4, 1993, pp. 535-541
10. Goggin, P.R., The Elastic Constants of Carbon-Fibre Composites, Journal of Materials
Science, Vol. 8, 1973, pp. 233-244.
11. Resnyansky, A.D. and Merzhievsky, L.A. "On Determination of Parametres of
Dependence of The Shear-Stress-Relaxation Time", [In Russian], Fizika Gorenia i
Vzryva, Vol. 28, No. 3, 1992, pp. 123-124.
II - 561
IMPACT DAMAGE AND FAILURE MECHANISMS IN

STRUCTURE RELEVANT COMPOSITE SPECIMENS
J. F. M. Wiggenraad and L. C. Ubels
National Aerospace Laboratory NLR, P.O. Box 153, 8300 AD Emmeloord, The Netherlands
SUMMARY: Impact damage and failure mechanisms, representative of stiffened composite

wing panels, were studied using small but structure relevant (SR) specimens. It was investigated
how laminate stacking sequence influences the location within a laminate of the major
delaminations that are generated by impact. Impact damages for six different lay-ups showed
major delaminations predominantly at the ply interfaces where they were expected.
Subsequently, the residual strength of the SR specimens was determined in compression. The
major impact induced delaminations were also the delaminations that propagated under loading,
driven by bending or buckling of the 0-degree plies. However, with increased loading, transverse
shear of the delaminated sublaminates became the ultimate failure mechanism. Thus it was
concluded that the residual strength of the SR specimens depends primarily on the reduced
global stability of the damaged laminate, and less on the individual stability of the delaminated
sublaminates. A failure criterion is needed, based on longitudinal compressive strain and
transverse shear strain in 0-degree dominated ply stacks.
KEYWORDS: design, impact, damage tolerance, modelling, testing
INTRODUCTION
The development of structural concepts for compression loaded, damage tolerant, stiffened
composite wing panels requires an experimental investigation of the failure mechanisms that
may occur. Insight into these mechanisms will lead to design guidelines for the structural
configuration considered, and can be incorporated in design optimization procedures. One of the
most critical states of damage to consider is the presence of one or more delaminations between
the plies of a laminate, caused by low velocity impacts. Such delaminations tend to spread when
the panel is loaded in compression, reducing the overall stability of the panel until ultimate
failure occurs. (Delamination "spread" is defined as the quasi-static or dynamic propagation of a
delamination under increased static loading, as opposed to delamination "growth" under fatigue
loading).
The fabrication and testing of full scale stiffened panels is an expensive procedure, so for
economic reasons most experiments are commonly performed on relatively simple specimens:
plain laminates with corresponding material properties. When testing a laminate in the form of a
small specimen supported by a frame, the damage resulting from the impact event as well as the
residual compression strength may deviate considerably from the experimental values that are
obtained when the laminate forms part of a larger structural configuration. Structural properties,
such as the different dynamic response during impact and the presence of multiple load paths in a
larger structure are thought to be responsible for the often superior damage tolerance observed
when testing structures instead of small specimens.
II - 562
At ICCM-9, a small Structure Relevant (SR) specimen was described [1], which is less
expensive to make and test than a full scale panel. When combined with a support to simulate the
stiffener [1-2], it maintains the essential design features of the panel configuration that is being
investigated, see Fig. 1. The SR specimen was shown to be useful in providing insight in
delamination spread mechanisms. In this study, delaminations were provoked by the insertion of
circular bronze foils. However, a proper evaluation of the damage tolerance of different design
concepts requires that genuine damage is induced in the SR specimens by impact. This impact
damage must correspond to the damage resulting from impacts on the stiffened panels that are
represented by the SR specimens. At ICCM-10 a method was presented to apply structure
relevant impact damage to SR specimens [3]. It was shown that the flexibility present in a larger
structure can be approximated by selecting the proper support conditions for the SR specimen. It
appeared that if the support conditions are such that the maximum deflection measured during
impact at the impact site is approximately the same for stiffened panel and SR specimen, this will
result in similar damage.
Fig. 1: Concept of structure relevant specimen

Based on the previous studies [1-3] it was believed that the failure mechanism related to the
damage configuration considered, consists of the spread of a few preferred delaminations out of
the multitude of delaminations and matrix cracks present in the impacted area, followed by
sequential buckling of the delaminated sublaminates. When the overall stability of the panel
reaches a critical value, ultimate failure occurs. As delamination spread occurs mainly in the
lateral direction (perpendicular to the stiffeners) it is expected to start from delaminations with
the widest span. The findings reported in [1-3] also revealed that the major impact induced
delaminations are formed along certain preferred ply interfaces. Hence, it was postulated in [3]
that by a deliberate positioning of these preferred interfaces deeper inside the laminate, thicker
sublaminates will be formed by the impact. As thicker sublaminates have a higher resistance
against buckling, delamination spread and ultimate failure are expected to be delayed when the
specimen is loaded by increased compression.
In the present study the results are presented of a test programme, in which SR specimens with
six different lay-ups were provided with impact damage and subsequently loaded in compression
up to failure. The objectives of this programme were to investigate whether the major impact
induced delaminations occur at the interfaces where they are expected, and to evaluate the
validity of the postulated failure mechanism.
II - 563
PANEL DESIGN AND SR SPECIMEN CONFIGURATION

The configuration of the SR specimens was based on a damage tolerant stiffened carbon/epoxy
panel concept developed by Boeing [4], now applied in the V-22 Osprey wing. This concept
comprises a soft (low axial stiffness) skin, doublers (extra plies added locally to the skin
underneath the stiffeners) and discrete (pre-cured) stiffeners, see Fig. 2. A panel design with
I-stiffeners was determined for an ultimate load level of 2000 N/mm, using the PANOPT
optimization code for stiffened panels [5], see Fig. 3. One of the six SR specimen lay-ups was
identical to the skin/doubler configuration of the panel design, while the other lay-ups were
variations thereof. The lay-ups of the six different SR configurations are listed in table 1 and are
shown in Fig. 4. In the present paper, results are discussed for five specimens of each lay-up.
Fig. 2: Boeings damage tolerant panel concept with soft skin, doublers and discrete stiffeners
Fig. 3: Design specifications, geometry and lay-up (dimensions in mm)

Table 1: Laminate configurations
II - 564
Low velocity impact damage was applied to the first three specimens of each lay-up with an
instrumented impactor, at the flat skin opposite to the edge of the doubler as shown in Fig. 4. The
impactor used had a mass of 2.312 kg and a semi-spherical tup with a diameter of 1.0 inch (25.4
mm). The impact energy was 35J for the first specimen and 50J for the second and third
specimens of each lay-up. The fourth specimen of each lay-up was statically indented up to the
same deflection as the maximum deflection measured during the 35J impacts. The fifth specimen
of each lay-up was left undamaged. All damaged specimens were C-scanned, and the first
specimen of each lay-up was dissected for a post-mortem investigation of the impact damage.
After impact damage was induced, the specimens were loaded in static uni-axial compression up
to failure (or up to a compressive strain of 0.0070 in case of several of the undamaged
specimens), mounted in the anti-buckling guide shown in Fig. 5. Displacements were measured
with LVDT's, and strains with strain gauges at locations also shown in Fig. 5. The failure mode
was established with post-mortem fractography for specimens '3' of each lay-up.
Fig. 4: Lay-ups of six SR-specimens with major impact induced delaminations (cross-section A
A refers to figure 8)
II - 565

Impact Damage
Impact parameters are not discussed, except to indicate that the static and dynamic (35J)
responses of the SR specimens were very close, as shown in Fig. 6 for Lay-up 1. This similarity
may warrant the approach to design for "static" rather than dynamic impacts, an observation
which may simplify optimization procedures in the future, as static responses are easier to
compute. The first specimen of each lay-up was dissected, and the major delaminations were
made visible with uv-light, illuminating a penetrating fluid (Fig. 7). The cross sections shown are
located at a distance of approximately 20 mm from the impact site to eliminate the smaller
delaminations. The location of the major delaminations relative to the stacking sequence are
indicated in Fig. 4 for all six lay-ups.
Fig 5: Test configuration of SR-specimens

In the following discussion the damage is described with reference to the impact direction: the
front is the flat, impacted side of the laminate. Most lay-ups contained two 0-degree ply stacks
and all lay-ups contained two 90-degree skin plies. The major impact induced delaminations
were expected to occur behind the 0-degree ply stacks and along the 90-degree plies when these
were nearby. Lay-up 1 shows two major delaminations behind the two 0-degree ply stacks, and
II - 566
along or close to the 90-degree plies which are located behind these ply stacks, as expected.
Lay-up 2 shows three major delaminations: two behind the 0-degree ply stacks, very deep inside
the laminate, but also one in front of the first ply stack, very close to the free surface. This
undesirable delamination is caused by the presence of a 0/90 degree ply interface in front of the
first ply stack. Lay-up 3 also shows two major delaminations behind the 0-degree ply stacks.
However, another large delamination is found in the wide space between the two thick 0-degree
ply stacks, behind a thinner stack of two 0-degree skin plies that pass through this space. Lay-up
4 shows two large delaminations behind the two doubler ply stacks, as well as two smaller
delaminations in front of each doubler ply stack. The two continuous 0-degree skin plies passing
in front of the two doubler ply stacks are apparently responsible for these smaller delaminations.
Lay-up 5 has a discrete doubler ply stack located on top of an uninterrupted flat skin. The
damage, very different in character compared to the damage of the other lay-ups, resembles
damage in a quasi-isotropic flat laminate, where the many ply angle changes (stiffness jumps)
result in a larger number of smaller delaminations. The widest ones are located along the two
90-degree plies. Lay-up 6 is identical with Lay-up 1, except for two bond films added
immediately in front of the two 90-degree plies. Just as for Lay-up 1, the major delaminations are
located behind the doubler ply stacks. The bond layers apparently had no benevolent effect on
the size of the impact damage.
Fig. 6: Dynamic response (35J) versus static response for lay-up 1

Ranking the six lay-ups for expected "damage tolerance", based on the location and size of the
major impact induced delaminations, and in correspondence with the postulated failure
mechanism of buckling sublaminates, the following order (from best to worst) was tentatively
established: Lay-ups 1, 6, 3, 2, 4 and 5. Apparently, to "place" delaminations deep inside a
laminate, and to avoid the formation of additional undesirable delaminations at other locations,
the number of ply angle changes should be limited. This can best be achieved by stacking the
0-degree doubler stacks adjacent to the 0-degree plies of the skin, by positioning the 90-degree
plies immediately behind the 0-degree ply stacks where they will certainly provoke the formation
of the major delaminations, and by surrounding these "planks" with uninterrupted 45 ply stacks.
II - 567
Fig. 7: Major delaminations in SR specimens

Failure Mechanisms
Specimens 2-5 of each lay-up were loaded in compression up to failure, or up to a compressive
strain of 0.0070 for some of the undamaged specimens (to prevent damage to the test
instrumentation). The maximum loads achieved are shown in table 2. The size of the total C-scan
damage area is also presented for these specimens. Most configurations had C-scan areas of
approximately 1800 mm. The C-scan areas for Lay-up 4 are somewhat smaller at 1500 mm,
and the C-scan areas of Lay-up 5 (with a discrete doubler on top) are showing a much larger
scatter, from 1450 mm to 2450 mm. This scatter is related to the large delamination occurring
between doubler and skin, which is less consistent than the "enclosed" delaminations of the other
configurations.
It appears from table 2 that the maximum loads of the undamaged specimens (specimens 5 of
each lay-up) are much higher than those of the damaged specimens, all corresponding to axial
compressive strains in excess of 0.0070. It should be noted that the axial stiffness of Lay-up 5 is
less than the stiffness of the other lay-ups, because its discrete doubler stack on top of the skin is
"softer" than 011 (to limit the stiffness difference between soft skin and stiff doubler). The failure
loads of indented specimens 4 are, like the damage areas, of the same order of magnitude as
those of the impacted specimens 2-3. For five of the six lay-ups the failure loads of the damaged
specimens were very close. For Lay-up 4 the difference between specimens 2 and 3 was much
larger (20%). This difference is due to the different failure modes observed for these two
specimens: bending of the damaged area toward the window in the anti-buckling frame support
(Fig. 5), or bending backward against the support. It must be concluded that the SR specimens do
not provide residual strength values representative of the stiffened panel configuration. Unlike
for the thicker (and stiffer) SR specimens described in [1-2], the absence of the stiffener can not
be fully compensated by the influence of the support, which can push but not pull. Hence, the
final failure mode of most specimens was by buckling in the direction away from the support
into the window. Two of the higher failure values were observed for specimens which bent in the
opposite direction, toward the support (specimens 2 of Lay-up 3 and Lay-up 4).
II - 568
Table 2: Damage areas and failure loads
However, to validate the postulated failure mechanism, of major delaminations growing and
sublaminates buckling, post-mortem cross sections (as indicated by A-A in Fig. 4a) were made
of specimens 3 of each lay-up, all of which had failed in the same mode: buckling away from the
support (opposite to the impact direction). Photographs of these sections are shown in Fig. 8 of
the skin/doubler zone, in which the 0-degree plies are visible as the lighter layers. The failure
loads of these specimens were 259/266J for Lay-ups 3 and 6, 229/230 J for Lay-ups 1 and 2, and
219/222 J for Lay-ups 4 and 5. Comparing this order of "damage tolerance" to the order
concluded above on the basis of the impact induced delaminations, there is not much difference:
Lay-up 3 was better than expected and changed positions with Lay-up 1, most likely because it is
an almost symmetric laminate and remained straight up to higher loads. Lay-up 6, thought to be
equivalent to Lay-up 1 because it has the same stacking sequence, while the added bond layers
had not limited the size of the damage, was actually better than Lay-up 1. The superior residual
strength of Lay-up 6 may be caused by a higher resistance against delamination spread or by the
improved stability due to the added thickness of the bond layers.
II - 569
Fig 8: Post-mortem cross-sections of SR specimens 3 in figure 4A (A-A)

Based on the most-mortem results it is concluded that the postulated failure mechanism is correct
in the sense that the major impact induced delaminations located behind the 0-degree doubler
plies are the delaminations that grow, driven by the buckling or bending of these ply stacks.
II - 570
However, the differences between the failure loads of the different lay-ups are smaller than
expected, indicating that buckling of the delaminated sublaminates is not the only acting failure
mechanism. From the photographs in Fig. 8 it is clear that ultimate failure occurs by transverse
shear of the unsupported 0-degree ply stacks. This failure mechanism does not seem to depend
much on the location of the 0-degree ply stacks, but rather on the amount of transverse shear
deformation that occurred. Increasing the global stability of the laminate, by maintaining
symmetry even in the doubler area (Lay-up 3) or by increasing the laminate thickness with bond
layers (Lay-up 6) apparently resulted in a higher damage tolerance. The global stability of an
impacted laminate is also increased by the presence of a stiffener. In fact, stiffened panels, made
according to the same design as shown in Fig. 2, and tested in compression with similar impact
damage [3], failed at much higher strains of 0.0062 and 0.0070 than the 0.0050-0.0055 observed
for the damaged SR specimens. The failure modes of the panels were not even related to these
damages, which illustrates the effect of the support of stiffeners in particular, and the damage
tolerance of Boeing's design concept in general.
CONCLUDING REMARKS
Impact damage and failure mechanisms were studied using small (unstiffened) structure relevant
(SR) specimens. By varying the stacking sequence of the skin/doubler laminate it was tried to
influence the location of the major impact induced delaminations within the laminate. It was
believed that delaminations located deeper inside a laminate would lead to improved damage
tolerance. Results for six different lay-ups were compared, and the major delaminations were
found at the ply interfaces where they were expected. In some cases, additional "undesirable"
delaminations were found also.
Subsequently, damaged and undamaged specimens for each lay-up were loaded in compression
while supported in an anti-buckling frame; the damage reduced the residual strength as expected.
For five of the six lay-ups the failure loads of the damaged specimens were very close. The
failure loads for the damaged specimens of the sixth lay-up were 20% apart due to the
occurrence of different failure modes. Specimens which were damaged by slow indentation
showed similar damage and similar residual strength as impact damaged specimens. This may
have important implications for design optimization procedures, as static responses are easier to
compute than dynamic responses.
A postulated failure mechanism was confirmed to a certain extent, in the sense that the major
impact induced delaminations were indeed the delaminations that propagated, driven by the
bending or buckling of the doubler ply stacks. However, transverse shear of the increasingly
unsupported 0-degree ply stacks became the ultimate failure mechanism. As a result of these
observations, it will be undertaken to develop a failure criterion based upon the interaction of
longitudinal compressive strain x and transverse shear strain xz for 0-degree dominated ply
stacks.
II - 571
ACKNOWLEDGEMENTS
The work on damage tolerance of composite structures described in this paper is part of a wider
research programme carried out at NLR. This particular task was carried out for the Netherlands
Agency for Aerospace Programs (NIVR).
REFERENCES
1. Labont, S. and Wiggenraad, J.F.M., "Development of a Structure Relevant Specimen for
Damage Tolerance Studies", presented at the 9th International Conference on Composite
Materials (ICCM-9), Madrid, Spain, 12-16 July 1993.
2. Labont, S. and Wiggenraad, J.F.M., "A Damage Tolerance Study Conducted with Structure
Relevant Specimens", NLR TP 93067 U, 1993.
3. Ubels, L.C. and Wiggenraad, J.F.M., "A Method to Apply Structure Relevant Impact Damage
to Small Structure Relevant Specimens for Damage Tolerance Studies", presented at the 10th
International Conference on Composite Materials (ICCM-10), Whistler, Canada, 14-18
August, 1995.
4. McCarty, J.E. and Roeseler, W.G., "Durability and Damage Tolerance of Large Composite
Primary Aircraft Structure", NASA CR-003767, 1984.
5. Arendsen, P., Thuis, H.G.S.J. and Wiggenraad, J.F.M., "Optimization of Composite Stiffened
Panels with Postbuckling Constraints", 4th CADCOMP, Southampton, UK, 1994.
II - 572
IMPACT DAMAGE GROWTH AND FAILURE OF

CARBON-FIBRE REINFORCED PLASTIC
SKIN-STRINGER PANELS
Emile Greenhalgh, Sunil Singh, Donald Roberts
Structural Materials Centre, DRA, Farnborough, Hants, UK, GU14 6TD
SUMMARY: This paper describes an investigation into low velocity impact damage in
skin-stringer panels. The effect of panel geometry and impact site on the damage growth and
compressive strength was investigated. Impact damage beneath a stringer was less detrimental
to strength than damage in the bay. In five of the six panels, failure was precipitated by a
combination of bay buckling and skin/stringer detachment, except in one panel where massive
lateral delamination growth led to a large reduction in strength. For impact damage in a bay, a
wide-bay/thick-skin design was the most damage tolerant whilst for impact damage beneath a
stringer, a high buckling strain design was the most tolerant. The results suggest that, at
current in-service strain levels, impact damage may be tolerated in some cases.
KEYWORDS: structures, impact, fractography, delamination, skin-stringer, buckling
INTRODUCTION
The specific in-plane properties of carbon-fibre composites make them invaluable for use in
aircraft structures. However, they are limited by their inherent susceptibility to impact and
other out-of-plane loadings. In coupons a low velocity impact can reduce the compressive
properties by up to 50% [1]. However, damage in coupons is more severe than that in
structures since, in the latter, much of the incident energy is absorbed through elastic
structural response [2]. Only through tests on structural elements (skin-stringer panels) can the
damage processes and failure of impacted composite structures be understood.
There has only been limited work conducted on the effect of impact damage on skin-stringer
panels. Madan showed that the greatest strength reduction was from bay impacts [3].
However, most studies have investigated softer skinned panels; i.e. containing few 0 plies
[4,5]. In these structures, impact over the stringers led to the greatest reductions in strength.
This paper describes the second part of an investigation into impact damage in skin-stringer
panels. In the first part of the work panels of different geometries were impacted at 15J at
various sites [2]. For impacts over the panel bay, the energy was mainly absorbed through the
formation of delaminations and matrix cracking. However, as the impact site approached the
stringer, or the bay aspect ratio increased, more energy was absorbed through elastic response,
leading to less severe damage states. Generally, the damage distribution through the thickness
was conical. In the bays, it was roughly circular in area whilst over the stringer feet it was
elliptical, with the major axis parallel to the stringers. The damage from impact over the
stringer centreline was negligible.
II - 573
In this part of the investigation identical panels were impacted either over the bay or the
stringer foot, and then tested to failure in quasi-static compression. The effect of panel
geometry and impact location were addressed, and compared with results from similar panel
tests with embedded defects [6].
EXPERIMENTAL DETAILS
The stiffened carbon-fibre reinforced panels (Figure 1), each made up of a quasi-isotropic
skin ([+45/-45/0/90]nS) and three I-section stringers, were manufactured. Seven panels, of
three different geometries were studied. The first (Type 1) was designed to buckle at a
compressive strain of 6000 whilst the second and third panels were designed to buckle at
3700. In the second case (Type 2) this was achieved by increasing the stringer spacing and
in the third case (Type 3) by decreasing the skin-thickness. One panel of each type was
compressively tested with impact damage either in the bay or in the stringer foot regions.
Further details of the manufacture and preparation of the panels are given elsewhere [2].
Figure 1: Geometries and impact sites in the skin-stringer panels

The panels were impacted using instrumented drop-weight equipment at an incident energy of
15J; all the impacts were applied to the skin on the outer (non-stringer) face. The panels were
impacted at the mid-length, 150mm from the panel ends, either in the left bay, or over the
tapered section of the central stringer foot. The seventh panel was subjected to a 30J impact at
the mid-length over the centreline of the left-hand stringer foot. The panel was of the Type 1
design and had originally been used to study the effect of artificial defects on the performance
of skin-stringer panels [6], but did not fail during testing.
After impacting, the dent depth and damage area (ultrasonic attenuation of 6dB or more
compared with best material) were measured. The panels were then compressively tested at a
rate of 0.005mm/s and the strains, displacements and loads recorded. Displacement
transducers and Moir interferometry were used to monitor out-of-plane deformations of the
damaged regions and the stringer deflections. Following testing to failure, the panels were
photographed, and examined to determine the damage growth and failure processes [7].
II - 574
When compared to the previous studies [2], the impact forces, deflections and damage areas
were found to be repeatable, particularly for impacts over the stringer feet. For identical
impacts, the damage was generally similar in shape and distribution through the thickness.
However, for impacts over the bay, the damage areas differed by as much as 26%.
Table 1: Results from the compressive tests on the skin-stringer panels
Panel Type
Type 1
Type 2
Type 3
Impact Location
Bay
Foot
Centreline
Bay
Foot
Bay
Foot
Damage Area (mm)
2141
778
50
675
824
534
1071
Buckling Strain ()
-5803
-7100
-4400
-4500
-3530
-4600
Failure Strain ()
-3662
-5803
-7464
-4543
-4964
-3597
-4697
Failure Load (kN)
-529
-803
-995
-734
-840
-460
-554
The results of the compressive testing of the skin-stringer panels are shown in Table 1. In
general, damage in the bay was more critical than damage beneath the stringer foot,
particularly for the Type 1 panel. Impact damage beneath the stringer centreline had a
negligible effect on the panel strength and did not initiate failure of the structure.
Figure 2 shows the typical damage development from an impact in the bay.
Figure 2: Development of the damage in a Type 3 panel with a bay impact

The image width is equivalent to the bay width and each Moir fringe is equivalent to an
out-of-plane deflection change of 0.2mm. The applied strain is shown below each image.
II - 575
For the panels with impacts in the bay the early damage development was similar for all three
types. Upon loading the impacted region formed a blister and the bay displaced outwards,
away from the stringers. The blister was elliptical, with the major axis perpendicular to the
applied load, and was highest in the Type 3 panel. However, the blister was smaller than the
damage observed from the non-destructive analysis. The damage grew laterally in a stable
manner from the impact site across the bay width. In the panels with 120mm bay widths, at an
applied strain of about -3500 the growth became more rapid, extending towards and
beneath the stringers. However, in the panel with a 148mm wide bay, the damage growth was
limited, and the damage front did not reach the stringers before panel buckling.
The Type 1 panel failed at a strain well below the design strain (-3662). The Type 2 and 3
panels failed at and beyond buckling, at strains of -4543 and -3597 respectively. The
Type 2 panel exhibited significant post-buckling performance and the buckling strain was
only slightly depressed. However, the Type 3 panel exhibited a significant fall in buckling
strain, and failed during panel buckling.
The performance of the panels containing impact damage beneath the stringer foot were very
different to those containing damage in the bays. As the load was applied the damaged region
started to displace outwards, with the largest displacement in the bay adjacent to the impact
site. There was little damage growth observed although there was some lifting of the stringer
foot from the skin. The Type 1 panel failed as it buckled at a strain of -5803 whilst the
Type 2 and 3 panels both buckled at values greater than their design strains. The Type 3 panel
failed soon after buckling (-4697) and the Type 2 panel had significant post-buckling
performance, failing at a strain of -4964.
Failure analysis (fractography) showed that five out of the six panels which had failed due to
impact damage, had failed in a similar manner. After the initial growth stages, the
combination of damage beneath the stringer foot and bay buckling precipitated detachment of
the stringer from the skin. The led to skin crippling and catastrophic panel failure. In the
Type 1 panel with damage in the bay, massive unstable delamination spread across the panel
width, leading to catastrophic failure of the skin.
DISCUSSION
Damage Growth and Failure Mechanisms
Before discussing the effect of various parameters, the mechanisms for damage growth and
structural failure should be considered. The overriding factor in controlling these mechanisms
was the location of the impact site; the failure processes of the panels with impact damage in
the bay was very different to that of the panels with damage beneath the stringer feet. An
impact over the stringer centreline had an negligible effect on the panel failure.
The failure processes for a panel containing impact damage in the bay is shown in Figure 3.
As the load was introduced the impacted bay deflected outwards, away from the stringer face
(Figure 3a). The damaged region also started to deflect, but inwards, towards the stringer face
(Figure 3b) and the damage started to grow from the impact site (Figure 3c). The damage
growth was predominantly transverse, growing towards the stringers. Initially the growth was
stable, but there reached a point where it became unstable and rapidly spreading across the
bay and beneath the stringers.
II - 576
Figure 3: Failure mechanisms for panels with an impacted bay

The later stages of damage growth were dependant on the panel type. In the panel designed to
the higher buckling strain (Type 1) the damage continued to spread across the panel width
(Figure 3d), rapidly developing into the undamaged bay and beneath all the stringers. Once
the damage had spread over 80% of the panel width, compressive failure of the remaining
skin initiated in the impacted bay and spread across the entire panel width (Figure 3e).
In the panels designed to the lower buckling strain (Types 2 and 3), the failure mechanism
was different. As the damage grew towards the stringers, the bays started to buckle. As in the
panels containing single plane defects in the bay [6] the presence of the damage reduced the
panel buckling strain. As the bays buckled, opening forces were promoted beneath the
stringers, leading to local detachment of the stringers from the skin (Figure 3f). This loss of
local support led to skin instability (crippling) and compressive failure (initiating at the
impact site), and finally the stringers detached and failed in flexure (Figure 3g).
All three panels which were impacted over the stringer foot failed by the same damage
process (Figure 4). Although the early stages of damage growth were different to those
observed in the panels containing impacts in the bay, the later stages of growth and failure
were similar to those observed in the Type 2 and 3 panels. Firstly, as the load was introduced,
the bay adjacent to the impacted stringer started to deflect outwards (Figure 4a), eventually
II - 577
leading to lifting of the stringer foot. There was no transverse damage growth, but damage
spread longitudinally, along the stringer foot (Figure 4b). The rate of damage growth was
much less than that observed in the panels containing damage in the bay, leading to higher
strains to failure. Eventually, the panels buckled which promoted partial stringer detachment
(Figure 4c). As with the impacts in the bay, this led to skin crippling, compressive failure
(initiating from the impact site) and failure of the central stringer (Figure 4d).
Figure 4: Failure mechanism for panels with an impacted stringer foot

The failure of the panel containing the impact damage on the stringer centreline was not
attributed to the impact damage. The failure initiated as massive lateral delamination growth
across most of the panel width, initiating from the artificial defect. This was followed by
stringer detachment and compressive failure of the skin.
Effect of Impact Location
As shown in Figure 5, the impact location had the largest effect on the panel performance.
Firstly, consider the Type 1 design of which there were three panels tested. Since the failure
of the panel containing the damage beneath the stringer centreline was not due to the impact,
this can be treated as a lower bound on the undamaged strength of this design. Considering
this, a 15J impact over the stringer foot led to a 12% drop in strength whilst a similar impact
in the bay led to a 51% drop in strength, the latter representing a significant drop in strength.
For the other two panel types the undamaged strengths were not known. However, there was
less reduction in strength between the stringer feet and bay impacts (9% and 23% for the Type
2 and 3 panels respectively) than was observed in the Type 1 panel (37%). The reduction in
the Type 3 panels was attributed to the bay damage reducing the buckling strain which
promoted stringer detachment. In the Type 2 panels, the failure mechanisms due to damage in
the bay and stringer foot were very similar, leading to the negligible drop in strength.
II - 578
(Type 2)
(Type 1)
(Type 3)
0
Failure Strain ( )
1000
2000
3000
4000
3662
3597
5000
6000
7000
4543
4964
4697
5803
8000
7464
Bay
Stringer Foot
Stringer Centreline
Design Strain
Figure 5: Failure strains for skin-stringer panels containing impact damage

Effect of Panel Geometry
For impact to the bay, there was a clear effect of panel type on the failure process. The Type 1
panel did not buckle and failed due to massive delamination growth. The Type 2 panel
exhibited little damage growth and failed by a combination of bay buckling and stringer
detachment. In the Type 3 panel, the failure was attributed to stringer detachment which was
promoted by the reduced buckling performance of the damaged bay. The higher failure strain
in the Type 2 panel was due to the slower development of the damage. This was attributed to
the impact damage being distant from the stringer feet and the out-of-plane deflection of the
damaged region being less than that in the other panel types. This subsequently led to two
factors which improved the damage tolerance for this design. Firstly, the reduced growth led
to the damage having less of an influence on buckling. Secondly, the damage had less effect
on the stringers, not promoting stringer detachment to the same extent in the Type 3 design.
For the damage beneath the stringer feet, there was less effect of panel geometry. Firstly, the
damage had little effect on buckling, therefore not promoting failure to the same extent as
from bay damage. The largest effect of geometry was observed in the Type 3 panels, again
due to the sensitivity of this design to stringer debonding.
Comparison with Embedded Defects
In earlier work [6], Type 1 panels containing Teflon discs between the plies were tested.
These defects were located at either of two different ply interfaces and were either in the bay
or beneath the stringer foot. The results were compared to the current work to gauge the
validity of modelling impact as single plane defects. There were some similarities between the
growth processes for the two types of damage. In both instances the damaged bay firstly
deflected outwards and the initial growth was transverse to the loading direction, leading to
the damage extending towards the stringer feet. However, at a given applied strain, the height
and width of the impact damage were much less than those of the artificial damage. This was
attributed to the impact damage consisting of small, deep defects which could not deflect
outward to the same extent that the relatively large, shallow single-plane defects could.
II - 579
As the damage front approached the stringer feet, the growth processes were accelerated. In
the panel containing impact damage in the bay this led to catastrophic compressive failure.
However, in the panels containing artificial defects, the growth promoted stringer detachment
only after panel buckling. For similar damage areas, the impacted panel had 60% of the
strength of the panel containing an artificial defect. In the panels containing artificial defects,
the growth was dominated by blistering of the surface plies and was subsequently dominated
by stable mode I growth [8,9]. In the impacted panel the damage was mainly deeper and mode
II dominated, so stable damage growth was limited. Furthermore, the local bending of the
damaged region subsequently led to unstable mode II growth from the impact site. When the
damage was beneath the stringers, the out-of-plane constraint on the skin inhibited mode I
dominated growth, as from the artificial damage, but would have had little effect on mode II
dominated growth, as from the impact damage. For both impact damage and artificial defects
beneath the stringer, the strength reduction was less than that from those in the bay.
IMPLICATIONS IN-SERVICE AND FOR DESIGN

The current results allow some prediction of the severity of the damage and suggest that in
many cases an impact over the stringer can be tolerated. Furthermore, damage beneath the
stringer had little effect on panel buckling and only for designs sensitive to skin/stringer
debonding did damage beneath the stringer become significant. When impact damage was
present in the bays, rapid unstable growth, sometimes leading directly to failure, occurred at
strains above -3500 (below most aircraft limit loads).
Based on studies from material testing, the current philosophy is to gauge the severity of an
impact by the damage area; the larger the damage area, the lower the failure strain, and this
relationship did hold for stringer foot impacts. However, for impacts to the bay, large damage
areas did not necessarily correspond to low failure strains since the panel geometry had a
significant effect on the failure processes.
Understanding of the damage growth and failure processes can be utilised to produce impact
tolerance in new structural designs. The driving force for impact damage growth was a
combination of mode II forces generated by bending of the damaged region, and mode I
forces from the blistering of the surface plies. However, unlike embedded defects, relatively
little stable damage growth was observed prior to panel buckling. The wide-bay/thick-skin
design exhibited the slowest damage development before failure and consequently had the
highest strength.
A measure with which to rank the damage tolerance is the load intensity/unit areal weight [6]
as shown in Figure 6.
All these panel designs were less tolerant of damage within the bay than beneath the stringer
foot, particularly for the high buckling strain structures (Type 1). Impacts over the stringer
foot did not pose a significant threat whilst those over the stringer centreline had no observed
effect on strength. It should be noted that all these designs had relatively hard skins; a
significant proportion of the load was carried by the skin. In soft skin designs (most of the
load is carried by the stringers) impact damage beneath the stringers may be more critical.
For the low buckling strain designs, the wide-bay/thick-skin approach (Type 2) was clearly
the more damage tolerant, particularly for impacts to the bay. Comparing all three panel types,
II - 580
Load Intensity/Unit Areal Weight (kN m/kg)
for bay impacts this design (Type 2) also out-performed the high strain buckling design
(Type 1). However, for stringer foot impacts, where the panel buckling precipitated failure in
all panel types, the high buckling strain design was the most damage tolerant.
300
Bay
Stringer Foot
Stringer Centreline
272kN m/kg
250
220kN m/kg
201kN m/kg
200
150
180kN m/kg
176kN m/kg
149kN m/kg
145kN m/kg
100
50
0
(Type 1)
(Type 2)
(Type 3)
Figure 6: Load intensity/unit areal weight for skin-stringer panels

CONCLUSIONS
The following conclusions have been drawn from investigations of the damage, growth and
structural failure in damaged skin-stringer panels under compressive loading:
1. Generally, the forces, displacements and damage areas were reproducible, except that the
damage areas from impacts over the bay which exhibited a large degree of scatter.
2. There was no direct relationship between impact damage area and failure strain.
3. In panels containing impact damage in the bays, the damage firstly spread laterally
towards the stringers, which in some cases significantly reduced the panel buckling strain.
During buckling, this led to stringer detachment, skin crippling and catastrophic failure. In
the high buckling strain design, massive unstable delamination growth occurred,
promoting failure.
4. In panels containing impact damage beneath the stringer foot, the local substructure
constrained and limited damage growth. Failure eventually occurred when the panel
buckled; the combination of the out-of-plane deflection of the bay and the presence of
damage beneath the stringer led to stringer detachment and catastrophic failure.
5. Panels containing impact damage in the bay exhibited a strength reduction of up to 37%
with respect to panels containing impact damage beneath the stringer foot. Impact damage
directly beneath the stringer centreline had a negligible effect on panel strength.
6. For the low buckling strain designs, a wide-bay/thick-skin was the most damage tolerant.
Of all the panels tested, this design was the most tolerant to impacts in the bay whilst the
high buckling strain design was the most tolerant of impact damage beneath the stringer
foot.
7. Panels containing single plane embedded delaminations were 60% stronger than panels
containing impact damage of the same dimensions.
8. At current in-service strain levels, impact damage may be tolerated in some cases.
Certainly for low buckling strain designs, there was no significant reduction in strength.
At current design strains for primary structures, damage may be tolerated in many cases.
II - 581
ACKNOWLEDGEMENTS
The authors acknowledge the support of DTI CARAD and the MOD Corporate Research
Programme.
The contributions of Mr S Lahiff, for designing the panels, Mrs A Dewar for panel
manufacture, Miss S Minshull and Mr R Thompson for assistance in panel testing and NonDestructive Evaluation Section are also acknowledged. Dr Sarah Bishop (MoD), Mr R Sharp
(BAe) and Prof G Davies (Imperial College) are also acknowledged for their general advice.
REFERENCES
1.
Gadke M. et al., GARTEUR Damage Mechanics for Composite Materials;

Analytical/Experimental Research on Delaminations, Proceedings of the 74th AGARD
Structures and Materials Panel, Patras, Greece, 1992
2.
Greenhalgh, E.S., Bishop, S.M., Bray, D., Hughes, D., Lahiff, S. and Millson, B.,
Characterisation of Impact Damage in Skin-Stringer Composite Structures, Composite
Structures, To be Published, 1997.
3.
Madan, R., Influence of Low Velocity Impact on Composite Structures. In Composite

Materials; Fatigue and Fracture, Edited by K OBrien, ASTM STP 1110, 1991.
4.
Avery, J., Allen, M., Sawdy, D. and Avery, S., Survivability Characteristics of
Composite Compression Structure. Proceedings of the 8th DOD/NASA/FAA
Conference, Fibrous Composites in Structural Design, Edited by Lenoe et al, Part 2,
1989.
5.
Ishikawa, T., Matsushima, M., Hayashi, Y. and Noguchi, T. Compression After Impact
Properties of CF/PEEK(APC2) and Conventional CF/EPOXY Composite Structures,
Proceedings of the 8th International Conference on Composite Materials, Honolulu,
Hawaii, 1991, pp28S1-28S10.
6.
Wiggenraad, J., Aoki, R., Gadke, M., Greenhalgh, E., Hachenberg, D., Wolf, K. And
Bubl, R., Damage Propagation in Composite Structural Elements - Analysis and
Experiments on Structures, Composite Structures, To be Published, 1997.
7.
Greenhalgh, E. and Cox, P., A Method to Determine Propagation Direction of

Compressive Fracture in Carbon-Fibre Composites, Composite Structures, Vol 21, 1992,
pp1-7.
8.
Greenhalgh, E., Characterisation of Mixed-Mode Delamination Growth in Carbon-Fibre

Composites, PhD Thesis, Imperial College of Science, Technology and Medicine,
London (to be published), 1997.
9.
Singh, S. and Greenhalgh, E., Micromechanisms of Interlaminar Fracture in

Carbon-Fibre Composites at Multidirectional Ply Interfaces, Proceedings of the 4th
International Conference on Deformation and Fracture of Composites, Manchester, UK,
24-26 March 1997.
II - 582
SIMULATION OF IMPACT DEFORMATION,

DAMAGE & FRACTURE IN COMPOSITE
LAMINATES
Da-Zhi Jiang 1 ,Wei Shen 2 and Xing-Ye Wang1
1
Department of Material Engineering & Applied Chemistry, National University of Defense

Technology, Changsha, Hunan, 410073, P.R. China
2
Department of Mechanics, Huazhong University of Science and Technology, Wuhan, Hubei,
430074, P. R. China
KEYWORDS: impact damage, finite element method, composite laminates
II - 583
II - 584
II - 585
II - 586
II - 587
II - 588
II - 589
II - 590
II - 591
II - 592
FABRICATION AND MECHANICAL PROPERTIES OF

GLASS MATRIX COMPOSITES WITH DIFFERENT
FIBRE ARCHITECTURES
Heinrich Kern, Volker Winkler and Aldo R. Boccaccini
Fachgebiet Wekstofftechnik, Technische Universitt Ilmenau, PF 100565
D-98684 Ilmenau, Germany
SUMMARY: Processing routes have been developed for the fabrication of glass matrix
composites with 2-dimensional woven fibre mats as reinforcement. Carbon and metallic
(stainless steel 316L) fibre mats were investigated. In order to achieve infiltration of the
matrix material into the fibres tows and an adequate adhesion between matrix and fibres, an
electrophoretic deposition (EPD) technique was used. The infiltration capability using both a
glass powder suspension and a silica colloidal solution was investigated and compared. Thus,
the processing routes involve three steps: 1) the infiltration of the fibre mats using
electrophoretic deposition, 2) the preparation of composite green bodies by stacking of
infiltrated prepregs and 3) the consolidation of the composites by a high-temperature
densification process involving pressureless sintering or hot-pressing. A high-level of matrix
infiltration in the intra-tow regions was achieved when a silica colloidal solution was used.
These samples could be densified by pressureless sintering. On the other hand, when a glass
powder suspension was used, there was not complete infiltration of the fibre mats. Porosity in
these samples could not be eliminated even after a hot-pressing stage.
KEYWORDS: glass matrix composites, woven fibre reinforcement, electrophoresis,

composites processing, metallic fibre reinforcement, hot-pressing, sintering
INTRODUCTION
The brittle nature, and therefore, high susceptibility to catastrophic failure of glass and glassceramic materials limits their use in structural applications. One approach to the improvement
of the thermomechanical properties of these materials is to produce a composite. In this way,
the low-modulus, low-strength, brittle matrix is reinforced by a high-modulus, high-strength
and/or high-ductility second constituent in the form of continuous or chopped fibres,
whiskers, platelets or particulates [1].
The best results in terms of increasing the fracture tolerance and imparting a pseudo-ductile
fracture behaviour to brittle matrices is achieved with continuous fibre reinforcement [1,2].
The vast majority of work concerned with fibre reinforcement of glasses and glass-ceramics
has focused on using ceramic fibres, especially in unidirectional and cross-ply fibre
alignments [2-7]. The reinforcement of glass matrices by 2-dimensional fibre architectures,
including ductile (metallic) fibre preforms, has been much less investigated [8], despite the
advantages that these architectures may have over their unidirectional counterparts. These are
related to the more isotropic mechanical properties obtained with 2-dimensional reinforcing
elements over unidirectional fibre reinforcement [9].
II - 593
The use of metallic reinforcement may have still further advantages, including an improved
resistance to fibre damage during composite processing due to the intrinsic ductility of
metallic fibres and the possibility of exploiting their plastic deformation for composite
toughness enhancement [10]. However, the authors are only aware of the work of Russian
workers on the use of 2-dimensional metallic fibre mats to reinforce glasses [8].
When using fibre mats as reinforcement, it is normally extremely difficult to achieve
complete infiltration of the matrix material into the fibre tows, where the openings are of the
order of 100 nm. Electrophoretic deposition (EPD) of colloidal ceramic sols has been
shown recently to be a simple and inexpensive method for achieving complete infiltration of
tightly woven fibre preforms [9, 11,12].
The technique has been employed for infiltrating silica and alumina sols into ceramic fibre
preforms, such as SiC (Nicalon) [9,11] and alumina woven mats [12], showing that EPD
provides a more efficient infiltration and stronger adhesion between fibre and matrix than a
simple slurry dipping technique.
In the present study processing routes to fabricate two types of glass matrix composites with
2-dimensional woven fibre mats as reinforcement are developed. Carbon and metallic
(stainless steel 316L) fibre mats were investigated. EPD was used to deposit matrix material
onto the fibre preforms. For the metallic fibre mat reinforced composites, which were
consolidated by pressureless sintering of the glass matrix, a silica sol was used first to
infiltrate the metallic fibre mats.
It was anticipated that the silica infiltration would not only contribute to reduce the intra-tow
porosity but also, acting as coating for the metallic fibres, would minimise possible reactions
of the metal with the glass matrix during high-temperature densification and would provide an
adequate interface in order to improve the composite fracture behaviour. A borosilicate glass
powder suspension was used to fabricate the carbon fibre mats reinforced composites.
The results of the EPD technique for both the colloidal silica and the glass powder suspension
were compared in terms of their ability to infiltrate woven fibre mats.
Figure 1 shows the processing routes employed in this study to fabricate glass matrix
composite materials. The details are given below.
Materials
Woven carbon fibre reinforced composites
Commercially available (Tenax, Akzo AG) woven carbon fibre mats were used, with a fibre
diameter of 10m. The glass matrix employed was a borosilicate glass (DURAN, Schott
Glaswerke, Mainz, Germany). The glass powder was milled and sieved so that all particles
were < 40 m. The density of this glass is 2.2 g/cm3 and its thermal expansion coefficient is
3.3 x 10-6 1/C [13]. DURAN glass was chosen because it is a well-known matrix material for
the fabrication of glass matrix composites [3] and because its viscous flow and sintering
behaviour are known [13]. For the EPD experiments a powder aqueous suspension containing
II - 594
40vol% solid and 0.1 wt% PVA was prepared. The suspension exhibited a strong tendency
to sedimentation.
Woven metal fibre reinforced composites
From the variety of metal fabrics commercially available [14], a woven satin 100% 316L
stainless steel fabric (Bekitherm FA, Bekaert SA, Zwevegem, Belgium) was chosen, on the
basis of its suitability to be infiltrated by sol particles using electrophoretic deposition. There
are other fibre mats available also, made from alloys such as Inconel 601, which have better
high temperature properties than 316L, but for this study, the more cost-effective 316L
stainless steel was considered adequate to demonstrate the use of this particular processing
technology. The individual fibres have an average diameter of 14 m and the weaves have a
nominal thickness of 0.4 mm [14]. The maximum using temperature of the 316L fibres in an
oxidising atmosphere is 900C and the thermal expansion coefficient (20-700C) is 16.8 x
10-6 1/C [15,16].
Figure 1: Processing route for the fabrication of 2-dimensional woven fibre mat reinforced
glass matrix composites
Due to this relatively low thermal capability and limited oxidation resistance of the metallic
fibre used, a glass with low sintering temperature was chosen for the matrix. The glass
selected was a soda-lime glass, with a composition corresponding to that used in the German
flat glass production [17]. The density of the glass is 2.51 g/cm3 and it was available as a
powder with an average particle size of 80 m. Its sintering and pressureless densification
behaviour have been investigated previously and an optimised sintering temperature of 670C
was determined [17]. The thermal expansion coefficient of this glass is 9.8 x 10-6 1/C. A
commercially available silica sol (Nyacol 2040 NH4, Akzo-PQ Silica, Amersfoort, The
Netherlands) was used for the electrophoretic infiltration of the metallic fibre mats. This sol is
II - 595
stable at pH 9, has a solid content of 40vol% and the sol particles are spherical with an
average particle diameter of 20 nm. For the EPD experiments the sol was used as-received.
Infiltration of Fibre Mats by Electrophoretic Deposition (EPD)
The EPD process relies on the presence of small charged particles within a liquid suspension,
which, on the application of an electric field, will migrate toward and deposit on an electrode
[18]. Movement of the particle occurs because the particle surface is charged with respect to
the suspending medium. If the deposition electrode is replaced with a conducting fibre
preform, the suspended particles are attracted to the fibre preform and then deposited on and
within it. The movement of the ceramic particles in an aqueous suspension under an external
applied electric field will be governed mainly by factors such as the field strength, the pH of
the solution and the ionic strength of the solution [18]. In the present experiments, both the
glass powder particles and the silica particles were negatively charged, as they migrated to the
positive electrode.
Thus, the fibre mats were placed as the anode. A stainless steel plate served as the negative
electrode. For the DURAN glass powder suspension a voltage of 30 V was applied to the
electrodes that were 30mm separated. For the silica sol a much lower voltage (4 V) for the
same electrodes separation was found to be adequate to obtain a reasonable deposition rate, in
agreement with previous studies using SiC fibre mats [9,11]. The deposition time was varied
to find the optimum time for the complete infiltration of the sol into the intra-tow regions, and
for obtaining the desired thickness of the deposited layer in the case of the glass powder.
Controlled drying of the infiltrated/coated fibre mats is critical to the production of prepregs
of sufficient quality, minimising the generation of cracks. The powder coated fibre mats were
dried in normal atmosphere as no significant microcracking was expected due to the large
powder particle sizes. The sol infiltrated fibre mats were dried both in normal air and in a
humid atmosphere to investigate the effect ot drying conditions on microcracking
development. The dried fibre mats were impregnated with a vacuum resin and polished to a 1
m finish for scanning electron microscopy (SEM) examination. For the silica sol infiltrated
mats, SEM energy dispersive X-ray analysis (EDX) was employed to differentiate between
resin and silica in the intra-tow regions.
Forming of Green Bodies, Composite Consolidation, and Mechanical Tests
Woven carbon fibre reinforced composites
A hot-pressing densification stage was chosen for the consolidation of these composites, as it
was found (see Results below) that the EPD process did not yield complete infiltration of the
woven carbon fibre mats. Typically, ten coated fibre mats were stapled, introduced in a
graphite die and hot-pressed in vacuum at 900C under an applied pressure of 10 MPa. The
holding time was 30 minutes. In this way cylindrical samples (50 mm diameter, 5.4 mm
height) were obtained. From these samples test bars (5 x 4 x 40 mm) were prepared by cutting
and polishing. These bars were tested in three-point bending (30 mm span). The loaddisplacement diagramm was recorded and SEM was used to observe fracture surfaces.
II - 596
Woven metal fibre reinforced composites

The silica sol-infiltrated metallic fibre mats were used to prepare a glass matrix composite by
a simple uniaxial cold-pressing and sintering route. Typically, two fibre mats were
sandwiched between layers of the soda-lime glass matrix powder in a die (24 mm diameter) to
form the green bodies. Similar laminated composites, but containing woven fabrics of Si-TiC-O fibres sandwiched between thin mullite layers have been reported in the literature [19].
Uniaxial pressures of 90 MPa were employed to obtain samples of sufficient strength to be
handled. The specimens were heated at a rate of 10C/min up to the sintering temperature of
670C and the dwell time at temperature was 1 hour before a subsequent furnace cooling
stage.
The samples were fabricated solely to investigate the possibility of composite consolidation
by this simple, pressureless densification route, and to assess the presence of any reaction at
the silica/metal fibre and silica/ glass matrix interfaces. Selected sintered samples were cut,
polished and observed using SEM. In addition, several rectangular bars (20mm x 2mm x
3mm) were cut from the cylinders and used to investigate, in a qualitative manner, the fracture
behaviour of the composites in three-point flexion (span: 16 mm). Fracture surfaces were
observed using SEM.

Woven Carbon Fibre Reinforced Composites
Figure 2 is a SEM micrograph showing the microstructure of a woven carbon fibre reinforced
DURAN glass matrix composite after hot-pressing at 900 C. A poor infiltration of glass
matrix in the spaces between the fibres is observed. This means that the combination of EPD
and hot-pressing, as employed in this study, was unsuitable for obtaining pore-free materials.
Indeed, using a suspension of m-sized glass particles in the EPD cell was the reason why the
fibre mats could not be infiltrated fully, and the glass particles were only (firmly) adhered to
the surfaces of the fibre mat. Moreover, the viscosity of the glass at the hot-pressing
temperature (900 C) was not sufficiently low to allow an efficient infiltration. This effect
may be enhanced by a poor wettability of the carbon fibres by the viscous glass. Figure 3
shows the load-displacement diagramme of a sample during the three-point bending test.
Composite behaviour, i.e. non-catastrophic fracture, is observed. However, due to the
considerable porosity present, the absolute values of the mechanical properties achieved are
rather low. The optimization of the hot-pressing parameters should result in a reduction of the
intra- and inter-tow porosity in these samples and the atainment of better mechnical
properties. These properties should be comparable to those measured in commercially
availble unidirectional carbon fibre reinforced DURAN glass matrix composites (ultimate
fracture strength = 800-1200 MPa, Youngs modulus = 120 GPa [3]), with the advantage,
however, of being isotropic in 2 dimensions due to the particular fibre architecture used. It is
suggested, however, that an increase of the hot-pressing pressure will not necessarily lead to
better results in terms of glass infiltration, since this would tend to close the intra-and intertow spaces, making more difficult the flow and intra-tow penetration of the glass. Probably a
pre-treatment of the carbon fibres surfaces to improve their wettability by the viscous glass
would be an effective approach.
II - 597
Fig. 2: SEM micrograph showing the microstructure of a carbon fibre mat reinforced
DURAN glass matrix composite. Poor infiltration of the matrix in the spaces between the
fibres is observed
II - 598
II - 599
II - 600
II - 601
II - 602
THERMOELECTRIC PROPERTIES OF B4C-BASED

COMPOSITE CERAMICS PREPARED BY ARCMELTING
Takashi Goto, Jianhui Li and Toshio Hirai
Institute for Materials Research, Tohoku University
2-1-1 Katahira Aoba Sendai 980-77, Japan
KEYWORDS: B-C system solid solution, B4C-SiC composite, B4C-TiB2 composite, boronrich boride, eutectic reaction, arc-melting, thermoelectric property
II - 603
II - 604
II - 605
II - 606
II - 607
II - 608
II - 609
II - 610
II - 611
II - 612
FORMATION AND CHARACTERISATION OF FIBREMATRIX INTERFACES IN CONTINUOUS FIBRE

REINFORCED SILICON CARBIDE
Pamela M. Farries and Jean-Bernard Veyret
European Commission, Joint Research Centre, Institute for Advanced Materials,
PO Box 2, 1755 ZG Petten, The Netherlands.
SUMMARY: The effect of processing conditions on interface formation in a fibre reinforced

silicon carbide composite was investigated through infiltration of a uni-directional lay-up of
pitch-based carbon fibres with an aqueous ceramic powder suspension and subsequent
consolidation at high temperature. Interface formation by a conventional method - the
attempted preservation of a pyrolytic carbon coating applied to the fibres by CVD - is
contrasted to an in-situ method, involving the exploitation of the migration of yttria (included
with the SiC powder along with alumina as sintering aids) to the fibre-matrix interface by hotpressing at temperatures of 1750 and 1800C in an ambient argon atmosphere and sintering
without mechanical pressure at 1850C in an argon atmosphere. Basic indications of
mechanical properties determined from 4-point bending tests at room temperature are
included. Early indications are that fibre degradation is inhibited in hot-pressing producing an
interface which can debond.
KEYWORDS: silicon carbide, interfaces, fibres, liquid phase sintering, hot-pressing, YAG
(yttrium aluminium garnet)
INTRODUCTION
Commissioning of long fibre reinforced ceramics into industrial service is becoming an
achievable goal as properties are improved to meet target requirements, but this success
simultaneously shifts emphasis for improvement towards the cost of producing a composite of
consistent high quality. Consolidation methods without applied mechanical pressure are being
investigated, however, process development is still constrained by the need to create or
preserve a pre-designed fibre-matrix interface architecture, often generated by complex and
expensive fibre coatings, during the very high temperature consolidation cycle. These two
constraints are inextricably linked and serious process developments can be made only by
considering the implications to interface structure and chemistry throughout production and
service.
Carbon and boron nitride are the most common choices of material which satisfy the required
criteria for debonding and controlled sliding at the fibre/matrix interface of silicon carbide
based CMCs, so imparting a degree of toughness to a material comprising two brittle
components. Both C and BN are susceptible to oxidation and in systems using SiC fibres, the
gap left by the removal of the interphase is filled by a silicate layer produced from the
oxidation of the fibre which does not satisfy the debonding and sliding requirements [1]. This
is unfortunate when the layer was produced in-situ during composite production, such as in
II - 613
the case of Nicalon reinforced lithium aluminium silicate (LAS) glass, but is a much more
serious concern when the interface was deliberately deposited. The move towards oxide-oxide
composite systems has prompted investigation of porous oxide and refractory metal coatings
[2]. Sputter deposited yttria coatings have been utilised in metal matrix composites to protect
SiC monofilaments from reaction with an Mg-Li alloy[3]. Combining the metal matrix and
oxide-oxide composite experience with the observations that alumina and yttria sintering aids
in silicon carbide have a propensity to migrate to free surfaces at high temperatures [4-7] and
alumina is observed to evaporate at high processing temperatures [4,6] presents the possibility
of a novel mechanism for the production of a compatible oxide interface during the
processing and use of a SiC/SiC composite which may be investigated in tandem with
existing methods [8].
The conditions necessary for production of a silicon carbide matrix composite, reinforced
with pitch-based carbon fibre by infiltration of a uni-directional lay-up with an aqueous
ceramic powder suspension containing silicon carbide and the liquid phase forming additives
alumina and yttria, and consolidation by hot-pressing are investigated, with the eventual aim
of including Hi-Nicalon silicon carbide fibre. The interfaces produced when composite
samples are densified with and without the application of mechanical pressure are compared
through EDX analysis and a first indication of mechanical properties obtained from four point
bending. Background to the chemical and physical development of the interface during
processing is provided from the results of an extensive study into the effects of varying
alumina and yttria additive concentration, sintering temperature, hold time and sintering
environment on the density, microstructure and chemical composition of free sintered silicon
carbide [8]. The effect of these sintering variables on free sintered tows of un-coated and
pyrolytic carbon coated carbon fibres are also detailed.
Sample Preparation
Monolithic silicon carbide samples were prepared from -SiC powder with a mean particle
diameter of 0.6m - UF-15 from H.C. Starck - by pressure filtration of an aqueous powder
suspension (optimised at 47vol% powder using information obtained with Matec Applied
Sciences, Electrokinetic Sonic Analysis System) dispersed using an ultrasonic probe. Alumina
(Baco RA 207LS Grade) and yttria (H.C. Starck Fine Grade) were added in the 3:5
(alumina:yttria) YAG molar ratio and their combined mass was either 3 or 9wt% of the total
powder, Table 1. Amoco P25 pitch based carbon fibre was dipped as tows into the slurries
used for monolith production and wound as 1D lay ups for composite casting from a slurry of
composition B containing a reduced volume of solids (40%) and a small amount of wetting
agent to allow complete infiltration.
Table 1: Sample compositions and theoretical densities (TD)
Composition
Sintering aid content

(Al2O3+Y2O3) wt%
Molar % Y2O3 in mix
A
B
9
3
62.5
62.5
II - 614
Theoretical Density
g/cm3
(rule of mixtures)
3.30
3.24
Densification
Composite samples were hot-pressed at 1750 and 1800C in an ambient argon atmosphere.
Mechanical pressure of 27MPa was applied incrementally beginning when the temperature
reached 1500C and reaching its maximum at the peak sintering temperature. Thereafter both
temperature and pressure were maintained at maximum for 60 minutes. Heating and cooling
temperature ramps were set at 10/min.
Pressureless sintering was carried out in a graphite resistance furnace at temperatures between
1750 and 1900C under 10bar argon pressure for hold times of up to 120 minutes in a closed
graphite crucible. Sample-crucible contact was prevented by the use of the minimum amount
of coarse SiC powder required, except in the case where and sinter bed containing 1.33wt%
alumina (indicated in Table 2) was used.
Table 2: Pressureless densification conditions
Temperature (C)
1750
1800
1850
1900
A, B,
fibre+B
A, B,
Composite
A, B,
fibre+A,
fibre+B,
coated
fibre+B
A, B,
fibre+A,
fibre+B,
coated
fibre+B
Hold Time (mins)

30
60
120
A, B
A, B,
fibre+A
120
alumina bed
A, B
Analysis
Monolithic samples were subjected to phase analysis by X-ray diffraction (XRD), aluminium
and yttrium content analysis by atomic absorption spectroscopy and density measurement by
mercury immersion. Linear shrinkage and mass loss during sintering were also determined for
these samples. All samples were examined by scanning electron microscopy(SEM) and the
composite hot-pressed at the lower temperature tested in 4-point bending.

Monolithic Silicon Carbide Densification
The peak density achieved when sintering without a sinter bed was 2.97g/cm3 or 90% of the
theoretical density (TD) for samples of composition A sintered at 1900C. The samples
sintered for 120 minutes all showed a markedly improved final density figure when sintered at
temperatures above 1850C, when shrinkage values were correspondingly greater, as too were
mass loss figures. Atomic absorption spectroscopy revealed, however, that up to half of the
II - 615
aluminium originally added was being lost from the samples of composition A sintered at
1900C along with approximately 20% of the yttrium [8], this observation being backed up by
the detection of an yttria rich phase in XRD (2Y2O3:Al2O3 as opposed to the expected YAG
phase 3Y2O3:5Al2O3) indicating that a shift in the composition of the liquid phase had
occurred. The holding time at peak temperature was reduced in an attempt to minimise
alumina volatilisation and mass loss, Fig. 1.
2.7
3
Density (Mg/m)
2.6
2.5
2.4
Composition A
Composition B
2.3
2.2
0
30
60
90
120
Hold Time (minutes)
Figure 1: Density variation of monolithic SiC with hold time at 1850C

Trends projected from sintering carried out at 1850C for 30, 60 and 120 minutes, suggested
that the maximum density for composition A would be achieved at this temperature from a
high temperature hold of 45 to 60 minutes in length, while for composition B the optimum
soak time is around 90 minutes. Prolonged holds are damaging for both compositions. The
increased density drop for the samples containing larger amounts of alumina (and hence
having the capacity to lose a larger proportion of their mass) held for 120 minutes at high
temperature proves the influence of time on alumina volatilisation. Mass losses are reduced
significantly with a sintering hold time of 30 minutes but densification is no longer
satisfactory.
Addition of 1.33wt% alumina to a sinter bed of coarse silicon carbide resulted in the highest
density figure achieved - 92.4% TD for a sample of composition A for a two hour hold at
1850C. Calculating the density increase as a percentage of the original green density of the
sample provides a convenient method for monitoring densification efficiency under different
conditions, while neutralising the effect of any variation in green density, Fig. 2.
II - 616
Density Increase (%)
50
40
30
20
10
0
0.5hrs
no
bed
1hr no
bed
3w t% additives
2hrs
no
bed
2 hrs
Al in
bed
9w t% additives
Figure 2: Effect of alumina on densification efficiency at a sintering temperature of 1850C
As also demonstrated in Fig. 1, a hold time of 60 minutes at temperature produces a greater

density increase than a 120 minute hold, Fig. 2, the effect being especially noticeable for
samples containing 9wt% which under these circumstances achieve a smaller density increase
than samples containing 3wt% additives, again demonstrating the volatility of the sintering
additions. This order is reversed when a sinter bed containing alumina surrounds the samples,
presumably preventing the evaporation of alumina and emission of carbon dioxide produced
on its reaction with SiC [5], and allowing the full benefit of an increased volume of liquid
phase to be realised.
Microstructural Observations
Monolithic Silicon Carbide
Silicon carbide was seen to densify leaving only isolated porosity at temperatures of 1850C
and above where the hold time was at least 60 minutes. Samples of composition B sintered for
a shorter time or at lower temperature were dense only in a region extending 0.5mm from the
surface (overall dimensions, 15 x 15 x 5 mm) seeming to indicate densification occurring
from outside inward. When viewed in back scattered mode, the 1850 and 1900C samples
sintered for 120 minutes show an unusual distribution of light contrast areas, Fig 3(a), shown
to be rich in yttrium through Energy Dispersive X-ray analysis (EDX), the contrast effect
being associated with yttriums high atomic number. The central zone at the top of the
micrograph contains a few widely spaced areas of the yttrium rich material up to 100m in
length. The outer zone contains a more even dispersion of small areas rich in yttrium. The
effect was not seen for any composition A samples sintered at temperatures lower than
1850C, for a hold time shorter than 120 minutes or for any of the composition B samples.
Combination of these observations suggest that densification occurs from the surface to the
centre and when the hold time is long enough and temperature high enough, the excess of a
high yttrium content second phase begins to segregate out, possibly to flaws and grain
boundaries.
II - 617
(a)
(b)
Figure 3: (a) SiC monolith containing 9wt% alumina and yttria in the YAG ratio sintered for
120 minutes at 1850C (b) Carbon coated fibre and SiC+3wt% YAG sintered for 120 minutes
at 1850C
Examination of samples sintered in an alumina containing bed at 1850C showed
microstructures similar to those sintered without bedding powder at 1950C but without the
light contrast areas of yttria rich material.
Fibre Tows
Of the fibre tows sintered alongside the monolithic samples, the carbon coated fibre showed
the most interesting behaviour. Tows dipped in slurry of composition B and sintered for 120
minutes at 1850C did not appear to be badly damaged and the area previously occupied by
the pyrolytic carbon coating is now filled with an yttrium rich phase, Fig. 3(b). Uncoated
fibres sintered under the same conditions were slightly degraded by the matrix Carbon coated
fibre dipped in slurry containing 9wt% additives showed yttrium ingress well into the body of
the fibre [8]. Fig. 3(b) appears to illustrate that at lower temperatures and with less oxide
additives the pyrolytic carbon layer is consumed sacrificially and the an yttrium rich
interphase subsequently forms.
Composite Properties
Composite green bodies were manufactured with a fibre volume fraction designed to rise to
35% on densification of the matrix. Initial hot-pressing at 1750C, judged to be the lower
temperature limit for feasible densification of the matrix and coinciding with the lowest
temperature applied in the pressureless sintering study, produced a composite with only 80%
TD and a correspondingly low volume fraction of fibres - 27vol%. This was reflected in the
mechanical properties of the material, the failure stress reaching a maximum of 200MPa in 4point bending. Sonic resonance measurements however indicated a Youngs modulus of
around 170GPa, not significantly lower than the value predicted by the rule of mixtures. Hotpressing at 1800C produced a composite of 98% TD which has yet to be analysed.
Examination of the fracture surfaces of the bend tested material hot-pressed at 1750C and
cross-sections of this and a similarly prepared green body sintered with out pressure
II - 618
application at 1850C in a 10 bar argon atmosphere, give reason for optimism regarding
interface formation, even with this uncoated fibre.
Figure 4: Fibre pull-out on 4-point bend specimen - un-coated P25 fibre, hot-pressed at
1750C, 3wt% sintering additives
Clean fibre surfaces pulled up to 150m out of the fibre surface are shown in figure 4
indicating surprisingly little fibre damage. Cross sections prepared to allow analysis of the
fibre matrix interface confirm the assertion.
(a)
(b)
Figure 5: (a)Longitudinal section of material hot-pressed at 1750C (b) Cross section of
composite free sintered at 1850C, both reinforced with un-coated fibre
EDX spot analysis on the cross section shown in Fig 5(a) showed no significant trace of
silicon in the fibre centre, indicating that the fibres are being protected from reaction with the
oxide additives by the high pressure exerted during hot-pressing. No significant
concentrations of alumina or yttria were observed in the sample. The matrix of the free
sintered sample (Fig. 5(b)) is much more porous than that of the hot-pressed sample, but the
fibres still retain complete integrity of form even at the very high processing temperature and
without the benefit of applied mechanical pressure to inhibit reactions. A stronger silicon
signal is detected in these fibres but the presence of a significant carbon peak indicates that
the phase is far from SiC in composition. When carbon and silicon are present in equal
quantities the strong silicon signal swamps the peak due to carbon. The darker amorphous
annulus visible around these fibres is found to be rich in yttrium and to contain some
aluminium.
II - 619
CONCLUSIONS
Pressureless liquid phase sintering of silicon carbide with additions of Al2O3 and Y2O3 was
carried out successfully at temperatures at and above 1850C. Placing samples in an alumina
containing silicon carbide sinter bed further improved densification and inhibited the loss of
alumina from the sample and migration of the liquid phase.
The application of mechanical pressure did not lower the temperature required for SiC
densification as far as expected, but processing at 1750C did produce a composite with a
fibre matrix interface capable of debonding.
Yttria rich material is seen at the fibre matrix interface in composites densified by
pressureless means, but this migration appears to be hindered by the application of
mechanical pressure in hot-pressing.
ACKNOWLEDGEMENTS
This work was carried out under the Fourth Framework Programme of the European
Commission and the principal author is grateful to The European Commission for the
provision of a fellowship under the Human Capital and Mobility Programme.
REFERENCES
1. Thouless, M.D., Sbaizero, O. and Evans, A.G., Effect of Interface Mechanical Properties
on Pullout in a SiC-Fiber-Reinforced Lithium Aluminium Silicate Glass-Ceramic,
Journal of the American Ceramic Society, Vol. 72, No. 4, 1989, pp.525-32.
2. Davis, J.B., Lfvander, J.P.A. and Evans, A.G., Fiber Coating Concepts for BrittleMatrix Composites, Journal of the American Ceramic Society, Vol. 76, No. 5, 1995,
pp.1249-57.
3. Kieschke, R.R., Warwick, C.M. and Clyne, T.W., Sputter Deposited Barrier Coatings on
SiC Monfilaments for use in Reactive Metal Matrices - III. Microstructural Stability in
Composites Based on Magnesium and Titanium, Acta. metallurgica et materiala, Vol.
39, No. 4, 1991, pp.445-452.
4. Kerber, A. and Velken, S.V., Liquid Phase Sintered Silicon Carbide: Production
Properties and Possible Applications, Proceedings of the Fourth European Ceramic
Society Conference, Riccione, Italy, October 2-6, 1995, Vol. 2: Developments in
Processing of Advanced Ceramics - II, Galassi, C., Ed, pp.177-184.
5. Mulla, M.A. and Krstic V.D., Low-Temperature Pressureless Sintering of -Silicon
Carbide with Aluminium Oxide and Yttrium Oxide Additions, Ceramic Bulletin, Vol.
70, No. 3, 1991, pp.439-442.
6. Kahlman, L., Rundgren, K., Lidn, E., Nyberg, B. and Calstrm. E., Processing of
Liquid Phase Sintered SiC - Mechanical and Wear Properties, Proceedings of the Third
European Ceramic Society Conference, Madrid, Spain, September 12-17, 1993, Vol. 3:
Engineering Ceramics, Duran, P. and Fernandez, J.F., Eds., pp.477-482.
7. Kostic, E., Sintering of Silicon Carbide in the Presence of Oxide Additives Powder
Metallurgy International, Vol. 20, No. 6, 1988, pp.28-29.
8. Farries, P.M., Veyret, J-B. and Rawlings, R.D., Fibre Reinforced Silicon Carbide: Effect
of Densification Method and Parameters, Key Engineering Materials, Vols. 127-131,
1997, pp.295-302.
II - 620
STRUCTURAL ANALYSIS OF CERAMIC-MATRIX

COMPOSITES
Alain Burr1, Francois Hild1, Frederick A. Leckie2
1
Laboratoire de Mcanique et Technologie, E.N.S. de Cachan, C.N.R.S. , Universit Paris 6

2
Department of Mechanical and Environmental Engineering
College of Engineering, University of California, Santa Barbara, CA 93106
SUMMARY: A constitutive law is proposed for CeramicMatrix Composites which models

matrixcracking, sliding, fiberbreakage, and fiber pullout. These different mechanisms
induce loss of stiffness, inelastic strains, hysteresis loops, and crack closure. The features are
analyzed within the framework of Continuum Damage Mechanics by the introduction of
physical internal variables identified previously in material science investigations. The
procedure is applied to a SiC/SiC [0/90] laminate. The desription is applied to some structural
problems including a plate in tension penetrated by a hole and subjected to tension and the
Iosipesco specimen subjected to shear. It is observed that the SiC/SiC has the ability to
redistribute stress so that the presence of stress concentrations does not compromise the
performance of the component.
KEYWORDS: ceramic-matrix composite, matrix cracking, stress redistribution
INTRODUCTION
Composites consisting of a ceramic matrix reinforced by continuous ceramic fibers (CMC)
are candidates for application in components which operate at temperatures in excess of those
which are normal for metallic structures. In spite of the fact that the constituents of the CMC
are both brittle it has been demonstrated by Aveston et al. [1971] that following
matrixcracking, sliding occurs at the fibermatrix interface which causes inelastic
deformations. The presence of matrix cracks and inelastic deformations may impart to the
material the ability to redistribute stresses. The ability to redistribute stress is an important
property since design studies indicate that working stresses for CMC components are
sufficiently high for matrixcracking to be unavoidable in regions of stress concentration
occurring at the junctions and penetrations which are a feature of engineering components.
The intention then of the present study is to develop a continuum description of the damage
processes which is mechanismbased and which may be used to describe the behavior of
CeramicMatrix Composites (CMCs) under the conditions of multiaxial stress occurring in
practice. Since crack spacing at saturation is small [Beyerley et al., 1992] in most CMCs,
Continuum Damage Mechanics is an appropriate means of describing degradation since
changes in elastic moduli measured on a macroscopic level provide a simpler and more robust
means of measuring damage than does microscopic measurement of crack density, which
requires the average of many readings before reliable values are established [Jansson &
Leckie, 1993].
II - 621
By combining Continuum Damage Mechanics (CDM) [Lemaitre, 1992] with

micromechanical studies which are mechanism -based constitutive equations are developed
which lend themselves to the finite element procedures commonly used in practice The CDM
formulation applied to reinforced composites is written within the framework of the
Thermodynamics of Irreversible Processes [Coleman & Gurtin, 1967; Rice, 1971; Germain et
al., 1983] . The first step in establishing such a model is to identify the internal variables
which define the state of the material . The second is to determine the expression of the state
potential in terms of the state variables and the third one to define the evolution laws of the
internal variables.
The model is integrated into a finite element system (ABAQUS)and is used to estimate the
bebavior of a beam subjected to four point bending and a plate subjected to tensile stress
which has been penetrated by a hole. The calculations have been performed using data for a
SiC/ SiC.
THE TENSILE STRESSSTRAIN RELATIONSHIP FOR UNIDIRECTIONAL CMCS

The results of loading / partial unloading tensile tests on SiC/SiC composites have the form
shown in Fig 1. Compared to the initial elastic modulus the unloading modulus is
substantially less. Plastic straining is also observed but the magnitude is usually quite modest.
Postmortem analyses of broken specimens indicate the presence of arrays of microcracks in
the matrix which are accompanied by debonding and friction at the fibermatrix
interface.Matrixcracking is responsible for the decrease of stiffness observed in experiments
on CMCs and release of residual stresses due to processing, and sliding at the fiber-matrix
interface is the source of irreversible deformation.
1
2R
2
D)
E(1
(1)
D)
E(1
in
Fiber
Matrix
1
M
Fig 1 Typical Stress/Strain
2L
Interface
(2)
E = E(1D)
2l F
Fig 2 Unit Cell
The cell model illustrated in Fig. 2 was first proposed by Aveston and Kelly (1971). In the
unit cell shown in Fig. 2 the elastic moduli of the fiber and matrix are Ef and Em, respectively,
the volume fraction of the fiber is f and Rf is the fiber radius. The elastic modulus of the
II - 622
undamaged composite is E = f Ef + (1f) Em and average distance between cracks is denoted

by 2L. The debond length at the fiber/matrix interface is 2ld, and the interface is assumed to
have a constant shear strength whose value is unknown at this stage. Residual stresses are
introduced during processing so that the residual stresses in the matrix is an additional
unknown.
If the initial behavior of the elementary cell is isotropic and elastic, and Youngs modulus is
E, it can be shown that the stiffness loss depends on the crack density defined as a2 / 4LW
where W is the width of the element By assuming plane stress conditions, and that the crack
~
interactions can be neglected, a first approximation for the reduced elastic modulus E can be
written as
~
E
E =
1
a2
1 + 2 4LW
This relationship can be recast in the framework of CDM as

~
E
=1D
E
where
(1)
D=
a2
2 4LW
a2
1 + 2 4LW
is the damage variable associated with the crack density. When D is small, a first order
solution to En. (1) is given by
a2
D 2 4LW
so that the damage variable is proportional to the crack density.
The internal elastic energy density in the unit cell [Hild et al., 1996] can be found by uses the
two step approach introduced by Volterra [1907] .The first step is to calculate the contribution
of debonding and sliding at the interface using the 'cut and paste' approach in which the
unbroken part (2) is moved with respect to the broken part (1) by an amount s
over a
gives rise to a
length lF (Fig3) . Because of interfacial sliding, this displacement s
selfbalanced linear stress field along a length lF in parts (1) and (2) when the interfacial
behavior is assumed to be characterized by a constant shear strength. By integration over lF
and then averaging over the total length L, the elastic energy density associated with this
process is given by
II - 623
= s
The crack opening displacement s
expressed as
2
2 fE1(1f)E2 s lF

3
E

lF L
due to slip induces an irreversible or inelastic strain
fE1 s
= E L
The second step consists of an elastic loading of the damaged system so that the elastic energy
density is given by
(2)
1
=2
E (1D) ( )
where the current elastic modulus is E (1D).

The total elastic energy density is then the sum of the two elements of the energy densities so
that
(3)
II - 624
= 2 E (1D) ( )
2
2 fE1(1f)E2 s lF

+3
E

lF L
The associated forces are used to define the relevant forces driving each mechanism. Matrix
cracking is assumed to be driven by Y, which plays an identical role as the energy release rate
G in the framework of Linear Elastic Fracture Mechanics. From a micromechanical analysis
[Hild et al., 1996] , it can be shown that the backstress is dependent on the applied stress .
therefore the driving force of the inelastic strains can be taken as the stresses acting in the
same direction. The same assumption can be made when the evolution of the damage variable
d related to sliding is analyzed, i.e. the driving force of d can be chosen to be its associated
force y, or the applied stress .

In the present approach the growth laws of the internal variables (D,d,) are established from
macroscopic quantities measured in the course of unloading and reloading sequences. To this
end use is made of the solution of the response of the unit cell (Fig. 2) when subjected to an
unloading/reloading sequence during which the magnitude of the shear stress remains
constant.
The expressions obtained from the analysis for the residual stress 1 and the internal variables
D,d, in terms of the macroscopic quantities shown in Fig. 1 are given respectively by [Hild
et al., 1996]
(5.1)
1
E
in + 2
(5.2)
D=
(5.3)
d
4
(5.4)
1 ( M in 2 )
4
_
_
D D 2
M in
2
M in

( in + 2 )
2
M in
d M1(1D)
= 2
E(1D)
where 1E1/ E is the residual stress in the broken layer (1). Equations (5) are only valid when
a constant shear strength characterizes the interfacial behavior, and lastly Eqn. (5.4) is only
valid for monotonic loading conditions.
By performing a series of unloading/reloading sequences the internal variables can be
determined from experiment using Eqns (5). The residual stress 1 is calculated from Eqn.
(5.1) and it is a test of the effectiveness of the model that the same value of the residual stress
is obtained for each loading sequence. The values of D and d are given by applying Eqns (5.2)
and (5.3) respectively. The information is now available to complete the calculation for Eqn.
(5.4). The corresponding associated forces are obtained by the expressions given in Eqns. (4).
The relationship between the internal variables and the associated forces can then be
II - 625
investigated by knowing the driving forces of each state variable. It is this method which is
proposed to model the behavior of CMC laminates.
CMCS WITH MULTIDIRECTIONAL FIBER SYSTEMS

The onedimensional investigation can be extended to a [0/90] laminate composite and to a
[0/90] woven composite subjected to multiaxial plane stress states by establishing first the
properties of each layer and then using laminate analysis.
The components of each layer consist of the matrix, the fiber and the interface, with f being
the fiber volume fraction. The fiber direction defines the 12 axes. The axes xy correspond
to the principal axes of the strains in the ceramic matrix. The definition of the axes used at the
constituent, layer and composite levels are shown in Fig. 4. For the case of a 0/90 symmetric
layup Burr et all (1996) have established that the free energy has the expression
(6)
1
00
= 2 ( ) : E(D00 ,D90 ,Df1 ,D90 )
_ _
_ my my
f1
_
: ( ) + S(,d)
_ _
__
2
1
Fig 4 Orientation of axes

where D00 ,D90 are the matrix damage parameters and Df = D00 ; D90 is the set of
my
my
f1
f1
damage variables modeling fiberbreakage in the 0 and 90 plies respectively.The interface

slip parameters are d = d11 ; d22 ; d12
Finite Element Analysis

The resulting constitutive equations have been introduced into a UMAT routine of the finite
element code ABAQUS (Hibbitt et al 1993). Early calculations of beams in four- point
bending, proved to be simple to run, and indicated the ability of the CMC to redistribute stress
and agreed with experimental observation that elastic calculations are very conservative.
Components which were more challenging were the Iosipescu specimen (Fig 5) and a plate,
penetrated by a hole, subjected to a tensile stress parallel to the fiber directions.
When the availability of materials is restricted to planar form shear data shear data data is to
subject the Iosipescu to shear loadspecimen [Iosipescu, 1967]. The shear properties of the
material are obtained by plotting the average stress at the minimum section against the shear
strain measured by strain gauges placed at the center of the specimen.
II - 626
2b
[0/90]
Strain gauges
Fig 5: The Iosipescu specimen
It is known that the shear stress at the minimum section of this specimen is sensibly constant
when the material is elastic and isotropic, but it is not known if the constant shear stress
assumption is valid when cracking occurs. In addition to verifying the suitability of the
Iosipescu specimen for obtaining the properties of materials which crack as illustrated in Fig,
the tests provide an opportunity to measure the ability of the constitutive equations to predict
the behavior of a component in which the stress state is different from those used in the
identification procedure.
Shear Stress (MPa)
250
200
150
100
50
0
-1
Non-linear computation
Average value
Linear computation
-0.5
0
0.5
Normalized Coordinate, y/b
Fig 6: Stress distribution at minimum section

The prediction of the average stress against strain at the center of the specimen is agrees with
the experimental observations to within 5%. The results of the finite element analysis shown
in Fig. 6 indicate that the shear stress at the minimum section is essentially constant therefore
justifying the use of the Iosipescu specimen as a means of obtaining shear data. Experiments
on plates in tension which are penetrated by holes suggest that strength is dictated by the nett
cross section and that the effect of the initial elastic stress concentration is minimised by the
stress redistribution due to matrix cracking. The calculations indicate because of stress
redistribution that almost half of the ligament stress carries the same stress. The elastic stress
concentration factor is 4.1 which following the effects of stress redistribution reduces to 2.4.
The large effect of matrix cracking is evident.
II - 627
CONCLUSIONS
A CDM model is proposed for CMCs which is mechanismsbased. The laws which relate the
growth of the internal state variables to their associated forces have been derived from the
unloadingreloading paths during tensile experiments for two different directions. The ability
of the model to predict the response to another state of stress suggests the advantage of a
mechanismbased approach.
When applied to SiC/SiC [0/90] layups, the present model has 10 internal variables, viz.
three inelastic strains modeling sliding, three damage variables describing the amount of
debonding and four damage variables accounting for matrixcracking and fiberbreakage in
the two plies. It is shown that only two different experiments in tension are needed to identify
the growth laws of the ten internal variables. The model has the potential to be applied to
other material configurations (e.g., SiC/CAS, SiC/C, C/C, and presumably SiC/Al2O3) and
architectures (e.g., woven configurations). Furthermore, the general framework presented
herein has been applied to room temperature configurations and monotonic loading
conditions. However extensions to cyclic load histories as well as high temperature
applications can be included with minimal change to the state potential formulations.
Evolution laws will have to be modified slightly.
The reliability of the Iosipescu test is confirmed as a means of average stressstrain shear
data, and the constitutive equations are able to predict the shear properties correctly.
However, it is shown that the average shear properties may be slightly different from the
actual stressstrain shear data in the center of the ligament. Therefore the identification of the
shear properties based upon the measurements on an Iosipescu test are, strictly speaking, only
an approximation of the actual response in pure shear.
The ability of stress redistribution due to the nonlinearity of the stress/strain curve has been
shown in the case of the Iosipescu experiment. Stress redistribution important for structural
applications has been demonstrated for beam bending and tensile plate penetrated by holes.
REFERENCES
1.
Aveston, J., Cooper, G. A., Kelly, A., 1971, Single and Multiple Fracture, National
Physical Laboratory: Properties of Fiber Composites, IPC Science and Technology Press,
Surrey (UK), 15-26.
2.
Beyerley, D., Spearing, S. M., Zok, F. W., Evans, A. G., 1992, Damage, Degradation and
Failure in a Unidirectional Ceramic-Matrix Composite, J. Am. Ceram. Soc., 75, n 10,
2719-2725.
3.
Burr, A., Hild, F., Leckie, F. A., 1995, Micro-Mechanics and Continuum Damage
Mechanics, Arch. Appl. Mech., 65, n 7, 437-456.
4.
Coleman, D. B., Gurtin, M. E., 1967, Thermodynamics with Internal Variables, J. Chem.
Phys., 47, 597-613.
5.
Germain, P., Nguyen, Q. S., Suquet, P., 1983, Continuum Thermodynamics, J. Appl.
Mech., 50, 1010-1020.
II - 628
6.
Hibbitt, H. D., Karlsson, B. I., Sorensen, P., 1993, Abaqus, version 5.3.
7.
Hild, F., Burr, A., Leckie, F. A., 1996, Matrix Cracking and Debonding in CeramicMatrix Composites, Int. J. Solids Struct., 33, n 8, 1209-1220.
8.
Iosipescu, N., 1967, New Accurate Procedure for Single Shear Testing of Metals, J.
Mater., 2, 537-566.
9.
Jansson, S., Leckie, F. A., 1993, The Mechanics of Failure of Silicon Carbide
FiberReinforced GlassMatrix Composites, Acta Metall. Mat., 40, n 11, 2967-2978.
10. Lemaitre, J., 1992, A Course on Damage Mechanics, Springer-Verlag, Berlin (Germany).
11. Rice, J. R., 1971, Inelastic Constitutive Relations for Solids: An Internal Variable Theory
and its Application to Metal Plasticity, J. Mech. Phys. Solids, 19, 433-455.
12. Volterra, V., 1907, Sur lquilibre des corps lastiques multiplement connexes, Annales
Scientifiques de lEcole Normale Suprieure, Paris, 24, n 3, 401-518.
II - 629
FAILURE MECHANISM OF A MULTI-LAYER

COATING FOR THE OXIDATION PROTECTION
OF C/C COMPOSITES
Laifei Cheng, Litong Zhang, Yongdong Xu
State Key Laboratory of Solidification Processing,
Northwestern Polytechnical University, Xian Shaanxi 710072, P. R. China
SUMMARY: A multi-layer coating for carbon-carbon composites (C/C) with a Si-W outer
layer, a SiC barrier layer and a SiC transition layer was prepared by the combination of
Siliconization, CVD and Liquid Reaction method. The oxidation experiments at from 1600C
to 1700C were conducted. The coated C/C always gained in weight at beginning. When it
started to loss in weight, deep holes could be observed on the coating surface. At this time, the
thickness of the oxide film on the coating surface at different temperature was found to vary
little. Thermochemical analysis shows that the gas pressure at the interface of the coating and
its oxide film could be higher than the ambient pressure, and gas bubbles would be formed in
the oxidation process. The nucleation, growth, bursting and self-sealing processes of the
bubbles were studied.
KEY WORDS: C/C composites, coating, oxide film, failure mechanism
INTRODUCTION
Carbon-carbon composites (C/C) exhibit excellent structural properties above 1650C, and
are considered as the most promising candidate materials for the high thrust-weight ratio
turbine engines. As a key technology for the high-temperature structural applications of C/C,
oxidation protection is paid more and more attention recently. A number of studies have been
conducted on the short-time oxidation protection, including various oxidation inhibitors,
protection coating and sealants. For the long-time coating, not much progress has been made,
in spite of years of intense effort. The study of failure mechanism is very important to the
prolongation of protection life for a high-temperature coating and the development of a longtime protection coating system of C/C aimed at the components of advanced aircraft
engines [1].
At above limiting use temperature, the coating will be destroyed entirely in a short time
because of chemical compatibility of interfaces. At below cracking temperature, failure will
take place at the coating-substrate interface because of oxygen diffusion through the cracking.
At between intrinsic protection range temperature, failure is mainly influenced by the oxygen
diffusion through the coating [2, 3]. Can we think simply the time obtained by dividing the
coating thickness with the diffusion rate constant of oxygen at some temperature as the
protection life at this temperature? In this paper, we will answer this question.
II - 630
The substrate material used in experiments, cut off from aircraft brakes, is a 2D-C/C with a
density of 1.7g/cm3 . The substrates are of cubed shape with a size of 5mm5mm25mm.
The multi-layer coating consists of three layers. The SiC transition layer was prepared by
Siliconazition on the surface of the substrates to decrease the interfacial mismatch and
increase the adhesive strength. The SiC barrier layer was prepared on the transition layer by
CVD to prevent carbon outward diffusion and out layer inward penetration[4]. The Si-W
sealant layer was prepared by Liquid Reaction to provide an effective barrier to inward
diffusion of oxygen. The purity of the Si and W powder used to prepare the transition and
sealant layer is above 99.5%, and that of the MTS used in CVD is above 98%. The Si-W layer
was prepared in a vacuum furnace in which the spreading out of the Si-W layer can be
observed timely[5]. The work temperature and the vacuum of this furnace is respectively
1600C and 0.0133Pa. Oxidation tests of long-time in dry air at 1600C, 1650C and 1700C
were conducted in a furnace heated with MoSi2 rods. The thickness of the oxide film on the
coating surface was identified with EMPA and SEM.

Because the coated C/C firstly gains and then losses in weight in the oxidation, the time the
weight change is equal to zero can be defined as the protection life tC , which can be called the
critical protection life or effective protection time also.
The measured protection life at different temperature is listed in Table 1. It is shown that the
relation curve of the protection life with temperature fits Arrheniuss law
lg t C =
44104
25.311
T
(1)
Table1: Effect of temperature on the oxidation protection life

Temperature (C)
1600
1650
1700
Oxidation protection life (h)
168
30
When the oxidation time is beyond the critical protection life, the coated C/C losses in weight
rapidly. This telles us that catastrophic oxidation has teken place. The catastrophic oxidation
always startes from one or two points which appear to be black at high temperature. Deep
holes will be formed at the black points which become the diffusion channel of oxygen. Fig.1
is the SEM micrograph of a deep hole on the coating surface.
It is not difficult to find that the deep holes are produced by escape of gas bubbles from their
shape. The thermochemical diagram (Fig.2) shows that the gas pressure Pt at the interface of
the coating and its oxide film can be higher than the ambient pressure, and gas bubbles will be
formed in the oxidation process[6]. The gas products at the interface are mainly SiO. Partial
pressure of the oxides of W is negligible. They disssolve in the oxide film, and then diffuse
II - 631
towards the surface, and run away at last (Fig.3). The W dissolved in the liquid Si increases
the interfacial gas pressure greatly.
The high-valence W ion is a glass former. It will get into the molecule net of the silica and
make the oxide film become silica glass in the oxidation process (Fig.4 and Fig.5). From
Raman spectrum of the oxide film, we know that the silica glass is highly stable in more than
one hundred hours (Table 2). Formation of silica glass layer is very important to the long-term
coating. Of course, glass layer will drop its using temperature.
Because the oxide film is thinner and resistes to the nucleation and bursting of the gas bubbles
less at the beginning of oxidation, more bubbles will be formed. The holes leaved on the
oxide film by bursting of gas bubbles is smaller and less shallow. They will be sealed fastly
by the viscous flow of glass. The longer of the oxidation time, the thicker the oxide film and
the larger the resistance, then the more difficult for the bubbles to nucleate. Consequently,
The bubbles formed are fewer, and greater and deeper when they burst. At some
temperature, the viscous flow is limited. When the bubbles are large enough so that they can
not be sealed in time, holes appear on the oxide film. At the same time, new oxide film is
formed at the holes. Obviously, bubbles will prior necleate, grow and burst at the holes in
further oxidation as the new film formed is very thin. As a result, the holes get deeper and
deeper. When they penetrate through the coating at last, failure takes place. The failure
precess is summarizd in Fig.6.
Fig.1: SEM micrograph of a deep hole on the coating surface
II - 632
II - 633
Fig.4: X-ray diffraction pattern of the coating after oxidation
Fig.5: SEM micrograph of the coating after oxidation

1. C/C; 2. Transition layer; 3. Barrier layer; 4. Si-W layer; 5.Silica layer
Table 2: The effect of oxidation time on the network structure of the silica glass formed
on the coating surface after oxidation at 1600C
Oxidation time Composition of the molecule network structure /% Broken bonds
/h
Dimer
Sheet
Network
Pure network
/%
Before oxidation
22
0
34
44
18.5
45
21
0
69
10
19.9
168
16
11
30
43
16.6
II - 634
Fig.6: The failure process of the coating in oxidation
If the analysis about the faillure process is correct, there certainly is a critical oxide thickness
X(tC), correspending to the critical life. That is to say, the oxide thickness when the coating
fails depends on the diffusion rate constant and has nothing to do with temperature. Relation
of the critical thickness with the critical life fits a parbolic law [7,8].
Fig 7: Distribution of oxygen ion on the coating after oxidation for 168h at 1600(C
X (t C ) = Bt C
(2)
II - 635
For this coating, the critical thickness of the oxide film is founded to be always 40 m less or
more at different oxidation temperature (Fig.7). On the basis of critical thickness, we can
obtain a relation of the critical life with temperature as following
lg t C =
44104
25.340
T
(3)
This relation calculated is identical with the that put forward derrectly from the oxidation
tests. This demonstrates clearly the failure mechanism presented.
It is meant by X(t)C that a coating would fail because some defects are formed when the
thickness of the film reaches its critical value. X(t)C is a constant which is only influenced by
the coating materials, coating structure, coating method and coating process, but not related to
the oxidation temperature. The properties of the coating materials determine the
characteristics of the oxide film and the interfacial gas pressure. The critical thickness of the
film increases with increasing the interfacial gas pressure, decreasing the diffusion rate of
oxygen and increasing the diffusion rate of gas species. In the presence of a barrier layer in
the coating, the diffusion of oxygen along the deep holes would be blocked, and the critical
thickness of the film would be increased. Deep holes could be considered to be formed at the
sites where there are some original defects. Obviously, the defects are conditioned by the
non-homogeneity of the coating preparation. Consequently, improving homogeneity of the
coating is a very effective way to increase the critical thickness and prolong the protection
life. After the coating materials are selected, the critical thickness depends upon the quality of
the multi-layer coating. Only proper methods are selected or developed to prepare every layer
of the multi-layer coating, the quality could be ensured. Although the interfacial gas pressure
varies with temperature, the limited step of the interfacial gas reaction is the diffusion rate but
not the reaction rate. Clearly, it is the diffusion rate constant but not the critical thickness that
is influenced by temperature. Generally, the viscosity of the oxide film drops when
temperature rises. The viscosity drop, on the one hand, makes the holes easy to be sealed and
is favorable to increasing the critical thickness. On the other hand, it promotes the nucleation
and growth of bubbles, and unfavorable to increasing the critical thickness. It can be argued
that viscosity has not influence on the critical thickness in a certain range of temperature.
CONCLUSIONS
1.
The gases produced by the interfacial reaction between the outer layer and its oxide film
are responsible for the coating failure. When the gas pressure is higher than the ambient,
bubbles are formed at the interface.
2.
The oxide film is a silica glass. When the bubbles are large enough that holes caused by
their bursting can not be sealed in time, non-homogeneous nucleation, growth and
bursting of the gas bubbles take place at the holes in further oxidation. As a result, deep
holes are formed, which will become the channel of oxygen diffusion.
3.
It is the critical thickness of the oxide film but not the coating thickness that affects the
oxidation protection life. The relation of the protection life obtained from experiments is
highly identical with that calculated based on the critical thickness.
II - 636
ACKNOWLEDGMENTS
The author gratefully acknowledges the support of K. C. Wong Education Foundation, Hong
Kong.
REFERENCES
1.
Deborah, D. L., Carbon Fiber Composites, Butterworth-Heinemann, London, 1993
2.
Strife, J. R., Sheehan, J. E., Ceramic Coating for Carbon-Carbon Composites,

Ceramic Bulletin, Vol.67, No.2, 1988, pp. 369-374
3.
Luthra, K. L., Oxidation of Carbon-Carbon CompositesA Theoretical Analysis,

Carbon Vol.26, No.2, 1988, pp. 217-224
Yongdong, Xu, Litong, ZHANG and Laifei, CHENG, Three Dimensional Carbon
Fiber Reinforced Silicon Carbide Composites Prepared By Chemical Vapor
Infiltration, J. Chinese Ceramic Society, Vol.24, No.5, 1996, pp. 485-489
5.
Litong, ZHANG, Laifei, CHENG, Measurements for the Wetting Angel Between
Ceramics and Metals, Yuhang Cailiao Gongyi, No.99, 1988, pp. 49- 5
5
6.
Knacke, O., Kubaschewski, O. and Hwsselmann, K., Thermochemical Properties of

Inorganic Substance, Heidelberg, Berlin, 1991
7.
Deal, B, E., General Relationship for the Thermal Oxidation of Silicon, J. Appl.
7
Phys., Vol.36 ,No.12, 1965, pp. 3770- 8
8.
Schiroky, G. H., Oxidation Behavior of Chemically Vapor-Deposited Silicon

Carbide, Advanced Ceramic Materials, Vol. 2, No.2, 1987, pp. 137- 41
1
II - 637
SUBSURFACE DAMAGE MEASUREMENT ON

CARBON CARBON COMPOSITE MATERIALS IN
BRAKING APPLICATIONS
Yves Remond and Christiane Wagner
Groupe de Modlisation et Simulation en Plasturgie et Mcanique ECPM
UP22 CNRS - Universit Louis Pasteur 4, rue Boussingault, 67000 Strasbourg FRANCE
SUMMARY : This paper deals with the measurement of subsurface damage of composite
materials after braking solicitations. Several carbon carbon composites materials are studied.
They have been tested under industrial braking conditions. Different damage methods are also
used to estimate the braking effects on the mechanical behavior of these materials. In
particular, a meso-hardness test is adapted to the heterogeneity of carbon carbon composite
and to their porosity. Depending on the type of material, the results show the evolution of the
meso-harness as a function of the distance to the braking surface. In the case of the quasi
transverse isotropic behavior, we obtain a measurement of a damaged subsurface. An other
original compression-bending test is also used, which confirms this damaged subsurface
effect before wear occurs.
KEYWORDS: carbon-carbon, braking, wear, subsurface, damage, hardness, compression
INTRODUCTION
Carbon carbon materials present a good mechanical behavior under specific conditions as
high temperature. Besides, their low density is a main quality used for manufacturing brake
discs of heavy vehicles as planes or trains. In this case, the braking system is made of several
carbon carbon discs which are rubbing one against another - rotor against stator. Three
different carbon carbon composites have been tested under industrial conditions as braking to
a stop, and with neutral, wet or usual environment. Recalling the main results of thermal and
wear behavior, we develop here experiments on CC composite materials after series of
braking tests in order to measure the braking effects on the composite subsurface. For
homogeneous materials, several authors show the deterioration of the material, with plasticity
or microcraks, in proximity of the damaged surface [1], [2], [3]. Then we develop specific
tests to quantify a possible damage during and after wear (braking) solicitations.
CARBON CARBON COMPOSITE MATERIALS AND BRAKING

Description
Three carbon carbon composites materials are compared here [4].
Material A : Carbon carbon laminates with thick woven layers in different directions in the
disc plane. This material presents a quasi isotropic behavior in this plane.
II - 638
Material B :
Material C :
Carbon carbon laminates with long fibers in a random distribution in the disc
plane. As material A, this one presents a similar isotropic behavior in this
plane.
Tridirectional carbon carbon composite material with needle bonding normal to
the disc and presenting an orthotropic behavior.
Thermal and Wear Behavior
neutral
humid
Temperature (C)
800
laboratory
700
600
500
400
300 V = 66 m/s
200 random fibers CC (material B)
1
3
4
Depth (mm)
Wear rate ( mm 3/ MJ)
During the braking tests, Jeanne [5] measures the disc temperature and the wear rate. The
main test characteristics are: a normal pressure of 1.3 MPa and angular velocity of 365 rad.s-1.
With these parameters, the surface energy density reaches 300 J/cm2. An example of
temperature evolution is given in Fig. 1 (material B). The measurements are realized at
distance of 1 mm, 3.5 mm and 6 mm to the friction surface. The highest temperature varies
with the type of environment up to 1000C with the experimental conditions. In the same
way, the thermal gradient varies too and presents a low variation under wet environment.
neutral
humid
laboratory
600
400
200
0
Fig. 1: Temperature evolution (material B)

and wear rate of carbon carbon composites (A, B and C) during braking
A:laminate
B:random
C:3D
0,6
A:laminate
0,5
-20
Damage
Stress (MPa)
-40
-60
0,4
0,3
0,2
0,1
-80
0,0
-0,90 -0,75 -0,60 -0,45 -0,30 -0,15 0,00
Strain (%)
10 20 30 40 50 60 70 80
Stress (MPa)
Fig. 2: Mechanical behavior (material A) and damage in compression test

The whole wear rate after different braking tests are presented below. For each material, this
higher wear rate does not occur for the same environment. That shows the great complexity of
the phenomena associated with thermal, chemical and mechanical processes [7], [8], [9], [10].
II - 639
Modeling of the Mechanical Behavior

The three carbon carbon composite materials present a similar macroscopical behavior. Fig. 2
shows a characteristic compression behavior (material A) in the disc plane with cyclic loading
and unloading to free stress. They present an elasto plastic behavior with a linear damage
evolution.
As we can see in Fig. 3, the Young modulus normal to the disc plan is very small.
The in plane elastic behavior is determined by :
ei = ii / E 0i(1 d i)
i
i
with i=1,2 and di=1-Ei/Eio
The damage evolution is described by :
d = f S ) = ( S e ) / b
(
with : S = Sup t
Beside a typical non linear modeling is obtained by the equations :

g and p = g

p =
L
~
~
~
where g is the coupled plastic criterion : g( ) = II L(p) with : =s/(1-D)
and L the hardening function which must be identified for each material.
E1=E2
(GPa)
E3
b
dcritique rupture pl.rupt. (%) rupture (MPa)
e
(GPa) (MPa) (MPa)
(%)
Material A
30
20
137,5
0,7
0,9
0,15
75
Material B
32
0,7
10
66,7
0,6
0,52
0,075
50
Material C
18
1,25
20
44,5
0,45
1,18
0,1
40
Fig. 3: Mechanical characteristics of CC
MESO-HARDNESS
A specific meso-hardness test is developed here taking into account the particular geometry of
the three carbon carbon materials. A spherical indenter is chosen and optimized. Its diameter
must be proportional to the surface elementary cell size in order to measure the mesohardness corresponding to the homogeneous behavior and not the meso-hardness of the
constituents. The carbon carbon samples are then machined with a succession of steps whose
thickness is 0.05 mm. For each step we measure twelve meso-hardness under progressive
loading. The mass used in the test are 0.3 kg, 1 kg and 15 kg. The optimal diameter of the
spherical tip is 6.35 mm. Each test lasts about one minute.
II - 640
During the meso-hardness test, the indenter penetrates the sample in a very small thickness,
much smaller than the thickness of a fiber layer. In that case, it is possible to obtain the
evolution of the meso-hardness in the material, due to the braking solicitations.
The results are analyzed with two classical concepts of harness : HB is the Brinell hardness
taking into account the permanent or plastic deformation of the material under loading : hp.
HR is the Rockwell hardness taking into account the total deformation, id est the elastoplastic deformation of the material under the indenter : hep.
With F, the load, D, the diameter of spherical tip and d the diameter of the indentation area,
we have :
HB =
F
F
, HR =
Dh p
Dh ep
hep=[D-(D2-d2)1/2]/2
The measure of hp is obtained during the test with unloading to free stress.
Results are presented in Fig. 4a, 4b and 4c.
Material A :
The meso-hardness from the braking surface to the middle of the sample
describes a periodical evolution where the periodic distance is equal to the
thickness of the fiber woven layers. It is difficult to observe a particular
evolution of the hardness near the braking surface.
Material C :
The hardness evolution does not present a characteristic variation between the
surface and the middle of the sample. We note that the accuracy of the results
is not good. That is an indicator of the heterogeneity of this material coming
from the needle bonding normal to the disc plane.
Material B :
The results are presented here in term of damage, with D a scalar parameter
equal to the ratio of the meso-hardness in the middle of the sample and the
local meso-hardness [6], [4].
D = 1
H
H*
Lematre [6] showed a similar evolution of the damage measure by classical tests as tension
test D=1-E/E0 and hardness test D=1-H/H* for isotropic material. If E0 represents the initial
Youngs modulus, the meso-hardness H* is more complex to obtain. In the case of an
elastoplastic behavior with hardening, the damage must appear at the same time as hardening.
So, if H represents the hardness of the material after loading, and H0 the hardness before
loading, we should obtain :
H> Ho with
H0=k p ,
H= k (R+p) (1-D) ,
R is the hardening variable and p the plastic low threshold stress.

The measure of damage must be so :
II - 641
H
with H*= k(R+p) hardness of the material with hardening but without damage,
H*
obtained by calculus. For the carbon carbon composite material B under braking solicitation,
we observe a hardness H at the surface smaller than H0 in the middle of the sample, H<H0.
That means that damage appears without hardening, this phenomenon is similar to the damage
evolution in elastic fatigue test for isotropic metals. Consequently, we propose the hypothesis
that under braking solicitations, appears a damage evolution mainly due to thermal and
physicochemical phenomena, and without hardening.
D = 1
15
Hardness
Mean hardness
12
Rockwell Hardness (GPa)
Rockwell Hardness (GPa)
14
10
8
6
Material A
4
0,0
0,5
1,0
1,5
2,0
2,5
Hardness
Mean hardness
14
13
12
11
10
9
8
3,0
Material C
0,0
0,2
Depth (mm)
(a)
0,6
(b)
HB (sample 1 et 2)
HR (sample 2)
HR (sample 1)
HR (sample 3)
0,20
Material B
0,15
Damage
0,4
Depth (mm)
0,10
0,05
0,00
0
2
3
Depth (mm)
(c)
Fig. 4: Meso hardness ((a) and (b)) and damage (c)
The model of damage measurement is then :
D = 1
H
H*
with H*=k p and H=k p (1-D)

The damage evolution shows a decreasing from the surface of the material.
II - 642
0,8
1,0
In the middle of the disc, the variation of measurement H corresponds to the heterogeneity
of the carbon carbon with a relative ration : H/H* = 7.5%.
Near the surface, the ratio H/H* measured reaches 20%. Between 7.5% and 20%, it is
possible to consider this ratio as a measurement of damage in a subsurface of this carbon
carbon composite material. The depth of this subsurface is a=0.4 0.1 mm. The thickness of
this damaged subsurface will probably vary as a function of the nature of the braking
solicitation (frequency, energy level...). In case of very dissipate braking, the damaged
subsurface will disappear in wear.
COMPRESSION BENDING TEST

The brake discs in material B are damaged only in a small thickness near the surface. We tried
to define a macroscopical test sensitive to a variation on rigidity in the subsurface. The
principle of this test is shown in Fig. 6a and 6b. Parallelepipedic samples are taken out of the
brake discs after braking test.
If a sample is not damaged, its rigidity center and its geometrical center match each other in a
cross section. If not, appears a distance between these two points, that generates a bending
couple when the sample is loaded in compression. Using strain gauges in all the lateral
surfaces of the sample, it is possible to measure the bending effect due to the damaged
subsurface. A specific experimental setup (Fig. 6b) is then constructed in order to apply the
compression load with a good accuracy (0.01 mm) exactly in the center of the cross-section.
A steel ball permits the free rotation of the head sample.
geometrical center
= rigidity center
y
y = l
/2
geometrical
l center rigidity
/2
center
strain
gauges
CC sample
wedges
a
l
undamaged
CC material
l
damaged
CC material
Fig. 6: Samples and compression-bending setup

This experimental setup has been tested on model materials for which a subsurface with a
different rigidity was prepared and then well known. Its accuracy permits to detect a damaged
subsurface with the following characteristics : a.d = 0.02 mm with a : the thickness of the
damaged zone and d : the mean damage in this subsurface.
II - 643
With Ei, the Young modulus of the CC sample in the compression direction, and l/2- the
distance between the center of rigidity and the center of the cross section, we can write :
E (x )dS = 0
s
(1)
Suppose that the damage evolution is linear, as shown with the meso-hardness results, we
obtain :
Ei(x) = Eoi( 1-d(x))
for x [l-a , l ]
o
and Ei(x) = E i
for x [0 , l-a ].
Assume too that :
Ei(x) = E0i [l +dc[(l -x)/a- l]]
1,2x10
Rotation (rad/mm)
1,0x10
8,0x10
6,0x10
4,0x10
2,0x10
-4
d=0,4
-4
d=0,15
d=0,2
d=0,1
-5
-5
-5
Bending
of material B
measured by
compressive-test
-5
Damaged depth
0,0
0,0
0,1
0,2
0,3
0,4
0,5
Damaged depth (mm)
Fig. 7: Identification of the damaged depth
It means that the critical damage is obtained at the braking surface, so previous relation (1),
could be written :
2 = ( dma(a-2 l) + l
) / (l -dm a )
Note that this scalar equation permits only to obtain one scalar unknown, the measure of by
the rotation of the head sample permits to find the thickness of the damaged subsurface.
The results of this compression bending test are shown in Fig. 7.
All specimens in material B present a bending effect corresponding to the values of a and d :
a.dm= 0.05 mm
Then if we consider a linear evolution of the damage in the subsurface, with a critical damage
dc=0.2 measured with the meso-hardness results, we obtain a thickness of this subsurface
equal to a=0.35mm.
II - 644
DISCUSSION AND CONCLUSION

Note that the measure of the bending effect must be absolutely realized in the beginning of the
rotation. If not, a complementary bending couple appears due to the progressive lateral
displacement of the setup induced by the bending strains.
The both results of meso-hardness and compression bending test developed here are in good
agreement. They show that, for the carbon carbon composite materials B, it can exist a
damaged subsurface induced by specific braking solicitations. Naturally, this subsurface is
progressively turning into wear particles, during a braking test. However, it seems to be
necessary to take into account the real damaged behavior of materials in order to have a better
understanding of their surface deterioration. The main difficulties which limit these approach
are the heterogeneity of carbon carbon composites and their porosity, which stop us
developing others methods to measure this subsurface.
REFERENCES
1.
A.T. Alpas, J.D. Embury, "The role of subsurface deformation and strain localization on
the sliding wear behaviour of laminated composites", Wear, 146, 1991, pp. 285-300
2.
J.D. Chen, J.H. Chern Lin, C.P. Ju, " Effect of load on tribological behaviour of carboncarbon composites ", Journal of Materials Sciences 31 (1996), pp. 1221-1229
3.
M. Cirino, R.B. Pipes, K. Friedrich, "The wear behavior of continuous fiber polymer
composite", ICCM 6th, Vol 5, London 1987, pp. 311-323
4.
C. Wagner, "Matriaux composites carbone carbone et freinage : tude de

lendommagement au voisinage dune surface de contact", Thse de doctorat de
lUniversit Louis Pasteur, Strasbourg, France, 1996
5.
J.F. Jeanne, " Etude de l'endommagement surfacique de matriaux composites type

Carbone/Carbone sous sollicitations tribologiques ", Thse de Doctorat, Ecole Centrale
de Nantes, 1994
6.
J. Lemaitre, J. Dufailly, R. Billardon, C. R. Acad. Sc. Paris, "Evaluation de

l'endommagement par mesures de microduret ", t.304, Srie II, N12, 1987,
7.
F. Durand, D. Rouby, G. Fantozzi, "Characterization of the high-temperature

mechanical behaviour of carbon materials ", Carbon vol 32, No 5, pp.857-865, 1994
8.
J. Xiaoyu, " Frictionnal contact analysis of composite materials, " Composite science
and technology 54, 1995, pp. 341-348
9.
F.M. Kustas, R.R. Hanson, J.L. Summer, " Tribological performance of 3D carboncarbon composite, Part I : Ambient Environment ", Lubrification Engineering, Vol.51, 7
(1995), pp. 599-604
10.
C. Ahearn, B. Rand, " Modification of the fibre-matrix bonding in a brittle carboncarbon composite by controlled oxidation ", Carbon, Vol 34, N2, 1996, pp. 239-249.
II - 645
MICROSTRUCTURAL CHARACTERISATION OF 3-D

WOVEN SIC/SIC-BASED COMPOSITES AFTER
TENSILE TESTING AT ROOM AND ELEVATED
TEMPERATURE IN DIFFERENT ATMOSPHERES
Ian J. Davies1, Takashi Ishikawa1, Masaki Shibuya2, and Tetsuro Hirokawa3
1
Airframe Division, National Aerospace Laboratory,

6-13-1 Ohsawa, Mitaka-Shi, Tokyo 181, Japan
2
Corporate Research and Development, Ube Industries Ltd.,
1978-10 Kogushi, Ube-Shi, Yamaguchi 755, Japan
3
Research and Development Department, Industrial Textile Division, Shikibo Ltd.,
1500-5 Shibahara-Minani, Yokaichi-Shi, Shiga 527, Japan
SUMMARY: 3-D woven SiC/SiC-based composite consisting of surface-modified Si-Ti-C-O

fibres and polymer conversion-derived matrix was tested in tension in vacuum (1200-1380oC)
and air (1000-1200oC). In addition, several specimens were surface-sealed using a proprietary
technique prior to mechanical testing. Resultant tensile property/strain curves are discussed
together with fracture surface characteristics such as main failure mode and fibre pull-out
length distribution. The effect of test condition on intra-fibre bundle spatial distribution of
fibre fracture characteristics is also mentioned. Composite properties at room and elevated
temperature in vacuum (unsealed) and air (surface-sealed) were generally consistent with a
stable fibre/matrix interface. Unsealed specimens tested at elevated temperature in air
exhibited properties consistent with extensive fibre/matrix interface oxidation.
KEYWORDS: microstructural characterisation, SiC/SiC, tensile test, stress/strain, fibre pullout, fracture mirror
INTRODUCTION
Ceramic matrix composites (CMCs) which have shown potential for use in high temperature
structural applications include those based on the SiC/SiC system. These composites contain
fibres with either Si-C-O or Si-Ti-C-O structures comprising of -SiC nanoparticles partially
surrounded by aromatic carbon layers in an amorphous SiOxCy matrix for the Si-C-O fibre
and SiOxCyTiz matrix for the Si-Ti-C-O fibre. Matrix densification is generally achieved using
either chemical vapour infiltration (CVI) or polymer conversion (PC) techniques.
Significant research has been conducted into optimisation of the fibre/matrix interface which
is known to significantly influence high temperature mechanical properties (particularly for
oxidising atmospheres). The problem of high temperature fibre/matrix interface stability has
also been addressed through coating or sealing the composite surface with an oxidationresistant glass layer.
II - 646
The present work is concerned mainly with mechanical and microstructural characterisation
of a 3-D woven SiC/SiC-based composite (without an oxidation-resistant glass layer) after
tensile testing in vacuum and air from room temperature (RT) up to 1380 oC. In addition,
exploratory tests were also carried out on a limited number of specimens after the surface had
been sealed with an oxidation-resistant glass compound.
The composite examined in this work utilised continuous Si-Ti-C-O fibres woven into an
orthogonal 3-D configuration with fibre ratios in the x, y, and z directions being 1, 1, and
0.13, respectively and a fibre volume fraction estimated to be 40 %. The Si-Ti-C-O fibres
(Tyranno LoxM) had been surface-modified to produce an outer 10 nm SiOx-rich layer
surrounding an inner 40 nm carbon-rich layer whilst matrix densification used the PC method
and a precursor similar to polytitanocarbosilane (PTCS). After machining to a geometry
suitable for tensile testing, a limited number of specimens were surface-sealed using a
proprietary technique (Kawasaki Heavy Industries, Kakamigahara City, Japan).
Mechanical testing of unsealed specimens was carried out with the loading direction parallel
to the specimen y-axis (experimental details being given elsewhere) at RT and also at elevated
temperature in vacuum (13 Pa)(1200-1380 oC) and in air (1000-1200 oC). Surface-sealed
specimens were tested in a similar configuration up to 1200 oC in air. It should be noted that
tensile strain could not be accurately measured beyond 1300 oC in vacuum (limitation of the
specimen grip) and also for unsealed specimens at elevated temperature in air.
Experimental stress/strain curves were characterised by computer fitting of a high-order
polynomial curve after which the instantaneous tensile modulus, E, and cumulative damage
energy could be derived. Cumulative damage energy (CDE) indicates the energy absorbed by
the composite during failure and so may be used to qualitatively measure the toughness of
the composite, with:
S
1
Cumulative damage energy =
FdA
b. h. Lo
0
(1)
where b is the mean initial test specimen width (m), h is the mean initial test specimen
thickness (m), Lo is the specimen gauge length (m), F is the tensile force acting on the
specimen (N), S is the composite failure strain, and A is the longitudinal elongation (m). The
majority of energy absorbed during failure is known to result from fibre pull-out so that CDE
may also provide information concerning fibre pull-out properties and hence the status of the
fibre/matrix interface.
Microstructural observations to evaluate the general nature of the fracture surface were
initially carried out using a zoom lens optical microscope (OM) (Olympus SZH) at low
magnification (20). Additional investigation concerning the fracture surface in the composite
zy-plane was carried out at higher magnification (50) (Olympus STM5-BD). Fibre pull-out
length distributions were determined using this technique through measuring the projection of
fibre pull-out length in the zy-plane. Obscuration of fibres with small pull-out lengths behind
those with larger pull-out lengths was minimised through considering only fibres which had
II - 647
pulled-out nearest to the microscope lens. Fibre pull-out length distributions will also be
examined using scanning electron microscopy (SEM) at a later date.
Preliminary examination of individual fibre fracture surfaces was carried out for selected
specimens using SEM (JEOL JSM-6300F). In order to determine any spatial dependence of
fibre fracture properties within fibre bundles, fibres were examined at the end and centre of
the fibre bundle major axis for each specimen. In both cases the fibre bundle was scanned
from edge to edge along its minor axis with all fibres within an approximate 100 m
corridor being analysed. A minimum of 100 fibres were analysed in each area with the
relative position, fibre diameter, surface character, and fracture mirror size (if applicable)
being measured for each fibre. Only initial results are presented here with a more complete
analysis being given elsewhere.

Mechanical Properties
Representative tensile property/strain curves for composites examined in this paper have been
presented in Fig. 1. The unsealed composite exhibited excellent mechanical properties at RT
with a tensile strength of 389 (+/-48) MPa (Fig. 2) and tensile strain to failure of 1.14 (+/0.04) % (Fig. 3(a)). Such values, combined with a CDE of 2634(+/-340) kJm-3 (Fig. 3(b))
compare extremely favourably with those obtained for previous SiC/SiC-based composites
and suggest the fibre/matrix interface to possess near-optimal properties. Tensile strength, ,
and modulus, E, values obtained from Fig. 1 were broadly similar to those predicted by
modified Rule of Mixtures equations:
E = fEfVf + EmVm
(2a)
= ffVf + mVm
(2b)
and
where V is the volume fraction of the respective phase, f is the fibre orientation factor, and
the subscripts f and m refer to fibre and matrix properties respectively, with the following
assumptions:
The decrease in tensile modulus with increasing strain at small strain levels (0.2 %) was
attributed to a reduced contribution from the matrix component (due to microcracking). The
matrix in SiC/SiC-based composites is known to have a substantially lower tensile strain to
failure (0.1 %) compared to that of the fibres (1 %).
Mechanical properties from 0.4 % until just prior to failure were due almost entirely to
fibre and fibre/matrix interface properties.
Following on from the previous assumption, the composite failed at a tensile strain similar to
that of the fibres. However, the tensile strength of the composite was only 50 % of that
predicted from Eqn. 2b (with mVm0). A large proportion of this difference may be
attributed to fibre strength reduction upon inclusion in the composite whilst another factor in
II - 648
Tensile stress (MPa)
the present case may be the non-alignment of fibres along the loading axis (due to the
weaving procedure).
400
Legend: as given below
300
200
100
(a)
Room temperature
1200 oC/vacuum
1300 oC/vacuum
1200 oC/air/100 hours
1000 oC/air
1100 oC/air
1200 oC/air
200
150
100
sealed unsealed
Tensile modulus (GPa)
0
250
(b)
50
0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Tensile strain (%)
Fig. 1: Tensile property curves for unsealed and surface-sealed SiC/SiC-based composite: (a)
stress/strain, and (b) modulus/strain
It should also be noted that the composite initial tensile modulus was determined by
significant contributions from both fibre and matrix. This situation is different from the case
of SiC/SiC-based composites with CVI-derived matrices where the majority of the initial
tensile modulus contribution is due to the matrix.
Tensile properties were almost identical for unsealed composite at 1200 oC in vacuum as for
the RT case (Figs. 1-3) apart from a slight reduction in initial tensile modulus (attributed to
matrix degradation) indicating the fibre/matrix interface to be stable at 1200 oC in vacuum.
Further increases in test temperature for unsealed specimens in vacuum resulted in a 50 %
decrease in tensile strength by 1380 oC (compared to RT and 1200 oC). A decrease in
mechanical properties under these conditions would be expected as Si-Ti-C-O fibres are
known to degrade at elevated temperature, even in vacuum. Thus, a decrease incomposite
mechanical properties under such conditions does not necessarily imply a degradation of the
fibre/matrix interface.
II - 649
Tensile strength (MPa)
400
300
200
unsealed/vacuum
unsealed/air
sealed/air
100
0
RT
1000
1100
1200
1300
1400
Test temperature (oC)

Fig. 2: Tensile strength vs test condition for unsealed and surface-sealed SiC/SiC-based
composite. Note non-linear temperature scale
Although strain measurement was not possible for unsealed composite at elevated temperature
in air, qualitative observations indicated the tensile stress/strain curve to be approximately
linear with a strain to failure of <0.1 % and, as such, is consistent with mechanical
property/strain data presented in Fig. 1 for unsealed composite tested at RT after ageing in air
(1200 oC/100 hours). The implication is that mechanical properties of unsealed composite are
significantly reduced by exposure to oxygen at elevated temperature and this is borne out by
tensile strength data presented in Fig. 2. The exact mechanism for the reduction in mechanical
properties under these conditions has not been established for the composite under
investigation but would be consistent with oxidation of the fibre/matrix interface producing a
substantial increase in interface sliding stress and transformation to notch-sensitive behaviour
observed by previous authors. The fibre/matrix interface in the present composite is known to
consist essentially of a carbon-rich layer which makes such a scenario likely.
Tensile property/strain curves for surface-sealed composite tested at elevated temperature in
air have also been included in Fig. 1. Although it should be stressed that such data is
preliminary, the curves appear similar to those of unsealed composite at elevated temperature
in vacuum. Such a conclusion may also be inferred from comparison of respective tensile
strength (Fig. 2), strain to failure (Fig. 3(a)), and CDE (Fig. 3(b)) data. It is suggested that
surface-sealing of the specimen is effective in severely retarding oxidation of the fibre/matrix
interface, at least for short-term exposure (in the order of several minutes). These results are
also consistent with preliminary creep investigations (1100 oC in air) of surface-sealed
composite which have shown typical lifetimes of 6 x 104 s at 200 MPa and may indicate that
long-term exposure at elevated temperature in air is feasible for these materials.
II - 650
Tensile strain to failure (%)
1.5
1.2
0.9
0.6
unsealed/vacuum
sealed/air
(a)
unsealed/vacuum
sealed/air
0.3
(b)
0.0
Cumulative damage
energy (kJm-3)
3000
2000
1000
0
RT
1000
1100
1200
1300
Fig. 3: Parameters determined from tensile stress/strain curves of unsealed and surfacesealed SiC/SiC-based composite: (a) tensile strain to failure, and (b) cumulative damage
energy. Note non-linear temperature scale
Microstructural Investigation
The composite tensile tested at RT exhibited a brush-like fracture surface (Fig. 4(a)) with
extensive fibre pull-out characteristic of tough CMCs and reflected in the relatively high
CDE values achieved (Fig. 3(b)). The primary crack path in the composite tested at RT
followed the inter-fibre bundle porosity perpendicular to the loading axis, indicating such
pores to be a preferred route for crack propagation under these conditions. A similar fracture
mode was also observed in unsealed specimens tested at elevated temperature in vacuum and
also for surface-sealed specimens tested at elevated temperature in air, indicating crack
initiation and propagation to be similar for these cases.
In contrast to this, unsealed composite tested at elevated temperature in air possessed a
relatively smooth fracture surface (Fig. 4(b)) with little evidence of fibre pull-out, indicating
brittle failure to have occurred. As stated earlier, such a failure mode was attributed to
oxidation of the fibre/matrix interface. The primary crack path in this case showed no
preference for the inter-fibre bundle porosity and was consistent with crack initiation within a
fibre bundle followed by catastrophic crack propagation.
II - 651
Fig. 4: Optical micrographs illustrating fracture surfaces for unsealed SiC/SiC-based

composite after tensile testing: (a) room temperature, and (b) 1200 oC in air
Fig. 5(a) illustrates the effect of test temperature on fibre pull-out length distribution for
unsealed composite tensile tested in vacuum with the distributions being similar for the RT,
1200 oC, and 1350 oC cases. The shift in fibre pull-out length to smaller values at 1300 oC was
attributed to a transient deterioration in vacuum condition during testing. Previous researchers
found that exposure to oxygen at elevated temperature may increase the fibre/matrix interface
sliding stress by an order of magnitude compared to that at RT and thus even a minor decrease
in vacuum quality may have a noticeable effect on fibre/matrix interface-dependent
properties. A bimodal fibre pull-out length distribution was noted for the composite when
tested at 1380 oC with a transition point of approximately 370 m. Although not fully
understood at present, such a change in behaviour at 1380 oC would be consistent with
changes in the Si-Ti-C-O fibre (and fibre/matrix interface) at this temperature.
Preliminary fibre pull-out length distributions for surface-sealed specimens tested in air (Fig.
5(b)) show the pull-out length range to be similar to that of the unsealed composite tested at
RT and elevated temperature in vacuum (Fig. 5(a)). The general form of the distributions in
Fig. 5(b) is uncertain but at least one specimen (1200 oC) bears a resemblance to those seen in
Fig. 5(a). From mechanical property data presented earlier it would be reasonable to expect
fibre pull-out distributions for surface-sealed specimens at elevated temperature in air to be
similar to those of unsealed specimens in vacuum at the respective test temperature. Thus, the
fibre pull-out length data appears to add weight to the unsealed specimen fibre/matrix
interface being stable at elevated temperature in vacuum (at least up to 1350 oC) with similar
behaviour probable for surface-sealed specimens at elevated temperature in air.
II - 652
Preliminary fibre fracture surface characteristics of unsealed composite tested in air at RT,
1100 oC, and 1200 oC were examined using SEM. Fibre fracture surfaces could be
characterised into two main groups (rough fracture surface (RFS) and smooth fracture
surface (SFS)) with the RFS fibres being further sub-divided into three categories, i.e.,
(i)
(ii)
(iii)
(iv)
rough fracture surface with a fracture mirror origin at the fibre surface (RFS-S),
rough fracture surface with a fracture mirror origin within the fibre body (RFS-B),
rough fracture surface with no obvious fracture mirror (RFS-N), and
smooth fracture surface (SFS).
Cumulative failure
1.0
0.8
0.6
RT (n = 1015)
1200 oC (n = 464)
1300 oC (n = 182)
1350 oC (n = 235)
1380 oC (n = 490)
0.4
0.2
(a)
Cumulative failure
0.0
1.0
0.8
RT (n = 1015)
1000 oC (n = 221)
1100 oC (n = 262)
1200 oC (n = 188)
0.6
0.4
0.2
(b)
0.0
100
1000
Fibre pull-out length (m)
Fig. 5: Cumulative failure vs fibre pull-out length for SiC/SiC-based composite: (a) unsealed
tested in vacuum, and (b) surface-sealed tested in air
Fig. 6 illustrates the fraction of fibres with different fracture surface characteristics as a
function of test temperature and position with the fibre bundle for unsealed composite tested
in air. Although not included in Fig. 6, the fraction of RFS-B fibres in all cases was almost
insignificant compared to that of RFS-S fibres (2% for the RT specimen) and is consistent
with results of previous researchers. It may also be observed that the percentage of SFS fibres
was substantially larger at 1100 oC and 1200 oC compared to at RT. One explanation for the
presence of SFS fibres is that they represent fibres with such low fracture strength that a
fracture mirror boundary was not formed. However, the more likely scenario in the present
case is of the fibre and matrix being so strongly bonded following exposure to oxygen at
elevated temperature that the propagating crack passed from matrix to fibre with little or no
II - 653
change in velocity or direction. Further evidence for such a scenario is that all SFS fibers
possessed negligible fibre pull-out.
It was also noted that SFS fibres were preferentially positioned around the perimeter of the
fibre bundle, i.e., not in the centre. This can be more clearly observed in Fig. 7(b) with SFS
fibres being preferentially positioned towards lower and higher values of the fibre bundle
minor axis (which corresponds to the outer edge of the fibre bundle). In contrast to this, RFSS fibres appear mainly limited to the central portion of Figure 7(b) parallel to the major axis
(which corresponds to the central region of the fibre bundle). If the SFS fibres were indeed
due to oxidation of the fibre/matrix interface, as seems likely, then it would suggest oxygen to
have infiltrating the fibre bundle from the outer surface. Such a proposition is in contrast to
previous researchers who concluded that oxygen attacks the fibre/matrix interface along the
length of individual fibres. One explanation for this difference may be that the narrow (40
nm) carbon-rich layer at the interface in the present composite allows rapid SiO2 formation
and so limits oxidation to the composite surface, i.e., self-healing behaviour .
RFS-S
RFS-N
SFS
Centre
Edge
Fibre fraction
1.0
0.8
0.6
0.4
0.2
0.0
RT
1200
1100
Distance along minor axis (m)
Fig. 6: Fibre fracture characteristics at the centre and edge of a fibre bundle for unsealed
SiC/SiC-based composite tensile tested in air. Note non-linear temperature scale
200
(a)
(b)
160
120
80
40
RFS-S
RFS-N
SFS
40
80
120 0
40
80
120
Distance along major axis (m)
Fig. 7: Fibre fracture surface characteristics at the centre of a fibre bundle after tensile
testing: (a) room temperature, and (b) 1100 oC in air
II - 654
CONCLUSIONS
(i) The main characteristics of tensile property/strain curves were found to be consistent with
predictions obtained from the modified Rule of Mixtures equation together with several
assumptions.
(ii) At room temperature the composite exhibited extensive fibre pull-out with the main crack
path following the inter-fibre bundle porosity. This was accompanied by excellent
mechanical properties including a tensile strength of 389 (+/-48) MPa and strain to failure
of 1.14 (+/-0.04) % and suggested the fibre/matrix interface to have near-optimal
properties.
(iii) Unsealed composite tested at 1200 oC in vacuum possessed similar mechanical properties
(except initial tensile modulus) and fracture surface characteristics as for the room
temperature case. Increasing the test temperature further resulted in similar fibre pull-out
properties but decreased tensile strength (50 % at 1380 oC) compared to at room
temperature. It was proposed that the fibre/matrix interface was stable up to at least
1350oC in vacuum.
(iv) Unsealed composite tested in air at 1100 oC and 1200 oC had a relatively flat featureless
fracture surface with little evidence of fibre pull-out. This was reflected in poor
mechanical properties and attributed to oxidation of the fibre/matrix interface under such
conditions.
(v) Surface-sealed composite tested in air up to 1200 oC had similar mechanical, fracture
surface, and (possibly) fibre pull-out properties as for the unsealed composite in vacuum.
This indicated that the surface-sealant utilised was successful in dramatically reducing
oxygen diffusion into the composite under these conditions.
(vi) Most fibres in the composite tested at room temperature possessed rough fracture
surfaces with well-defined fracture mirrors originating at the fibre surface (RFS-S).
However, for the unsealed specimen tested in air at elevated temperature the majority of
fibre fracture surfaces were smooth with no evidence of a fracture mirror (SFS)
whilst RFS-S fibres were mainly limited to the central portion of the fibre bundle. These
results suggested oxidation to have proceeded from the outer surface of the fibre bundle
(as opposed to along the fibre/matrix interface of individual fibres).
ACKNOWLEDGEMENTS
This work was supported by funding from the Science and Technology Agency of Japan
whilst one of the authors (IJD) was supported as a Science and Technology Agency Fellow.
The authors gratefully acknowledge Y. Nomura and N. Suzuki for help with mechanical
testing.
II - 655
REFERENCES
1.
Bodet, R., Bourrat, X., Lamon, J., and Naslain, R., Tensile creep behaviour of a
silicon carbide-based fibre with low oxygen content, Journal of Materials Science,
Vol. 30, 1995, pp. 661-677.
2.
Bunsell, A. R., Berger, M. H., and Hochet, N., Structural and mechanical
characterisation of some alumina and SiC based fibres, High-temperature ceramicmatrix composites II (A. G. Evans and R. Naslain, Ed.), Ceramic Transactions, volume
58, American Ceramic Society, Westerville, 1995, pp. 85-94.
3.
Droillard, C., Voisard, P., Heibst, C., and Lamon, J., Determination of fracture
toughness in 2-D woven SiC matrix composites made by chemical vapor infiltration,
Journal of the American Ceramics Society, Vol. 78, No. 5, 1995, pp. 1201-1211.
4.
Ishikawa, T., Shibuya, M., Hirokawa, T., and Watanabe, N., Mechanical properties of
3-D fabric ceramic composites using new PC type matrix for improved oxidation
resistance, Proceedings of the 37th JSASS/JSME structures conference, July 12th14th, Fukuoka, Japan, Japan Society of Aeronautical and Space Sciences, Tokyo,
1995, pp. 29-32 (in Japanese)
5.
Naslain, R., Challenging ceramic matrix composites for applications in severe

environments, Advanced Composite Materials, Vol. 5, No. 1, 1995, pp. 35-44.
6.
Masaki, S., Moriya, K., Yamamura, T., Shibuya, M., and Ohnabe, H., Development
of Si-Ti-C-O fiber reinforced SiC composites by chemical vapor infiltration and
polymer impregnation and pyrolysis, High-Temperature Ceramic-Matrix Composites
II (A. G. Evans and R. Naslain, Ed.), Ceramic Transactions volume 58, American
Ceramic Society, Westerville, OH, 1995, pp. 187-192
7.
Ishikawa, T., Yamamura, T., Hirokawa, T., Hayashi, Y., Noguchi, Y., and
Matsushima, M., Strength and fracture toughness properties of oxidation resistant
high-temperature ceramic matrix composites, Proceedings of the 9th International
Conference on Composite Materials: Volume II (A. Miravette, Ed.), Woodhead
Publishing Co., Madrid, 1993, pp. 137-144.
8.
Davies, I. J., Ishikawa, T., Shibuya, M., and Hirokawa, T., Stress/strain behavior of
3-D woven SiC/SiC-based composites, submitted to Journal of the American
Ceramic Society, 1996
9.
Advanced technical ceramics - Mechanical properties of ceramic composites at room

temperature. Part 1: Determination of tensile strength, Report no. DD ENV 658-1,
British Standards, Manchester, U.K., March 1993
10.
A. G. Evans, Perspective on the development of high toughness ceramics, Journal of

the American Ceramic Society, Vol. 73, 1990, pp. 187-206
11.
Davies, I. J., Ishikawa, T., Shibuya, M., Hirokawa, T., Optical microscopy of 3-D
woven SiC/SiC-based composites, submitted to Composites Science and Technology,
1996
II - 656
12.
Davies, I. J., Ishikawa, T., Shibuya, M., and Hirokawa, T., Fibre properties in 3-D
woven SiC/SiC-based composites after tensile testing at room and elevated
temperature, in preparation
13.
Droillard, C. and Lamon, J., Fracture toughness of 2-D woven SiC/SiC CVIcomposites with multilayered interphases, Journal of the American Ceramic Society,
Vol. 79, No. 4, 1996, pp. 849-858
14.
Prewo, K., Tension and flexure strength of silicon carbide fibre-reinforced glassceramics, Journal of Materials Science, Vol. 21, 1986, pp. 3590
15.
Emig, G. and Wirth, R., Tensile strength of silicon carbide fibre bundles at elevated
temperatures, Journal of Materials Science, Vol. 30, 1995, pp. 5813-5818
16.
Thouless, M. D., Sbaizero, O., Sigl, L. S., and Evans, A. G., Effect of interface
mechanical properties on pullout in a SiC-fiber-reinforced lithium aluminium silicate
glass-ceramic, Journal of the American Ceramic Society, Vol. 72, No. 4, 1989, pp.
525-532
17.
Davies, I. J., Ishikawa, T., Shibuya, M., and Hirokawa, T., Damage characterisation
of 3-D woven SiC/SiC-based composites, Proceedings of the 21st Symposium on
Composite Materials (Oct. 31st-Nov. 1st 1996, Toyama, Japan), Japan Society for
Composite Materials, Tokyo, 1996, pp. 103-104
18.
Eckel, A. J., and Bradt, R. C., Strength distribution of reinforcing fibres in a Nicalon
fiber/chemically vapor infiltrated silicon carbide matrix composite, Journal of the
American Ceramic Society, Vol. 72, No. 3, 1989, pp. 455-458
19.
Naslain, R., The concept of layered interphases in SiC/SiC, High-Temperature

Ceramic-Matrix Composites II (A. G. Evans and R. Naslain, Ed.), Ceramic
Transactions volume 58, American Ceramic Society, Westerville, OH, 1995, pp. 23-39
II - 657
THE MECHANICAL PROPERTIES OF 3D WOVEN

CARBON FIBERS REINFORCED CERAMIC MATRIX
COMPOSITES
Wenwei Zheng, Xingye Wang, Jiayu Xiao, Fengrong Liu, Dazhi Jiang, Zhaohui Chen
Department of Materials Engineering and Applied Chemistry, National University of Defense

Technology, NUDT, Changsha, Hunan 410073, P.R. China
KEYWORDS: ceramic matrix composites, 3D woven structure, mechanical properties,

carbon fiber, precursor pyrolysis
II - 658
II - 659
II - 660
II - 661
II - 662
II - 663
II - 664
EFFECTS OF TEMPERATURE AND LAYUP ON THE

MIXED-MODE DELAMINATION TOUGHNESS OF A
CERAMIC MATRIX COMPOSITE MATERIAL
Jonathan J. Polaha, Tor W. Sherwood and Barry D. Davidson
Department of Mechanical, Aerospace and Manufacturing Engineering,

Syracuse University, Syracuse, NY 13244
KEYWORDS: delamination, fracture, single leg bending, mixed mode, crack tip element
II - 665
II - 666
II - 667
II - 668
II - 669
II - 670
II - 671
II - 672
II - 673
II - 674
TOUGHENING OF SILICON NITRIDE MATRIX

COMPOSITES BY THE ADDITION OF SI-C-N
NANOMETER PARTICLES
Xingui Zhou, Changrui Zhang, Zhaohui Chen, Kai Xie and Anchen Zhou
Department of Materials Engineering and Applied Chemistry,
Changsha Institute of Technology, 410073, Changsha, Hunan, P.R.China
SUMMARY: In this paper, Si3N4 matrix composites reinforced by Si-C-N nanometer
particles or SiC whiskers were fabricated using the hot-pressing technique. The mechanical
properties of the composites containing various amounts of these Si-C-N nanometer particles
were investigated. The maximum fracture toughness of these composites was 11.96 MPam1/2
and the flexural strength was 878.50 MPa after addition of 20 wt% of Si-C-N nanometer
particles. It showed that the Si-C-N nanometer particles can remarkably improve the
properties of Si3N4 matrix. Fracture surface and the micrographs of local intend areas were
examined by SEM, the results revealed that the reinforcing mechanisms acting in these
nanocomposites were crack deflection and crack branching by Si-C-N particles.
KEYWORDS: toughening, si-c-n nanometer particles, silicon nitride, fracture toughness,

ceramic matrix composites
INTRODUCTION
Si3N4 and SiC ceramic are well known for their high strength, hardness and thermal stability,
as well as the extreme brittleness that prevents their application in many critical components.
The mechanical properties of these ceramics have been improved by the microstructure
development, however, it seems to be difficult to expect the further improvement of
mechanical properties by this technique. The need for further improvement in the mechanical
properties has led to the development of high-strength and high-toughness ceramics, such as
fiber-, whisker- and particulate- reinforced ceramic matrix composites[1,2]. SiC was used as a
reinforcing material for Si3N4 more than 20 years ago[3], for the attempt to take advantage of
the characteristics of both materials. Owing to process problems and the cost of SiC whiskers
and fibers, the attention was being devoted to particlate composites. However, the use of SiC
as uniform particles cannot improve noticeably the properties of the Si3N4 matrix.
Recently, researchers have found that the properties of ceramic can be dominantly improved
by reducing particle sizes to nanometer levels, Niihara and his colleagues have been
investigating the Si3N4 based nanocomposites, in which the nano-size SiC particles are
dispersed mainly with the matrix grain[4-6], the flexural strength reaches 1550 MPa, the
fracture toughness reaches 7.5 MPam1/2. The materials are referred to as nanocomposites
and the approach is generally indicated as a new design concept in the structure ceramic
field. Research on nanometer particles toughening ceramic matrix composites becomes a new
hot point.
II - 675
In this paper, nanometer Si-C-N particle reinforced Si3N4 composites were made, the
mechanical properties were measured, the phase structure was studied by XRD, the main goal
was to improve the properties of Si3N4 matrix and to understand the reinforcing mechanisms
of these nano-size SiC dispersed composites.
Fabrication of Composites
Si-C-N nanometer particles used as the reinforcing materials were made in our laboratory,
through a Chemical Vapor Condensation (CVC) method, the average particle sizes were 70
nm, the total carbon in the particles was 24.08wt%, and the total nitrogen was 18.06wt%,
Fig.1 showed the shape. As another reinforcing materials SiC whiskers with an average
diameter of 0.6 mm were used, starting powders for the matrix material were Si3N4 (average
particle size 1.0 mm), and La2O3 (average particle sizes 2.0 mm), Y2O3 (average particle size
0.5 mm).
As the Si-C-N nanometer particles contained many agglomerates which obstructed the
densification of the composite, the particles must be through dispersion. This dispersion was
executed by ultrasonication and stirring in alcohol. The SiC whiskers were pretreated using
50vol%HF + 50vol%HNO3 for 30 minutes, then washed with deionized water. After the
treatment, the Si-C-N nanometer particles and SiC whiskers were dried and used for the
reinforcing materials.
The Si-C-N nanometer particles (or SiC whiskers) were mixed with the Si3N4 matrix powders
which had been obtained by mixing of Si3N4 powder and sintering-aid powders in advance.
The contents of the sintering-aid powders were 10wt% La2O3 and 10wt% Y2O3 with respect to
the weight of the matrix powders. The powders were wet milled in alcohol for 24h, Si3N4
grinding media were used in all milling step. After drying, the obtained powders were sieved
to under 100 mesh and were molded in a BN-coated graphite die for sintering. Hot-pressing
was carried out at 1800 for 1h in a flowing nitrogen atmosphere under a pressure of 27.5 MPa
for all composites. As comparison, monolithic Si3N4 with the same sintering-aid was also
fabricated under the same conditions. Table 1 showed the contents of the composites.
Fig.1: The shape of Si-C-N nanometer particles
II - 676
Table 1: The contents of the composites (wt%)

Sample
Si-C-N
SiC
whiskers
Si3N4
La2O3
Y2O3
S1
---
---
80
10
10
S2
10
---
70
10
10
S3
20
---
60
10
10
S4
---
20
60
10
10
Measurements of Composite Properties

Hot-pressed plate samples were cut into bars and polished for measuring the mechanical
properties of the composites. Flexural strength were measured by three-point bending test, the
sample size was 3mm x 4mm x 35mm, the span was 30mm. Fracture toughness was measured
by a Single-End Straight-notch Beam (SENB) method, the sample size was 2.5mm x 2.5mm x
30mm, the notches were machined using a diamond blade 0.15mm wide deeply cut in the
center.
Fracture toughness was also measured by an indentation fracture (IF) method. A surface flaw
was made in the Well-polished specimens plane perpendicular to the hot-pressing axis by a
standard diamond Vickers indenter using a 19.6 KN load with a loading time of 20 seconds,
KIC values were calculated using the following equation:.
KIC=0.142(c/a)-1.56(E/H)2.5Ha1/2/
(1)
Where c is the radius of crack length, a is the impression radius, H is the hardness value
calculated from the diagonal length, E is the Youngs modulus of the composites, and is the
constraint factor ( 3). An average elastic modules of 310GPa was assumed for the
composites.
Density of the composites was measured by the Archimedes method using deionized water as
the immersion medium, and their relative densities were evaluated using the theoretical
densities calculated from simple mixing theory.
Results and Discussion
The Density and hardness of the composites reinforced with Si-C-N nanometer particles and
SiC whiskers are listed in Table 2. All of the composites had densities higher than 3.2 g/cm3,
indicationing that the composites were densified using Y2O3-La2O3 sintering additives. The
differences in relative density among these composites may be attributed to the presence of
the Si-C-N nanometer particles, the results showed that the obstacle role of Si-C-N nanometer
particles to the sintering of Si3N4 was more than that of SiC whiskers, and the sintering
condition of Si-C-N nanometer particle reinforced Si3N4 composite wasnt very suitable,
which should be studied further. The density and content of the composites will affect the
hardness, with the density increasing, the hardness increases; with the SiC adding in, the
hardness increases, it was confirmed by the values of hardness in Table 2.
II - 677
Table 2: The Density of the composites

Sample
Measured
density
(kg/cm3)
Calculated
Relative density
density (kg/cm3)
(%)
Hardness
(GPa)
S1
3.53
3.49
100
14.20
S2
3.41
3.43
99.42
14.17
S3
3.29
3.39
97.05
13.26
S4
3.42
3.39
100
15.57
The room-temperature flexural strength of the composites measured by three-point bending

are listed in Table 3, varied from 689.9~878.5MPa, the fracture toughness values measured by
both the SENB method and indentation fracture method are also listed in Table 3, the fracture
toughness measured using SENB method varied from 8.27~11.24MPam1/2. The result values
of the indentation technique were lower than those of the SENB values, but the rules were the
same. The fracture toughness showed strong dependence on the amount of the Si-C-N
particles, the 10 wt% Si-C-N particles had no significant effect on the fracture toughness. In
contrast, the 20wt% Si-C-N powder addition significantly improved toughness.
It is well known that the porosity of the composites will significantly affect the mechanical
properties of Si3N4 matrix, it was very interesting that the density of S3 (reinforced with
20wt% Si-C-N nanometer particles) was the lowest, the relative density was only 97.05%, and
the hardness was also the lowest, but the flexural strength and the fracture toughness values
were the highest, even more than S4 (reinforced by 20wt% SiC whiskers). In Ref.7, the
fracture toughness of SiC particle reinforced Si3N4 was improved significantly by the addition
of SiC particles, but the flexural strength of the composite was decreased by introducing SiC
particles. In this paper with the nano size Si-C-N adding, both of the toughness and the
flexural strength were increased. It showed that the presence of Si-C-N nanometer particles
make remarkable improvement of Si3N4 matrix, it revealed a bright prospect of
nanocomposites.
Table 3: The mechanical properties of the composites
Item
Sample
S1
S2
S3
S4
II - 678
Flexural strength
(MPa)
689.9
785.4
878.5
804.8
Fracture toughness (MPam1/2)

SENB
IF
8.36
7.87
8.27
7.72
11.96
8.85
11.24
8.78
Fig.2: The fracture surfaces of the composites (SEM)
The fracture surfaces of the composite examined by SEM are shown in Fig.2, It showed that
intergranular fracture and roughfracture surfaces developed in all cases. The photos
indicated that the morphology of Si3N4 grains in the nanocomposites was strongly influenced
by the Si-C-N dispersions, the growth of of elongated Si3N4 grains was accelerated by the SiC
dispersion, as compared with monolithic Si3N4, the nanocomposites were composed of more
uniform and homogeneous elongated Si3N4 grains. SiC whisker pull-out was observed in
Fig.2d and elongated - Si3N4 grain pull-out in other photos in Fig.2.
The micrographs of local intend areas examined by SEM are shown in Fig.3, it can be seen
that the crack length of S3 and S4 was shorter than that of S1 and S2, this showed more
evidence of why the fracture toughness of S3 and S4 were higher than that of S1 and S2.
The X-ray diffraction analyses of the composites revealed that the major phases were -Si3N4
and -SiC, Fig.4 shows the XRD parttern. Work is in process to characterize further the effect
of phase structure of Si-C-N nanometer particles reinforced Si3N4 composites on their
mechanical properties.
II - 679
Fig.3: The micrographs of local intend areas (SEM)

.
CONCLUSIONS
1.
Si-C-N nanometer particles has noticeable reinforcement to Si3N4 matrix, it can improve
both of the fracture toughness and flexural strength.
2.
The obstacle role of Si-C-N nanometer particles to the sintering of Si3N4 was more than
that of SiC whiskers.
3.
The fracture toughness showed strong dependence on the amount of the Si-C-N
nanometer particles.
4.
The reinforcing mechanisms acting in the Si-C-N nanometer particle reinforced Si3N4
composites were crack deflection and crack branching by Si-C-N particles.
II - 680
Fig.4: The XRD of the composites
REFERENCES
1. Buljan, S.T. and Zilberstein, G., Advanced Structure Ceramics, edited by Becher, P.F.,
Suain, M.V. and Somiya, S., (MRS, Pittsburgh, PA), 1987, pp.273-281.
2. Wei, G.C. and Becher, P.F., Development in SiC-Whisker-Reinforced Ceramics, Am.
Ceram. Soc. Bull., Vol.64, No.2, 1985, pp.298-304.
3. Lange, F.F, Effect of Microstructure on Strength of Si3N4-SiC Composite System, J.
Am. Ceram. Soc., Vol.56, No.9, 1973, pp.445-490.
4. Niihara, K., New Design Concept of Strutural Ceramics: Ceramic Nanocomposites, J.
Ceram. Soc. Jpn., Vol.99, No.10, 1991, pp.974-982.
5. Niihara, K., Suganuma, K., Nakahira, A. and Izaki, K., Interfaces in Si3N4-SiC Nanocomposite, J. Mater. Sci. Lett., Vol.9, No.10, 1990, pp.598-599.
6. Niihara, K., Izaki, K. and Nakahirm, A., The Silicon Nitride-Silicon Carbide
Nanocomposites with High Strength at Elevated Temperatures, J. Jpn. Soc. Powder
Metall., Vol.37, No.2, 1990, pp.352-356.
7. Hironori, K, Takaaaki, S., Hiroshi, S. and Tadahiko, M., Toughening of Silicon Nitride
Matrix Composites by the Addition of Both Silicon Carbide Whiskers and Silicon Carbide
Particles, J. Am. Ceram. Soc., Vol.73, No.3, 1990, pp.678-683.
II - 681
NON LINEAR BEHAVIOR OF TWO WOVEN

CERAMIC MATRIX COMPOSITES
UNDER CYCLIC LOADING
Jean-Marie Morvan and Stphane Baste
Universit Bordeaux I, C.N.R.S. URA 867, Laboratoire de Mcanique Physique,
351, Cours de la Libration, 33405 TALENCE Cedex, France.
SUMMARY: The non-linear mechanical behavior of both a 2D SiC-SiC and a 2D C-SiC

woven ceramic matrix composites is investigated under cyclic loading by means of a
metrology coupling an ultrasonic method and classical strain measurements. It appears that
the loss of stiffness during monotonous loading is mainly due to transverse microcracks
density and that the changes in elasticity during cyclic loading are the result of the
opening/closure of these cracks. In order to take this effect into account, an internal variable
that represents the damage accumulated along the loading is introduced. The non linear
behavior of both materials is predicted under both monotonous and cyclic loading. The
constitutive laws of the crack density and opening/closure variable describe the threedimensional changes in elasticity under cyclic loading as well as the inelastic strains along the
whole test.
KEYWORDS: microcracking, anisotropic damage, ceramic matrix composites, cyclic

loading, inelastic strains
INTRODUCTION
The non-linear behavior of ceramic matrix composites (CMC) has been extensively studied
under monotonous loading [1]. Concerning cyclic loading, the knowledge is mostly
concentrated on cyclic fatigue and frictional sliding determination. It is now commonly
recognize that the macroscopic behavior of these materials under tensile loading is the result
of the combination of three main damage mechanisms [2]: the matrix microcracking normal
to the tensile axis, the deflection of these cracks at the fiber-matrix interface if the interface is
weak enough and the fibers fracture. The matrix microcracking induces a loss of stiffness.
The Mode II cracking prevents the composite from failing too early. The fiber-matrix
debonding leads to a fiber-matrix sliding with friction depending on the nature of the
interface. The interfacial shear stress is thus of first importance in the global behavior. The
combination of these mechanisms leads to a highly non linear behavior.
The stress-strain curves of these materials clearly show that the global strain is the sum of an
elastic and an inelastic strain. The inelastic strain comes from both the opening of the
transverse cracks and the interfacial fiber-matrix sliding [3]. Nevertheless this sliding is
subordinate to the direction of the load. Assuming the damage is the relative variation of the
stiffness tensor [4], it appears that the damage does not increase, or slightly, during the
cycles. The value of damage reached during monotonous loading is the upper limit of
II - 682
damage during the cycles. When a cyclic loading is performed, as the stress decreases during
unloading, a sliding opposite to the one created by an increase in stress can occur. The
transverse cracks that have been created during the monotonous loading are then prone to
close (Fig. 1). They are still present anyway but their closure simulates an increase of
stiffness as if the material could recover its mechanical properties. The cracks that remain
open represent the active part of the microcracking. The sample exhibits an apparent state of
damage while the damage accumulated is greater.
Debonding crack
l /2
d
2U
Transverse
Matrix Crack
Matrix
Debonding crack
Transverse
Matrix Crack
Crack Opening
Displacement
Fibre
Matrix
Fibre
2R
(a)
(b)
Fig. 1: States of the transverse cracks: (a) Open crack, (b) Close crack
In order to separate and to identify accurately the effects of the initiation and growth of
matrix microcracks under tensile loading as well as the effect of these cracks under cyclic
loading on both the stiffness and the inelastic strains, it is necessary to perform a strain
partition under load.
FORMULATION OF THE MODEL

With the cracked media replaced by an effective equivalent medium, it is possible to deduce
the variations of the elastic properties of the cracked material. By modeling the cracks as
2
2
2
2
elliptic cylinders; x1 a + x2 b = 1 , x 3 < , the effective elastic properties of a fibrous
composite containing cracks are [5]:
S = S0 +
(1)
where S is the compliances tensor of the material, S0 is the compliances tensor of the
uncracked material, is a fourth order tensor which terms depend upon both the cracks
whose aspect ratio =b/a tends to zero and the mechanical properties of the medium that
surrounds them and is the crack density. Cracks thickness 2b in Direction 3 is considered as
negligible. Length 2c in Direction 2 is higher than depth 2a in Direction 1.
Thus, the damage due to a single damage orientation can be described with one parameter
only: the crack density. This formulation was written for a single crack system. Actually the
degradation process exhibits several specific orientations. It is very common to observe three
II - 683
orthogonal crack systems in CMC; a transverse one due to matrix cracking, a longitudinal one
due to fiber-matrix debonding as well as an interlaminar one when delamination occurs [6]. It
is also important to keep in mind that these systems might interact. For the transverse crack
system, normal to the tensile axis, only three components of are different from zero [7]:
T33, T44 and T55. As the cracks are deviated in mode II in the fiber-matrix interphase, a
longitudinal crack system with crack density L must be taken into account. For this
longitudinal crack system, parallel to the tensile axis, the non zero components of the crack
tensor L are obtained by a simple index transformation: L22, L44 and L66. When
delamination occurs, the interstacks cracks that appear are modeled by a family of slit cracks
parallel to Plane (2, 3) and normal to Axis 1. The crack density is L. We then have for the
crack tensor I: I11, I55 and I66.
As a proportion of the transverse cracks can close during unloading, it is necessary to
differentiate an apparent state of cracking and the number of cracks that have been created
effectively [8]. The closing effect induces a damage deactivation leading to unilateral
behavior [9]. Obviously, the crack density parameter is very sensitive to the effect of the
opening/closure of the cracks. Just as the active cracks were defined, we can define an active
crack density:
= F.
(2)
where F is the proportion of active cracks and * is the density of cracks that have been
accumulated. F represents the proportion of cracks that remains open during unloading.
Taking into account the phenomenon of opening-closure of the cracks, assuming that the
effects of the various cracks systems are additive and adding a corrective term that takes into
account their interaction, the compliance tensor of the cracked material becomes:
S = S0 +
T
FT . T +
L
FL . L +
I
interaction
FI. I + S
(3)
Thus, the global behavior depends upon six functions: *i and Fi,(i=T,L,I). The compliances
variations are then simply a function of the various cracks systems that appear through the
variations of the crack densities and of the mechanical properties of both the uncracked and
damaged material. When cyclic loading is performed, the opening/closure functions Fi
modulate the compliances variations.
The inelastic strains represent a large part of the total strains in CMC. It is now commonly
recognized that they mainly come from the crack opening displacements, in other words, the
cracks thicknesses. In an extensometer length L, if there are n cracks, 2a deep and 2b thick,
the distance between them is 2a/*T. So:
*
n = L T 2a
(4)
and then:
inelastic
II - 684
inelastic
L = n U L = n 2b L = *
T
where
= 2 b 2a
(5)
During unloading, the effect of closure of the cracks must be taken into account. It is clear as
well that some cracks will not close completely, some have a residual thickness. This
thickness is most probably due to the roughness of the cracks edges coming from both wear
debris at the sliding interfaces and wedging of the cracks by grain bridging. That is why
another aspect ratio ' that takes this new thickness into account is introduced. Finally, the
variations of inelastic strains during the whole test can be described by:
inelastic
*
= T ( F( ) + )
(6)
The inelastic strains are thus a function of the transverse crack density and of the aspect ratio
of the cracks whether they are completely open or not. As for the compliances, the
opening/closure function modulates the inelastic strains variations during cyclic loading.
The opening of the cracks is strongly correlated to the sliding stress that takes place at the
fiber/matrix interface. The measurement of the inelastic strains and the identification of the
transverse crack density had lead us, with the help of a modified shear lag model [10], to
establish a new formulation to identify the interfacial sliding stress :
(7)
Eqn (7) simply relates to the radius of the fibers R, their Young modulus, Ef, the distance
between the cracks, l and the length along which the interface remains bonded, b (Fig. 1 (a)).
In order to study accurately the damage evolution in composites, an experimental device that
couples an ultrasonic immersion tank associated to a tensile machine and an extensometer has
been developed. It allows to study the loss of stiffness of the material under tensile stress
because it makes it possible to perform under load the angular investigation necessary to
identify the elastic tensor variation. As it gives access to the complete stiffness tensor, it
becomes possible to know which coefficients are affected during the damage [11]. It also
allows to know precisely which proportion of the global strain is either elastic or inelastic
[12].
45
Extensometer
Immersed Transducer
2
Contact Transducer
(C 44 )
Fig. 2: Under load strain partition equipped sample
II - 685
The characterization of anisotropic materials using an ultrasonic method gives access to the
purely elastic part of their behavior. The main principles of ultrasonic evaluation have been
given by Roux [13] for the elastic coefficients evaluation of homogeneous anisotropic
materials. The nine elastic constants Cij which fully describe the elastic behavior of an
orthotropic material are identified by the wave propagation velocities in three data planes:
Plane (1, 2), Plane (1, 3) and Plane (1, 45) described by the bisectrix of axis 2 and 3 (Fig. 2)
[14].
The two composites studied, a 2D SiC-SiC and a 2D C-SiC, are of woven type. They are
manufactured from preforms built up from multiple layers of respectively Silicon Carbide
and Carbon cloth by S.E.P. (Socit Europenne de Propulsion, France). The SiC matrix was
in both cases added by chemical vapor infiltration. Under tensile stress, they exhibit rapidly a
non linear behavior mainly due to the matrix microcracking. Because of the woven nature of
the samples, Directions 2 et 3 are initially symmetrically equivalent. Both specimens are thin
plate shaped, 3 mm in thickness and 2.7 g.cm-3 in density. They were submitted to tensile
stress in Direction 3 parallel to one of the bundle direction.
The stress-stain curves of the samples are shown on Fig. 3 and 4. Although the strain to
failure is about 0.8 per cent for the two composites studied, they reach this value in a different
manner. The 2D C-SiC exhibits quite immediately a non linear behavior whereas the 2D SiCSiC remains elastic until about 80 MPa as the linear variation of the strain testifies. About the
cycles performed, the lack of hysteresis in the 2D C-SiC curve is relevant of the fact that the
sliding occurs with little or no friction in the fiber-matrix interface whereas the large loops
with many changes in slopes that appear in the curve of the 2D SiC-SiC are relevant of the
interfacial debonding extent and emphasize the presence of sliding with friction that needs a
threshold of stress to occur. But in both cases, the permanent strains are far from negligible. It
is noteworthy that, until 120 MPa, the stress-strain curve of the 2D SiC-SiC shows that the
strain increases during the steps necessary for the ultrasonic evaluation while the loading
device is constant stress controlled.
250
350
200
300
250
150
ss (MPa)
100
200
ss (MPa)
150
100
50
0
Total Strain
(%)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Fig. 3: Stress-Strain Curve of a 2D SiC-SiC
Total Strain
(%)
50
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Fig. 4: Stress-Strain Curve of a 2D C-SiC
The accuracy and the reliability of the complete determination of the stiffness tensor together
with its variation during the test allow, by inverting this tensor, to get the tensor of elastic
compliances and its own variation. The variation of all the components of the compliance
tensor of the 2D SiC-SiC with their confidence interval is shown in Fig. 5. The compliances
variations exhibit three domains. Until 70 MPa, damage threshold of this composite, the
various compliances remain the same. From then, the matrix microcracking begins. The
compliances increase along the tensile axis is very large; S33 increases more than 300%. This
represents the increase in compliance due to the inter-bundles matrix cracks until saturation at
II - 686
about 120 MPa. Then, this variation becomes slighter as the matrix microcracking spreads
inside the bundles. The shear moduli related to the planes that include the tensile axis, S 44 and
S55, also increase a lot. The microcracking affects more slightly the elastic coefficients in
Plane (1, 2) normal to the cloth plane.
Fig. 6 shows the variations during the cycles of the compliances of the 2D C-SiC that were
the most influenced by the damage with their relative confidence intervals. The cycles are
shifted for clarity. S33, S44 and S55 relative to the transverse cracks system are the only
compliances to exhibit a variation during the cycles. S11 does not vary at all during the whole
test and nor S66 nor the off diagonal compliances, S12, S13 and S23. S22 varies during the
monotonous loading but do not change along the cycles. The progressive variation of S 33, S44
and S55 during the cycles points out a progressive closure of the cracks. As the cracks close,
they are active no more and they lose the effect they had on the stiffness constants. When the
sample is reloaded, the cracks re-open, become active again and every compliance reaches the
value it has before unloading.
The elastic strains are obtained with the generalized Hooke's law:
1
elastic
i
= ( C ij ) j = Sij j
(8)
II - 687
Therefore the elastic strain along the tensile axis is simply S33 times the applied stress. As an
extensometer indicates the total strain, the inelastic strain is then obtained simply:
II - 688
inelastic
total
elastic
(9)
Fig. 7 and 8 show respectively the relative part of the elastic and inelastic strains on the total
strain during the monotonous loading of the 2D SiC-SiC and of the 2D C-SiC. Fig. 8 also
shows what happen during the cycles in the 2D C-SiC. In both cases, the inelastic strains are
far from negligible. Unlike the 2D C-SiC, the strain partition of the 2D SiC-SiC exhibits three
zones during monotonous loading. If for the 2D SiC-SiC, the inelastic strains do not appear
until the beginning of the inter-bundles matrix microcracking, it is clear that they are present
from the very beginning of the test in the case of the 2D C-SiC. There is no damage threshold
for the 2D C-SiC whereas the behavior remains linear elastic until about 80 MPa for the 2D
SiC-SiC.
The inelastic strains of the 2D SiC-SiC exhibit two very different increases that correspond to
the scale at which they occur; an increment of strain at constant stress bs which starts
around 80 MPa and stops at about 140 MPa due to inter bundles microcracking and a strain fs
which needs an increase in stress and which begins at 120 MPa when the cracking starts to
occur at the fibers scale inside the bundles. The strain partition on the cycles of the 2D C-SiC
indicates that the linear variation observed is the sum of the non linear variation of both the
elastic strain and inelastic strains. As it could have been awaited for, the elastic strain comes
back to zero when no stress is applied whereas the inelastic strains are still present in the form
of residual strains.
This emphasizes that the strain measured when the sample is fully unloaded represents only
the residual part of the inelastic strain because the reversible sliding has closed some of the
cracks. It is noteworthy that the residual strain value at the unloading point is far smaller than
the under load inelastic strain. This establishes the partly reversible nature of the inelastic
strain and therefore the necessity of doing the strain partition under load.
elastic
elastic +
fs
total
250
200
ss (MPa)
150
bs
100
50
Strain (%)
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Fig. 7: Under load strain partition of a 2D SiC-SiC

inelastic
elastic
350
total
Strain (%)
280
210
ss (MPa)
140
70
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Fig. 8: Under load strain partition of the 2D C-SiC with a cyclic loading at 280 MPa
II - 689
IDENTIFICATION OF THE MODEL

The experimental variations of the elastic constants and Eqn (1) give access to the
constitutive laws of the cracks density, Fig. 9 and 10. Only the monotonous loading is
necessary to identify the crack densities. From the variations of the compliances of both
samples, it results that every cracks density variation exhibits a linear behavior, the transverse
variation having the greatest slope whatever the material and the scale considered. The
interlaminar crack density I of the 2D C-SiC does not vary during the test meaning that no or
very small delamination occur between the stacks for this sample. The fact that the two
longitudinal cracks systems display a more stochastic behavior indicates that these damage
mechanisms may be more influenced by local variations of the composite microstructure.
About the 2D SiC-SiC, up to 80 MPa, as the damage has not begun, the crack densities are
equal to zero. Then, at the inter-bundle scale, a 10 MPa delay necessary for the mode II
cracking to start is highlighted.
During monotonous loading, all the transverse cracks are open. So the opening/closure
function F is equal to 1. Eqn (2) reduces to:
T = T
(10)
The comparison between the experimental data and the predicted variations are depicted in
Fig. 5 and 6. For the 2D SiC-SiC, a good correlation is obtained for the first three diagonal
compliances, S11, S22 and S33. The three remaining ones, S44, S55 and S66, exhibit a more or less
large deviation between the predictions and the experimental results. Anyway, concerning the
inter-bundle part of the cracking, up to about 120 MPa, the predictions for S44 and S55 are
good. For S66, the deviation probably indicates that the transverse cracks thickness must be
taken into account in the model. In the intra-bundle part of the prediction of the shear moduli,
the deviations might be explained this way: first, as the crack systems are not independent,
they might interact and second, one of the hypothesis of the model is that the cracks are
elliptic, but, actually, at the intra-bundle scale they are more likely to be circular because their
extent is limited by the fibers arrangement. This is emphasized by the similar variation of S44
and S55 in the intra-bundle part. Finally, concerning the off-diagonal compliances, the remark
about taking into account the transverse cracks thickness is also valid to improve the
prediction. For the 2D C-SiC, the comparison between the experimental data and the
predicted variations shows a good correlation for the monotonous loading represented by the
top of the cycles.
Clearly, during loading-unloading cycles, only the mechanism of the opening/closure of the
transverse microcracks will take place. The state of the longitudinal and interstacks cracks
does not, or slightly, vary. Thus the model of compliance variation can be simplified:
FL = 1 _ L = L
II - 690
and
FI = 1 _ I = I
(11)
At the top stress of a cycle, max, all the cracks are open. We get the maximum of active
cracks and thus, F=1. At zero stress, there is still a non zero proportion of open cracks leading
to residual strains. F is minimum but different from zero as all the cracks do not close.
On Fig. 11 is represented the opening/closure variable F of the 2D SiC-SiC. This step
function is representative of the existence of the threshold of stress needed to begin the
fiber/matrix sliding at the interface. Fig. 12 shows the evolution of F during the various cycles
of the 2D C-SiC. The proportion of open cracks can be described by a unique scalar function
describing the progressive opening /closure of the transverse cracks:
F = 1 ((1 r ) (1 + r ))
where
r = (Y 0 + Y ) (Y0 + 1)
and
Y = max
(12)
where Y 0 = - 0.25 is the threshold where all the cracks close and m =1.2 a constant.
So, for this material, the behavior during the cycles is fully described by four functions:*T,
L, I and F that can predict the compliances variations (Fig. 6).
The inelastic strains prediction (Fig. 13) during both monotonous and cyclic loading is rather
good as well once the residual opening of the cracks of the 2D C-SiC has been taken into
account ( Eqn (6)). Fig. 14 describes the evolution of the inelastic strains for the 2D SiC-SiC
at the various scales of its behavior as a function of the two transverse cracks densities and
the aspect ratio using Eqn (5). A 10 MPa delay for the sliding to begin can clearly be seen.
However, there is a gap between the experimental data and the prediction at the end of the
inter-bundles phenomenon because the two slidings coexist in a 40 MPa range.
About the 2D C-SiC, micrographs have indicated that the material contains a lot of porosities
and that many transverse cracks are situated in the SiC mantel around the bundles. The
opening/closure of these cracks certainly occurs without any friction because the opening
takes place on the pore side. This can explain the lack of hysteresis during the cycles. The
interfacial shear stress of the 2D SiC-SiC has been identified to be about 27 MPa using Eqn
(7) at the inter bundle scale and much higher at the fibers scale. This is in line with the
assumption of a strong cohesive zone inside the bundles.
II - 691
Fig. 11: Opening-closure variable for transverse microcracks of a 2D SiC-SiC
CONCLUSION
The behavior of ceramic matrix composites is strongly influenced by matrix microcracking.
The ultrasonic characterization through the complete determination of the stiffness tensor
along the whole test can detect all the damage mechanisms of CMC: the transverse matrix
microcracking as well as the presence of longitudinal cracks at the fiber-matrix interface.
Thus this technique is of great help in measuring the anisotropic damage and it also allows to
perform the strain partition under load which is the only way to separate the various damage
mechanisms. This technique gives a correct estimation of both the inelastic strains and the
elastic modulus drop. The problem of the three-dimensional description of the non linear
behavior of ceramic matrix composites can be solved by the measurement of the cracks
II - 692
density parameters and the closing-opening variable. Their knowledge can fully predict the
loss of stiffness due to matrix microcracking responsible for the non linear behavior of CMC.
These parameters, representative of the initiation and growth of the cracks, describe the
elastic part of the behavior. Furthermore, with the estimation of the cracks aspect ratio, the
inelastic strains can be predicted accurately.
REFERENCES
1. Evans A. G. and Zok F. W., "The physics and mechanics of fiber-reinforced brittle matrix
composites", J. Mat. Sci., Vol. 29, 1994, pp. 3857-3896.
2. Marshall D. B. and Evans A. G., "Failure mechanisms in ceramic-fiber/ceramic-matrix
composites", J. Am. Ceram. Soc., Vol. 68, No. 5, 1985, pp. 225-231.
3. He M. Y., Wu B.-X., Evans A. G. and Hutchinson J. W., "Inelastic strains due to matrix
cracking in unidirectional fiber-reinforced composites", Mech. Mater., Vol. 18, 1994, pp.
213-229.
4. Baste S. and Audoin B., "On internal variables in anisotropic damage", Eur. J. Mech.,
A/Solids, Vol. 10, No. 6, 1991, pp. 587-606.
5. Laws N., Dvorak G.J. and Hejazi M., "Stiffness changes in unidirectional composites
caused by crack systems", Mech. Mater., Vol. 2, 1983, pp. 123-137.
6. Camus G., Guillaumat L. and Baste S., "Evolution of Damage in a 2D Woven C/SiC
Composite under Mechanical Loading. Part I: Mechanical Characterization", C. Sci. and
Tech., to be published.
7. Laws N., "A note on interaction energies associated with cracks in anisotropic solids",
Phil. Mag., Vol. 36, 1977, pp. 367-372.
8. Baste S., El Guerjouma R. and Audoin B., "Effect of Microcracking on the Macroscopic
Behavior of Ceramic Matrix Composites, Ultrasonic Evaluation of Anisotropic Damage",
Mech. Mater., Vol. 14, 1992, pp. 15-31.
9. Chaboche J.-L., "A Continuum Damage Theory with Anisotropic and Unilateral
Damage", La Recherche Arospatiale, Vol. 2, 1995, pp. 139-147.
10. Danchaivijit S. and Shetty D. K., "Matrix cracking in ceramic matrix composites", J. Am.
Ceram. Soc., Vol. 76, No. 10, 1993, pp. 2497-2504.
11. Audoin B. and Baste S., "Ultrasonic Evaluation of Stiffness Tensor Changes and
Associated Anisotropic Damage in a Ceramic Matrix Composite", J. Apple. Mech., Vol.
61, 1994, pp. 309-316.
12. Baste S. and Morvan J.-M., "Under Load Strain Partition of a Ceramic Matrix Composite
Using an Ultrasonic Method", Experimental Mechanics, Vol. 36, No. 2, 1996, pp. 428434.
13. Roux J., "Elastic wave propagation in anisotropic materials", Proceedings IEEE 1990
Ultrasonics Symposium, Honolulu, Dec. 1990, pp. 1065-1073.
14. Baste S. and Hosten B., "valuation de la matrice d'lasticit des composites orthotropes
par propagation ultrasonore en dehors des plans principaux de symtrie", Rev. Phys.
Appl., Vol. 25, 1989, pp. 161-168.
II - 693
FRACTURE IN OFF-AXIS AND ANGLE-PLY FIBER

REINFORCED CERAMICS
Alex S. Selvarathinam & Y.J. Weitsman
Department of Mechanical & Aerospace engineering & Engineering Science
307 Perkins Hall, University of Tennessee, Knoxville, Tennessee-37996-2030
KEYWORDS: ceramic composites, fracture, oblique reinforecement
II - 694
II - 695
II - 696
II - 697
II - 698
II - 699
II - 700
II - 701
II - 702
THE RESEARCH OF SELF-REINFORCED HP

SILICON NITRIDE
Li Fengmei
Institute of Aeronautical Materials, Beijing 100095, China
SUMMARY: For the self reinforced HP Si3N4 ceramics the effects of (Y2O3 + La2O3) wt%
and the ratio between Y2O3 and La2O3 on the mechanical properties and microstructure were
studied. While Y2O3 : La2O3 = 1 : 1 (wt%) the flexure strengths and fracture toughness reach
to a maximum at same time and the content of (Y2O3 + La2O3) approaches 20 wt% the flexure
strength goes through a maximum 715 MPa (at 1350 C), the content of (Y2O3 + La2O3)
approaches 16 wt%, the fracture toughness goes through a maximum, 7.8 MPaM1/2 (at room
temperature). Analysis of toughening processes indicated that both crack deflection and
Si3N4 bridging or pullout mechanisms can contribute to the fracture toughness of HPSi 3N4
ceramics.
KEYWORDS: self reinforce, silicon nitride, mechanical properties, microstructure
INTRODUCTION
Silicon nitride has the potential to become the material of choice for advanced application
because of a combination of excellent mechanical and thermal properties. In general, fibers or
whiskers are introduced to overcome the brittleness of unique ceramics, but there are several
processing problems eliminating its application. For example, the component of different
shape is difficult to well distributed. Selfreinforced silicon nitride or in situ toughened Si3N4
is based upon microstructural engineering through the controlled nucleation and growth of the
Si3N4 grains. During sintering, Si3N4 grains with high aspect ratios were formed, therefore,
the results are obtain similar to the Si3N4 toughened by fibers or whisker. Silicon nitride is a
highly covalent material that requires the use of sintering additives to reach full densification.
During sintering silica on the silicon nitride surface react with sintering additives to a glass.
At high temperatures, the Si3N4 dissolves into the glass and precipitates in the form of
Si3N4 . The morphology of the Si3N4 can vary from equiaxied to highly elongated grains
depending on the characteristics of the oxynitrid glass as well as other factors such as Si3N4
powder, sintering additives, and processing condition. In this paper , the effects of Y2O3 and
La2O3 sintering additives on the mechanical properties and microstructure of Selfreinforced
HP silicon nitride are reviewed.
The characteristics of used Si3N4 powder are shown in Table 1. Sintering additives are Y2O3
and La2O3, its pureness is 99.9%. The preparation process of Si3N4 ceramic parts are shown in
Figure 1. The trial compositions (Si3N4, Y2O3 and La2O3) were mixed. Mixing time was 24 h
II - 703
in the polyethylene bottle with HPSN milling medium and pure ethanol as mixing fluid. After
mixing, the ethanol was evaporated from the powder slurries. The powders were dried further
in a air oven at 110 C and then sieved through a 100 mesh stainless steel screen . Composite
bars were obtained by hot pressing the mixture in a graphite die at 1780 C, 25 MPa applied
pressure in a nitrogen atmosphere, in 1 h. The initial experimental work of this study involved
the determination of the content and ratio of sintering additives. For this, ten trial
compositions were chosen Table 2. The composite blocks were loaded in three point bend at
1350 C and room temperature at a constant displacement rate of 0.5 mm/min in air.
Table 1: Characteristics of Si3N4 raw power
/(+)
%
90
Diameter
m
0.3
BET
m2/g
20
Fe
0.16
Impurity
Al
0.07
wt%
Si
0.6
W
0.11

Tow toughening mechanisms are shown in Fig. 2, Fig. 3, grain bridging or pullout and crack
deflecting. As this bridge zone develops, a closure stress is placed on the propagating crack
that effectively increased the fracture resistance of the material. this increased fracture
resistance is result of crack bridging and pullout (Fig. 2). As the propagating crack collides
with elongated grains, the crack would be deflected (Fig. 3). The extremely high fracture
toughness is produced by an interaction between the propagating crack and the elongated
grains in the material. Many investigators have demonstrated that the fracture toughness of
Si3N4 is strongly dependent upon grains morphology. In materials containing elongated grains
with high aspect ratios and large diameters, fracture toughness can reach higher values.
Mitomo et al[1] and Matsuhiro at al[2] have shown that the fracture toughness of Si3N4 depends
on the size and fraction of these elongated grains. Figure 4 summarizes the relationship
II - 704
between fracture toughness and grain size for Y2O3 - Al2O3 - Si3N4 and Y2O3 MgO - Si3N4
systems processed at different temperature. Figure 5 summarizes the relationship between the
relative fracture toughness and the percentage of elongated grains.
Fig. 4: The relationship between fracture toughness and the diameter of elongated Si3N4
grain
Fig. 5: The relationship between the relative fracture Fig. 6: Y2O3 -La2O3 binary diagram
toughness and the percentage of elongated Si3N4 grains
The selection of sintering additives not only is a density criterion, but also is a key parameter
in controlling the morphology of silicon nitride grains. The chemical composition of the glass
determines the temperature and the rate of the to transformation. The glass composition
also determines the rate of mass transport, which is directly related to the kinetics of grain
growth. The effects of the ratio between Y2O3 and La2O3 on the HP Si3N4 mechanical
properties are shown in Fig. 7, while Y2O3 : La2O3 = 1 : 1 (wt%), the flexure strengths and
fracture toughness reach to a maximum at same time, flexure strength approaches 676 MPa
(at 1350 C), fracture toughness is 6.13 MPa.M1/2 (at room temperature). From Y2O3 and
La2O3 binary diagram (Fig. 6), we know, while Y2O3 : La2O3 is close to 1 : 1, compound
II - 705
(LaYO3) was forming easy, sintering additives were at grain boundary in the compounds, not
in glass, therefore, raising the softening temperature of glass and improving the Si3N4
mechanical properties.
Table 2: Effects of the content of (Y2O3 + La2O3 ) wt% and the ratio between Y2O3 and La2O3
on the HP Si3N4 mechanical properties
Si3N4
87
87
87
87
87
95
90
84
80
75
Composition
Y2O3
0
4
6.5
9
13
2.5
5
8
10
12.5
( wt % )
La2O3
13
9
6.5
4
0
2.5
5
8
10
12.5
fRT
Y2O3:La2O3 (MPa)
0 : 13
586.8
4:9
773.0
1:1
581.1
9:4
365.7
13 : 0
703.0
1:1
413.5
1:1
660.0
1:1
880.9
1:1
890.7
1:1
801.4
f1350
(MPa)
447.3
669.4
676.6
589.4
128.0
212.0
380.4
686.1
715.6
595.04
K1cRT
(MPa.M1/2)
5.78
5.49
6.13
5.82
5.05
4.60
5.67
7.78
7.68
7.93
Fig. 7: The relationship between the ratio of sintering additives ( Y2O3 : La2O3 ) and Si3N4
mechanical properties
The effects of the content of sintering additives (Y2O3 + La2O3) wt% on the HP Si3N4
mechanical properties and microstructure are shown in Fig. 8 and Fig. 9. While the content of
sintering additives (Y2O3 + La2O3) approaches 16 wt%, the fracture toughness goes through a
maximum, 7.8 MPa.M1/2 (at room temperature). While the content of sintering additives
(Y2O3 + La2O3) approaches 20 wt%, the flexure strength goes through a maximum, 715 MPa
(at 1350 C). Fig. 9 is scanning electron micrographs of a self-reinforce Si3N4. Fig. 9
illustrates how the microstructure can be varied by the content of sintering additives when
materials are made from the same starting powder and processed under identical conditions.
In these materials, the grains with highest aspect ratios have been produced in the material
containing 16-20 wt% (Y2O3 + La2O3) sintering additives. The materials made with 5 or 25
wt% sintering additives (Y2O3 + La2O3) contained mainly a large number of fine grains.
These results clearly show that the grain morphology is strongly influenced by the content of
sintering additives. The main benefit of controlling the formation of elongated grains is a
dramatic improvement in the mechanical properties of silicon nitride. The improved fracture
resistance and mechanical performance is the result of a reinforcing phenomenon from the
whiskerlike grains, similar to the behavior observed in whisker-reinforced ceramics[3].
II - 706
Fig. 8: The relationship between the content of sintering additives (Y2O3 + La2O3 ) and HP
Si3N4 mechanical properties
Fig. 9: SEM morphology of fracture surface of different content of sintering additives

(Y2O3 + La2O3) for HP Si3N4
CONCLUSIONS
Toughening mechanisms of self-reinforced Si3N4 are crack bridging, grain pullout and crack
deflecting. In materials containing elongated grains with high aspect ratios and large
diameters, fracture toughness can reach higher values.
While Y2O3 : La2O3 close to 1 : 1 , the compound (LaYO3) was forming easy, therefore, the
mechanical properties were improved. While Y2O3 : La2O3 =1 : 1, the flexure strength of the
II - 707
material made with 20 wt% (Y2O3 + La2O3 ) sintering additives goes through a maximum,
715 MPa (at 1350 (C), the fracture toughness of the material made with 16 wt% (Y2O3 +
La2O3) reaches a maximum, 7.8 MPa.M1/2 (at room temperature).
REFERENCES
1. Mitomo M, Uenosono S. J. Am. Ceram. Soc. 1992, 751038
2. Matsuhiro K, Takashashi T. Ceram. Eng. Sci. Proc. 1989, 1080716
3. Becher P F. J. Am. Ceram. Soc. 1991, 74225569
II - 708
MICROSTRUCTURAL STUDY OF THE SINTERED

COMPOSITE COMPOSED OF TIN AND NI
Shujie Li 1,2,3, Yuping Li 3,4, J. H. Maas 3, J. Boeijsma 3 and B. H. Kolster 2,5
1
Currently: Beijing University of Aeronautics and Astronautics, Department of Materials

Science and Engineering, Beijing 100083, P.R.China
2
Former Foundation for Advanced Metals Science, The Netherlands
3
University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
4
Currently: College of Traditional Chinese Medicine and Pharmacology,
Division of Analytical Chemistry, Beijing 100013, P.R.China
5
Currently: PIM-Tech b.v., P.O. Box 192, 7360 AD Beekbergen, The Netherlands
SUMMARY: Pure TiN powder was mixed with pure Ni powder at a volume ratio of 1:1.
Then, the mixed powder was cold compacted. The compacts were pressureless sintered at
1150, 1250 and 1350 C respectively for 1 hour in vacuum. X-ray diffraction analysis and
microstructural analysis by SEM and EDX were carried out with the sintered samples. It is
shown that the lattice constant of Ni increases with increasing sintering temperature, but the
crystal structure of TiN is stable under the experimental conditions. The Ti content of the Ni
solid solution determined from the lattice constant value is very close to the one obtained by
EDX quantitative analysis. The experimental results indicate that interdiffusion takes place at
the TiN/Ni interface during sintering.
KEYWORDS: ceramic matrix composite, cermet, interface
INTRODUCTION
Various kinds of cermets are widely employed in industry due to their favorable properties
which are different from those of ceramics and metals. As one of their representatives, the
cermet composed of WC and Co has been used as a wear-resistant material for several
decades. However, this cermet is expensive because both tungsten and cobalt are not abundant
materials. Furthermore, the high specific gravity of this material makes the parts too heavy to
use in some cases especially for aeronautical and astronautical applications. Hence,during the
last 30 years many attempts have been made to find alternatives with low specificgravity and
relatively comparable mechanical properties coupled with favorable cost [1-9]. Although
much significant progress has been reported [1-9], the cermet composed of WC and Co has
not been replaced in industry so far. Nevertheless, the effort to look for the alternatives has
not stopped.
As reported in Ref. 10-12, in order to achieve this goal, Al2O3 and Ni were selected as the
starting materials due to their favorable properties and easy availability. However, the
problem associated with these two components is the poor wettability between solid Al2O3
and liquid Ni [13], which hinders the manufacturing of this cermet by liquid phase sintering.
In order to improve the sintering ability of this material, attention was focused on introducing
interlayers which could adhere strongly to both the ceramic phase and the metallic phase. It is
II - 709
reported that TiN was found to be the best suited to achieve this goal. The TiN interlayer was
introduced by coating the Al2O3 powder with a layer of TiN using CVD process. Then, Al2O3
powder coated with TiN (denoted by Al2O3-TiN) was mixed with Ni powder. The mixture
was hot-pressed to manufacture the composite, which is expressed by Al2O3-TiN+Ni [10-12].
In order to investigate the mechanism that TiN interlayer remarkably enhances the sintering
ability and mechanical properties of the material, it is essential to study the two interfaces
(TiN/Ni and Al2O3/TiN) involved in this system. This paper deals with the TiN/Ni interface.
Pure TiN powder (particle size: 5-25m, purity: 99.9 Wt%) was mixed with pure Ni powder
(particle size: 1-3m, purity: 99.9 Wt%) at a volume ratio of 1:1 for 45 minutes using a
powder mill. The powder was loaded in a plastic cylindrical vessel containing TiN balls 10
mm in diameter for mixing. The mixed powder, 1 gram for each sample, was cold compacted
at 750 MPa for 60 seconds. Neither lubricant nor binder was used during compacting. The
compacts were pressureless sintered at 1150, 1250 and 1350 C respectively for 1 hour in
vacuum of 710-3 Pa (510-5 Torr). After sintering for the required duration, the samples were
left in the furnace in vacuum to cool gradually to room temperature.
The sintered billets were ground lightly with 1000 mesh sand paper to remove the surface
layer. Then, they were washed in alcohol using an ultrasonic bath. After drying, X-ray
diffraction was carried out on all these samples. The polished sections were observed by SEM
and the composition of the various phases at the sections was analyzed qualitatively and
quantitatively by EDX (Energy Dispersive X-ray Microanalysis System).

Fig. 1 shows the X-ray diffraction patterns of the samples sintered at the various temperatures.
For comparison, the X-ray diffraction pattern of the compact with the same composition is
also included in the figure. This figure indicates that with increasing sintering temperature
every diffraction peak of Ni shifts to the left side of the figure (low angle), whereas, the
diffraction peaks of TiN do not shift. So, according to Bragg's law, the lattice constant a of Ni
increases with increasing sintering temperature. Its value as a function of sintering
temperature is presented in Fig. 2.
II - 710
II - 711
Fig. 2: Lattice constant a of the Ni solid solution as a function

of sintering temperature (For pure Ni: a=3.5238)
A conceivable explanation for the expansion is that dissociation of TiN takes place during
sintering [14] resulting in the formation of Ti atoms (or Ti ions), then the Ti atoms diffuse
into Ni lattices leading to the expansion, as the atomic radius of Ti (2.00 ) is larger than that
of Ni (1.62 ). The higher the sintering temperature, the more Ti atoms diffuse into Ni
lattices, consequently, the larger the lattice constant value will be. Fig. 1 also shows that there
is no diffraction peak broadening meaning that the Ni solid solution is homogeneous, which is
not surprising for the high temperatures and long duration of sintering applied and the small
particle size of Ni.
The quantitative relationship between the lattice constant of Ni solid solution and its Ti
content is given in Fig. 3 [15]. Using this figure the Ti content of the Ni solid solutions can be
obtained as illustrated in Fig. 4, which shows the Ti content as a function of sintering
temperature acquired from the data presented in Fig. 2. From Fig. 4, one can see that the
amount of Ti diffused into Ni lattices is significant.
Fig. 5 shows the scanning electron micrographs of the cermet pressureless sintered at 1350 C
for 1 hour. Two phases are visible in the micrographs. Various areas from each phase were
analyzed by EDX. A representative EDX spectrum for the metallic phase corresponding to the
area indicated by the arrow in Fig. 5A is presented in Fig. 6. This figure confirms that the
metallic phase consists of Ni and Ti. The quantitative EDX analysis gives the Ti content as
11.1 0.1 At% (The uncertain range is standard deviation). This value is very close to the one
given in Fig. 4. A representative EDX spectrum for the ceramic phase corresponding to the
area indicated by the arrow in Fig. 5B is presented in Fig. 7. This figure indicates that the
ceramic phase consists of Ti, N and a small amount of Ni as well. Quantitative EDX analysis
shows that the Ni content is below 6 At%.
II - 712
Fig. 3: Lattice constant a of the Ni solid solution as a function of Ti content
Fig. 4: Ti content of the Ni solid solution as a function of sintering temperature
Fig. 5: Scanning electron micrographs of the cermet composed of 50 Vol% TiN + 50 Vol% Ni
pressureless sintered at 1350 C in vacuum for 1 hour. The metallic phase (A) and the
ceramic phase (B) were analyzed by EDX
II - 713
Fig. 6: EDX spectrum analyzed at the metallic phase

as indicated by the arrow in Fig. 5A
Fig. 7: EDX spectrum analyzed at the ceramic phase as indicated

by the arrow in Fig. 5B
II - 714
CONCLUSIONS
Based on the aforementioned experimental results, it can be concluded that during sintering of
the compacts composed of 50 Vol% TiN + 50 Vol% Ni at 1150-1350 C in vacuum for 1
hour, Ti atoms diffuse into Ni lattices forming homogeneous solid solution. This leads to the
expansion of the lattice constant. The higher the sintering temperature, the more Ti atoms
dissolve, and the more obvious expansion occurs. Only a small amount of Ni diffuses into
TiN and dissolves in it. The amount of the Ni is so small that the X-ray diffraction patterns
show a stable crystal structure of TiN within the temperature range tested.
It is believed that this interdiffusion also takes place at the TiN/Ni interface involved in the
Al2O3-TiN+Ni system. For the hot-pressed cermet Al2O3-TiN+Ni, this interdiffusion is
beneficial to the adhesion at the TiN/Ni interface. Therefore, it can be an explanation of the
mechanism that TiN interlayer remarkably enhances the sintering ability and mechanical
properties of the Al2O3-TiN+Ni system.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge that this project is financially supported by former
Foundation for Advanced Metals Science, the Netherlands, University of Twente, the
Netherlands, and National Natural Science Foundation of China.
REFERENCES
1.
McHugh, C. O., Whalen, T. J. and Humenik, M., "Dispersion-strengthened Aluminum

Oxide", Journal of American Ceramic Society, Vol. 49, No. 9, 1966, pp. 486-491.
2.
Gazza, G. E., Barfield, J. R. and Preas, D. L., "Reactive Hot-pressing of Alumina with
Additives", American Ceramic Society Bulletin, Vol. 48, No. 6, 1969, pp. 606-610.
3.
Rankin, D. T., Stiglich, J. J., Petrak, D. R. and Ruh, R., "Hot-pressing and Mechanical
Properties of Al2O3 with a Mo-dispersed Phase", Journal of American Ceramic
Society,Vol. 54, No. 6, 1971, pp. 277-281.
4.
Whitney, E. D., "New Advances in Ceramic Tooling in the USA", Powder Metallurgy
International, Vol. 10, No. 1, 1978, pp. 16-21.
5.
Naerheim, Y., "Strength, Toughness and Thermal shock Resistance of Al2O3-Ti

Cermets", Powder Metallurgy International, Vol. 18, No. 3, 1986, pp. 158-162.
6.
Brook, R. J., "Additives and the sintering of Ceramics", High Tech Ceramics,
Vincenzini, P., Ed., Elsevier Science Publishers BV, Amsterdam, 1987, pp. 757-761.
7.
Trabelsi, R., Treheux, D., Goeuriot-Launay, D., Goeuriot, P., Thevenot, F., Orange, G.
and Fantozzi, G., "Friction, Wear Resistance and Mechanical Properties of an alumina-g
Aluminum Oxynitride Composite (Aluminalon)", High Tech Ceramics, Vincenzini, P.,
Ed., Elsevier Science Publishers BV, Amsterdam, 1987, pp. 2683-2695.
II - 715
8.
Lee, M. and Borom, M. P., "Rapid Rate Sintering of Al2O3-TiC Composites for CuttingTool Applications", Advanced Ceramic Materials, Vol. 3, No. 1, 1988, pp. 38-44.
9.
Goeuriot, P., Goeuriot-Launay, D. and Thevenot, F., "Oxidation of an Al2O3-g AlON

Ceramic Composite", Journal of Material Science, Vol. 25, 1990, pp. 654-660.
10.
Li, Shujie, Investigation of Powder Metallurgical Route for Manufacturing Al2O3/Ni

Cermets with Interlayers, Quick Service drukkerijen Nederland B. V., Enschede, The
Netherlands, 1992, ISBN 90-9005531-2.
11.
Li, Shujie, Khosrovabadi, P. B. and Kolster, B. H., "Effect of a Titanium Nitride

Interlayer on the Densification , Properties and Microstructure of Cermets Based on
Alumina and NickelPart 1: Densification and Properties", International Journal of
Refractory Metals and Hard Materials, Vol. 11, No. 3, 1992, pp. 181-192.
12.
Li, Shujie, Khosrovabadi, P. B. and Kolster, B. H., "Effect of a Titanium Nitride

Interlayer on the Densification, Properties and Microstructure of Cermets Based on
Alumina and NickelPart 2: Microstructure", International Journal of Refractory
Metals and Hard Materials, Vol. 11, No. 3, 1992, pp. 193-200.
13.
Feidorlqink, U. M., Principles of Powder Metallurgy, The Publishing House for

Metallurgical Industry, Beijing, 1974, (Chinese translation from Russian), p. 329.
14.
Mnatsakanyan, E. D., Vil'k, Yu. N., Paputskii, Yu. N., Kozlovskii, L. V. and Ordan'yan,
S. S., "Interfacial Reaction and Mass Transport of the Components in The Al2O3-TiN
System During Sintering", Sov. Powder Metall. Met. Ceram. (English Translation), Vol.
24, No. 11, 1985, pp. 812-817.
15.
Pearson, W. B., A Handbook of Lattice Spacings and Structures of Metals and Alloys,
Pergamon Press Ltd, Oxford, London, Edinburgh, New York, Paris, Frankfurt,
Reprinted with corrections 1964, p. 779 and p. 791.
II - 716
STRUCTURAL EVOLUTION OF C/C-SIC

COMPOSITES THROUGH THE MAIN PRODUCTION
STEPS ON THE LIQUID SILICON INFILTRATION
(LSI) ROUTE
Herbert Mucha1, Akira Kamiya2, Bernhard Wielage1
1
Technical University Chemnitz-Zwickau, Institute of Composites and Surface

Technology, D-09107 Chemnitz, Germany
2
National Industrial Research Institute Nagoya, 1-1 Hirate-cho, Kita-ku, Nagoya,
462, Japan
SUMMARY: Carbon fiber reinforced carbon- and SiC-matrix composites can be made by
different production methods and are displaying characteristic microstructures. This paper
focuses on the analysis of structural aspects induced by the fabrication process of C/C-SiC
composites fabricated by the Liquid Silicon Infiltration (LSI) method which was developed
by DLR (German Aerospace Research Establishment).
The LSI method consists in three main production steps (CFRP, C/C and C/C-SiC state) and
the evolution of the composites morphology and structure is traced through these production
steps from the carbon fiber reinforced plastics (CFRP) state to the finally processed C/C-SiC
composite. Conventional optical microscopy, EPMA, REM and TEM serve as tools to
disclose the microstructural organization and elemental informations in order to understand
their origins and the implications on the finally processed composite materials.
INTRODUCTION
Several methods are well suited to produce carbon fiber reinforced SiC-matrix composites.
Examples are the chemical vapour infiltraton (CVI), various Si-polymer pyrolysis routes with/
without filler and the Liquid Silicon Infiltration (LSI) method [1] which is developed at DLR
(German Aerospace Research Establishment). It is based on a C-polymer pyrolysis and
subsequently followed by an infiltration of liquid silicon. Each of these methods offers merits
for particular economical or technological requirements. The LSI-method as a one shot-near
net shape fabrication process is a quick and economical alternative to the other processes for
many technical applications in near future [2]. The upscaling from the laboratory to a
technical scale proved feasible. A large number of high performance real parts as e.g. an
intake ramp for hypersonic propulsion systems or brakes are produced by this way and
confirm the high level of technological development of the LSI-method [3,4,5].
Starting from the raw components the material has to undergo several transformations before
the product reaches the final C/C-SiC condition. The following discussion puts emphasis on
the elucidation of structural changes which take place accompanied with the three main
processing steps which are Carbon Fiber Reinforced Plastic (CFRP), Carbon/Carbon (C/C)
and C-fiber reinforced ceramic state (C/C-SiC).
II - 717
For tracing the structural changes a specialized sample preparation is applied adequate to
handle the strong heterogeneity of the composites. Through all the material states this is a
basic requirement for the optical microscopy, REM (Raster Electron Microscope), EPMA
(Electron Probe Micro Analyser) and TEM (Transmission Electron Microscope)
investigations. The samples architecture is shown clearly by the optical microscopy. EPMA
serves informations about elemental distributions within the C/C-SiC samples, wheras REM
and imaging TEM disclose further details of the microstructural organization. This way the
fiber/matrix interfaces , the reaction zones or even C-fiber damages are depicted.
THE LIQUID INFILTRATION METHOD

The LSI method [1,2,3] is a relatively fast fabrication process that consists of three processing
steps, which are passed through just one time during the whole fabrication cycle (Fig. 1).
Starting materials are stacks of 2D carbon fiber cloth (plain weave) and high carbon yield
polymeric C-precursors which have to fulfill particular requirements induced by the
processing.
1. CFRP forming (5-20bar, Polymerization, 200oC, inert atmosphere):
The first step of composite fabrication applies the Resin Transfer Moulding to inject the
polymer into the stack of carbon fiber cloth. The subsequent polymerization is performed in a
furnace applying a well defined time-temperature profile. It is accompanied with development
of little matrix porosity but significant matrix shrinkage which causes a periodic primary
segmentation within the fiber bundles.
2. C/C forming (Pyrolysis, 1 bar, N2, 900oC):
The CRFP composite undergoes a thermolysis with well controled time-temperature profile
during the next fabrication step. A great matrix volume shrinkage is accompanied by
substantial structural changes. A multiply connected translaminar channel system and a
second segmentation of fiber bundles is formed. The characteristic micro crack pattern
depends essentially on the fiber matrix binding strength.
3. C/C-SiC forming ( Liquid Silicon infiltration, vacuum, 1550oC):
Capillary forces are the driving forces of the infiltration of liquid elemental Si into the C/C
channel system under vacuum conditions, after preheating the C/C composite. The
accompanied diffusion controlled reaction between carbon and silicon is strongly dependend
on the pore size distribution of the infiltrated C/C-composite but also on the particular
modification of carbon being present within the reaction zone. The time-temperature profile is
very essential for tailoring the desired product properties. Fig. 1 schematically displays such a
profile (Brennkurve) [3].
II - 718
Fig. 1: Schematic Time-Temperature Profile (Brennkurve) of the LSI process [3]

EXPERIMENTAL
All samples independently whether CFRP, C/C or C/C-SiC state are cut from plate like
composites using a diamond saw under well defined angles with the laminate layers. The
ceramographical preparation makes use of supporting rings inserted into the conductive
embedding material. Grinding applies various SiC papers down to 14 m grain size. Polishing
on cotton cloth using OP-S suspensions (0,25 m grain size) follows subsequently. To
prevent electron induced matrix degradation of CFRP samples a gold coating is applied and
REM is operated in such cases at low voltages (<5kV). The TEM preparation additionally
required dimple grinding (diamond paste of 3 m) and subsequent conventional ion milling
(PIPS).
RESULTS
CFRP State
The CFRP real structure (Fig. 2a,b) results from the matrix shrinkage and superimposed
shrinkage impediment along the C-fiber axes. An almost periodic division of the fiber bundles
in small segments is seen. Preparation of samples parallel to the laminate layers discloses the
existence of two perpendicular crack systems which are almost independent of each other
(Fig. 2b). The polymerisation and degassing conditions (one parameter e.g. is the pressure) as
well as the type of raw materials exercise a strong influence on the development of open and
closed porosity within matrix rich areas and inside the fiber bundles.
The matrix rich areas usually display a homogeneously dispersed closed porosity. A careful
mechanical sample preparation partially retains an additional foamy phase located at crack
sites (Fig. 3a) or independent of cracks (Fig. 3b).
II - 719
(a)
(b)
Fig. 2: Optical micrograph of CFRP composite: a) perpendicular b) parallel to laminate
layer
(a)
(b)
Fig. 3: REM micrograph of foamy matrix (CFRP) a) adjacent to crack b) independent of
crack
(a)
(b)
Fig.4: C/C state sample (a) overview (b) foamy matrix component
II - 720
C/C State
The applied pyrolysis induces very severe structural changes to the CFRP-composite caused
by the mass loss (60% carbon yield) during pyrolysis (Fig. 4a). A multiply connected open
channel system develops and the closed porosity in bulk matrix areas grows significantly.
The increased crack width in matrix rich areas is evident as well as the periodical occurence
of regions within the fiber bundles where the average distance between neighbouring fibers
increased (zone of reduced fiber density). This periodicity corresponds to that of the primary
segmentation as observed in the CFRP-state. To some extent a fiber rearrangement takes
place (larger fiber distance) in areas of primary segmentation but also a displacement of a
foamy matrix component into the bulk matrix cracks is observed (Fig. 4b). The foamy matrix
component usually is separated from bulky matrix or fibers by very fine cracks (Fig.4a).
Additionally a second segmentation takes place within the C-fiber bundles but these cracks
are not filled by any foamy matrix component. Polarized light microscopy which is performed
additionally gives indications that stress induced graphite is formed locally within the fiber
bundles even at processing temperature as low as 900oC.
C/C-SiC State
A characteristic of C/C-SiC is the segmented wall structure which develops from the open
channel system of the C/C state by the infiltration of liquid silicon (Fig. 5a). This channel
system extends over large distances and is shown most clearly if the cut section is in parallel
to the laminate layers (Fig. 5b). The infiltration and the accompanied chemical reaction shows
a locally different behaviour (Fig. 6a) which is dependent not only of the width of the pores
but also correlated to the carbon modifications reacting with Si and their lateral distribution
within the reaction zone. C-matrix modifications of different appearance were detected in the
C/C state. After the infiltration the former crack areas still can be identified in large scale by
optical microscopy. Their structure reflects the former course of reaction between Silicon and
Carbon. The REM investigations of C/C-SiC materials require a gold coating in order to
prevent charging but clearly reveal topographical inequalities at higher magnifications. They
originate in the different hardness of the substrates surface and become effective during the
mechanical sample preparation (Fig.6b). The reaction zones mostly display a cloud like shape
and consist of a brittle material (Fig. 7a) which is retained only under dedicated sample
preparation. The reaction between silicon and carbon involves both fiber and matrix. Even
submicrometric pores are efficiently infiltrated.
A typical crack area which generated from a zone of reduced fiber density by Si infiltration
(Fig. 6b) shows sharp topographical steps in central areas and a less sharp and brittle type
adjacent to the carbon fiber bundles or to bulky matrix areas as displayed in Fig. 7a under
higher magnification. The gold coating, and the limited accuracy of conventional EDS
(Energy Dispersive X-ray Spectroscopy) equipment for the detection of light elements, as
carbon, give a preference to WDS (Wavelength Dispersive X-ray Spectroscopy) EPMA
investigations of the elemental distribution. Since no sample charging was observed, no gold
coating had to be applied. The Si elemental maps (as Fig.7b) and topographical images show
some general features as pure Si areas (white) embeded in Si-rich areas of large grain size
(central grey), Si-supply channels (dark grey) and fine grained areas (grey) with concentration
gradients. They are adjacent to the unreacted carbon zones (dark black). Transmission
electron micrgraphs usually show that even narrow infiltrated pores consist in a core of coarse
grained SiC at almost the same position as the former pore and a reaction zone consisting of
II - 721
fine grained SiC surrounding it (Fig. 8a). Elemental Si also can be detected locally but
preferentially within the wide pores. The fine grained SiC area occasionally displays
microcracks ending at closed porosity. Closed porosity in the C/C state preferentially
accumulates near fiber surfaces (Fig. 8b) and modifies the resulting fiber matrix binding
strength. The closed C/C state porosity usually is retained during the silicon infiltration
leading to voids in the C/C-SiC reaction zones. TEM imaging proves that even
submicrometric cracks which are connected to the open channel system are infiltrated
efficiently. Their reaction zones have the same characteristic appearance as the larger ones.
Another interesting observation can be made in front of the reaction zones inside the carbon
fibers. Analogous effects were found in the carbon matrix (Fig 8a), too. The fibers show
indications of damages within a range of several hundred nanometers in front of the reaction
zone in apparently unreacted areas of the fibers. Whereas the segregations in the fibers seem
oriented along the fiber axis, the segregations in the matrix appear in a more spherical shape.
(a)
(b)
Fig. 5: Segmented wall structure of C/C-SiC a) perpendicular b) parallel to laminate layers
(a)
(b)
Fig. 6: C/C-SiC reaction zones (a) optical micrograph (b) REM micrograph
II - 722
(a)
(b)
Fig. 7: C/C-SiC (a) reaction zone (REM) (b) Si-elemental map (EPMA-WDS)
(a)
(b)
Fig. 8: C/C-SiC (a) crack structure (TEM) (b) accumulated closed porosity (TEM)
DISCUSSION
It is likely that the WDS Si-elemental map (Fig. 7b) is taken from an area which before the Si
infiltration took place resembled that of Fig. 4b but showed additional open porosity. The
narrow cracks separating the foamy matrix from the regular composite are assumed to be
connected with the open porosity. Under this condition liquid silicon is supplied by the the
open pores and distributed to the narrower cracks. The higher Si concentration compared to
other crack areas can be explained by this argument. The foamy matrix in central parts of the
zone of reduced fiber density serves as easily accessible carbon reservoir for the Si-C
reaction. Due to the direct impregnation of the foamy matrix by liquid silicon the diffusion
distances are short and the course of reaction is much more advanced compared to those
regions where the reaction is based on the long distance diffusion of the reaction partners.
Finally in such areas a recrystallization can take place and form large crystallites, as seen by
TEM micrographs (Fig. 8a) or less pronounced in mass contrast or topo images. The hardest
component is the coarse grained phase (SiC), whereas the weakest is elemental silicon. The
correlated topographical steps can be detected on suitably prepared samples. Only pores
II - 723
which are directly infiltrated can be filled up by this method; closed porosity will remain
closed porosity, independent of whether the carbon of the surrounding reacts to SiC or not. As
a hypothesis, the segregations within the fibers in front of the reaction zone may grow due to
a Si gas phase infiltration inside the nanoporosity of the fiber.
A suitable segmentation and pore size distribution is essential for good infiltration results. The
fiber-matrix binding strength is one of the controlling parameters of the segmentation width.
Hardening during CFRP forming is accompanied with a matrix shrinkage and a first
segmentation of the fiber bundles inside the 2D (0o-90o) CFRP composite. Subsequent
pyrolysis causes further shrinkage and an additional segmentation. Wanner [6] created a
model to describe the periodicity of this secondary segmentation based on the internal stress
distribution of the composite.
The controlled release of gaseous reaction products has to take place in order to reduce the
size of bubbles and not to accumulate too much closed porosity inside the composite. The
corresponding time-temperature profiles (Brennkurve) are tailored for each precursor and
based on thermogravimetrical and thermoanalytical investigations.
CONCLUSIONS
Tracing the structure of the starting CFRP-material through the subsequent manufacturing
steps (Pyrolysis and Si infiltration) the foamy matrix component leads to coarse grained SiC
areas when infiltrated directly. If not infiltrated directly this porosity is conserved as an area
of weak binding. The importance of an effective control of crack size and porosity during
C/C-SiC fabrication is confirmed.
ACKNOWLEDGEMENTS
The authors gratefully wish to acknowledge the support of the German Aerospace Research
Establishment (DLR) which supplied the samples and made it possible to perform a part of
this work in its laboratory for microstructure analysis.
REFERENCES
1.
2.
3.
4.
5.
6.
II - 724
Kochendrfer, R.; Liquid Silicon Infiltration - A fast and low cost CMC manufacturing
process, ICCM 8, 1991, Honolulu
Krenkel, W.; Kochendrfer, R.; The LSI Process - A cost effective processing
technique for ceramic matrix composites; Proc. of ICAM, Progress in Advanced
Materials and Mechanics ,Tagung12.8.-15.8.1996, Peking
Krenkel, W.; CMC Design Consequences, AGARD Workshop Introduction of
Ceramics into Aerospace Structural Composites, Antalya, 18-23 April 1993
R. Kochendrfer, W. Krenkel; CMC intake ramp for hypersonic propulsion systems;
Proc. HT-CMC2, Santa Barbara, Ca, 21.8.-24.8.1995, S. 13-22
W. Krenkel; CMC Materials for high performance brakes; ISATA Conference on
supercars, Aachen, 31.10.-4.11.1996
A.
Wanner;
Gefge
und
mechanische
Eigenschaften
des
Kohlenstoffaser/Kohlenstoffmatrix-Verbundwerkstoffs
CF222,
Dissertation,
Universitt Stuttgart 1991
THE FABRICATION OF TWO-DIMENSIONAL

CARBON FIBER REINFORCED ALUMINA
COMPOSITES BY PRESSURE INFILTRATION
Hsien-Kuang Liu and Bor-Horng Lin
Department of Mechanical Engineering, Feng-Chia University
100 Wenhwa Road, Taichung, Taiwan, R.O.C.
SUMMARY: Using pressure infiltration can fabricate two-dimensional (2-D) textile ceramic
matrix composites with uniform and dense microstructure. Based upon this processing
technique, composites composed of silica sol, alumina particles, and 2-D satin weave carbon
fibers were fabricated. The related processing parameters studied in this work include fiber
surface treatment and fiber volume fraction. The woven carbon fabrics were treated by a
liquid oxidation reaction method in order to study their surface conditions on mechanical
properties of 2-D composites. Fiber volume fractions are varied by changing weave density of
2-D carbon woven fabrics. Material characterizations have been conducted by a three-point
bending test and an indentation fracture toughness test. Therefore, the aim of this work is to
optimize processing parameters for obtaining better mechanical properties of composites.
KEYWORDS: pressure infiltration, 2-D fabrics, liquid oxidation reaction
INTRODUCTION
In practical applications, a structure component might be subjected to mutidirectional loads,
resulting in the need to arrange the reinforcing fibers in more than one direction. Textile
composites using tailored preforms, such as woven, knitted, or braided fabrics, are versatile to
accommodate a variety of requirements, such as stability (strength/stiffness) or extensibility in
desired directions and relative ease of handling in fabrication [1]. The advantages have led to
the current interest in fabricating ceramic matrix composites based on textile preforms.
However, the microstructure of textile ceramic matrix composites (CMCs) is significantly
influenced by the processing route. For example, textile CMCs fabricated by the chemical
vapor infiltration (CVI) technique normally contain high porosity which results in the
stiffness reduction of the composite [2].
Several processing routes for CMCs have their respective advantages. SiC matrix ceramic
composites fabricated by CVI [3] have improved mechanical properties. The advantage of
CVI is that chemical vapors can react and produce pure ceramic matrix at low temperature
around 1200oC. 2-D Nicalon/BN/SiC composites processed by the isothermal chemical vapor
infiltration (ICVI) method [4] have better load transfer efficiency under aggressive
environment than typical SiC/C/SiC composites. 2-D woven fabric SiC/SiC composites with
multilayered interphases fabricated by CVI method [5,6] have both high strength and
toughness due to strong interphases and large amounts of matrix cracks at saturation,
II - 725
respectively. 2-D woven fabric SiC/SiC composites can be produced by a slurry infiltration
method [7] associated with multiple impregnation at a lower temperature. 3-D C/SiC and
C/Si3N4 composites made by polymer pyrolysis and repeated infiltration [8] gain their unique
properties. A common advantage of these fabrication routes is that the processing
temperature can be lowered to 1200oC. However, a common disadvantage of these
fabrication routes is long processing time. Based on economical point of view, to cut down
the processing time should be one of the important issues for fabrication of CMCs in the near
future. To achieve the above goal, Liu and Parvizi-Majidi [9] tried to fabricate 3-D ceramic
matrix composites with satisfactory density and uniformity in a single infiltration using the
addition of solid particles to the sol during sol-gel processing of ceramic matrix composites.
Therefore, the purpose of this work is to introduce a pressure infiltration method which could
fabricate 2-D woven carbon fabric reinforced alumina composites with dense microstructure.
The woven carbon fabrics were treated by a liquid oxidation reaction method in order to
evaluate the influence of their surface conditions on mechanical properties of 2-D woven
fabric CMCs. To optimize processing parameters for obtaining better mechanical properties
is the goal in this study.
MATERIAL PROCESSING
Sol and Woven Fabric Preparation
The silica sol was prepared using a modified recipe of Klein's work [10] . In this recipe,
tetraethyl-orthosilicate (TEOS), ethanol, deionized water and 7 wt% of HNO3 were mixed
and stirred at a volume ratio of 1: 1: 1.6 : 0.06. All the reaction of the mixture was all at
room temperature. 25 grams of the alumina powders ( - Al2O3, Bayer process, average
particle diameter of 0.45 m, purity of particle of 99.9%, and a density of 2.7g/cm3) were
added to the sol to prepare the infiltrate. The infiltrate was further mixed both in ultrasonic
bath and electromagnetic blender in turns until the powders in the infiltrate were dispersed
uniformly and well wetted.
2-D plain carbon fabrics were used as preforms. The carbon fiber is Toho T-300 fiber with
tensile strength of 3200 MPa. The first kind of fabric has fiber volume fraction (Vf) of
33.7%. Each fiber yarn has 3000 fibers. Fiber has density of 1.78 g/cm3 and diameter of 8
m. Weave density is warp 5 yarns/cm and weft 5 yarns/cm. Fabric weight is 194 g/m2. The
second kind of fabric has fiber volume fraction of 21.5%. Weave density is warp 2 yarns/cm
and weft 2 yarns/cm. Fabric weight is 87 g/m2. Each fabric was cut as a circle of 50 mm
diameter. 12 layers of the woven fabrics were stacked together and stitched by a string as a
preform with a thickness 3.67 mm (first kind) and 2.78 mm (second kind), respectively. The
preform was further processed by pressure infiltration to obtain composites. The woven
fabrics were either untreated (as-received) or treated by liquid oxidation reaction in order to
compare the effect of fiber surface conditions on mechanical properties of the composites.
The treated procedure is as follows. Carbon fabrics were dipped in 35wt% nitric acid solution
at 70oC for 6 hours. Then the fabrics were washed by deionized water and dried at 150oC .
II - 726
Pressure Infiltration
The pressure infiltration apparatus was shown in Fig. 1, with an inner diameter of 50 mm in
the cylinder, and a base cavity for collecting the liquid. The filter assembly consists of nitrate
cellulose membrane filter paper with selected pore size, a wire cloth-stainless steel, and a
perforated solid steel disk. The filter assembly was placed on the top of the base cavity. The
wire cloth-stainless steel was used to keep the filter paper from rupture. Cylinder, filter
assembly, and base were tightly sealed by thread and O-ring. Carbon fabrics were first put in
the cylinder and attached to the filter assembly closely. A volume of 175 ml of infiltrate was
then poured into the cylinder to infiltrate fabrics. An MTS testing machine set at constant
crosshead speed mode was used to provide the infiltration pressure on the plunger. The use of
a constant crosshead speed insured a gradual increase in pressure which consolidates powders
inside the fabrics uniformly, and prevented filter paper from rupture. The maximum values of
pressure were recorded: 46.7 MPa for the composite with untreated carbon fibers and 21.0
MPa for the composite with treated carbon fibers.
Fig. 1: Schematic diagram of pressure infiltration apparatus.

After first infiltration, the specimen was ejected and dried in an oven at 60oC for 24 hours.
After the drying process, the composite was soaked in silica sol until saturation, and finally
the composite green density was calculated from its dimensions and weight. Green
composites were then hot pressed in a furnace (FCPHP-R-5, FRET-20, HIGH MULTI 5000,
Multi-purpose High Temperature Furnace, Japan) under temperature 1600oC and pressure 25
MPa for one and half hours by first evacuated to 10-5 torr, and then flowing nitrogen gas.
After hot pressing, the percentage theoretical densities of composites were measured. In
addition, microstructure and mechanical properties of composites were also evaluated.
MATERIAL CHARACTERIZATION
The hot pressed composites were cut as specimen to evaluate mechanical properties and
examine microstructure. Specimen cross sections were directly examined by a scanning
electron microscope for bonding between fibers and matrix. Specimen before and after
mechanical testing were embedded in urea, then polished with diamond paste and examined
by an optical microscope. Specimen after mechanical testing were photographed by camera
to observe their macroscopic failure.
II - 727
The indentation fracture technique was adopted to determine the fracture toughness of
composite materials. The Vickers diamond pyramid indenter was used to produce the crack
patterns onto the well polished cross sections of specimen, which are on matrices between
fiber layers. The indentation load is 5 kgw, which provides consistent indentation patterns
with radial crack lengths substantially larger than the indentation diagonals. Thus, fracture
toughness (KIC) can be calculated by measuring the lengths of diagonals and radial cracks by
using the following equation
KIC = (E/H)1/2 (P/c3/2)
(1)
where is a constant around 0.016, E is modulus of composites, H is hardness of composites,

P is indentation load, and c is half of the crack length.
A three point bending test was adopted to measure flexural strength and
modulus of
composites. An MTS testing machine was used to conduct the test. The span L, width w,
and height h of the specimen were designed to be 40 mm, 9 mm, and 5 mm, respectively;
detailed dimensions were measured for each specimen for further calculation of the strength
and modulus. The width of the specimen is made larger compared to that of typical
unidirectional ceramic composites to avoid scattering of the data. Crosshead speed was set at
0.008 mm/sec. The relationships of applied load (P) versus displacement () were recorded
during the test. The ultimate load (Pmax) were taken for the calculation of flexural strength
(max) and modulus (E) according to the following equations
max = 3PmaxL/2wh2
(2)
E = PmaxL3/4wh3
(3)

Effect of Liquid Oxidation Reaction
In order to enhance the bonding between carbon fabrics and alumina matrix as well as
interlayer shear strength, liquid oxidation reaction was adopted to modify the surface
condition of carbon fabrics. Nitric acid was used in the liquid oxidation reaction. The
existence of shallow grooves on the surface of one-hour treated fibers was observed. After 6
hour treatment, the grooves become deeper; furthermore the outside layer have been etched
as some fragments and then peeled off until the surfaces becomes rougher and rugged. This
phenomena is beneficial for interfacial bonding due to more contacts between matrix particles
and fibers. By liquid oxidation reaction, some active oxide functional groups are attached to
the surface of carbon fabrics, such as hydroxyl group, ester group, carboxylic acid group,
and carbonyl group. These functional groups were verified by IR spectrum for the surface of
treated fibers [11]. After 300 minute treatment, the distribution of those functional groups is
as follows: 80% for hydroxyl, 6% for carbonyl, 6% for carboxylic acid, and 8% for ester,
respectively. Those active oxide groups may react with matrix particles and silica gels
leading to chemical bonding between fibers and matrix. Treated carbon fabrics with active
functional groups on their surfaces can have high affinity with silica in the matrix. Some of
those groups have negative charge, such as hydroxyl. They attract hydrogen atoms with
positive charge in silanol group on the surface of silica. Hence, during hot pressing, those
II - 728
active functional groups can react with silanol via dehydration process and result in siloxane,
then chemical bonds were formed [12]. Furthermore, silica may react with alumina and
produce mullite with lower melting point resulting in better interfacial bonding. The
evidences can be supported from several figures that discussed below.
Figs. 2(a) and 2(b) show cross sections of hot pressed composites with treated and untreated
carbon fabrics, respectively. After cutting and polishing, the former shows better infiltration
of matrix in the intrayarn pores inferring better bonding between fibers and matrix. While the
latter shows worse infiltration corresponding to poor bonding at the interface. Fracture
surfaces perpendicular to treated carbon fibers of 2-D carbon/alumina composites are flat
indicating complete bonding between fibers and matrix (Fig. 3(a)). However, it seems that
incomplete bonding between fibers and matrix occurs for composites with untreated carbon
fibers because the surface of composites is likely in a molten state as shown in Fig. 3(b). The
incomplete bonding may be due to the reaction of sizing on the carbon fabrics with silica gels.
By our observation, it is found that the distribution of alumina matrix is inhomogeneous and
sintering effect is nonuniform in two kinds of pore, interyarn pores and intrayarn pores. This
may lead to the reduction of strength of composites. The content of matrix materials in
intrayarn pores is less because they are much smaller compared to interyarn pores so that
infiltration is more difficult. Furthermore, during hot pressing, fibers around intrayarn pores
constrains matrix shrinkage more than fibers around interyarn pores. Both reasons result in
poor bonding between fibers around intrayarn pores and matrix particles, but better bonding
between fibers around interyarn pores and matrix particles as shown in Figs. 3 (a) and (b).
The influence of liquid oxidation reaction on fracture toughness of 2-D carbon/alumina
composites is shown in Table 1. Composites with untreated fibers have higher fracture
toughness than those of composites with treated fibers. It is because that the former have poor
bonding at fiber/matrix interface so that propagation of cracks may go through interface. Its
porosity is somewhat higher than that of the latter as can be compared by the final densities
from Table 1. The above reasons explain why composites with untreated fibers obtain better
fracture toughness. But, in Table 2, it is shown that composites with untreated fibers obtain
both lower flexural strength and modulus of elasticity as compared to those of composites
with treated fibers due to better bonding between fibers and matrix.
(a)
(b)
Fig. 2: 2-D carbon/alumina composites with cross sections parallel to the hot pressing
direction showing (a) better infiltration in the intrayarn pores among treated carbon fibers
inferring better bonding between matrix and fibers in this region, (b) worse infiltration in the
intrayarn pores among untreated carbon fibers
II - 729
(a)
(b)
Fig. 3: Fracture surfaces of 2-D carbon/alumina composites perpendicular to the fibers
showing (a) complete bonding between treated carbon fibers and matrix, (b) incomplete
bonding between untreated carbon fibers and matrix
Table 1: Results of fracture toughness of 2-D carbon fabric/Al2O3 ceramic matrix

composites
Table 2: Flexural strength and modulus of 2-D carbon fabric/Al2O3 ceramic matrix
composites
Effect of Fiber Volume Fraction

The fracture toughness of composites increase with the increase of fiber volume fraction as
shown in Table 1. It is known that fibers in ceramics can inhibit the propagation of cracks
resulting in the increase of fracture toughness. The influence of volume fraction of fibers Vf
on both flexural strength and modulus of elasticity is shown in Table 2. For lower fiber
II - 730
volume fraction, less constraint of matrix shrinkage during sintering from fiber fabrics results
in better densification and less porosity of matrix particles. This leads to higher strength.
However, the sparse fiber yarns in the fabrics can distort during their preparation or
infiltration of the composite resulting in its lower modulus of elasticity.
The failure procedure of 2-D carbon/alumina composites with untreated carbon fibers can be
described by Fig. 4. Arabic numerals represent the sequence of the procedure. Cracks initiate
at the matrix between fabric layer (1) and propagate at the interface between fiber layer and
matrix. When cracks meet fiber yarns perpendicular to the paper plane (weft yarns), they
propagate through intrayarn pores (2). Later, cracks would meet the fiber yarns parallel to
the paper plane (warp yarns) and continue to propagate at the interface between fiber yarns
and matrix (3). Eventually, as the energy of cracks is large enough, it breaks the fiber yarns
parallel to the paper plane (4) and composites fail. According to the observation, most of
cracks in the composite with untreated carbon fibers locate at interlayer matrix, between weft
and warp yarns, inside weft yarns, and on warp yarns. The distribution of cracks is uniform.
On the other hand, cracks in the composite with treated carbon fibers are located at interlayer
matrix and inside weft yarns near matrix-fiber interface. The distribution of cracks is
nonuniform and dense around large pores. An overview of failure of 2-D carbon/alumina
composites can be seen in Fig. 5. Some transverse microcracks can be found in the
photographs. Interlayer delamination is the major failure mode of composites with treated
fibers (Fig. 5(a)) because interlaminar shear strength of 2-D ceramic composites is
intrinsically weak and cracks may propagate through matrix between layers instead of
stronger interface between fibers and matrix. But local kinking is the major failure mode of
composites with untreated fibers (Fig. 5(b)) because cracks could propagate at weaker
interface under large compressive stresses and then local fibers would bend or kink without
the constraint of matrix.
Fig. 4: Failure procedure of 2-D carbon/alumina composites with untreated carbon fibers;
Arabic numerals representing the sequence of the procedure
II - 731
Fig. 5(a): Overview of failure of 2-D carbon/alumina composites showing first major failure
mode delamination (fibers treated)
Fig. 5(b): Overview of failure of 2-D carbon/alumina composites showing second major
failure modes local kinking (fibers untreated)
II - 732
CONCLUSIONS
2-D carbon/alumina composites were successfully fabricated by the pressure infiltration
method. The fracture toughness increases but flexural strength decreases with increasing fiber
volume fraction. Composites with untreated carbon fabrics have high fracture toughness of
7.0 MPa.m1/2, lower flexural strength of 30.4 MPa, and lower modulus of elasticity of 649.5
MPa. But composites with treated carbon fabrics by liquid oxidation reaction have lower
fracture toughness of 6.5 MPa.m1/2, higher flexural strength of 51.7 MPa, and higher modulus
of elasticity of 2081.4 MPa. Better bonding between treated fibers and matrix was found by
scanning electron microscopic observation.
ACKNOWLEDGMENTS
The authors are grateful to National Science Council, Taiwan R.O.C. for financial support
under contracts : NSC 83-0405-E-035-017 and NSC 84-2216-E-035-027.
REFERENCES
1.
Chou, T. W. and Ko, F., Textile Structure Composites, Elsevier Science Publishers, New
York, 1989.
2.
Kuo, W. S. and Chou, T. W., "Elastic Response and Effect of Transverse Cracking in
Woven Fabric Brittle Matrix Composites," J. Am. Ceram. Soc., Vol. 78, No. 3, 1995, pp.
783-792.
3.
Caputo, A. J., Stinton, D. P., Lowden, R. A. and Besmann, T. M., "Fiber Reinforced SiC
Composites with Improved Mechanical Properties," Am. Ceram. Soc. Bull., Vol. 66, No.
2, 1987, pp. 368-372.
4.
Leparoux, M., Vandenbulcke, L., Goujard, S., Robin-Brosse, C. and Domergue, J. M.,
"Mechanical Behavior of 2D-SiC/BN/SiC Processed by ICVI," Proceedings of The Tenth
International Conference on Composite Materials, Whistler, British Columbia, Canada,
August 14-18, 1995, Vol. IV: Characteriztion and CMCs, Poursartip, A. and Street, K. N.
Eds, pp. pp. 633-640.
5.
Droillard, C. and Lamon, J., Fracture Toughness of 2-D Woven SiC/SiC CVIComposites with Multilayered Interphases, J. Am. Ceram. Soc., Vol. 79, No. 4, 1996,
pp. 849-858.
6.
Droillard, C., Voisard, P., Heibst, C. and Lamon, J., "Determination of Fracture
Toughness in 2-D Woven SiC Matrix Composites Made by Chemical Vapor
Infiltration", J. Am. Ceram. Soc., Vol. 78, No. 5, 1995, pp. 1201-1211.
7.
Shin, D. W. and Tanaka, H., Low-Temperature Processing of Ceramic Woven

Fabric/Ceramic Matrix Composites, J. Am. Ceram. Soc., Vol. 77, No. 1, 1994, pp. 97104 .
II - 733
8.
Nakano, K., Kamiya, A., Nishino, Y., Imura, T. and Chou, T. W., "Fabrication and
Characterization of Three-Dimensional Carbon Fiber Reinforced Silicon Carbide and
Silicon Nitride Composites," J. Am. Ceram. Soc., Vol. 78, No. 10, 1995, pp. 2811-2814.
9.
Liu, H. K. and Parvizi-Majidi, A., "The Effect of Solid Particles Addition in Sol-Gel
Processing of Ceramic Matrix Composites," Ceram. Eng. & Sci. Proc., Vol. 13, 1992, pp.
642-649 .
10. Klein, L. C., "Sol-Gel Processing of Silicates," Ann. Rev. Mater. Sci., Vol. 15, 1985, pp.
227-248 .
11. Chang, W. C., Effects of Processing Parameters on the Properties of Carbon/Carbon
Composites, Ph.D. Dissertation, National Tsing-Hua University, Taiwan, 1995, pp. 5-6,
5-7.
12. Su, P. S., The Material Technology of Nanometer Particles, Fu-Han Publishing Inc.,
Taiwan, 1990.
13. Liu, H. K., Kuo, W. S. and Lin, B. H., "Pressure Infiltration of Sol-Gel Processed Short
Fiber Ceramic Matrix Composites," in preparation , 1996.
II - 734
APPLICATION OF PRECERAMIC POLYMER FOR

SHAPE FORMING OF MULTIPHASE SIC/SI3N4
CERAMIC PARTS
Yongqing Li, Zhaohui Chen, Changrui Zhang, Haifeng Cheng, Ancheng Zhou and
Xiaoheng Bao
Department of Material Engineering and Applied Chemistry, National University of Defence
Technology, Changsha, Hunan, PRC, 410073
SUMMARY: The purpose of the present work is to fabricate shape-complex SiC/Si 3N4
multiphase ceramic parts (such as blade) using preceramic polymer (Polycarbonsilane PCS)
as shape-forming binders. SiC/Si3N4 multiphase ceramics are formed by in-situ pyrolysis and
sintering from PCS/Si3N4 green body. Divinylbenzene(DVB) is used as active diluent agent to
prepare slurry of PCS/DVB/Si3N4 at room temperature, and as cross-linking agent to make
solid green body of PCS/DVB/Si3N4. It is necessary to control precisely rate of pyrolysis
temperature of 250-450 C under 0.3 C/min in order to obtain a pyrolysis with body of
deformation-free and crack-ffee. Before sintering, decreasing the carbon-rich content in the
pyrolysis body, the density of sintered body can be enhanced.
KEYWORDS: precursor, SiC/Si3N4 ceramic, shaping, pyrolysis, sintering
INTRODUCTION
SiC/Si3N4 multiphase ceramic is one of the most promising candidate materials for high
temperature structural components in heat engines for its attractive characteristics such as low
density and excellent resistance to corrosion, wear, oxidation and heat-shocking at high
temperature. Once it was used in heat engine, fuel would be saved, the working temperature,
efficiency and power of the engine would be elevated. It may be the bottle-neck to prepare
such engine parts as turbine blade and rotor to realize the practicability of a ceramic engine.
The shaping technologies in common use in nowadays (e.g. injection and casting) are not
much promising because of their shortcomings as high cost, long cycle, poor property of
product and so on. The authors of this paper prepared ceramic blades using organosilicon
precursor as shaping binder, and discussed the effects of processing parameters on the
structure and performance of the pyrolyzed and sintered blades.
Preceramic polymer such as polycarbosilane (PCS) can be used as binder to fabricate ceramic
part [1-5]. Unlike conventional binders (PE,PP etc.) that are totally burned out, during
pyrolysis, the preceramic polymer binder can be chemically transformed into a continuous
ceramic phase. For example, while PCS is used as a binder of Si3N4 powder, PCS can be
changed into nanophase SiC at over 1000 C, and a multiphase ceramic can be in-situ formed
in the final part.
II - 735
Polycarbosilane (PCS), as a preceramic polymer (dark black brittle melt point 125-135 C)
was dissolved into divinylbenzene (DVB) which acted as active dilute agent, to form viscous
liquid for binder solution at room temperature. -Si3N4 powder and Sintering aids
(Al2O3/Y2O3) were incorporated into PCS/DVB solution by ultrasonic dispersion to form a
slurry of PCS/DVB/Si3N4/Al2O3/Y2O3, then the slurry was cast into a mold. Heat treatment
for curing of slurry was 130 C, 7 hours at 1.6 MPa nitrogen pressure for preventing DVB
from evaporation. First, Green body of ceramic parts was pyrolysed in nitrogen flow at 0.3~1
C/min from room temperature to 1000 C, held for 1 hour at 1000 C; Then, the pyrolysed
body was cooled to room temperature in the furnace. Second, the pyrolyzed body was heated
in nitrogen flow at 20 C/min from room temperature to 1600 C, held 1 hour at 1600 C to
crystalline amorphous SiC, and to react Si3N4 with free carbon derived from pyrolysed PCS,
and then sintered at 10 C/min from 1600 C to 1930 C, held for 120 minutes at 1930 C and
1.0 MPa nitrogen pressure.
Density of specimens was determined by Archimedes immersion technique. Phase
identification of sintered body was performed by X-ray diffraction using CuKoc-radiation.
Microstructure of pyrolysed body and sintered body was examined by scan electron
microscopy (SEM). Fracture strength was evaluated using three-point bending over a 20 mm
span. The ratio span to height specimens is 5.
RESULT AND DISCUSSION

Preparation of Slurry and Preforms
The slurry was composed of Si3N4 powder, PCS, DVB, sintering aids and shaping aids
(crosslinking initiator and coupling agent). The role of DVB in the slurry was: 1) a diluent to
reduce the viscosity of the slurry, and 2) a solidifying agent to react with Si-H bonds in PCS
(initiated by dicumylphene, pop above 120 C) and self-crosslink due to the vinyl groups in
its molecular. The effects of the concentration of DVB on the property of the crosslirlked
preform were listed in Table 1. According to these results, the ratio of PCS to DVB should be
1.5:1. Other experimental data proved that middle-molecular-weight PCS was appropriate and
coupling agent (e.g. titanate serials) should added into the slurry before ball milling to
improve the disconcentration of ceramic particles in precursor solution. For the same aim,
further dispersing treatment by means of super sonic should be adopted.
Table 1: Effects of PCS/DVB ratio on the facade of pyrolysed parts
PCS/DVB
outward appearance
of pyrolysed parts
2:1
metamorphosed
obviously
1.5:1
without distortion
no visible defects
1:1
cracked
evidently
Different compositions of binder solution were used to prepare crosslinked preform and the
crosslinking degrees of each sample were compared in Table 2. The data showed that
middlemolecular-weight PCS was seemly and the incorporation of about 2wt.% of pop as
initiator would promote the crosslinking reaction. IR analysis of the slurry, the crosslinked
preform and the precursor saluted in xylene after extracting the preform revealed that most of
II - 736
Si-H bonds did not crosslink as we expected. So, the microstructure of the crosslinked product
was cured DVB net with PCS chains running through it.
Table 2: Crosslinking degrees of preforms with different binder compositions
1
LPCS
MPCS
DVB
DCP
Crosslinking degree (%)
100
50
31.4
100
100
50
3
42.4
50
42.6
100
50
3
60.8
100
2
97
NOTE: All preforms were cured at 180 C in 1. 0-2. 0MPa N2 for 8h. The crosslinking degrees were
determined by measuring the maintained masses of the preforms after immersing them into xylene at
room temperature for 24hrs and then drying them at 80 C in vacuum for 2hrs. MPCS=middle
molecular weight PCS, LPCS=low molecular weight PCS
As discussed above, DCP was effective to initiate the self-polymerization of DVB and partial
crosslinking of Si-H with vinyl group. TG analysis showed that DCP began decomposition at
about 120 C. Besides, the boiling point of DVB is 138 C. So, the crosslinking reaction
should be performed at 120-138 C and nitrogen pressure (1.0-2.0 MPa) should be adopted to
avoid the volatilisation of DVB and the oxidation of PCS.
Pyrolyzation of the Preforms
TG analysis (10 C/min, Ar) of shaped preforms revealed that most of the loss occurred below
550 C, especially from 250 C to 550 C (see Figure 1). The said weight loss was due to the
volatilisation of low-molecular-weight oligomer and pyrolyzation of the crosslinked
precursor. it was important to control the heating rate this temperature scale to reduce the
amount of the volatiled and pyrolysed gas and the possibility of deforming, cracking and
bulging. Heating rates of 0.3, 0.5 and 1.0 C/min were introduced to optimize the heating rate.
Results in Table 3 showed that defect-free pyrolyzed parts could be obtained at a heating rate
lower than 0.3C/min.
Fig 1: TG curves of crosslinked samples

Note: 1: DVB/DCP 2: LPCS/DVB/DCP 3: MPCS/DVB/DCP. 4. LPCS/DVB/DCP/Si3N4
II - 737
Table 3: The property of ceramic parts pyrolysed at different heating rates

heating rate (C/min)
1
0.5
0.3
property of pyrolyzed parts

heavily distorted and cracked
slightly distorted, cracked at large cross section
no deformation, crack-free
TG curves in Figure 1 also showed that the weight loss above 800 C was negligible. This
result implied that the inorganization of the precursor was almost complete before 800 C.
The pyrolyzed ceramic bars that were further treated up to 1500 C (holding for 120 min)
showed very slight shrinkage and weight loss (about 2 wt.%). Their bending strength (6010
MPa) were almost the same as that before the treatment. So it was necessary to elevate the
sintering temperature to improve the strength of the parts.
As concluded above, the microstructure of the cured preceramic binder was a inter-joined
3dimensioned network with PCS chains running through crosslinked DVB moleculars, and in
the network there was about 12-18 wt.% of uncured PCS that would melt at elevated
temperature to cause distortion of the preform. This kind of deformation could be reduced by
extracting 8-10 wt.% of the soluble PCS out of the preform with xylene as solvent, without
distorting the preform during extraction. The extracted preforms were heated at a faster
temperature rate, and defect-free pyrolyzed parts were obtained with highly reduced linear
shrinkage. The reasons for such a improvement might be that absence of low molecular
weight PCS decreased the volatilization during pyrolyzation and gangways for pyrolyzed gas
were left after extraction.
Sintering of Pyrolyzed Parts
Pyrolysed body has too low density, much more porosity and poor strength to use. It is
desired to sinter for dense SiC/Si3N4 ceramic parts. In the sintering process, free carbon
derived from pyrolysis of PCS/DVB will react with sintering aids (Al2O3, Y2O3, La2O3, etc.)
to decrease the amount of liquid phase for sintering, and react with SiO2 on the surface of
Si3N4 powder to increase the content of SiC. Those are retarded to densely sinter. Thus
decarbonation is important prior to sintering.
Table 4: Weight loss and density of Sintered body by deferent pretreatment
II - 738
Table 4 shows weight loss and density of Sintered body by different pretreatment before
sintering. The result indicates that deformation and crack had been found in the sintered
specimen which the removal of carbon-rich by oxidating was at 700 C for 120 minutes, but
no defect had been found in the sintered specimens when oxidation condition changed to 500
C for 180 minutes. That is because oxygen reacts not only with carbon-rich, but also with
amorphous SiC derived from PCS. Large amount of SiO2 formed on the surface of pore, in
the sintering process, liquid-phase sintering aid was so heterogeneous distributing that
nonhomogeneously shrinkage of sintering resulted deformation and crack. Otherwise, gradual
change of carbon-rich content from the surface to the middle will influence the well
distributed of composition and microstructure of sintering body. The improvement of relative
density for sintered body which had been treated with NH3 is not pronounced due. to low
efficient of removal of carbon-rich. Density of sintered body which had been treated by
extraction of PCS and removal of carbon-rich in NH3 atmosphere is high due to low content
of SiC, carbon-rich derived from PCS/DVB. Thus the main factors which influenced densely
sintering are the content of PCS/DVB in the green body, SiC and carbon-rich in as-fired body
Effect of sintering temperature on density of the sintered body is shown as Table 5. Increasing
sintering temperature, density of sintered body can be enhanced. Increasing size of blade,
density of sintered body decreased. A multiphase SiC/Si3N4 dense ceramic part, with a
relative density of 93 % and no defect (such as deform, crack etc.), is obtained at 1930 C for
120 minute.
Table 5: Effect of sintering temperature on the density of sintered body
Microstructure and Strength of Sintered Body

Figure 2 shows X-ray diffraction patterns of sintered specimen of PCS/DVB/Si3N4. This
result indicates that -SiC and -Si3N4 grain phase was in-situ formed in the final sample. SiC come from crystallization of amorphous silicon carbide derived from PCS. -Si3N4 come
from transformation of -Si3N4.
II - 739
II - 740
KEY COMPONENTS OF HIGH TEMPERATURE

STABLE HIGH OXIDATION RESISTANT CERAMIC
MATRIX COMPOSITES OF SIC/SIC TYPE WORKING
OUT AND TECHNOLOGICAL WAYS OF THEIR
MAKING
A. M.Tsirlin1, V. G.Gerlivanov1, N. A.Popova1, Yu E. Pronin1,
E. K. Florina1and S. P. Gubin2
1
State Scientific Center of Russian Federation State Scientific-Research

Institute of Chemistry and Technology Organoelement Compounds, 38 Shosse
Entusiastov, 111123 Moscow, Russia.
2
Institute General and Neorganic Chemistry of Russian Academy of Science,
31 Lenin Prospect, 117071 Moscow, Russia.
SUMMARY: In a short form the grounding of the task priority is given. Necessity of
complex key components of ceramic matrix composites (CMC): polymer precursors, fibers
and interface coatings - working out is the important line in this area. Some positive new
results in the chemistry and technology of high temperature stable high oxidation resistant
CMC for civil aviation and automotive industry have been received. The easy production
process of oxygenless polycarbosilane has been developed, high pressure and oxygenated
initiators were not used. A new method to synthesise oxygenless polymetallocarbosilanes
(PMCS) to stabilize ceramic fine structure has been proposed and preliminary checked. New
PMCS having metals in cluster form (nano-particles) without any oxygen links between metal
and silicon atoms are called cluster polymetalocarbosilanes (ClPMCS). They have improved
structure and are cheaper in making also. Preliminary research stages in oxygenless polymer
curing with unsaturated organosilicon compounds and complex interface molybdenum
disilicide - silicon carbide coatings have been carrying out. They have made possible to get
specimen fiber with - 0,3%O2 and 0,1-5% of Ti or Zr and have shown some perspectives in
more economical technology and better oxidation resistance.
KEYWORDS: oxygenless ceramic polymer precursors, cluster polymetallocarbosilanes,

unsaturated curing organosilicons, tetravinylsilane, molybdenum disilicide, interface coating
INTRODUCTION
A high temperature stable high oxidation resistant CMC for a long work at 1500 1700C in
civil aviation and automotive engines is a subject of priority working out nowadays [1]. From
working temperature 1100-1200 (now maximum) to 1500-1700C transition make it possible
to decrease fuel consumption and increase flight distant by 25% and to duplicate specific
thrust. Cooling systems in some cases become unnecessary and in Diesel engine as well
(adiabatic cycle) and so on [2,3]. One of good potential variants is CMC of SiC/SiC type with
coreless SiC fiber. Unfortunately all components of this material now known: polymer
II - 741
precursors, commercial (Nicalon and Tyranno) and pilot (radiation or unsaturated organic
vapour curing) grade fibers, matrix compositions and structures, interface coatings - can not
support work for a long time at temperatures more than 1100C and too expensive. It is
necessary to delete oxygen in all the components, to stabilize fine ceramic structure at high
working temperatures, to work out very high oxidation resistant complex interface coating. It
is very important to depreciate the production of components and the material on the whole.
This aspects are the subjects of many publications and have been intensely discussed at the
last international conferences (ICCM-9, ICCM-10, HT-CVC-1, HT-CMC-II, MICC-93) in
1993-96. There is a wide range of difficult problems to overcome. The authors think
especially important to work out all key components of this material in the frame of one
complex program as for example it is not possible to develop oxygen resistant interface
coating or rational stabilizing methods without well finished oxygenless fiber. This is a base
for their research program and this paper.
MAIN LINES OF RESEARCH AND RESULTS

Polymer Precursors
PCS was first synthesized on the 70-th [4] by pyrolysis and thermal rearranging of
polydimethyl silane (PDMS) in autoclave in inert environment (10 MPa, 470C). This method
is too expensive for industrial production. A lot of attempts to avoid high pressure and PDMS
using through the help of initiators and other organosilicon compounds have not given
positive results as far as PCS quality is concerned. The oxygen is introduced with initiators
and stoichiometry and structure composition of PCS is distorted (Fig.1).
Elution time
Fig.1. MWD of synthesized PCS
1-high pressure, 2-low pressure, boron
initiated, 3- low pressure, titanium initiated,
4- low pressure without initiators
Fig.2. Iron cluster particle distribution by

size in high pressure polyethylene.
Content of Fe,%mas. 1-0,7; 2-1,0;
3-5,0;4-7,0 (model system)
It is the studying of PDMS and PCS syntheses, chemical composition and molecular structure
of initial and final products that have given the optimal regime in the temperature-pressure-
II - 742
time-composition system. This regime makes it possible not to use oxygen containing
initiators and any exotic initial products and to have as good PCS as autoclave one at handy
temperature, pressure and process duration (Table 1). A considerable depreciation of PCS is
reached which as is known becomes then 3- fold much in fibre and (4-5)- fold much in the
composite production.
Table 1: Polymer precursors characteristics
Characteristics
PCS autoclav.
PBCS
PTiCS
PCS(without in.)
ClPMCS
[HSiC3]/
[SiC4]
O2
0,9-0,95
0,8-0,85
0,2-0,3
0,9-0,93
0,95-1,1
0,3 - 0,8
1,0 - 3,0
4,0 - 6,0
0,3 - 0,8
< 0,5
Content, % mass.
H at Si
Init.(metal)
0,65 - 0,75
0,55 - 0,65
0,30 - 0,60
0,60 - 0,70
0,92 - 0,98
0,5 - 1,0
5,0
0,5 - 5,0
Mn
830
1050
890
1060
1220
MWD
Mw
Mz
1310
3010
7300
2070
1830
2290
9100
250000
4830
3580
Ceramic Phase Structure Stabilizers, Polymetallocarbosilanes

As it is generally known effective stabilizers of fine ceramic structure are getero-atoms of
many elements oxygen and nitrogen as well. By-effects such as fusibility, thermal stability,
sintering ability and so on changing are positive using high melting metals (Ti, Zr, Hf, B, Mo
and some other). The problem is working out of a method to introduce the metal into the
polymer providing with the homogeneous distribution just on the synthesis stage and disusing
of oxygen. The first demand was only fulfilled in the practical method of metal-stabilizers
introducing through thetraalkoxi-compounds of metals. The Tyranno - fiber of Ube
Industries (Japan) was made with tetraalkoxititane. But ceramic phase in this fiber is of Si-CO-Ti type as Si-O-Ti link (oxygen bridge) takes place. More over active Si-H - links valuable
for curing are spent in this reaction. So it is necessary to work out essential new way of metal
introducing in the PCS or new method of PMCS synthesis.
The last task is carrying out by the authors on the base of an experience of linear monophase
metallopolymers making: polyethylene, polypropylene, polytetrafluorethylene [5]. It is quite
different from traditional methods to introduce metal particles of 5-100 m in size into
polymer to form homogenized and aggregatively stable but two-phase systems. Two-phase
metallopolymers do not have so high physical and mechanical properties as initial polymer
because their ordering (crystallinity) become lower. They cannot be used for spinning fiber
also.
The new method introduce the metals in cluster form into micro voids which is natural in
every polymer. These voids are connected with polymer molecules packing and located
between crystalline and amorphous parts of micro volumes. Clusters (nano-particles) of
metals are formed in situ as a result of high velocity thermodestruction of organometallic
compound solutions during their input in polymer solutions or melts [6]. Carbonyls, acetates,
phormiates and some other classes of metal compounds are applicable. Fig. 2 gives a notion
of the method possibility. It shows Fe particles size distribution (small angle scattering of Xray (XSAS)) in the high pressure polyethylene at different general metal concentrations. In
the most interesting for ceramic stabilization 1-3% Fe particle sizes is in narrow interval 5-50
. Some testing methods have shown a separate metal phase is absent at these concentrations
and active interaction of metal and polymer is observed.
II - 743
As PMCS are vastly different from linear polymers in structure and thermal properties and
high melting metals are used to stabilize them so it was necessary to use other initial
compounds or synthesize them. Also the technological method of thermodestruction and the
stage of synthesis of cluster PMCS (ClPMCS) to introduce organometallic compound solution
is different for each of them. This compound should have full solubility in synthesis solvent
and be fast and clearly decomposed to metal and volatile ligand. The introduction of metals
on the intermediate stage of syntheses gives the most effective results because they take part
in chemical processes of regrouping and polycondensation. The cyclopentadienyl complexes
and carbonyls of metals seem the best to fulfil these demands in the case of ClPMCS. The Ti
and Zr complexes syntheses were made as follows. On the first stage sodiumcyclopentadienyl
was synthesized from sodium suspension and cyclopentadien in toluene. On the second stage
it was transformed into dicyclopentadienylchlorides of Ti and Zr by reaction with
metalochlorides in toluene or benzene (inert atmosphere):
2C5H6 + 2Na 2C5H5Na + H2
MCl4 + 2NaC5H5 2NaCl + (C5H5)2 MCl2 ,
M = Ti, Zr
[1]
PMCS by the traditional method and ClPMCS through organometallic complexes syntheses
can be illustrated by following general schemes:
PMCS:
CH3
CH3
CH3
CH3
CH3
|
|
|
|
|
(SiCH2)n (SiCH2)m + fM(OC4H9)4 (SiCH2)f (SiCH2)(n-f) ( SiCH2)m
|
|
|
|
|
O
H
CH2
H
CH2
|
(H9C4O)2MO
ClPMCS:
CH3
H3 C CH3
CH3
CH3
|
| |
|
|
-HCl |
(SiCH2)n (SiSi)m + f(C5H5)2MCl2 (SiCH2)q (M) (SiCH2)p
|
| |
|
|
|
H
CH2
H
H3C CH3
[2 ]
[3]
ClPMCS syntheses were carried at 250-280 C and higher dicyclopentadienylchlorides being

introduced into fused oligocarbosilanes (dimetildichlorsilane pyrolysis products) mixture at
the period when about 10% of Si yet stayed in a Si-Si- fragments. After metals introduction
the process speeds up but without any negative by-effect . Si-Si-fragments fully disappear, no
insoluble or high fusible products are formed, only high vacuum topping is necessary. To
study ClPMCS by nuclear -resonance method the model FePMCS was synthesized via
ferrumpentacarbonyl with the same method of synthesis.
The cluster polymers average molecular mass were 1200-2500 (gel-penetration
chromatography), base skeleton (IR-spectroscopy) like the usual PCS formed by Si-C-Si
(1050 and 1350 cm-1) links with hydride SiH (2100 cm-1) and methyl Si-CH3 (1250 and
1400 cm-1) side substituents at Si-atoms. NMR (29Si)-spectroscopy showed high regularity in
ClPMCS structure: groups HSiC3/SiC4 relationship ( = -16 m.f. and 0 m.f.) is about unity,
oxygen containing fragments such as [SiSi(C2)O] and [OSi(C2)O] (= -6,5 and -22 m.f.)
were not discovered. Metals are present in the polymers in the form of cluster particles not
II - 744
more then 10-50 in size in the concentration of 0,1-10% (different initial loading) (XSAS,
X-ray fluorescence). Preliminary data show that till concentration of 3% metal does not
prevent stable spinning of fiber from fused polymer at 250-280 C. Ceramic yield is up to
75%.
As it can be seen in Table 1 the new (cluster) method of PMCS synthesis provides the same
metal introduction without using deficit Si-H - links and more regular polymer structure. It is
possible to suppose the evenly dispersed through the polymer clusters act as active centre of
polymer molecular embranchment without Si-H-links using. This differs from the method [7],
according to which a treating of hidrydesilanes with a catalytic quantity of Ti and Zr complex
compounds lead to crosslinking by Si-H links interaction. Certainly the mechanism of cluster
(nano-particles) action in the polymer matrix of the PCS-type is a question of series studying.
Oxygenless Curing of Polycarbosilanes
Several methods of this process implementation are known. The most practical now with and e- rays treating [8, 9] are difficult (i.e. expensive ) in technology. It is necessary to use
unique equipment, flame- and explosion-proof systems, cumbersome schemes for preparing
and recovering organic containing gas mixtures. Technological difficulties are not small with
the ammonia atmosphere using also [10]. The most valuable from the technological view
point would be methods of curing in the atmosphere of pure nitrogen. These methods are
based as a rule on unsaturated organic or organoelement compounds introducing into PCS.
The heating to the temperature which is higher then the synthesis one cures PCS with this
active unsaturated links help [11]. The difficulties are in provision of nearly ideal (molecular)
dispersion of additives, their action in rather near temperature interval with high selectivity.
Nevertheless this way maybe would be the most effective.
So as the first laboratory results showed that the using of monosilanes with vinyl and allyl
substituents such as tetravinyl-, tetraallyl-, trivinylmetyl-, trivinylfenyl- and other
monosilanes is very attractive. Advantage and peculiarity on the tetravinylsilane example can
be seen from the following scheme:
CH3
CH3
CH=CH2
|
|
2{(CH2Si)n (CH2Si)m } + CH2=CHSiCH=CH2
|
|
|
CH2
H
CH=CH2
CH2=HC CH=CH2
\ /
CH3
CH2 CH 2 Si CH2 CH2
CH3
|
|
|
|
(SiCH2)n (SiCH2)m
(SiCH2)m (SiCH2 )n
|
|
|
|
CH2
CH3
CH3
CH2
[ 4 ]
Unlike other methods there is no gas evolution in this one. It is very important to have dense
and regular structure of the ceramic product. Then as reactivity of vinyl groups in the same
molecule is different they can react with PCS for example in two stages which is useful for
large items. It is handy to cure these items preliminary at not top temperature for some
II - 745
strengthening in the forming equipment. Then it become possible to remove them into high
temperature furnaces to finish the curing and carrying out the heat treating.
An analogous method uses an active amino groups by introducing into PCS some
polysilazanes [11] but it is necessary to optimize a residual nitrogen:
CH3
R
CH3
CH3 R
CH3
_
| | |
|
t, Ar(N2)
|
| | | |
(CH2Si)m + (N-Si-N)p + (SiCH2)k {(H2CSi)mNSiN(SiCH2)k}p [5]
\
/ | \
/
-2H2
|
|
|
R=H, CH3, CH=CH2...
CH3
H H CH3 H H
One more way of oxygenless curing is now in authors working out. That is to use cluster
(nano-particles) of metals not only for structure stabilizing but for curing as well. They
possess a very great surface energy and are possible to tear not only Si-H, but Si-C and C-H
links also and to initiate regroup and polycondensation processes at additional thermal
treating:
H
H3C
CH2
H
H3C CH2
|
| |
|
Cl
|
| | |
CH2SiCH3 + HSiCH2 + H3CSiCH2 CH2SiCH2SiCHSiCH2 [6]
|
|
| (H2, CH4)
|
|
|
CH2
CH2
CH2
CH2 CH2 CH2
A lot of complicated variants of chemical interaction arise of course, so very fine control of
physical and chemical parameters is necessary.
CMC Interface Coatings
This CMC component provides high strength and toughness and long time work at high
temperatures in aggressive environment. Its polyfunctionality is illustrated on Fig.3. Massive
SiC matrix in the air defends itself by a fine selfcuring film of SiO2. But it is not enough to
defend CMC as a whole. On mechanical force influence microcracks arise in the matrix. They
give way for the oxygen into the material right to fiber surface but thin fiber is not possible
with the oxide film to defend. More over they move across the fibber and can cut them
adhesion between fiber and matrix being strong enough. A fast and even catastrophic brittle
destruction of material will follow at rather small load. So it is necessary to have as a
minimum two layers on the fiber surface: dense high oxidation resistant (protective) and
friable or weak adhesive contributing to local defoliation at crack front-fiber surface point
(rheological). The system fiber-coating-matrix components do not as a rule compatible in
thermal expansion coefficients that demands to add in the coating one or more buffer
(compensating) layers. The scheme of Fig. 3 shows that there is no protection from initial
oxygen in the fiber. No good makes the initial oxygen in the matrix also as its interaction with
SiC lead to gas evolution that is harmful for ceramic dense structure:
SiC(s) + 1,5O2(g) = SiO2(s) + CO(g)
SiC(s) + O2(g) = SiO (g) + CO(g)
II - 746
[7]
Attempts to use for high oxidation resistant CMC coatings C, BN, their combinations with
TiC and SiC were not successful [12-14 et al.]. High thermostable oxides - Al2O3, Zr2O3- do
not hold stable dense structure for a long time at cycle load especially. Practically all of high
temperature stable carbides, borides, nitrides, aluminides and silicides are not
thermodynamically compatible with SiC and long time oxidation resistant enough. This is
why the authors selected as a main research and technology way to use molybdenum
disilicide (MoSi2) for protective layers and polymer precursor SiC - for friable rheological
layers.
MoSi2 characteristics are rather well known [15]. The highest oxidation resistance and
thermodynamic and thermal expansion compatibility with SiC among refractory compounds,
2000 C melting point, complementary positive results with W, Ge, Re et al. silicides
alloying. Besides it is cheaper and more handy for machining than SiC and Si3N4. Its
limitations are more density, lower creep characteristic, fast oxidation process at about 500 C
if pour and microcracks are present. These limitations are difficult to overcome in the matrix
and not so difficult in the coating case of using the material. Rheological layer for CMC
SiC/SiC of SiC or Si3N4 is obviously reasonable. Polymer technology provides flexible
process control with type and composition of polymer precursors, degree of dilution, number
of treating cycles. This in no case excludes the CVD and CVI methods using at some
technological stages.
SiO2
FRDWLQJ

PDWUL[
FUDFN
IURQW
GHIROLDWLRQ
]RQH
ILEHU
D
PDWUL[

ILEHU
FRDWLQJ
Fig. 3: CMC structure and cracks spreading scheme

rheological layer, 2- protective layer, 3- buffer layer
The experiments were carried out in two types furnaces: with the direct and indirect heating
and with the cameras of cold and hot walls accordingly (Fig. 4). Metering and heating means
provided 2500 and 1200 C accordingly both from about 1.105 to 1 Pa. Gas mixture preparing
system gave possibility to have mixtures of organometallic compounds and metal chlorides
with the hydrogen, nitrogen and inert gases of 0,01 - 30% in composition.
II - 747
Fig. 4: Furnace scheme for interface coatings applying

1-gas inlet, 2-to vacuum pump outlet, 3- copper water-cooled electrodes, 4- graphite heater,
5- treated sample, 6- silicon carbide heater, 7- steel vessel, 8- vacuum pump
2
3
4
Fig.5: Coating on Nicalon fiber
1,2molybdenum: end view (2980) , side surface (4000), 3,4friable SiClayer (1%
PCS solution, pyrolysis at 1000 0C in inert atmosphere) end view (2200,17600)
Fig.6.Coating on SiC/C fiber

1,2 molybdenum: surface and structure (6750), end view(503), 3,4 two-layer
coating of silicon and molybdenum: side surface (2020), end view(4020)
II - 748
CONCLUSIONS
The results support the authors conception of complex program necessity in working out key
components of high oxidation resistant for long time work CMC for civil technical products.
New methods for production of oxygenless polymer precursors (PCS and ClPMCS) are an
important step in this program. Next one must be the finishing of oxygenless technologically
cheap curing methods that will give real possibility to have cheaper oxygenless fiber and to
carry active work on interface coatings. This chief line does not exclude parallel experimental
work on the all particular subjects of the program.
ACKNOWLEDGMENTS
The authors are very grateful to State Scientific Center of Russin Federation GNIIChTEOS,
Moscow, for providing resources for this work and to International Science and Technology
Center for support in presenting this paper.
REFERENCES
1.
Ho, C.Y. and El-Rahaiby, S.K.,Assessment of the Status of Ceramic Matrix Composites
Technology in the United States and Abroad. Ceramic Engineering & Science
Proceedings, vol.13, 1992, No 78, pp. 317.
2.
Bulter, E.G.,Needs and Market Prospects for Ceramic Fibres in the Aero Engine
Industry, ECCM-6, EuroJapanese Colloquium on Ceramic Fibres, Bunsell, A.R. and
Kimpara, I., Eds., Woodhead Publishing, Ltd., Abington Hall, England, 1993, pp.
81108.
3.
Tsirlin, ..,Nepreryvnye Neorganicheskie Volkna dlya Compozitsionnyh Materialov,

oskva, tallurgiya, 1992, 237 s.
4.
Yajima, S.,Silicon Carbide Fiber, Handbook of Composite, vol.1, Watt, W. and

Perov, B.V., Eds. Elsevier Publishers, New York, 1985, pp. 201237.
5.
Gubin, S.P. i Kosobudskii, I.D.,tallicheskie Klastery v Polimernyh Matritsah,

Zhurnal Uspehi Himii, tom LII, vyp. 8, s. 1350-1364.
6.
Kozinkin, .V., Vlasenko, V.G., Gubin, S.P., Shuvaev, .. i Dubovtsev, I.A. Klastery
v polimernyh matritse. 2. Issledovanie sostava i stroeniya zhelezosoderzhashih klasterov
v polietilenovoi matritse, Zhurnal Neorganicheskie Materialy, 1996, tom 32, 4, s. 422428.
7.
Seyferth, D., Wood, T., Tracy, H. and Robiso J.,NearStoichiometric Silicon Carbide
from an Economical Polysilane Precursor, Journal American Ceramic Society, Vvol.75,
No 5, 1992, pp. 13001302.
8.
Ichikawa, H., Okamura, R. and Seguchi, T.,Oxygen-Free Ceramic Fibers from

Organosilicon Precursors and E-beam Curing, HighTemperature Ceramic Matrix
Composites, vol. II, pp. 6574.
II - 749
9.
Hasegawa, Y.,New Curing Method for Polycarbosilane with Unsaturated Hydrocarbons

and Application to Thermally Stable SiC Fibre, Journal Composites Science and
Technology, No 51, 1994, pp. 161166.
10. Patent, Japan, 03 40,814 (CI D 01 F 9/10), Seguchi, T., Nishii, M., Okamura, K., et al.,
1991.
11. Patent, United States, 4 780 337 (CI B 32 B 9/00), Seyferth, D., 1988.
12. Nair, S.V., Wang, YuLin,Failure Behaviour of a 2-di Woven SiC Fiber/ SiC Matrix
Composite at Ambient and Elevated Temperatures, Ceramic Engineering & Science
Proceedings, vol.13, 1992, No 78, pp. 433441.
13. V., Harkins, B., Beyermann, W., Keiser,J., et al.,Evolution of SiC/SiC composites for
Heat Exchanger Applications, Ceramic Engineering & Science Proceedings, vol.13,
1992, No 78, pp. 503519.
.
14. Yamamura, T., Tyranno Fibers, ECCM-6, EuroJapanese Colloquium on Ceramic
Fibres, Bunsell, A.R. and Kimpara, I., Eds., Woodhead Publishing, Ltd., Abington Hall,
England, 1993, pp. 187202.
15. Journal Material Sciense & Engineering, vol. A 155,1992,No 12, (Papers, Presented at
First High Temperature Structural Silicides Workshop).
II - 750
MICROSTRUCTURE OF THE TZP PARTICULATE

COMPOSITES
Marek Faryna1, Zbigniew Pedzich2, Lidia Litynska1 and Krzysztof Haberko2
1
Institute of Metallurgy and Materials Science, Polish Academy of Sciences,

30-059 Krakow, ul. Reymonta 25, Poland
2
Department of Special Ceramics, Faculty of the Materials Science and Ceramics
University of Mining and Metallurgy, 30-059 Krakow, al.Mickiewicza 30, Poland
SUMMARY: The aim of this paper is to present results of the microstructure investigations
in the tetragonal zirconia polycrystals (TZP) - tungsten carbide system. Microscopic
observations compared with phase composition investigations confirmed that the reaction
between composite components occured. They also revealed close adhering of the matrix and
carbide grains with no discontinuities. Crystallographic correlations between zirconia and
carbides grains was shown.
KEYWORDS: tetragonal zirconia polycrystals, tungsten carbides, particulate composites,
INTRODUCTION
Yttria stabilized tetragonal zirconia polycrystals (Y-TZP) are well known for their good
mechanical properties. These are due to the tetragonal to monoclinic martensitic
transformation at the crack tip advancing through the material. However, still better properties
could be obtained by incorporation WC inclusions into the TZP-matrix. Our previous works
[1,2] pointed out that hardness, fracture toughness, Youngs modulus and, especially, wear
resistance of the material can be improved. It is worth to notice that dense composite bodies
(> 98 % theoretical density) were obtained by the pressureless sintering.
Thermodynamic calculations [3] show that under the potentially necessary sintering
conditions reaction between ZrO2 and WC could not result in the formation of any of the
tungsten oxides. However, the reaction (1)
ZrO2 + 6WC = ZrC + 3W2C + 2CO
(1)
could proceed towards the right hand side if the CO partial pressure would be lower than
indicated in the Table 1.
The samples were sintered in the carbon bed. So, the following reaction had also to be
considered.
ZrO2 +3C = ZrC + 2CO
(2)
II - 751
Table 1: Limiting CO partial pressure (in atms)

Temperature, C
Reaction 1
Reaction 2
1400
0.95
0.054
1500
5.20
0.20
1600
23.8
0.65
1700
93.0
1.89
Composites processing
The 2.8 mole % of Y2O3 - 97.2 mole % ZrO2 solid solution was selected as the matrix
material. The starting powder was prepared by the coprecipitation - calcination technique [4].
The commercially available tungsten carbide powder (Baildon, Poland) was applied.
Homogenization of the oxide and carbide powders was performed by one hour attrition
milling in ethyl alcohol using 2 mm Y-TZP balls as the grinding media. Volume of the
carbide additives was 20 %. Dry powders with no lubricating additives were uniaxially
pressed under 50 MPa and isostatically repressed under 350 MPa. By this procedure
cylindrical samples of 23 mm diameter and 10 mm thickness were prepared. The samples
were placed in a carbon bed and sintered in the argon atmosphere at temperature ranging from
1400 to 1700C with the two hour soaking at each temperature. Sintering was performed in a
furnace equipped with the tungsten heating elements.
Experimental methods
X-ray diffraction was used to establish the phase composition of the bodies sintered at
different temperatures. Scanning electron microscope PHILIP XL30 equipped with an energy
dispersive spectrometer LINK ISIS was used to investigate morphology of the thermally
etched surfaces as well as the chemical composition of the composites. Transmission electron
microscope PHILIPS CM20 equipped with energy dispersive spectrometer LINK eXL was
applied to determine crystalline structure and chemical composition from both TZP grains and
WC inclusions. Thin foils were prepared by ion milling of the samples using the Duo Mill
GATAN 600 equipment.

Figures 1 and 2 show X-ray patterns of the composites sintered at the whole range of
temperature. In the as received surfaces of the bodies (Fig. 1.), ZrC is present even at the
lowest sintering temperature. W2C appears in samples sintered at 1700C. This fact indicates
that ZrC is formed due to reaction (2). The limiting CO partial pressure for the reaction (2) at
1400C is 0.054 atm. Under such conditions reaction (1) should also proceed at 1400C. This
is not the case, plausibly because of the kinetic effect.
Fig. 2. Shows the X-ray diffraction patterns of the samples having thin near surface layer
(~100 microns) removed. It is different from that taken from the as received surface (Fig. 1.).
In the composite sintered at the 1400C only ZrO2 and WC were detected. At higher sintering
temperatures phases not introduced into the starting material, ZrC and W2C appear. They
II - 752
undoubtedly result from the reaction (1). In the composite achieved at the 1700C W 2C is the
dominant carbide phase.
Fig. 1: X-ray diffraction patterns of the as received surfaces of the composites sintered at
indicated temperatures. T - tetragonal zirconia phase
Fig. 2: X-ray diffraction patterns of the composites sintered at indicated temperatures.

Surface layer of about 100 microns was removed by grinding and polishing
SEM micrographs of the composite microstructures are shown in Fig. 3. In the samples
sintered both at 1400C and 1600C the boundaries between zirconia and carbide grains are
distinct and well defined. The increase of the sintering temperature up to 1700C results in the
rims surrounding carbide inclusions. X-ray energy dispersion microanalysis shows that in
these rims both zircon and tungsten are present.
II - 753
Fig. 3: SEM micrographs of the thermally etched composite surfaces. Sintering temperatures:
1400, 1600 and 1700C (starting from top of the figure)
II - 754
TEM observations of the sample sintered at 1400C revealed that not only WC (Fig. 4.) but
also W2C are present (Fig. 5.). The identification of the phases is based on indexing of the
electron diffraction patterns. Diffraction pattern taken from the border zone between ZrO2 and
WC grains (Fig. 4.D.) indicates the existence of the following crystallographic relationships:
[0001] WC
[001] t-ZrO2
[1 010] WC
[010] t-ZrO2
The presence of the W2C grains in material achieved at 1400C confirms that reaction (1)
proceeds at this temperature.
Fig. 4: A. Transmission electron micrograph the TZP-WC composite sintered at 1400C, B.

selected difraction pattern (SADP) of tetragonal zirconia phase, C. SADP of hexagonal WC,
D. SADP of boundary between ZrO2 and WC grains
II - 755
Fig.5: A. Transmission electron micrograph the TZP-WC composite sintered at 1400oC, B.

selected difraction pattern (SADP) of tetragonal zirconia phase, C. SADP of orthorhombic
W2C
In the composite sintered at 1700C W2C is the dominating tungsten carbide phase. Fig. 6.
shows the TEM micrograph in which the tetragonal (Fig. 6B.) and monoclinic (Fig. 6C.) ZrO2
as well as W2C (Fig. 6D.) grains coexist. The presence of martensitic transformation and
modulated structure characteristic for the tetragonal phase (the state just before
transformation) suggests that tetragonal to monoclinic transformation takes place during the
thin section preparation. This has been confirmed by the X-ray diffractometric measurements,
where no monoclinic phase was observed in the as received surfaces. Crystallographic
correlation observed between tetragonal and monoclinic ZrO2:
[100] t-ZrO2
[100] m-ZrO2
[001] t-ZrO2
[010] m-ZrO2
confirms Wen et. al. investigations of the 3Y-TZP system [5].
II - 756
Fig. 6: A. Transmission electron micrograph the TZP-WC composite sintered at 1700oC, B.

selected difraction pattern (SADP) of tetragonal zirconia phase, C. SADP of monoclinic and
tetragonal zirconia, D. SADP of orthorhombic W2C.
The crystallographic correlations between tungsten carbide phases and the matrix has also
been found:
[100] W2C
[001] t-ZrO2
[010] W2C
[010] t-ZrO2
In all sintering temperatures applied in this work close adhering of the matrix and carbide
grains with no discontinuities occured. ZrC distinct in X-ray patterns was not identified by the
electron diffraction. It is probably localized in the rims surrounding tungsten carbides grains.
CONCLUSIONS
Reaction between composite components under applied sintering condition occured. It results
in the W2C and ZrC formation inside the material. Reaction between the zirconia matrix and
the carbon bed is limited only to the near surface layer (~100 microns).
It was revealed that by means of the TEM microscopy carbide/matrix contacts were tight. No
microcracks were observed. A crystallographic correlations between ZrO2 and both WC and
W2C were observed. It indicates that under the applied sintering conditions components tend
II - 757
to arrange in way that a minimum interface energy is achieved. This is substantiated by the
crystallographic arranging of the composite phases.
ACKNOWLEDGEMENT
This work was supported by the Polish State Committee for Scientific Research under the
grant no. 7 T08A 052 10.
REFERENCES
1.
Haberko, K., Pdzich, Z., Piekarczyk, J., Rg, G., Zirconia - Tungsten Carbide
Particulate Composites. Part I: Manufacturing and Physical Properties, in Fourth EuroCeramics, vol. 4, Basic Science - Trends in Emerging Materials and Applications, A.
Bellosi (Eds.), Gruppo Editoriale Faenza Editrice S. p. A., 1995, pp. 29-36.
2.
Pdzich, Z., Haberko, K., Faryna, M., Zirconia - Tungsten Carbide Particulate
Composites. Part II: Microstructure and Wear Resistance, in Fourth Euro-Ceramics, vol.
3, Basic Science - Optimization of Properties and Performance by Improved Design and
Microstructural Control, S. Meriani, V. Sergo (Eds.), Gruppo Editoriale Faenza Editrice
S. p. A., 1995, pp. 433-438.
3.
Haberko, K., Pdzich, Z., Rg, G., Buko, M. M., Faryna, M., The TZP Matrix - WC
Particulate Composites, European Journal of the Solid State and Inorganic Chemistry,
Vol. 32, 1995, pp. 593-601.
4.
Pdzich, Z., Haberko, K., Coprecipitation Conditions and Compaction Behaviour of YTZP Nanometric Powders, Ceramics International, Vol. 20, No. 2, 1994, pp. 85-89.
5.
Wen, S., Ma, L., Guo, J., Yen, T., Transmission Electron Microscopic Observations of
the Martensitic Phase Transformation in Tetragonal ZrO2, Journal of the American
Ceramic Society, Vol. 69, No. 7, 1986, pp. 570-572.
II - 758
FEM ANALYSIS OF MICRO-INDENTATION TEST

FOR C/C COMPOSITES
H. Serizawa 1, S. Sato 2, H. Tsunakawa 3 and A. Kohyama 4
1
Joining and Welding Research Institute, Osaka University,

11-1, Mihogaoka, Ibaraki, Osaka 567, Japan
2 Department of Mechanical Engineering, Fukushima National College of Technology,
30 Nagao, Kamiarakawa, Taira, Iwaki-shi, Fukushima 970, Japan
3 Engineering Research Institute, The University of Tokyo,
Yayoi, Bunkyo-ku, Tokyo 113, Japan
4 Institute of Advanced Energy, Kyoto University, Gokasho, Uji, Kyoto 611, Japan
SUMMARY: To investigate the mechanism of energy loss in load - displacement of indenter

curve (indentation curve) under fiber indentation test of unidirectionally reinforced C/C
composites, the fiber indentation tests were performed and the indentation curve was analyzed
by an axis symmetrical finite-element method (FEM) model including a newly designed fibermatrix interfacial bonding. In this model, only elastic deformations were considered for both
fiber and matrix, and the interfacial bonding was assumed to yield not to fracture. From this
FEM analysis, a hysteresis loop was obtained in the indentation curve and the interfacial
debonding was founded to start from the inside of specimen under the indentation test. And
the indentation curve calculated by the FEM analysis was coincide with the experimental
results, so compressive Youngs modulus of carbon fiber in C/C composite was founded to be
predictable from the FEM analysis.
KEYWORDS: C/C composite, carbon fiber, indentation test, interface, finite element
method, Youngs modulus, transmission electron microscopy, graphite
INTRODUCTION
Mechanical properties of interface between fibers and matrices are known to be quite
influential in determining properties of fiber reinforced composites. Some of the previous
reports have indicated the effectiveness of controlling fiber-matrix interface on improving
properties of composite materials [1]. So the effects of surface treatment or coating on fibers
for several composites have been studied [2] and many experiments have been carried out to
measure the interfacial mechanical properties by various methods, those were pull-out test [3],
push-out test [4], protrusion test [5] and multiple fracture test [6]. And most of these studies
have done by using simple model composites reinforced with single thick fiber. But practical
carbon reinforced carbon (C/C) composites are reinforced with a large number of very thin
fibers and have many pores and cracks in matrix [7]. Therefore, to measure the mechanical
properties of practical C/C composites in the micro area is very difficult and the micro
indentation test is considered to be one of the suitable test method.
There were some reports of the indentation tests about carbon materials [8], but in these
reports maximum load and displacement of indenter were too large to indent only one fiber in
II - 759
the practical C/C composites. So, to study mechanical properties of fiber, matrix and their
interface in C/C composites, the authors have applied a newly developed ultra micro
indentation test machine to fiber indentation tests in C/C composites [9-11]. Where, any
indications of plastic deformation were not observed on specimen surfaces after indentation
test by scanning electron microscopy (SEM), although load - displacement of indenter curve
(indentation curve) showed large hysteresis loop, that indicated the existence of some energy
loss under the test [10]. However this experimental result could not be understood by any
models including the indentation tests of carbon materials as far as the authors knowledge. So,
in this study, fiber indentation tests were performed and a new finite-element method (FEM)
model was provided for analyzing the deformations of fiber, matrix and their interface.
EXPERIMENTAL
Material Used and Specimen
Material used was unidirectionally reinforced C/C composite, heat treated at 1873 K. Carbon
fiber used was mesophase pitch-based fiber and its characteristics are shown in Table 1.
Matrix precursor was a mixture of green coke and phenolic resin (80/20 in volumetric ratio).
And a fiber volume fraction and a density of C/C composite were 45.0 % and 1.69 Mg/m3,
respectively.
Table 1: Characteristics of fibers used
Diameter (m)
Tensile Modulus (GPa)
Tensile Strength (Gpa)
Heat Treatment Temperature (K)
9.8
542
2.92
2473
Schematic flow of specimen preparation for fiber indentation test is shown in Fig. 1. C/C
specimen, whose size was 20l x 5w x 1t mm, was fixed at the center of acrylic pipe by resin.
And next, the fixed specimen was sliced perpendicular to the fiber direction with about 200
m thickness by low speed diamond saw. Then the thickness of sliced specimen was reduced
80 m by mechanical grinding and the pipe and resin were removed in acetone. Finally the
size of specimen for fiber indentation test became 5l x 1w x 0.08t mm.
Fiber Direction
C/C
Sliced by
Low Speed
Diamond Saw
Resin Acrylic Pipe

Specimen
Mechanical
Grinding
about 200m
Thickness
under 100m
Thickness
Inclination Angle of Side Triangle
Remove Resin
and Acrylic Pipe
Specimen Holder
Indenter Tip
Fig. 1: Schematic flow of specimen preparation and shape of indentation tip
II - 760
Fiber Indentation Test

Fiber indentation tests were carried out by utilizing the dynamic ultra indentation test machine
[8]. The specimen was placed on a specimen holder made of SUS and fibers in C/C
composites were loaded by triangular diamond pyramid indenter tip, where maximum loads
were 0.2 N. The shape of triangular diamond pyramid indenter tip is shown in Fig. 1. The
inclination angle of side triangle was 68 degrees.
Structural Inspections
Macroscopic structure, including fiber volume fraction, was determined by both optical
microscopy and SEM. And microstructural inspections of interface were performed with high
resolution transmission electron microscopy (TEM : JEOL-4000FX), where an acceleration
voltage was 400 kV. Specimens for TEM observation were prepared by dimple grinding and
Ar ion milling, and were observed parallel to the fiber direction.
RESULTS OF FIBER INDENTATION TEST

Typical examples of indentation curves under fiber indentation tests are shown in Fig. 2. For
all cases, the indentation curves showed large hysteresis loops and came back to the origin.
Under indentation tests of glasslike carbon, same hysteresis loops were reported [8]. However
maximum load and displacement of indenter were over 10 N and 100 m under the carbon
indentation test, respectively. In addition to that, cracks were observed at the specimen
surface after carbon indentation tests by SEM. But, after the fiber indentation tests, the
interfacial debonding and the indentation mark could not be observed at the specimen surface.
It indicates that the origin of the hysteresis loops under the fiber indentation tests was
different from that under the carbon indentation tests, that is, the loops under the fiber
indentation tests were not resulted from the essential property of carbon.
And micro steps can be detected in Fig. 2 and some of them are pointed out by arrows, those
are considered to be indications of interfacial debonding. So, for the exact understanding of
the indentation curves and the deformations of fiber, matrix and interface in C/C composites
under fiber indentation test, a new FEM analysis including the interfacial debonding seems to
be needed.
0.25
UD-C/C Composites (HTT=1873K)

Specimen Thickness = 80m
Load (N)
0.20
0.15
0.10
0.05
0.00
Displacement of Indenter (m)
Fig. 2: Indentation curves under fiber indentation test in C/C composites
II - 761
FEM MODEL
Shape of FEM Model
By using an axis symmetrical FEM model shown in Fig. 3, the deformations of fiber, matrix
and interface in C/C composite under fiber indentation test were calculated. For the analysis
of practical composites, the correlation between adjacent fibers must be taken into account,
but in this study, two dimensional models were used for the first step of FEM examination
including the interfacial debonding. So C/C composites reinforced with only single carbon
fiber were used for this calculation, and the matrix phase was assumed to be carbon layer. The
width of this layer was assumed to be 2 m, that reflects a fiber volume fraction of specimen
for fiber indentation test correctly (Fig. 4).
Indenter Tip
C/C
Matrix
Fiber
z
Holder
Fig. 3: FEM model of fiber indentation test

Carbon Matrix
Carbon Fiber
Fiber Volume Fraction : 45%

Radius of Carbon Fiber : 5m
C/C Composites
Carbon Layer
Thickness of Carbon Layer : 2m
Fig. 4: Thickness of carbon layer around fiber
II - 762
Under the indentation test, the effect of the area of contact between indenter tip and specimen
on stress distribution was considered to be very large [10]. For this axis symmetrical FEM
model, the shape of indenter tip was a cone not a triangular pyramid. Therefore, in this
analysis, indenter tip was assumed to be a cone with 145.6 degrees of the apical angle so that
the area of contact in case of the cone was as same as in case of the triangular pyramid at the
same displacement of indenter, shown in Fig. 5.
Triangular Pyramid
h
a = 2 3 h tan 68
68
a
b=
h
cos68
r1
Cone
Area of Contact
1
AT = 3 ab = 3 3 h 2 sin 68 tan 68
2
r1 = h tan
r2
2
r2 =
h
cos
Area of Contact
r
AC = r2 2 1 = h 2 sin tan
r2
AT = AC 72.8
Fig. 5: Shape of indenter tip

Interfacial Bonding Between Fiber and Matrix
Fig. 6 shows a typical example of TEM image at interface between fiber and matrix. A
vertical direction of this figure is fiber direction. From this figure, the thickness of interface
seems to be about 30 nm. And a structure of carbon at interface was observed to be graphite
layers parallel to the fiber direction. That is, c-axis direction of this graphite was parallel to
radial direction of the fiber. And, for graphite structure, bonding strength at c-axis direction is
very week because there are van der Walls bonds between graphite layers [7]. So the
debonding of fiber-matrix interface is considered to be as same as the debonding between the
graphite layers.
Fig. 6: TEM image at interface between fiber and matrix
II - 763
In the most previous analysis of micro indentation test, an extra interphase, which was made
from reaction of fiber and matrix, was set up independently [12]. For C/C composites, the
extra interphase seems to be the graphite layer. Now, in this FEM analysis, the graphite layers
were assumed to be springs between nodes of fiber and matrix element in the FEM model.
And debonding was assumed to occur when a length of the spring, that was 30 nm in the
beginning of the FEM analysis, changed to 45 nm, which indicates that the distances between
each graphite layers expand to one and half times. Where stiffness of the spring parallel and
transverse to the fiber direction were determined to 50 GPa and 20 GPa, respectively, from
mechanical properties of graphite [13].
And two kinds of the stress-transmission method in the spring were assumed to be as shown
in Fig. 7 after the debonding at fiber-matrix interface. In Fig. 7, the case 1 is a usual method
of the interfacial debonding on FEM analysis under the fiber indentation test. Under the
condition of case 1 on the axis symmetrical FEM model, a debonding at one point on fiber
surface seems to be equivalent to a debonding on one circumference of fiber surface including
the debonding point. But this phenomena is considered not to exactly express a three
dimensional phenomena of the debonding at fiber-matrix interface, because under the three
dimension the debonding at one point on fiber surface seems to be gradually spread on the
circumference including the debonding point. So a newly designed interfacial bonding was
constructed shown in case 2 of Fig. 7. Under the condition of case 2, the bonding force of the
spring between fiber and matrix was assumed to linearly decrease after the beginning of the
interfacial debonding, that is considered to indicate the gradual spread of the interfacial
debonding. That is, the interfacial bonding was presumed to yield not to fracture. And the
slope of the bonding force in an unloading under the fiber indentation test was assumed to be
as same as that in loading, because the bonding between fiber and matrix was considered to be
supported by the residual bonding interface that has the same stiffness in both loading and
unloading.
Connection between Fiber and Matrix : Spring
Case 1 : Spring Fracture Model
Connection Break

Force=0, Stiffness=0
Force
Case 2 : Spring Yield Model

Connection Break

Spring Yield
Displacement
Force
Displacement
Fig. 7: Condition of debonding at fiber-matrix interface

Condition of FEM Model
After the fiber indentation test, the interfacial debonding and the indentation mark could not
be observed at the specimen surface by SEM. So, in this FEM analysis, only the elastic
deformations of fiber, matrix and C/C composites were considered. That is, the origin of the
large hysteresis loop in Fig. 2 was assumed to be only an energy loss caused by the fibermatrix interfacial debonding. The specimen holder and the indenter tip were presumed to be
rigid. The condition of FEM analysis is as shown in Table 2. The Youngs moduli of C/C
II - 764
composite were determined from the measured values [14], except to the longitudinal
Youngs modulus of unidirectional C/C composites. The radial Youngs modulus of fiber and
the isotropic Youngs modulus of matrix was assumed to be same as the transverse Youngs
modulus of unidirectional C/C composites. And the Poissons Ratio and the shear modulus
were determined from the previous researches [15] except for xy and zx of fiber, which was
decided in reference of the Poissons Ratio of composite and fiber.
On the other hand, the carbon fiber was considered to be mainly received the compressive
stress at the fiber direction under the fiber indentation test, and the compressive Youngs
modulus of fiber has been reported to be the quarter of the tensile Youngs modulus [16]. And
the longitudinal Youngs modulus of fiber are known to have the strain dependence. So the
longitudinal Youngs modulus of carbon fiber was assumed to be the quarter of the calculated
value from the longitudinal Youngs modulus of unidirectional C/C composites [14]
according to the simple law of mixture, and the longitudinal Youngs modulus of C/C
composites was determined from this Youngs modulus of fiber.
Table 2: Condition of FEM model
Analysis Model
Axis Symmetrical FEM Model
Specimen Size :
Radius of Fiber :
Indenter Tip :
Radius 2500 m, Thikness 80 m

5 m
Cone with 145.6 of the Apical Angle
@
Thickness
: 2 m (from Fiber Volume Fraction)
of Carbon Layer
Elastic Properties of Each Element
Exx: 80GPa xy:0.16 Gxy:10GPa
Fiber :
Eyy: 8.9GPa yz:0.30 Gyz:20GPa
Ezz: 8.9GPa zx:0.16 Gzx:10GPa
Density 2.08 Mg/m 3
Carbon Layer :
E: 8.9GPa : 0.30
Density 1.70 Mg/m 3
C/C :
Exx: 40GPa xy:0.20 Gxy:8.9GPa

Eyy: 8.9GPa yz:0.30 Gyz:15GPa
Ezz: 8.9GPa zx:0.20 Gzx:8.9GPa
Density 1.68 Mg/m 3
RESULTS OF FEM ANALYSIS

In this FEM analysis, only a displacement of indenter can be controlled. So the FEM analyses
with the two kinds of interfacial bonding, that were case 1 and 2, were done in loading from 0
to 2.5 m displacement of indenter and in unloading from 2.5 to 0 m displacement. The
indentation curves calculated from FEM model with the interfacial bonding of case 1 and 2
are shown in Fig. 8 (a) and (b), respectively. In Fig. 8 (a), the load was largely decreased over
0.6 m displacement of indenter, and the interfacial debonding started after the 0.6 m
displacement in the FEM model with the interfacial bonding of case 1. But the large decrease
of load was not detected in the experimental results. On the other hand, for the FEM model
with the interfacial bonding of case 2, the load was continuously increased in loading
although the interfacial debonding began after the 0.6 m displacement. That is, the
indentation curve of case 2 showed the hysteresis loop as same as the experimental results. So
II - 765
the FEM model with the interfacial bonding of case 2 is considered to be suitable for the
analysis of the fiber indentation test of C/C composites.
And, in this FEM analysis of case 2, the start point of the interfacial debonding was the inside
not the surface of specimen. To put it in the concrete, the interfacial debonding started at the
interface under 5 m depth from the upper side of specimen surface. And the area of
interfacial debonding gradually expanded into the lower side of specimen surface with
increasing the displacement of indenter, but the interfacial debonding at both the upper and
lower sides of specimen surface did not occur at the 2.5 m displacement of indenter. This
result from the FEM analysis was in good agreement with the result by the SEM observation
of specimen surface after the fiber indentation test. Furthermore, in this FEM analysis of case
2, the maximum value of load at the 2.5 m displacement of indenter was close to the
experimental results. Therefore the compressive Youngs modulus of carbon fiber in C/C
composite was founded to be predictable from the FEM analysis with a newly designed
interfacial bonding.
0.30
0.30
(a) UD-C/C Composites
0.25
(Fiber Indentation Test)
0.20
0.15
Load (N)
Load (N)
0.25
Calculated Data : Case 1
(b) UD-C/C Composites

(Fiber Indentation Test)
0.20
0.15
0.10
0.10
0.05
Calculated Data : Case 2
0.05
0.00
0.0
0.5
1.0
1.5
2.0
2.5
0.00
0.0
0.5
1.0
1.5
2.0
2.5
Fig. 8: Indentation curves calculated from FEM model with the interfacial
bonding of case 1 (a) and case 2 (b)
CONCLUSIONS
To investigate the mechanism of the energy loss in the indentation curve under the fiber
indentation test of C/C composites, the fiber indentation tests were performed and the
indentation curve was analyzed by an axis symmetrical FEM model including a newly
designed fiber-matrix interfacial bonding . The conclusions could be summarized as follows.
(1)
From the FEM analysis, the hysteresis loop was obtained in the indentation curve and
the interfacial debonding was founded to start from the inside of specimen under fiber
indentation test.
(2)
The indentation curve calculated by this FEM analysis was coincide with the
experimental results, so the compressive Youngs modulus of carbon fiber in C/C
composite was founded to be predictable from this FEM analysis.
II - 766
ACKNOWLEDGMENTS
The authors would like to express their sincere appreciations to Dr. E. Tsushima and Mr. J.
Takayasu (TONEN Co.) for providing materials and to Mr. M. Fukunaga (Nippon MARC
Co., Ltd.) for fruitful discussion.
REFERENCES
1.
Leparoux, M., Vandenbulcke, L., Goujard, S., Robin-Brosse, C. and Domergue, J.M.,
Mechanical Behavior of 2D-SiC/BN/SiC Processed by ICVI, Proceedings Tenth
International Conference on Composite Materials, Whistler, British Columbia, Canada,
August 14-18, 1995, Vol. IV: Characterization and Ceramic Matrix Composites,
Poursartip, A. and Street, K.N., Eds, pp. 633-640.
2.
Grande, D.H., Mandell, J.F. and Hong, K.C.C., Fiber-Matrix Bond Strength Studies of
Glass, Ceramic, and Metal Matrix Composites, Journal of Materials Science, Vol. 23,
1988, pp.311-328.
3.
Lawrence, P., Some Theoretical Considerations of Fibre Pull-Out from an Elastic

Matrix, Journal of Materials Science, Vol. 7, 1972, pp.1-6.
4.
Marshall, D.B., An Indentation Method for Measuring Matrix-Fiber Frictional Stresses

in Ceramic Composites, Journal of the American Ceramic Society, Vol. 67, 1984,
pp.C-259-C-260.
5.
Kagawa, Y. and Honda, K., A Protrusion Method for Measuring Fiber/Matrix Sliding
Frictional Stresses in Ceramic Matrix Composites, Ceramic Engineering and Science
Proceedings, Vol. 12, 1991, pp. 1127-1138.
6.
Marshall, D.B. and Evans, A.G., Failure Mechanisms in Ceramic-Fiber/CeramicMatrix Composites, Journal of the American Ceramic Society, Vol. 68, 1985, pp. 225231.
7.
Thomas, C.R., Essentials of Carbon-Carbon Composites, The Royal Society of

Chemistry, 1993.
8.
Sakai, M., Hanyu, H. and Inagaki, M., Indentation-Induced Contact Deformation and
Damage of Glasslike Carbon, Journal of the American Ceramic Society, Vol. 78, 1995,
pp.1006-1012.
9.
Hamada, K., Sato, S., Tsunakawa, H. and Kohyama, A., Interfacial Microstructure and
Mechanical Properties of C/C Composites, Proceedings Tenth International
Conference on Composite Materials, Whistler, British Columbia, Canada, August 1418, 1995, Vol. VI: Microstructure, Degradation, and Design, Poursartip, A. and Street,
K.N., Eds, pp. 423-430.
10.
Serizawa, H., Kohyama, A., Watanabe, K., Kishi, T. and Sato, S., Elastic FEM
Analysis of Fiber Push-Out Test for C/C Composites, Materials Transactions, JIM,
Vol. 37, No. 3, 1996, pp. 409-413.
II - 767
11.
Watanabe, K., Kohyama, A., Sato, S., Serizawa, H., Tsunakawa, H., Hamada, K. and
Kishi, T., Evaluation of Interfacial Shear Strength of C/C Composites by means of
Micro-Indentation Test, Materials Transactions, JIM, Vol. 37, No. 5, 1996, pp. 11611165.
12.
Tsai, H.C., Arocho, A.M. and Gause, L.W., Prediction of Fiber-Matrix Interphase
Properties and their Influence on Interface Stress, Displacement and Fracture Toughness
of Composite Material, Materials Science and Engineering, A126, 1990, pp. 295-304.
13.
Kelly, B.T., Physics of Graphite, Applied Science Publishers, London and New Jersey,
1981.
14.
Serizawa, H., Kohyama, A. and Sato, S., Effect of Heat Treatment Temperature on
Youngs Modulus and Internal Friction of C/C Composites, Proceedings of the Ninth
International Conference on Composite Materials, Madrid, Spain, July 12-16, 1993,
Vol. 2, Miravete, A., Ed., pp. 524-531.
15.
Goggin, P.R., The Elastic Constants of Carbon-Fibre Composites, Journal of

Materials Science, Vol. 8, 1973, pp. 233-244.
16.
Shinohara, A.H., Sato, T., Saito, F., Tomioka, T. and Arai, Y., A Novel Method for
Measuring Direct Compressive Properties of Carbon Fibers Using A Micro-Mechanical
Compression Tester, Journal of Materials Science, Vol. 28, 1993, pp. 6611-6616.
II - 768
PREPARATION AND PROPERTIES OF 3D-CARBON

FIBER WOVEN PREFORM REINFORCED SIC AND
AL2O3 HYBRID MATRIX COMPOSITES
Jiayu Xiao, Fengrong Liu, Wenwei Zheng, Xingye Wang, Dazhi Jiang, Zhaohui Chen
Department of Materials Engineering and Applied Chemistry,
National University of Defence Technology, NUDT, Changsha, Hunan, 410073, P.R. China
KEYWORDS: carbon fiber, woven preform, SiC, A12O3, hybrid matrix composite,
precursor, sol-gel
II - 769
II - 770
II - 771
II - 772
II - 773
II - 774
CONTINUOUS FIBRE REINFORCED COMPOSITES

WITH MULLITE MATRIX DERIVED FROM
COLLOIDAL PRECURSORS
Jaili Wu, Frank R. Jones and Peter F. James
Department of Engineering Materials, Sir Robert Hadfield Building, Mappin Street
University of Sheffield, Sheffield S1 3JD England UK
SUMMARY: In this paper, the preparation of carbon and SiC continuous fibre reinforced
composites with mullite matrix derived from colloidal-size alumina powder and silica colloids
is described. The prepregs was prepared by one infiltration of a fibre tow with high solids
yield colloidal sols of 40 vol% containing an oxide composition corresponding stoichiometric
mullite, 3Al 2O3 2SiO 2 . The composites were consolidated by hot-pressing at 1550 C for 0.5
h at 15 MPa to form dense mullite matrix. The microstructures of the unidirectional
composites were characterised by XRD and SEM. The structure of the interfaces was
examined by TEM. The properties of the composites: flexural strength, modulus and work of
fracture were measured by three point bending.
KEYWORDS: SiC, carbon, continuous fibre, colloidal, mullite, matrix, composites,

unidirectional, mechanical properties, interface
INTRODUCTION
Continuous fibre reinforced ceramic matrix composites have been extensively investigated for
the improvement of the strength and toughness of ceramics, which can be used as structural
materials at high temperatures. Densification of the ceramic matrix and control of the
interface between fibre and matrix are vital in the production of composites with useful
properties [1, 2]. Mullite combines excellent strength and creep resistance at high
temperatures with good chemical and thermal stability and a low thermal expansion
coefficient of 4.5-5.610 6 K 1 [3]. Therefore, mullite-containing composites have a potential
as a high performance material [4]. In previous work [5-7] it was shown that a sol containing
40% solids by volume, of the mullite composition, 3Al 2 O 3 2SiO 2 , could be prepared using
-alumina powder and colloidal silica sol. This sol was transformed into a gel from which a
mullite monolith was obtained by pressureless sintering at 1600C for 2 h. A density of 3.10
10 3 kg / m 3 for the bulk sample was obtained, which corresponds to 98% of the theoretical
density of 3.1710 3 kg / m 3 for mullite. The microstructures of the bulk materials have been
characterised and the densification mechanism analysed [6, 7]. It was demonstrated that a
unidirectional carbon fibre reinforced crystalline mullite matrix composite could be fabricated
by one infiltration with this sol and consolidated by hot-pressing [5].
In this paper, the fabrication of composites using colloidal precursors for a mullite matrix is
described, in which either high strength carbon or Nicalon silicon carbide fibres were
II - 775
employed as the reinforcement. The matrix was stoichiometric mullite, 3Al 2 O 3 2SiO 2 . The
microstructures of the composites were characterised by X-ray diffraction (XRD) and
scanning electron microscopy (SEM). The fibre-matrix interfacial characteristics were
examined by TEM and related to the properties, fracture behaviour and processing conditions
of the composites.
EXPERIMENTAL
Prepregs of the unidirectional fibre reinforced mullite matrix composite samples were
prepared by the single-stage infiltration of the fibre tows with a colloidal sol corresponding to
an oxide composition of 3Al 2 O 3 2SiO 2 , which has been described elsewhere [5], using
Ludox colloidal aqueous silica sol (Dupont AS40, 40 wt% of silica particles of 5-50 nm in
diameter) and -alumina powder (Alcoa A1000 SG, with an average size of 0.5 m) as
starting materials. Briefly, the as-received -alumina powder was dispersed in Ludox
colloidal silica sol to produce a homogenous sol. The viscosity of the sol was sufficiently low
for the efficient impregnation of the fibre tow. The high solids concentration (alumina-silica
mixture) of 40 vol% (68 wt%) enabled single-stage infiltration to be employed for the
formation of a dense matrix.
Grafil LXA unsized high strength continuous carbon fibre tow has a diameter of 8 m, a
strength of 4 GPa, an elastic modulus of 224 GPa, a ultimate strain of 1.78% and an average
density of 1.8110 3 kg / m 3 . The Nicalon-207 silicon carbide is a polyvinyl alcohol sized
continuous fibre yarn which has a strength of 2.6 GPa, an elastic modulus of 189 GPa, a
ultimate strain of 1.37%, an average density of 2.5510 3 kg / m 3 and a diameter of 14 m.
Nicalon-207 SiC polyvinyl alcohol sized fibre yarn was de-sized by passing through hot water
at 90 C, and then stored in vacuum ready for the preparation of the composites.
The composite prepregs were fabricated by infiltrating the continuous fibres using the above
sols without additional binder in a single stage. The prepreg was dried in air at room
temperature for 48 h and in an oven at 90C for a further 48 h. By careful control of the solids
concentration, composites with a fibre volume fraction in the range of 30-60% could be
prepared. In the graphite die assembly, the composite stack of prepregs was sealed by two
thick layers of graphite powder to ensure that a reducing atmosphere of CO existed during
hot-pressing. The composite samples were consolidated by hot-pressing at 1550C, for 0.5 h
at 15 MPa for the formation of the mullite matrix. All hot-pressed composite samples were
allowed to cool naturally to room temperature overnight within the furnace.
The densities of the composite samples, c were determined by mercury porosimetry using a
Micromeritics Pore Size 9320 with a pressure range of ~174 MPa. The volume fraction of
fibres in the composites and the crack spacing in the polished surface of the composites were
determined by computer-controlled image analysis using an AMS OPTOMAX V. The
maximum theoretical density of each composite (assuming no porosity), 0c was calculated
using the following equation:
0c = Vmm + Vf f
II - 776
(1)
where m is the theoretical density of matrix; f is the theoretical density of fibre; Vf is the
measured volume fraction of fibre in composite; and Vm is the measured volume fraction of
matrix in composite. The phases present in the matrix of the composites were identified by
XRD on a Philips 1710 X-ray Diffractometer with the scanning speed of 2 (2)/min. SEM
was performed on a JEOL JSM6400. TEM was performed on a Philips EM400T. Thin
sections for TEM were prepared by ion beam milling a specimen which had been
mechanically dimpled to less than 30 m in thickness using fine grinding paste.
The flexural strengths, flexural moduli, and work of the fracture of the composites at room
temperature were determined in three point bend on a Mayes Universal Tester at a
displacement rate of 0.2 mm/min. The span of the lower support pins was 30 mm, and a span
to specimen thickness ratio greater than 20 was employed. Rectangular test specimens cut
from hot-pressed cylindrical composite samples were ground and polished to the dimensions
of 37 mm in length, 2.0 mm in width and 1.5 mm in thickness. The load-deflection curve for
the composites was recorded to fracture. The mechanism of failure was identified by visual
examination after the test and by SEM examination of the fracture surface of tested
specimens.
The flexural strengths and moduli were determined according to British Standard Methods of
Testing (10). The work of fracture was calculated from the area under the load-deflection
curve. The ultimate failure strain, , was calculated by an approximate eqn. (7).
RESULTS
Microstructures of Composites
Figure 1 shows the XRD pattern for the hot-pressed carbon fibre composites hot-pressed at
1550 C, for 0.5 h at 15 MPa. The composite matrix is predominantly crystalline mullite,
although a trace of unreacted alumina is evident. For the SiC/M composite hot pressed under
the same conditions as the C/M composite, XRD gave a similar result as in Fig. 1. Fig. 2
shows a polished section through the C/M composite approximately parallel to the fibre
direction. A dense mullite matrix with a uniform distribution of carbon fibres (Vf =0.52) was
revealed. No large pores were observed, which was consistent with the measured density of
95% of the calculated theoretical density of 2.4610 3 kg / m 3 for the composite. Mechanical
polishing resulted in some degree of damage or "pluck-out" at the fibre-matrix interface
(white areas indicated by arrows) suggesting that little or no chemical bonding had occurred.
Some regularly spaced microcracks were observed in the matrix, which are perpendicular to
the fibre direction in fig. 2, which are believed to be induced by thermal mismatch between
the fibres and mullite matrix. Thermally induced cracks are produced by stresses resulting
from a mismatch in the expansion coefficients of the carbon fibres and the mullite matrix on
cooling from the high processing temperature to room temperature. The measured thermal
expansion coefficient is 4.8310 6 K 1 for the bulk samples of mullite and that of the carbon
fibres is about zero [8]. The thermally induced cracks and the interfacial pores are believed to
be the main contributors to the 5% porosity within the C/M composite.
The thermal expansion coefficient for the SiC fibres at 4.0-5.010 6 K 1 [8] is similar in
magnitude to that of mullite. Thus, the potential for thermal mismatch cracking in the
SiC/mullite composites is much lower as suggested by the observation of few cracks in a
II - 777
random polished section through the SiC/M composite (not shown). A polished section
perpendicular to the fibre direction (Fig. 3) shows that some fibres were in contact.
Fibre/matrix Interfaces in Composites
The interfaces between the fibre and matrix in the composites were examined by TEM as
shown in Fig. 4. A sharply defined interface was observed between the carbon fibre and the
mullite matrix in the C/M composite (not shown). Some crevices appear to occur along the
interface which would further weaken the interfacial bonding. The crevices are believed to be
the consequence of the thermal expansion mismatch between fibre and matrix during cooling.
No obvious chemical interaction between fibre and matrix was evident. Some mullite grains
near the fibre surface were observed to have grown parallel to the fibre direction. Dislocations
were also observed in the mullite grains or along grain boundaries. Evidence for a thin
continuous interfacial layer between the SiC fibres and the mullite matrix in the SiC/M
composite was revealed by TEM (Fig. 4). Energy dispersive spectroscopy (EDS) [7] indicated
the layer mainly consisted of silica and carbon (amorphous or crystallised). It is possible that
the SiC fibre degraded to form a silica layer at high temperatures.
Densification and Mullitisation of Composite Matrices
Highly densified composites were produced including C/M, and SiC/M which have densities
ranging from 95 to 98% of the theoretical density (Table 1). Glass viscous sintering is the
main matrix densification mechanism as discussed previously (5-7). From XRD and TEM
analysis, substantial mullitisation (approximately 90% or more by volume) could be achieved
in the matrix. Complete mullitisation was not achieved at 1550 C for 0.5 h because of the
slowness of the interdiffusion processes [9].
During the three point bending test, most of the composites initially showed an elastic
response with deflection increasing linearly with increase in load, followed by an extended
non-elastic regime. Generally after the maximum value of load was reached, the degree of
subsequent extension was strongly dependent on the nature of the interface between fibre and
matrix. Figure 5 is representative examples of the stress vs. deflection curves for the carbon
fibre mullite composite, C/M and SiC fibre mullite composite, SiC/M, compared with typical
results for the matrix materials alone (dotted lines). A summary of hot-pressing parameters
together with the properties of the composites is given in Table 1. Vf is the measured fibre
volume fraction, c / c0 the ratio of measured to theoretical density expressed as a percentage,
the ultimate flexural strength, E the flexural modulus, W the work of fracture, and the
ultimate failure strain. c and Ec are the ultimate tensile strength and the modulus of the
composites calculated theoretically according to theory [11]. In the flexural test, employed
here, only the outside surface experiences a maximum stress. All failures were initiated on the
tensile face of the specimen.
Referring to Table 1, the highest average flexural strength (737128 MPa) was observed for
the carbon fibre reinforced mullite matrix composite hot pressed at 1550 C for 0.5 h at 15
MPa (C/M). This composite also exhibited the highest work of fracture, 482122 kJ / m 2 , and
ultimate failure strain of 0.370.13%. These values were considerably higher than those for
unreinforced bulk mullite of 150 MPa for , 5.18 kJ / m 2 for work of fracture and 0.001% for
[11]. Samples of bulk mullite were prepared by pressureless sintering of gels of -alumina
and Ludox silica at 1600 C for 2 h to 3.110 3 kg / m 3 , approximately 98% of the theoretical
II - 778
density for mullite, 3.1710 3 kg / m 3 [9]. The SiC fibre mullite matrix composite (SiC/M) hot
pressed at 1550 C for 0.5 h also had a high strength, of 63098 MPa, W of 293111
kJ / m 2 , and of 0.300.1%.
Fracture Surfaces Observations
The nature of the fibre/matrix interface can be also revealed by examining the fracture
surfaces of composite specimens after the three point test. Extensive fibre pull-out was
observed in the fracture surfaces of both carbon fibre and SiC fibre reinforce mullite matrix
composites. A typical fracture surface of SEM micrograph from SiC composite is shown in
Fig. 6.
DISCUSSION
Strengths and Ultimate Failure Strains
In comparison with the bulk unreinforced matrix materials, all the composites demonstrated
an enhanced flexural strength and a change in their fracture behaviour. The flexural strength
of these materials correlate directly with the quality of the interfacial bond strength. As
expected those materials with a weak interface exhibited tough behaviour (Table 1). However,
there is a discrepancy between the experimental values and the theoretical predictions based
on the mixture laws (Table 1). The experimental values of strength are lower than the
theoretical predictions even assuming that there are no contributions from the matrices. Thus,
other parameter needs to be considered. When exposed to high temperatures, carbon fibres
oxidise, and the composition and structure of SiC fibres change. Consequently the retained
strength of the fibres is affected and also the strength of the composite. Nicalon-SiC fibres
generally show reaction even in a reducing atmosphere [12, 13] as follows:
SiO x C y (SiC fibre ) SiO 2 + C
(2)
The reduction in strength above 1000 C is caused by microcrystallisation of the fibre

structure [12], although the degradation can be alleviated by employing reducing atmospheres
[13]. Both a silica phase and the presence of carbon have been detected in the interfacial
regions within the SiC/M composite using EDS [7], which are believed to be the products of
the above reaction. A severe reduction in the strength of the SiC composites because of the
high processing temperatures was also observed, however the longer processing time at 1550
C was not necessary to lead to a reduction in the SiC fibre strength according to the Raman
spectroscopy analysis on the SiC fibre composite specimen [14].
Interfaces, Fracture Behaviour and Fibre Pull-out
A glassy silica phase in the interfacial regions between the SiC fibres and matrices of the
SiC/M was observed by TEM and defined by EDS. Silica is expected to form on the fibre
surface as a product of reaction in equ. 2. A weak interface was observed for the SiC
fibre/mullite composite, SiC/M hot-pressed for 0.5 h as shown by the extensive fibre pull-out
after the bending test (Fig. 6). In this case the interfacial glass layer between the fibres and
matrix in the composite may not be sufficiently thick to form a strong bond.
II - 779
The influence of the porosity of the composites on the nature of the interface should also be
considered. From SEM and TEM observations, some pores and crevices were detected along
the interfacial region between the fibres and the mullite matrix within C/M. These pores may
have resulted partly from incomplete densification and partly from the volume change during
the mullitisation in the matrix, and the crevices may have formed as a result of a relative
displacement between fibre and matrix on cooling from high temperature because of thermal
expansion mismatch. These pores and crevices could contribute to the weak interfacial
bonding, resulting in extensive fibre pull-out on failure (Fig. 5) with a substantial failure
strain of 0.370.13%.
CONCLUSIONS
Carbon fibre and SiC fibre continuous reinforced mullite matrix composites were successfully
fabricated by colloidal processing, using -alumina powder and colloidal silica sol as
precursors by one stage infiltration followed by hot-pressing at 1550 C. Highly densified
composites were obtained.
The microstructures of the composites were examined by SEM and TEM. Parallel cracks
were observed in the carbon fibre reinforced composites, resulting from thermal expansion
mismatch. For the C/M composite a sharply defined interface with little sign of interaction
was observed between carbon fibres and mullite matrix, and some crevices were also
observed along the interface. A highly densified and nearly crack-free mullite matrix was
observed in the SiC/mullite composite, SiC/M. For this composite a thin continuous
interfacial silica layer was also observed by TEM and EDS, which may result from reaction at
the SiC fibre surface at high temperatures.
Using the three point bending test, the properties of the composites were determined. The
highest flexural strength of 737128 MPa, with the highest work of fracture, 482122
kJ / m 2 , and ultimate failure strain of 0.370.13% were achieved for the C/M. The SiC/M also
had a high of 63098 MPa, W of 293111 kJ / m 2 , of 0.300.1%. The fibre
reinforcement to mullite is highly effective.
The amount of silica phase along the interface between SiC fibre and matrix is thought to be
the main factor controlling the interfacial bonding strength in the SiC fibre composites. The
thin silica layer at the interface as shown by TEM and EDS for the SiC/M composite may be
account for the relatively weak interfacial bonding observed.
REFERENCES
1.
2.
3.
4.
Sheppard, L.M. "Enhancing Performance of Ceramic Composites," Am. Ceram. Soc.

Bull., Vol. 71 1992, pp617.
Cornie, J.A., Chiang, Y.M., Uhlmann, D.R., Mortensen, A. and Collins, J.M.,
"Processing of Metal and Ceramic Matrix Composites," Am. Ceram. Soc. Bull., Vol. 65
1986, pp293.
Hawkes, W.H., "The Production of Synthetic Mullite," Trans. Brit. Ceram. Soc., Vol.
61 1962, pp689.
Sacks, M.D., Lee, H.W. and Pask, J.A., "A Review of Powder Preparation Methods and
Densification Procedures For Fabricating High Density Mullite," Ceram. Trans. Vol. 6.
II - 780
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
Mullite and Mullite Matrix Composites, Edited by S. Somiya, R.F. Davis, J.A. Pask,
1990, pp167.
Wu, J., Jones, F.R. and James, P.F., "Mullite Matrix Continuous Fibre Reinforced
Composites by Sol-gel Processing", Ceram. Trans., Vol. 46: Advances in Ceramic
Matrix Composites II, 1995, pp177.
Wu, J., Chen, M., Jones, F.R. and James, P.F., "Characterisation of sol-gel derived
alumina-silica matrices for continuous fibre reinforced composites", J. Euro. Ceram.
Soc. Vol. 16 1996, pp619.
Wu, J., PhD thesis, University of Sheffield 1995.
Prewo, K., "Fiber-Reinforced Ceramics: New Opportunities for Composite Materials",
Am. Ceram. Soc. Bull., Vol. 68 1989, pp395.
Pask, J.A., Zhang, X.M., Tomsia, A.P. and Yoldas, B.E., "Effect on Sol-Gel Mixing on
Mullite Microstructure and Phase Equilibria in the -Al 2 O 3 SiO 2 System", J. Am.
Ceram. Soc., Vol. 70 1987, pp704.
British Standard Methods of Testing, "Method 1005, Determination of flexural
properties--Three point method", (British Standards Institution) BS 2782: Part 10:
Method 1005, EN63 1977.
Kelly, A., Strong Solids, Clarendon Press, Oxford, 2nd Ed, 1973.
Simon, G. and Bunsell, A.R., "Creep behaviour and structural characterisation at high
temperatures of Nicalon SiC fibres", J. Mater. Sci., Vol. 19, 1984, pp3658.
Bender, B.A., Wallace, J.S. and Schrodt, D.J., "Effect of thermochemical treatments on
the strength and microstructure of SiC fibres", J. Mater. Sci., Vol. 26, 1991, pp970.
Wu, J. and Colomban, Ph., "Raman spectroscopy study on continuous fibre reinforced
ceramic matrix composites", in J. Raman Spectrosc. to be published..
Table 1: Summary of the hot-pressing conditions and properties of composites

Hot-pressing
parameter
Fibre/Matrix
Number tested
Vf (2%)
c / c0 (%)
(MPa)
E (GPa)
W (kJ / m 2 )
c (GPa)
Ec (GPa)
: standard deviation
1550 C 0.5 h 15
MPa
SiC/M
6
60
1550 C 0.5 h 15
MPa
C/M
6
52
98
63098
20811
293111
1.6
149
95
737128
12764
482122
2.2
160
II - 781
Figure 1: XRD of mullite matrix carbon fibre reinforced composites of C/M hot-pressed at
1550 C for 0.5 h at 15 MPa
Figure 2: SEM of polished section parallel to the fibre direction through the C/M composite
hot-pressed at 1550 C for 0.5 h at 15 MPa
II - 782
Figure 3: SEM of polished cross section perpendicular to the fibre direction through the
SiC/M composite hot-pressed at 1550 C for 0.5 h at 15 MPa
Figure 4: TEM of the interfacial structure between fibre and matrix for the mullite matrix
Nicalon-SiC fibre reinforced composite, SiC/M hot-pressed at 1550 C for 0.5 h at 15 MPa
II - 783
Figure 5: Typical load vs. deflection plots tested by three-point-bending; behaviour of matrix
shown by dotted curves. C/M: Carbon fibre mullite composite C/M, SiC/M: Nicalon-SiC fibre
mullite composite
Figure 6: SEM of the fracture surface of SiC/M composite hot pressed at 1550 C for 0.5 h at
15 MPa, parallel to the fibre direction
II - 784
MECHANICAL PROPERTIES OF SPIDER SILK

Sueo Kawabata1, Frank K. Ko2, Mari Inoue3, Masako Niwa3 and John W. Song4
1
Department of Materials Sciences, University of Shiga Prefecture, Shiga, Japan

2
Fibrous Materials Research Center, Department of Materials Engineering,
Drexel University, Philadelphia, PA 19104, USA
3
Nara Women University, Nara, Japan
4
USArmy Natick Research and Development , Engineering Center, Natick, MA. USA
SUMMARY: The mechanical properties of Nephila clavipes spider silk are presented to
illustrate the unique combination of strength and toughness and the anisotropy of spider silk.
The tensile properties of the silk of Nephila clavipes spider were characterized by simple
elongation, transverse compression and torsion under various humidity conditions using an
ultra-sensitive testing system. It was founded that, although the properties of the spider silks
vary from species to species, they all have a high level of combined strength and toughness as
well as shear resistance much greater than that of some of the toughest synthetic fibers such as
the aramid fibers.
KEYWORDS: spider silk, tensile modulus, compressive modulus, shear modulus,

micromeasurement instrument, anisotropic properties, combined strength and toughness
INTRODUCTION
Strength and toughness are usually considered mutually exclusive properties for materials. In
spite of the progress made over the past century in polymeric fiber science and technologies,
the search for a truly strong and tough fiber continues. It is of practical and scientific interest
to explore the limit of strength and toughness of fibrous materials; and to examine the factors
which contribute to the development of a combination of strength and toughness in materials.
The answers to these questions may be found in nature.
In the world of natural fibers, spider silk has long been recognized as the wonder fiber for its
unique combination of high strength and rupture elongation. An earlier study, as shown in
Figure 1, indicated spider silk has strength as high as 1.75 GPa at a breaking elongation of
over 26% [1]. With toughness more than three times that of aramid and industrial fibers,
spider silk continues to attract the attention of fiber scientists and hobbyists alike.
Considering the remarkable mechano-chemical properties of spider silk and fueled by the
recent progress in biotechnology, there is a revival of interest in using spider silk as a model
for the engineering of high energy absorption fibers [2]. Because of the fineness of spider silk,
on the order of 4m, the characterization of the mechanical properties of spider silks are
limited to tensile mode. Little is known about the response of spider silks to other modes of
deformation in the transverse direction and in torsion. It is the objective of this paper to
present, for the first time, the tensile, transverse compression and torsional stress-strain
properties of the spider silk from Nephila Clavipes spiders. This was made possible by using
II - 785
an ultra sensitive micromeasurement fiber testing system developed by Kawabata [3]. From
these experimental data, the engineering properties: tensile modulus, transverse compressive
modulus, and shear modulus of the spider silk was determined.
Specific Stress (g/den)
30
High Modulus Fibers

(glass, graphite,
aramid, ceramic,
steel, etc)
20
Spider Silk
Industrial Fibers
(tire cord, ropes)
10
Commercial Fibers
0
0
10
20
30
40
50
60
Elongation (%)
Figure 1: Tensile stress-strain behavior of agiope aurentia spider silk
SPIDER SILK AND SPECIMEN PREPARATION

There are over one million species of spiders on earth. Less than one percent of the spiders
are orb weavers or silk makers. Properties of spider silks very by species and by functions.
Depending upon the species an individual spider has more than five different silk glands
where silks of different properties are produced before extrusion through three pairs of
spinnerets. The drag line originates in the ampullate glad and is extruded through the anterior
spinnerets. Spider dray line is the strongest and toughest of the silks a spider makes. Spider
drag line is the structural framework of a spider web and the life line of a spider. Chemically
spider silk is a fibroin composed of twenty-odd alpha amino acids connected by the amide
linkage (-COHN-). Alanine , serine and proline account for eighty percent of the amino acids.
Drag lines of Nephila Clavipes (golden silk spider) and Argiope Aurentia (black and yellow
spider) are among the strongest spider silks that we know. The strength of the drag line of
Nephila Clavipes obtained by forcible silking was reported to be about 8 grams/denier and the
strength of the drag line of Argiope Aurentia was over 12 grams/denier. In fact the strength
and toughness of the silk from Nephila Clavipe was recognized for a long time by the South
Sea Islanders who use the silk to make various kinds of bags and fish nets.
The spider silks used in this study were forcible spun from Nephila Clavipes spiders. They
were provided by the US Army Natick R&DEC. The diameter of the silk, averaging over 117
measurements using the scanning electron microscope, is 3.67 +/- 0.47m. Figure 2 shows a
SEM picture of the silk. To facilitate handling of this delicate silk fiber, each fiber was
mounted carefully on a cardboard frame with an opening equal to the gage length of the
specimen. Thirty eight (38) samples were prepared for tensile test. Fifty (50) samples were
prepared for transverse compression test whereas a total of twenty nine (29) samples were
II - 786
prepared for torsion test. In order to investigate the effect of water on the mechanical
properties of the spider silk half of each group of fibers were tested in ambient condition
(20oC, 65% RH) and the other half of the specimen were tested in wet condition. The
diameter of each fiber was measured and recorded prior to testing. All the tests were
performed with the micromeasurement system developed by Kawabata, including a micro
tensile tester, torsion tester, and single fiber transverse compression tester 3.
Figure 2: SEM picture of spider silk
TENSILE PROPERTIES
The tensile properties of the spider silk were measured with the Kawabata micro tensile tester
equipped with a special function "floating control". In order to minimize mechanical noise
from the tester, a DC servo-motor system and simple transmission mechanism were used.
The motor driving power is transmitted to the driving shaft through a helical rack-and-pinion
mechanism similar to the driving mechanism of a microscope body tube. As shown in Figure
3, the control stroke of the driving chuck is 30 cm; however, in most of the fiber tensile
testing, the dynamic stroke is measured in millimeters, requiring very precise measurement.
Accordingly, the servo control system can be switched from the large stroke to the more
precise small stroke zone at any portion of the large stroke travel. The control accuracy in the
small stroke zone is 10 times more accurate, enabling a smoother control of fiber strain in the
"floating zone". The maximum load capacity of this tester is 20 Kg and the maximum testing
speed is 10 cm/sec. The position of the driving chuck can be controlled by an external servocommand signal (such as a sinusoidal signal) at any stage of the tensile testing process.
II - 787
Figure 3: Floating-control tensile tester

The tensile testing were carried out under ambient and wet conditions with a cross head speed
of 0.3 cm/min and a gage length of 2.53 cm. As shown in Figure 4, the stress-strain response
of the spider silk tested under ambient condition assumed a sigmoidal shape with an initial
modulus of 12.7 GPa. The breaking strength and elongation are 0.85 GPa and 20%
respectively. When tested in water, there is no significant change in ultimate strength,
however, a reduction of the initial modulus to 0.54 GPa and increased in ultimate elongation
to 36% was observed. This phenomenon may be attributed to the rupturing of the secondary
bonds which hold the helical molecular chains together resulting in greater elongation due to
the uncoiling of the molecular chain.
Figure 4: Tensile stress-strain properties of spider silk
TRANSVERSE PROPERTIES
The compression tests in the transverse fiber diameter direction were carried out by placing a
single fiber between a flat and mirror-finished steel plate and a mirror finished 0.2 mm square
II - 788
compression plane. Because of the fineness of the spider fiber, a combination of sensitive
instrumentation and mechanistic analysis are required in order to assure accurate
measurement of the compressive stress-stain properties. For a compression force F, deforming
the fiber cross-section as shown in Figure 5, its corresponding diameter change, U, can be
calculated based on the following equations:
U =
4F
(S11 - 13 2s33)
4F
(S11 - 13 2s33)
(0.19 + sin h -1
R
)
b
(1)
where
B2
(2)
R = fiber in undeformed state;

S11 = 1/ET
S22 = 1/EL
ET = transverse modulus
EL = longitudinal modulus
F = compression force/unit length of fiber
For polymers with highly oriented molecular-chain such as spider silk it is approximated that
S33 << S11 and 13 is small; therefore, Equation (1) may be reduced to
4F
U =
B2
1
ET
(0.19 + sin h -1
R
)
b
(3)
4F
ET
(4)
Figure 5: Diametrical compression of single fiber

The predictive accuracy of equation (3) have been verified using the Finite Element Method
[3]. The FEM analysis also provides the stress distribution within the fiber cross-section. In
the central area of the cross section, the maximum principal stress is:
II - 789
F
a = 2R
(5)
and the maximum stress occurs in the area nearest the point of contact, approximated by
m 2a
(6)
The compression force, F, is obtained by a moving electromagnetic coil acting against a

strong fixed permanent magnet, which permits precise control of the compression rod.
The resolution of the displacement measurement is 0.05 m. The influence of the stiffness of
the rod, etc. on the measured curve has been confirmed as negligible. Although U can be
measured to a high degree of accuracy, the initial region of the F-U curve has some
uncertainty, which causes a degree of uncertainty in the measurement of U values. In order to
eliminate the effect of this uncertainty on ET, the use of the slope of the F-U curve was
applied as follows: From Equations 3 and 4, U is expressed only as a function of ET for a
fixed value F. For fixed F1 and F2,
U1 = f (ET)|F = F1
(7)
U2 = f (ET)|F = F2
(8)
where F1 and F2 are taken after passing the initial region to obtain the corresponding U1 and
U2, respectively, as well as the slope of the curve between F1 and F2. The following relation
may then be derived:
U = g(ET)
(9)
or
(10)
ET = g-1 (U)
where U = U2 - U1 and g(ET) = f (ET)|F = F2 - f (ET)|F = F1
The silk fibers were subjected to transverse cyclic loading at a compressive speed of 0.3
m/sec. under ambient and wet conditions, The compressive modulus of the fiber tested in
ambient condition was 0.58 GPa. and the fiber experienced a high degree of permanent
deformation ( ~20%). As shown in Figure 6. similar to the tensile behavior, a drastic
reduction in compressive modulus was also observed when the test was performed under wet
testing conditions.
II - 790
Figure 6: Compressive stress-strain behavior of spider silk
TORSIONAL PROPERTIES
As shown in Figure 7, a single fiber having both ends reinforced by a paper backing using
ceramic adhesives is hung on a top hook connected to a highly sensitive torque detector
supported by two torque wires made of 0.2 mm piano wire. The bottom end is connected to a
bar, and both ends of the bar are inserted into slits of a servo-driven cylindrical tube. The full
scale of the torque meter is 0.0025 gf-cm/10 volt.
The specimens were tested in both ambient and wet conditions with a gage length of 1.33 mm
at a torsion speed of 0.3 rad/sec. Because of the low level of torsional load, great care must
be taken to contain the mechanical noise. Figure 8 shows the smoothed out torsional stressstrain curves illustrating the effect of moisture on the shear rigidity of the fibers. In both
cases, a significant hysteresis was observed. A shear modulus of 3.58 GPa was determined for
the silk tested in ambient condition. On the other hand, under wet condition, the silk fiber has
low resistance to deform after overcoming the frictional moment. This appears to be
consistent with the tensile deformation mechanism of the helical molecular structure.
II - 791
Figure 7: Torsion tester for single fiber
Figure 8: Torsional stress-strain behavior of spider silk
II - 792
CONCLUSIONS
Spider silk is an excellent model for the development of strong and tough materials. With a
highly anisotropic mechanical behavior, the sensitivity of the initial modulus and elongation
but not ultimate strength to moisture indicated an interesting molecular composite structure
composed of well ordered fibrillar structures oriented in a primitive helical structure.
An ultra sensitive fiber testing system has been introduced and its capability demonstrated
with the ultra-fine spider silk. This provides a useful tool in determining the engineering
properties of single fibers as shown in Table 1.
Table 1: Engineering properties of spider silk
Tensile Modulus:
EL (GPa)
Transverse Compressive
Modulus: ET (GPa)
Shear Modulus
G (GPa)
12.71
(25.6%)*
0.579
(41.3%)*
3.58
(19.7%)*
* Coefficient of Variation
ACKNOWLEDGEMENTS
This work was supported in part by the US Army Research Office through an ASSERT
program to the second author for materials supply and training of graduate students.
REFERENCES
1.
2.
3.
Ko, Frank, Nonlinear Viscoelasticity of Aramid Fibers, Ph. D Thesis, Georgia Institute of
Technology, Atlanta, Ga., 1977.
Kaplan, D., Adams, W. W., Farmer, B., and Viney, C., editors, Silk Polymers: Materials
Science and Biotechnology, .ACS Symposium Series 544, American Chemical Society,
DC, 1994
Kawabata, S., Micromeasurement of Mechanical Properties of Single Fibers, PP. 311328, in Modern Textile Characterization Methods, Raheel, M,., editor, Marcel Dekker,
1996
II - 793
NATURAL-FIBRE-MAT-REINFORCED
THERMOPLASTIC COMPOSITES BASED ON FLAX
FIBRES AND POLYPROPYLENE
S. Garkhail 1, R. Heijenrath 1, M. van den Oever 2, H. Bos 2 and T. Peijs 1
1
Centre for Polymer and Composites, Eindhoven University of Technology

P.O. Box 513, 5600 MB Eindhoven, The Netherlands
2
ATO-DLO , Agrotechnological Research Institute, Wageningen, The Netherlands
SUMMARY: Natural-fibre-mat-reinforced thermoplastic (NMT) composites based on flax

fibre mats and a Polypropylene (PP) matrix were manufactured using (i) a film-stacking
method and (ii) a paper making process. The influence of fibre length and fibre content on
stiffness and strength is reported and compared with data for glass-mat-reinforced
thermoplastic (GMT) composites, including the influence of using maleic-anhydride grafted
PP. The data is also compared with existing micromechanical models like Kelly-Tyson and
Cox-Krenchel for strength and stiffness, respectively. A good agreement was found between
theory and experiment in case of stiffness while in case of strength the experimental values
fall well below the theoretical predictions. Results indicated that NMTs are of interest for
low-cost engineering applications and can compete with commercial GMTs, especially when
a high stiffness per unit weight is desirable. Results also indicated that the key area for future
development lies not only in improved adhesion but mainly in improving the fibre strength.
KEYWORDS:
manufacturing
natural
fibre,
thermoplastic,
composites,
mechanical
properties,
INTRODUCTION
In the last decade, research activities in the area of thermoplastic composites have shifted
towards the development of 'cost-performance' engineering materials. Especially, glass-matreinforced thermoplastic (GMT) materials [1], being stampable sheet products based on
commodity resins such as polypropylene (PP) and moderate loadings of relatively long glass
fibres in random array have proven to be very successful in high volume markets such as the
automotive industry. Because of their excellent price-performance ratio, E-glass fibres are by
far the most important fibres for these type of composites. However, these fibres do have
some disadvantages regarding (thermal) recycling issues and abrasive wear of processing
equipment. Moreover, glass fibres are non-renewable and can cause skin irritations during
processing and assembly of fibre-reinforced parts.
Nowadays, ecological concern has resulted in a renewed interest in natural materials and
issues such as recyclability and environmental safety are becoming increasingly important for
the introduction of new materials and products. An interesting alternative for the use of glass
fibres as reinforcement in engineering materials are natural fibres based on lignocellulose
such as flax, sisal and jute [2]. These vegetable fibres are renewable, nonabrasive, can be
thermally recycled, and show less concern with safety and health. In addition, they exhibit
excellent mechanical properties, especially when their low density and price in comparison to
II - 794
E-glass fibres is taken into account. Although these fibres are abundantly available, especially
in developing countries such as Bangladesh and India, the applications are still rather
conventional, i.e. ropes, matting, carpet backing and packaging materials. Moreover, in the
last few decades the uses of natural fibres for these type of applications has been declined due
to the introduction of synthetic fibres, such as nylon and polypropylene. Hence, also in the
economic interests of developing countries, there is an urgent need for new application areas
for these natural fibres.
The first natural-fibre-reinforced composites were based on thermoset matrices, such as
unsaturated polyester or phenolic resins, together with sisal and jute [3,4]. More recently,
developments shifted to thermoplastic matrix composites [5-7]. This research focuses on the
development of natural-fibre-mat-reinforced thermoplastic (NMT) composites, being GMTlike materials based on natural fibres [8]. In this development study flax fibres are used as
reinforcements, mainly because of their excellent mechanical properties when compared with
other natural fibres and also because of availability in different forms in Western-Europe. PP
is used for a number of reasons. First, it is easy to process and one of the cheapest polymers
on the market. Secondly, it has a low processing temperature, which is essential because of
the relatively low thermal stability of natural fibres (200-250C). Finally, it has the perfect
ability to protect the hydrophilic natural fibre because of its strong hydrophobic and apolar
character. A clear disadvantage of this apolar character for composite applications is its
limited wettability as well as poor interfacial bonding with reinforcing fibres. This
disadvantage can, however, be overcome by functionalization of the polymer, which has
proven to be very effective in enhancing fibre/matrix adhesion in composite systems based on
polyolefins [9,10]. For the optimisation of the mechanical performance of the flax/PP
composites it is expected that both flax content and the flax fibre length should be as high as
possible. However, high fibre volume fractions and large fibre lengths will limit the
processability of such composites. Therefore, a balance between mechanical performance and
processability should be found.
In this research the influence of fibre length and volume fraction is investigated on random
flax-mat-reinforced PP composites, which were manufactured using the film-stacking method
and the suspension impregnation method, respectively. The latter method allows for a
systematic variation of the fibre length, since no fibre break-up occurs during processing. The
use of conventional melt-processing methods for the production of short fibre reinforced
compounds such as extruders will degrade the fibre length and are therefore not suitable for
such a systematic study. In addition, the influence of maleic-anhydride PP (MA-PP) on the
mechanical performance of flax/PP composites was also studied. Flax fibres, as well as glass
fibres, contain functional OH-groups that are able to interact chemically with the maleic
anhydride grafted polypropylene. From this an improved interfacial bond strength between
the flax fibre and the modified PP is expected. In order to get a better insight in the
importance of all these different parameters, the experimental results were compared with
model predictions using micromechanical models for random short-fibre-reinforced
composites [11-15].
II - 795
EXPERIMENTAL
Materials
In this study random flax fibre mats of Eco Fibre Products B.V. (The Netherlands) in
combination with an isotactic-polypropylene (PP) matrix of Shell (XY6500T) with a melt
flow index of 35 were used. In order to study the effect of improved fibre/matrix adhesion on
composite performance, 5 wt.% of a maleic-anhydride-modified polypropylene (MA-PP)
(Polybond 3002, BP Chemicals Ltd.) was added to the homopolymer. NMT composite
plates with different fibre contents were manufactured using the film-stacking method. In this
film-stacking method, pre-dried flax fibre mats and PP film were stacked alternately. PP as
well as films based on a blend of PP and MA-PP were made using film-blowing equipment.
Impregnation was achieved by applying heat and pressure in a hot-press (200C, 25 bar for 15
min.). The obtained composite plate was cut into tensile specimens according to ASTM-D
638 specifications. The Youngs modulus of the PP matrix was approximately 1.6 GPa and
the yield stress was about 32 MPa. Random MA-PP/flax composites were manufactured in a
similar way. The modification of PP with a commercially available MA-PP (Polybond 3002,
BP Chemicals Ltd.) was performed in a twin-screw extruder (Werner & Pfliederer ZSK25).
For the suspension impregnation process a dispersion of flax fibres and PP-fibres (Youngs
modulus = 1.6 GPa and yield stress = 29 MPa) was made in an ethanol/water (1:1) mixture. In
separate production runs, three different flax fibre lengths were used (3, 6 and 25 mm).
Significantly differing fibre lengths were removed before use. After drying of the materials at
room temperature for 24 hours and at 60C for one hour, this lofted mixture of flax fibres and
PP fibres was consolidated in a hot-press. For the suspension impregnation process random
flax/MA-PP composites were made by treating the fibres with MA-PP (Hostaprime HC5,
Hoechst). Degreasing of the flax fibres was executed via extraction of the flax fibres with an
ethanol/toluene (2:1) mixture, refluxed for three hours. The extraction fluid was replaced by a
new, fresh fluid after which it was refluxed for one hour. The fibres were subsequently
washed for approximately 30 minutes with cold ethanol, followed by 6 litres of cold water
over a bchner-funnel. The extracted fibres were dried at 60C in an oven with circulating air
for 24 hours. Afterwards they were immersed in a solution of MA-PP copolymer in hot
toluene (100C) for 10 minutes. The concentration of copolymer in solution was
approximately 2 wt.% on the fibres. After treatment, the fibres were extracted with toluene for
one and a half hour to remove all components not chemically bonded to the flax fibres.
Finally, the fibres were dried at 60C for 24 hours. These treated flax fibres were used in the
suspension impregnation process, yielding a composite plate using the compression moulding
method described above.
Test methods
Uniaxial tensile tests on random flax mat composites were performed on a Frank tensile
machine, type 81565, according to ASTM standards (D638M). The specimens were cut in a
dog-bone shape with dimensions 200mm x 20mm x 1.8mm. An extensometer was used to
monitor the elongation of the tested specimen. The uniaxial tests on random flax/PP
composites, made by the paper making process, were performed on a Zwick tensile testing
machine, type Z010. The specimens were rectangular cross-sections having dimensions of
100mm x 20mm x 2mm. Again, an extensometer was used to monitor the elongation of the
tested specimen.
II - 796

Influence of fibre length
Composite stiffness
The results of the tensile tests on the manufactured flax-fibre-reinforced composites, with
varying fibre lengths, are given in Figure 1 and 2. Figure 1 shows the Youngs modulus for
the systems based on PP and MA-PP together with the Cox-Krenchel model predictions. Cox
[11] introduced a fibre length efficiency factor into the rule-of-mixtures equation for the
composite stiffness Ec, to account for the ineffective loading of fibres over their stress
transfer length. Since the fibre is only partly utilised in the case of a short fibre composite,
these effects have to be taken into account when modelling the mechanical performance of
such composites. The expression used by Cox yields:
Ec = LE V f E f + (1 V f ) Em
(1)
In the case of stiffness related problems Coxs shear lag model is used for the calculation of
the fibre efficiency factor LE [16-21] the assumption of elastic fibres in an elastic matrix,
leading to:
tanh( L / 2)
]
L/2
(2)
2G m
2
[
]1/ 2
D E f ln(R / r)
(3)
LE = [1
where =
The R/r factor can be related to the fibre volume fraction Vf by:
ln(R / r) = ln( / i Vf )
(4)
so that Equation 3 can be rewritten as
2G m
2
[
]1/ 2
D E f ln( / i Vf )
(5)
where c i depends on the geometrical packing arrangement of the fibres, r is the fibre radius
and R is related to the mean spacing of the fibres. In this study a square packed fibre
arrangement is assumed, following Thomason et. al [20], where similar equations were used
for the modelling of glass/PP composites. The inter-fibre spacing in the composite is 2R and
ci equals 4. Besides the ineffective fibre length, another reason why the fibre can only be
partially utilised is the fibre orientation. In order to take this into account, the theory of Cox
was extended by Krenchel [12], who took fibre orientation into account by adding a fibre
orientation factor o into the rule-of-mixtures equation.
E c = o LE Vf E f + (1 Vf )E m
(6)
II - 797
The Krenchel orientation factor o allows for the introduction of a fibre orientation
distribution. If transverse deformations are neglected, o is given by:
o = a n cos 4
(7)
where an is the fraction of fibres with orientation angle n with respect to the reference axis.
For a two-dimensional (in-plane) random orientation of the fibres it can be shown that
o = 3/8. In a three-dimensional random fibre orientation the fibre orientation factor yields the
value of 1/5. For thin section laminates with fibre lengths greater than the thickness of the
sample, fibres are expected to be oriented mainly in two directions. However, deviations are
expected to occur due to out-of-plane oriented fibres or bend fibres. In the model a modulus
of 45 GPa for the flax fibre and a modulus of 1.6 GPa for the PP matrix is used. An effective
fibre length of 25 mm for the random flax mat material, together with a distribution function
for the fibre diameter is used. A rather good agreement is found between the model
predictions and experimental data. No effect of improved adhesion on the modulus of flax/PP
composites is observed indicating good wetting of the fibres. Only the data for a fibre length
of 3 mm shows a somewhat increased composite stiffness with the use of MA-PP.
Presumably, the higher interfacial bond strength in case of flax/MA-PP leads to a slight
decrease of the critical fibre length (from ~3.5 mm to ~2.5 mm) and consequently a small
improvement in composite stiffness might occur. At fibre lengths of 6 and 25 mm this
improvement is not observed since in both cases the fibre length is well above the critical
fibre length. No fibre length dependence is observed for the developed flax/MA-PP as the
fibre lengths are higher than critical lengths thus showing the curve in the plateau zone.
Fig. 1: Tensile modulus of the flax/PP composites (o) and the flax/MA-PP composites ()
as a function of the flax fibre length. The dashed line represents the Cox-Krenchel
prediction
Composite strength
In Figure 2 both the results of the tensile strength measurements as well as the model
predictions for flax/PP and flax/MA-PP, as a function of the fibre length, are plotted. In the
case of tensile strength modelling of discontinuous fibre composites, Kelly and Tyson [14]
extended the rule-of-mixtures equation for composite strength in a way similar to Coxs
rule-of-mixtures for composite stiffness:
II - 798
uc
= O LS Vf f + (1 Vf )
(8)
where LS is the fibre length efficiency factor and o is the fibre orientation factor, similar to
the Cox-Krenchel model, in order to take off-axis fibre orientation into account. For the fibre
length efficiency factor, Kelly and Tyson used [14] :
LS =
1
(
Vf
Li Vi
[ 2L
i
] + [Vj (1
j
Lc
)] )
2L j
(9)
The first summation term in Equation 9 accounts for the contribution of all fibres of subcritical length (L<Lc). The second summation term incorporates the strength contribution
from fibres whose lengths are super-critical (L>Lc). A combination of Equation 8 and 9
results in the Kelly-Tyson model [14,15] for the prediction of the strength ( uc ) of a polymer
composite reinforced with short off-axis fibres:
uc = o ( [
i
Li Vi
L
] + [ j Vj (1 c )] ) + (1 V f ) m
D
2L j
j
(10)
In the case of stiffness, the related Cox-Krenchel model, including a theoretical orientation
parameter for in-plane random fibre orientations (o = 3/8), can be used quite effectively for
the prediction of the stiffness of random fibre composites. However, previous studies have
shown that the Kelly-Tyson model for strength yields far too high values for the strength of
random fibre composites. In general this discrepancy is accounted for by adopting the fibre
orientation efficiency parameter o and as a result no unambiguous value for o is found in
the literature for the prediction of random composite strength. In this paper we used a value of
0.2 for the fibre orientation efficiency parameter o of our random flax/PP composites. This
value, having no physical meaning, was reported by Thomason et al. [20] and was obtained
by fitting o to the experimental strength data of a similar type of composite based on
glass/PP. Moreover, a fibre strength of 750 MPa is used together with an interfacial bond
strength ( ) of 10 MPa for flax/PP and 16 MPa for flax/MA-PP. The interfacial shear
strength value of 16 MPa for the flax/MA-PP system is taken similar to the shear yield stress
of the pure PP matrix as calculated from the Von Mises yield criterion (29/3 MPa), which
means that for this composite system perfect adhesion, or a matrix dominated rather than
interface dominated shear failure mode, is assumed. First of all, when looking at Figure 2, it
becomes clear that the experimental results at high fibre lengths (6 and 25 mm) of the flax/PP
as well as the flax/MA-PP composites are significantly lower than the Kelly-Tyson
predictions for tensile strength. With respect to the effect of improved interfacial bonding the
model predicts only a small increase in strength in the case of improved adhesion. Similar to
the stiffness there appears to be no significant influence of the flax fibre length on the
measured composite strength. Only the experimental results of the 3 mm flax/(MA-)PP
composites are more or less comparable with the model prediction. Due to the easier way of
mixing and consequently better impregnation, the composite with 3 mm long fibres is of a
higher quality in comparison to the 6 and 25 mm ones. Moreover, during the chopping of the
fibres into the desired length the weak internal interaction between the elementary fibres
results in extensive fibrillation of the fibres and a decrease of the flax fibre diameter and
consequently, an increase in fibre aspect ratio. As a result, the stress transfer becomes more
effective, yielding a composite with relatively better properties.
II - 799
Fig. 2: Tensile strength of the flax/PP composites () and the flax/MA-PP composites () as a
function of the flax fibre length. The dotted line and the dashed line represent the Kelly-Tyson
prediction for the flax/PP and the flax/MA-PP composites, respectively
Influence of fibre volume fraction
Composite stiffness
Figure 3 shows the tensile modulus of the flax/(MA-)PP composites together with the CoxKrenchel prediction and the commercially known glass-mat-reinforced thermoplastic (GMT)
materials, as a function of fibre volume fraction. Based on the data obtained, it can be
concluded that the stiffness of flax-fibre-reinforced PP is comparable to E-glass fibre
composites. Especially, when the relatively low density of the flax fibre is taken into account,
the stiffness per unit weight approaches or even surpasses that of GMT materials. Moreover,
as a result of the relatively low price of flax fibres compared to E-glass fibres these materials
are particularly of interest from a cost-performance point of view. From this it can be
concluded that flax/PP composites can compete with glass fibre based GMT materials in
stiffness critical applications.
Fig. 3:
II - 800
Tensile modulus of the flax/PP composites () and the flax/MA-PP composites ()

as a function of Vf. The dashed line represents the Cox-Krenchel prediction
A rather good agreement is found between the experimental data and the predictions using the
Cox-Krenchel model. In the model a modulus of 45 GPa for the flax fibre and a modulus of
1.6 GPa for the PP matrix is used, together with an effective fibre length of 25 mm for the
random flax mat material and a distribution function for the fibre diameter.
Composite tensile strength
Figure 4 shows the tensile strength of the manufactured flax/PP and flax/MA-PP composites,
together with the Kelly-Tyson predictions for flax/PP and flax/MA-PP as well as the
commercially glass fibre based GMT materials, as a function of the fibre volume fraction. In
the model the same material parameters are used as in the Cox-Krenchel model for
the prediction of composite stiffness. Again, a fibre strength of 750 MPa is used. These
micromechanical calculations also indicate that the difference in interfacial bond strength in
the case of flax/PP ( = 10 MPa) and flax/MA-PP ( = 16 MPa) has no significant influence
on the predicted composite strength (see Figure 4).
Fibre volume fraction [-]

Fig. 4: Tensile strength of the flax/PP composites () and the flax/MA-PP composites ()
as a function of Vf. The dotted line and the dashed line represents the Kelly-Tyson predictions
for flax/PP and flax/MA-PP composites, respectively. The solid line represents the
commercially available GMT
Clearly, GMT materials show superior strengths compared to flax-fibre-reinforced PP
composites for all volume fractions. With respect to the effect of improved interfacial bonding
in the case of flax/MA-PP composites no significant difference in tensile strength was found.
Moreover, based on micromechanical calculations no significant improvements can be
expected since already an upper limit matrix dominated shear strength value is used for the
calculation of the theoretical strength of flax/MA-PP. Obviously, the strength of the flax
based composites is intrinsically limited by the relatively low strength of the technical flax
fibres (~750 MPa). Therefore the present study shows that the key area for the improvement
of the tensile strength of the flax/PP composites lies not only in the interphase modifications
for improved adhesion but mainly in fibre modifications for improved fibre strength.
II - 801
CONCLUSIONS
For the production of flax/PP random composites two production methods similar to the
commercially known methods for glass/PP composites were used. First, the so-called filmstacking method based on flax non-woven fibre mats and secondly, a suspension
impregnation process using chopped short flax fibres were used. Material parameters that
were studied for the optimisation of the mechanical performance of such composites were (i)
fibre volume fraction, (ii) fibre length and (iii) fibre matrix adhesion. In order to get a better
insight in the importance of these different parameters for the optimisation of composites, the
experimental results were compared with model predictions using micromechanical models
for random short-fibre-reinforced composites. As expected, there is a significant effect of
fibre volume fraction on mechanical properties, whereas, no real experimental evidence for
the anticipated increase in mechanical performance with increasing fibre length was found.
Based on the experimental results it can be concluded that the tensile modulus of the flaxfibre-reinforced composites is comparable to E-glass-fibre-reinforced composites. Especially,
when focusing on the specific properties as a result of the low density of flax, the stiffness per
unit weight of flax-fibre-reinforced composites approaches that of glass-fibre-reinforced
materials. Moreover, as a result of the relatively low price of flax fibre compared to E-glass
fibres these materials might be of interest from a 'cost-performance' point of view. However,
due to the relatively low tensile strength of flax fibres compared to E-glass fibres, the tensile
strength of technical flax-fibre-reinforced composites is significantly lower than that of their
glass fibre counterparts. In short, it can be concluded that thermoplastic NMT composites
based on a PP matrix and flax fibres can compete with E-glass reinforced GMT materials in
stiffness critical structures, whereas for strength critical applications these materials still need
to be optimised.
ACKNOWLEDGEMENT
The authors would like to acknowledge Dr. G. Pott of CERES Fibres BV for providing the
flax fibre mats used in the present study.
REFERENCES
1.
Berglund, L.A. and Ericson, M.L. in: Polypropylene: Structure, blends and composites,
Vol. 3, Ed. J. Karger-Kocsis, Chapman & Hall, London (1995) 202.
2.
Morton, W.E. and Hearle, J.W.S., Physical Properties of Textile Fibres, 2nd ed., John
Wiley & Sons, Inc., New York (1975).
3.
Satyanarayana, K.G. et al. in: Composite Structures, Vol. 1, Ed. I.H. Marshall, Elsevier
Appl. Sci.., London (1981) 618.
4.
Chawla, K.K. and Bastos, A.C. in: Proc. 3rd. Int. Conf. Mech. Behaviour of Materials,
Vol. 3, Pergamon Press, Toronto (1979) 191.
5.
Kuruvila, J., Thomas, S., Pavithran, C. and Brahmakumar, M., J. Appl. Polymer Sci.,
47, (1993) 1731.
II - 802
6.
Selzer, R., Advanced Composite Letters, 4(3), (1995) 87.
7.
Sanadi, A.R., Caulfield, D.F., Jacobson, R.E. amd Rowell, R.M., Ind. Eng. Chem. Res.,
34 (1995) 1889.
8.
Heijenrath, R., Peijs, T., Advanced Composite Letters, 5 (3) 1996.
9.
Rijsdijk, H.A., Contant, M. and Peijs, A.A.J.M., Composites Science & Technology,
48, (1993) 161.
10.
Felix, J., Ph.D Thesis, Enhancing Interactions Between Cellulose Fibres and Synthetic
Polymers, Chalmers University of Technology, Goteborg, Sweden 1993.
11.
Cox, H.L., Brit. J. Appl. Phys., 3 (1952) 72.
12.
Krenchel, H. in: Fibre reinforcement, Akademisk Forlag, Copenhagen, 1964.
13.
Folkes, M.J. in: Short fibre reinforced thermoplastics, Research Studies Press,
Chichester, (1985) 16.
14.
Kelly, A. and Tyson, W.R., J. Mech. Phys. Solids, 13 (1965) 329.
15.
Kelly, A. and Macmillan, N.H. in: Strong Solids, Clarendon Press, Oxford (1986) 269.
16.
Ericson, M. and Berglund, L., Composite Science Technology, 43 (1992) 269.
17.
Ericson, M. and Berglund, L., Composite Science and Technology, 49 (1993) 121.
18.
Thomason, J.L. and Vlug, M.A., Composites, Part A, 27A (1996) 477.
19.
Rosenthal, J., Polymer Composites, 13 (1992) 462.
20.
Thomason, J.L.,Vlug, M.A., Schipper, G., Krikor, H.G.L.T, Composites, Part A, 27,
(1996) 1075.
21.
Pan, N., Polymer Composites, 14 (1993) 85.
II - 803
MECHANICAL PROPERTIES OF WOVEN FABRIC

CARBON/MAGNESIUM-COMPOSITES
O.ttinger, W.Schff, C.Hausmann, T.Heyne and R.F.Singer
Department of Materials Science
University of Erlangen
Martensstr.5, D-91058 Erlangen, Germany
SUMMARY: The mechanical behaviour of a woven fabric carbon/magnesium-composite

which is manufactured by a gas pressure infiltration technique is the subject of this research
work. Apart from the investigation of the influence of the fibre volume content (50-70 vol.-%)
on the bending strength, all other mechanical testing (bending, tensile and compression test) is
carried out by using composites with a constant fibre volume fraction of about 60 vol.-%. The
main points of the mechanical characterisations are the orientation and temperature
dependence, as well as the thermal long term stability of the 0/90 woven fabric C/Mgcomposites. The investigated two-dimensionally reinforced magnesium composites show a
typical orthotropic mechanical behaviour. The 0/90 strength is temperature independent up to
300C. Moreover, the outstanding mechanical properties stay unchanged even after a long
thermal exposure (200C; 1000h).
KEYWORDS: Woven fabric composite, carbon magnesium composite, MMC, mechanical

properties, thermal stability, orientation influence, temperature influence.
INTRODUCTION
Continuously reinforced metal matrix composites have received significant attention in the
last few years. Compared to conventional engineering materials they offer high specific
stiffness and strength, tailored thermal expansion and conductivity. Especially the low density
(1.8 g/cm), as well as the remarkable in-axis tensile, fatigue strength and stiffness make
unidirectionally carbon fibre reinforced magnesium (C/Mg) extremely attractive for advanced
aerospace and automotive applications [1-5]. However, C/Mg-composites have particularly
poor off-axis properties [4,6]. Fibre coatings and adapted fibre/alloy systems are able to
reduce the anisotropic behaviour of the composites slightly [6-8]. A much more effective way
to surmount this specific weakness, which is pursued in the present paper, is to use twodimensional (2D-) woven fabrics. The results reported in this paper are part of a more
comprehensive study with different C/Mg-systems [9].
EXPERIMENTAL
Materials
The specific subject of this research work is the AM20/T300J-composite system, which is
manufactured by a gas pressure melt infiltration technique [9, 10].
II - 804
A magnesium matrix with 2 wt.-% aluminium (AM20: 2 wt.-% Al, 0.6 wt.-% Mn, balance
Mg [11]) was chosen because of the comparatively moderate fibre/matrix-reaction (Al4C3)
resulting in an optimised interfacial strength [8, 12]. The commercial 2D satin 1/4 woven
fabrics are made of the high-tenacity carbon fibre T300J produced by Toray. The woven
fabric shows 7 filaments per cm in warp and fill direction and weigh 285 g/m. The fibre
preforms were manufactured without binder by stacking several woven fabric sheets
(typically 180x80 mm) in the infiltration form. Depending on the desired fibre volume
fraction a certain amount of woven fabric layers, which can be calculated by Eqn.1, was used.
d 2 N L N F x
100%
VF =
2h
with:
VF:
d:
NL:
N F:
x:
h:
(1)
Fibre volume fraction

Fibre diameter (here: d=7.1m)
Amount of woven fabric layers
Fibres per filament (here: NF=3000)
Filaments per cm
Thickness of the composite plate
Mechanical testing
The mechanical characterisation including bending, tensile and compression tests was
performed with an 100kN Instron 4505 machine.
The bending strength specimens had the dimensions of 100x10x2 mm and were tested
corresponding the German standard DIN 29971. According to this test method the influence
of the following factors on the bending strength were examined:
- Fibre volume fraction
- Orientation of the fibres
- Temperature in the range of room temperature to 400C
- Isothermal exposure at 200C and a duration of 1h to 1000h
The composite 0/90 tensile strength was determined by using flat specimens with a total
length of 175 mm. The gage length was 50 mm with a cross section of 8x2 mm. The 45
tensile samples had no gage length and a dimension of 100x10x2 mm. All tensile tests were
performed at a crosshead speed of 0.5 mm/min.
The compression strength was measured according to the German standard DIN 50106 for
unreinforced metals. The cylindrical specimens had a diameter of 6 mm and height of 9 mm.
For comparison the Celanese compression test method [13] with flat specimens of 112 mm in
length and a cross section of 6.3x2 mm were used. For both methods the crosshead speed was
0.5 mm/min.
II - 805

Influence of the fibre volume content on the bending strength of 2D-AM20/T300J
The influence of the fibre volume fraction on the bending strength and stiffness can be seen in
Fig. 1.
Bending strength (MPa)
Bending strength bB
700
130
600
500
110
400
Bending stiffness Eb
300
90
200
AM20/T300J
(0/90)
100
0
50
55
60
65
70
70
Bending stiffness (GPa)
150
800
50
Fibre content (vol.-%)

Figure 1: Influence of the fibre content on the 0/90-bending strength and stiffness of 2DAM20/T300J composites
According to the Rule of Mixture both the 0/90-bending strength and stiffness show
approximately a linear dependence of the fibre volume content in the range of 50-70 vol.-%.
Large variations of the bending strength with about 70 vol.-% can be observed. However
stable mechanical properties are obtained at volume fractures below 60 vol.-%. Therefore all
further mechanical composite testing are carried out with a constant moderate fibre volume
fraction of about 57 vol.-%.
Orientation dependence on mechanical properties of 2D-woven fabric reinforced
magnesium
As seen in Fig. 2, the two-dimensionally reinforced magnesium composites show a typical
orthotropic mechanical behaviour which has been previously observed with other woven
fabric composites [14]. Both the tensile and the bending strength reach the highest values in
the 0 and 90, respectively the fill and the warp direction. The total tensile strain at
composite fracture does not exceed 1%. In contrast to the 0/90 response, the woven fabric
reinforced composite show much lower strength in the 45direction. Bending and tensile
strength reach a minimum. The tensile strength is just slightly higher than the unreinforced
matrix AM20. The total fracture strain of the 45 composite samples is about 5% and half as
big as the unreinforced matrix. The 45 tensile test specimens show shear failure, i.e. the
crack surface is parallel to fibre orientation. The 0/90 tensile specimens fail normal to the
loading direction.
To analyse the orientation dependence of the compression strength of the 2D-AM20/T300J
two types of fibre orientations are tested: Load direction perpendicular (Fig. 3a) and parallel
(Fig. 3b) to the woven fabric layers.
II - 806
Bending-, tensile strength (MPa)
y 90
45
700
600
500
AM20/T300J, bB
400
300
AM20/T300J, Rm
200
AM20, Rm
100
100
200
300
400 500
600 700
Bending-, tensile strength (MPa)

Figure 2: The orthotropic mechanical behaviour of 0/90-reinforced AM20/T300J composites
(57 vol.-% fibre content) in comparison to isotropic tensile behaviour of the unreinforced
matrix AM20
Table 1: Influence of the fibre orientation on the compression strength of 2D-AM20/T300J
(57 vol.-% fibre content)
Fibre orientation (related to the
compression direction)
0/90 parallel
perpendicular
unreinforced
Compression strength
[MPa]
505 84
583 48
225 [7]
As Table 1 indicates, the compression strength of the composite is more than twice as high as
that of the unreinforced AM20, whereas the difference between loading the composite
perpendicular and the parallel to the fibre orientation is rather small. However, both fracture
modes can be clearly distinguished, as shown in Fig. 3.
a)
Shear failure
b)
Shear- and delamination failure
5mm
Figure 3: Influence of the fibre orientation on the fracture mode of the compression
specimens. The white arrows are pointing out the shear and delamination cracks
II - 807
With an orientation of the woven fabric laminates perpendicular to the compression direction
only shear failure in the plane of the maximum shear stress can be seen (Fig. 3a). The
compression specimen loaded parallel to the fibre direction demonstrates a different
behaviour (Fig. 3b): Beside the shear-cracks additional delamination cracks parallel to the
woven fabric layers occur. This is the reason why the attained compression strength is
somewhat smaller. Compared to these results the Celanese compression test method of the
0/90-bending strength parallel to the laminates showed without exception a slightly higher
strength of about 580 MPa instead of 505 MPa.
Mechanical properties at elevated temperatures and after thermal exposure
Bending-, Tensile strength (MPa)
Fig. 4 illustrates the temperature influence on the mechanical behaviour of 0/90-reinforced

AM20/T300J in comparison to unreinforced AM20.
800
Bending strength bB
700
(0/90)
VF = 57%
600
500
400
300 Tensile strength Rm
AM20
(unreinforced)
Tensile strength Rm
200
100
0
100
200
300
400
Temperature T (C)
500
Figure 4: The influence of the temperature on the mechanical behaviour of 0/90-reinforced

AM20/T300J
Regarding the low fibre volume content of only half of 57 vol.-% in loading direction
outstanding bending and tensile strength can be achieved even at elevated temperatures. In
contrast to unreinforced AM20, which tensile strength decreases distinctly with increasing
temperatures, the composite strength is constant up to 300C. Only at higher temperatures
(e.g. 400C) the strength declines and the composite strength will not be longer determined by
the fibre/matrix interface strength but by the very low matrix strength itself [9].
In contrast to many other magnesium matrix composites with e.g. alumina short-fibres [15]
the excellent 2D-AM20/T300J-composite strength still exist even after long thermal exposure,
as one can see in Fig 5.
Compared to the as-cast specimens neither the bending strength nor the bending stiffness is
reduced. The main reason for this excellent thermal stability is that carbon and magnesium is
thermodynamical inert (the Gibbs enthalpy values G0 for the formation of magnesium
carbides MgC2 and Mg2C3 are positive from casting to room temperature) [16] and that the
II - 808
800
(0/90)
VF = 57% 90
700 Bending strength bB

600
500
80
400
Bending stiffness Eb
200
100
0
70
Thermal treatment
T=200C (air atmosphere)
300
60
10
100
Bending stiffness (GPa)
Bending strength (MPa)
moderate reaction between the alloying element aluminium and the carbon fibre is completed
during the composite processing.
1000 10000
Thermal exposure duration (h)

Figure 5: The influence of the thermal exposure duration at 200C on the bending strength
and stiffness of woven fabric reinforced AM20/T300J
Comparison of the temperature influence on the composite strength of different woven
fabric reinforced composite
Bending strength bB (MPa)
Fig. 6 gives a comparison of the temperature influence on the bending strength of different
woven fabric reinforced composites. All composites are reinforced with woven fabric made of
high tenacity carbon fibres with similar fibre strength. The fibre volume content of all
composites is approximately 60 vol.-%.
1000
(0/90)
VF 60%
C/EP
(Fibredux
914C-833-40)
800
600
C/Mg
(T300J/AM20)
400
C/C
200
0
(1501 G)
100
200
300
400
500
600
Temperature T (C)
Figure 6: Comparison of the temperature influence on the bending strength of different woven
fabric reinforced composites [17, 18]
At low temperatures (up to 100C) the C/EP-system (carbon fibre reinforced epoxy) shows
the highest bending strength whereas these properties get quickly lost at elevated temperatures
[17]. In contrast, the strength of C/C composites (carbon fibre reinforced carbon) is relatively
low at room temperature but increases slightly up to temperatures of 2000C in an nonoxidising atmosphere [18]. However, the woven fabric reinforced AM20/T300J shows
II - 809
medium strength at room temperature, but has superior properties in the temperature range of
180C to 400C. Therefore, the 2D-AM20/T300J-system is able to close the strength gap
between C/EP and C/C-composites and makes C/Mg-composites extremely attractive for the
internal combustion engine and other high temperature components. Some near-net-shape cast
prototypes of woven fabric reinforced AM20/T300J are shown in Fig.7:
Figure 7: Woven fabric reinforced AM20/T300J-prototypes manufactured at the university of

Erlangen
CONCLUSIONS
The results of the present investigation lead to the following conclusions:
1. Strength and stiffness of 0/90 AM20/T300J increase linearly with the fibre volume
fraction.
2. The 0/90-reinforced AM20/T300J system shows an orthotropic mechanical behaviour.
The 0/90-tensile strength is more than three times higher than unreinforced AM20. The
tensile strength with 45 fibre orientation is only slightly higher than the unreinforced
matrix.
3. Bending and tensile strength of 0/90 reinforced AM20/T300J is constant up to 300C.
The composite system shows an excellent thermal stability even after long thermal
exposure (200C, 1000h).
4. Compared to other woven fabric reinforced composites (C/EP and C/C) the discussed
C/Mg-system has superior strength in the temperature range of 180C to 400C. Therefore
2D-AM20/T300J-system is an very promising material for high temperature composites
as required in internal combustion engines.
REFERENCES
1. A.P.Diwanji, I.W.Hall: Effects of Surface Treatment and Elevated Temperature Testing
on Mechanical Properties and Fibre/Matrix Interface of Gr/Mg Composites, Proc. of the
2nd Annual Joint Conference on the American Society for Composites, Metals Park,
American Society for Composites, 1988, p.351-360
II - 810
2. M.Rabinovitch, J.C.Daux, J.L.Raviart, R.Mevrel: Carbon-Reinforced Magnesium and

Aluminium Composites Fabricated by Liquid Phase Hot Pressing, Proc. of 4th European
Conf. on Composite Materials, Elsevier Applied Science, 1990, p.405-410
3. Y.Kagawa, E.Nakata: Some Mechanical Properties of Carbon Fibre-Reinforced
Magnesium-Matrix Composite Fabricated by Squeeze Casting, Journal of Materials
Science Letters 11 (1992), p.176-178
4. O.ttinger, M.Gruber, C.Grau and R.F.Singer, Properties and Microstructures of Carbon
Fiber Reinforced Magnesium Alloys, High Performance Composites: Commonalty of
Phenomena, TMS, (1994), p.275-289
5. D.Wurm, O.ttinger, R.F.Singer: The influence of Carbon Fibre Type and Stress-Ratio on
the Fatigue Behaviour of Unidirectionally Reinforced C/Mg-Composites, in: Metal Matrix
Composites, The Tenth International Conference on Composite Materials
(ICCM10),Vol.1, 1995, p.537-544
6. M.Rabinovitch, M.H.Vidal Setif, J.C.Daux, J.L.Raviart, R.Mevrel, H.Abiven, J.F.Peltier:
Proprietes Mecaniques de Composites Magnesium/Carbone Elabores Par Pressage Chaud
sous Moyenne Pression, Proc. of Int. Symp. On Advanced Materials for Ligthweight
Structures, Netherlands, Eur. Space Agency, ESA, 1992, p.135-139
7. M.Rabinovitch, M.H.Vidal Setif, J.C.Daux, J.L.Raviart, J.L.Gerard, R.Mevrel, M.Lancin,
O.Perez: An Approach to Optimize Interfaces in Carbon/Aluminium and
Carbon/Magnesium Metal Matrix Composites Elaborated by Hot Pressing, Proc. of the 9th
Int. Conf. on Composite Materials, Vol. Metal Matrix Composites, 1993, p.683-690
8. O.ttinger, C.Grau, R.Winter, R.F.Singer, A.Feldhoff, E.Pippel, J.Woltersdorf: The
Effect of Aluminium Additions on the Interfacial Microstructur and Mechanical
Properties of C/Mg-Composites, Proc. of the 10th Int. Conf. on Composite Materials, Vol.
VI, 1995, p.447-454
9. O.ttinger: Herstellung, Mikrostruktur und Eigenschaften von kohlenstofflangfaserverstrkten Magnesiumlegierungen, Fortschrittberichte VDI Reihe 5, Grund- und
Werkstoffe Nr.450, 1996
10. O.ttinger and R.F.Singer, An Advanced Melt Infiltration Process for the Net Shape
Production of Metal Matrix Composites, Z. Metallkunde 84, (1993), p.827-831
11. Hydro Magnesium: Datasheet Magnesium, High Purity Casting Alloys, Stabbek ,Norway,
1994
12. A.Feldhoff, E.Pippel, J.Woltersdorf: Interface Reactions and Fracture Behaviour of FibreReinforced Mg/Al Alloys, to be published in Journal of Microscopy, Vol.185, Part 2,
1997
II - 811
13. ASTM Standards and Literature References for Composites Materials D 3410-852,
Standard Test Method for Compressive Properties of Unidirectional or Crossply FiberResin Composites, Philadelphia, American Society for Testing and Materials, 1987,
p.131-140
14. K.Schulte, Faserverbundwerkstoffe mit Polymermatrix - Aufbau und mechanische
Eigenschaften,
DLR-Forschungsbericht
DLR-FB
92-28,
Kln,
Deutsche
Forschungsanstalt fr Luft und Raumfahrt, 1992
15. C.Khler, K.U.Kainer: Thermal Exposure Effects on the Structure and Mechanical
Properties of Short-Fibre Reinforced Magnesium Alloys, Proc. of the 9th Int. Conf. on
Composite Materials, Vol. Metal Matrix Composites, 1993, p.347-354
16. I.Barin, O.Knacke: Thermochemical Properties of Inorganic Substances, Berlin, Springer,
1973
17. Ciba-Geigy: Datasheet Firedux 914, Cambridge, 1989
18. SGL Carbon Group: Datasheet Sigrabond, Meitingen, 1994
II - 812
MODELLING STRESS TRANSFER IN A WOOD

COMPOSITE
Rowan Paton
Department of Materials Engineering, Monash University, Clayton, Victoria, 3168, Australia
SUMMARY: A model has been developed for the longitudinal elastic modulus of a new
structural wood composite, Scrimber. The model has been developed using a micromechanics
of composites approach, and is based on the Krenchel model. The morphology of Scrimber
has been studied in order to estimate the effects of strand discontinuity, strand misorientation,
strand damage during pressing, and wood densification during production of the composite.
The prediction of the model was within 2% of the measured modulus for a single batch of
Scrimber. Similar models should be very useful for other structural wood composites.
This paper concentrates on the effect of strand discontinuity on the modulus. The strand
discontinuity factor was estimated using a measured strand aspect ratio distribution, modified
by strand end shape criteria, and a form of the shear-lag theory of stress transfer.
KEYWORDS: wood composites, modelling, stress transfer, modulus
INTRODUCTION
Wood composites are a class of materials which are formed by breaking down wood into
smaller wood elements which are then bonded together again in a different form. A class of
wood composites increasingly used for structural purposes are the "Artificial Lumber" or
"Composite Lumber" materials such as Parallel Strand Lumber and Laminated Veneer
Lumber, materials which have similar properties to wood itself. These have been developed to
meet the demand for timber-like products at a time when the supply of large old-growth trees,
which are the most suitable for sawn timber production, is steadily decreasing.
Despite the common use of the term wood composites to describe these materials, and the
strong links with the polymer composites industry provided by groups like the US Forest
Products Laboratory, the industry has tended to regard these materials as an exotic species of
timber, rather than studying these materials as composites, an engineered material. Although a
great deal of theory has developed to predict the properties of many composites, there appears
to have been little serious use of such theory to predict the properties of wood composites.
A new form of composite lumber, Scrimber, has been developed in Australia. Scrimber is
produced from strands produced by passing juvenile Radiata Pine logs through a series of
rolls. This splits the log in the grain direction, and produces batches of partially-connected
strands called increments. The width and thickness of these strands typically varies from 1 to
10 mm, and their length typically varies from 10 to 200x the width or thickness. These are
dried, coated with adhesive, and laid up into a preform. The preform is then consolidated
before being compressed and cured in a radio frequency heated press. The principal direction
II - 813
of the strands in the beam is close to parallel to the long direction of the Scrimber beam.
Scrimber has an obvious grain and grain direction.
Scrimber is in many ways similar to a unidirectional fibre composite. The basic starting point
for the prediction of the longitudinal modulus of such a composite is often the Rule of
Mixtures:
Ec = EfVf + EmVm
(1)
where E is the modulus, V the volume fraction, and the subscripts "c", "f" and "m" refer to the
composite, fibres and matrix respectively. In the derivation of this equation it is assumed that
the fibres are continuous; that the fibres are perfectly aligned in the longitudinal direction; that
there is perfect bonding between fibre and matrix; and that the constituents are linearly elastic
for the stresses expected.
In Scrimber, as in many practical composites, the fibres are neither perfectly aligned nor
continuous. Krenchel (Theory 1) [1] was one of the first to predict the modulus of such
composites. He built on the Rule of Mixtures model, introducing two "fibre efficiency"
factors. These factors (here referred to as , for fibre orientation, and st, for stress transfer)
allow for the effect of fibre orientation and fibre discontinuity on the composite modulus. His
equation has been rewritten using the more usual terminology introduced above:
Ec = o st EfVf + EmVm
(2)
This approach is now a standard method for predicting the modulus of short fibre composites,
and has been used in this work to model the modulus of Scrimber. Feedlog offcuts, an
increment of strands, and two cured beams from a single batch of Scrimber were collected
and tested in compression and shear, measured, and examined to provide data for the model.
Predictions of the model were then compared with measured Scrimber modulus. This paper is
primarily concerned with the estimation of the stress transfer factor st for the model.
PREVIOUS WORK
Where the wood elements, (in Scrimber these are referred to as strands) are discontinuous, the
stress and strain in these strands are not uniform, and it is necessary to characterise the stress
transfer into the strands to determine how this affects the modulus of the wood composite. No
useful work specifically concerning stress transfer between wood composite elements could
be located. However a good model for the transfer of stress between fibres in a short-fibre
composite, the "shear-lag" theory, which exists in many forms, was first presented by Cox [2].
His and other variants of the theory are reviewed in Holister and Thomas (Chapter 2) [3]. For
this theory it is assumed that: the fibre and matrix are elastic and isotropic; a perfect bond
exists between fibre and matrix; there is no load transfer through the ends of the fibre; and
fibre shear strain is negligible compared to matrix shear strain.
Consider a tensile load is applied in the fibre longitudinal direction to an idealised composite
consisting of aligned short cylindrical fibres, with length L and radius r, in an elastic matrix.
II - 814
The tensile strain in the composite as a whole, and in the matrix, is considered uniform1 and
called c. The shear-lag theory holds that the shear stress on a fibre surface at a point near the
end of the fibre varies along the length of the fibre, being directly related to the difference
between the fibre tensile strain and the bulk composite tensile strain at each position along the
fibre. Cox showed that the tensile stress in the fibre, x, varies along the length of the fibre, L,
as:
cosh L x
2
x = Ef c 1
=
cosh L
2
Ef
Gm
Af
s
r
c
2 Gm
EfAf ln. s
r
(3)
= Fibre modulus in the x direction

= Matrix shear modulus
= Fibre cross-sectional area
= Mean separation of fibres normal to their length
= Fibre radius
= Matrix strain in the x direction
Fig. 1: Stress profile in a short fibre according to the Cox shear-lag theory
The form of this stress distribution is shown in Figure 1. Cox also showed that the average
stress in a fibre is:
tanh L
2
av = Ef c 1
L
2
(4)
If the effective modulus of the fibres in the x (longitudinal) direction, E eff, is defined as:
Eeff = av
c
(5)
then the short fibre stress transfer efficiency factor, st, can be defined as:
L
Eeff = av = 1 tanh 2
st =
Ef max
L
2
(6)
A reasonable approximation where the fibre ends of adjacent fibres are evenly distributed along the length of
the composite and the fibre crosssectional area is small compared to the composite crosssectional area.
II - 815
Since tanh L/2 1 where L/2 > 3, it can be shown that st is dependent on L. Space does
not permit here, but it can be shown also that L is related to the "fibre" aspect ratio, L/d,
where d is the fibre diameter.
While the shape of the stress transfer profile shown in Figure 1 may well represent the
situation in wood composite elements, these equations have been derived assuming a rigid,
circular fibre, not a good description of wood composite elements. Another form of the shearlag theory will now be examined.
For the purposes of investigating the stress transfer mechanism, a wood composite may be
thought of as similar to a laminate (like plywood or laminated veneer lumber) made up from
discontinuous bonded laminae. The geometry for stress transfer into a lamina in such a
laminate is exactly analogous to that for a double lap bonded joint. The pioneering work in
this area was done in 1938 by Volkerson, who presented a version of the shear-lag theory
suitable for bonded lap-shear joints which is still considered accurate for double-lap joints
where adherend bending is not a big factor. This arrangement of the original theory by
Volkerson is from Adams and Peppiatt [4]. Figure 2 shows the geometry.
Fig. 2: Geometry of lap joint for Equation 2.10

E and G are the elastic and shear moduli respectively, b is the lap joint width and P is the load
on the joint; the subscripts 1, 2, 3 refer to adherend 1, adherend 2, and the adhesive.
Volkerson showed that 1x, the tensile stress on adherend 1, is:
1 1 cosh L sinh x
1x = P 1 1 cosh x
sinh L
b1
(7)
where:
=
G3 E1 1 + E2 2
E1 E2 1 2 3
E2 2
E1 1 + E2 2
The form of the stress profile described by Equation 7 is shown below in Figure 3.
II - 816
(8)
Fig. 3: Stress transfer profile in a lap joint

Volkerson's theory is based on the same assumptions as Cox's theory, although of course the
assumptions are expressed in terms of the adherends and the adhesive, rather than the fibre
and matrix. It can be demonstrated that the right-hand side of the two stress profiles is very
similar. The rate of stress build up in a fibre or ply depends on the value of or (in
Equation 7 the value of also has a very, very small effect), and for the conditions below
Equations 3 and 7 give indistinguishable results to three significant figures. This shows that
where stress transfer in a composite can be described by the Volkerson theory, the value of
calculated using this theory could be substituted for in Equation 6 to predict st for the
composite.
P1
L/2 x L
L > 25
b1
= 2Efc
(9)
Three of the assumptions behind the Volkerson theory, namely that the adherend and adhesive
remain elastic, that a perfect bond exists between adherend and adhesive, and that there is no
load transfer through the ends of the adherend, may be valid for wood composites at low
strain. However the assumption that adherend shear strain is negligible compared to adhesive
shear strain is clearly invalid, as the shear moduli of wood and adhesive are similar, and the
wood element is normally much thicker than the adhesive. Adams and Peppiatt recognised
that this is so for many practical lap joints, and presented a modification of the Volkerson
theory which incorporates the effect of shear strains in the adherends. In their modified
theory, Equation 8 is replaced by:
=
2G1 G2 G3 E1 1 +E2 2
E1 1 E2 2 1 G2 G3 + 2 G1 G3 + 23 G1 G2
(10)
This equation could be used to calculate the stress transfer profile for wood elements in a
wood composite, and thus could be used to estimate st for a wood composite using Equation
6, rewritten here in terms of rather than :
tanh L
2
st = 1
L
2
(11)
II - 817
If Equations 10 and 11 are used to estimate st for a wood composite, it must be remembered
that these equations assume no stress transfer through the ends or the sides of the adherends
(wood elements). This assumption will be examined later.
ASPECT RATIO OF SCRIMBER STRANDS

In most short fibre composites and many wood composites the fibre or wood element is an
easily identifiable unit with a simple and consistent geometry. The Scrimber strand is a more
complex entity, with a large variety of shapes and sizes. One of the first problems in
describing Scrimber strands is to decide where one strand ends and the next begins; an
increment can contain both independent strands and partially connected clumps. It has been
assumed here that the wood in a clump is likely to lie together in the Scrimber, and to transfer
stress as a unit. Therefore, for the purposes of stress transfer, a single strand was considered to
be any part of the increment which was completely separate from the rest; clumps of partially
connected pieces were considered to be a single strand. A representative strand is depicted in
Figure 4.
Fig. 4: Representative strand shape

The majority of strand ends were of the tapered type. Cracks were a feature of almost every
strand, especially the clumps. Many of these cracks may be penetrated by adhesive and
bonded during the pressing stage, but others are expected to persist in the cured composite.
The presence of unbonded cracks within a strand is expected to lower the rate of stress
transfer between strands, but no practical quantitative method of predicting the effect of this
on Scrimber modulus could be conceived. Blunt steps were found on only a small minority of
strands. Their effect would be the same as that of blunt-ended strands, discussed below, but
on a much smaller scale; it was not possible to quantify it.
For stress transfer, it is considered that the most important feature of the strand is whether or
not its end is bonded. When a strand end is unbonded, there is no stress transfer through the
end, and shear-lag theory can be used to calculate st for that strand using the strand
dimensions. When a strand end is bonded, most of the stress is expected to be transferred
through the end. Adhesive bondlines in Scrimber are very thin, with a typical thickness of
0.1mm, and the adhesive modulus Ea is approximately half the strand modulus EL. Therefore,
where strand ends remain bonded, there is no significant variation in the fibre stress near the
fibre end, and st 1. Since Scrimber made up from end-bonded strands should have the
same initial modulus as a hypothetical Scrimber made from continuous strands, for the
purpose of calculating st, end-bonded strands can be considered to have an effectively
infinite aspect ratio.
Whether or not a strand end in Scrimber is bonded appears to depend on its degree of taper.
During the pressing stage, if the strand has a sufficiently fine taper, adjacent strands are able
to deform around the strand end, and to remain in contact with it. This is depicted in Figure 5.
II - 818
Fig. 5: Effect of strand end shape on end bonding and stress transfer.
Observation of the shapes of voids found at strand ends in Scrimber suggests that, for adjacent
strands to be able to deform around the strand end and remain in contact with it, the strand
end included angle needs to be less than approximately 30.
On this basis, the strands in the sample increment were classified according to their end
shapes and measured. The great majority of strands had at least one tapered end. Where the
strand-end included angle was more than 30, the end was considered to be blunt, and thus
likely to be unbonded if incorporated into Scrimber. Approximately 25% of the strands by
weight had two tapered ends. These were assigned an "infinite" aspect ratio, and their total
weight fraction only was recorded. For strands with two blunt ends width, thickness, length
and weight fraction were recorded. For strands with one blunt end only, the same dimensions
were recorded; however in this case an effective length of twice the actual length was
assigned and used to calculate an effective aspect ratio.
In total, three sets of data were recorded. In the first data set, as explained above, the actual
strand length is recorded for strands with both ends blunt; the effective length for strands with
one end blunt, and an "infinite" length for strands with two tapered ends.
A second set of aspect ratio data was recorded to investigate the effect on predicted Scrimber
modulus of all strand ends being considered unbonded. In this set, the actual strand length
was recorded for strands with one and both ends blunt. Since individual measurements had
not been taken of strands with both ends tapered, these were excluded from this data set,
which was recorded as a crude type of worst-case.
It was necessary to record a third aspect ratio data set because of a manufacturing practise at
Scrimber International. The most common reason for strands having a blunt end was that this
end was part of the original guillotined log end. These guillotined blunt ends were found close
to the increment ends, which placed them on the surface of the cured beam. Since the average
thickness of an increment after pressing was 5.5 mm, most of these blunt ends were found
within that distance of the surface of the cured beam. It was regular Scrimber International
practice to plane some material from all surfaces of the beam before use, and accordingly
approximately 7mm was planed from the surface of beams 2270/4 and 2270/5 during
production of the compression specimens. The compression specimens tested in this work
therefore included few guillotined ends, and their modulus could be expected to reflect this.
For the third set of aspect ratio data, strands with the guillotined blunt end were re-evaluated,
and considered to have a tapered end in its place, since those strands, after incorporation in
the Scrimber beam, would have gained a tapered end during planing. Strand effective lengths
II - 819
were modified accordingly. This third set of aspect ratio data is considered to be the most
representative of the strands making up the compression specimens for Scrimber batch 2270.
VERIFICATION OF THE SHEAR-LAG THEORY FOR WOOD COMPOSITES

To verify that the shear-lag theory as developed by Adams and Peppiatt [4] for non-rigid
adherends adequately predicts stress transfer in wood composites, the predictions of that
theory were tested against strain measurements near the end of a wood element in a simple
wood composite. Laminated Veneer Lumber (LVL), was chosen as a suitable composite
because it has a simple and consistent wood element geometry and similar glue-lines to
Scrimber. Most importantly, the stress state within a loaded section of LVL was expected to
be more predictable than in other wood composites, especially Scrimber, and the wood
elements within LVL were large enough to easily apply bonded foil strain-gauges.
The LVL chosen was made up from 15 rotary-peeled 3mm plies of Radiata, from the same
general area as the Scrimber feedlogs. The plies, each 1200 mm long, were bonded with
phenol formaldehyde adhesive; the outer three plies on each face were scarf jointed, while the
remainder were laid butt to butt with a small gap, typically 1 mm. The longitudinal direction
of each ply coincided with the longitudinal direction of the specimen.
A 600mm long piece of LVL was cut as a test piece. The test specimen was 47mm thick
(normal to the ply plane) and 10mm deep. A ply end near the centre of the test specimen, in
an area generally free from other butt joints or grain irregularities, was chosen for
measurement of the strain build-up. Foil strain gauges were bonded to the edge of the ply near
its end. The gap between the two ply ends was 1mm; approximately 70% of the gap was filled
with adhesive spew incorporating bubbles. Stress transfer in the LVL test specimen was
expected to be similar to that in a double-lap-shear specimen. The geometry of the test
specimen, considered as a double lap-shear specimen and using the notation of Adams and
Peppiatt from Figure 2, is set out in Figure 6.
Figure 6. LVL considered as a double lap shear specimen.
II - 820
The stress transfer Equation 7 from Adams and Peppiatt has been rewritten in terms of the
notation in Figure 6. It predicts that for 0 x < L/2:
2x =
F 1 cosh x + 1 1 cosh L sinh x

sinh L
22 b
(12)
where (Eqn 10):

=
2G1 G2 G3 E1 1 + E2 2
E1 1 E2 2 1 G2 G3 + 2 G1 G3 + 23 G1 G2
A modification must be made to Equation 10 in order to predict a sensible value of for the
LVL experimental configuration. Equation 10 is based on the assumption that both adherends
are free to deform in shear. Since the gap between the ply ends is very small, adherend 1 is
effectively rigid, and G1 G2. In addition, because the adhesive is very thin, 2G3 23G2,
and in this case E1 = E2. Therefore the equation can be simplified to:
=
2G2 1 + 2
E1 1 2
(13)
Using Equation 13, was calculated for the LVL experimental configuration shown in
Figure 6. The predicted value of for the LVL specimen was 0.26. The strain-gauged LVL
specimen was loaded in tension to 20 kN, in 1 kN steps. At each step, individual gauge
readings were recorded. All strain/load plots were linear for specimen loads up to 8 kN. For
this initial linear section, Figure 7 shows the relative strain in the ply under each of the strain
gauges plotted against the distance of the gauge from the ply end.
Extrapolation of the measured strain profile to the ply end indicates that the strain at the ply
end was approximately 50% of the maximum ply strain; ie. roughly 50% of the load was
transferred into the ply through its end. This load is thought to have been transferred by
tensile stresses in the adhesive spew, which as noted previously filled approximately 70% of
the inter-ply gap. In addition, the measured rate of strain transfer (controlled by the value of
in the shear-lag model) for the ply is less than predicted, and therefore the predicted value of
appears to have been too high. The rate of stress transfer into the ply could be expected to
have been reduced by the peeling cracks observed in every ply. The presence of these cracks
apparently lowers the ply shear stiffness G2 and consequently has been overestimated.
A shear lag profile can be shown to closely fit the measured strain transfer profile if = 0.15
and = 0 at x = 4.5 mm (which corresponds to the strain at the ply end being 52% of the
maximum strain), as demonstrated in Figure 7. This suggests that the chosen shear-lag theory
can accurately predict the form of the stress transfer profile in a wood element if the amount
of stress transferred through the end is known and the effect of wood element internal
cracking on the effective shear modulus of the wood element can be estimated.
II - 821
Figure 7. Fitting of the shear-lag prediction to the measured strain profile.

It must be emphasised that the stress transfer profile described above applied only where
specimen load was less than 8kN. The strain gauge measurements also indicated partial
failures of the ply end bond at 8kN and again at 15kN. The strain transfer profiles above these
loads showed progressively less strain transfer through the end of the ply.
This experiment confirmed that the stress profile in a wood composite element fits the shearlag form very well, and that the appropriate shear-lag theory can be used to predict st for
wood composites. It also showed that estimates of , and therefore st, however, from wood
element geometry and properties are likely to be too high if the wood element contains
internal unbonded cracks. A porous spew fillet 1mm thick, bonding approximately 70% of the
strand end surface, was apparently able to transfer 52% of maximum ply stress through a very
blunt ply end. The result suggests that, in a lightly-loaded wood composite, for a wood
element with a well-bonded end, almost all of the stress could be expected to be carried
through the end. This justifies the proposal made earlier that end-bonded strands could be
treated as effectively continuous for the estimation of st.
ESTIMATION OF THE STRESS TRANSFER FACTOR FOR SCRIMBER

In detailed investigations of the microstructure of Scrimber it was found that the strand crosssection circumference was similar to that of a square of the same area, and that at least 40% of
the strand surface was unbonded. Therefore for the purposes of predicting stress transfer
between strands in Scrimber, the typical strand was assumed to be square in cross-section, and
bonded on two opposite sides only. (The unbonded surface fraction was thus assumed to be
50%.) The strand was also assumed to be in the centre of the Scrimber compression specimen,
as shown in Figure 8.
II - 822
Figure 8. Geometry and notation for application of shear-lag theory to Scrimber.

The notation used is the same as for the LVL specimen in Figure 6. The strand cross-sectional
dimensions were calculated as shown below, where As is the measured strand cross-section
area and Ds is the batch 2270 densification factor:
22 =
As
Ds
(14)
For each strand, the variable (i) was calculated from Equation 13 and the stress transfer
efficiency factor, st(i), was then calculated from Equation 11. P(i) was defined as the strand
mass divided by the sample mass. For Scrimber batch 2270, st was calculated for each set of
strand aspect ratio measurements using:
i=n
st = st(i) P (i)
i=0
(15)
For Set 3, believed to be the most representative set of aspect ratio measurements for the
sample increment, st was calculated to be 0.98. For Set 1 (not corrected for the absence of
guillotined ends in the Scrimber specimens) st = 0.96; and for Set 2 (the crude worst case),
st = 0.94.
The estimate of st = 0.98 for batch 2270 reached above is probably too high, as it relies on
the assumption that the Scrimber strands were internally intact. However investigations of
Scrimber microstructure found that internal unbonded cracks were likely to be present in the
majority of strands. No practical method of predicting the effect of these internal cracks could
be conceived. The quantitative effect of these cracks can only be guessed, using the LVL
stress transfer experiment as a guide. Given that relatively minor internal cracking lowered
the effective value of in the LVL to 60% of the predicted value, it seems possible that the
existence of cracks within the majority of Scrimber strands might reduce the effective value
of for those strands to half or even a third of its predicted value, giving rise to an estimate
for st as low as 0.95.
II - 823
MODEL PREDICTIONS AND CONCLUSIONS

The Krenchel-type model introduced in as Equation 2 was found to need modification as
follows. The symbols are defined below.
Es = d Ds Ew
(16)
EB (Scrimber) = st o (d Ds Ew) Vs + Ea Va
(17)
This makes the full model:
Values for all the variables in this relationship were found for Scrimber batch 2270. Using
this proposed model, the average modulus of the Scrimber batch 2270 compression specimens
was predicted to be 11.05 GPa at standard conditions. This compares well with the average
measured modulus for those specimens of 10.8 GPa at standard conditions.
st
o
d
Ds
(stress transfer efficiency factor) = 0.98.

(orientation efficiency factor) = 0.875.
(strand damage factor)
= 0.88.
(strand densification factor)
= 1.27.
Ew
Vs
Ea
Va
(feedlog modulus)
(strand volume fraction)
(adhesive modulus)
(adhesive volume fraction)
= 12.9 GPa.
= 0.875.
= 7.6 GPa.
= 0.03.
Although it was impractical to attempt to design an experiment to test the effect of strand
aspect ratio, and therefore stress transfer, directly, the apparent accuracy of the whole model
allows some confidence in the predicted value of the various factors.
Strand discontinuity was found to have only a small effect on Scrimber modulus. The high
strand unbonded surface fraction of the batch studied also had little effect on modulus. Strand
misorientation and damage were found to be the major factors reducing Scrimber modulus.
This method described here of estimating the effect of fibre discontinuity in wood
composites is relatively simple and demonstrates the usefulness of applying standard
composites theory, allied with a good grasp of the composite microstructure, to the modelling
of wood composites. Procedures for estimating the efficiency factors could easily be modified
for application to other wood composites.
REFERENCES
1.
Krenchel, H., "Fibre Reinforcement", Akademisk Forlag, Copenhagen, 1964.
2.
Cox, H.L., "The elasticity and strength of paper and other fibrous materials",
British Journal of Applied Physics, Vol. 3, 1952, pp 72-79.
3.
Holister, G.S., and Thomas, C., "Fibre Reinforced Materials", Elsevier, London, 1966.
4.
Adams, R.D., and Peppiatt, N.A., "Effects of Poisson's ratio strains in adherends on
stresses of an idealised lap joint", Journal of Strain Analysis, Vol 8, 1973, pp 134-139.
II - 824
MICROCELLULAR GRAPHITIC FOAM PROCESSING

Kristen M. Kearns1 and David P. Anderson2
1
Wright Laboratory Materials Directorate, US Air Force, 2941 P St. Ste. 1,

Wright-Patterson Air Force Base, Ohio 45433-7750 USA
2
University of Dayton Research Institute, 300 College Park Avenue,
Dayton, Ohio 45469-0168 USA
SUMMARY: This paper reviews the experimental work to develop processing conditions
for pitch foams. The foams are produced by dissolving a gas into the pitch under pressure.
After the appropriate amount of time has lapsed the pressure is released, the gas nucleates
bubbles, and these bubbles grow, forming the pitch into the foam structure. The resultant
foamed pitch is then stabilized in an oxygen environment. At this point a rigid structure
exists with some mechanical integrity. The foam is then carbonized to 800C followed by a
graphitization to 2500C. The shear action of the bubbles growing aligns the graphitic planes
along the foam struts to provide the ideal structure for good mechanical properties. Variations
of the blowing temperature, blowing pressure and saturation time results in foams of various
open pore sizes. This study showed the final temperature and the duration of the hold time
most effected the foam structure. A lower final temperature and a longer hold time produced
a foam with a uniform porosity foam with a very small bubble size.
KEYWORDS: structural foams, graphitic foams, open-cell foams, processing
INTRODUCTION
Structural composites are based on disconnected carbon fibers integrated in some type of
matrix. The strength and stiffness of commercial carbon fibers are due to the graphitic
morphology that originates from the melt spinning of the precursor pitch. If an interconnected
network of struts could be produced which possessed a similar morphology to the carbon
fiber, a new generation of composite reinforcement could emerge. Foams are an example of
such materials with an open cell structure being the desired architecture. Microcellular, open
cell foams can be produced from anisotropic pitch with graphitic planes aligned along the
struts. The process sequence includes blowing, stabilizing, carbonizing, then graphitizing the
foam, similar to the process for manufacturing pitch-based carbon fibers. A foam could be
blown into a mold for net-shape composites or co-processed with fibers for anisotropic
reinforcements. Model graphitic foams were predicted [1] to have a compression modulus of
approximately 2 GPa with a density of about 0.1 gm/cm3. Not many other foams or core
materials have a density and compression modulus near these values.
The original process for fabricating microcellular foams was developed by Suh et al. [2-3]
They demonstrated microcellular foams could be produced from amorphous polymers by
saturating the polymer with a gas then heating above the glass transition temperature. The
sudden release of pressure resulted in a foam of uniformly distributed pores. Dutta, et al. [4]
modified this process by saturating with nitrogen under pressure to get better solubility of
nitrogen into pitch. In this process, the solubility of nitrogen into the pitch is critical to foam
II - 825
fabrication. Therefore, they first melted the pitch in a steady flow of nitrogen. The pitch was
then saturated with nitrogen gas and held for 10 minutes before dropping the pressure and
temperature suddenly.
The current method of fabricating microcellular foams is a variation of both the Suh and Dutta
work. It includes a slight initial pressurization before the pitch is heated. Thus, the pitch is
under a slight pressure during heating and is further pressurized at the final temperature. For
this process pressure, temperature and time are the controllable parameters and can be varied
to produce foams with various magnitudes and types (closed or open) of porosity. Therefore,
this study was done to determine the effect of these parameters on the final foam structure.
EXPERIMENTAL
Materials
Foams were processed from AR pitch manufactured by Mitsubishi Gas Chemical Co. The
pitch is produced by the catalytic polymerization of naphthalene and supplied in a pellet form.
Manufacturing data claims the softening temperature is 239C and the material is 100 percent
anisotropic. The glass transition temperature occurs over a temperature range of 230 to
260C.
Processing
The pitch pellet was jet-milled into particles of an average size of 1-3 microns.
Approximately 10 grams of jet-milled pitch was pressed into a 5.7 cm diameter puck. The
pitch puck was then placed in a Parr pressure reactor and heated (under a slight nitrogen
pressure) at 3C/min to the desired final temperature. When the puck reached the final
temperature additional nitrogen was added to obtain the final desired pressure and the entire
system was held at these conditions for a determined amount of time. The pressure was
abruptly vented to the atmosphere. Figure 1 shows a process cycle, with a sample rate of one
data point per 5 seconds. This processing created the foam structure to be further analyzed.
Further processing was required to transform it from a pitch to a carbonaceous foam. After
blowing, the foam was quickly removed from the reactor and placed in a 150 to 175C oven.
The foam was then cooled to room temperature and weight and dimensional measurements
were taken. The foams were then oxygen stabilized in a forced air oven at 220C until a
weight gain of approximately 7 percent had been reached. Samples were then carbonized with
a 60C/hour heating rate to 850C in a nitrogen environment.
II - 826
8000
300.00
PTC1 Deg.C
6000
200.00
5000
4000
150.00
3000
100.00
Pressure, kPa
Temperature, C
7000
Pressure kPa
250.00
2000
50.00
1000
0.00
0
0
25
50
75
100
125
Time, min
Figure 1: Foam Blowing Process Cycle

Characterization
Physical evaluations included weights and dimensional measurements. These measurements
were taken after puck fabrication, after foaming, and after carbonization to give a nondestructive indication of any differences produced by the various processing conditions. The
completely processed foams were examined under both optical and scanning electron
microscopy (SEM) and measurements of the pore size, pore shape, and amount of open
porosity were made. The relative pore size was measured using the SEM images. All samples
were examined at the same magnification and pore sizes measured using arbitrary rulers in the
imaging software program. Averages were calculated using 101 randomly chosen pores. The
standard deviation of the pore size was also calculated and assumed to be a measure of the
uniformity of the foam porosity.
Finally, compression tests were performed on samples from the experimental conditions that
produced testable materials. The test configuration basically followed ASTM D695, however
sample sizes were dictated by the size of the fabricated foam. From the tests both
compression strength and modulus were calculated.
Design of Experiment
A 24 half factorial design of experiment was performed to look at the effect of four process
parameters; the initial pressure, final temperature, final pressure, and the length of the hold
time at the final conditions. Table 1 provides the experimental values for each parameter and
Table 2 the conditions of each experimental run.
Table 1: Experimental Values
Initial Pressure (Pi):
Final Temperature(Tf):
Final Pressure (Pf):
Hold Duration (t):
1380 kPa (-)

250C (-)
6900 kPa (-)
15 minute (-)
3450 kPa (+)

270C (+)
10,350 kPa (+)
30 minute (+)
II - 827
Table 2: Experimental Run Conditions

Sample #
1
2
3
4
5
6
7
8
Pi
+
+
+
+
-
Tf
+
+
+
+
-
Pf
+
+
+
+
-
t
+
+
+
+
-

For all experiments, foams were created with varying microstructures and compression
properties. Figure 2 shows SEM images of the extreme microstructures produced by varying
the process parameters. The foam in Figure 2a was processed with higher initial pressure, the
lower final temperature and final pressure, and the longer hold time. Figure 2b was processed
with higher initial pressure and higher final temperature, lower final pressure, and the shorter
hold time. The sample represented in Figure 2a had much more uniform porosity with a
smaller bubble size than the non-uniform, large sized pores shown in Figure 2b. Figure 2a
represented the more desired foam structure.
500 m
500 m
(a)
(b)
Figure 2: SEM Images of foam samples,
(a) small, uniform porosity, (b) large, non-uniform pores.
The volumetric change due to foaming was measure by comparing the pre-foamed puck
volume to the foam volume before stabilization. Each process parameter was plotted versus
the volumetric change to determine which had the most influence, Figure 3. As can be seen
the only parameter that had any effect is the hold duration at the final temperature after the
final pressure has been added. A shorter hold time resulted in a larger expansion in the
foamed pitch. The longer time at the final conditions allowed the system to equilibrate and
more nitrogen was able to diffuse into and uniformly disperse throughout the pitch. When the
pressure was released the nitrogen nucleated and grew bubbles and quickly produced an opencell structure. As the open cell structure formed, the nitrogen was able to diffuse out of the
foam and bubble growth was terminated.
The final temperature had a minor secondary effect on the volumetric change. The foams
produced with a lower final temperature did not expand as much as the ones made at the
II - 828
higher temperature. At the lower temperature, the pitch had a higher viscosity and could keep
more nitrogen trapped. When the pressure was released the nitrogen expanded and quickly
created an open cell structure. The open cells allowed the nitrogen to vent to the atmosphere
and minimized volume expansion. At the higher temperature, less nitrogen was available for
bubble growth and an open structure was not produced until much later in the foaming
process. Therefore, a larger expansion in volume occurred at the higher final temperature.
Volumetric Increase, %
500
400
300
200
100
0
0
1000
2000
3000
500
400
300
200
100
0
240
4000
250
260
270
Final Temperature, C
(a)
(b)
Initial Pressure, kPa
500
400
300
200
100
0
5000
280
7000
9000
11000
500
400
300
200
100
0
10
15
20
25
30
Final Pressure, kPa
Hold Duration, min.
(c)
35
(d)
Figure 3: Volumetric Increase from puck to foam as a function of (a) initial pressure,
(b) final temperature, (c) final pressure, and (d) hold duration.
The average pore size was plotted against each of the process parameters, Figure 4, to
determine what were the effects of each parameter on the pore size. The results showed the
final temperature had the largest influence on pore size. The higher final temperature
produced foams with larger size pores than the lower final temperature. At the lower
temperature the pitch has a much higher viscosity and thus a much greater ability to keep the
nitrogen trapped. The greater amount of nitrogen trapped inside the pitch the more bubbles
that could nucleate and/or the more nitrogen there is available to grow the bubbles. The
viscosity of the pitch was high enough to restrict the bubble growth to keep the pores small.
When the results from the volume increase were compared to the relative pore size it was
determined the smaller pore size also had the lower volume increase. Apparently the nitrogen
nucleated bubbles and as they grew they formed an open-cell foam structure that allowed the
II - 829
40
35
30
25
20
15
10
5
0
Avg. Pore Size
Avg. Pore Size
nitrogen to diffuse out before the bubbles grew to a larger size or coalesced to form larger
pores. This would allow for a high density of very small bubbles.
1000
2000
3000
4000
40
35
30
25
20
15
10
5
0
240
250
260
270
(a)
(b)
40
35
30
25
20
15
10
5
0
5000
Avg. Pore Size
Avg. Pore Size
280
7000
9000
Final Pressure, kPa
11000
40
35
30
25
20
15
10
5
0
10
15
20
25
30
35
Hold Duration, min.
(c)
(d)
Figure 4: Average pore size as a function of (a) initial pressure,
(b) final temperature, (c) final pressure, and (d) hold duration.
Compression tests were only done on material from experiments 3,4,5,7 in Table 1. These
were the only experiments that produced samples large enough to make compression samples.
The compression strength was plotted versus process parameters and shown in Figure 5. The
higher final temperature produced lower strength foams than the lower final temperature.
This trend could be correlated with the relative pore size. The larger pore size produced a
foam with lower compression strength. This was due to the smaller strut size in the larger
pore foam. As the bubble grows the strut becomes more elongated and thinner than the struts
of the smaller pore foam. The pore size can be explained by the difference in the viscosity of
the pitch at 250C and 270C. At 270C the pitch molecules were more mobile and the
nitrogen did not saturate the pitch as much as it did at 250C.
II - 830
70000
Compression Strength, kPa
70000
60000
50000
40000
30000
20000
10000
0
60000
50000
40000
30000
20000
10000
0
1000
2000
3000
4000
240
250
260
270
(a)
(b)
70000
70000
280
60000
50000
40000
30000
20000
10000
0
60000
50000
40000
30000
20000
10000
0
5000
7000
9000
Final Pressure, kPa
11000
10
15
20
25
Hold Time, min
30
35
(c)
(d)
Figure 5: Compression strength as a function of (a) initial pressure,
final temperature, (c) final pressure, and (d) hold duration.
CONCLUSIONS
From the experiments the amount of pressure does not significantly contribute to the final
foam structure. Foams processed at the high and low levels had the same structure.
Therefore, the pressures that were studied were sufficient to diffuse nitrogen into the puck.
Further studies need to be conducted to determine the lower limit on the amount of pressure
required to produce a foam structure.
The final temperature did produce some variation in the foam materials. At the lower
temperature the pitch is within the glass transition temperature range. In this range the
material had more integrity and was able to restrict gas flow and bubble growth. When the
foam was at 250C and the pressure was released the pitch was viscous enough to restrict
bubble growth and produce a small pore size foam. At 270C the pitch viscosity was low
enough so the nitrogen was able to diffuse in and out easily. When the pressure was released
the nitrogen nucleated bubbles and the bubbles grew rapidly leaving a large cell foam behind.
The hold time produced variations in the different foam samples. A more uniform pore
distribution was produce with the longer hold time. This could be attributed to the system
have more time to reach an equilibrium at the final processing conditions. After the final
II - 831
temperature is reached and the final pressure is applied, the 30 minute hold time allowed the
pitch system to equilibrate and the nitrogen to further diffuse into the pitch. The more
nitrogen in the pitch the more gas available to nucleate and grow bubbles. Experiments with
additional hold times need to be conducted to determine the effect longer and shorter hold
times have on the foam structure.
The most desirable foam structure was produced with a final temperature of 250C and these
conditions were maintained for 30 minutes before the 10350 kPa of pressure was suddenly
released. These parameters allowed nitrogen to diffuse into the pitch, yet did not weaken the
pitchs ability to restrict bubble growth to produce the small pores. This foam structure had
the desired uniform, small pore size. The compression strength of this structure was the
highest value achieved for all of the foam experiments.
REFERENCES
1.
Hall, R.B. and Hager, J.W., Graphitic Foams as Potential Structural Materials, 21st
Biennial Conference on Carbon Extended Abstracts, 1993, pp. 100-101.
2.
Colton, J.S. and Suh, N.P., The Nucleation of Microcellular Thermoplastic Foam with
Additives: Part I: Theoretical Consideration, Polymer Engineering and Science,
Vol.27, No.7, 1987, pp.485-499.
3.
Colton, J.S. and Suh, N.P., Nucleation of Microcellular Foam: Theory and Practice,
Polymer Engineering and Science, Vol.27, No.7, 1987, pp.500-503.
4.
Dutta, D., Hill, C.S., and Anderson, D.P., Processing, Structure, and Morphology of
Graphitic Carbon Foams Produced from Anisotropic Pitch, Novel Forms of Carbon II,
MRS Symposium Proceedings, Vol. 349, pp. 61-66.
5.
Hager, J.W., Anderson, D.P., and Roy A.K., Structure and Properties of Carbon Foams
from Pitch and Phenolic Precursors, 22nd Biennial Conference on Carbon Extended
Abstracts, 1995, pp.106-107.
6.
Hager, J.W., Anderson, D.P., and Roy, A.K., Progress in Open-Celled Carbon
Foams, Proceedings of 40th International SAMPE Symposium (Closed Sessions),
1995, pg. 42-48.
7.
Hager, J.W., and Anderson, D.P., Idealized Ligament Formation and Geometry in
Open-Celled Foams, 21st Biennial Conference on Carbon Extended Abstracts, 1993,
pp. 102-103.
8.
Anderson, D.P., Gunnison, K.E., and Hager, J.W., Ligament Structure of Open-Cell
Carbon Foams and the Construction of Models Based on that Structure. Novel Forms
of Carbon, MRS Symposium Proceedings, Vol. 270, 1992, pp.47-52.
II - 832
DEVELOPMENT AND CHARACTERISATION OF A

NEW HIGH DAMPING COMPOSITE: MG2SI/MG
Christine Mayencourt and Robert Schaller
Institut de Gnie Atomique, Ecole Polytechnique Fdrale de Lausanne,
1015 Lausanne, Switzerland
SUMMARY: A new, lightweight, high-damping material has been processed and

characterised. It is a magnesium based composite with unidirectional Mg2Si fibres. This
natural composite has been produced by unidirectional solidification of a Mg-Si eutectic alloy
(1.34 wt% Si) or near eutectic alloy (2 wt% Si). Mechanical strength was measured through
tensile tests and the damping capacity was deduced from internal friction measurements in a
low frequency torsion pendulum. The results show that these composites have a tensile
strength comparable with industrial magnesium cast alloys such as AZ63 and a damping
capacity which is 10 to 100 times higher. The mechanical properties and damping capacity are
interpreted in terms of sample microstructure with emphasis on the effects of matrix and fibre
orientation on the dislocation mobility.
KEYWORDS: damping capacity, mechanical strength, magnesium, unidirectional fibres,

dislocations, microstructure
INTRODUCTION
Modern technology tends to suppress mechanical vibrations for social and economic reasons.
Mechanical vibrations produce undesirable noise and can affect the precision of a machinetool or cause fatigue-induced rupture of devices. These mechanical vibrations can be reduced
in many ways, for instance by using external dampers or by increasing the inertial masses.
However, these solutions are not always applicable, particularly in transport systems, which
require lightweight and small structures [1]. Thus, contemporary transport systems need lowdensity materials which exhibit simultaneously a high-damping capacity and a high
mechanical-strength. These properties are often incompatible [2] because the microscopic
mechanisms which are involved in strengthening and in damping are not independent [3]. It
is therefore important to develop new materials such as two-phase composites in which each
phase can play a specific role: strengthening or damping.
Pure magnesium is known to exhibit a very high damping capacity [4], but its mechanical
strength is low [5]. This metal can be hardened by introducing second phase precipitates
which are able to pin dislocations, in order to avoid plastic deformation and thereby to
increase the yield stress. However, if the mechanical properties are better in magnesium
alloys than in pure magnesium, the damping capacity is lost in most of them [5]. A pure
magnesium matrix reinforced with high-strength fibres is hence a good candidate for a highdamping composite. Actually the damping capacity of magnesium is due to dislocation
vibration [6]. In order to obtain a high damping, this implies that the dislocation loops, which
II - 833
are defined between the pinning centres, must be relatively free to vibrate. Thus, the
magnesium matrix has to be relatively pure. In other words, in order to keep the damping
capacity high, attention must be paid to the choice of the alloying elements, whose solubility
in magnesium has to be low enough.
In this context, Mg - Si alloys have been studied as a system of low solubility because silicon
is almost insoluble in magnesium (solubility lower than 30 ppm) [7]. Using the eutectic
composition properties, a Mg-Si eutectic alloy (1.34 wt% Si) or near eutectic alloy (2 wt%
Si) is converted into a magnesium matrix composite with Mg2Si fibres by unidirectional
solidification. The obtained natural composite is formed by a pure magnesium matrix
reinforced with a low fraction of Mg2Si fibres. So the dislocation loops can be long enough to
vibrate and dissipate energy while pinning by fibres avoids plastic deformation.
This paper presents Mg2Si / Mg composites which were processed by unidirectional
solidification and characterised by optical and transmission electron microscopy, by tensile
tests and by mechanical spectroscopy, i.e. measurements of the internal friction and of the
elastic shear modulus as a function of temperature for different samples.
Mg2Si / Mg composites were processed by unidirectional solidification of two different
alloys. The grade A was an eutectic alloy (1.34 wt% Si) of low purity and the grade B was a
near eutectic alloy (2 wt% Si) of medium purity, the mean impurity concentration is about
0.15%. Unidirectional solidification was performed using the same process as the Bridgman
method for single crystal growth [8]. In a controlled atmosphere, a graphite crucible
containing the alloy is moved in a vertical thermal gradient furnace. If the solidification
surface is plane, the solid phase growth is regular and a two phase composite expands with
vertical Mg2Si fibres in a pure magnesium matrix. This solidification mode can occur if the
alloy concentration is the eutectic concentration or near eutectic concentration. The
solidification rate (v) and the distance between fibres () are related by [9]:
v = const.
2
(1)
Plate shape specimens were cut from the solidified rods by spark machining, in such a way
that the sample axis is parallel to the solidification direction, i.e. parallel to the fibres.
The mechanical properties have been studied by classical tensile tests which provide the 0.2%
yield stress, the tensile strength and the rupture strain.
Damping properties were deduced from mechanical spectroscopy measurements [10]. When a
material vibrates under low stress excitation (much lower than the elastic yield stress)
anelastic deformation is responsible for energy dissipation.
In a free-decay inverted torsion pendulum, the internal friction (IF) at low frequency (~2Hz)
and at a strain amplitude of about 10-5 is given by:
IF =
II - 834
1 A0
ln
n A n
(2)
where A0 and An are the amplitude of the first and the nth oscillations. The shear modulus (G)
was determined by the following equation [11]:
G=
8"If 2
a4
(3)
where a is the sample thickness, " its length, I the pendulum inertia momentum and f the
vibration frequency.
A forced, inverted torsion pendulum, which allows one to vary the frequency in a wider range
(10-3 to 10 Hz), was also used. In this case, the internal friction is given by:
IF = tan()
(4)
where is the phase lag between the strain and the applied stress.
RESULTS
Sample Microstructure
Unidirectional samples were processed from both alloys. Fig. 1 displays two pictures obtained
by optical microscopy on an unidirectional sample solidified at 0.5 mm/min. One can see that
the Mg2Si fibres are parallel to the direction of solidification.
(a)
(b)
Fig. 1: a) Lengthways image of an unidirectional sample. b) Transversal image of the same
sample. Mg2Si fibres are parallel to the solidification direction
Depending on the solidification rate (or crucible speed in the furnace), thermal gradient and
alloy purity, the sample structures can be rather different. If all the conditions are proper
balanced, i.e. a high thermal gradient, a low solidification rate and a high alloy purity [8, 9],
the Mg2Si fibres are parallel to the direction of the thermal gradient (Fig. 1). Moreover, the
distance between fibres depends on the solidification rate as predicted by Eqn 1, with a
different constant for each alloy (Fig. 2).
II - 835
Grade A alloy
Grade B alloy
[mm/min]
10
v=cstA
2
v=cstB
8
7
6
5
4
3
8 9
8 9
10
v [mm/min]
Fig. 2: Distance between fibres versus the solidification rate in a log-log graph. According
with the error margins, the points follow the Eqn 1, with two different constants for both
alloys, because of the different compositions
In addition, the matrix also exhibits a crystallographic orientation with the c-axis of the
hexagonal structure parallel to the solidification direction. Fig. 3 is a Transmission Electron
Microscopy (TEM) diffraction pattern associated with a magnesium grain, cut perpendicularly
to the solidification direction. The hexagonal symmetry indicates that the matrix, which has
an hexagonal structure, has its c-axis parallel to the specimen axis, i.e. the basal planes are
perpendicular to the solidification direction.
Fig. 3: MET diffraction pattern for a magnesium grain. The c-axis of the hexagonal structure
is parallel to the solidification direction (photo by Christian Verdon)
Without good conditions, randomly oriented short fibres are obtained, but the matrix
nonetheless exhibits an oriented c-axis. A disoriented matrix can be obtained by natural
solidification. Unidirectional specimens were obtained from grade B alloy for all tested rates,
i.e. from 0.2 to 8 mm/min. Unidirectional specimens were obtained from the grade A alloy
only for low solidification rates, between 0.2 and 0.7 mm/min.
Tensile tests have been performed on these composites. Fig. 4 displays the mean ultimate
tensile strength of different samples, in which the Mg2Si fibres are parallel to the tensile
II - 836
stress, as a function of the distance between fibres. One can see that the tensile strength is
higher when the distance between fibres is lower, i.e. when the solidification rate is higher.
For the shortest distances between fibres, the tensile strength is comparable with the industrial
magnesium cast alloys such as AZ63 [5].
Grade A alloy
Grade B alloy
Tensile strength [MPa]
200
150
100
50
0
0
Distance between fibres [m]
10
12
Fig. 4: Mean ultimate tensile strength of unidirectional samples versus distance between
fibres. The tensile strength is higher when the distance between fibres is smaller. The dotted
line corresponds to the level of the as cast AZ63 alloy
Internal Friction Spectrum
The Mg2Si / Mg natural composites are characterised by a high damping capacity. Fig. 5
displays the internal friction (IF) and the elastic shear modulus (G) as functions of the
temperature, measured in a specimen annealed during 4 hours at 600 K. Curves were
measured by means of a low frequency torsional pendulum on a sample of grade A alloy
solidified at 0.5 mm/min, but the damping general features are similar in all studied samples
and with all measurement methods.
The internal friction is high and rather constant. It exhibits a broad maximum at about 200 K
and a smaller one at about 400 K. The shear modulus exhibits an unusual behaviour: although
modulus usually increases with decreasing temperature, this material presents a decreasing
modulus with decreasing temperature between 430 and 350 K. Both the damping and the
shear modulus exhibit large hysteresis during thermal cycling: damping is higher and modulus
lower during heating than cooling. When the temperature is kept constant after heating, the
damping decreases continuously and the modulus increases until they reach the values of
isothermal equilibrium.
II - 837
IF *1000
40
G [GPa]
20
IF
19
30
18
20
17
10
16
15
14
100
200
300
T[K]
400
500
Fig. 5: Internal friction (IF) and shear modulus (G) versus temperature in a sample annealed
4 hours at 600K. IF is high and presents two maxima at 200 and 400K. G presents an unusual
decrease between 430 and 350K. Both curves exhibit a large hysteresis
Effects of the Measurement Parameters on the Internal Friction

One way to understand the microscopic origin of the damping is to study the internal friction
spectra, in particular the maxima of internal friction. Actually the microscopic phenomena,
such as twinning or gliding of dislocations interacting with points defects, induce specific
behaviour of the internal friction as functions of different parameters, as the strain amplitude
or the frequency [10, 12]. The study of these dependencies generally allows one to determine
some characteristic values of the phenomenon such as its activation energy and thus to deduce
the involved phenomenon.
The study of the internal friction at different frequencies does not reveal a clear shift of the
maxima, although a modification of the curve shape is observed. Fig. 6 displays the internal
friction measured at room temperature (290K) as a function of the strain amplitude. The curve
indicates a clear dependence on the amplitude. This reveals that the involved microscopic
mechanism is complex.
12
IF *1000
10
8
6
4
2
0
5
10
15
20
25
30
-6
35x10
Fig. 6: Internal friction versus strain amplitude at 290K. The internal friction depends on the
strain amplitude
II - 838
Moreover, the internal friction also depends on the heating or cooling rate. Fig. 7a displays
the internal friction and the shear modulus versus the temperature at different heating rate

( T ). One can observe that the internal friction is higher and the modulus lower when the
heating rate is higher. Furthermore, Fig. 7b indicates that this dependence is complex. When

T 0, the damping decreases linearly with , but the linear extrapolation for T tending to zero

(dotted line) is higher than the measured value for T = 0. When heating is stopped, the
stabilisation of the internal friction take a longer time than the stabilisation of the temperature.

So it is possible to distinguish two effects: linear dependence with T value (3) and time
dependence at constant temperature (2). These two contributions are added to the intrinsic
internal damping (1).
17
35
IF *1000
45
30
15
14
25
48
46
16
G [GPa]
40
2
1
0.5
IF
IF *1000
50
44
42
40
1
20
a)
13
160
180
200
220 240
T [K]
260
280
300
38
b)
0.0
0.5
. 1.0
1.5
2.0
Fig. 7: (a) Internal friction (IF) and shear modulus (G) versus temperature at different
heating rates: 0 K/min, 0.5 K/min, 1 K/min, 2K/min. The internal friction increases and the
shear modulus decreases when the heating rate increases, (b) Internal friction (IF) versus

heating rate ( T ) at 220K. The internal friction is composed of three parts. (1) Intrinsic
damping, (2) time dependant damping and (3) heating rate dependant damping
Effects of the microstructure on the internal friction
To enhance the composite properties, it is important to localise and to understand the
mechanism of damping. Another way to obtain this information is to study the damping
behaviour as a function of the sample structure.
Fig. 8 displays the internal friction versus the temperature measured with the forced torsion
pendulum in different samples, in which the c-axis of the hexagonal structure of the Mg
matrix is parallel to the sample axis. One can see that the Mg2Si fibre phase morphology has
no obvious influence on the damping capacity.
Fig. 9 presents the internal friction spectra of three specimens: one specimen (1) where the
Mg matrix c-axis and the Mg2Si fibres are parallel to the specimen axis, a second one (2)
where only the Mg matrix c-axis is aligned along the specimen axis, the Mg2Si fibres
exhibiting no preferential orientation, and a third one (3) in which neither the Mg matrix nor
the Mg2Si fibres show any orientation. In the two first cases, where the Mg matrix c-axis is
oriented, the internal friction level is high. On the contrary, the internal friction is lower by a
factor two or three in the sample which does not exhibit any structural orientation. One can
deduce that neither the distance between fibres nor the fibre orientation modify significantly
II - 839
the internal friction level. On the contrary, the orientation of the matrix has a great influence
on damping.
50
IF *1000
40
30
20
10
0
100
200
300
T [K]
400
500
Fig. 8: Internal friction versus temperature in unidirectional samples of grade B alloy with
different solidification rate: (1) 8 mm/min, (2) 5 mm/min, (3) 2 mm/min, (4) 1 mm/min, (5) 0.7
mm/min, (6) 0.5 mm/min, (7) 0.2 mm/min. The distance between fibres has no clear effect on
the damping capacity
40
IF *1000
30
2
20
10
3
0
100
200
300
T[K]
400
500
Fig. 9: Internal friction versus temperature in an unidirectional fibres/matrix sample (1), in

an oriented c-axis matrix sample (2) and in a disoriented sample (3). Both the oriented c-axis
matrix samples IF spectra have similar shape and height, but disoriented sample has much
lower amplitude
DISCUSSION
Sample Structure
Both alloys can be unidirectionally solidified, with Mg2Si fibre structures in agreement with
Eqn 1, but the rates which are required to solidify unidirectional composites and the constant
II - 840
which appear in Eqn 1 are different. This means that the involved solidification mechanism is
the same in both cases and correspond with the well known theory of unidirectional
solidification [8, 9], but the involved parameters are merely different.
The mechanism of unidirectional solidification is based on the fact that both phases solidify
together at the same time. The distance between fibres depends on the solidification rate as
shown in Eqn 1, because it corresponds to an equilibrium between the needed energy to
initiate a new phase and the energy to grow the existing phases. The major difference between
the two alloys is the impurity concentration. In fact, the grade A alloy contains more
impurities. All impurities are sites which might seed new phases, so the constant appearing in
Eqn 1 depends on the alloy purity. Moreover, these impurities disturb the regular growth of
the fibres, and the unidirectionality can be obtained only if the solidification rate is slow
enough. A pure enough alloy is hence necessary to obtain a small distance between fibres. As
the grade B alloy is purer than the eutectic alloy, finer structures can be obtain with this alloy.
The preliminary tensile tests are very promising. Although a large dispersion of the results,
one can clearly see that the tensile strength is enhanced when the distance between fibres is
small, i.e. when the solidification rate is high. These tests were performed on prismatic
samples with a small cross section of 6 mm2. Some edge effects are certainly responsible for
the dispersion of the results, because the stress accumulation at corners facilitate the
formation of cracks. Furthermore, the presence of oxides induces some porosity during the
solidification, which decreases the mechanical strength.
Future work will be to optimise the processing of this composite and to find the best sample
configuration for tensile tests. But already, for the lowest distances between fibres, the tensile
strength is comparable with the industrial magnesium cast alloys such as AZ63.
Internal Friction Spectrum
Modulus Anomaly
Materials usually have a decreasing modulus as temperature increases. In these composites,
on the contrary, the shear modulus increases between 350 and 430 K. This anomaly can be
interpreted as follows: At the solidification temperature, the composite is at the equilibrium as
it concerns the internal stresses, i.e. no internal stress appears between matrix and fibres.
When the composite is subjected to an external stress such as torsion, the fibres are in
extension. The high Young's modulus of the fibres acts as a restoring force, so the shear
modulus of the composite is enhanced. During cooling, due to the thermal expansion
coefficient mismatch, the internal stresses at the interface between fibres and matrix increase.
These stresses can be first relaxed by dislocation creation and motion, but when the
temperature decreases below a given temperature this relaxation is not sufficient, and the
internal stresses compress the fibres. These compressed fibres are bent, they are no longer in
extension under the external stresses and thus they can have no more effect on the shear
modulus. So the shear modulus of the composite decreases to the modulus of the pure
magnesium.
II - 841
In other words, at high temperature the fibres participate in the modulus: the modulus is high
and at low temperature the fibres do not participate in the modulus, hence, the modulus is
low. This phenomenon occurs during each thermal cycle and explains the modulus anomaly.
Damping Behaviour
The samples exhibit a high damping, as was desired. But, the internal friction level depends
strongly on the sample microstructure (Fig. 8). In order to manage the internal friction level, it
is thus important to localise and understand the mechanisms which lead to the high damping.
Fig. 6 and 7 indicates that the involved phenomenon is complex, nevertheless some possible
interpretations can be explored.
Fig. 8 and 9 show that only the matrix orientation significantly influences the internal friction,
but neither the fibre orientation nor the distance between fibres does. Hence, the observed
effects may be due to a microscopic phenomenon which occurs in the matrix, such as motion
of dislocations, twinning or defect movement. It is well known [2, 4, 6] that internal friction
in pure magnesium is due to dislocation motion in the basal planes. Therefore, the composites
have a rather good damping capacity if the applied stress is in a preferred orientation to
induce dislocation vibration and if the dislocations are free enough to vibrate. This explains
the fact that the damping capacity is higher when the c-axis of the matrix is perpendicular to
the stress, i.e. when the c-axis is aligned along the sample axis, as shown in Fig. 9.
In other words, if the matrix is c-axis oriented and if the magnesium matrix is pure, the
damping must be high. As the matrix c-axis orientation is guaranteed when the fibres are
unidirectional, mechanical properties of this composite can be optimised without care to the
damping capacity. As presented in the introduction, the high-damping capacity and the highmechanical strength can be reached in these composites, because the involved mechanisms
are independent.
However, a large dispersion in the internal friction occurs between each measured sample
(Fig. 8). Previous works [4, 6] have shown that the dislocation motion in the basal plane is
controlled by their interaction with points defects, more precisely, the interaction consist in
pinning and depinning of the dislocations from the points defects. Thus internal friction
spectra are very sensitive to the impurity concentration. Fig. 6 confirms this interpretation, but
the shape modification with the frequency indicates that the interaction between dislocations
and point defects is not a pure pinning-depinning mechanism. The dependence of the internal
friction on the frequency permits one to suppose a dragging of point defects of higher
mobility.
Transient Internal Damping

The internal friction depends on the heating or cooling rate ( T ). Such a dependence, called
transient internal damping, has already been observed in some composites [13]. This is due to
the thermal expansion coefficient mismatch between the matrix and the fibres. But, the

dependence on T value is more complex than effects commonly reported in literature. Usual

dependence is linear, but here there are two different effects depending on the T value. Fig. 7

shows that when T 0 damping decreases linearly with T , as studied in literature. The main

characteristics of such an effect is a dependence with T /(0) where 0 is the applied stress
amplitude and is the circular frequency.
II - 842
However, Fig. 7 displays also that the internal friction measured in isothermal conditions (at

T =0) is lower than expected by extrapolation of the spectra measured for T 0. Moreover,
when heating is stopped, the internal friction takes a longer time to stabilise than the
temperature. Internal friction is then composed of three contributions: the intrinsic damping
measured in isothermal conditions, a contribution which depends on the temperature

stabilisation time and a contribution which depends on the heating or cooling rate T .
CONCLUSION
A new composite material has been developed and characterised. It is a natural composite
formed by unidirectional solidification of a Mg-Si eutectic alloy (1.34 wt% Si) of low purity
(grade A) or near eutectic alloy (2 wt% Si) of medium purity (grade B). Both the fibres and
the matrix c-axes can be aligned in the direction of solidification. The distance between fibres
depends on the solidification rate and on the alloy purity. The grade B alloy give access to the
finest fibre structure, because of its higher purity. The unidirectional fibres enhance the
mechanical strength up to some industrial magnesium cast alloys such as AZ63. The matrix
orientation provides a very high damping capacity over a wide range of temperatures, without
influence of the distance between fibres. These natural composites exhibit mechanical
properties comparable with the ones of classical magnesium alloys with a damping capacity
100 time higher.
ACKNOWLEDGEMENTS
The authors would like to thank Mr. Christian Verdon for his observations in Transmission
Electron Microscopy. This work has been partially supported by the Swiss Priority Program
for Materials.
REFERENCES
1.
R. Van Fleteren, Magnesium for automotive applications, Advanced Materials &

Processes 5/96, 33
2.
K. Sugimoto, Basic and Applied research on High-Damping Alloys for Application to

Noise Control, Mem. Inst. Sci. Res., vol.35, 31-44, 1978
3.
R. Schaller and J. Van Humbeeck, High Damping Materials, Materials Science Forum
Vol. 119-121, 803-808, Switzerland, 1993
4.
S.M. Seyed Reihani, C. Esnouf, G. Fantozzi and G. Revel, Effect of impurities on the
internal friction spactrum of magnesium after palstic deformation, J. de physique Tome
42, C5, 145-150, 1981
5.
Metal Handbook, Vol.2, Ninth Edition, 1979
II - 843
6.
K. Sugimoto, K. Niiya, T. Okamoto and K. Kishitake, A study of Damping Capacity in

Magnesium Alloys, Trans. JIM, Vol. 18, 277-288, 1977
7.
T.B. Massalski, Binary alloy phase diagrams, vol.2, American Society for Metals, Ohio,
1986
8.
W. Kurz, D.J. Fischer, Fundamentals of Solidification, Trans Tech Publications, 1984
9.
M.C. Flemings, Solidification Processing, Materials Science and Engineering Series,

1974
10. A.S. Nowick and B.S. Berry, Anelastic relaxation in crystalline solids, Materials science
series, 1972
11. S. Timoshenko, Rsistance des matriaux, Librairie polytechnique Branges, Paris, 1963
12. G. Gremaud, The Hysteretic Damping Mechanisms Related to Dislocation Motion, J.
Phys., Tome 48, C8 15-30, 1987
13. X. Zhou, R. Fougres and A. Vincent, Internal damping effects due to the thermal
expension mismatch between aluminium and silicon in an Al 11.8%Si alloy, J. Phys.,
2185-2201, 1992
II - 844
INTERFACE STUDY OF COMPOSITE BANDS

FORMED BY POWDER PLATING AND ROLLING
(PPR)
Sun Yong1, Zhang Shuhong1, Zhang Daiming1, Qian Tiancai2, Gao Yun2, Li Bin2
1
Department of materials, Kunming University of Science and Technology,

Kunming 650093, Yunnan, P.R.China
2
Yunnan Analytic & Testing Centre, Kunming 650051,Yunnan, P.R.China
SUMMARY: Different kinds of metal powder have been plated on a band of mild steel and
then both of them rolled toghter. With this new method, some composite bands such as Custeel, Al-steel, Sn-steel, etc., have been made. TEM, EPMA, XPS and XRD are used to study
the interface among powder particles, the boundary of grains inside powder particles particles
and the interface between composite-layered and matrix. The metal powders are deformed by
force and their surface areas increase two or three times. There exist some pieces of oxidation
film at the interface between powder particles, and there is a lot of dislocation-substructures in
the grain inside powder particles. The SADP changes into multiple spots or small arcs, and
take on a form of a discontinuous ring distribution. After annealing, a diffusion phenomenon
can be observed at the interface among powder particles and between composite-layer and
matrix. Three conclusions are given.
KEYWORDS:
metals composite, surface coating, powder processing, multi-layers
interfaces, laminar materials
INTRODUCTION
It is well known that there are many methods for forming composite [1-3]. There also has
been a lot of research focused on the structures and properties of composite [4-6]. PPR
(Powder Plating and Rollingis a new method for forming composite bands [7]. With PPR,
which has been patented, metal powders (chosen as composite-layered metal) are plated onto
the other metal band (chosen as the matrix) and then both of them are rolled together to obtain
a composite band. In consideration of the difference of the presently reported technology from
others, i.e. the composite-layered metal is not a sheet but powder, the interfaces of composite
band formed by PPR are different from those of composites formed by other methods. In this
paper, the main purpose is to study the interface among powder particles, the boundary of
grain inside powder particles and the interface between the composite-layer and matrix. The
effect of annealing temperature on the interface and microstructure of composite bands is also
discussed. These will provide useful information for controlling the properties of composite in
real production.
II - 845
EXPERIMENT
Plating a mild-steel band with different kinds of metal powder, and then rolling them together,
a system of composite bands, such as Cu-steel, Al-steel, Sn-steel can be prepared. For
different metal powder, a variant of annealing temperatures are employed in a vacuum
furnace. Cutting out the composite bands into pieces, the samples for observing microstructure
of composite bands are ready.
By rubbing the matrix off the sample, transforming the composite-lamella into thin foils. The
thin foils are prepared by the electrochemistry-polishing method and examined in TEM of
2000-EX with accelerating voltage of 200 kv.
Samples for observing in EPMA are prepared by the way the same with the sample for
observing in OM and then the samples are examined in EPMA of EPMA505.
Some samples are analyzed by using XPS and XRD.

Interfaces Among Powder Particles
Different from traditional composite materials, surface composite layer is an even substance
formed of a multitude of metal powder particles that are elongated in all directions, one on the
top of another in the deforming process. Therefore, there exist a large number of interfaces
among particles. The binding at the interfaces is of importance. Before being pressed tight,
powder particles with bigger diameter will begin to undergo a deformation while powder
particles with smaller diameter will slide themselves into the gaps among the particles
Fig.1: Deforming process of metal powder particles Fig. 2: TEM picture of the interfaces of aluminum powder particles
with bigger diameter. With deforming pressure increasing, powder particles will gradually
flatten out, close to each other, the gaps becoming smaller and smaller (See Fig.1a). Finally,
the powder particles overlap each other and combine into continuous lamella. The appearance
shows that metal particles have been cold welded into an even and continuous thin layer
covering the surface of the matrix metal (See Fig.1b). An analysis finds that the surface areas
of metal powder particles expand two or three times after undergoing deformation. The
oxidation film wrapping the metal powder particles bursts, exposing a lot of clean and non-
II - 846
oxidation surfaces. It is these clean non-oxidation surfaces that enable the metal particles to
weld together closely. Table 1 presents the statistics of the deformation of metal powder
particles.
Fig.2 shows that the broken fragments of aluminum oxide film can be observed at the
interfaces of aluminum powder particles.
Table 1: Statistics of the deformation of the metal powder particles
metal
powder
Al
Cu
Sn
original
diameter
(mm)
0.079
0.055
0.055
thickness after
deformation
(mm)
0.01
0.005
0.005
average diameter average elongation increase rate of

after deformation
rate
surface areas
(mm)
(%)
(%)
0.185
134
174
0.149
170
267
0.152
176
282
Boundary of Grain Inside Powder Particle

It can know from the above analysis that metal powder particles undergo a considerably big
deformation in the composite processing with PPR. During the composite processing, the
grains inside powder particle undergo a considerably big deformation too. The result of XRD
examination shows that the grain's size inside a powder particles is about 1200--1400
before deformation while the size becomes about 240--280 after deformation. This shows
that deformation affects a strong crushing function on metal grains. An observance reveals
that there exist concentrated dislocation groups (See Fig.3a) inside grains. These
dislocation lines tangle together, forming dislocation-substructures. The results of SADP can
(a)
(b)
Fig. 3: Dislocation line and SADP inside grains
of copper powder particle
Fig. 4: Dislocation outcrop copper

powder particle etched tanks at the
interface
of
the
dislocationsubstructure
inside
grain
II - 847
be seen in Fig.3(b), in which the diffraction spots change into multiple spots or small arcs,
and take on a form of a discontinuous ring distribution. This means that the crystals inside the
grains are crushed and been super-fineness under the large deformation forces, and these
dislocation-substructures are of small phase difference in selected area. Fig.4 clearly shows
the dislocation outcrop etched tanks at the interface of the dislocation-substructure inside the
grain. After recrystalization annealing treatment, the original grain boundaries and dislocation
line are hard to find under TEM. The substructure inside grain is replaced by annealing twin
crystals or normal crystals (See Fig. 5 and Fig. 6).
Fig. 5:Texture of copper grains after annealing Fig. 6: Texture of tin grains after annealing
Binding Interface Between Composite-Layer and Matrix Metal
By using XPS, the characteristic binding energy peaks of Cu and Al can be found at the
binding surface of the steel matrix (See Fig.7 and Fig.8). The analysis results show that under
a strong external force the powder particles and the matrix are made to be close to each other
to atomic size, producing a metallic bond binding between them.
The observation about the composite bands which has undergone diffusion annealing with
EPMA also reveals that the composite layer metal is closely bond with the matrix and the
metal atoms on the both sides of the interface can cross the interface between powder
composite layer and matrix by means of diffusion to form a diffusion bonding. Fig.9 and
Fig.10 present the second electronic images of the Al-steel, Cu-steel composite bands by
using EPMA. The results of a linear scanning show that there is a clear component gradient
changes at the binding interface, which shows clearly that there is a diffusion occurring at the
interface between the composite layer and the matrix under the diffusion annealing
temperature but no metal compound is formed. An examination of the bonding strength of the
interface also reveals the good bonding.
II - 848
Fig. 7: XPS analytical spectrum line at the bonding surface of the steel matrix of aluminumsteel composite band
Fig. 8: XPS analytical spectrum line at the binding surface of the steel matrix of copper-steel
composite band
Fig. 9 Binding interface of aluminum-steel

band
Fig. 10: Binding interface of copper-steel

composite band.
II - 849
CONCLUSIONS
(1) In the preparation of PPR, the surface areas of metal powder particles increase two or
three times, which makes the oxidation film of powder particles burst and fall down, exposing
a lot clean and non-oxidation surface. It is these clean non-oxidation surfaces that ensure a
bind among metal particles and between grain composite layer and matrix.
(2) After deformation, there exist much interfaces of powder particle inside composite layer
and there exist much boundary inside grain in every powder particle. Therefore, it is
important a good bind among the interfaces of powder particles and the boundaries of grains.
(3) A proper annealing treatment can make atoms spread cross the interface, thus causing the
disappearance of the interfaces of powder particles which have overlaped each other, form an
even continuous metal thin layer by recrystallizing. The dislocation-substructures inside
grains is replaced by normal crystals. A diffusion bonding is formed at the interfaces between
composite-layer and matrix through atomic diffusion.
ACKNOWLEDGEMENT
This project is supported by the National Natural Science Foundation of China
REFERENCES
1. M. M. Michorius, Chemical vapor deposition finds more and more uses, Proces.
Technologie, 6(9), 1990, pp.18-21
2. A.Kelly, Novel forming methods for composites, Inst. Phys. Conf. Ser.,111,1990, pp.5-12
3. Kang xiancheng, The forming methods and using for metals composite sheets,
Guowaikeji, 10, 1990, pp.1-4
4. A. Lawley, Characterization of microstructure in metallic and composite materials, Dep.
Mater. Eng., Drexel Univ., Philadelphia, USA, Report 1987
5. W. Wlosinski, Bonded multi-layer materials-the materials of the future, Trans. JWRI,
17(1), 1988, pp.263-72
6. Shi Q. N., Sun, Y., Zhang D. M., Zhang S. H., A Study of Rolling Force in Composite
Rolling Process with Metallic Powder and Metallic Sheets, MSMM96, Beijing,1996.
7. Sun Y., Shi Q. N., Zhang S. H., Zhang D. M., The Microstructure of Composite Bands
Formed by Powder Plating and Rolling ( PPR ), Journal of Materials Processing
Technology, Vol. 63, Nos. 1-3, 1997, pp.438-441.
II - 850
MECHANICAL PERFORMANCE OF COMPRESSION

MOLDED AND INJECTION MOLDED BLENDS OF
LARC POLYIMIDE WITH POLYETHERIMIDE (PEI)
David Olivero1, Donald Radford1, Sarah Hummel1, Charles Smith2 and Betty Tung3
1. Composite Materials, Manufacture and Structures Laboratory (CMMS), Department of
Mechanical Engineering, Colorado State University, Fort Collins, CO 80523
2. Department of Industrial Science, Colorado State University, Fort Collins, CO 80523
3. IMITEC, Inc., 1990 Maxon Rd., Schenectady, NY, 12308
SUMMARY: Two high temperature polyimides from NASA-Langley, LaRC-SI and
LaRC-IAX, were injection molded into tensile coupon shapes. A custom injection die was
developed with built-in electric heaters to generate the required die temperatures. LaRC-SI
and LaRC-IAX tensile coupons were produced by both injection molding and compression
molding for performance comparison. Compression molding and injection molding yielded
similar mechanical performance. However, non-optimal processing conditions and
contamination reduced strength and strain to failure in some specimens. To evaluate trade-offs
in blending LaRC-SI and LaRC-IAX with lower cost thermoplastics, PEI blends were studied,
prepared by both compression and injection molding. Overall, mechanical properties followed
a rule of mixtures relationship, with variations resulting from non-optimal processing rather
than blending. Preliminary thermal analysis highlighted extremely different blending related
to process conditions. Compression molding resulted in a binary polymeric system while
injection molding resulted in a single, completely mixed phase.
KEYWORDS: LaRC-SI, LaRC-IAX, PEI, polyimide thermoplastic, injection molding
INTRODUCTION
Two high temperature thermoplastics, LaRC-SI and LaRC-IAX, have recently been
developed by NASA-Langley. These aromatic polyimides exhibit high strength and a high
glass transition temperature (240C), and have been identified for use as adhesives, films and
composite matrix material [1]. Composite processing has already been demonstrated,
particularly in prepregs of IM7 [2,3] Additional features of both LaRC-SI and LaRC-IAX
include good chemical and oxidation resistance and high dielectric strength.
Thus far, LaRC-SI and LaRC-IAX, in powder form, have been successfully compression
molded. However, compression molding can be a limiting and high cost manufacturing
process. This fact, plus the high cost of LaRC-SI and LaRC-IAX limits the materials'
applicability. Injection molding of these polyimides is thus of interest as a significant
improvement in manufacturing potential. This paper discusses the results of preliminary
injection molding trials with LaRC polyimides from pelletized stock.
To further increase the viability of this material, both LaRC-SI and LaRC-IAX have been
blended with ULTEM 1000, an amorphous polyetherimide (PEI) with a glass transition
temperature of 212C. PEI is currently much less expensive than LaRC, yet with comparable
II - 851
physical properties and only slightly reduced mechanical performance. Blending of these
thermoplastics has therefore been investigated as a way to increase the economy of LaRC-SI
and LaRC-IAX.
Tensile tests were undertaken to investigate the effects of both processing method and PEI
blend ratio on the mechanical performance of LaRC-SI and LaRC-IAX. To investigate the
presence of any thermal degradation, as well as the degree of mixing of PEI with each LaRC
polyimide, differential scanning calorimetry (DSC) traces of specimens produced by both
compression and injection molding are presented.
EXPERIMENTAL
Compression Molding of LaRC-SI, LaRC-IAX and Blends with PEI
Compression molding of LaRC-SI/PEI and LaRC-IAX/PEI blends was carried out with a 20
ton hot press. Powders of LaRC-SI and LaRC-IAX were used to produce all compression
molded specimens. At the time of this study, PEI was not available as a powder. Sheets of PEI
were therefore milled into small flakes, and dry blends of the thermoplastics were well mixed.
Unfortunately, this process of milling PEI introduced impurities which, after molding,
resulted in stress concentrating defects in some tensile coupons.
Compression molding of all blends was carried using process parameters obtained with past
experience compression molding LaRC-SI. To produce blends with good consolidation and
no thermal degradation, a processing temperature of 320C was used. An aluminum die with a
63.5 mm x 177.8 mm x 25.4 mm cavity was used to prepare all specimens. The resulting
63.5 mm x 177.8 mm sample was subsequently machined into three separate ASTM D-638
type #1 tensile coupons. Machined specimen edges were sanded to eliminate machining
induced surface flaws.
Injection Molding of LaRC-SI, LaRC-IAX and Blends with PEI
A two-ton injection molder was used to investigate the injection molding potential for
LaRC-SI and LaRC-IAX. The molder, pictured in figure 1, was a screw injector with
individual temperature control of the hopper, preheat, screw, nozzle and mold zones.
Figure 1: Two-ton injection molder at CSU

All process zones in the molder are capable of temperatures approaching 400C, which were
found to be necessary to mold these high temperature thermoplastics, except for the water
heated mold, which has a maximum temperature near 100C. To prevent freezing at the mold,
II - 852
a steel die was custom modified with electric cartridge heaters to enable mold temperatures
above 220C. The geometry of the die cavity was milled to ASTM D-638-89 type #1 tensile
coupon specifications, with the exception of a draft angle milled into the cavity walls to
facilitate part ejection.
As moisture absorption is an ongoing concern in injection molding plastics, all pellet blends
were dried in a recycling desiccant hopper at 140C for at least 15 hours just prior to injection
molding. A given blend ratio was determined to be successfully molded once 20 specimens
with good quality, consistency and reasonable absence of contaminants were continuously
produced. This procedure allowed the investigation of a variety of blend ratios, from 0%
polyimide to 100% polyimide (by volume) in 25% increments, with the limited quantity of
LaRC-SI and LaRC-IAX at hand. A portion of the over 200 tensile coupons test coupons
produced are pictured in figure 2. Three specimens were produced for each material, blend
ratio and processing method combination. That is, a given blend specimen was produced by
both compression molding and injection molding for process comparison.
Tensile Testing
Tensile tests were performed at room temperature (25C) with an ATS universal tester fitted
with wedge lock grips. In some cases the grip sections of the tensile coupons were roughened
to prevent slippage in the grips. A load rate of 5 mm/min was used in all tests, as per ASTM
standard D-638. A clip gage extensometer with a 25.4 mm gage length was used to measure
strain. Stress and stiffness for each coupon were calculated based upon the average gage
thickness, which generally varied less than 0.05 mm across a given gage section.
Figure 2: LaRC-SI and LaRC-IAX tensile test specimens

Differential Scanning Calorimetry (DSC)
Of great interest is the effect of processing method and conditions on the resulting blend of
PEI with LaRC-SI or LaRC-IAX. In particular, the degree of mixing of the two constituents,
investigated by looking at changes in glass transition temperature, gives an initial indication
of blend quality. To investigate processing effects, two samples were prepared and tested with
a Seiko DSC 220C analyzer, one sample taken from an injection molded sample and one from
a compression molded sample. Each sample was scraped with a razor blade from the grip
section of a 50% LaRC-SI, 50% PEI coupon after tensile testing.
II - 853

Both LaRC-IAX and LaRC-SI, as well as blends with PEI, were successfully injection
molded. Further, blends were found to retain the strength of the neat polymers, indicating the
miscibility of the two constituents. A significant amount to data scatter was obtained,
however, and is attributed to premature failures in some of the specimens. Typical problems
in injection molded specimens included splay marks, white streaks and incomplete cavity fill.
These problems can be related to the inability to fully optimize processing parameters such as
temperature and pressure with the limited material at hand. Compression molded specimens
were far more consistent in their quality, due to extensive processing experience with LaRC
polyimides. Contamination was a significant problem for both processing methods, however,
and was generally intermittent and unpredictable.
LaRC-SI blends exhibited greater ductility and toughness than LaRC-IAX blends. LaRC-SI
blends were characterized by plastic flow associated with neck down, while LaRC-IAX
blends were almost exclusively brittle in their fracture behavior. For the process parameters
employed, the stress-strain behaviors observed in these materials are shown in figure 3
(a and b), which can be considered representative of mechanical behavior of LaRC-SI and
LaRC-IAX.
2 5 C - 1 0 0 % L a R C - IA X - In je c t io n M o ld e d
140
120
Stress (MPa)
100
80
60
40
20
0
0 .0 0
0 .0 2
0 .0 4
0 .0 6
0 .0 8
0 .1 0
0 .1 2
0 .1 4
0 .1 6
0 .1 8
S tra in (m m /m m )
(a)
25 C - 100 % L aR C -S I - In jec tio n M o ld e d
140
120
Stress (MPa)
100
80
60
40
20
0
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
S tra in (m m /m m )
(b)
Figure 3 (a & b): Relative stress-strain performance of LaRC-SI and LaRC-IAX
Higher failure strengths were usually obtained with LaRC-SI than with LaRC-IAX.
The highest ultimate tensile stresses obtained with LaRC-IAX blends were 110-115 MPa,
while in LaRC-SI blends was 135 MPa. However, as seen in figure 3, the failure strength of
II - 854
LaRC-IAX is similar to the yield point of LaRC-SI. LaRC-SI blends thus appear to be
roughly 10-20% stronger, although this reduction may originate from reduced flaw
sensitivity.
LaRC-SI/PEI Blend Tensile Tests
Figure 4 relates the elastic modulus and ultimate tensile strength of LaRC-SI blends with PEI
to the blend composition. Based on investigations of specimen failure mode, it was observed
that low strength specimens failed prematurely at a point defect or impurity. Figure 4 thus
shows the best performance obtained for each blend ratio, as these results are considered the
most accurate assessment of the material.
130
5.5
120
110
4.5
100
90
3.5
80
70
Inj. Molded Modulus

Inj. Molded Strength
Comp. Molded Strength
Linear (Inj. Molded Modulus)
2.5
Comp. Molded Modulus

Linear (Comp. Molded Modulus)
2
0%
Ultimate Tensile Strength (MPa)
140
Elastic Modulus (GPa)
6.5
25%
50%
75%
60
50
100%
Volume Percent LaRC-SI
Figure 4: Elastic modulus and ultimate tensile strength vs. blend ratio for LaRC-SI/PEI
The strength of injection molded and compression molded LaRC-SI blends are very similar,
indicating no strength tradeoff with injection molding. For each processing method, as the
volume percent of LaRC-SI increases, the ultimate tensile strength increases in a linear
fashion, from 110 MPa to over 130 MPa. Injection molded specimen modulus is consistently
lower than the corresponding modulus of compression molded specimens. The origin of this
difference is unclear, although it is less than 10% in most cases.
LaRC-IAX/PEI Blend Tensile Tests
Figure 5 relates the elastic modulus and ultimate tensile strength in LaRC-IAX blends with
PEI to blend composition in the same manner as figure 4. No clear trend in ultimate tensile
strength is apparent with respect to blend composition in either compression or injection
molded LaRC-IAX specimens. Blends of LaRC-IAX failed in a predominantly brittle fashion
(recall figure 3a), and impurities had a significant effect on the ultimate strength due to flaw
sensitivity. It should also be noted that the semi-crystalline nature of LaRC-IAX may have
II - 855
played a role in making the generation of consistent blends more difficult than with LaRC-SI,
which is amorphous. Further, the injection molding process was not as well optimized as with
LaRC-SI, and so specimens of acceptable quality were more difficult to achieve.
130
5.5
120
110
4.5
100
90
3.5
80
70
Inj. Molded Modulus

Inj. Molded Strength
Comp. Molded Strength
Linear (Inj. Molded Modulus)
2.5
Comp. Molded Modulus

Linear (Comp. Molded Modulus)
2
0%
Ultimate Tensile Strength (MPa)
140
Elastic Modulus (GPa)
6.5
25%
50%
75%
60
50
100%
Volume Percent LaRC-IAX
Figure 5: Elastic modulus and ultimate tensile strength vs. blend ratio for LaRC-IAX/PEI
Injection molded specimen modulus, shown in figure 5, is consistently lower than the
corresponding modulus of compression molded specimens, just as in figure 4. The difference
between the two processing methods is again less than 10% in most cases. The general trend
is a slight increase in modulus as the amount of LaRC-IAX is increased. Data roughly follow
a rule of mixtures relationship, with no significant deviations.
Differences Between Compression Molded and Injection Molded Specimens
Compression and injection molding processes resulted in blends with similar yet highly
scattered mechanical performance. Failures outside the gage section (indicating brittleness
and high flaw sensitivity) were more prevalent in compression molded specimens. This is
related to the presence of contaminants in the milled PEI prior to processing, which
introduced defects and provides an explanation for the increased number of failures outside
the gage length. Some injection molded specimens failed in this fashion, also at point defects
related to contaminants. An example of an out-of-gage failure is seen in figure 6, a cross
sectional micrograph of an internal flaw generated fracture surface in a 25% LaRC-IAX, 75%
PEI injection molded part. The flaw is located at the center of the image.
All samples were observed to have elastic moduli in the range 3.2-3.7 GPa. Changes in
modulus did not correlate with ultimate tensile strength. Failure strain, however, was clearly
observed to increase with strength, as one would intuitively expect based on a Hookean
relationship between stress and strain, given a more or less constant modulus.
II - 856
1 mm
Figure 6: Cross-section micrograph of brittle fracture surface generated by internal flaw
outside gage section. Specimen is 25% LaRC-IAX, 75% PEI, injection molded.
Differential Scanning Calorimetry (DSC)
Figure 7 and 8 show traces generated by DSC analysis of compression and injection molded
specimens, respectively. Figure 7 shows two distinct "dips" in the DSC trace, one located at
the Tg of PEI (212C) and one at the Tg of LaRC-SI (239C). The sample thus contains both
polymer species, as separate phases rather than an actual blend. Note in figure 8, however,
that in the injection molded specimen there is only one intermediate glass transition occurring,
at 225C. This indicates complete mixing of the two species, and the creation of a miscible
blend of PEI and LaRC-SI. Thus, the compression and injection molding processes seem to
produce blends that are fundamentally different. This difference may be due to the increased
mixing, as well as the higher temperature and pressure of injection molding.
DDS
DSC
ENDOTHE
Figure 7: DSC trace of 50% LaRC-SI, 50% PEI compression molded specimen
II - 857
DDS
DSC
ENDOTHE
Figure 8: DSC trace of 50% LaRC-SI, 50% PEI injection molded specimen
CONCLUSIONS
Both LaRC-SI and LaRC-IAX polyimides have been successfully injection molded into
ASTM type #1 tensile coupons. Mechanical performance of these injection molded coupons
compares favorably to compression molded specimens of LaRC-SI and LaRC-IAX.
The polyimides have also been successfully blended with PEI and found to retain mechanical
properties while significantly reducing the quantity of the thermoplastic polyimide.
Contamination and non-optimized processing have limited the mechanical performance
observed in all blends. These conditions contributed to premature failures of a number of
specimens during tensile testing. Further, it was found with differential scanning calorimetry
that the compression molding parameters resulted in specimens which were essentially two
phase materials, while injection molded specimens were completely blended.
Future efforts involving LaRC-SI/PEI and LaRC-IAX/PEI blends should involve commercial
PEI powders to eliminate contamination, and optimization of injection molding parameters.
Also, in-depth thermal analysis is planned to evaluate the various blends and changing
process conditions. Finally, testing of non-mechanical performance of such blends is also
necessary since one of the principal strengths of the high temperature thermoplastic
polyimides is chemical resistance.
ACKNOWLEDGMENTS
Appreciation is extended to Mr. Brent Sigmon for his invaluable assistance with injection
molding. This work was supported by IMITEC, Inc. of Schenectady, NY.
REFERENCES
1. D. Progar and T. St. Clair, "LARC-IA: A Flexible Backbone Polyimide," NASA
TM-102586, 1990.
2. T. Hou et al., "Processing and Properties of IM7/LARC-IAX2 Polyimide Composites,"
Journal of Advanced Materials, Vol. 27, No. 2, 1996, pp. 11-18.
3. M. Rommel, L. Konopka and P. Hergenrother, "Process Development and Mechanical
Properties of IM7/LaRC PETI-5 Composites," 28th Int. SAMPE Technical Conference:
Technology Transfer in a Global Community, 1996, pp. 1-13.
II - 858
ENGINEERING COMPOSITE ELECTRE MATERIALS

L.S. Pinchuk1, V.A. Goldade1, I.M. Vertyachikh1 and V.N. Kestelman2.
1
Metal-Polymer Research Institute of Belarussian Academy of Sciences

32a Kirov Str., 246652 Gomel, Republic of Belarus
KVN International, 632 Jamie Circle, King of Prussia, PA 19406, USA.
SUMMARY: Modification of polymer materials by electric fields is now part of advanced

materials science. Polymer electrets are efficient means of rising serviceability of joints with
small clearances between conjugated parts. Electret field makes it possible to regulate
wetting and spreading, sorption and diffusion, as well as capillary permeation of liquids.
Electrets in machines protect metal-polymer joints against corrosion, increase adhesion of
composite components and articles, improve strength, enhance technological effectiveness of
composite formation, improve tribological characteristics of friction joints. Electret plastics
bear mechanical loads in machines and are the source of electric field. Electrets as materials
of a new generation respond to the latest trends in technology for simultaneous increase in
serviceability and reduction in machine size and mass.
KEYWORDS: electrets and materials, polarization, metal-polymer electrets, wetting and

spreading, strength and adhesion, corrosion protection, tribological characteristics
INTRODUCTION
Electric polarization is intrinsic physical state of the polymers being dielectrics in their
majority. Polarization-retaining dielectrics, i.e. electrets or electric analogues of magnets are
characterized by ability to generate a constant electric field.
The electrets are traditionally used in gas filters, microphones, systems of electronic focusing
[1]. Electrets were not used in joints contributing much into machine serviceability, including
bearings, seals, anti-corrosion systems. Nevertheless, experience in different technologies
showed that electrets could improve working capacity of machine joints with a small
clearance between conjugated parts [2,3]. The processes of lubrication, capillary permeation,
spread of liquids, etc. (where surface phenomena are very important) can be controlled using
a relatively weak electret field.
Technologists are usually not paying attention to the fact that electric fields always generate
polarization in dielectrics. Polarization emerges even when electric field is used to solve some
mechanical or physic-chemical problems to match dielectric components at composites
forming.
More attention has been recently paid to ecological, environment protection problems. The
problems arise from the growing industry, which environmental effect is commensurable with
the natural processes. The former leads to changes in water, soil, air, violates equilibrium in
nature. Application of electrets can influence some aspects of the problem. Use of electric
field source in machines is ecologically pure. An experience has been attained in efficient
II - 859
purification of industrial wastes and ejections, rise f combustion effectiveness, reduction of

hostile media leakage by transfer of materials into the electret state.
Electrets perform two functions in machines: 1. the mechanical load carrier and 2. electric
field source. Their application responds to the recent technological trends and permits to rise
serviceability, reduce dimensions and material consumption in machines. These materials are
fairly competitive, their composition and technology are new and patented.
The object of the present paper is to describe new perspectives of electret composite materials
production and efficient use in machine joints governing serviceability of equipment.

Polarization Methods
Methods of polymer composite electric polarization realize mostly the technology of thermoand corona-electrets. In principle, they can be performed at any stage of article production [4].
It has been found out that polymer composites can be polarized by charged fillers [5].
Polyvinylbutyral (PVB) impregnation with corona-discharged oxide powders results in a
maximal characteristic of the electret charge (Fig.1) during thermally stimulated
depolarization (TSD) of the filled films . Space charge of the film and sheet materials can be
formed without sources of electric voltage.
It has been shown earlier that a stable electret state emerges when the polymer layer between
short-circuited electrodes of unlike metals in metal 1-polymer-metal 2 systems undergoes
thermal treatment [2,6-8]. The new type of electret state was called metal-polymer electret
(MPE). It has also been established that during their formation and operation MPE can
display and realize functional potentialities of composites, in particular, strength, corrosion
resistance, as well as frictional, electric and other characteristics [9].
The use of fillers influences electric polarization of polymer composites. PVB samples were
thermally treated under 393 K in contact with short-circuited copper and aluminum
electrodes. Due to conductive or electret fillers graphs (Fig. 2) fall into two groups . The
first group fillers (curves 1-3) features high charge density (up to 10-3 C/cm2). The
conducting filler particle bears, presumably, the function of auxiliary electrodes at
polarization and the amount of charge carriers transmitted from the electrode into polymer
matrix increases too. The second group (4,5) is characterized by the drop of charge with
increasing content of dielectric filler. The latter does not participate in electric conductivity at
sample polarization. The degree polarization current and TSD charge decrease with increase
of filling degree.
The dependencies (Fig. 2) have a maximum at about 1 mass% of the filler content. Small
quantities of foreign impurities contribute to formation of a more perfect permolecular
structure in polymer binders and to electret charge formation. Further increase in filling
degree causes disordering of permolecular structures accompanied by TSD charge reduction.
II - 860
When C > 1% (1-3) the charge grows and then reduces due to reduced quantity of polarized
polymer. The described process is most prominent when the sample is filled with dielectrics
(4,5).
Wetting and Spreading
Liquid wetting and spreading over polymer surface can be regulated by the latter polarizing.
Pentaplast (PPl) films were transferred into electret state by heating between steel electrodes
under 150 oC and electric field of E = 104-106 V/m (thermo-electrets) during 1 h or by corona
discharge (corona-electrets). The technique of measuring liquid spreading over the sample
surface is described in [10].
Fig. 1. TSD currents in SiO2-filled PVB (up to 20 mass%) versus

temperature and filler charge: 1 - corona-discharged, 2 - original
Fig. 2. TSD charge density in PVB samples depending on filler

content: 1 carbon fabric, 3 Al and Cu powders, 4 basalt
fiber, 5 fiber glass
The increase of sample polarization and electret charge reduces the initial velocity of the
liquid spread (Table 1) and the actual contact area between the drop and sample (Fig. 3). The
polarizing charge creates energy barrier in experimental conditions . To overcome this barrier
the motive force of spreading is spent.
II - 861
Fig. 3. Increase in diethylene glycole drop radius r spreading over pentaplast sample
surface as related to time t when the drop was placed on the sample and electret charge
density (C/m2): 1 - 0, 2 - 10-5, 3 - 0.003, 4 - 0.36, 5 - 0.58. r = r r0, where r and r0 are the
radius in moments t and 0.
Table 1. The initial velocity v of DEG spreading over PPl versus sample surface charge
density .
Sample
, C/m2
v, 10-4 mm/sec
Reference
2.8
Thermo-electret
4.96.10-6
2.78.10-3
1.4
0.6
Corona-electret
0.36
0.4
The experimental dependence of edge wetting angle on the polarizing charge density is
described by the Eqn 1:
cos = cos 0 - 1.4104 - 210-9 2
(1)
where and 0 are the liquid equilibrium edge wetting angles for, correspondingly, electret
and nonpolarized samples; o is the sample surface charge density.
Figure 4 shows the scheme for a device to investigate capillary motion of a liquid in contact
gaps between steel sample 7 and polymer electret coating 6 on the substrate 4. The liquid
flows over to the studied joint from a glass tube 1 where its level reduces by a value h.
Capillary pressure of the liquid penetrated into the joint and that of the column of the height
h in tube 1 are equal. Based on the above procedure an Eqn 2 has been derived [3]
W 12
2
(2)
dSpg
co s
where S is tube 1 cross section, - liquid density, g- gravity force acceleration, W is the
work to be spent for the liquid to separate from the coating, 12 - surface tension at the liquid
gas surface.
h =
II - 862
Fig. 4. Measurement diagram (a) of liquid capillary permeation into clearance between solids
and dependence (b) of liquid level h variation in measured capillary on the sample charge
surface density : a) 1 - glass tube, 2- water, 3 -connecting tube, 4 - substrate, 5 -conduit, 6 coating, 7 - sample, 8 - punch in the coating; b) 1- polymethyl methacrylate (PMMA), 2 fluoroplast-3 (F-3).
Nonmonotonous character of curves 1 and 2 in Fig. 4 is the result of superimposed reduction
velocities W and cos when the electret charge grows. On the initial inclining portion of the
curves, the reduction of W dominates. At the same time physical adsorption of liquid
molecules is intensified, which rises the double electric layer on the electret surface. When
the layer is fully formed, W values stabilize. This corresponds to the minimum of h = f()
dependence. The ascending portion of the curves is the result of cos reduction under
constant W.
Curves 1 and 2 correspond to difference in parameters of PMMA and hydrophobic F-3
wetting with water.
Sorption and Diffusion
The theory of diffusion based on the mechanism of diffusing particles jumping over the solid
vacancies [11] suggests the following Eqn 3:
D=
a2
exp( W / kT )
0
(3)
where D is diffusion coefficient, a is the solid lattice constant, about 10-13 s is the particle
oscillation period, W - energy of diffusion activation, k - gas constant, T - absolute
temperature. If diffusion proceeds in a solid dielectric possessing a polarized charge, then
W = U + E + Ee, where U is the energy of vacancy formation, E is energy barrier height, Ee energy acquired by a particle in the electret field. Diffusion proceeds in the electret faster or
slower as compared to the polarized dielectric depending on Ee sign.
The polarized charge influences first of all the diffusing substance sorption by the electret.
This influence behavior is conditioned by formation of the double electric layer in vicinity of
II - 863
the electret surface. Sorption of organic solvent vapors was studied using PVB films
thermally treated in contact with shorted copper and aluminum electrodes. The sample
polarized charge density determined by the TSD method was about 10-5 C/cm2. Both
polarized and reference (thermally treated in contact with aluminum electrodes) samples were
endured in diethylene glycole (DEG) and benzene vapors under T1 = 303 K and = 323 K (T2
is about TC of PVB). The experimental technique is described in [12].
Fig. 5. Mass (M) of PVB samples versus endurance time (t) in DEG vapors. Temperature, K:
1, 2 -323, 3, 4 - 303. Samples: 1, 3 - reference, 2 and 4 - electret.
In Fig. 5 it is seen that DEG sorption by the electret under T2 (curve 2) is much less than in
[1]. When T1 < Tc the difference in sorption is not large (3 and 4). The similar results were
obtained for the samples treated in benzene vapors (Table 2).
Table 2. Increase of PVB sample mass (M) depending on time (t) of endurance in benzene
vapors under 323 K.
t, 103 s
M, 10-2 g for the samples

reference
electret
4.5
0.55
6.4
1.10
7.3
10.7
1.30
0.01
1.50
0.05
13.9
1.66
0.14
15.2
1.73
0.24
When T < Tc, the reduction in segmental mobility of macromolecules bounds sorption due to
lower velocity of conformation transformations. This probably levels the difference in
sorption of solvent vapors by electret and nonelectret films (curves 3 and 4 in Fig. 5). Upon
the formation of the double electric layer from oriented dipoles near the electret surface, an
equilibrium concentration of the solvent is established in the sample. The film permolecular
structure in the process of electric polarization spurs sorption reduction in both polar (DEG)
and nonpolar (molecular dipole moment equals zero) solvents. It appears that benzene
II - 864
nonpolar molecules acquire the dipole moment in the electret field and interact with the
electret following the described mechanism. Difference in DEG and benzene diffusion is
attributed to DEG stronger solution capacity as compared to PVB.
Adhesion
The effect of polarized charge on polymer adhesion to metals was first studied in the metalpolymer-metal systems [13]. Strength of splice adhesion was determined by lamination
method . Strength increased 1.5-2 times when the adhesive joint formation was accompanied
with the polymer transformation into MPE. Adhesion increase is much effected by activated
in electric field diffusion of the metal-polymer contact reaction products into the polymer.
A model of a composite material is suggested. The composite incorporates a thermoplastic
polymer binder and dispersed particles of metal oxides which differ by the value of the
standard electrochemical potentials [14]. Fuming-oxides (FO) were used as the filler. They
are a by-product of lead-zinc production and contain Zn, Pb, Mg, Cu and other oxides. During
material formation FO reduction to metals is proceeding. The reduction of the sample ohmic
resistance at heating, as well as the results of X-ray diffraction analysis are the proof of it. It
was observed that with increasing time of treatment the intensity of peaks characterizing the
oxide crystalline structure reduced, while the intensity of peaks corresponding to pure metal
increased.
When a certain concentration of the conducting filler is reached, the formation of conducting
bridges is probable in the composite and the contact between particles is unnecessary for the
polarizing current to run. Microcircuits metal 1-polymer-metal 2 are formed between both the
filler particles in the polymer matrix bulk and between a separate particle and metal electrode
contacting the sample. Thus, electrochemical interaction of the composite components is
realized in the process of thermal treatment which increases adhesion (Fig. 6).
Fig. 6. Strength of adhesive joints (a) and protective capability (corrosion rate, b) from FO
containing polymer composites with Al foil versus FO content and filler type: 1 - PE, 2 - PVB,
3 - PPl.
Rearrangement in the polymer permolecular structure accompanying polarization causes, as a
rule, improvement of their strength and impairment of deformation characteristics. So,
thermally electrified polycarbonate and PTFE samples show an increased by 15-60% breaking
tensile stress and several times relaxation time of mechanical stress [4].
II - 865
Strengthening of polymer composites by transferring their components into electret state has
been verified experimentally [5]. Breaking tensile stress of composites (PE, PA, PVC, PPl,
PTFE matrices;, reinforcing elements - basalt, phenylone, glass, Dacron fibers) increases 1.62.5 times due to filler and matrix electric polarization. Adhesion of polymer binder to filler
exerts a substantial effect on composite strength. Force of fiber glass peeling off from PVCbased fiber glass plastic increases at binder polarization from 3.4 to 4.4. kN/m. For PPl it
increases from 1.7 to 2.6. kN/m .
Protection against Corrosion
One of crucial factors for the polymer-based rustproof systems is the barrier action of polymer
elements insulating metal from environment. It is believed that, when a structural polymer
barrier is disturbed, it does not influence metal corrosion in electrolytes. Nevertheless, the
polymer electret field effect on both barrier characteristics of polymers and kinetics of
electrochemical processes in metal-polymer systems should be taken into account. The
polarized charge effects the kinetics of liquid spreading over polymer electrets and sorption of
liquid media.
Diffusion of liquids into MPE coating g on metals has been studied. If liquid concentration
C0 on the coating surface is supposed to be constant during diffusion, then the amount of
liquid sorbed by MPE to the moment t is [15]
Q=
2 C0
Dt exp( )
(4)
where D diffusion coefficient of nonpolarized polymer; -MPE surface charge density

determined by TSD. The parameter 3104 is determined experimentally for each polymerliquid combination. The equation describes satisfactorily diffusion into MPE with > 5.10-7
C/cm2. Probable reduction of coating diffusion permeability using polarization is limited by
the ultimate charge density of MPE 10-5 C/cm2 due to electrodes blocked by products of
metal-polymer reaction [16].
The initial corrosion of protected by polymer coating metal is penetration of corrosive
medium along the coating-metal interface leading to the coating peeling off. Peeling is
accompanied by the electrode metal potential bias to the negative domain. This evidences
that under the coating there occurs either retardation of cathodic oxidation or acceleration of
metal ionization. Both processes can be controlled using electret coatings.
Table 3 demonstrates test results of polymer coatings on steel with a through punch in the
protecting film reaching the substrate. Coatings of group A were formed following the
classical vibro-vortex method. Melting of group B coatings was accomplished in electric field
of E = 20 kV/cm strength by connecting upper electrode with the positive pole of a highvoltage source and the substrate with the ground. When melting group C coatings, the
polymer layer was covered by a copper foil closed in the steel substrate. After cooling the foil
was peeled off.
II - 866
Table 3. The coating peeling area S and change of the substrate electrode potential /\U
depending on endurance time t in a strong NaCl solution.
Coating
Group
Coating material
S (cm2) during t (days)
U (V) 5
days
1 day
PPl
PVB
0.25
3.03
0.74
6.24
2.03
10.32
2.76
3.97
0.24
PPl
PVB
0
0.01
0.01
0.02
0.02
0.04
0.02
0.07
0.04
0.08
0
0.02
PPl
PVB
0
0.01
0.01
0.02
0.02
0.05
0.02
0.07
0.03
0.08
0
0.02
It is obvious that Group B thermo-electrets which contact the substrate by the negatively
charged surface peel off slower and their substrate unrefining is less as compared to
nonelectret ones (group A). Similar results showed MPE-based coatings (C) formed without
outer electric source.
Triboengineering Characteristics
Among phenomena accompanying polymer friction, those of electrical nature are the least
studied, though their influence is not of minor importance.
Crucial factors governing friction and wear of polymers being either a polarized body or
electric field source are the following. Electrization potential U of the polymer element in a
metal-polymer friction pair succeeds to saturate rather quickly (in a few seconds). Change of
U sign during initial friction can be attributed to loss of moisture by the element surface
layers. Air humidity and the presence of a liquid film on the friction surface exerts a
considerable effect on U. Polymer in the field of electrization charge transfers under frictional
heating into tribological electret state [7]. It determines the kinetics and intensity of physicchemical processes in the friction zone. Most important experimentally recorded parameters
of triboelectret state are the efficient surface density of charge o and TSD current spectrum.
The latter helps to estimate the correlation between homo- and heterocharge in the
triboelectret, as well as activation energy of the charge formation and declining, time of
relaxation, etc.
MPE wear was examined by removing from the friction surface the layers which contacted
during polarization Cu and Al electrodes. The experimental technique is described in [18].
MPE charge is concentrated mainly in the surface layer (20-60 m). It was recorded by TSD
that o near the copper electrode is higher than near the aluminum one. Wear of the friction
surface contacted Al at polarization is lower than that contacted Cu (Fig. 7). Obviously MPE
wear proceeds more intensive y with o growth. Such a regularity is violated only with thin
(15-20 m) surface layers displaying elevated concentration of chemical structural defects.
This is also typical, though to a less degree, of nonelectret samples.
II - 867
Fig. 7: The dependence of MPE wear rate on the thickness of the layer removed from the
sample surface contacting Al (1) and Cu (2) at polarization. 3 - reference sample. Sample
materials: a - PVB, b - PPl, c - PA.
Variation of MPE tribological characteristics are due to space charge formation accompanied
by structural rearrangement of the near-electrode layer. Rise in degree of nonpolarized PPl
samples crystallinity constitutes 4% from the copper electrode side and 6% from the
aluminum one. Upon removal of 80-100 m thick layer the polarized and nonpolarized
samples do not differ in crystallinity degree. It may be so that space charge generated in the
near-electrode regions during crystallization hamper macromolecule packing. Therefore, the
larger charge value localized near copper electrode corresponds to a smaller degree of
crystallinity.
Novel triboengineering materials and friction joints have been developed [19-20]. They
incorporate MPE as electric field generator to improve friction and wear parameters.
CONCLUSIONS
Polymer electrets are used to regulate technological processes of composite material
formation. They involve sorption, diffusion, wetting, spreading, saturation and etc. Electret
composites contribute to protecting machine parts against corrosion, raise of components
adhesion, material hardening and improvement of friction joint tribological characteristics.
Electret composites are the new materials effecting the engineering developments.
ACKNOWLEDGEMENTS
The authors are grateful to Dr. Yu.I. Voronezhtsev for experiments with MPE and discussion
of results.
II - 868
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
Electrets, Ed. by Sessler G.M., Springer-Verlag, Berlin, 1987.

Goldade V.A. and Pinchuk L.S., Electret Plastics: Physics and Materials Science,
Nauka i Tekhnika, Minsk, 1987.
Pinchuk L.S., Hermetology, Nauka i Tekhnika, Minsk, 1992.
Voronezhtsev Yu. I., Goldade V.A. , Pinchuk L.S.and Snezhkov V.V. , Electric and
Magnetic Fields in Technology of Polymer Composites, Nauka i Tekhnika, Minsk,
1990.
Vertyachikh I.M., The Development of Engineering Polymer Composite Materials with
Electret Components, Ph.D. Thesis, Gomel, 1990.
Goldade V.A., Pinchuk L.S.and Voronezhtsev Yu.I., Electric Polarization of Polymers
in Metal-Dielectric- Metal Systems, Proc. ISE-6 Conf., IEEE, New York,
1988,
pp.419-423.
GoldadeV.A., Voronezhtsev Yu.I. and Pinchuk L.S., On Mechanism of Polymer
Polarization in Metal 1-Polymer- Metal 2 Systems, Proc. Intern. Conf. Plastko- 90,
Ostrava, 1990, pp. 76-80.
Goldade V.A.and Pinchuk L.S., New technology of Electrically Polarized Polymers
Creation during Heat Treatment in Contact with Different Metals, Proc. Polymer
Process. Soc., Stuttgart,1995, pp. 6-9.
Goldade V.A. and Pinchuk L.S., New Spheres of Metal- Polymer Electrets Technical
Application, Proc. ISE-9, IEEE, Shanghai, 1996, pp. 1079-1080.
Plevachuk V.G., Vertyachikh I.M., Goldade V.A.and Pinchuk L.S., Effect of the
Charge of a Polymer Electret on the Spreading of Liquids, Polymer Science, A, Vol.
37, No. 10, pp. 1071-1074.
FrenkelYa. I., Kinetic Theory of Liquids, USSR Acad. of Sci. , Moscow, 1945.
Vertyachikh I.M., Goldade V.A., Neverov A.S.and Pinchuk L.S., The Effect of
Polymer Electret Electric Field on Sorption of Organic Solvent Vapors, High-Mol.
Comp., 1982, Vol. 24B, No. 9, pp. 683-687.
Belyi V.A., Goldade V.A., Neverov A.S.and Pinchuk L.S., Polymer Strength in
Splices of Unlike Metals, Polymer Mechanics, 1977, No. 4,pp. 740-742.
Goldade V.A., On Adhesive Strength and Protective Capacity of Polymer Composite
Coatings, Mechanics of Composite Materials, 1981, No. 5, pp. 842-845.
Voronezhtsev Yu.I., Effect of Polymer Coating Electric Polarization on its Diffusion
Permeability, Trans. of BSSR Acad. of Sci., Phys. Techn. Sci., 1986, No.4, pp. 42-44.
Goldade V.A., Voronezhtsev Yu. I. and Pinchuk L.S., On Some Peculiarities of
Polymer Electric Polarization in Metal 1-Polymer-Metal 2 Systems, High-Mol. Comp.,
1988, Vol.30B, No. 6, pp. 427-431.
Klimovich A.F.and Guzenkov S.I., Electric Phenomena at Polymer Friction. III,
Regularities of Electret State Formation, Soviet Friction and Wear, 1989, Vol. 10, No.
5, pp. 779-785.
Voronezhtsev Yu.I., Goldade V.A., Pinchuk L.S. and RechitsG.V., Influence of
Charge Distribution on Friction Characteristics of Polymer Electret, Soviet J. of
Friction and Wear, 1984, Vol. 5, No. 11, pp. 110- 113.
USSR Patent No. 1165588, B29C, Device to Manufacture Articles from Polymer
Bulk Materials, 1985.
USSR Patent No. 1055913, F16C, Sliding-Contact Bearing, 1983.
II - 869
MULTI-AXIAL WARP KNITTED LAYERS A TEXTILE FOR REINFORCING CONCRETE

Dr.-Ing. Gerd Franzke 1, Prof. Dr.-Ing. habil. Peter Offermann 1, Dipl.-Ing. Thomas
Bischoff 2, Prof. Dr.-Ing. Burkhard Wulfhorst 2
1
Institut fr Textil- und Bekleidungstechnik der TU Dresden,

Mommsenstrae 13, 01069 Dresden, Germany.
2
Institut fr Textiltechnik der RWTH Aachen,
Eilfschornsteinstrae 18, 52062 Aachen,Germany.
SUMMARY: The aims described before will be achieved by presenting the historical
propagation in civil engineering from asbestos to glass fibres respectively alcaline resistant
glass fibres. A short summary of existing technologies for processing short fibres with
concrtee matrices will show today`s state of the art. Both aspects will be compared with the
historical development of fibre reinforced plastics technologies. Derived from this comparison
the necessary technology transfer from plastics to concretes will be obvious. Following this
idea it will be presented how a new composite material has been born by reinforcing concrete
with textiles. The results of a project including production of multiaxial layer fabrics, matrix
optimization, production oft test specimen and determination of mechanical properties will
prove the technical and economical qualification of the new composite material in comparison
with short fibre reinforcement.
KEYWORDS: textile reinforced concrete, alkali-resistant glass fibre, fibre tests, warp knitted
multi-axial layer fabric WIMAG, stitch-bonded variable layer fabric NVG,
INTRODUCTION
The reinforcement of mineral matrix systems using fibres has long been a state-of-the-art
technique. The best known and quantitively most widespread are asbestos fibres and building
components reinforced with these. The need to find a substitute for asbestos fibres worldwide
resulted in a shift of interest to, among other substances, glass fibres. In order to use these in a
concrete matrix it was essential that the fibres be resistant to the alkaline environment if they
were to retain their very good mechanical qualities for a sufficient duration. The development, 25
years ago, of alkali-resistant glass fibres enabled glass fibres in the form of short strands tobe
used for reinforcing concrete matrices, a technique which has become firmly established in the
construction sector and is applied in a wide variety of ways. Some improvement is needed if
optimal advantage is to be taken of these fibres, particularly in view of the short length of the
fibres and the fact that their orientation is not always appropriate for achieving the bearing
capacity required. The effectiveness of the fibres is to be improved by using glass filament yarns
and orienting them in accordance with the function of the particular building component in
II - 870
question. A technological variant is the production of textile fabrics that meet these requirements.
At present knitting technology with its variable stitch-bonding techniques is ideal for the
manufacture of such structures. Initial tests have been carried out at the Institute for Textiles and
Clothing Technology, TU Dresden, and at the Institute for TextileTechnology, RWTH Aachen,
as a part of AiF project no. 9272/B [1]. The results of these tests form the basis for this talk.
FIBRE TESTS
Reinforcment materials in a concrete building component are subject to media influences right
from the start of the manufacturing process until the end of their service life. These include the
alkaline influence from the concrete matrix and the acidic influence of rainwater seepage at
defective points. In Germany, very high demands have to be met proving the durability of new
building materials before their introduction in the construction industry. Tests concerning the
influence of the above-mentioned media on the qualities of the reinforcement materials are
therefore of particulare interest.
This research project made use of the special glass fibres NEG AR H-340V, 650 tex produced by
Nippon Electric Glass. These fibres have a raised zirconium content (ZrO2), providing good
resistance in fine concrete. These fibres are referred to as alkali-resistant (AR) glass fibres. The
yarns produced by Nippon Electric Glass are the only such fibres suited treatment as textiles and
they were selected for these tests for this reason [1,2].
The experiments testing resistance to artificial rainwater (pH 6) and to sodium hydroxide (pH
12,1), which corresponds to the pH value of a special concrete matrix, were conducted according
to the established norms. The fibre material was taken from a spool and laid out in test groups in
developing trays which were subsequently filled with the solutions. The trays were then covered
to prevent evaporation, thus ensuring constant pH values.
The test materials were then stored in a climatic chamber. After 24 hours, 7 days and 28 days
they were removed, air-dried for three hours and then subjected to bending pressure. The force:
bending ratio was determined in accordance with bending pressure test DIN 53834, part 1.
Figure 1 shows the specific tensile strength of the AR glass fibre yarns in relation to duration of
storage and the medium applied.
The results show that with the specimens stored in normal climate conditions only minor
changes in durability are recorded, so that comparisons can be made between these and the
specimens exposed to the special conditions as described above. After drying, the samples stored
in the solutions have a slightly adhesive structure. The fibre lengths stored in rainwater, in
particular, lead to higher degrees of durability as a result of these adhesive qualities which have
the effect of bonding fibre and matrix together.
II - 871
30
25
20
15
10
5
0
Fig. 1: Specific tensile strength of AR glass filament yarns under the influence of media and
duration of exposure
The fibre lengths stored in sodium hydroxide solution show only minor changes in durability,
thus confirming their good resistance to alkaline media.
PRODUCTION OF MULTI-AXIAL LAYER FABRICS FOR REINFORCEMENT

CONCRETE
OF
Two multi-axial types of textiles will be manufactured for the initial basic tests on the suitability
of long-fibre from AR-glass reinforcements in concrete matrices. The two manufacturing
techniques which will be compared are MALIMO stitch bonding, with the stitch-bonded variable
layer fabric (NVG) manufactured applying this technique, and the LIBA technique, together with
its product, the warp knitted multi-axial layer fabric (WIMAG).
Production Of Warp Knitted-Multi-Axial Layer Fabrics (Wimag)
The warp knitted multi-axial layer fabrics are used succesfully in the field of plastic composites
technology. This is due to the cost-effective production and excellent mechanical properties of
WIMAG. Depending on the size of the machine, the textile (cf. Fig. 2) may comprise up to 8
layers of fibres and an additional 2 covering layers of fibre mats, which are fixed in place by
warp knitted mesh threads.
Fig. 2 also illustrates the warp knitting machine (LIBA system) with its four weft insertion
systems, as has been used at the Department of Textile technology of the Technical University of
Aachen (ITA) for manufacturing WIMAG.
II - 872
Fig. 2: Warp Knitted Multi-Axial Layer Fabrics (LIBA-Technique)

Here the reinforcing threads (2) are drawn off the warping creel (1) and inserted by the weft
inserting systems in the required directions. The weft inserting systems oscillate between the
needle transport chains (3), which fix the deposited fibres to the machine during transportation.
The weft inserting systems can be mounted in a fixed position (90) or variably (30 to 60) or
(-30 to -60) to the direction of production. 0 threads can be fed directly from warp beams (6).
It is also possible to supply non-woven fabrics or fibre mats as covering layers via rollers (5).
The warp knitted multi-axial layer fabric, designed in this manner, is fixed in place in the knitting
unit (8). The warp knitting threads are removed from the warp beams (7). The finished WIMAG
(9) is wound on at the end onto a cloth beam (10).
PRODUCTION OF STITCH-BONDED VARIABLE LAYER FABRICS (NVG)

During the manufacture of stitch-bonded materials (cf. Fig. 3), a parallel weft yarn sheet is first
of all laid with the aid of a weft guiding carriage. The continuous advance feed of the transport
chain during laying causes this yarn sheet to be inserted at an angle of about 88, and not
precisely at 90.
Using two variable thread guiding systems (V1 and V2), parallel yarn sheets can be inserted at
an angle of between 0 and about 80, and in addition, the distance between the threads can be
altered. This allows the reinforcing textile to be reinforced locally in certain zones, which lie
within the subsequent building component, for example in the zones where force is applied.
II - 873
The weft thread sheet is fed to the knitting elements, together with the other thread layers. The
reinforcing threads are fixed in place with the aid of a sewing yarn system which creates knitted
meshes. Thus, this procedure combines elements of sewing (passing through and combining
thread layers) and of knitting (creating knitted meshes from a large number of individual threads
at the same time).
Fig. 3: Malimo stitch-bonding technique

The Malimo stitch-bonding technique is applied at the Institute of Textile and clothing
Technology at the Technical University of Dresden (ITB).
TESTS ON THE TEXTILE PROPERTIES OF THE REINFORCEMENT

The physical tests show that the processing of the reinforcement fibres into a net-shaped
structure has been done satisfactorily, although the meshes sometimes have rather irregular
thread patterns. With 90 oriented rovings the WIMAG fabrics were found to have somewhat
better properties than the NVG fabrics (cf. Fig. 4). With 45 orientation hardly any difference
was found in the test values for the two types of fabric [2].
II - 874
Fig. 4: Comparision of maximum tensile strength of the reinforcement textiles

Since working with these substances as textiles is not without ist difficulties, the fabrics achieve
a yarn strength of only 43% (as cited by the Manufacturer). Causes of this include the slightly
wavy formation of yarn resulting from the open structure and from defects in the material used.
THE COMPOSITE ELEMENTS

Three different net-shaped structures (S1, S2, S3) with 90 and +/- 45 oriented reinforcing
fibres, made from AR fibre glass, is manufactured from the two types of textiles described
above; the different fabric types can be compared. Net-shaped structures have been selected in
order to ensure that the textile is impregnated by the relatively high viscous concrete.
A special-purpose matrix material has been developed on the basis of experience derived
using types of short-staple reinforced concrete; this material attacks fibre glass less by
forming a medium which is less alkaline around the fibre glass. These cement matrices are
based on Portland cement with highly efficient alkali binding components. This type of
cement was therefore selected for trials with the following additives (weights), on the
recommendation of industry [3]:
Cement
Sand
Water
Liquefier
LP 1-87 Addiment
1
1
0,42 to 0,48
0,22 to 0,04
0,003
part
part
parts
parts
parts
II - 875
THE PRODUCTION OF THE TEST SPECIMENS

The test pieces are manufactured with aid of a positioning unit which allows the reinforcing
textiles to be guided into the mould tool in a defined manner (cf. Fig. 5).
Fig. 5: Production of Test Specimens

The divible mould tools are fixed in place above the postioning rods. Once the bottom halves
of the mould tools have been positioned and filled, the reinforcing composite is fixed in place
with the aid of clamping elements. The tension applied merely serves to ensure that the textile
is laid evenly; in other words, no elastic deformation of the reinforcing textile is generated.
Once the composite has been fixed in place, the top and bottom of the mould tool are screwed
together and filled with concrete. The complete mould tool can now be raised from the
positioning unit and placed in a climatic chamber to be stored for 28 days.
RESULTS OF THE MECHANICAL TESTS

The first series of tests, in which the two more open glass structures were tested, proves that
the concrete with 0.42 parts by weight of water and 0,035 parts by weight of cement liquefier
can withstand the highest maximum bending pressure. The specimens were tested after only
10 days curing time. At this stage the composite matrial still does not possess its ultimate
properties. Nevertheless, the calculated bending tensions should only be viewed initially for
the purpose of comparison in order to determine the ideal make-up of the concrete. In this
series of trials, the concrete samples with the WIMAG glass reinforcing of greater density
possess a maximum bending strength of 9.1 N/mm, as opposed to 5.08 N/mm in the case of
non-reinforced concrete.
The optimised aggregate parts were used in the actual final of trials. The test pieces, which
were produced applying the same process, were now stored for 28 days. The results of 4-point
bending test are illustrated in Fig. 6.
II - 876
25
20
15
10
5
0
Fig. 6: Results of the 4-Point-Bending Test

It is clear that the modulus of rupture (MOR) of the textile reinforced concrete secimens (21.0
N/mm) is up to four times higher than of non-reinforced concrete (5.1 N/mm). This
maximum value was achieved using the WIMAG AR glass structures of greatest density
(WGS3). The more open (S2 or S1) or NVG structures lie somewhat below the maximum
values.
THE PRODUCTION OF A DEMONSTRATOR COMPONENT

For demonstration purposes, a girder used a window lintel, geometrically measuring 700 mm
x 140 mm x 140 mm. In order to minimise its weight, the building component is built as a
hollow concrete profile with a styrodur core measuring 700 mm x 100 mm x 100 mm. The
girders were made by the Institute of Supporting Frameworks and Building Materials at the
Technical University of Dresden, using the optimised concrete composition and with the aid
of double moulds (cf. Fig. 7).
II - 877
Fig. 7: Production of a Window Linter

The reinforcing structures are first impregnated separately with the concrete matrix in a tray,
before then being inserted in the mould and evenly aligned with a 2 kg weight at either end.
The matrix material is fed into the bottom of the mould, and the styrodur core is then inserted;
it is centred using U-shaped spacer pieces at either end of the mould. The remaining hollow
spaces are then filled, and air pockets are removed by compressing the matrix. After curing,
the girders are removed from the mould, impregnated in the water bath and stored for 28 days
at climatic conditions of 20 C and 55 % relative humidity.
Following the storage of the window lintels, the sides which were facing up during
manufacture are ground. The girders are then tested in 4-point bending tests, conforming to
DIN 51300 and DIN 51302. The breaking load, deformation and extension of the girders are
ascertained.
The suitability of the different types of composites for reinforcing concrete building
components can be judged taking the modulus of ruptue (MOR) as an example (cf. Fig. 8). It
clearly has a greater bearing capacity than concrete which is either non-reinforced or else only
reinforced using chopped strands. The extension and deformation values of the reinforced
windows lintels are also calculated as being clearly higher than the comparable values for the
girders which were either non-reinforced or else only short-staple reinforced. The final
appraisal of the results of the trails will have to be carried out following completion of the
final trials [4].
II - 878
12
10
8
6
4
2
0
Fig. 8: Results of the Mechanical Testing of the Demonstrator Components
CONCLUSION AND PROGNOSIS

The results of this research project show that sheets made up of AR glass fibre threads
applied as reinforcements for concrete components can lead to better use of materials and
hence reduce costs. A proportion of glass amounting to 1% by volume applied in this way
produces the same durability as a 3-5% proportion applied as short-staple glass fibres. By
orienting the yarns in accordance with the demand to be placed on the building component in
question, a higher degree of efficiency is achieved in the reinforcement material. The results
also show however, that there is still more room for improvement in the properties of the
fabric sheetsas regards their ability to be treated as textiles. Improvements in this sphere
would lead to further enhancement of the properties of building components with respect to
durability and the minimisation of weight.
More highly developed alkali-resistant glass filament yarns are currently being tested by
Nippon Electric Glass with regard to their textile properties and as reinforcement materials for
very thin concrete elements. In the future it is intended to establish rules on the dimensions of
reinforcing sheets made of alkali-resistant glass fibre threads.
In order to solve the problems regarding the textile properties of these products, such as
improvement of their processing potential, enhancement of fibre/matrix adhesion, and
development of reinforcement structures appropriate to loading and processing demands, it is
essentials that there be intensive cooperation between manufacturers of fibres, textile machine
II - 879
producers, the textile industry, and research institutions, always bearing in mind, of course,
the needs of the end consumer, the construction industry.
We should like to express our thanks to the Gesamttextil Research Committee for ist
support of the work presented here (AiF-No. 9272 B) and other on-going research. This
support is in the form of a research grant awarded by the Working Group of Industrial
Research Associations from funds provided by the Federal Ministry of Trade and Commerce.
REFERENCES
1.
Wulfhorst, B., Bischoff, Th., Offermann, P., Franzke, G.: Verwirkte

Verstrkungsgelege fr das textile Bauen; Abschlubericht zum AiFForschungsvorhaben Nr. 9272, RWTH Aachen und TU Dresden, 1996
2.
Franzke, G., Bischoff, Th.: Textilarmierung fr mineralische Baustoffe; Vortrag zur

3. Dresdner Textiltagung 1996, Dresden
3.
Weiser, W.: Interal Information of the Fachvereinigung Faserbeton e. V. Dsseldorf,

Germany, Juli 1994
4.
Bischoff, Th., Franzke, G.: Multi-Axial Warp Knitted Layers a Textile for Reinforcing
Concrete Strasbourg, France, International Congress GRC 95, 9-11 October 1995
II - 880
PROCESSING AND ELECTRICAL PROPERTIES OF

PTCR COMPOSITE MATERIALS
Z. X. Xiong and K. Z. Baba-Kishi
Department of Applied Physics, The Hong Kong Polytechnic University, Hong Kong
SUMMARY: This paper reports the processing and electrical properties of a new type of
PTCR composite materials, which has been made into electric heating cable and plate used in
industrial and domestic area. The additives of carbon black and semiconductive ceramic
powders were mixed with the matrix of polyethylene powder or epoxy resin. The SEM
micrographs of the composites displayed that CB, SBT and silver powders were randomly
distributed in the PE or ER matrix. A conducting model of the composite is proposed and its
equivalent parallel-series electric circuit is schematically shown, to explain the mechanism of
the PTCR effect of the composite.
KEYWORDS: processing, electrical property, PTCR composite and SEM
INTRODUCTION
Polymer matrix composite materials not only exhibit PTCR (Positive Temperature Coefficient
of Resistivity) effect but also behave perfect flexibility [1,2]. A new type of PTCR composite
materials was reported here, which has been made into electric heating cable and plate used in
industrial and domestic areas.
The PTCR effect of this kind of composite appears when a volume fraction of conductive
additive was so large that the percolation threshold of conduction can take place [3,4], usually
about 20 - 50 weight percentage of carbon-black is enough. The resistivity of the composite at
room temperature is mostly equal to 102 to 105 m.
EXPERIMENTAL
As conductive and semiconductive additives, commercial carbon black (CB) powder, silver
(S) powder or/and (Sr0.2Ba0.8)(Ti0.999Nb0.001)O3 ceramic (SBT) powder were mixed into
polyethylene (PE) powder or epoxy resin (ER) by roll-milling at 150 ~ 180 C for 30 minutes.
The samples were then pressed by a hydropress at 100 ~ 130 C for 5 minutes. Ga-In solution
electrodes were applied. The samples were irradiated in an electron accelerator, 2 MeV for
energy, with different irradiation dose. The resistivity of the samples was measured by a
multimeter as a function of temperature with a heating speed of 1 C/minute. The
microstructure of the composite materials was observed with a scanning electron microscope
(SEM).
II - 881

Cooled down to room temperature at a slow speed, sample 1B exhibited lower resistivity at
room temperature (Rr) and behaved stronger PTCR effect than that of sample 1G that was
cooled at a high speed (Fig. 1). However, there was little difference between sample 1B1 and
1G1 in which a special reagent was added. Partially substituting the CB powder with SBT
powders uncoated and coated with a thin nickel layer, the authors produced samples 1B2 and
1G2. The later showed lower Rr and better PTCR effect; moreover, it took short time to reach
a thermostatic temperature.
10
1G
1B
1G1
1B1
1G2
1B2
Log p
8
7
6
5
4
3
0
50
100
150
T ('C)
Fig. 1 Temperature dependence of resistivity of the composites
8
2A1
7
2A2
2A3
2A4
Log p
2A5
5
4
3
2
1
0
20
40
60
80
T ('C)
100
120
140
Fig. 2 Effect of additive content on PTCR curve of the composites
II - 882
The PTCR effect of the samples was also dependent on the cross-linking degree or
crystallinity of the PE matrix (Fig. 2). Sample 2A1 in which about 60 % PE matrix was crosslinked revealed stronger PTCR effect than sample 2A2 in which the PE matrix was free of
cross-linking. Though influencing Rr, the CB content did not evidently change the PTCR
effect of samples 2A3 and 2A4, in which 30 wt % and 50 wt % CB powders has been mixed
with the ER matrix, respectively. Nevertheless, the PTCR effect of sample 2A5, which
consisted of 30 wt % CB powder and 10 wt % silver powder in the ER matrix, was obviously
weakened by the addition of the silver powder; meanwhile, the sample 2A5 behaved
remarkably low resistivity within a wide temperature range. The PTCR effect of the material
was influenced by the dose of electron irradiation when its degree of cross-linking was
relatively low, as sample A1 (Fig. 3) with a, b, c and d representing irradiation dose of 2.0,
1.5,1.0 and 0 x 10 5 Gy, respectively.
11
10
9
Log p
8
7
A1a
A1b
A1c
A1d
A2a
A2b
A2c
A2d
6
5
4
3
2
0
20
40
60
80
100
120
140
T ('C)
Fig. 3 Effect of electron irradiation dose on PTCR curve of the composites

The micrographs of the composites displayed that CB, SBT and silver powders were
randomly distributed in the PE or ER matrix as shown in Fig. 4.
Fig. 4 SEM images of the PTCR composites
II - 883
CONDUCTING MODEL AND EQUIVALENT CIRCUIT

Based on the experimental results, a conducting model of the sample is proposed (Fig. 5)
and a parallel-series equivalent electric circuit is schematically shown in Fig. 6. Since the
Fig.5 A conducting model of the sample
Fig.6 A equivalent electric circuit of the sample
composite contained three phases: insulating polymer, semiconductive ceramics and

conducting carbon-black, the conduction band of the composite with 0-3 connectivity are thus
characterized by three parameters: the barrier height of the insulating polymer, Ei, the barrier
height of the semiconductive ceramics, Es, and the conduction level of the carbon-black, Ec.
Therefore, the total resistance of the composite, Rt, results from three kinds of resistance:
resistance of the polymer, Ri, resistance of the semiconductive ceramics, Rs, and, resistance
of the carbon-black, Rc. The heights and widths of the barriers vary, in fact, at different
micro-regions in the composite. The resistivity should be calculated by integration within the
whole material to obtain more accurate results.
The temperature dependence of the resistivity can be roughly divided into three segments
corresponding to three temperature regions. At room temperature the resistivity of the
composite was usually small, which might be due to the relatively close distance between the
conductive powders of the carbon-black or the semiconductive ceramics. The powders
distributed randomly in the matrix of the polymer act as conducting bridges or network in the
insulating polymer matrix. Moreover, the tunnel effect might also take place within a thin
enough layer of the polymer at relatively low temperature when an electric field is applied. As
the temperature is increased, however, the polymer matrix is expanded faster than that of the
additives. The numbers of the conducting paths of the network are decreased, and the
resistivity is thus increased sharply. So the PTCR effect appears within this temperature
range. At a higher temperature, usually above the glass transition point of the polymer, the
polymer matrix begins to melt. The concentration of conduction carriers is increased quickly
at high temperature, and thus, the resistivity is decreased.
II - 884
CONCLUSION
A new type of PTCR composites consisting of conductive carbon-black and semiconductive
ceramic powders in the matrix of PE or epoxy resin has been developed, which has been made
into electric heating cable and plate used in industrial and domestic areas. The amounts of
constitutes and the processing parameters have effects on the electric properties of the
composite materials.
REFERENCES
1.
J. Meyer, Polym. Eng. Sci., 14 (1974) 706.
2.
D. S. WcLachlan, et al., J. Am. Ceram. Soc., 73 (1990) 2187.
3.
M. Sahimi, Application of Percolation Theory, Tayler & Francis Plishers, London,

1994.
4.
D. Stauffer and A. Aharony, Introduction to Percolation Theory, Tayler & Francis

Publishers, London, 1991.
II - 885
SUBJECT INDEX
3D woven structure
658
A12O3
769
acrylonitrile / butadiene copolymers
254
adhesion
859
alkali-resistant glass fibre
870
aluminum-fiber
301
anisotropic damage
682
anisotropic properties
785
aramid fiber reinforced epoxy matrix composite 503
arc-melting
603
artificial inserts
438
artificial viscosity
397
AS-4/PEEK
177
asymptotic expansion method
453
B4C-SiC composite
B4C-TiB2 composite
ballistic damages
ballistic impact
ballistic limit velocity
B-C system solid solution
beam end constraint
biomimetic composite
bismaleimide
bolt
boron-rich boride
braking
brittle composite
broken ellipsoidal inhomogeneity
buckling
603
603
485
532, 542
522
603
103
10
185
156
603
638
203
364
1, 438, 573
C/C composite
630, 759
carbon
775
carbon fibre
301, 342, 658, 759, 769
carbon magnesium composite
804
carbon/epoxy
225
carbon-carbon
638
ceramic matrix composites 621, 658, 675, 682, 694
709
cermet
709
CF/PEEK composite
114, 503
cluster polymetallocarbosilanes
741
coating
630
colloidal
775
combined strength and toughness
785
commingled prepreg
114
complex frequency equations
94
compliance
194
composite armor
464
composite laminates
583
composite layered plate analytical modeling 522
composite plate
36
composite progressive failure
397
composites processing
593
compression
638
II - 886
compression after impact

compression loading
compressive modulus
conservation laws
continuous fibre
corrosion protection
coupling effect
crack
crack growth
crack growth rate
crack propagation
crack tip element
crack tip stress singularity
craze
creep strength
cyclic loading
cylindrical bending
27, 503
438
785
522
775
859
103
342, 431
332
215
234
665
387
312
167
682
94
damage
203, 342, 374, 513, 638
damage mechanics
364
damage mechanisms
114
damage tolerance
10, 17, 397, 513, 562
damping capacity
833
debonding
312
delaminated composite plates
47
delamination
1, 36, 234, 290, 353, 415, 431, 438
503, 532, 573, 665
delamination onset
185
delamination resistance
244, 254
delamination tests
322
design
474, 562
discontinuous plies
353
discrete stiffener theory
84
dislocations
833
dynamic plate behavior
47
electrets and materials
electrical property
electrophoresis
energy release rate
epoxy
equivalent inclusion method
eutectic reaction
experimental design
experiments
extreme value theory
fabric orientation
fabrics
failure behavior
failure criteria
failure mechanism
fatigue
fatigue behaviour
fatigue crack growth
fatigue damage
859
881
593
215
342
364
603
27
1
254
447
542
301
156
225, 630
124, 135, 156, 185
114, 145, 177
194
211
fatigue life model

203
fatigue strength
167
fiber orientation distribution
10
fiber pull-out and push-out
387
fiber reinforced plastics
167
fibre breakage
532
fibre orientation
415
fibre pull-out
646
fibre reinforced composites
273
fibre tests
870
fibre-reinforced polymer-matrix laminated
composites
145
fibres
613
finite element analysis
1, 353
finite element method
290, 583, 759
finite element modelling
397
fractography
225, 573
fracture
10, 342, 387, 665, 694
fracture mechanics
353
fracture mechanism
312
fracture mirror
646
fracture toughness
215, 234, 424, 431, 447, 675
free vibration
36, 55
free-edge effects
453
friction
387
fusion bonding
1
glass fabric
glass fibre
glass matrix composites
glass/polyester composite
global
gradient design concept
graphite
graphitic foams
447
244, 342
593
135
438
464
759
825
hardness
heterogeneous materials
hierarchical model
high speed testing
high temperature
high velocity impact
higher-order shear deformation theory
homogenisation
homogenisation approach
hot-pressing
hybrid element
hybrid matrix composite
hybrid short-fiber
hypervelocity impact damage
hysteresis loop
638
332
552
405
215, 301
522
55
374
552
593, 613
290
769
301
474
124
impact
impact damage
impact loading
impact perforation
impact strength
indentation test
inelastic strains
initial defect
1, 74, 495, 562, 573

17, 583
503
522
234
759
682
353
injection molding
851
interface
613, 709, 759, 775, 845
interface coating
741
interfacial constitutive laws
47
interfacial crack
387
interferometry
438
interlaminar delamination toughness
322
interlaminar shear stress
185
laminar materials
laminate
laminated composite
laminated composite beam
LaRC-IAX
LaRC-SI
large deflection
lateral loading
liquid phase sintering
load carrying capacity
load frequency
loading rate
local
low velocity impact
low-energy impact
845
55, 431
552
103
851
851
74
74
613
364
124
215
438
513
177
magnesium
833
magnesium alloy
424
manufacturing
794
manufacturing defects
322
marine
244, 415
material surface law
453
matrix
775
matrix cracking
621
maximum stress failure criterion
495
mean field theory
364
mechanical properties 312, 658, 703, 775, 794, 804
mechanical strength
833
metallic fibre reinforcement
593
metal-matrix composite
211
metal-polymer electrets
859
metals composite
845
method of cells
47
microcracking
682
micromeasurement instrument
785
micromechanics
47, 332, 364
microstructural characterisation
646
microstructure
703, 833
mixed mode
665
MMC
804
mode I
215, 234
modelling
342, 562, 813
modulus
813
Moir
438
molybdenum disilicide
741
mullite
775
multi-layers
845
natural fibre
natural frequency
nondestructive inspection
794
36, 103
503
II - 887
non-linear dynamic response

notch
notched sample
64
301
10
oblique reinforecement
off-axis loading
open-cell foams
orbital debris
orientation influence
oxide film
oxygenless ceramic polymer precursors
694
552
825
474
804
630
741
PAEK resin
301
parallelism
374
particle or short-fiber reinforced composites
364
particle size
312
particulate composites
751
particulate filled polypropylene
312
PEI
851
perforation
542
phase separation
229
pin
156
plain weave
84
plastic flow
312
plastic zone
353
plate
55
ply orientation
103
polarization
859
polyamide composites
405
polycarbonate
452
polyethylene
542
polyimide thermoplastic
851
pores
322
post failure
532
post-failure behavior
495
post-impact fatigue
17
potential energy release rate
290
powder metallurgical technique
424
powder processing
845
power law equation
194
precursor
735, 769
precursor pyrolysis
658
p-Ritz method
55
processing
27, 825, 881
progressive modeling
156
PTCR composite
881
pultruded tubing
74
pyrolysis
735
quality
quasi-static modeling
random fiber composites
rate effect
rectangular cutout
reinforcement cracking
residual strength
residual velocity
resin transfer molding
ring stiffened composite cylindrical shell
II - 888
27
74
194
234
84
364
135
522
27
84
sandwich beam
sandwich plates
scanning acoustic microscope
self reinforce
SEM
shaping
shear
shear modulus
sheet molding compounds
shock-wave propagation
SiC
SiC particles
SiC/Si3N4 ceramic
SiC/SiC
Si-C-N nanometer particles
silicon carbide
silicon carbide fiber
silicon nitride
single leg bending
singular element
singular interlaminar stresses
sintering
skin-stringer
slot
small arms body armor
S-N curve
sol-gel
specific damping capacity
spider silk
static strength
statistical model
stick-slip phenomenon
stiffness degradation
stiffness loss
stitch-bonded variable layer fabric NVG
stochastic process modelling and analysis
strain energy release rate
strain rate dependence
strength
strength criteria
strength ratio
stress redistribution
stress singularity
stress transfer
stress/strain
structural foams
structures
subdomain decomposition techniques
subsurface
surface coating
survivability
T300/976
temperature influence
tensile behaviour
tensile modulus
tensile test
tension-tension fatigue
test data evaluation
testing
495
513
503
703
881
735
225
785
194
552
769, 775
424
735
646
675
613
211
675, 703
665
290
453
593, 735
573
301
464
177
769
94
785
167
332
234
495
203
870
145
353, 387
405
859
273
10
621
290
813
646
825
573
374
638
845
485
177
804
405
785
646
114
273
562
tetragonal zirconia polycrystals

751
tetravinylsilane
741
textile reinforced concrete
870
thermal effect
124
thermal stability
804
thermoelasticity
374
thermoelectric property
262, 612
thermoplastic
1, 124, 794
thermoset
415
thick annular composite plates
64
through thickness stress
353
time-temperature superposition principle
167
Timoshenko beam theory
103
tip mass
103
toughening
675
toughness
387, 415
transition behavior law
453
transmission electron microscopy
759
tribological characteristics
859
tungsten carbides
751
ultrasonic C-Scan
unidirectional
unidirectional fibres
unidirectional laminates
unsaturated curing organosilicons
1
244, 775
833
453
741
unstable fracture
variational principle
vibration damping
vibratory and fatigue characteristics
vinyl ester
viscoelasticity
viscoplasticity
voids
234
290
94
485
244, 254
94, 215
47, 215
322
wall effect
453
warp knitted multi-axial layer fabric WIMAG 870
water absorption
503
wear
638
Weibull distribution
177
wetting and spreading
859
wood composites
813
woven fabric composite
447, 804
woven fibre reinforcement
593
woven glass fibres
415
woven glass roving reinforcement
254
woven preform
769
woven roving
244
YAG (yttrium aluminium garnet)
Youngs modulus
613
759
II - 889
AUTHOR INDEX
ABACHI, P.
ABE, Kenichi
ADDESSIO, Frank L.
AL-HMOUZ, I.
ALLEN, D.H.
ANDERSON, David P.
ANDERSONS, J.
ANSART, Thierry
ARENDTS, F.J.
ASP, Leif E.
ATODARIA, D.R.
BABA-KISHI, K. Z.
BAO, Xiaoheng
BASTE, Stphane
BEAUMONT, P.W.R.
BGUELIN, Ph.
BERTHET, Florentin
BILLET, Jean Louis
BISCHOFF, Thomas
BOCCACCINI, Aldo R.
BOEIJSMA, J.
BOLDUC, M.
BONORA, N.
BOS, H.
BRANDT, Fredrik
BRUNET, Louis J.
BURCHILL, Peter J.
BURR, Alain
CARDEW-HALL, M.
CARMAN, Greg P.
CHANG, Fu-Kuo
CHEN, C.C.
CHEN, Zhaohui
CHENG, Haifeng
CHENG, Laifei
CHO, Young-Tae
COMPSTON, P.
CRAFT, William J.
CRASTO, Allan S.
CUNTZE, Ralf G.
CZARNECKI, Gregory J.
DAI, Jin
DAVIDSON, Barry D.
DAVIES, Ian J.
DAVIES, P.
DEVOS, Pierre
DEVRIES, F.
DIAO, Xiaoxue
DIMANT, R.A.
DU, Shanyi
DUBNIKOVA, I.L.
DUMONTET, Hlne
EL-KARMALAWY, M.
FARRIES, Pamela M.
II - 890
424
522
47
124
94
825
135
27
438
322
194
881
735
682
342
405
27
453
870
593
709
532
203
794
322
74
244, 254
621
94
17
397
55
658, 675, 735
769
735
630
364
244, 415
513
185
273
263
332
665
646
415
27
374
114
342
332
312
453
145
613
FARYNA, Marek
FLORINA, E.K.
FRANZKE, Gerd
GANESAN, R.
GAO, Yun
GARKHAIL, S.
GERLIVANOV, V.G.
GESHURY, Amotz J.
GOLADE, V. A.
GOTO, Takashi
GREENHALGH, Emile
GUBIN, S.P.
HABERKO, Krzysztof
HABOUSSI, Mohamed
HAHN, H. Thomas
HAUSMANN, C.
HEIJENRATH, R.
HEYNE, T.
HILD, Francois
HIRAI, Toshio
HIROKAWA, Tetsuro
HOA, S.V.
HOJO, Masaki
HUANG, C.L.D.
HUGHES, Derke R.
HUMMEL, Sarah
INOUE, Mari
ISHIGURO, Takeshi
ISHII, Hitoshi
ISHIKAWA, Takashi
JAMES, Peter F.
JAR, P.-Y.B.
JIANG, Dazhi
JING, X.T.
JONES, F.R.
JUN, Xiao
KAINER, K.U.
KALYANASUNDARAM, S.
KAMIYA, Akira
KASANO, Hideaki
KAUSCH, H.H.
KAWABATA, Sueo
KEARNS, Kristen M.
KELKAR, Ajit D.
KERN, Heinrich
KESTELMANN, V. N.
KIM, Patrick
KIM, Ran Y.
KIM, Young-Wann
KITIPORNCHAI, S.
KO, Frank K.
KOHYAMA, A.
KOLSTER, B.H.
KOMAI, Kenjiro
KOMATSU, H.
751
741
870
145
845
794
741
464
859
603
573
741
751
453
17
804
794
804
621
603
646
145
234
64
513
851
785
167
364
1, 646
775
244, 415
583, 658, 769
447
775
431
424
94, 225
717
522
405
785
825
513
593
859
10
185
84
55
464, 785
759
709
503
542
Proceedings of ICCM-11, Gold Coast, Australia, 14th-18th July 1997
KONDO, Kyohei
KORSGAARD, J.
KUROKAWA, Tomoaki
KUSAKA, Takayuki
LHOSTIS, G.
LAI, Ren-Cheng
LECKIE, Frederick A.
LEE, Ya-Jung
LEE, Young-Shin
LN, F.
LESSARD, Larry B.
Li, Bin
LI, Fengmei
LI, Jianhui
LI, Shujie
LI, Shunling
LI, Yongqing
LI, Yuping
LIEW, K.M.
LIM, C.W.
LIN, Bor-Horng
LITYNSKA, Lidia
LIU, Fengrong
LIU, Hsien-Kuang
LOWE, Adrian E.
MAAS, J.H.
MAI, Y.-W.
MALKE, R.
MALLICK, P.K.
MATSUSHIMA, Masamichi
MAYENCOURT, Christine
MINOSHIMA, Kohji
MITROVIC, Milan
MIYANO, Yasushi
MORDIKE, B.L.
MORVAN, Jean-Marie
MUCHA, Herbert
MURAVIN, D.K.
NAKADA, Masayuki
NEMES, James A.
NEWAZ, G.M.
NIWA, Masako
NWOSU, Sylvanus N.
OCHIAI, Shojiro
OFFERMANN, Peter
OLIVERO, David
OSHMYAN, V.G.
TTINGER, O.
PAN, Jin
PATER, R.H.
PATON, Rowan
PEDZICH, Zbigniew
PEIJS, T.
PETROSSIAN, Z.
PINCHUK, L. S.
POLAHA, Jonathon J.
POPOVA, N.A.
POTTER, B.D.
PRONIN, Yu E.
290
135
234
234
374
36
621
495
84
374
156
845
703
603
709
431
735
709
55
55
725
751
658, 769
725
94, 225
709
114
301
194
1
833
503
17
167
424
682
717
312
167
74
203
785
263
234
870
851
312
804
211
215
813
751
794
353
859
665
741
215
741
PUTATUNDA, S.K.
QIAN, Tiancai
QIAN, W.
RADFORD, Donald
RAUO, Boko
REMOND, Yves
RESNYANSKY, A.D.
ROBERTS, Donald
ROMENSKY, E.I.
SATO, S.
SCHFF, W.
SCHALLER, Robert
SCHMITT-THOMAS, Kh. G.
SCZEPANIK-WEINMANN, M.
SELVARATHINAM, Alex S.
SERIZAWA, H.
SHEN, Wei
SHERCLIFF, H.R.
SHERWOOD, Tor W.
SHIBUYA, Masaki
SHOKRIEH, Mahmood M.
SHORTLIFFE, G.D.
SHYU, Yau
SIMPSON, Gary J.
SINGER, R.F.
SINGH, Sunil
SMITH, Charles
SONG, John W.
SOULAT, D.
STOLL, U.
STRAZNICKY, P.V.
SUN, C.T.
SUN, Yong
SUN, Y.Q.
TAI, Nyan-Hwa
TAKAHASHI, K.
TAKAO, Y.
TANAKA, Kazuto
TANIMOTO, T.
TAO, Jie
TENNYSON, R.C.
TODO, M.
TOHGO, Keiichiro
TRINH, Khanh
TSENG, Chih-Ming
TSIRLIN, A.M.
TSUNAKAWA, H.
TUNG, Betty
TYLKO, S.
UBELS, L.C.
VAN DER OEVER, M.
VAN HOOF, J.
VERTYACHIKH, I. M.
VEYRET, Jean-Bernard
WAGNER, Christiane
WAN, Hong
WANG, Biao
WANG, J.B.
WANG, W.X.
194
845
387
851
785
638
552
573
552
759
804
833
301
438
694
759
583
342
665
646
156
474
495
244, 254
804
573
851
464, 785
374
438
532
387
845
447
177
405, 542
215
503
10
431
474
405
364
397
177
741
759
851
532
562
794
532
859
613
638
211
332
447
215
II - 891
WANG, Xingye
WANG, Xing-Ye
WEILAGE, Bernhard
WEITSMAN, Y.J.
WIGGENRAAD, J.F.M.
WILLIAMS, Todd O.
WINKLER, Volker
WISNOM, Michael R.
WORSWICK, M.J.
WU, Jaili
WULFHORST, Burkhard
XIAO, Jiayu
XIAO, X.R.
XIE, Kai
XIONG, Z. X.
XU, Yongdong
YAGI, Kazuhiro
YANG, Deming
YANG, Z.M.
II - 892
658, 769
583
717
694
562
47
593
353
532
775
870
658, 769
124
675
881
630
290
211
103
YANG, Zhen-Guo
YE, L.
YEH, Meng-Kao
YIP, Ming-Chuen
YOKOYAMA, T.
YONG, Li
YU, Hui-Chia
YUAN, F.G.
ZHANG, Changrui
ZHANG, Daiming
ZHANG, Litong
ZHANG, Shuhong
ZHANG, S.
ZHENG, S.R.
ZHENG, Wen-Wei
ZHOU, Anchen
ZHOU, Xingui
ZHUO, Yue
301
114
36
177
103
431
177
215
675, 735
845
630
845
145
447
658, 769
675, 735
675
211

ICCM 11v2

Uploaded by

Copyright:

Available Formats

ICCM 11v2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

ICCM 11v2

Uploaded by

Copyright:

Available Formats

ELEVENTH INTERNATIONAL CONFERENCE

Dynamic Response Behaviour

Residual Strength of GRP at High Cycle Fatigue

A Singular Hybrid Finite Element Analysis of Interlaminar Crack Problems

Ceramic Matrix Composites

Structural Analysis of Ceramic-Matrix Composites

Preparation and Properties of 3D-Carbon Fiber Woven Preform Reinforced SiC

New and Non-Traditional Materials

Proceedings of ICCM11, Gold Coast, Australia, 14th-18th July 1997

COMPRESSION AFTER IMPACT (CAI) PROPERTIES

Volume II: Fatigue, Fracture and Ceramic Matrix Composites

However, one major disadvantage of CF/PEEK was also

Fig.1: Single hat panel: shape

Fig. 2: Three hat panel: shape

Proceedings of ICCM11, Gold Coast, Australia, 14th-18th July 1997

Fig. 3: Photo of impact test by drop weight

Fig.4: Delamination image for CPA04

Volume II: Fatigue, Fracture and Ceramic Matrix Composites

w/o Autoclave, 1 Hat

Compression tests were conducted after 12 strain

Fig. 6: Strain gage map for CPA panel

Fig. 7: Photo of compression test for CPA

Proceedings of ICCM11, Gold Coast, Australia, 14th-18th July 1997

200 < Prdc. Bkl. Str.

Fig. 8: Stress-strain behavior for panel

Fig. 9: Stress-strain behavior for panel

Single Hat: w/o Autoclave

Normalized Impact Energy (J/mm)

Fig. 10: CAI strengths vs.

Volume II: Fatigue, Fracture and Ceramic Matrix Composites

Fig. 11: Final Failure Mode of CPA 04

Fig. 12: Final Failure Mode of CPA 03 panel

PREDICTION OF BUCKLING BEHAVIOR BY FINITE ELEMENT ANALYSIS

Proceedings of ICCM11, Gold Coast, Australia, 14th-18th July 1997

Outside of delaminated regions, skin-stiffener attachment flange was merely modeled as a

Volume II: Fatigue, Fracture and Ceramic Matrix Composites

Proceedings of ICCM11, Gold Coast, Australia, 14th-18th July 1997

Engineering Mechanics Research Corporation, Users Manuals of NISA-II, Version 91,

Volume II: Fatigue, Fracture and Ceramic Matrix Composites

DAMAGE TOLERANCE OF BIOMIMETIC

School of Engineering and

KEYWORDS: biomimetic composite, damage tolerance, notched sample, strength ratio,

Proceedings of ICCM11, Gold Coast, Australia, 14th-18th July 1997

2 + R 2 + 3R 4 (KT 3)[5R 6 7R8 ] , x d

A11 A22 A12 + A11 A22 A12

Fig. 1: Lay-up of the biomimetic composite laminate

Volume II: Fatigue, Fracture and Ceramic Matrix Composites

(0 / 45 / 90/ -45 )2s

RESULTS AND DISCUSSION

Proceedings of ICCM11, Gold Coast, Australia, 14th-18th July 1997

Table 2: Elastic properties of the unnotched laminates (calculated values in parentheses)

49.38 (52.0) / 47.98 (52.0)*

* in two normal directions, parallel to surface fibers

Volume II: Fatigue, Fracture and Ceramic Matrix Composites

Proceedings of ICCM11, Gold Coast, Australia, 14th-18th July 1997

Volume II: Fatigue, Fracture and Ceramic Matrix Composites

Eriksson, I., Aronsson, C.-G., Strength of tensile-loaded graphite/epoxy laminates

Proceedings of ICCM11, Gold Coast, Australia, 14th-18th July 1997

EFFECT OF LOADING PARAMETERS ON