DEVELOPMENT OF PROBABILISTIC MODELS FOR LONG TERM RELIABILITY

OF SANDWICH COMPOSITES IN SALINE FREEZE/THAW ENVIRONMENT FOR

CIVIL ENGINEERING APPLICATIONS

Dissertation

Submitted to

The School of Engineering of the

UNIVERSITY OF DAYTON

In Partial Fulfillment of the Requirements for

The Degree of

Doctor of Philosophy in Engineering

By

Sadra Emami

UNIVERSITY OF DAYTON

Dayton, Ohio

December 2017
DEVELOPMENT OF PROBABILISTIC MODELS FOR LONG TERM RELIABILITY

OF SANDWICH COMPOSITES IN SALINE FREEZE/THAW ENVIRONMENT FOR

CIVIL ENGINEERING APPLICATIONS

Name: Emami, Sadra

APPROVED BY:

Elias Toubia, Ph.D., P.E. Charles Browning, Ph.D.

Advisor Committee Chairman Committee Member
Associate Professor Professor
Department of Civil and Department Chair of Chemical and
Environmental Engineering and Materials Engineering
Engineering Mechanics

Kellie R. Schneider, Ph.D. Donald A. Klosterman, Ph.D.

Committee Member Committee Member
Assistant Professor Associate Professor
Department of Engineering Department of Chemical and
Management, Systems and, Materials Engineering
Technology

Steven Donaldson, Ph.D.

Committee Member
Associate Professor
Department of Civil and Environmental
Engineering and Engineering Mechanics

Robert J. Wilkens, Ph.D., P.E. Eddy M. Rojas, Ph.D., M.A., P.E.

Associate Dean for Research and Innovation Dean, School of Engineering
Professor
School of Engineering

ii
 © Copyright by

Sadra Emami

All rights reserved

2017

iii
 ABSTRACT

DEVELOPMENT OF PROBABILISTIC MODELS FOR LONG TERM RELIABILITY

OF SANDWICH COMPOSITES IN SALINE FREEZE/THAW ENVIRONMENT FOR

CIVIL ENGINEERING APPLICATIONS

Name: Emami, Sadra

University of Dayton

Advisor: Dr. Elias Toubia

Structural engineers commonly use balsa wood and PVC foam as core materials for

sandwich composite structures. The long-term reliability and damage mechanism of these

composite sandwich structures under severe environmental conditions are still unclear.

These liability concerns prevent most civil structural engineers from considering this

material in infrastructure applications. Also, they must account for the unique nature of the

civil construction industry and offer an advantage over traditional building materials such

as steel and concrete. Introducing a strength reduction factor (ɸ) will increase the chance

for using sandwich composites in civil engineering applications. Using a probabilistic

approach is a means to reach to this goal. In this project, the long-term effects of

freeze/thaw in a saline environment were simulated by exposing sandwich composites to

100 days of freeze/thaw exposure (-20oC to 20oC) in the presence of a 3% NaCl saline

solution. The sandwich panels were comprised of using balsa wood (SB100) and foam

iv
core (Airex C70.55) with fiberglass/vinyl ester face sheets, fabricated with vacuum assisted

resin transfer molding (VARTM).Samples were tested for core shear, core compression,

and peel tests. Results confirmed that exposure reduced the balsa wood core properties

significantly, however, PVC foam core shear modulus increased by 25%, and the

compression modulus reduced by 25%. Simulated lifetime core shear fatigue curves were

also developed and evaluated. A substantial reduction in tensile, shear, and compression

properties was observed for the face sheets constructing the sandwich panels. The thermal

cycling events degraded the matrix binding, the warp and fill fibers, thus impairing the

structural integrity of the cross-ply laminate. After testing, a reliability-based approach was

used to examine the “strength reduction” factors (ɸ) of the core materials. Resistance

factors of 0.79 and 0.95 were obtained for balsa and foam core composite sandwich

materials, respectively. The results of this research will ultimately lead to a probabilistic

analysis model that will eventually act as a benchmark to reliably predict the performance

of sandwich core materials.

v
This dissertation is dedicated to my beloved family and friends who have provided me
with the support, encouragement, and strength I needed throughout my graduate career.

vi
 ACKNOWLEDGMENTS

I sincerely would like to thank my advisor, Dr. Elias Toubia, for giving me the

opportunity to work in this area, and motivating me throughout my Ph.D. program. His

excellent insight into science and engineering, management’s skills, generosity, and

positive attitude have positively affected my academic and social lives. He was extremely

supportive throughout my Ph.D. thesis. Without his technical and funding support, it would

have been impossible to carry out this work.

Special thanks go to materials engineering department chair, Prof. Charles

Browning for his support, transferring his extreme knowledge in composites and helps. In

addition, I would like to thank Dr. Donald Klosterman, who was the faculty advisor,

SAMPE chapter president, and my Ph.D. committee member. He was extremely helpful,

nice, and collaborative. Also, I would like thanks, Dr. Kellie Schneider and Dr. Steven

Donaldson for their helpful advice.

I would like to acknowledge administrative assistant of the department, Dana

Tobias. And staff members from the University of Dayton Research Institute (UDRI),

specifically to Ms. Mary Galaska, Ms. Marlene Houtz, Mr. Dale Grant, Carl-William O

Sjoblom, and all staffs in the smart lab. And the composite advantage LLC for molding the

sandwich composite panels. Finally, the invaluable contribution made by my dad, mom

towards my academic success was my biggest source of inspiration to push forward.

vii
 TABLE OF CONTENTS

ABSTRACT......………………………………………………………………….…...….iv
DEDICATION………………………………….………………………………...………vi
ACKNOWLEDGMENTS ...................................………………………………...……..vii
LIST OF FIGURES…………………………………………………………..……......….x
LIST OF TABLES……………………………………………………………......……..xiv
CHAPTER 1 INTRODUCTION…………………………………....…….…….…….1
1-1 Problem Statement….………………………….………..............….......4
1-2 Significance of the Research……………………………………….......5
CHAPTER 2 BACKGROUND…………………..…..……………...………….....….7
2-1 Reliability in Civil Engineering Composite Materials………………....7
Traditional and Probabilistic Composite
Structural Design Process…………………………………………8
Steps in Quantifying Randomness……………………………….10
Parameters to Calculate Reliability………………………………11
2-2 Raw Materials Composition and Structure……………………………15
Balsa Wood Structure……………………………………………15
PVC Closed Cell Foam Structure………………………..………18
2-3 Composite Durability…………………………………...……………22
Glass Fiber Laminate Durability…………………………………22
Sandwich Composite Durability…………………………………23
CHAPTER 3 THEORETICAL APPROACH………..………………………....…...26
3-1 Selection of Probability Distribution………………………….…..…..27
3-2 Limit State Functions.……………………..…….…………………….30

viii
CHAPTER 4 EXPERIMENTAL APPROACH………………………………….....35
4-1 Durability Study of Sandwich Composite and Laminate……...……...35
Panels and Laminates Preparation…………………………….....35
4-2 Environmental Experiment Setup………………………….……….....38
4-3 Test Setup………………………………………………………...…...40
GFRP Laminate Testing…............................................................44
4-4 Characterization Methods……………………………………………...46
CHAPTER 5 RESULTS AND DISCUSSION…..………………...………..…….…50
5-1 Balsa Wood Core Sandwich Composites Properties……………...…..50
5-2 Foam Core Sandwich Composite Properties………………….…...….59
5-3 Laminate Tests Results……………………………………….…...…..68
Effect on Shear Properties……………………………………….69
Effect on Tensile Properties…………………………………..….71
Effect on Compressive Properties………………………………..73
Root Cause Failure Analysis……………………………………..74
Damage State in Woven Roving Architecture………………..….78
Matrix and Fiber Damage…………………………………….….80
5-4 Model Results………………………………...……………………….83
Parameters Estimation……………………………………….…..95
CHAPTER 6 SUMMARY OF SIGNIFICANT RESEARCH
RESULTS AND FINDINGS………...………………………….....…98
CHAPTER 7 CONCLUSIONS AND RECOMMENDATIONS…..………….........102
BIBLIOGRAPHY………………….…………...………………..…………………..…104

ix
 LIST OF FIGURES

Figure 1. A typical sandwich composite..…………...…...……………...…………….2

Figure 2. Probabilistic analysis concept [55]…...…………………..……………........9
Figure 3. Three-dimensional presentation of the problem [55]…………...…..…..…14
Figure 4. Bascule FRP composite footbridge – Fredrikstad in Norway [11]….....….16
Figure 5. Simplified microstructure of balsa wood showing middle
lamella and tracheid cells with P, S1, S2 and S3 cell wall layers [16]….....17
Figure 6. Panel composed of balsa blocks joined by PVA base adhesive [19]….…..18
Figure 7. PVC repeat unit………………………………………..…………………..19
Figure 8. Schematic showing an approximately cellular microstructure for
mechanical modeling. a) open-cell foam, b) closed-cell foam [24]……….20
Figure 9. Chord slope calculation method…………………………………………...32
Figure 10. VARTM process of molding the panels……………………………….….37
Figure 11. A cycle pattern for F/T thermal cycling…………………………………...39
Figure 12. Left: F/T machine with samples inside; Right: flexural
samples (top) and compression sample (bottom) …………………………39
Figure 13. Four-point bending fixture for fatigue testing…………………...…..….…42
Figure 14. A sample under compression test…………………………………….……43
Figure 15. Peel test a) steel T-shape tabs bonded to the specimen.
b) specimen with interface cut area, c) specimen ready for test …….…....44
Figure 16. Iosipescu test setup…………………………………….………………..…45
Figure 17. a) Tension and b) compression test setup…………………………..…..….46
Figure 18. DMA fixtures, a) 3-point bending, b) shear……………………….………48
Figure 19. Specimens inside of water tank and ultrasonic scanning probe………...…48

x
Figure 20. Areal weight of balsa wood core sandwich after F/T exposure……….…..50
Figure 21. Molded balsa wood after a) 50 cycles, b) 200 cycles……………………...51
Figure 22. Balsa core shear strength (left) and modulus (right)
-ASTM C393 and D7250……………………………….…………………52
Figure 23. 0 cycle core shear failure mode……………………………………………52
Figure 24. 200 cycles core shear failure mode………………………………….….…53
Figure 25. Flexure results for balsa core sandwich samples……………………….…53
Figure 26. Balsa wood fatigue failure mode……………………………………..……54
Figure 27. Balsa wood core shear fatigue life prediction for exposed
(200 cycles) and unexposed Specimens (0 cycles), R = -0.1 at 4 Hz.......…55
Figure 28. Balsa Wood Texture a) Unexposed b) Exposed Samples (Optical
Microscopy)………………………………………………….………..…..55
Figure 29. Balsa Core Flatwise Compression Strength (Left) and Compression
Modulus (Right)-ASTM C 365 Preliminary Testing at 0 and 200
cycles…………………………………………………………………..…..56
Figure 30. Load-stroke curve for exposed and unexposed specimens………………..57
Figure 31. Balsa wood fracture toughness for unexposed (0 cycles) and
exposed (200 cycles) specimens (ASTM E2004)………………...……….57
Figure 32. a) A specimen under peel test, b) failure modes of exposed and
unexposed specimens……..…………………………………………....….58
Figure 33. Foam core shear strength and shear modulus -ASTM C393
and D7250……………………………………………..……………….......59
Figure 34. Shear four-point test………………………………………..……………...60
Figure 35. Four-point flexure results for foam core sandwich samples …………..….60
Figure 36. SEM: a) unexposed Foam and b) exposed Foam after 200 F/T
cycles, X-Ray CT-Scan: c) foam Core Sandwich
Specimen after F/T Exposure…………………………………………..….61
Figure 37. DMA testing results, a) shear mode testing,
b) 3-point bending mode………………………………………….….……63
Figure 38. Fatigue failure mode of PVC foam………………………………….…….64

xi
Figure 39. Foam core shear fatigue life prediction for exposed (200 cycles)
and Unexposed Specimens (0 cycles), R = -0.1 at 4 Hz………………..…64
Figure 40. Foam core flatwise compression strength and compression modulus
ASTM C 365 preliminary testing at 0 and 200 cycles…………….………65
Figure 41. Load-stroke curve for exposed and unexposed specimens…………..…....66
Figure 42. PVC foam Fracture Toughness for Unexposed (0 cycles)
and Exposed (200 cycles) Specimens (ASTM E2004)………………....…66
Figure 43. a) A specimen in middle of peel test, b) specimens after peel test………..67
Figure 44. Overall results of exposed (200 cycles) and unexposed (0 cycles)
specimens………………………………………………………………….69
Figure 45. Shear failure modes for a) unexposed; and b) exposed specimens……..…70
Figure 46. Percent change in shear modulus and strength………………………….…70
Figure 47. Micrograph of the cross section after failure at the V-notch location a)
unexposed: showing only interlaminar cracks, b) Exposed: showing
interlaminar and intralaminar cracks…………………………..…………..71
Figure 48. Percent change in tensile modulus and strength………………..………….72
Figure 49. Tensile test failure modes; unexposed and exposed specimens…………...72
Figure 50. Longitudinal and transverse Stress-Strain curves for
exposed and unexposed specimens…………………………………...…..73
Figure 51. a) Percent change in modulus and strength, b) Compression failure
modes…………………………………………………………………..….74
Figure 52. DMA results for composite, a) 1st DMA heating cycle, b) 2nd DMA
heating cycle…………………………………………………………….....77
Figure 53. X-Ray computed tomography showing damage in exposed samples a)
surface plies, b) longitudinal cross-section, c) Transverse cross
section……………………………………………………………………...79
Figure 54. Ultrasonic C-scan results (102 mm wide specimens); a) unexposed:
grid pattern, b) exposed: linear pattern…………………………………....80
Figure 55. Local magnified optical image at weave junction showing matrix
failure between damaged warp fibers and fill fibers after exposure…..…..81
Figure 56. SEM micrograph showing degradation of fiber/matrix interface and
fiber fracture…………………………………………………...….…….…81
Figure 57. SEM picture of damaged E-glass fiber after exposure……………...……..82
xii
Figure 58. Shear strength statistical plots of balsa core material…………..…..…..….86
Figure 59. Shear modulus statistical plot of balsa core material………………..….…87
Figure 60. Shear strength statistical Plot of foam core material……............................88
Figure 61. Shear modulus statistical plot of foam core material……….......................89
Figure 62. Compression strength statistical plot of balsa core material…………...….90
Figure 63. Compression modulus statistical plot of balsa core material……...………91
Figure 64. Compression strength statistical Plot of foam core material……...……….92
Figure 65. Compression modulus statistical plot of foam core material……..……….93

xiii
 LIST OF TABLES

Table 1. Probability distributions functions [59]………………………………….….…29

Table 2. Properties of composite constituents……………………….…...………..……36
Table 3. Number of samples and dimensions for sandwich composites.……...………..40
Table 4. Number of samples and dimension for GFRP laminates……........................…41
Table 5. Core shear fatigue matrix- balsa wood core………………...……………....…42
Table 6. Core shear fatigue matrix- Airex C70.55 PVC foam……………..……………42
Table 7. Mechanical test results………………………………..………...…………..….68
Table 8. Summarized Tg results from DSC………………………………...…………...78
Table 9. Statistical values for balsa and foam core shear test
of sandwich composites………………………………...……………………...83
Table 10. Statistical values for Balsa and foam core compression
test of sandwich composites……………………………………....………..….84
Table 11. Data results of OSL for Normal, Lognormal and Weibull distribution…….....85
Table 12. Selected distributions for balsa core and foam mechanical properties...…...…94
Table 13. Selected distributions for variables ………………………………….……......95
Table 14. Resistance factor results……………………………………………..…….......96
Table 15. F/T coefficient introduced in this research ……….……………………....…..97

xiv
 CHAPTER 1

INTRODUCTION

Composite sandwich construction consists of two face sheets with a high level of

strength and stiffness, which are adhesively bonded to both sides of a lightweight core

sheet. In general, face sheets are made of composite laminates such as glass fiber or carbon

fiber with resin; the core is typically honeycomb, end-grain balsa wood, or closed-cell

polymer foam. Each component has a different function: face sheets carry bending and in-

plain forces, while core sheets carry the transverse shear load. This function of sheets helps

prevent face-sheet buckling and holds the face sheets in positions far away from the neutral

axis in order to maximize the flexural stiffness of the structure [1]. Figure 1 shows a typical

sandwich composite

1
 Figure 1. A typical sandwich composite

Sandwich structures are currently being utilized extensively in structural

applications such as aerospace, aircrafts, marine vessels, tanks, refrigerator containers,

transportation structures, and bridge decks where minimum structural weight and

maximum stiffness/strength are essential. One of the main applications of composite

sandwich materials is in civil engineering. Transportation infrastructure systems

deteriorate at an alarming rate in the U.S. and many other countries. For example, of the

nearly 600,000 bridges in the U.S. transportation system, about 143,000 are older than 50

years and unsuitable to meet the current or projected traffic demands [2]. Therefore, more

durable bridge decks are highly desired by transportation agencies. In addition to

durability, two issues need to be addressed in making composite sandwich materials

applicable to civil engineering techniques. The materials must accommodate the unique

nature of the civil construction industry and must offer an advantage over traditional

building materials such as steel and concrete. Additionally, liability concerns prevent the

employment of less known products if the basis for their technical design data and long-

term durability behavior has not been established. Consequently, for civil applications,

2
composites are deemed to be less reliable than conventional construction materials (steel,

concrete, etc.) for which the design methods, standards, and supporting databases already

exist. These issues are best addressed in the context of probability-based limit state design

[2]. This approach is capable of quantifying the inherent risk of a design and assessing the

sensitivities of design variables, which will be applied throughout this work.

Furthermore, several factors affect the properties of composites. In general, the

mechanical and physical properties of polymer composites are adversely affected by

several environmental factors. Consequently, the ability to forecast changes in material

properties as a function of environmental exposure is required in designing extended

service lives. For example, bridge decks or other structures and infrastructures are

subjected to a wide range of repetitive loadings (fatigue) and changes in temperature and

moisture levels. Load carrying ability can be significantly reduced by the presence of

moisture in the polymer foam or balsa core, which is absorbed during use or storage in

humid environments. Hygrothermal conditions have been observed to cause degradation

in the composite through changes such as plasticization, hydrolysis, and saponification of

the resin, including fiber-matrix debonding and matrix micro-cracking in the composite

and even pitting and crack of fiber [3–5]. Moreover, the presence of deicing salts in winter

must be considered as part of the bridge deck service environment. Cold temperature is

another concern in assessing composite sandwiches, since the concrete deterioration in

most often observed in freezing temperatures and the durability of fiber-reinforced

composite is an advantage [6].

One of the advantages of fiber-reinforced materials that helps designers is the

mechanical behavior of materials. In fact, it can be designed with several parameters such

3
as a different choice of resin to fiber ratio, fiber orientation in different lamina or stacking

sequences [7]. Reliability prediction is difficult for these types of materials, as there are a

wide range of failure mechanisms and random variables that influence those mechanisms

which must be taken into consideration while modeling the reliability of composites [8].

This larger number of variables, as compared to isotropic materials, shows that the strength

of composite materials is more uncertain, and due to anisotropic properties, they require

many variables for description [9].

1-1 Problem Statement

The long-term goal of this research is to assess the feasibility of utilizing sandwich

composite materials for infrastructure applications in freeze-thaw environments. Hence,

the first goal is to understand the durability of these materials through the following

uncertainties:

• How will the mechanical properties of the composite be affected by exposure

to a freeze-thaw environment?

• What will happen to the microstructure of the composite components after

long-term exposure to freeze-thaw conditions?

• What are the mechanisms of degradation?

Additionally, one of the current barriers to the implementation of these materials in

civil engineering applications is the lack of technical knowledge regarding their long-term

durability; thus, structural engineers do not have access to the design standards for these

materials as they do for traditional building materials such as wood, steel, and concrete.

Therefore, the purpose of this study is to develop probability models that describe the

4
reliability of sandwich composite materials, focusing on freeze-thaw conditions in this

study. The specific questions to be addressed are the following:

• Are these materials potentially reliable for long-term application in freeze-

thaw environments?

• Which probabilistic analysis model is appropriate for these composites under

freeze-thaw conditions?

1-2 Significance of the Research

The National Academy of Engineers has issued a list of grand challenges, which

outlines various initiatives for engineering achievement, including the restoration and

improvement of urban infrastructure [1]. In 2013, the American Society of Civil

Engineering issued a report card for the U.S. infrastructure, with bridges earning a grade

of C+ [2]. The average age of bridges in this country is 42 years, and it is estimated that an

investment of over $75 billion is required to repair these antiquated structures. It is

important that this infrastructure is not only repaired but also improved, with new

technologies that are more beneficial than legacy systems. This work aims to help

determine the feasibility of utilizing sandwich composite materials in infrastructure

applications.

Structures such as bridge decks or other civil infrastructure are subjected to a wide

range of repetitive loadings (fatigue) and changes in temperature and moisture levels. The

load carrying ability of sandwich composites can be significantly reduced by the presence

of moisture in the polymer foam or balsa core, which may be absorbed during use or storage

5
in humid environments [4–5]. The F/T effect, as well as the presence of deicing salts in

winter, must be considered as part of the bridge deck service environment [5].

One advantage of using sandwich composites in bridge decks is their weight.

Sandwich composite bridge decks weigh only 10–20% of a structurally equivalent

reinforced concrete deck. A lighter weight (load) for the structure can translate into savings

across the structure, as the size of structural members and foundations can be reduced

accordingly. Fabricating deck panels in a factory and shipping them to the bridge site also

offers several advantages over cast-in-place concrete. For example, quality can be closely

monitored in a controlled environment, the possibility of weather delays is minimized, and

most significantly, bridge downtime can be substantially reduced.

6
 CHAPTER 2

BACKGROUND

2-1 Reliability in Civil Engineering Composite Materials

Most observable phenomena in the world contain a certain amount of uncertainty;

that is, they cannot be predicted with certainty. Repeated measurement of a physical

phenomenon generates multiple outcomes. Among these many outcomes, some are more

frequent than other. The occurrence of multiple outcomes without any pattern is described

in terms of uncertainty. Engineers always attempt to account for the uncertainty in their

analysis and design. In engineering, satisfactory performance is often given in terms of the

probability of successfully satisfying some performance criterion. This probabilistic

assurance of performance is referred to as “reliability” [54].

Since the performance of composite structures is greatly influenced by the

manufacturing quality control procedures, operating conditions, and environmental effects,

the use of a deterministic approach (that is, the current industry approach) to design and

analyze composite structures may undermine their high utilization and optimization

potential. This may result in higher safety factors. Moreover, the reliability of such an

approach cannot be explicitly qualified.

7
 One of the present barriers to the implementation of composite materials in

infrastructure applications is a lack of technical knowledge regarding their long-term

durability; thus, structural engineers do not have access to the design standards for these

materials as they do for traditional building materials such as wood, steel, and concrete.

Therefore, the purpose of this study is to use probabilistic design in order to predict and

quantify the structural strength and performance of certain FRP composites, particularly

sandwich composites.

Traditional and Probabilistic Composite Structural Design Process

Traditional design approaches for composite materials have employed

deterministic criteria for failure analysis. Their values have been established after many

years of experiments and calibration by judgment; however, they are not suited to new

materials with particular features, such as composites. Safety factors are generated by this

process. They are derived on the basis of past experience but do not guarantee safety or

satisfaction in terms of performance. Moreover, they do not provide information regarding

the way in which the different parameters of the system influence safety. Thus, it is difficult

to design a system on the basis of uniform distribution of safety level. Design safety is

ensured by keeping the resistance of structure (RN) greater than the load on the structure

(SN) with a specified margin of safety (SF), which can be presented as

Nominal SF = RN/SN

Both the resistance and load in this formula are conservative and may fail to convey

the actual margin of safety in a design.

8
 In the probabilistic design of engineering system, the future is predicted with the

help of information from the past, which includes experience and judgment. New

approaches are required to predict the reliability of composite structures, since strengths

and stresses are random variables. The basic probabilistic approach can be summarized in

the following: (1) quantification of all input variables required for structural analysis, (2)

definition of the resulting stress and strength of the structure for each possible failure mode,

and (3) evaluation of the resulting probability of structural failure. Figure 2 illustrates this

process.

Figure 2. Probabilistic analysis concept [55]

The left-hand side presents the input data required for the determination of the

applied stress distribution, with each stress possibly having a different statistical

distribution. Alternatively, the right-hand side depicts the various capabilities of the

structure. The middle part indicates the associated failure probabilities for each failure

mode, which are applied stress and resistive component strength distribution.

9
 Engineers often do not know the underlying distribution of random variables, and

thus, they have to rely on test data and observation for the estimation of the parameters of

the distribution functions. By collecting more information, engineers can reduce statistical

uncertainties. This can be accomplished by compiling test data, building a visual

representation of the test data, such as histograms, or by simulations and experience.

Steps in Quantifying Randomness

Mathematical modeling and representation of a random variable is an elementary

task in any probabilistic formulation, which needs to be conducted systematically.

The first step is data collection, for which an appropriate data population is

required. In this process, understanding, mean, variance, the coefficient of variation

(COV), and skewness can describe the randomness of a sample population. A complete

description can be obtained by plotting the information graphically in the form of a

histogram. In preparation of the histogram, the number of bins or width of the interval plays

an important role in extracting the information related to randomness.

In order to make it simple yet comprehensive, analytical models can be emplyed.

For this purpose, it is necessary to treat discrete and continuous random variables

separately. Probability density function (PDF) and cumulative density function (CDF) are

usually used to calculate the probability of random variables. The nature of structures is

based on the continuous random variables. Commonly used continuous random variables

include the normal or Gaussian distribution, lognormal random variable, and Weibull

random distribution. There are several methods to determine distribution and parameters

from observed data such as probability paper and statistical tests, including the chi-square

10
test and K–S test. In order to estimate the parameters of a distribution, the method of

moment and method of maximum likelihood are the most common.

Parameters to Calculate Reliability

Instead of using the safety factors for the resistance alone, as in the working stress

method, or for the loads alone, as in the ultimate strength method, it is more rational to

apply the safety factor to both resistance and load, as is done in the case of concrete or steel

load and resistance factor design (LRFD) [56]. This concept has been developed for

conventional building materials, wood-based building materials, for example. ASTM

D5457–15 is used to calculate the reference resistance of wood-based materials and

structural connections [57].

The LRFD method depends on the applied load and resistance variables, which are

both described below:

Applied Load: Loads (applied stress) to the structure are random processes. Civil structures

are often designed for a 50–year lifetime, which means that the maximum loads occurring

over a 50–year period must be taken into consideration. The loads are represented by

stochastic processes and vary continuously with time or take on discrete values. The

instantaneous values of the loads are often considered as the mean values. Permanent loads

include the weight of the structure and immoveable fixtures (walls, carpet, and roof). These

loads have a low variance due to the fact that the permanent load is closely tied to the

construction materials. A permanent load is also known as a dead load.

Sustained loads change at discrete times, but remain relatively constant between

changes. Sustained loads include those produced by occupancy (weight of people,

11
furniture, vehicles, equipment) as well as environmental changes (temperature, humidity,

snow).

Transient loads have relatively shorter durations and occur infrequently.

Earthquake and severe wind loads fall under this category.

Multiple loads are likely to act on the structure at any given time. However, it is

unlikely that multiple types of maximum loads will be experienced at the same time, such

as maximum earthquake load and maximum snow load. Several methods have been used

to combine the effects of multiple load types.

The Ferry Borges–Castanheta (FB–C) [58] model assumes that each load is

characterized by a sequence of independent load events, with each event having an equal

duration. The service life of the structure is then divided into increments equal to the

duration of the load events. The occurrence rate of the load determines the probability of

the load occurring at an arbitrary time interval. If the load does occur, the probability

distribution of the load magnitudes is applied to the time interval. This process is carried

out for each load type experienced by the structure. The contributions from all the load

effects are combined to predict the maximum combined load that is likely to occur.

In this research project, the main effort is directed toward the development of a

resistance factor for composite materials rather than stochastic load models, as these

models have been already implemented in the ASCE7-10 (Minimum Design Loads for

Buildings and Other Structures standard) [6]. Structural engineers are well acquainted with

these minimum design loads.

12
Component Strength: Calculation of component strength is the most complicated section

in a composite probability of failure. The first contributions were in the form of

probabilistic strength of materials in aircraft applications. They used the β method (the

method with the establishment of target reliability) in order to laminate plates [59–60]. To

implement the reliability concept in civil engineering, the criteria for design strength for

composite sandwiches should be similar in appearance to the design strength for steel and

wood constructions in order to facilitate acceptance and implementation in the civil

structural engineering practice.

In order to ensure compatibility with civil engineering practice, a few authors

approached the probabilistic design at the structural civil engineering level, called Load

and Resistance Factor Design (LRFD). LRFD enables designs to be targeted for a desired

quantitative measure of reliability and minimizes the likelihood of overdesign or under

design [61–62]. The choice of target reliability level takes into account the possible

consequences of failure in terms of risk to life, injury, potential economic losses, and the

degree of societal inconvenience.

Depending on target reliability, the probability of failure can be calculated with this

formula:

𝛽 = −Ф−1 (𝑃𝑓 )

Ф( ) = Cumulative Distribution Function (CDF) of the random variable

probabilistic model

𝑃𝑓 = Failure probability corresponding to specified reference period

13
 In order to make the right choice for target reliability, the reference period,

consequences of failure, and cost of safety measures should be analyzed for the specific

case under consideration. It can be calculated with the limit state function, 𝑔(𝑅, 𝑆) [63].

A technically accurate description of the stress–strength curve overlap is presented

in Figure 3, which indicates the stress and strength along the horizontal and vertical axes

respectively. The line drawn represents the scenarios where stress = strength, or 𝑔(𝑅, 𝑆) =

𝑅 − 𝑆 = 0. This line separates the failure region (g < 0) from the safe region (g > 0). The

function 𝑔(𝑅, 𝑆) is commonly referred to as the limit state function. The probability of

failure is defined as the volume under the surface shown in the failure region where g < 0.

Figure 3. Three-dimensional presentation of the problem [55]

The essence of the structural reliability problem is the probability integral:

14
 where X = {x1,x2,.., xn}T is a vector of random variables that represent uncertain

quantities influencing the state of the structure, fx(X) is the probability density function

(PDF), and g(X) ≤ 0 denotes a subset of the outcome space, where a failure occurs [64].

For mathematical analysis, it is necessary to describe the failure domain g(X) < 0 in an

analytical form, which is widely referred to as the limit state function (LSF).

Due to the difficulty in directly computing this integral, many methods exist for

resolving it approximately [65–66].

In 2014, Jawaheri Zadeh and Nanni [67] introduced a method to bypass the loading

variables and weigh the resistance of two structural elements with the same ultimate limit

state directly against each other. They employed the concept of reliability index to develop

comparative reliability. They compared the flexural and shear strength reduction factor for

fiber reinforced polymer (FRP) reinforced concrete, with conventional steel-reinforced

concrete beams, and the results were the same as available reduction factors.

2-2 Raw Materials Composition and Structure

Balsa Wood Structure

Balsa wood end-grain has significant advantages for infrastructure application,

especially as core materials in sandwich composites. This is due to the relatively high

stiffness to weight and strength to weight ratio [10]. An example of using balsa wood as a

core material is the 56m span Bascule FRP Composite Footbridge – Fredrikstad. The

bridge deck is a sandwich construction with carbon fiber (CFRP) laminates and a balsa

wood core, with embedded heating cables for defrosting during the wintertime. The deck

15
is strong enough to carry a car with up to 2.0–ton axle load. It is the largest moveable bridge

in Scandinavia, which utilizes FRP composites as the main load carrying materials [11].

Balsa wood structure is composed of several different cell orientations, mostly

called tracheids. About 80–90% of the volume is arranged longitudinally; parenchyma are

arranged radially and constitute about 8–15% of the balsa wood volume and sap channels

[12].

Figure 4. Bascule FRP composite footbridge: Fredrikstad, Norway [11]

Tracheids have an irregular hexagonal cross section, consisting of long tubes. The

materials of these structures chiefly are cellulose, hemicelluloses, and lignin, which

together form the elementary fibrils [13]. The latter, aggregated into larger units by

hydrogen bonds, constitute the microfibrils that are the basic elements of the cell wall

layers. The tracheid cells consist of the primary wall (P) and secondary wall (S) layers, as

presented in Figure 5. The secondary wall is further sub-divided into the S1, S2, and S3

layers. The S2 layer (of approximately 2 lm thickness) accounts for roughly 85% of the

16
secondary wall thickness. The parenchyma rays penetrate the tracheids radially and are

responsible for misalignments of the latter along the natural axis of the tree [14]. Sap

channels, responsible for fluid transport in the tree, have thinner cell walls and are

relatively larger in diameter (150–250 lm compared to 30–40 lm for tracheids). Variation

of density occurs across the trunk due to the different growth of the early/spring wood and

late/summer wood during each annual ring [15].

Figure 5. Simplified microstructure of balsa wood, showing middle lamella and tracheid

cells with P, S1, S2, and S3 cell wall layers [16]

Balsa wood is a natural product. This indicates that it cannot be controlled in the

manner of industrial materials. Therefore, there are no accurate material properties and

often a single value is utilized for their properties [10]. The density of balsa wood sheets is

usually used to categorize its mechanical properties. Balsa panels are not directly produced

17
from the trunk of the tree. Their production involves joining the small cubes of balsa

together with adhesive. Their fiber direction is perpendicular to panel plate, as presented

in Figure 6.

Figure 6. Panel composed of balsa blocks joined by PVA base adhesive [19]

The density is an average of assembled end-grain blocks with few centimeters size,

each of which may have several different densities. This can induce significant

uncertainties in its properties. [17–19]. Quantifying the uncertainty properties and their

influence on the safety of structures can prove to be a solution for ensuring the quality and

reliability of balsa core sandwich composites.

PVC Closed Cell Foam Structure

According to Gibson and Ashby [20], a foam material consists of a cellular

structure that has interconnected small solid struts and or plates, forming an open or closed

cell foam. The bone of humans and animals are examples of foam structures in nature.

18
Styrofoam is an example of artificial foam used for insulation. Additionally, foam panels

are commonly used in surfboards and car bumpers. Structural application of foams as core

materials in wind blades and in bridges has created a considerable amount of demand for

them.

One of the main types of foam structure comes from PVC. Polyvinyl chloride

(PVC) polymer is a widely utilized amorphous thermoplastic that is produced by

polymerization of vinyl chloride monomers. The basic repeating unit of the PVC polymer

chain is presented in Figure 7.

Figure 7. PVC repeat unit

The PVC polymer contains 57% chlorine (Cl) by weight. The presence of the Cl

atom causes an increase in the interchain attraction and, in turn, increases the hardness and

stiffness of the polymer. Moreover, the combustion of PVC produces dangerous fumes,

which can be fatal on incineration [21]. Initially, the usage of PVC was restricted due to its

brittle nature. Additives, such as plasticizers, were introduced to the polymer, which made

the PVC more flexible and easier to be processed [22]. PVC has a softening point at about

80 ⁰C and is resistant to liquids such as salt water and antifreeze mixtures, which makes it

suitable for domestic water piping [23].

19
 Polymer foams are classified as cellular structures that can either be open or closed

[24]. Open-cell foams have the simplest structure, consisting of beam-like elements, which

define each cell that provides an open grid-like structure (refer to Figure 8a). Closed-cell

foams contain a combination of the beam-like structure of open-cell foams and membranes

that close off the open sections of the cell (Figure 8b). The fraction of polymer in the beam

elements and in the membrane is considered to substantially influence the stiffness of the

foam. For polymer foams, most of the solid plastic is located on the edges.

Figure 8. Schematic showing an approximately cellular microstructure for

mechanical modeling: a) open-cell foam, b) closed-cell foam [24]

PVC foams are widely used in sandwich structures, varying from pure insulation

applications to structural core materials used in marine vessels, aerospace structures, and

wind turbine blades. PVC foams dominate the market in terms of polymer foam core

material.

20
 In order to produce the PVC foam, a PVC plastisol consisting of isocyanates (for

crosslinking the molecular PVC chains), a blowing agent (for initiating the foaming

process), and a stabilizer are mixed together at temperatures below 100 ⁰C. The mixture is

then placed into the water, which reacts with the isocyanates in order to initiate the cell

nucleation and expansion. The foam is then allowed to cure in a mold in order to form its

final rigid structure. Since the cross-linking and foaming processes occur simultaneously,

properties of solid cross-linked PVC are not available [25].

Closed-cell foams are made up of polarized molecules that are otherwise known as

dipoles, and have a very low moisture absorption [27]. They have an interpenetrating

polymer network (IPN), a polymer comprising two or more networks that are at least

partially interlaced on a polymer scale but not covalently bonded to each other. The

network cannot be separated unless the chemical bonds are broken [26]. The two or more

networks can be considered to be entangled in such a way that they are concatenated and

cannot be pulled apart, but not bonded to each other by any chemical bond. The mixture of

polyvinyl chloride and polyurea has a good bond strength. Closed-cell PVC foam forms a

solid shape due to its linear structure. However, due to this structure, it is more brittle than

the open-cell PVC foam.

21
2-3 Composite Durability

Glass Fiber Laminate Durability

Glass Fiber Reinforced Plastic (GFRP) composites are extensively being used in

civil engineering applications, namely, pedestrian bridges, piling bumpers, and offshore

and marine structures. This GFRP is fabricated through the Vacuum Assisted Resin

Transfer Molding (VARTM) process, and it is commonly comprised of low cost woven

roving (WR) E-glass fibers infused with a vinyl ester resin. Their end use applications

require continuous exposure to thermal cycling conditions, and their in-service

environment is generally combined with the aqueous environment or seawater exposure.

In cold regions, ice chemicals and rock salts are frequently used on highway and pedestrian

bridges for ice and snow removal. In the last two decades, several researchers have

investigated the F/T cycling effects on composite materials [42–47]. Significant reduction

in mechanical properties was reported when thermal cycling occurred in an aqueous

environment [42–52]. When F/T is combined with salt water, two mechanisms are most

likely to occur. The first one is related to matrix cracking under thermal cycling, and the

second to diffusion through the resin down to the fibers. Resin degradation, fiber-matrix

debonding, and fiber stress corrosion were found to be the major failure mechanisms in

GFRP when immersed in sea water [53]. Experimental data available in the literature

related to WR GFRP under F/T exposure reported the existence of limited mechanical tests.

This study provides a full spectrum of mechanical testing, progressively backed by several

comprehensive non-destructive inspection techniques in order to better understand damage

initiation in a similar construction. Based on a recent literature survey conducted by Sousa

et al. [49], it was felt that a detailed root cause failure analysis is required to assist the

22
development of a mechanistic model on the basis of experiment related to the material

degradation in the observed loss of stiffness and strength. Understanding the cause, nature,

and the way damage is induced by thermal cycling, one would be able to quantify the

reduction and properly predict the structural performance of the WR GFRP construction.

If such root cause failure analysis is properly investigated, specifically by locating the size

and orientation of the defects, then a valid multi-scale durability and damage tolerance

model can be generated and implemented in any finite element software package in order

to predict the material state of composite structures under typical service environments.

This experimental research is an attempt to provide durability data and a

degradation model of sandwich composites by simulating the real environmental exposure

of infrastructure. Additionally, due to fact that the performance of composite structures is

largely influenced by quality control procedures, operating conditions, and environmental

effects, the use of a deterministic approach (that is, the current approach) to design and

analyze composite structures may undermine their high utilization and optimization

potential, which results in higher safety factors. Use of probabilistic design to predict and

quantify the structural strength and performance of certain FRP composites, particularly

sandwich composites, can help design engineers and also act as a reference to guide those

applying this technology.

Sandwich Composite Durability

Composite sandwich structures are currently being used in infrastructure

applications such as seawalls, piles, bridge decks, tanks, wind turbine blades, and so on.

These structures are constructed of lightweight sandwich constructions with face

23
sheets/skins, separated by a lightweight core material infused through the low-cost Vacuum

Assisted Resin Transfer Molding (VARTM) process. The face sheets are commonly

composed of woven roving or knitted e-glass reinforced vinylester composites. The core is

constructed of closed cell polymeric foams or fiber reinforced core materials. End-grain

wood balsa core is still being used in the spar caps and shear webs of offshore wind turbine

blades and other related transportation applications. This type of construction is in demand

in applications where a high strength and stiffness to weight ratio is an essential feature of

the design. In the northern region of the U.S. and Canada, these structures are often treated

by de-icing agents for the purpose of removing snow and ice. These agents contain a

significant amount of saline concentration, which is transported to the composite structure

by direct splashing or air-borne debris. The Freeze-thaw (F/T) cycling effect that results

from these agents will be discussed in this project.

In a similar manner as any structural material, fiber reinforced composite materials

are susceptible to degradation due to severe environmental exposure. Effect of F/T and

thermal cycling on FRP composites were investigated by several researchers [28–29].

Hygrothermal and thermal cycling effects were found to cause damage such as resin

cracking and fiber–matrix interface debonding. Glass–transition temperature and

mechanical properties of the FRP materials dropped significantly when the F/T exposure

was combined with the salt water environment [30-32]. Pitting and cracking of glass fibers

were also reported [29].

A few studies have investigated sandwich construction under the F/T environment.

Zinno et al. [33] studied the effect of F/T thermal cycling on phenolic-impregnated

honeycomb sandwich with glass fiber skin. The authors reported less than 15% reduction

24
in the elastic modulus of the e-glass/face sheets and 20% reduction in the core shear

ultimate strength. Other researchers studied the effect of seawater environment and

mechanical degradation of foam materials (PVC and polyurethane) [32, 34–35]. The

authors reported an increase in the fracture toughness of the foam after exposure. Avilés

and Aguilar‐Montero [34] discovered that face sheets greatly prevent water ingress into the

sandwich structure; however, when moisture diffuses due to the concentration gradient, the

speed of diffusion is accelerated in the foam, along the thickness of the core [34]. This

effect could plasticize the resin and cause cracking. These cracks will weaken the bond at

the interface layer between the face sheet and the core. Consequently, this could affect the

core shear strength and modulus, that is, the most critical mechanical properties of

sandwich structures.

The fracture toughness of composite sandwich structures in seawater was also

investigated [36–41]. Cantwell et al. [37] claimed that fracture toughness increased after

seawater exposure for balsa wood core sandwiches due to fiber bridging and greater energy

absorption. Singh and Davidson [40] observed that low temperature increased stiffness,

shear strength, and fatigue life of PVC foam core in sandwiches; however, the seawater

effect was minimal and influenced the static failure mode. Recently, Siriruk et. al. [41]

revealed that seawater alone reduced the interfacial toughness between the foam and the

facing; however, the combined effects of freezing may be negligible.

25
 CHAPTER 3

THEORETICAL APPROACH

The foundation of probabilistic design involves basing the design criteria on

reliability targets instead of deterministic criteria. In the application of design-based

parameters, parameters such as applied loads, material strength, and operational are

researched and/or measured, and subsequently statistically defined. A probabilistic

analysis model is developed for the entire system and solutions performed to yield failure

probabilities.

Based on the obtained experimental data, the following processes will be assumed

in this research:

1) The identification of potential failure modes of the structure under

anticipated loading conditions: the failure mode can change from location to location with

regard to the structure, with some areas being subject to more than one failure mode

2) The establishment of an acceptable probability of failure (or minimum

reliability) for each failure mode and/or location on the structure, based on available code

designs for civil engineering applications.

26
 3) The application of existing structural analysis methods utilized to model the

internal response (stresses) of the structure on the applied loads and to model the stiffness

of the structure.

4) Statistically, quantify defined random design variables that affect both the

stress and strength of the structure.

5) The application of probabilistic analysis methods to determine the

probability of failure at predetermined locations. There are four available methods for this

purpose:

 Single and two-variable failure probability determination

 Limit state approximation

These methods will be examined; and the one that produces the most accurate

results will be reported as the appropriate method.

Here, we utilize the model system probability of failure as a function of individual

location failure probabilities: once failure probabilities of individual locations have been

calculated, the final step involves the determination of an overall probability of failure for

the concerned structural component.

3-1 Selection of Probability Distribution

The mathematical modeling and representation of a random variable constitutes a

primary task for any probabilistic formulation that requires to be conducted systematically.

The first step of this process entails data collection; this further necessitates having an

appropriate data population. For this process, understanding, mean, variance, and skewness

can constitute descriptions of the randomness of a sample population.

27
 Normal or Gaussian distribution, lognormal random variable, and Weibull random

distribution constitute commonly employed continuous random variables. There are

several methods to determine the distribution and parameters from the observed data, such

as probability papers and statistical tests such as chi-square test and KS test. Method of

moment and method of maximum likelihood constitute the most frequently employed

methods for the estimation of the parameters of a distribution [54].

The cumulative probability distribution function (CDF) selected to represent the

properties of material contributes a significant effect on the calculated reliability of a

structure. If the right probability distribution is not employed, the results can vary by over

an order of magnitude. The reason for this pertains to the tail type of different CDFs that

are commonly utilized to model the properties of materials [55].

To address the problem of tail sensitivity, several methods are available. For

applications pertaining to civil engineering, LRFD is developed to address this issue. For

respective conventional building materials, different distribution are employed. For

example, for steel, a lognormal distribution was utilized for materials properties [56]. For

concrete structures, MacGregor et al. suggested the application of a normal distribution

[57]. However, for engineering wood materials, a Weibull distribution with two parameters

was introduced to describe these materials’ properties [58]. The reference resistance of

wood-based materials for LRFD is also well described in ASTM D5457-15 [59].

For fiber reinforced plastics (FRP), two-parameter Weibull distribution was

employed in accordance to the composite materials handbook MIL17; in their calculation,

the Weibull distribution is first selected and if it was not rejected, it is utilized [60]. If it

was rejected, two other distributions (normal and lognormal) are examined.

28
 In this study, normal, lognormal, and two-parameter Weibull distribution were

examined; to test the goodness of fit, several methods exist. As mentioned earlier,

probability distributions are similar in the center and the sensitivity is concentrated in the

tails; the Anderson-Darling test is sensitive in respect to tails [61] and was employed to

test the goodness of fit in this study.

The Anderson-Darling statistics is defined in equation (4),

𝑛
1−2𝑖
An2 = ∑ [ {ln[𝐹0 (𝑥(𝑖) )] + ln[1 − 𝐹0 (𝑥(𝑛+1−𝑖) )]}] − 𝑛 Equation (4)
𝑖=1 𝑛

Where x (i) = ith-order statistics of the data set;

F0 (x) = CDF

N = sample size

For this goodness of fit method, the Observed Significance Level (OSL), which

constitutes a measure of fitness, is illustrated in Table 1.

Table 1. Probability distributions functions [59]

Commonly, a five percent significance level is utilized; therefore, the null

hypothesis is rejected if the OSL is less than 0.05.

29
 Next step would be applying probability plots. A probability plot was formulated

for each category to examine the distribution; further, the R2 results for each one are

offered.

3-2 Limit State Functions

The calculation of the reduction factor, Ф, was performed based on the target

reliability value of 3 as it exhibits a reasonable safety factor for the structure. The target

reliability should not be too small, because the structure will not be safe, or too big as the

cost of the structure will increase accordingly in this case. The limit state of shear strength

is illustrated in equation (5); the variables in this equation are as follows: load (p), width

(b), and sandwich thickness (d). Sallowable constitutes the maximum allowable stress for the

target reliability value of three.

𝑝
𝑔(𝑥) = (2𝑏𝑑) − 𝑆𝑎𝑙𝑙𝑜𝑤𝑎𝑏𝑙𝑒 = 0 )5(

The core shear modulus formula was calculated based on the following formula:

𝑈(𝑑−2𝑡)
𝐺= (6)
b(𝑑−𝑡)2

Where, t is the laminate thickness; U is transverse shear modulus and it can be calculated

based on the following equation:

𝑃(𝑆−𝐿)
𝑈= P(2𝑆3 −3𝑆𝐿2 +𝐿3 )
(7)
4[𝛥− ]
96𝐷

Where, S = support span, L = load span; Δ is beam mid-span deflection; D signifies

the bending stiffness; D signifies dependence on Ef , facing modulus; C = core thickness;

the following equation was used to calculate D:

30
 𝐸𝑓 (𝑑3 −𝑐 3 )𝑏
D= (8)
12

The average amount of 3.5 × 106 psi was used for Ef, the facing modulus of glass

fiber reinforced plastic.

The limit state equation used for the shear modulus is depicted in equation (9); G

allowable is the maximum allowable shear modulus for the target reliability of three.

𝑈(𝑑−2𝑡)
𝑔(𝑥) = − 𝐺𝑎𝑙𝑙𝑜𝑤𝑎𝑏𝑙𝑒 = 0 (9)
b(𝑑−𝑡)2

For compression strength calculation, the following limit state equation was used,

where a and b signify the length and width of the sample respectively:

𝑝
𝑔(𝑥) = 𝑎.𝑏 − 𝑠𝑎𝑙𝑙𝑜𝑎𝑏𝑙𝑒 = 0 (10)

Limit state equation for compression modulus is calculated based on the

following equation:

𝑓𝑐 ((𝑝0.003 −𝑝0.001 ) .𝑡)

𝐸𝑧 = ((𝛥0.003−0.001 ) .𝐴)
(11)

Where, p and Δ are applied load and recorded deflection for 0.001 and 0.003

deflections. A is the area of the sample. Figure 9 shows the method to calculate the chord

slope.

31
 Figure 9. Chord slope calculation method

The limit state equation to calculate the compression shear modulus is based on

the following equation:

𝑓𝑐 ((𝑝0.003 −𝑝0.001 ).𝑡)

𝑔(𝑥) = 𝐸𝑧 = ((𝛥0.003−0.001 ).𝐴)
− 𝐺𝑎𝑙𝑙𝑜𝑤𝑎𝑏𝑙𝑒 = 0 (12)

To simulate a design similar to civil engineering practice, the LRFD approach is

necessary. LRFD enables designs to be targeted on a desired quantitative measure of

reliability and minimize the probability of overdesigning or under designing [65–72], and

the approach should be similar in appearance to the design of wood and steel to be accepted

and implemented in the civil engineering industry. It is specified as Ф𝑅𝑛 ; it constitutes the

product of the resistance factor, Ф and nominal strength, Rn.

The nominal strength depended on several factors and is defined based on the

following formula, given in equation (13):

𝑅𝑛 = 𝐶1 𝐶2 … 𝐶𝑘 𝑅0 (13)

32
 Ro is determined based on a laboratory test, mostly from standards such as ASTM

or other international standards test methods. It is based on the 5th percentile of the

distribution of strength. Ci factors account for environmental or geometry such as moisture,

temperature, freeze-thaw, size effect, or similar effects. [63]

To calculate the resistance factor, an approximation was introduced by

Ellingwood et al. [63]. It is shown in equation (14):

µ
Ф = (𝑅𝑅 ) exp[−𝛼𝑅 𝛽𝑉𝑅 ] (14)
𝑛

In this equation, µR signifies the mean of R (in the end-use condition), VR the

covariance of R, 𝛼𝑅 ; the sensitivity coefficient depends on the PDF of R, and is typically

equal to 0.6, while 𝛽 constitutes the reliability index. For the calculation of Rn, several

methods related to different kinds of distribution are viable. Further, the equation Ф.Rn =

Sallowable is extremely useful for the calculation of the resistance factor.

Using a target reliability value of three, the allowable load was determined by the

First Order Second Moment (FOSM) method to calculate Rn. For Weibull distribution,

ASTMD5457 introduced a formula that is very useful, which is shown in equation 15:

Rn = RP × Ω ×KR (15)

Rp is termed as the distribution percentile estimate; equation 16 is utilized to

calculate this factor, and it is dependent on Weibull distribution parameters:

RP= ƞ. [-ln (1-P)] 1/α (16)

Where,

ƞ = Weibull scale parameter,

33
 p = percentile of interest expressed as a decimal (for example, 0.05)

α = Weibull shape parameter.

The shape (α) and scale (ƞ) parameters of the two-parameter Weibull distribution

can be calculated for shear and compression strength. Parameters’ estimation can be done

through gamma function, but in this research, parameters were calculated through Minitab

software.

The data confidence factor, Ω, accounts for uncertainty associated with data sets.

This factor, which is a function of the coefficient of variation, sample size, and reference

percentile, is applied as a multiplier on the distribution estimate. There is an available table

that provides data confidence factors appropriate for lower fifth-percentile estimates [60].

The reliability normalization factor, KR, is employed to adjust the distribution

estimate (for example, R0.05) to achieve a target reliability index (β). The reliability

normalization factor comprises the ratio of the computed resistance factor to the specified

resistance factor, adjusted by a scaling factor. These data sets are available in standard

ASTM D5457 [60].

34
 CHAPTER 4

EXPERIMENTAL APPROACH

4-1 Durability Study of Sandwich Composite and Laminate

To study the effect of environmental conditions on sandwich composites, two types

of common core materials utilized in infrastructure applications were selected and molded

with E glass fiber as the face sheet.

Panels and Laminates Preparation

The cost-effective Vacuum Assisted Resin Transfer Molding (VARTM) process

was applied to mold all sandwich composite panels. This process is generally employed

for large civil engineering structural applications. The materials selected to manufacture

the sandwich composite panels are as follows:

35
 Core materials

• End Grain Balsa SB100 (density 150 kg/m3), 25.4 mm thick

• Airex C70.55 PVC foam (density 60 kg/m3), 25.4 mm thick

Face sheets

• Fabric: E-glass plain weave fabric (813 g/m2)

• Resin: Vinylester Resin (Derakane 610C)

The typical properties of composite constituents are illustrated in table 2. Further,

in figure 10, a panel during VARTM process has been depicted.

Table 2. Properties of composite constituents

Property Glass fiber Vinyl ester Balsa wood PVC foam

Tensile strength (MPa) 13500 71 13 1.3

Tensile modulus (GPa) 80 3.53 3.51 0.045

Elongation (%) 4.6 4.5 - 16 (shear)

Density (g/cc) 2.62 1.07 0.151 0.06

Thickness (mm) 0.54 - 25.4 25.4

36
 Figure 10. VARTM process of molding the panels

The materials employed for face sheet consisted of six layers of E-glass plain weave

fabric (813 g/m2) reinforced with Derakane 610C-200 epoxy-vinyl ester resin (Ashland

Inc.). This resin is currently used for several infrastructure applications due to its durability,

toughness, and fire resistant properties. It is comprised of a liquid blend of vinyl ester

monomers in styrene, in which the styrene content is approximately 40–50 wt%. Although

this type of thermosetting polymer resin is often referred to in the industry as an “epoxy

vinyl ester” resin, it does not actually contain any reactive epoxy groups, nor does it cure

by the same polymerization mechanism as most epoxy resin systems. It is synthesized

through the reaction of epoxy monomers with methacrylic acid that replaces all the epoxide

groups with methacrylate groups. This reaction leads to the formation of a carbon-carbon

double bond (referred to as vinyl) on the ends of each monomer that is attached to the

central monomer through an ester linkage, thus the term “vinyl-ester”. The main purpose

is to facilitate the initiation of the polymerization at room temperature with the utilization

of peroxide-based initiator systems as a result of peroxide’s reaction with the carbon-

37
carbon double bond. However, the presence of the ester group further facilitates the

increased possibility of hydrolysis decomposition in comparison to epoxy resin systems

that do not contain the ester linkage.

The initiator system utilized was Cadox 50 at a level of an amount of 1.25 wt%.

This material primarily comprises a solution of methyl ethyl ketone peroxide (MEKP).

Organic peroxide was employed as a catalyzer (Cadox 50).

These molded cross-ply laminates were allowed to cure for 30 days under ambient

conditions (23 °C and 50% RH). Nominal properties of composite constituents are listed

in Table 1. The fiber volume fraction was approximately 53% with a total measured

laminate thickness of 3.24 ± 0.1 mm.

The measured thickness of each face sheet was 3.24 ± 0.1 mm, with a 53% fiber

volume fraction. For compression and facing cleavage tests, 3 plies of E-glass fabrics

constituted the face sheets of the sandwich panels. The facing cleavage test standard ASTM

E2004 was selected in this study, since it targets infrastructural applications, such as

durable rigid walls and relocatable shelter wall structures.

4-2 Environmental Experiment Setup

In the presence of a saline water solution (3% salt), 200 cycles of freeze/thaw (F/T)

(– 20 °C to 20 °C) cycles were conducted on all samples (Figure 10). The F/T slab tester

by Qualitest TM was utilized to expose all samples. To facilitate the initiation of the process

of degradation in the core material, all cut specimens were placed in stainless steel pans

inside the test chamber and submerged halfway along the thickness (for laminates was full

immersion). This immersion approach serves to simulate the long-term exposure and saline

38
water diffusion in the core material. Figure 11 depicts the 12-hour period of F/T cycles

employed in this study. The total aging process for the exposed samples lasted for 100

days. Figure 12 exhibits the samples placed inside the F/T machine.

Figure 11. A cycle pattern for F/T thermal cycling

Figure 12. Left: F/T machine with samples inside; Right: flexural samples (top)

and compression sample (bottom)

39
4-3 Test Setup

Post the process of curing, samples were cut and extracted from panels for the

application of different test methods based on a designed test matrix as illustrated in

Table 3.

Table 3. Number of samples and dimensions for sandwich composites

Dimensions No. of samples with No. of samples with

Evaluation Test Length, width, and balsa core foam core
thickness (mm) exposed/unexposed exposed/unexposed
Sandwich beam
flexural and shear
stiffness (ASTM
406.3 × 76.2 × 32 30/30 30/30
C393, ASTM
D7250)

Core Shear Fatigue 406.3 × 76.2 × 32 9/9 9/9

Test

Balsa Core Flatwise

Compression
Strength and
Compression 76.2 × 76.2 × 32 30/30 30/30
Modulus (ASTM C
365)

Standard Test
Method for Facing
Trapezoidal
Cleavage of 6/6 6/6
(see Figure 3)
Sandwich Panels
(ASTM E2004)
Note: Exposed = 200 cycles, unexposed = 0 cycle

In order to examine the skin properties of sandwiches, laminates were prepared

according to dimensions in Table 4.

40
 Table 4. Number of samples and dimension for GFRP laminates

Evaluation test Dimension (mm) samples

Shear V-Notched Beam (ASTM D5379-12)

76.5 × 19 × 3.24 12
IOSIPESCU

Laminate Tensile Test (ASTM D3039) 254 × 25.4 × 3.24 12

139.7 × 6.35 × 12
Laminate Compression Test (ASTM D6641)
3.24

A four-point bending test fixture illustrated in Figure 13 was employed for the

purpose of this investigation. This fixture restrains any longitudinal rocking motion, while

it serves to fatigue the sandwich samples. All specimens were 406 mm long, 76.2 mm wide,

and 32.7 mm thick, including the core thickness (25.4 mm). Preliminary static testing was

conducted at 0 cycle (ambient specimens) and post 200 cycles (exposed) in accordance to

ASTM C393 [13] and D7250 [14]. The total number of samples is provided in Table 2.

The fatigue testing was performed for R = – 0.1 (R= max. load/min. load), compression-

compression fatigue was performed at a testing frequency of 4 Hz. Since this research

focusses on infrastructural applications, stiffness reduction is critical, a failure protocol

constituted a 30 % reduction in stiffness (30% increase in stroke) at the constant load

amplitude regime (Table 5 and 6).

41
 Figure 13. Four-point bending fixture for fatigue testing

Table 5. Core shear fatigue matrix: balsa wood core

Test Type Load Max. load (N) Min. load (N) Mean (N) Amplitude (N)

Static Pavg 8896 N/A N/A N/A

Fatigue 0.75 Pavg 6672 667 3670 3002

Fatigue 0.6 Pavg 5782 578 3180 2602

Fatigue 0.5 Pavg 5337 533 2935 2402

Table 6. Core shear fatigue matrix: Airex C70.55 PVC foam

Test Type Load Max Load (N) Min Load (N) Mean (N) Amplitude (N)

Static Pavg 4381 N/A N/A N/A

Fatigue 0.75 Pavg 3287 328 1810 1479

Fatigue 0.6 Pavg 2629 262 1446 1184

Fatigue 0.5 Pavg 2188 218 1203 985

42
 The compression test was conducted with an MTS machine; Figure 14 depicts a

sample that comprises three layers of laminate on each side under the compression test.

Figure 14. A sample under compression test

The peel test setup was accomplished according to ASTM E2004, for both balsa

and a foam core specimen; Figure 15 displays the setup condition for this test. To prepare

the samples, two pieces of steel were bonded with an adhesive (LORD 7075) as depicted

in Figure 14a: one side of the sample interface of core/laminate was cut with a saw as

shown in Figure 14b. In Figure 15c, a sample assembled on MTS machine has been

illustrated. The cut portion is always placed on the upper side.

43
 a)

b) c)

Figure 15. Peel test a) steel T-shape tabs bonded to the specimen;

b) specimen with interface cut area; c) specimen ready for the test

GFRP Laminate Testing:

The following experiments was applied to predict the mechanical properties of

exposed and unexposed laminate specimens. Two strain gages were placed in the center of

44
the Iosipescu specimens (ASTM D5379-12) to achieve strain values of + 45 and – 45

degrees; the final shear strain was calculated through the average of the absolute values of

individual strain gage results. The shear strain of 0.05% to 0.15% was utilized to calculate

the shear modulus. Figure 16 illustrates a coupon under the test comprising the strain

gauges.

Figure 16. Iosipescu test setup

For tensile (ASTM D3039) and compression (ASTM D6641) tests, gages were

installed in the center of the coupon, longitudinal and transverse direction strain values

were calculated with regard to the tensile test. For the compression test, gages were

installed on each side of the specimens to detect any possible buckling while testing. Shear

strain values between 0.05% to 0.25% were utilized for both tests to calculate the shear

modulus. Figure 17 depicts the tensile and compression test setups respectively.

45
 a) b)

Figure 17. a) Tension test setup; b) compression test setup

4-4 Characterization Methods

To study the effect of F/T on the microstructure and physical properties of polymer

and composite, the following characterization experiments were applied prior to and

subsequent to the provision of environmental exposure; the results of analyses were also

compared:

1) Thermomechanical Analysis (TMA)

TMA was employed to measure the following:

46
  Glass transition temperature

 Residual curing

 Coefficient of thermal expansion

Operating parameters for the test were maintained at 25–130 °C, 5 °C/min, cool,

re-reheat; the TMA model was constituted by TA Instrument Q400; the sample size was

determined as 10 x 10 mm.

2) Dynamic Mechanical Analysis (DMA)

This technique was employed to calculate the viscoelastic properties of polymers.

In this technique, a sinusoidal stress is applied and the strain in the material is measured

that allows the determination of the complex modulus. The temperature of the sample or

the frequency of the stress is often varied, contributing to variations in the complex

modulus. This approach can be utilized to locate the following:

 Glass transition temperature

 Storage modulus (E’)

 Loss modulus (E”)

The temperature underwent a change from 25 °C to 180 °C, and a 1 Hz frequency

was performed. The DMA unit was constituted by a TA instrument Q800. This test was

applied to the balsa wood and foam core materials for core type materials. Two types of

fixture, 3-point bending and shear were utilized; for laminates, merely a 3-point bending

fixture was utilized. The samples were of the following dimensions: 10 mm x 60 mm; the

thickness was 5 mm. Figure 18 illustrates the fixtures.

47
 a) b)

Figure 18. DMA fixtures, a) 3-point bending, b) shear

3) Ultrasonic Pulse Reflector

The ultrasonic reflector plate with the transducer frequency as 5 MHz, scan/index

length as 15 inches, scan/index increment value 0.02 inches were applied to ascertain,

whether the samples test pieces were free of discontinuities or defects. Figure 19 depicts a

specimen during the performance of the ultrasonic scanning tests in the water tank.

Figure 19. Specimens inside the water tank and an ultrasonic scanning probe

4) X-Ray Computed Tomography (CT) Scanning

48
 This technology possesses the ability to exact three-dimensional cross-sectional

images of the entire part. To perform the X-ray CT-scan Inspection, HMXST high-

performance Real-Time X-ray, X-Tek real-time HMX160 was applied to obtain images

from different cross sections of the specimens. Several samples of sandwich composites

and laminates prior to and subsequent to the environmental exposure were tested.

5) Scanning Electron Microscopy (SEM)

A Hitachi S-4800 High-Resolution Scanning Electron Microscope (HRSEM) was

employed to examine the microstructure, defects, and any kind of degradation on the

surface of samples.

6) Optical Microscopy

The DSRL and the Keyence Optical Microscopy was utilized to observe phase

alterations and the microstructure of polymers and fiber.

49
 CHAPTER 5

RESULTS AND DISCUSSION

All exposed sandwich samples reported in the following sections were removed

from the environmental F/T chamber and allowed to dry for 72 hours before testing.

Following shows the results:

5-1 Balsa Wood Core Sandwich Composites Properties

The areal weight of balsa wood composite sandwiches shows in figure 20, the
samples reached to saturation point in around 150 cycles of F/T.

12
10
8
Kg/m2

6
4
2
0
0 50 100 150 200
Cycles

Figure 20. Areal weight of balsa wood core sandwich after F/T exposure

50
 Figure 21 shows a resin molded balsa wood, in 50 and 200 cycles of F/T, in 200

cycles lot of cracks has been expanded on the surface and inside of balsa wood.

a) b)

Figure 21. Molded balsa wood after a) 50 cycles, b) 200 cycles

Figure 22 below shows the balsa core shear strength and modulus before and after

200 F/T cycles (total of 60 samples). Results are presented in bar charts with a 95%

confidence interval of the mean. Referring to Figure 4, a 13.9% reduction in shear strength

was observed and a coefficient of variation (COV) of 13% at 0 cycle and 21% at 200 cycles.

The shear modulus decreased by only 4% after exposure with larger COV in modulus of

20% at 0 cycles, and 41% at 200 cycles. An analysis of variance (ANOVA) revealed that

the reduction in core shear strength was statistically significant (P-value= 0.008), whereas

the core shear modulus was not statistically significant (P-value= 0.6).

51
Figure 22. Balsa core shear strength (left) and modulus (right) -ASTM C393 and D7250

The failure mode for unexposed and exposed specimens has shown in figure 23 and

24 respectively.

Figure 23. 0-cycle core shear failure mode

52
 Figure 24. 200-cycles core shear failure mode
The core shear failure modes for the unexposed samples formed a 45-degree angle

in the balsa wood core and along the end grain (cells). For the exposed samples, the

fibrillation mechanism of balsa cells was noticeable at the shear span of the four-point bend

test (Figure 24). This large fibrillation is due to the excessive absorption of moisture and

degradation due to F/T cycling. The thermal cycling decreased the shear strength of the

balsa wood core and slightly increased the ductility of the sandwich structure. Figure 25

presents the average trendline of exposed and unexposed specimens when tested according

to ASTM C393 [15].

Figure 25. Flexure results for balsa core sandwich samples

53
 As illustrated in Figure 25, the core shear stress vs. displacement plots show better

toughness for the exposed specimens (200 cycles). Figure 26 shows fatigue failure mode

for balsa wood. Three distinct crack events occurred while cycling the specimens, 1) crack

initiation and propagation just below the load point, 2) core shear along the wood cells

(with fibrillation for the exposed samples) and 3) bottom skin/core delamination.

Figure 26. Balsa wood fatigue failure mode

The simulated lifetime core shear fatigue is shown in Figure 27. It is interesting to

observe that the curve fit trendline of the exposed samples showed an increase in fatigue

life at high-cycle levels. This observation justifies the higher toughness presented in Figure

25. This toughness could be regarded as the energy dissipated due to friction in between

the fibrillated cells while cycling the specimens at 4 Hz. Similar trend was also observed

by other researchers [13] for tension-tension fatigue of wood perpendicular to the grains.

54
 Figure 27. Balsa wood core shear fatigue life prediction for exposed (200 cycles) and

unexposed specimens (0 cycles), R = -0.1 at 4 Hz

A nondestructive technique was sought to justify the increase in core shear fatigue

life of the balsa core. Figure 28 shows the optical microscopy images of two compressed

balsa wood sandwich samples at 0 cycle and 200 cycles. The fibrillation of the wood cell

structures is clearly visible after 200 cycles of F/T exposure (Figure 28 b). It seems that

under F/T cycling, water absorbed by the wood cells (unfilled with resin) through capillary

action caused expansion and contraction inside the cellular wall structure of the balsa. This

sequentially disintegrates the wood cells into several parts and indirectly created a longer

path for the crack to travel along the core while fatiguing the specimen.

Figure 28. Balsa wood texture a) unexposed b) exposed samples (optical microscopy)

55
 The results of compression strength on the sandwich specimens with balsa core

before and after F/T exposure is shown in Figure 29. A 36% reduction in compression

strength was observed with a COV of 21% at 0 cycle and 27% at 200 cycles. The

compression modulus decreased by 33% after exposure with a COV of 33% and 30%

respectively. The reduction in balsa core compression strength and modulus were

statistically significant.

Figure 29. Balsa core flatwise compression strength (left) and compression

modulus (Right) - ASTM C 365 preliminary testing at 0 and 200 cycles

Figure 30 and 31 show the facing cleavage test results in terms of fracture toughness

(J/m2) for the Balsa wood core samples. A total of 12 samples were tested, with 32%

increase occurred after exposure (200 cycles). The COVs for the unexposed and exposed

were 48% and 30%, respectively. The failure modes experienced in the Balsa core samples

were pure face sheet/core interface failure mode. Due to the large COV and data scatter,

an ANOVA was conducted. Even though the mean increased after exposure (200 cycles),

56
the result (P-value = 0.185) did not support that a strong evidence exists to conclude that

the fracture toughness means are different between the two exposures.

Figure 30. Load- stroke curve for exposed and unexposed specimens

Figure 31. Balsa wood fracture toughness for unexposed (0 cycles) and exposed (200

cycles) specimens (ASTM E2004)

57
 Figure 32 shows a sample under test and failure modes. Not so much different

happened between failure modes of exposed and unexposed specimens.

a)

b)

Figure 32. a) A specimen under peel test, b) failure modes of exposed and unexposed

specimens

58
5-2 Foam Core Sandwich Composite Properties

Figure 33 below shows the core shear strength and modulus before and after 200

F/T cycles for the PVC foam core (Airex C70.55) sandwich samples. The 95% confidence

interval shows 3.2% increase in core shear strength and approximately 30% increase in

core shear modulus. The core shear strength had a covariance of 3.8 % at 0 cycle and 4.8

% at 200 cycles. The shear modulus had a covariance of 7.3 % at 0 cycle and 5.5 % at 200

cycles after 200 cycles.

Figure 33. Foam core shear strength and shear modulus -ASTM C393 and D7250

A specimen after the shear test is shown in figure 34, pure core shear failure mode

was observed in exposed and unexposed (ambient) specimens. The failure occurred at 45-

degree angle in the PVC foam. Typical core shear strength vs. displacement plots (average

of 30 samples) are shown in Figure 35. The 200 cycles (exposed samples) plot clearly

59
indicates an increase in the core shear stiffness. Contrary to the balsa core, no increased in

toughness is detected after exposure.

Figure 34. Shear four-point test

Figure 35. Four-point flexure results for foam core sandwich samples

To better understand the increase in core shear strength and stiffness after exposure,

several nondestructive/destructive techniques were attempted to examine the

microstructure change and variation in the PVC foam network. Figure 36 shows the SEM

60
micrographs of foam core before and after exposure. Examining the foam microstructure

after exposure (Figure 36b), the network shows negligible dimensional change with few

cracks in between the cellular structures of the foam. The X-ray computed tomography

picture shown in Figure 36c quantify the resin thickness absorbed by the foam core. This

Figure shows the absorbed resin confined by the foam cells creating a “key” anchorage

zone that could potentially increase the core shear strength and improve the face sheet core

connection. During the cycling events, the resin absorbed by the foam was continuously in

the glassy state, this in turns restricts chain mobility and segmental motion inside the

polymer microstructure, which increases the modulus of the thermoset resin.

Figure 36. SEM: a) unexposed foam and b) exposed foam after 200 F/T cycles, X-Ray

CT-Scan: c) foam core sandwich specimen after F/T exposure

61
 The SEM and X-ray computed tomography results were not sufficient to clarify the

increase in stiffness of the PVC foam core. The dynamic mechanical analysis (DMA) in

shear sandwich mode was used next to assess the morphology and viscoelastic transition

response of the foam core. Figure 37 below shows the variation in the storage modulus and

loss modulus with temperature. It is clearly shown that the storage modulus of the foam

increased after 200 cycles of F/T exposure, which is consistent with the results of the

ASTM C393 and D7250 tests. This implies that the freeze/thaw exposure increased the

stiffness of the foam network. Interestingly, the loss modulus also increased from the

freeze/thaw exposure, yet there was no significant change in the glass transition

temperature (as taken from the peak in the E” curve). However, a standard DSC test

conducted on the foam showed that the Tg increased by 4°C after exposure, which could

be explained by a higher crosslink density after exposure. This combination of results is

not fully understood, but it does imply the freeze/thaw exposure produced a structural

change in the foam that could be from chemical and/or physical means. This result is not

an example of the anti-plasticization effect observed in some polymer-plasticizer systems,

which involves a decrease in Tg and increase in the glassy state storage modulus. In the

current study, the Tg remained the same, while both E’ and E” increased.

62
 Figure 37. DMA testing results, a) shear mode testing, b) 3-point bending mode

The failure mode of a fatigue test is shown in Figure 38. The failure mode for both

the exposed and unexposed specimen was same with a 45-degree crush. The fatigue plot

for the Airex C70.55 PVC foam is shown in Figure 37. During the fatigue test, three stages

of fatigue crack growth occurred in the foam core specimens. The fatigue life of the

exposed and unexposed (ambient) specimens is almost identical (Figure 39).

63
 Figure 38. Fatigue failure mode of PVC foam

Figure 39. Foam core shear fatigue life prediction for exposed (200 cycles) and

unexposed specimens (0 cycles), R = -0.1 at 4 Hz

The results of compression strength on foam core samples before and after F/T

exposure has shown in figure 40, 1.5% decrease in compression strength and 2% increase

in compression modulus were recorded after 200 cycles of F/T. Covariance percent for

64
unexposed specimens was 9.42% and 12% for strength and modulus respectively, this

amounts for exposed specimens was 6% and 11%.

Figure 40. Foam core flatwise compression strength and compression modulus

ASTM C 365 preliminary testing at 0 and 200 cycles

As the compression and shear strength of foam does not change so much, it means

the average there is no difference in the dissipated energy after F/T effect. These results

can confirm same damping properties and consequently same fatigue life. However, after

F/T, 27% increase in shear modulus happened, means after F/T lower fatigue life is

expected as shown in figure 38.

6 specimens of exposed and 6 unexposed were tested under peel test; the load-

stroke curve is shown in figure 41, in total 8.1% in energy density occurred after exposure

65
to freeze-thaw, the covariance percent for unexposed and exposed was 8.6 and 8.3 percent

respectively. (Figure 42)

Figure 41. Load-Stroke curve for exposed and unexposed specimens

Figure 42. PVC Foam fracture toughness for unexposed (0 cycles) and exposed (200

cycles) specimens (ASTM E2004)

66
 Figure 43a shows a specimen in middle of peel test, figure 43b shows the failure

modes of exposed and exposed specimen, no significant change was observed in these

specimens.

a)

b)

Figure 43. a) A specimen in middle of peel test, b) specimens after peel test

67
5-3 Laminate Tests Results

Table 7 shows the average mechanical test results and Coefficient of Variation

(COV) for each standard test (average of 12 samples per standard test). All mechanical

characterization tests were conducted using a calibrated test machine (INSTRON 4208).

The COV for strength and modulus for all specimens were less than 15.3 %.

Table 7. Mechanical test results

Average Average
Sample Modulus Strength
Test Direction Modulus Strength
Type COV% COV%
(GPa) (MPa)

Shear - 4.76 7.8 67.4 2.6

Tension Longitudinal 31.09 4.7 453.6 7.8

Unexposed
Tension Transverse 32.7 14.9 - -

Compression Longitudinal 31.11 5.5 424.6 3.3

Shear - 3.87 9.6 58.3 6.8

Tension Longitudinal 24.9 4.91 347.14 12.85

Exposed
Tension Transverse 30.2 15.3 - -

Compression Longitudinal 27.7 2 352.2 9.6

Normalized moduli and strengths for shear, tension, and compression are shown in

Figure 44. Large reductions occurred in the tensile and compressive strengths of the GFRP

material. The shear and tensile longitudinal moduli were also reduced after F/T exposure.

In general, the compression strength in WR fabric is lower than the tensile strength; this is

68
due to the already kinked fibers; however, this is not the case after exposure (Table 6, row

6 and 8). The tensile test results are mainly dominated by the fiber properties and clearly,

indicate that the longitudinal fibers and the structural integrity of the WR have been

endangered due to the environmental conditions. A mechanistic model, which describes

the mechanism of deterioration of the exposed laminates, is described in the root cause

failure analysis section.

Figure 44. Overall results of exposed (200 cycles) and unexposed (0 cycles)

specimens

Effect on Shear Properties

Effect of exposure on shear properties of the laminates was studied in accordance

with the v-notched beam test (ASTM D5379-12) [12]. Figure 45 shows typical shear failure

mode that occurred at the mid-span of the V-notched beam. The average reduction in shear

69
modulus and strength after 100 days of exposure is 18.5% and 13.5 %, respectively (Figure

46). Similar reductions were reported by Guzman and Brøndsted [13] after 8 years of

saltwater exposure.

Figure 45. Shear failure modes for a) unexposed; and b) exposed specimens

Figure 46. Percent change in shear modulus and strength

Following the shear test, failed samples were cut in between the V-notch locations

(across the length of the sample) and dissected using an optical image technique. Figure 46

shows the final fracture states and internal cracks for the unexposed and exposed

specimens. The unexposed failed specimens exhibited interlaminar cracks and separation

in between the resin and warp fibers. The cracks were oriented along the warp direction.

70
For the exposed failed samples, a high density of intralaminar cracks occurred along the

fill fibers as well as their interfaces (Figure 47). This clearly shows that the resin has been

degraded due to the F/T cycling.

Figure 47. Micrograph of the cross section after failure at the V-notch location a)

Unexposed: showing only interlaminar cracks, b) Exposed: showing interlaminar and

intralaminar cracks

Effect on Tensile Properties

In this test, a 23% reduction in the tensile strength and 19.7% in the longitudinal

modulus occurred after exposure (Figure 48). These values are significantly higher than

the COV listed in Table 11. Since the ASTM D3039 [14] test results are mainly dominated

by the fibers, this presumably indicates that the fibers were damaged in the longitudinal

and transverse directions of the WR construction. In general, when a delaminated

composite material is subjected to tension, the residual strength is reduced by only 10 to

15 % [15]. Lateral strains (90-degree direction) were used to monitor and extract the

stiffness of the laminate in the transverse direction. The transverse modulus decreased by

10 %. The failure modes are shown in figure 49. In addition, it is interesting to note that

71
Guzman and Brøndsted [13] reported similar reductions in transverse tensile modulus and

strength (single glass fiber) after 8 years in salt water environment.

Figure 48. Percent change in tensile modulus and strength

Figure 49. Tensile test failure modes; unexposed and exposed specimens
72
 Figure 50 shows the tensile stress-strain curves in the longitudinal and transverse

directions. For the exposed samples, the transverse strains are more pronounced than the

unexposed samples. These highly deviated curves designate some sort of stiffness

degradation due to a loss of the structural integrity of the laminate, especially at the weave

junctions (nodes) between the 0 and 90-degree fibers of the WR construction. This could

be a direct result of a weakened matrix that binds the warp fiber to the fill fiber. More in-

depth validation of this key failure mechanism is explained in the non-destructive testing

section.

Figure 50. Longitudinal and transverse stress-strain curves for exposed and unexposed

specimens

Effect on Compressive Properties

The combined loading compression test fixture was used to study the effect of

exposure on the compressive properties of the GFRP laminates. A total of 12 samples were

tested in accordance with ASTM D6641 [16]. The exposed samples showed an average

reduction of 17.1% in strength and 12% in modulus. One can notice the high dependency

73
of the matrix degradation and interfacial failure on the compression strength versus the

shear strength observed in Figure 46. Typically experienced failure modes are shown in

figure 51.

Figure 51. a) Percent change in modulus and strength, b) Compression failure modes

Root Cause Failure Analysis

Dynamic Mechanical Analysis (DMA) and Differential Scanning Calorimetry (DSC)

To study the internal damage or defects that occurred due to these environmental

conditions, a TMA and DMA tests were conducted on unexposed and exposed laminates.

The DMA results are given in Figure 52. Figure 52a shows the results obtained

from the 1st heating cycle (25-140°C @ 5 C/min). After cooling the sample in the

apparatus, a 2nd heating cycle was run on the same specimen. The reheat cycle results

(Figure 10B) show the condition of the material after being post cured. This analysis is

74
commonly run for thermosets because usually they are not fully cured in the original

manufacturing process.

Polymers generally are viscoelastic materials, exhibiting simultaneous elastic and

viscous behavior in response to time-varying stress. The amount of each type of behavior

depends on the state of the polymer structure (chemical composition, crosslink density)

and the temperature. The storage modulus (E’) characterizes the elastic energy stored by

the polymer, while the loss modulus (E”) represents the energy lost to viscous dissipation.

Polymers that are highly crosslinked usually have high values of E’. Also, the shape of the

E’ and E” curves can be used to locate the temperature at which the polymer structure

transitions from a glassy state (at low temperature) to a rubbery state (at higher

temperature).

Referring to Figure 52a, the sample subjected to 200 freeze-thaw cycles exhibited

a 20% decrease in E’ near room temperature. This could be due to a variety of reasons,

such as plasticization of the matrix, chemical decomposition of the matrix, physical

degradation of the fiber-matrix interface, etc. Chemical decomposition of the matrix could

be explained by hydrolysis of the ester linkage in the vinyl ester matrix, although hydrolysis

reactions usually slow at the temperatures encountered. Therefore, the change may be due

to physical degradation of the resin and/or fiber-matrix interface from the freeze-thaw

exposure.

The E” curves provide more information about the phase structure of the resin.

Cured vinyl ester resins usually are comprised of two polymers phases: 1) the main

crosslinked structure which is comprised of the vinyl ester-styrene copolymer network, and

75
2) a phase rich in homopolymerized linear polystyrene. The peaks in the E” curve

correspond to the temperature at which each phase transitions from glassy to rubbery state.

The glass transition of polystyrene is approximately100°C, which corresponds closely to

“phase 1” peak. The glass transition of the copolymer network is generally observed to be

higher: in this case, it occurs at 115°C and is labeled as “phase 2” in the 0-cycle sample.

It is interesting that the 200-cycle sample does not exhibit the phase 2 transition in the same

location. Most likely, the phase 2 transition was reduced to about 75°C (labeled as phase

2’, which is the shoulder in the E” curve, as well as the steep drop in E’). Absorbed water

would be expected to plasticize the main network because it has some polarity (e.g. ester

bonds), while water should have little or no impact on the highly nonpolar polystyrene

phase. Indeed, the phase 1 peak is almost identical in both samples.

Figure 52B shows the results after post-curing the samples. Any water in the matrix

is expected to be removed in the 1st heating cycle. The value of E’ for the 200-cycle sample

remained lower (17.5%) than the control sample, which indicates that there was some

nonreversible degradation caused by the freeze-thaw cycles. The E” peak of the

polystyrene phase (phase 1) was largely unaffected by the post-curing process in both 0

and 200 cycle samples. In the 0-cycle sample, the E” peak of the vinyl ester phase shifted

to a slightly higher temperature (120°C), which can be attributed to residual curing from

the 1st heating cycle. In the 200-cycle sample, the vinyl ester E” peak is seen as a faint

shoulder around the same temperature as the 0-cycle sample. This indicates that the vinyl

ester network was no longer plasticized, but was not as prevalent as the 0-cycle sample.

As with the E’ result, this also indicates some difference in the network or fiber-matrix

interface caused by the freeze-thaw cycle.

76
 b

Figure 52. DMA results for composite, a) 1st DMA heating cycle, b) 2nd DMA heating

cycle.

The DSC results for a neat resin samples are summarized in Table 8. This technique

was used to verify that the samples were not fully cured after VARTM processing, and to

measure the Tg of the resin independently of the presence of fibers. The Tg is strongly

related to the nature of the crosslinked network, and therefore it can provide information

about the decomposition of the polymer. The DSC traces for the 1st heat showed a Tg

77
around 67°C for the 0-cycle sample and 63°C for the 200-cycle sample. This difference

was attributed to plasticization of the matrix from the freeze-thaw cycle, as observed in

DMA. However, upon reheating the Tg increased to essentially the same value (96-97°C)

in both cases, due to post-curing experienced in the prior cycle. This implies that there was

no permanent chemical decomposition of the polymer, for example by hydrolysis. The

permanent damage detected by DMA (drop in storage modulus) is therefore attributed

primarily to physical degradation of the polymer (cracks) and/or between the polymer and

fiber. The small change to phase 2 detected by DMA is not able to be detected by DSC.

Table 8. Summarized Tg results from DSC.

Tg (°C)
# cycles
1st heat 2nd heat

68.0 96.3
0
66.5

62.4 97.4
200
64.3

Damage State in Woven Roving Architecture

To investigate the size and location of the internal defects, a 3D X-Ray CT-Scan

test was conducted on the exposed specimens. Figure 53 shows resin cracking at preferred

sites. In addition, intralaminar cracks occurred in the mid-section of the FRP laminate,

specifically in resin rich areas. The damage is shown in Figure 52 (transverse cross-section)

is approximately 1.2 mm in length and 0.7 mm thick. These defects (cracks and fractures)

78
are a source of delamination initiation that further degrades the component strength and

eventually leads to local stiffness reduction.

c)
a) b)

Figure 53. X-Ray computed tomography showing damage in exposed samples a) surface

plies, b) longitudinal cross-section, c) transverse cross section

To test the structural integrity of the connections between the fill and warp fibers,

a pulse ultrasonic C-scan technique was performed on two exposed and unexposed

specimens. Figure 54 shows the C-scan images of all tested samples. The grid pattern

(signals at nodes) is clearly shown in the unexposed samples (figure 53a), whereas the

linear pattern is clearly visible in the exposed samples. This change in pattern indicates that

these connections are impaired and ineffective in transferring the loads, including racking

or shear loads. In addition to the shear stiffness loss of the cracked matrix, the deterioration

of these nodes is believed to affect the tensile and shear moduli of the laminate. To interpret

this phenomenon, one could envision the laminate structure like a picture frame (hinged

truss) with some degree of rotational stiffness at the four pins (nodes). Two of these pins

79
could relate to a horizontal spring simulating the lateral stiffness supplied by the matrix. If

a load (tension of shear) is applied to the truss, the capability of the truss to resist load and

deformation (axial or racking) is controlled by the rotational stiffness supplied by the nodes

(bond at the junction of fill and warp fibers) and the axial stiffness of the spring. This

simple mechanistic model helps to understand the basic phenomena governing the behavior

of these systems.

a) b)

Figure 54. Ultrasonic C-scan results (102 mm wide specimens); a) unexposed: grid
pattern, b) exposed: linear pattern

Matrix and Fiber Damage

To map out the damage and frequency of defects that occurred in the laminate after

exposure, a high-resolution optical image technique was used to inspect their locations.

Figure 55 shows damaged fibers and matrix bond failure between the fill and warp fibers.

This failure pattern commonly occurred at the kink/weave location between the roving. It

seems that the volume changes due to the thermal cycling initiated high non-uniform

residual stress at the sharpest curvature location and significantly degraded this

node/junction between the fibers. The authors believe that this damage will be recurrent

for any plain weave type no matter what type of reinforcement is used.

80
 Figure 55. Local magnified optical image at weave junction showing matrix failure

between damaged warp fibers and fill fibers after exposure

Figure 56 shows the microstructure of an exposed specimen with a high-resolution

Scanning Electron Microscopy (SEM). A significant amount of fiber/matrix degradation is

observed. The mismatch in the coefficient of thermal expansion between the resin and the

E-glass fiber is believed to cause this interfacial failure mode. In addition, fiber fractures

are clearly shown in figure 55.

Figure 56. SEM micrograph showing degradation of fiber/matrix interface and

fiber fracture

81
 Karbahari et. al. reported that aqueous solution in contact with glass fiber surface

produced free-alkali hydroxide groups that degraded the silica structure of the glass fiber

when exposed to an alkaline media [11]. The effect is through breaking of Si-O bonds in

the glass network and was believed that surface loss and pitting occurred in areas of high

pH solution. In general, alkali solution such as cement environment is known to degrade

the glass fiber through several processes such as pitting, etching, leaching, and

embrittlement [17-19]. Figure 57 shows a surface of a filament of glass fiber in an exposed

sample. Several defects (black spots) are observed, including localized areas around these

spots which might be related to coupling agent or polymer swelling. Research on this

subject is critically needed to better understand this damage type.

Figure 57. SEM picture of damaged E-glass fiber after exposure

82
5-4 Model Results:

Tables 9 and 10 present the statistical values for each of the random variables

pertaining to shear and compression tests respectively.

Table 9. Statistical values for balsa and foam core shear test of sandwich composites

Test method Statistical Core shear strength, MPa Width (b), beam thickness

value mm (d), mm
0 cycle 200 cycles

Balsa Core Flexural Mean 2.044 1.759 76.12 32.207

and Shear Stiffness

Stan-dev 0.277 0.395 0.304 0.279
(ASTM C393, ASTM

D7250)

Balsa Core Shear Mean 171.73 164.68 76.12 32.207

Modulus
Stan-dev 35.88 70.38 0.03 0.279

Foam Core Flexural Mean 0.998 0.956 75.79 32.562

and Shear Stiffness

Stan-dev 0.03745 0.030 0.462 0.144
(ASTM C393, ASTM

D7250)

Foam Core Shear Mean 18.829 23.551 75.79 32.562

Modulus
Stan-dev 1.370 1.262 0.462 0.144

83
 Table 10. Statistical values for balsa and foam core compression tests of sandwich

composites

Test Method Statistical Core shear strength, MPa Width (b), beam

value mm thickness
0 cycle 200 cycles
(d), mm

Balsa core Mean 10.745 6.847 77.19 76.91

Compression strength
Stan-dev 2.261 1.913 1.036 1.041

Balsa Core Mean 505.52 340.58 77.19 76.91

Compression
Stan-dev 172.28 106.11 1.036 1.041
Modulus

Foam Core Mean 0.995 0.956 76.84 76.375

Compression
Stan-dev 0.095 0.192 0.713 0.605
Strength

Foam Core Mean 48.546 42.71 76.84 76.375

Compression
Stan-dev 6.245 4.355 0.713 0.605
Modulus

The OSL values for normal, lognormal, and Weibull distribution for each one of

the materials were calculated, and they are provided in Table 11.

84
 Table 11. Data results of OSL for normal, lognormal, and Weibull distribution

Sample Coefficient of OSL

Material properties
size variation Normal Lognormal Weibull

Balsa core shear strength 22 13.5 0.385 0.229 0.608

Balsa Core Shear Modulus 22 41.8 0.3171 0.622 0.102

Balsa Core Compression

31 21.04 0.373 0.025 0.738
Strength

Balsa Core Compression

31 33.8 0.396 0.724 0.491
Modulus

Foam Core Shear Strength 30 3.75 0.637 0.600 0.608

Foam Core Shear Modulus 30 7.16 2.81E-05 9.42E-06 7.86E-07

Foam Core Compression

24 9.42 0.0192 0.0490 0.000664
Strength

Foam Core Compression

24 12.59 0.719 0.5860 0.683
Modulus

The results of applying normal, lognormal and Weibull distribution on data has

been shown in Figures of 58 to 65:

1. Shear Test Results

The shear test results are shown in Figures 58 to 61.

85
Figure 58. Shear strength statistical plots of balsa core material

86
Figure 59. Shear modulus statistical plot of balsa core material

87
Figure 60. Shear strength statistical plot of foam core material

88
Figure 61. Shear modulus statistical plot of foam core material

89
2. Compression Test Results

The compression test results are shown in Figures 62 to 65.

Figure 62. Compression strength statistical plot of balsa core material

90
Figure 63. Compression modulus statistical plot of balsa core materials

91
 test

Figure 64. Compression strength statistical plot of foam core material

92
Figure 65. Compression modulus statistical plot of foam core material

93
 Based on the results of the Anderson-Darling test and probability distribution plot,

the following distributions, presented in Table 12, were selected:

Table 12. Selected distributions for balsa core and foam mechanical properties

Balsa Core Sandwich Foam Core Sandwich

Shear strength Weibull Lognormal

Shear modulus Weibull Weibull

Compression strength Lognormal Weibull

Compression modulus Weibull Normal

Weibull distribution signifies the most commonly employed distribution in the

composite industry; further, it is also recommended by Handbook MIL 17 [61, 64]. This

does not imply that Weibull distribution will necessarily form the most suitable fit for these

composites. However, due to a lack of available data with a sufficient size unlike steel and

concrete, the most common distribution in the composite forms a better alternative for us

and will facilitate the modeling of an appropriate resistance factor and end-use condition

for purposes of practical design. Moreover, the design codes of wood structures are based

on Weibull distribution and will help LRFD similar to it. In this thesis, a distribution was

selected based on the best distribution fit.

The distribution for each one of the variables is illustrated in Table 13.

94
 Table 13. Selected distributions for variables

p Δ b d

Balsa Shear strength Weibull - Normal Normal

Balsa Shear modulus Weibull Normal - -

Balsa Compression Strength Lognormal - Normal Normal

Balsa Compression Modulus Weibull Normal Normal Normal

Foam Shear Strength Lognormal - Normal Normal

Foam Shear Modulus Weibull Normal - -

Foam Compression Strength Weibull - Normal Normal

Foam Compression Modulus Normal Normal Normal Normal

Parameters Estimation

Table 14 demonstrates the results of maximum load, nominal strength, and the

resistance factors for the target reliability value of three and for each type of test methods.

95
 Table 14. Resistance factor results
Nominal Max allowable Resistance
Test Method
strength, Rn stress, S (MPa) factor, Ф

Bals Core Shear Strength 1.351 1.06 0.790

Balsa Core Shear Modulus 99.97 73.972 0.74

Balsa Core Compression Strength 6.218 5.164 0.83

Balsa Core Compression Modulus 160.96 129.465 0.804

Foam Core Shear strength 0.923 0.883 0.957

Foam Core Shear Modulus 16.089 17.235 1

Foam Core Compression Strength 0.805 0.756 0.94

Foam Core Compression Modulus 36.056 32.47 0.902

Based on the results in Table 14, it is possible to determine the environmental factor

or F/T (Cf-t) for both balsa core and foam core composite sandwiches by the application of

equation 2. Their values are depicted in Table 15.

96
 Table 15. F/T coefficient introduced in this research

Types of Composite Sandwich F/T coefficient (Cf-t)

Balsa Core Shear Strength 0.73

Balsa Core Shear Modulus 0.23

Balsa Core Compression Strength 0.48

Balsa Core Compression Modulus 0.79

Foam Core Shear Strength 1

Foam Core Shear Modulus 1

Foam Core Compression Strength 0.95

Foam Core Compression Modulus 0.942

The coefficient of F/T for foam core sandwiches is equal to 1, which means that

freeze-thaw did not have any significant effect on the foam shear strength, but the effect of

F/T on balsa core sandwiches was substantial. These outcomes are in agreement with the

experimental results.

97
 CHAPTER 6

SUMMARY OF SIGNIFICANT RESEARCH RESULTS AND FINDINGS

The degradation and change in the microstructure of the core materials were

characterized using several nondestructive and destructive techniques. The following

results and observations were made:

1- The most significant reduction due to F/T cycling combined with saline

environment occurred in the compression strength and modulus, with a 36%

and 33%, respectively. In addition, the balsa wood core shear strength decreased

by 14%

2- The F/T cycling increased the PVC foam shear modulus by 25%. This was

consistent with the DMA testing, which revealed a significant increase in

storage modulus of the foam after exposure. In addition, the DSC results

showed a modest increase in Tg which could be related to a higher crosslink

density of the foam.

98
 3- Fatigue life of balsa wood increased after exposure. The optical microscopy

showed fibrillation of exposed wood cells; this could be attributed to the

mechanical fatigue from expansion and contraction. The friction in between the

cell walls of the balsa wood increased its damping characteristics; in addition,

it is believed that the fibrillation mechanism delayed the crack propagation

while fatiguing the specimens. This behavior was justified by the increase in

loss modulus of the DMA tests performed in both the bending and shear modes.

4- The facing cleavage of sandwich panels test (ASTM E2004) showed minimal

effect on fracture toughness of the balsa wood and PVC foam core materials

after F/T exposure.

The experimental results and evaluation of a WR E-glass fabric/epoxy vinyl ester

composite submerged in saline solution and exposed to 100 days of F–T environment was

studied. State-of-the-art NDI testing techniques were used to investigate the nature of the

degradation, including all constituent materials, interfaces, and fabric construction. A

mechanistic model describing the failure mechanism and damage initiation and

propagation as related to observed material degradation and reduction is presented. The

following conclusions have been drawn:

1. The conditioned specimens showed a substantial material degradation when

compared to identical control specimens. The largest reduction was observed in the

tensile properties of the GFRP. A reduction of 23% in strength and 20% in modulus

was detected. The average reductions in shear modulus and strength were around

18.5% and 13.5%, respectively, whereas an average reduction of 17.1% in

compression strength and 12% in compression modulus was observed.

99
2. The shear failure postmortem optical microscopy images showed delamination and

intralaminar cracks in the exposed samples.

3. The DMA testing indicated plasticization of the matrix from the exposure. The

plasticization effects could be removed by heating the sample (thereby removing

water), but there was a permanent decrease in storage modulus that is attributed to

physical degradation of the polymer and/or fiber-matrix interface. Most likely

hydrolysis of the resin was not involved. DSC results corroborated these findings.

These results are consistent with the formation of internal cracks inside the GFRP

laminate.

4. The fiber/matrix interface has traditionally been viewed as the main damage due to

thermal cycling. However, when F–T is combined with saline exposure, the warp

fibers specifically at the weave junctions experienced extensive damage and

fracture.

5. The X-ray computed tomography showed a preferred site location of defects in the

WR construction. The frequency of these defects was mainly located in the resin

rich areas between the fill and warp fibers.

6. The ultrasonic C-scan and optical microscopy showed that the volume change due

to F–T cycling and reduction of through-thickness tensile strength of the matrix

binding the warp and fill fibers at the weave junction impaired the structural

integrity of the FRP material.

7. The observed strength and stiffness reductions after accelerated exposure were

similar to what was reported in the literature on GFRP immersed in seawater for up

to eight years. The testing scheme and protocol proposed in this study could be used

100
 as an accelerated testing method for future evaluation of durability in composite

materials.

8. The results of this research are expected to apply to plain weave fabric regardless

of the fiber types used in the reinforced polymer composites. A multi-scale

durability and damage tolerance model would take advantage of the damage types

presented in this paper, including the frequency, sizes, shapes, locations, and

distribution of the defects.

101
 CHAPTER 7

CONCLUSIONS AND RECOMMENDATIONS

Composite sandwich structures are currently used in civil engineering applications

such as seawalls, bridge decks, tanks, and wind turbine blades. However, a major concern

in this regard has been whether the sandwich composite, especially the core material, offers

long-term durability and lifetime reliability when exposed to various outdoor

environments. Both novel and existing structures demand more extensive investigation

with regard to the selection of the strength reduction factor (also referred to as safety factor)

or other factors imposed on their design equations. The current values of these factors have

been typically determined based on subjective judgment, and therefore they still require

concrete evidence for validation. This validation can be obtained through a reliability

analysis, which relates the probability of failure to the safety factor and the amount of

tolerated load, providing a basis for their calibration to attain the targeted level of safety.

For this purpose, two types of sandwich core composites were designed and

fabricated using a woven E-glass reinforced vinyl ester resin. This construction included

relatively thick face sheets that promoted core shear failure in all beam specimens. The

effect on core mechanical properties (shear and compression) before and after 100 days of

F/T exposure in saline solution (3% NaCl) was investigated. Shear strength, modulus, and

102
lifetime simulated fatigue testing properties of the balsa wood (SB100) and PVC foam

(Airex C70.55) was assessed.The main conclusions from this study are the following:

 The F/T effect reduced the mechanical properties of balsa core sandwich

composite materials significantly; however, this effect for foam core

sandwiches was not substantial.

 Say something about whether the use of models was successful, and which

one(s) provided the best fit.

 State the resulting safety factors here.

In summary, the novelty of the work is the following. Introducing a strength

reduction factor (ɸ) will increase the chance for sandwich composites in civil engineering

applications. The probabilistic approach used in this study is a means to reach to this

goal. Currently these factors are not available for designing sandwich composites in civil

engineering applications. This work provided a basis for using this approach and calculated

initial values for consideration. In addition, this study contributed new and detailed

experimental data on the behavior of balsa and foam core composites in freeze/thaw saline

environment, which will augment what little is currently published in the literature on this

subject.

103
 BIBLIOGRAPHY

1- Sharma, Nitin, Ronald F. Gibson, and Emmanuel O. Ayorinde. "Fatigue of foam

and honeycomb core composite sandwich structures: a tutorial." Journal of

Sandwich Structures & Materials 8, no. 4 (2006): 263-319.

2- Zureick, Abdul-Hamid, B. Shih, and E. Munley. "Fiber-reinforced polymeric

bridge decks." Structural Engineering Review 3, no. 7 (1995): 257-266.

3- Barjasteh, Ehsan. Isothermal and hygrothermal agings of hybrid glass

fiber/carbon fiber composite. 2011.

4- Ishai, Ori, Clement Hiel, and Michael Luft. "Long-term hygrothermal effects on

damage tolerance of hybrid composite sandwich panels." Composites 26.1 (1995):

47-55.

5- Li, Xiaoming, and Y. Jack Weitsman. "Sea-water effects on foam-cored

composite sandwich lay-ups." Composites Part B: Engineering 35.6 (2004): 451-

459.

6- Sun, Wei, et al. "Damage and damage resistance of high strength concrete under

the action of load and freeze-thaw cycles." Cement and Concrete Research 29.9

(1999): 1519-1523.

104
7- Barbero, Ever J. Introduction to composite materials design. CRC press, 2010.

8- Soares C. Reliability of components in composite materials. Reliab Eng Syst Safe

1997;55(2):171–7.

9- Chiachio,M ,Chiachio J, Rus G, Reliability in composites – A selective review and

survey of current development, Composites Part B: Engineering 2012;43(3); 902–

913

10- Dimitrov, Nikolay, and Christian Berggreen. "Probabilistic fatigue life of balsa

cored sandwich composites subjected to transverse shear." Journal of Sandwich

Structures & Materials 17, no. 5 (2015): 562-577.

11- http://www.fireco.no/?news=bascule-frp-composite-footbridge-fredrikstad

12- Jozsef, Bodig. Mechanics of wood and wood composites. 1982.

13- Vural, M., and G. Ravichandran. "Dynamic response and energy dissipation

characteristics of balsa wood: experiment and analysis." International Journal of

Solids and structures 40, no. 9 (2003): 2147-2170.

14- Kahle, E., and J. Woodhouse. "The influence of cell geometry on the elasticity of

softwood." Journal of materials science 29, no. 5 (1994): 1250-1259.

15- Osei-Antwi, Michael, Julia De Castro, Anastasios P. Vassilopoulos, and Thomas

Keller. "Shear mechanical characterization of balsa wood as core material of

composite sandwich panels." Construction and Building Materials 41 (2013): 231-

238.

16- Soden, P. D., and R. D. McLeish. "Variables affecting the strength of balsa wood."

The Journal of Strain Analysis for Engineering Design 11, no. 4 (1976): 225-234.

105
17- Da Silva, Andre, and Stelios Kyriakides. "Compressive response and failure of

balsa wood." International Journal of Solids and Structures 44, no. 25 (2007): 8685-

8717.

18- Bond, I. P., and Martin P. Ansell. "Fatigue properties of jointed wood composites

Part I Statistical analysis, fatigue master curves and constant life diagrams." Journal

of materials science 33, no. 11 (1998): 2751-2762.

19- Kline, Donald Edgar, Frank E. Woeste, and B. A. Bendtsen. "Stochastic model for

modulus of elasticity of lumber." Wood and fiber science 18, no. 2 (2007): 228-

238.

20- Gibson, Lorna J., and Michael F. Ashby. Cellular solids: structure and properties.

Cambridge university press, 1999.

21- Zinno, A., et al. "Experimental characterization of phenolic-impregnated

honeycomb sandwich structures for transportation vehicles." Composite Structures

93.11 (2011): 2910-2924.

22- Toubia EA, Emami S, Klosterman D. Degradation mechanisms of balsa wood and

PVC foam sandwich core composites due to freeze/thaw exposure in saline

solution. Journal of Sandwich Structures & Materials. 2017 Apr

28:1099636217706895.

23- Toubia EA, Emami S, Klosterman D. Failure mechanism of woven roving

fabric/vinyl ester composites in freeze–thaw saline environment. Journal of

Composite Materials. 2016 Nov 30:0021998316681860.

24- Toubia EA, Emami S, durability of sandwich composite structures due to freezing

and thawing and ice chemicals, CMAX 2016

106
25- Siriruk, Akawut, Dayakar Penumadu, and Y. Jack Weitsman. "Effect of sea

environment on interfacial delamination behavior of polymeric sandwich

structures." Composites Science and Technology 69.6 (2009): 821-828.

26- Elias A. Toubia, Abraham Elmushyakhi, Influence of core joints in sandwich

composites under in-plane static and fatigue loads, Materials & Design, Volume

131, 5 October 2017, Pages 102-111, ISSN 0264-1275.

27- Abraham Elmushyakhi and Elias Toubia. 2017. "Axial Fatigue Characterization of

Core Joints in Sandwich Composite Structures," Proceedings of the American

Society for Composites: Thirty-Second Technical Conference, Purdue University,

West Lafayette, Indiana, October 23-25, 2017. Lancaster, PA, USA: DEStech

Publication, Inc., Electronic product

28- Toubia, E., Morgan, A. and Elmushyakhi, A. "Post-Fire Residual Mechanical

Properties of a Composite Laminate — Experimental Study," FL, December 11-

14, 2017. CAMX – The Composites and Advanced Materials Expo.

29- Elmushyakhi, Abraham. "In-Plane Fatigue Characterization of Core Joints in

Sandwich Composite Structures." Electronic Dissertation. University of Dayton,

2017. OhioLINK Electronic Theses and Dissertations Center.

30- Buffel, Bart, Frederik Desplentere, Kris Bracke, and Ignace Verpoest. "Modelling

open cell-foams based on the Weaire–Phelan unit cell with a minimal surface

energy approach." International Journal of Solids and Structures 51, no. 19 (2014):

3461-3470.

31- https://netcomposites.com/guide-tools/guide/core-materials/pvc-foam/

107
32- Alemán JV, Chadwick AV, He J, Hess M, Horie K, Jones RG, Kratochvíl P, Meisel

I, Mita I, Moad G, Penczek S. Definitions of terms relating to the structure and

processing of sols, gels, networks, and inorganic-organic hybrid materials (IUPAC

Recommendations 2007). Pure and Applied Chemistry. 2007 Jan 1;79(10):1801-

29. Saenz, Elio. Fatigue and fracture of foam cores used in sandwich composites.

Florida Atlantic University, 2012.

33- Kamalloo A, Emami S, Touranian H. Enhancing mechanical strength of inorganic

geopolymer by using phenol resin. International Journal of Engineering-

Transactions B: Applications. 2012 Apr 30; 25(4):325.

34- Emami S, Nooranian H, Kamalloo A. Flexural strength and microstructure of

alkali-resistant glass fibre reinforced calcium aluminate-phenol resin composite.

Advances in Cement Research. 2011 Jan 13; 23(1):11-5.

35- Sousa, João M., João R. Correia, Susana Cabral-Fonseca, and António C. Diogo.

"Effects of thermal cycles on the mechanical response of pultruded GFRP profiles

used in civil engineering applications." Composite Structures 116 (2014): 720-731.

36- Ishai, Ori, Clement Hiel, and Michael Luft. "Long-term hygrothermal effects on

damage tolerance of hybrid composite sandwich panels." Composites 26, no. 1

(1995): 47-55.

37- Toubia EA, Emami S. Experimental Evaluation of Structural Steel Coating

Systems. Journal of Materials in Civil Engineering. 2016 Jul 1; 28(12):04016147.

38- Li, Xiaoming, and Y. Jack Weitsman. "Sea-water effects on foam-cored composite

sandwich lay-ups." Composites Part B: Engineering 35, no. 6 (2004): 451-459.

108
39- Gomez, Jose, and Brian Casto. "Freeze-thaw durability of composite materials." In

First International Conference on Composites in InfrastructureNational Science

FoundationNational Sicence Foundation. 1996.

40- Wu, Hwai-Chung, Gongkang Fu, Ronald F. Gibson, An Yan, Kraig Warnemuende,

and Vijay Anumandla. "Durability of FRP composite bridge deck materials under

freeze-thaw and low temperature conditions." Journal of bridge engineering 11, no.

4 (2006): 443-451.

41- Toubia EA, Emami S. Performance Comparison of Structural Steel Coating

Systems. Ohio Department of Transportation, Office of Statewide Planning &

Research; 2016 Feb.

42- Zinno, A., A. Prota, E. Di Maio, and C. E. Bakis. "Experimental characterization

of phenolic-impregnated honeycomb sandwich structures for transportation

vehicles." Composite structures 93, no. 11 (2011): 2910-2924.

43- Aviles, F., and M. Aguilar‐Montero. "Moisture absorption in foam‐cored

composite sandwich structures." Polymer Composites 31, no. 4 (2010): 714-722.

44- Manujesh, B. J., Vijayalakshmi Rao, and MP Sham Aan. "Moisture absorption and

mechanical degradation studies of polyurethane foam cored E-glass-reinforced

vinyl-ester sandwich composites." Journal of Reinforced Plastics and Composites

33, no. 5 (2014): 479-492.

45- Scudamore, R. J., and W. J. Cantwell. "The effect of moisture and loading rate on

the interfacial fracture properties of sandwich structures." Polymer composites 23,

no. 3 (2002): 406-417.

109
46- Cantwell, W. J., G. Broster, and P. Davies. "The influence of water immersion on

skin-core debonding in GFRP-balsa sandwich structures." Journal of reinforced

plastics and composites 15, no. 11 (1996): 1161-1172.

47- Neşer, Gökdeniz. "Seawater exposure effect on the fracture toughness of low-

density woods/GRP sandwich systems." Journal of wood science 56, no. 2 (2010):

154-159.

48- Kolat, Koray, Gökdeniz Neşer, and Çiçek Özes. "The effect of sea water exposure

on the interfacial fracture of some sandwich systems in marine use." Composite

structures 78, no. 1 (2007): 11-17.

49- Singh, Abhendra K., and Barry D. Davidson. "Effects of temperature, seawater and

impact on the strength, stiffness, and life of sandwich composites." Journal of

Reinforced Plastics and Composites 30, no. 3 (2011): 269-277.

50- Siriruk, A., D. Penumadu, and A. Sharma. "Effects of seawater and low

temperatures on polymeric foam core material." Experimental mechanics 52, no. 1

(2012): 25-36.

51- Rivera, J., and V. M. Karbhari. "Cold-temperature and simultaneous aqueous

environment related degradation of carbon/vinylester composites." Composites

Part B: Engineering 33.1 (2002): 17-24.

52- Shi, Jia-Wei, Hong Zhu, Gang Wu, and Zhi-Shen Wu. "Tensile behavior of FRP

and hybrid FRP sheets in freeze–thaw cycling environments." Composites Part B:

Engineering 60 (2014): 239-247.

53- Gomez, Jose, and Brian Casto. "Freeze-thaw durability of composite materials."

First International Conference on Composites in Infrastructure. 1996.

110
54- Wu, Hwai-Chung, et al. "Durability of FRP composite bridge deck materials

under freeze-thaw and low temperature conditions." Journal of bridge engineering

11.4 (2006): 443-451.

55- Lord, Harold W., and Piyush K. Dutta. "On the design of polymeric composite

structures for cold regions applications." Journal of Reinforced Plastics and

Composites 7.5 (1988): 435-458.

56- Marouani S, Curtil L and Hamelin P. Ageing of carbon/epoxy and

carbon/vinylester composites used in the reinforcement and/or the repair of civil

engineering structures. Compos B Eng 2012; 43: 2020–2030

57- Aniskevich, K., et al. "Mechanical properties of pultruded glass fiber reinforced

plastic after freeze–thaw cycling." Journal of Reinforced Plastics and Composites

(2012): 0731684412462755.

58- Sousa, João M., et al. "Effects of thermal cycles on the mechanical response of

pultruded GFRP profiles used in civil engineering applications." Composite

Structures 116 (2014): 720-731.

59- Dutta PK and Hui D. Low-temperature and freeze–thaw durability of thick

composites. Compos B Eng 1996; 27:371–379

60- Wu L, Murphy K, Karbhari VM, et al. Short-term effects of sea water on E-glass/

vinylester composites. J Appl.Polym Sci 2002; 84: 2760–2767.

61- Karbhari VM and Chu W. Degradation kinetics of pultruded E-glass/vinylester in

alkaline media. ACI Mater J 2005; 102: 34–41.

111
62- Guzman, V. Alzamora, and Povl Brøndsted. "Effects of moisture on glass fiber-

reinforced polymer composites." Journal of Composite Materials 49, no. 8 (2015):

911-920.

63- Mahadevan, Sankaran, and A. Haldar. "Probability, reliability and statistical

method in engineering design." (2000).

64- Zureick, Abdul-Hamid, Richard M. Bennett, and Bruce R. Ellingwood.

"Statistical characterization of fiber-reinforced polymer composite material

properties for structural design." Journal of Structural Engineering 132.8 (2006):

1320-1327.

65- Galambos TV, Ellingwood B, MacGregor JG, Cornell CA. Probability based load

criteria: assessment of current design practice. Journal of the Structural Division.

1982 May; 108(5):959-77.

66- MacGregor, James G. "Load and resistance factors for concrete design." Journal

Proceedings. Vol. 80. No. 4. 1983.

67- Gromala DS, Sharp DJ, Pollock DG, Goodman JR. Load and resistance factor

design for wood: the new US wood design specification. InProc., Int. Timber

Engineering Conference 1990 (pp. 311-318).

68- M. W. Long and J. D. Narciso, ‘Probabilistic Design Methodology for Composite

Aircraft Structures’, Dot/Faa/Ar-99/2, 1999.

69- ASTM D5457 – 15, Standard Specification for Computing Reference Resistance

of Wood-Based Materials and Structural Connections for Load and Resistance

Factor Design

112
70- Handbook-MIL-HDBK M. 17-3F: Composite Materials Handbook, Volume 3-

Polymer Matrix Composites Materials Usage, Design, and Analysis. US

Department of Defense. 2002.

71- J.Ferry Borges, M. Castanheta , Structural Safety,(2nd edition) Laboratorio

Nacional de Engenharia Civil, Lisbon (1972)

72- Ellingwood BR. Probability-based LRFD for engineered wood construction.

Structural Safety. 1997 Jan 1; 19(1):53-65.

73- King RL. Statistical methods for determining design allowable properties for

advanced composite materials. In The 15th Reinforced Plastics Congress 1986

1986 (pp. 79-85).

74- American Society of Civil Engineers. Minimum design loads for buildings and

other structures. Vol. 7. Amer Society of Civil Engineers, 2010.

75- L.Yang, "Reliability of composite laminates." Mechanics of structures and

machines 16.4 (1988): 523-536.

76- C. Lind, "The design of structural design norms." Journal of Structural Mechanics

1.3 (1972): 357-370

77- Ellingwood, Bruce R. "Toward load and resistance factor design for fiber-

reinforced polymer composite structures." Journal of Structural Engineering 129.4

(2003): 449-458.

78- Nowak, Andrzej S., and Kevin R. Collins. Reliability of structures. CRC Press,

2012.

79- Ditlevsen, Ove, and Henrik O. Madsen. Structural reliability methods. Vol. 178.

New York: Wiley, 1996.

113
80- M.Chiachio, J.Chiachio and G.Rus, ‘Reliability in Composites - A Selective

Review and Survey of Current Development’, Composites Part B: Engineering,

43 (2012), 902–13

81- Zadeh, Hany Jawaheri, and Antonio Nanni. "Reliability analysis of concrete

beams internally reinforced with fiber-reinforced polymer bars." ACI Structural

Journal 110.6 (2013): 1023.

114