Abaqus Example Problems Manual: Abaqus Version 6.6 ID: Printed On

Abaqus Example Problems Manual
Abaqus Version 6.6 ID:

Printed on:
Abaqus
Example Problems Manual
Volume II

Printed on:
Legal Notices
CAUTION: This documentation is intended for qualied users who will exercise sound engineering judgment and expertise in the use of the Abaqus
Software. The Abaqus Software is inherently complex, and the examples and procedures in this documentation are not intended to be exhaustive or to apply
to any particular situation. Users are cautioned to satisfy themselves as to the accuracy and results of their analyses.
Dassault Systmes and its subsidiaries, including Dassault Systmes Simulia Corp., shall not be responsible for the accuracy or usefulness of any analysis
performed using the Abaqus Software or the procedures, examples, or explanations in this documentation. Dassault Systmes and its subsidiaries shall not
be responsible for the consequences of any errors or omissions that may appear in this documentation.
The Abaqus Software is available only under license from Dassault Systmes or its subsidiary and may be used or reproduced only in accordance with the
terms of such license. This documentation is subject to the terms and conditions of either the software license agreement signed by the parties, or, absent
such an agreement, the then current software license agreement to which the documentation relates.
This documentation and the software described in this documentation are subject to change without prior notice.
No part of this documentation may be reproduced or distributed in any form without prior written permission of Dassault Systmes or its subsidiary.
The Abaqus Software is a product of Dassault Systmes Simulia Corp., Providence, RI, USA.
Dassault Systmes, 2010
Abaqus, the 3DS logo, SIMULIA, CATIA, and Unied FEA are trademarks or registered trademarks of Dassault Systmes or its subsidiaries in the United
States and/or other countries.
Other company, product, and service names may be trademarks or service marks of their respective owners. For additional information
concerning trademarks, copyrights, and licenses, see the Legal Notices in the Abaqus 6.10 Release Notes and the notices at:
http://www.simulia.com/products/products_legal.html.

Printed on:
Locations
SIMULIA Worldwide Headquarters
SIMULIA European Headquarters
Rising Sun Mills, 166 Valley Street, Providence, RI 029092499, Tel: +1 401 276 4400,
Fax: +1 401 276 4408, simulia.support@3ds.com http://www.simulia.com
Gaetano Martinolaan 95, P. O. Box 1637, 6201 BP Maastricht, The Netherlands, Tel: +31 43 356 6906,
Fax: +31 43 356 6908, simulia.europe.info@3ds.com
Technical Support Centers

United States
Australia
Austria
Benelux
Canada
China
Czech & Slovak Republics
Finland
France
Germany
Greece
India
Israel
Italy
Japan
Korea
Latin America
Malaysia
New Zealand
Poland
Russia, Belarus & Ukraine
Scandinavia
Singapore
South Africa
Spain & Portugal
Taiwan
Thailand
Turkey
United Kingdom
Fremont, CA, Tel: +1 510 794 5891, simulia.west.support@3ds.com

West Lafayette, IN, Tel: +1 765 497 1373, simulia.central.support@3ds.com
Northville, MI, Tel: +1 248 349 4669, simulia.greatlakes.info@3ds.com
Woodbury, MN, Tel: +1 612 424 9044, simulia.central.support@3ds.com
Beachwood, OH, Tel: +1 216 378 1070, simulia.erie.info@3ds.com
West Chester, OH, Tel: +1 513 275 1430, simulia.central.support@3ds.com
Warwick, RI, Tel: +1 401 739 3637, simulia.east.support@3ds.com
Lewisville, TX, Tel: +1 972 221 6500, simulia.south.info@3ds.com
Richmond VIC, Tel: +61 3 9421 2900, simulia.au.support@3ds.com
Vienna, Tel: +43 1 22 707 200, simulia.at.info@3ds.com
Huizen, The Netherlands, Tel: +31 35 52 58 424, simulia.benelux.support@3ds.com
Toronto, ON, Tel: +1 416 402 2219, simulia.greatlakes.info@3ds.com
Beijing, P. R. China, Tel: +8610 6536 2288, simulia.cn.support@3ds.com
Shanghai, P. R. China, Tel: +8621 3856 8000, simulia.cn.support@3ds.com
Synerma s. r. o., Psry, Prague-West, Tel: +420 603 145 769, abaqus@synerma.cz
Vantaa, Tel: +358 46 712 2247, simulia.nordic.info@3ds.com
Velizy Villacoublay Cedex, Tel: +33 1 61 62 72 72, simulia.fr.support@3ds.com
Aachen, Tel: +49 241 474 01 0, simulia.de.info@3ds.com
Munich, Tel: +49 89 543 48 77 0, simulia.de.info@3ds.com
3 Dimensional Data Systems, Crete, Tel: +30 2821040012, support@3dds.gr
Chennai, Tamil Nadu, Tel: +91 44 43443000, simulia.in.info@3ds.com
ADCOM, Givataim, Tel: +972 3 7325311, shmulik.keidar@adcomsim.co.il
Lainate MI, Tel: +39 02 39211211, simulia.ity.info@3ds.com
Tokyo, Tel: +81 3 5442 6300, simulia.tokyo.support@3ds.com
Osaka, Tel: +81 6 4803 5020, simulia.osaka.support@3ds.com
Yokohama-shi, Kanagawa, Tel: +81 45 470 9381, isight.jp.info@3ds.com
Mapo-Gu, Seoul, Tel: +82 2 785 6707/8, simulia.kr.info@3ds.com
Puerto Madero, Buenos Aires, Tel: +54 11 4312 8700, Horacio.Burbridge@3ds.com
WorleyParsons Advanced Analysis, Kuala Lumpur, Tel: +603 2039 9000, abaqus.my@worleyparsons.com
Matrix Applied Computing Ltd., Auckland, Tel: +64 9 623 1223, abaqus-tech@matrix.co.nz
BudSoft Sp. z o.o., Pozna, Tel: +48 61 8508 466, info@budsoft.com.pl
TESIS Ltd., Moscow, Tel: +7 495 612 44 22, info@tesis.com.ru
Vsters, Sweden, Tel: +46 21 150870, simulia.nordic.info@3ds.com
WorleyParsons Advanced Analysis, Singapore, Tel: +65 6735 8444, abaqus.sg@worleyparsons.com
Finite Element Analysis Services (Pty) Ltd., Parklands, Tel: +27 21 556 6462, feas@feas.co.za
Principia Ingenieros Consultores, S.A., Madrid, Tel: +34 91 209 1482, simulia@principia.es
Simutech Solution Corporation, Taipei, R.O.C., Tel: +886 2 2507 9550, lucille@simutech.com.tw
WorleyParsons Advanced Analysis, Singapore, Tel: +65 6735 8444, abaqus.sg@worleyparsons.com
A-Ztech Ltd., Istanbul, Tel: +90 216 361 8850, info@a-ztech.com.tr
Warrington, Tel: +44 1 925 830900, simulia.uk.info@3ds.com
Sevenoaks, Tel: +44 1 732 834930, simulia.uk.info@3ds.com
Complete contact information is available at http://www.simulia.com/locations/locations.html.

Printed on:
Preface
This section lists various resources that are available for help with using Abaqus Unied FEA software.
Support
Both technical engineering support (for problems with creating a model or performing an analysis) and
systems support (for installation, licensing, and hardware-related problems) for Abaqus are offered through
a network of local support ofces. Regional contact information is listed in the front of each Abaqus manual
and is accessible from the Locations page at www.simulia.com.
SIMULIA Online Support System
The SIMULIA Online Support System (SOSS) provides a knowledge database of SIMULIA Answers. The
SIMULIA Answers are solutions to questions that we have had to answer or guidelines on how to use Abaqus,
SIMULIA SLM, Isight, and other SIMULIA products. You can also submit new requests for support in the
SOSS. All support incidents are tracked in the SOSS. If you are contacting us by means outside the SOSS
to discuss an existing support problem and you know the incident number, please mention it so that we can
consult the database to see what the latest action has been.
To use the SOSS, you need to register with the system. Visit the My Support page at www.simulia.com
to register.
Many questions about Abaqus can also be answered by visiting the Products page and the Support
page at www.simulia.com.
Anonymous ftp site
To facilitate data transfer with SIMULIA, an anonymous ftp account is available on the computer
ftp.simulia.com. Login as user anonymous, and type your e-mail address as your password. Contact support
before placing les on the site.
Training
All ofces and representatives offer regularly scheduled public training classes. We also provide training
seminars at customer sites. All training classes and seminars include workshops to provide as much
practical experience with Abaqus as possible. For a schedule and descriptions of available classes, see
www.simulia.com or call your local ofce or representative.
Feedback
We welcome any suggestions for improvements to Abaqus software, the support program, or documentation.
We will ensure that any enhancement requests you make are considered for future releases. If you wish to
make a suggestion about the service or products, refer to www.simulia.com. Complaints should be addressed
by contacting your local ofce or through www.simulia.com by visiting the Quality Assurance section of
the Support page.

Printed on:

Printed on:
CONTENTS
Contents
Volume I
1.
Static Stress/Displacement Analyses

Static and quasi-static stress analyses
Axisymmetric analysis of bolted pipe flange connections

Elastic-plastic collapse of a thin-walled elbow under in-plane bending and internal
pressure
Parametric study of a linear elastic pipeline under in-plane bending
Indentation of an elastomeric foam specimen with a hemispherical punch
Collapse of a concrete slab
Jointed rock slope stability
Notched beam under cyclic loading
Uniaxial ratchetting under tension and compression
Hydrostatic fluid elements: modeling an airspring
Shell-to-solid submodeling and shell-to-solid coupling of a pipe joint
Stress-free element reactivation
Transient loading of a viscoelastic bushing
Indentation of a thick plate
Damage and failure of a laminated composite plate
Analysis of an automotive boot seal
Pressure penetration analysis of an air duct kiss seal
Self-contact in rubber/foam components: jounce bumper
Self-contact in rubber/foam components: rubber gasket
Submodeling of a stacked sheet metal assembly
Axisymmetric analysis of a threaded connection
Direct cyclic analysis of a cylinder head under cyclic thermal-mechanical loadings
Erosion of material (sand production) in an oil wellbore
Submodel stress analysis of pressure vessel closure hardware
Using a composite layup to model a yacht hull
1.1.1
1.1.2
1.1.3
1.1.4
1.1.5
1.1.6
1.1.7
1.1.8
1.1.9
1.1.10
1.1.11
1.1.12
1.1.13
1.1.14
1.1.15
1.1.16
1.1.17
1.1.18
1.1.19
1.1.20
1.1.21
1.1.22
1.1.23
1.1.24
Buckling and collapse analyses
Snap-through buckling analysis of circular arches

Laminated composite shells: buckling of a cylindrical panel with a circular hole
Buckling of a column with spot welds
Elastic-plastic K-frame structure
Unstable static problem: reinforced plate under compressive loads
Buckling of an imperfection-sensitive cylindrical shell
Abaqus ID:exa-toc
Printed on: Wed February 24 -- 17:09:53 2010
1.2.1
1.2.2
1.2.3
1.2.4
1.2.5
1.2.6
CONTENTS
Forming analyses
Upsetting of a cylindrical billet: quasi-static analysis with mesh-to-mesh solution

mapping (Abaqus/Standard) and adaptive meshing (Abaqus/Explicit)
Superplastic forming of a rectangular box
Stretching of a thin sheet with a hemispherical punch
Deep drawing of a cylindrical cup
Extrusion of a cylindrical metal bar with frictional heat generation
Rolling of thick plates
Axisymmetric forming of a circular cup
Cup/trough forming
Forging with sinusoidal dies
Forging with multiple complex dies
Flat rolling: transient and steady-state
Section rolling
Ring rolling
Axisymmetric extrusion: transient and steady-state
Two-step forming simulation
Upsetting of a cylindrical billet: coupled temperature-displacement and adiabatic
analysis
Unstable static problem: thermal forming of a metal sheet
Inertia welding simulation using Abaqus/Standard and Abaqus/CAE
1.3.1
1.3.2
1.3.3
1.3.4
1.3.5
1.3.6
1.3.7
1.3.8
1.3.9
1.3.10
1.3.11
1.3.12
1.3.13
1.3.14
1.3.15
1.3.16
1.3.17
1.3.18
Fracture and damage
A plate with a part-through crack: elastic line spring modeling

Contour integrals for a conical crack in a linear elastic infinite half space
Elastic-plastic line spring modeling of a finite length cylinder with a part-through axial
flaw
Crack growth in a three-point bend specimen
Analysis of skin-stiffener debonding under tension
Failure of blunt notched fiber metal laminates
Debonding behavior of a double cantilever beam
Debonding behavior of a single leg bending specimen
Postbuckling and growth of delaminations in composite panels
1.4.1
1.4.2
1.4.3
1.4.4
1.4.5
1.4.6
1.4.7
1.4.8
1.4.9
Import analyses
Springback of two-dimensional draw bending

Deep drawing of a square box
2.
1.5.1
1.5.2
Dynamic Stress/Displacement Analyses

Dynamic stress analyses
Nonlinear dynamic analysis of a structure with local inelastic collapse

Detroit Edison pipe whip experiment
vi
Abaqus ID:exa-toc
2.1.1
2.1.2
CONTENTS
Rigid projectile impacting eroding plate

Eroding projectile impacting eroding plate
Tennis racket and ball
Pressurized fuel tank with variable shell thickness
Modeling of an automobile suspension
Explosive pipe closure
Knee bolster impact with general contact
Crimp forming with general contact
Collapse of a stack of blocks with general contact
Cask drop with foam impact limiter
Oblique impact of a copper rod
Water sloshing in a baffled tank
Seismic analysis of a concrete gravity dam
Progressive failure analysis of thin-wall aluminum extrusion under quasi-static and
dynamic loads
Impact analysis of a pawl-ratchet device
High-velocity impact of a ceramic target
2.1.3
2.1.4
2.1.5
2.1.6
2.1.7
2.1.8
2.1.9
2.1.10
2.1.11
2.1.12
2.1.13
2.1.14
2.1.15
2.1.16
2.1.17
2.1.18
Mode-based dynamic analyses
Analysis of a rotating fan using substructures and cyclic symmetry

Linear analysis of the Indian Point reactor feedwater line
Response spectra of a three-dimensional frame building
Brake squeal analysis
Dynamic analysis of antenna structure utilizing residual modes
Steady-state dynamic analysis of a vehicle body-in-white model
2.2.1
2.2.2
2.2.3
2.2.4
2.2.5
2.2.6
Eulerian analyses
Rivet forming
Impact of a water-filled bottle using Eulerian-Lagrangian contact
2.3.1
2.3.2
Co-simulation analyses
Closure of an air-filled door seal

Dynamic impact of a scooter with a bump
2.4.1
2.4.2
vii
Abaqus ID:exa-toc
CONTENTS
Volume II
3.
Tire and Vehicle Analyses

Tire analyses
Symmetric results transfer for a static tire analysis

Steady-state rolling analysis of a tire
Subspace-based steady-state dynamic tire analysis
Steady-state dynamic analysis of a tire substructure
Coupled acoustic-structural analysis of a tire filled with air
Import of a steady-state rolling tire
Analysis of a solid disc with Mullins effect and permanent set
Tread wear simulation using adaptive meshing in Abaqus/Standard
Dynamic analysis of an air-filled tire with rolling transport effects
Acoustics in a circular duct with flow
3.1.1
3.1.2
3.1.3
3.1.4
3.1.5
3.1.6
3.1.7
3.1.8
3.1.9
3.1.10
Vehicle analyses
Inertia relief in a pick-up truck

Substructure analysis of a pick-up truck model
Display body analysis of a pick-up truck model
Continuum modeling of automotive spot welds
3.2.1
3.2.2
3.2.3
3.2.4
Occupant safety analyses
Seat belt analysis of a simplified crash dummy

Side curtain airbag impactor test
4.
Mechanism Analyses
Resolving overconstraints in a multi-body mechanism model

Crank mechanism
Snubber-arm mechanism
Flap mechanism
Tail-skid mechanism
Cylinder-cam mechanism
Driveshaft mechanism
Geneva mechanism
Trailing edge flap mechanism
Substructure analysis of a one-piston engine model
Application of bushing connectors in the analysis of a three-point linkage
Gear assemblies
5.
3.3.1
3.3.2
4.1.1
4.1.2
4.1.3
4.1.4
4.1.5
4.1.6
4.1.7
4.1.8
4.1.9
4.1.10
4.1.11
4.1.12
Heat Transfer and Thermal-Stress Analyses
Thermal-stress analysis of a disc brake
5.1.1
viii
Abaqus ID:exa-toc
CONTENTS
A sequentially coupled thermal-mechanical analysis of a disc brake with an Eulerian

approach
Exhaust manifold assemblage
Coolant manifold cover gasketed joint
Conductive, convective, and radiative heat transfer in an exhaust manifold
Thermal-stress analysis of a reactor pressure vessel bolted closure
6.
Fluid Dynamics and Fluid-Structure Interaction
Conjugate heat transfer analysis of a component-mounted electronic circuit board

7.
5.1.2
5.1.3
5.1.4
5.1.5
5.1.6
6.1.1
Electrical Analyses
Piezoelectric analyses
Eigenvalue analysis of a piezoelectric transducer

Transient dynamic nonlinear response of a piezoelectric transducer
7.1.1
7.1.2
Joule heating analyses
Thermal-electrical modeling of an automotive fuse

8.
Mass Diffusion Analyses
Hydrogen diffusion in a vessel wall section

Diffusion toward an elastic crack tip
9.
8.1.1
8.1.2
Acoustic and Shock Analyses
Fully and sequentially coupled acoustic-structural analysis of a muffler

Coupled acoustic-structural analysis of a speaker
Response of a submerged cylinder to an underwater explosion shock wave
Convergence studies for shock analyses using shell elements
UNDEX analysis of a detailed submarine model
Coupled acoustic-structural analysis of a pick-up truck
Long-duration response of a submerged cylinder to an underwater explosion
Deformation of a sandwich plate under CONWEP blast loading
10.
7.2.1
9.1.1
9.1.2
9.1.3
9.1.4
9.1.5
9.1.6
9.1.7
9.1.8
Soils Analyses
Plane strain consolidation

Calculation of phreatic surface in an earth dam
Axisymmetric simulation of an oil well
Analysis of a pipeline buried in soil
Hydraulically induced fracture in a well bore
Permafrost thawingpipeline interaction
ix
Abaqus ID:exa-toc
10.1.1
10.1.2
10.1.3
10.1.4
10.1.5
10.1.6
CONTENTS
11.
Abaqus/Aqua Analyses
Jack-up foundation analyses

Riser dynamics
12.
11.1.1
11.1.2
Design Sensitivity Analyses

Overview
Design sensitivity analysis: overview
12.1.1
Examples
Design sensitivity analysis of a composite centrifuge

Design sensitivities for tire inflation, footprint, and natural frequency analysis
Design sensitivity analysis of a windshield wiper
Design sensitivity analysis of a rubber bushing
13.
12.2.1
12.2.2
12.2.3
12.2.4
Postprocessing of Abaqus Results Files
User postprocessing of Abaqus results files: overview

Joining data from multiple results files and converting file format: FJOIN
Calculation of principal stresses and strains and their directions: FPRIN
Creation of a perturbed mesh from original coordinate data and eigenvectors: FPERT
Output radiation viewfactors and facet areas: FRAD
Creation of a data file to facilitate the postprocessing of elbow element results:
FELBOW
Abaqus ID:exa-toc
13.1.1
13.1.2
13.1.3
13.1.4
13.1.5
13.1.6
TIRE AND VEHICLE ANALYSES
3.
Tire and Vehicle Analyses
Tire analyses, Section 3.1

Vehicle analyses, Section 3.2
Occupant safety analyses, Section 3.3

Printed on:
TIRE ANALYSES
3.1
Tire analyses
Symmetric results transfer for a static tire analysis, Section 3.1.1

Steady-state rolling analysis of a tire, Section 3.1.2
Subspace-based steady-state dynamic tire analysis, Section 3.1.3
Steady-state dynamic analysis of a tire substructure, Section 3.1.4
Coupled acoustic-structural analysis of a tire lled with air, Section 3.1.5
Import of a steady-state rolling tire, Section 3.1.6
Analysis of a solid disc with Mullins effect and permanent set, Section 3.1.7
Tread wear simulation using adaptive meshing in Abaqus/Standard, Section 3.1.8
Dynamic analysis of an air-lled tire with rolling transport effects, Section 3.1.9
Acoustics in a circular duct with ow, Section 3.1.10
3.11

Printed on:
TIRE RESULTS TRANSFER
3.1.1
SYMMETRIC RESULTS TRANSFER FOR A STATIC TIRE ANALYSIS
Product: Abaqus/Standard
This example illustrates the use of the *SYMMETRIC RESULTS TRANSFER option as well as the
*SYMMETRIC MODEL GENERATION option to model the static interaction between a tire and a rigid
surface.
The *SYMMETRIC MODEL GENERATION option (Symmetric model generation, Section 10.4.1
of the Abaqus Analysis Users Manual) can be used to create a three-dimensional model by revolving an
axisymmetric model about its axis of revolution or by combining two parts of a symmetric model, where one
part is the original model and the other part is the original model reected through a line or a plane. Both
model generation techniques are demonstrated in this example.
The *SYMMETRIC RESULTS TRANSFER option (Transferring results from a symmetric mesh or
a partial three-dimensional mesh to a full three-dimensional mesh, Section 10.4.2 of the Abaqus Analysis
Users Manual) allows the user to transfer the solution obtained from an axisymmetric analysis onto a
three-dimensional model with the same geometry. It also allows the transfer of a symmetric three-dimensional
solution to a full three-dimensional model. Both these results transfer features are demonstrated in this
example. The results transfer capability can signicantly reduce the analysis cost of structures that undergo
symmetric deformation followed by nonsymmetric deformation later during the loading history.
The purpose of this example is to obtain the footprint solution of a 175 SR14 tire in contact with a at
rigid surface, subjected to an ination pressure and a concentrated load on the axle. Input les modeling a
tire in contact with a rigid drum are also included. These footprint solutions are used as the starting point in
Steady-state rolling analysis of a tire, Section 3.1.2, where the free rolling state of the tire rolling at 10 km/h
is determined and in Subspace-based steady-state dynamic tire analysis, Section 3.1.3, where a frequency
response analysis is performed.
Problem description
The different components of the tire are shown in Figure 3.1.11. The tread and sidewalls are made of
rubber, and the belts and carcass are constructed from ber-reinforced rubber composites. The rubber is
modeled as an incompressible hyperelastic material, and the ber reinforcement is modeled as a linear
elastic material. A small amount of skew symmetry is present in the geometry of the tire due to the
placement and 20.0 orientation of the reinforcing belts.
Two simulations are performed in this example. The rst simulation exploits the symmetry in the
tire model and utilizes the results transfer capability; the second simulation does not use the results
transfer capability. Comparisons between the two methodologies are made for the case where the tire
is in contact with a at rigid surface. Input les modeling a tire in contact with a rigid drum are also
included. The methodology used in the rst analysis is applied in this simulation. Results for this case
are presented in Steady-state rolling analysis of a tire, Section 3.1.2.
The rst simulation is broken down into three separate analyses. In the rst analysis the ination of
the tire by a uniform internal pressure is modeled. Due to the anisotropic nature of the tire construction,
the ination loading gives rise to a circumferential component of deformation. The resulting stress eld
3.1.11

Printed on:
is fully three-dimensional, but the problem remains axisymmetric in the sense that the solution does not
vary as a function of position along the circumference. Abaqus provides axisymmetric elements with
twist (CGAX) for such situations. These elements are used to model the ination loading. Only half the
tire cross-section is needed for the ination analysis due to a reection symmetry through the vertical
line that passes through the tire axle (see Figure 3.1.12). We refer to this model as the axisymmetric
model.
The second part of the simulation entails the computation of the footprint solution, which
represents the static deformed shape of the pressurized tire due to a vertical dead load (modeling the
weight of a vehicle). A three-dimensional model is needed for this analysis. The nite element mesh
for this model is obtained by revolving the axisymmetric cross-section about the axis of revolution.
A nonuniform discretization along the circumference is used as shown in Figure 3.1.13. In addition,
the axisymmetric solution is transferred to the new mesh where it serves as the initial or base state in
the footprint calculations. As with the axisymmetric model, only half of the cross-section is needed
in this simulation, but skew-symmetric boundary conditions must be applied along the midplane of
the cross-section to account for antisymmetric stresses that result from the ination loading and the
concentrated load on the axle. We refer to this model as the partial three-dimensional model.
In the last part of this analysis the footprint solution from the partial three-dimensional model is
transferred to a full three-dimensional model and brought into equilibrium. This full three-dimensional
model is used in the steady-state transport example that follows. The model is created by combining
two parts of the partial three-dimensional model, where one part is the mesh used in the second analysis
and the other part is the partial model reected through a line. We refer to this model as the full threedimensional model.
A second simulation is performed in which the same loading steps are repeated, except that the full
three-dimensional model is used for the entire analysis. Besides being used to validate the results transfer
solution, this second simulation allows us to demonstrate the computational advantage afforded by the
Abaqus results transfer capability in problems with rotational and/or reection symmetries.
Model definition
In the rst simulation the ination step is performed on the axisymmetric model and the results are
stored in the results les (.res, .mdl, .stt, and .prt). The axisymmetric model is discretized
with CGAX4H and CGAX3H elements. The belts and ply are modeled with rebar in surface elements
embedded in continuum elements. The ROUNDOFF TOLERANCE parameter on the *EMBEDDED
ELEMENT option is used to adjust the positions of embedded element nodes such that they lie exactly on
host element edges. This feature is useful in cases where embedded nodes are offset from host element
edges by a small distance caused by numerical roundoff. Eliminating such gaps reduces the number
of constraint equations used to embed the surface elements and, hence, improves performance. The
axisymmetric results are read into the subsequent footprint analysis, and the partial three-dimensional
model is generated by Abaqus by revolving the axisymmetric model cross-section about the rotational
symmetry axis. The *SYMMETRIC MODEL GENERATION, REVOLVE option is used for this
purpose. The partial three-dimensional model is composed of four sectors of CCL12H and CCL9H
cylindrical elements covering an angle of 320, with the rest of the tire divided into 16 sectors of
C3D8H and C3D6H linear elements. The linear elements are used in the footprint region. The use of
3.1.12

Printed on:
cylindrical elements is recommended for regions where it is possible to cover large sectors around the
circumference with a small number of elements. In the footprint region, where the desired resolution
of the contact patch dictates the number of elements to be used, it is more cost-effective to use linear
elements. The road (or drum) is dened as an analytical rigid surface in the partial three-dimensional
model. The results of the footprint analysis are read into the nal equilibrium analysis, and the full
three-dimensional model is generated by reecting the partial three-dimensional model through a
vertical line using the *SYMMETRIC MODEL GENERATION, REFLECT=LINE option. The line
used in the reection is the vertical line in the symmetry plane of the tire, which passes through the axis
of rotation. The REFLECT=LINE parameter is used, as opposed to the REFLECT=PLANE parameter,
to take into account the skew symmetry of the tire. The analytical rigid surface as dened in the partial
three-dimensional model is transferred to the full model without change. The three-dimensional nite
element mesh of the full model is shown in Figure 3.1.14.
In the second simulation a datacheck analysis is performed to write the axisymmetric model
information to the results les. The full tire cross-section is meshed in this model. No analysis is
needed. The axisymmetric model information is read in a subsequent run, and a full three-dimensional
model is generated by Abaqus by revolving the cross-section about the rotational symmetry axis. The
*SYMMETRIC MODEL GENERATION, REVOLVE option is again used for this purpose. The road
is dened in the full model. The three-dimensional nite element mesh of the full model is identical to
the one generated in the rst analysis. However, the ination load and concentrated load on the axle are
applied to the full model without making use of the results transfer capability.
The footprint calculations are performed with a friction coefcient of zero in anticipation
of eventually performing a steady-state rolling analysis of the tire using the *STEADY STATE
TRANSPORT option, as explained in Steady-state rolling analysis of a tire, Section 3.1.2.
Since the results from the static analyses performed in this example are used in a subsequent timedomain dynamic example, the input les contain the following features that would not ordinarily be
included for purely static analyses:
The TRANSPORT parameter is included with the *SYMMETRIC MODEL GENERATION

option to dene streamlines in the model, which are needed by Abaqus to perform streamline
calculations during the *STEADY STATE TRANSPORT analysis in the next example problem.
The TRANSPORT parameter is not required for any analysis type except *STEADY STATE
TRANSPORT.
The hyperelastic material that models the rubber has a *VISCOELASTIC, TIME=PRONY option
included. This enables us to model viscoelasticity in the steady-state transport example that follows.
As a consequence of dening a time-domain viscoelastic material property, the *HYPERELASTIC
option includes the LONG TERM parameter to indicate that the elastic properties dened on the
associated data lines dene the long-term behavior of the rubber. In addition, all *STATIC steps
include the LONG TERM parameter to ensure that the static solutions are based upon the long-term
elastic moduli.
3.1.13

Printed on:
Loading
As discussed in the previous sections, the loading on the tire is applied over several steps. In the rst
simulation the ination of the tire to a pressure of 200.0 kPa is modeled using the axisymmetric tire model
(tiretransfer_axi_half.inp) with a *STATIC analysis procedure. The results from this axisymmetric
analysis are then transferred to the partial three-dimensional model (tiretransfer_symmetric.inp) in
which the footprint solution is computed in two sequential *STATIC steps. The rst of these static steps
establishes the initial contact between the road and the tire by prescribing a vertical displacement of
0.02 m on the rigid body reference node. Since this is a static analysis, it is recommended that contact
be established with a prescribed displacement, as opposed to a prescribed load, to avoid potential
convergence difculties that might arise due to unbalanced forces. The prescribed boundary condition
is removed in the second static step, and a vertical load of
1.65 kN is applied to the rigid body
reference node. The 1.65 kN load in the partial three-dimensional model represents a 3.3 kN load
in the full three-dimensional model. The transfer of the results from the axisymmetric model to the
partial three-dimensional model is accomplished by using the *SYMMETRIC RESULTS TRANSFER,
REVOLVE option. Once the static footprint solution for the partial three-dimensional model has
been established, the *SYMMETRIC RESULTS TRANSFER, REFLECT option is used to transfer
the solution to the full three-dimensional model (tiretransfer_full.inp), where the footprint solution is
brought into equilibrium in a single *STATIC increment. The results transfer sequence is illustrated
in Figure 3.1.15.
Boundary conditions and loads are not transferred with the *SYMMETRIC RESULTS TRANSFER
option; they must be carefully redened in the new analysis to match the loads and boundary conditions
from the transferred solution. Due to numerical and modeling issues the element formulations for the
two-dimensional and three-dimensional elements are not identical. As a result, there may be slight
differences between the equilibrium solutions generated by the two- and three-dimensional models.
In addition, small numerical differences may occur between the symmetric and full three-dimensional
solutions because of the presence of symmetry boundary conditions in the symmetric model that are
not used in the full model. Therefore, it is advised that in a results transfer simulation an initial step be
performed where equilibrium is established between the transferred solution and loads that match the
state of the model from which the results are transferred. It is recommended that an initial *STATIC
step with the initial time increment set to the total step time be used to allow Abaqus/Standard to nd
the equilibrium in one increment.
In the second simulation identical ination and footprint steps are repeated. The only difference is
that the entire analysis is performed on the full three-dimensional model (tiretransfer_full_footprint.inp).
The full three-dimensional model is generated using the restart information from a datacheck analysis
of an axisymmetric model of the full tire cross-section (tiretransfer_axi_full.inp).
Contact modeling
The default contact pair formulation in the normal direction is hard contact, which gives strict
enforcement of contact constraints. Some analyses are conducted with both hard and augmented
Lagrangian contact to demonstrate that the default penalty stiffness chosen by the code does not
3.1.14

Printed on:
affect stress results signicantly. The augmented Lagrangian method is invoked by specifying the
AUGMENTED LAGRANGE parameter on the *SURFACE BEHAVIOR option. The hard and
augmented Lagrangian contact algorithms are described in Contact constraint enforcement methods in
Abaqus/Standard, Section 34.1.2 of the Abaqus Analysis Users Manual.
Solution controls
Since the three-dimensional tire model has a small loaded area and, thus, rather localized forces, the
default averaged ux values for the convergence criteria produce very tight tolerances and cause more
iteration than is necessary for an accurate solution. To decrease the computational time required for
the analysis, the *CONTROLS option can be used to override the default values for average forces and
moments. The default controls are used in this example.
Results and discussion
The results from the rst two simulations are essentially identical. The peak Mises stresses and
displacement magnitudes in the two models agree within 0.3% and 0.2%, respectively. The nal
deformed shape of the tire is shown in Figure 3.1.16. The computational cost of each simulation is
shown in Table 3.1.11. The simulation performed on the full three-dimensional model takes 2.5 times
longer than the results transfer simulation, clearly demonstrating the computational advantage that can
be attained by exploiting the symmetry in the model using the *SYMMETRIC RESULTS TRANSFER
option.
Input files
tiretransfer_axi_half.inp
tiretransfer_symmetric.inp
tiretransfer_symmetric_auglagr.inp
tiretransfer_full.inp
tiretransfer_full_auglagr.inp
tiretransfer_axi_full.inp
tiretransfer_full_footprint.inp
tiretransfer_symm_drum.inp
tiretransfer_full_drum.inp
tiretransfer_node.inp
Axisymmetric model, ination analysis (simulation 1).

Partial three-dimensional model, footprint analysis
(simulation 1).
Partial three-dimensional model, footprint analysis using
augmented Lagrangian contact (simulation 1).
Full three-dimensional model, nal equilibrium analysis
(simulation 1).
using augmented Lagrangian contact (simulation 1).
Axisymmetric model, datacheck analysis
(simulation 2).
Full three-dimensional model, complete analysis
(simulation 2).
Partial three-dimensional model of a tire in contact with
a rigid drum.
Full three-dimensional model of a tire in contact with a
rigid drum.
Nodal coordinates for the axisymmetric models.
3.1.15

Printed on:
tiretransfer_axi_half_ml.inp
tiretransfer_symmetric_ml.inp
tiretransfer_full_ml.inp
Table 3.1.11
Axisymmetric model, ination analysis (simulation 1)

with Marlow hyperelastic model.
Partial three-dimensional model, footprint analysis
(simulation 1) with Marlow hyperelastic model.
(simulation 1) with Marlow hyperelastic model.
Comparison of normalized CPU times for the footprint analysis (normalized

with respect to the total No results transfer analysis).
Ination
Footprint
Total
Use results transfer

and symmetry
conditions
No results
transfer
0.005(a) +0.040(b)
0.265(c) +0.058(d)
0.368
0.347(e)
0.653(e)
1.0
(a) axisymmetric model

(b) equilibrium step in partial three-dimensional model
(c) footprint analysis in partial three-dimensional model
(d) equilibrium step in full three-dimensional model
(e) full three-dimensional model
3.1.16

Printed on:
sidewall
tread
bead
belts
carcass
Figure 3.1.11
Tire cross-section.
Embedded surface elements

carrying rebar
2
3
Figure 3.1.12
Axisymmetric tire mesh.
3.1.17

Printed on:
Z
R
3
1
Figure 3.1.13
Partial three-dimensional tire mesh.
Z
R
3
1
Figure 3.1.14
Full three-dimensional tire mesh.
3.1.18

Printed on:
sidewall
Embedded surface
elements carrying rebar
tread
bead
Axisymmetric model
carcass
belts
2
3
Results
transfer
Results
transfer
Z
R
Z
R
3
1
3
1
Partial 3-D model
Figure 3.1.15
Full 3-D model
Results transfer analysis sequence.
Z
R
3
1
Figure 3.1.16
Deformed three-dimensional tire (deformations scaled by a factor of 2).
3.1.19

Printed on:
ROLLING TIRE
3.1.2
STEADY-STATE ROLLING ANALYSIS OF A TIRE
This example illustrates the use of the *STEADY STATE TRANSPORT option in Abaqus (Steady-state
transport analysis, Section 6.4.1 of the Abaqus Analysis Users Manual) to model the steady-state dynamic
interaction between a rolling tire and a rigid surface. A steady-state transport analysis uses a moving reference
frame in which rigid body rotation is described in an Eulerian manner and the deformation is described in
a Lagrangian manner. This kinematic description converts the steady moving contact problem into a pure
spatially dependent simulation. Thus, the mesh need be rened only in the contact regionthe steady motion
transports the material through the mesh. Frictional effects, inertia effects, and history effects in the material
can all be accounted for in a *STEADY STATE TRANSPORT analysis.
The purpose of this analysis is to obtain free rolling equilibrium solutions of a 175 SR14 tire traveling
at a ground velocity of 10.0 km/h (2.7778 m/s) at different slip angles on a at rigid surface. The slip angle
is the angle between the direction of travel and the plane normal to the axle of the tire. Straight line rolling
occurs at a 0.0 slip angle. For comparison purposes we also consider an analysis of the tire spinning at a
xed position on a 1.5 m diameter rigid drum. The drum rotates at an angular velocity of 3.7 rad/s, so that a
point on the surface of the drum travels with an instantaneous velocity of 10.0 km/h (2.7778 m/s). Another
case presented examines the camber thrust arising from camber applied to a tire at free rolling conditions.
This also enables us to calculate a camber thrust stiffness.
An equilibrium solution for the rolling tire problem that has zero torque, T, applied around the axle is
referred to as a free rolling solution. An equilibrium solution with a nonzero torque is referred to as either a
traction or a braking solution depending upon the sense of T. Braking occurs when the angular velocity of the
tire is small enough such that some or all of the contact points between the tire and the road are slipping and
the resultant torque on the tire acts in an opposite sense from the angular velocity of the free rolling solution.
Similarly, traction occurs when the angular velocity of the tire is large enough such that some or all of the
contact points between the tire and the road are slipping and the resultant torque on the tire acts in the same
sense as the angular velocity of the free rolling solution. Full braking or traction occurs when all the contact
points between the tire and the road are slipping.
A wheel in free rolling, traction, or braking will spin at different angular velocities, , for the same
ground velocity,
Usually the combination of and
that results in free rolling is not known in advance.
Since the steady-state transport analysis capability requires that both the rotational spinning velocity, , and
the traveling ground velocity, , be prescribed, the free rolling solution must be found in an indirect manner.
One such indirect approach is illustrated in this example. An alternate approach involves controlling the
rotational spinning velocity using user subroutine UMOTION while monitoring the progress of the solution
through a second user subroutine URDFIL. The URDFIL subroutine is used to obtain an estimate of the free
rolling solution based on the values of the torque at the rim at the end of each increment. This approach is
also illustrated in this example.
A nite element analysis of this problem, together with experimental results, has been published by
Koishi et al. (1997).
3.1.21

Printed on:
ROLLING TIRE
Problem description and model definition
A description of the tire and nite element model is given in Symmetric results transfer for a static tire
analysis, Section 3.1.1. To take into account the effect of the skew symmetry of the actual tire in the
dynamic analysis, the steady-state rolling analysis is performed on the full three-dimensional model, also
referred to as the full model. Inertia effects are ignored since the rolling speed is low (
10 km/h).
As stated earlier, the *STEADY STATE TRANSPORT capability in Abaqus uses a mixed
Eulerian/Lagrangian approach in which, to an observer in the moving reference frame, the material
appears to ow through a stationary mesh. The paths that the material points follow through the
mesh are referred to as streamlines and must be computed before a steady-state transport analysis can
be performed. As discussed in Symmetric results transfer for a static tire analysis, Section 3.1.1,
the streamlines needed for the steady-state transport analyses in this example are computed using
the *SYMMETRIC MODEL GENERATION, REVOLVE option.
This option generates the
three-dimensional mesh by revolving the two-dimensional tire cross-section about the symmetry axis
so that the streamlines follow the mesh lines.
The incompressible hyperelastic material used to model the rubber in this example includes a
time-domain viscoelastic component, which is enabled by the *VISCOELASTIC, TIME=PRONY
option. A simple 1-term Prony series model is used. For an incompressible material a 1-term Prony
series in Abaqus is dened by providing a single value for the shear relaxation modulus ratio, , and
its associated relaxation time, . In this example
= 0.3 and = 0.1. The viscoelastici.e., material
historyeffects are included in a *STEADY STATE TRANSPORT step unless the LONG TERM
parameter is used. See Time domain viscoelasticity, Section 19.7.1 of the Abaqus Analysis Users
Manual, for a more detailed discussion of modeling time-domain viscoelasticity in Abaqus.
Loading
As discussed in Symmetric results transfer for a static tire analysis, Section 3.1.1, it is recommended
that the footprint analyses be obtained with a friction coefcient of zero (so that no frictional forces
are transmitted across the contact surface). The frictional stresses for a rolling tire are very different
from the frictional stresses in a stationary tire, even if the tire is rolling at very low speed; therefore,
discontinuities may arise in the solution between the last *STATIC analysis and the rst *STEADY
STATE TRANSPORT analysis. Furthermore, varying the friction coefcient from zero at the beginning
of the steady-state transport step to its nal value at the end of the steady-state transport step ensures that
the changes in frictional forces reduce with smaller load increments. This is important if Abaqus must
take a smaller load increment to overcome convergence difculties while trying to obtain the steady-state
rolling solution.
Once the static footprint solution for the tire has been computed, the steady-state rolling contact
problem can be solved using the *STEADY STATE TRANSPORT option. The objective of the rst
simulation in this example is to obtain the straight line, steady-state rolling solutions, including full
braking and full traction, at different spinning velocities. We also compute the straight line, free rolling
solution. In the second simulation free rolling solutions at different slip angles are computed. In the rst
and second simulations material history effects are ignored by including the LONG TERM parameter on
3.1.22

Printed on:
ROLLING TIRE
the *STEADY STATE TRANSPORT steps. The third simulation repeats a portion of the straight line,
steady-state rolling analysis from the rst simulation; however, material history effects are included by
omitting the LONG TERM parameter. A steady ground velocity of 10.0 km/h is maintained for all the
simulations. The objective of the fourth simulation is to obtain the free rolling solution of the tire in
contact with a 1.5 m rigid drum rotating at 3.7 rad/s.
In the rst simulation (rollingtire_brake_trac.inp) the full braking solution is obtained in the rst
*STEADY STATE TRANSPORT step by setting the friction coefcient, , to its nal value of 1.0 using
the *CHANGE FRICTION option and applying the translational ground velocity together with a spinning
angular velocity that will result in full braking. The *TRANSPORT VELOCITY and *MOTION options
are used for this purpose. An estimate of the angular velocity corresponding to full braking is obtained
as follows. A free rolling tire generally travels farther in one revolution than determined by its center
height, H, but less than determined by the free tire radius. In this example the free radius is 316.2 mm
and the vertical deection is approximately 20.0 mm, so
296.2 mm. Using the free radius and the
effective height, it is estimated that free rolling occurs at an angular velocity between
8.78 rad/s
and
9.38 rad/s. Smaller angular velocities would result in braking, and larger angular velocities
would result in traction. We use an angular velocity
8.0 rad/s to ensure that the solution in the rst
steady-state transport step is a full braking solution (all contact points are slipping, so the magnitude of
the total frictional force across the contact surface is
).
In the second steady-state transport analysis step of the full model, the angular velocity is increased
gradually to
10.0 rad/s while the ground velocity is held constant. The solution at each load
increment is a steady-state solution to the loads acting on the structure at that instant so that a series
of steady-state solutions between full braking and full traction is obtained. This analysis provides us
with a preliminary estimate of the free rolling velocity. The second simulation (rollingtire_trac_res.inp)
performs a rened search around the rst estimate of free rolling conditions.
In the third simulation (rollingtire_slipangles.inp) the free rolling solutions at different slip angles
are computed. The slip angle, , is the angle between the direction of travel and the plane normal to
the axle of the tire. In the rst step the straight line, free rolling solution from the rst simulation is
brought into equilibrium. This step is followed by a *STEADY STATE TRANSPORT step where the
slip angle is gradually increased from
0.0 at the beginning of the step to
3.0 at the end of
the step, so a series of steady-state solutions at different slip angles is obtained. This is accomplished
by prescribing a traveling velocity vector with components
and
on the
0.0 in the rst steady-state transport step and
3.0
*MOTION, TRANSLATION option, where
at the end of the second steady-state transport step.
The fourth simulation (rollingtire_materialhistory.inp) includes a series of steady-state solutions
between full braking and full traction in which the material history effects are included.
The fth simulation (rollingtire_camber.inp) analyzes the effect of camber angle on the lateral thrust
at the contact patch under free rolling conditions.
The nal simulation in this example (rollingtire_drum.inp) considers a tire in contact with a rigid
rotating drum. The loading sequence is similar to the loading sequence used in the rst simulation.
However, in this simulation the translational velocity of the tire is zero, and a rotational angular velocity
is applied to the reference node of the rigid drum using the *TRANSPORT VELOCITY option. Since
a prescribed load is applied to the rigid drum reference node to establish contact between the tire and
3.1.23

Printed on:
ROLLING TIRE
drum, the rotation axis of the drum is unknown prior to the analysis. Abaqus automatically updates the
rotation axis to its current position if the angular velocity is dened using the *TRANSPORT VELOCITY
option. The rotational velocity of the rigid surface can also be dened using the *MOTION, ROTATION
option. In that case the position and orientation of the axis of revolution must be dened on the data
lines in the steady-state conguration and, therefore, must be known prior to the analysis. The position
and orientation of the axis are applied at the beginning of the step and remain xed during the step.
When the drum radius is large compared to the axle displacement, as in this example, it is a reasonable
approximation to dene the axle in the original conguration without signicantly affecting the accuracy
of the results.
Figure 3.1.21 and Figure 3.1.22 show the reaction force parallel to the ground (referred to as rolling
resistance) and the torque, T, on the tire axle at different spinning velocities. The gures compare
the solutions obtained for a tire rolling on a at rigid surface with those for a tire in contact with a
rotating drum. The gures show that straight line free rolling,
0.0, occurs at a spinning velocity
of approximately 9.0 rad/s. Full braking occurs at spinning velocities smaller than 8.0 rad/s, and full
traction occurs at velocities larger than 9.75 rad/s. At these spinning velocities all contact points are
slipping, and the rolling resistance reaches the limiting value
Figure 3.1.23 and Figure 3.1.24 show shear stress along the centerline of the tire surface in the
free rolling and full traction states for the case where the tire is rolling along a at rigid surface. The
distance along the centerline is measured as an angle with respect to a plane parallel to the ground passing
through the tire axle. The dashed line is the maximum or limiting shear stress, , that can be transmitted
across the surface, where p is the contact pressure. The gures show that all contact points are slipping
during full traction. During free rolling all points stick.
A better approximation to the angular velocity that corresponds to free rolling can be made by
using the results generated by rollingtire_brake_trac.inp to rene the search about an angular velocity of
9.0 rad/s. The le rollingtire_trac_res.inp restarts the previous straight line rolling analysis from Step 3,
Increment 8 (corresponding to an angular velocity of 8.938 rad/s) and performs a rened search up to
9.04 rad/s. Figure 3.1.25 shows the torque, T, on the tire axle computed in the rened search, which
leads to a more precise value for the free rolling angular velocity of approximately 9.022 rad/s. This
result is used for the model where the free rolling solutions at different slip angles are computed.
Figure 3.1.26 shows the transverse force (force along the tire axle) measured at different slip angles.
The gure compares the steady-state transport analysis prediction with the result obtained from a pure
Lagrangian analysis. The Lagrangian solution is obtained by performing an explicit transient analysis
using Abaqus/Explicit (discussed in Import of a steady-state rolling tire, Section 3.1.6). With this
analysis technique a prescribed constant traveling velocity is applied to the tire, which is free to roll along
the rigid surface. Since more than one revolution is necessary to obtain a steady-state conguration, ne
meshing is required along the full circumference; hence, the Lagrangian solution is much more costly
than the steady-state solutions shown in this example. The gure shows good agreement between the
results obtained from the two analysis techniques.
Figure 3.1.27 compares the free rolling solutions with and without material history effects included.
The solid lines in the diagram represent the rolling resistance (force parallel to the ground along the
3.1.24

Printed on:
ROLLING TIRE
traveling direction); and the broken lines, the torque (normalized with respect to the free radius) on
the axle. The gure shows that free rolling occurs at a lower angular velocity when history effects are
included. The inuence of material history effects on a steady-state rolling solution is discussed in detail
in Steady-state spinning of a disk in contact with a foundation, Section 1.5.2 of the Abaqus Benchmarks
Manual.
Figure 3.1.28 shows the camber thrust as a function of camber angle. The lateral force at zero
camber and zero slip is referred to as ply-steer and arises due to the asymmetry in the tire caused by the
separation of the belts by the interply distance. Discretization of the contact patch is responsible for the
non-smooth nature of the curve, and an overall camber stiffness of 44 N/degree is reasonably close to
expected levels.
Figure 3.1.29 shows the torque on the rim as the rotational velocity is applied with user subroutine
UMOTION, based on the free rolling velocity predicted in user subroutine URDFIL. As the torque on the
rim falls to within a user-specied tolerance of zero torque, the rotational velocity is held xed and the
step completed. Initially, when the free rolling rotational velocity estimates are beyond a user-specied
tolerance of the current rotational velocity, only small increments of rotational velocity are applied. The
message le contains information on the estimates of free rolling velocity and the incrementation as the
solution progresses. The angular velocity thus found for free rolling conditions is 9.026 rad/s.
Acknowledgments
SIMULIA gratefully acknowledges Hankook Tire and Yokohama Rubber Company for their cooperation
in developing the steady-state transport capability used in this example. SIMULIA thanks Dr. Koishi of
Yokohama Rubber Company for supplying the geometry and material properties used in this example.
Input files
rollingtire_brake_trac.inp
rollingtire_trac_res.inp
rollingtire_slipangles.inp
rollingtire_camber.inp
rollingtire_materialhistory.inp
rollingtire_drum.inp
rollingtire_freeroll.inp
rollingtire_freeroll.f
Three-dimensional full model for the full braking and

traction analyses.
Three-dimensional full model for the rened braking and
traction analyses.
Three-dimensional full model for the slip angle analysis.
Three-dimensional full model for the camber analysis.
Three-dimensional full model with material history
effects.
Three-dimensional full model for the simulation of rolling
on a rigid drum.
Three-dimensional full model for the direct approach to
nding the free rolling solution.
User subroutine le used to nd the free rolling solution.
Reference
Koishi, M., K. Kabe, and M. Shiratori, Tire Cornering Simulation using Explicit Finite Element
Analysis Code, 16th annual conference of the Tire Society at the University of Akron, 1997.
3.1.25

Printed on:
ROLLING TIRE
Drum
Road
Figure 3.1.21
Rolling resistance at different angular velocities.
Drum
Road
Figure 3.1.22
Torque at different angular velocities.
3.1.26

Printed on:
ROLLING TIRE
Lower shear limit

Upper shear limit
Shear stress
Figure 3.1.23
Shear stress along tire center (free rolling).
Shear stress
Lower shear limit
Upper shear limit
0.30
Shear stress (MPa)
0.20
0.10
0.00
0.10
0.20
0.30
70.00
80.00
90.00
100.00
110.00
Angle (degrees)
Figure 3.1.24
Shear stress along tire center (full traction).
3.1.27

Printed on:
ROLLING TIRE
Torque
Figure 3.1.25
Torque at different angular velocities (rened search).
Standard
Explicit
Transverse Force (kN)
2.4
2.0
1.6
1.2
0.8
0.4
0.0
0.5
1.0
1.5
2.0
2.5
Slip angle (degrees)
Figure 3.1.26
Transverse force as a function of slip angle.
3.1.28

Printed on:
3.0
ROLLING TIRE
Force - no convection
Force - convection included
Torque/R - no convection
Torque/R - convection included
Figure 3.1.27 Rolling resistance and normalized torque as

a function of angular velocity (R=0.3162 m).
Camber thrust
Figure 3.1.28
Camber thrust as a function of camber angle.
3.1.29

Printed on:
ROLLING TIRE
Torque on Rim
Figure 3.1.29 Torque on the rim for the direct approach

to nding the free rolling solution.
3.1.210

Printed on:
DYNAMIC TIRE
3.1.3
SUBSPACE-BASED STEADY-STATE DYNAMIC TIRE ANALYSIS
This example illustrates the use of the *STEADY STATE DYNAMICS, SUBSPACE PROJECTION option
to model the frequency response of a tire about a static footprint solution.
The *STEADY STATE DYNAMICS, SUBSPACE PROJECTION option (Subspace-based steady-state
dynamic analysis, Section 6.3.9 of the Abaqus Analysis Users Manual) is an analysis procedure that can be
used to calculate the steady-state dynamic response of a system subjected to harmonic excitation. It does so
by the direct solution of the steady-state dynamic equations projected onto a reduced-dimensional subspace
spanned by a set of eigenmodes of the undamped system. If the dimension of the subspace is small compared
to the dimension of the original problem (i.e., if a relatively small number of eigenmodes is used), the subspace
method can offer a very cost-effective alternative to a direct-solution steady-state analysis.
The purpose of this analysis is to obtain the frequency response of a 175 SR14 tire subjected to a harmonic
load excitation about the footprint solution discussed in Symmetric results transfer for a static tire analysis,
Section 3.1.1. The *SYMMETRIC RESULTS TRANSFER and *SYMMETRIC MODEL GENERATION
options are used to generate the footprint solution, which serves as the base state in the steady-state dynamics
calculations.
Problem description
A description of the tire being modeled is given in Symmetric results transfer for a static tire analysis,
Section 3.1.1. In this example we exploit the symmetry in the tire model and utilize the results transfer
capability in Abaqus to compute the footprint solution for the full three-dimensional model in a manner
identical to that discussed in Symmetric results transfer for a static tire analysis, Section 3.1.1.
Once the footprint solution has been computed, several steady-state dynamic steps are performed.
Both the *STEADY STATE DYNAMICS, DIRECT and the *STEADY STATE DYNAMICS,
SUBSPACE PROJECTION options are used. Besides being used to validate the subspace projection
results, the direct steady-state procedure allows us to demonstrate the computational advantage afforded
by the subspace projection capability in Abaqus.
Model definition
The model used in this analysis is essentially identical to that used in the rst simulation discussed
in Symmetric results transfer for a static tire analysis, Section 3.1.1, with CGAX4H and CGAX3H
elements used in the axisymmetric model and rebar in the continuum elements for the belts and
carcass. However, since no *STEADY STATE TRANSPORT steps are performed in this example,
the TRANSPORT parameter is not needed during the symmetric model generation phase. In addition,
instead of using a nonuniform discretization about the circumference, the uniform discretization shown
in Figure 3.1.31 is used.
The incompressible hyperelastic material used to model the rubber includes a viscoelastic
component described by a 1-term Prony series of the dimensionless shear relaxation modulus:
3.1.31

Printed on:
DYNAMIC TIRE
with relaxation coefcient

and relaxation time
. Since the material is incompressible,
the volumetric behavior is time independent. This time domain description of the material must be
written in the frequency domain to perform a *STEADY STATE DYNAMICS analysis. By applying a
Fourier transform, the expressions for the time-dependent shear modulus can be written in the frequency
domain as follows:
where
is the storage modulus,
is the loss modulus, and is the angular frequency. Abaqus
will perform the conversion from time domain to frequency domain automatically if the Prony series
parameters are dened with the *VISCOELASTIC, FREQUENCY=PRONY option. See Time domain
viscoelasticity, Section 19.7.1 of the Abaqus Analysis Users Manual, for a more detailed discussion on
frequency domain viscoelasticity.
Loading
The loading sequence for computing the footprint solution is identical to that discussed in Symmetric
results transfer for a static tire analysis, Section 3.1.1, with the axisymmetric model contained in
tiredynamic_axi_half.inp, the partial three-dimensional model in tiredynamic_symmetric.inp, and
the full three-dimensional model in tiredynamic_freqresp.inp. Since the NLGEOM=YES parameter
is active for the *STATIC steps used in computing the footprint solution, the steady-state dynamic
analyses, which are linear perturbation procedures, are performed about the nonlinear deformed shape
of the footprint solution.
The rst frequency response analyses of the tire are performed using the *STEADY STATE
DYNAMICS, SUBSPACE PROJECTION option. The excitation is due to a harmonic vertical load
of 200 N, which is applied to the analytical rigid surface through its reference node. The frequency
is swept from 80 Hz to 130 Hz. The rim of the tire is held xed throughout the analysis. Prior to
the subspace analysis being performed, the eigenmodes that are used for the subspace projection are
computed in a *FREQUENCY step. In the frequency step the rst 20 eigenpairs are extracted, for
which the computed eigenvalues range from 50 to 185 Hz.
The accuracy of the subspace analysis can be improved by including some of the stiffness associated
with frequency-dependent material propertiesi.e., viscoelasticityin the eigenmode extraction step.
This is accomplished by using the PROPERTY EVALUATION parameter with the *FREQUENCY
option. In general, if the material response does not vary signicantly over the frequency range of
interest, the PROPERTY EVALUATION parameter can be set equal to the center of the frequency span.
Otherwise, more accurate results will be obtained by running several separate frequency analyses over
3.1.32

Printed on:
DYNAMIC TIRE
smaller frequency ranges with appropriate settings for the PROPERTY EVALUATION parameter. In
this example a single frequency sweep is performed with PROPERTY EVALUATION=105 Hz.
The main advantage that the subspace projection method offers over mode-based techniques
(Mode-based steady-state dynamic analysis, Section 6.3.8 of the Abaqus Analysis Users Manual) is
that it allows frequency-dependent material properties, such as viscoelasticity, to be included directly in
the analysis. However, there is a cost involved in assembling the projected equations, and this cost must
be taken into account when deciding between a subspace solution and a direct solution. Abaqus offers
four different parameter values that may be assigned to the SUBSPACE PROJECTION option to control
how often the projected subspace equations are recomputed. These values are ALL FREQUENCIES, in
which new projected equations are computed for every frequency in the analysis; EIGENFREQUENCY,
in which projected equations are recomputed only at the eigenfrequencies; PROPERTY CHANGE, in
which projected equations are recomputed when the stiffness and/or damping properties have changed
by a user-specied percentage; and CONSTANT, which computes the projected equations only once
at the center of the frequency range specied on the data lines of the *STEADY STATE DYNAMICS
option. Setting SUBSPACE PROJECTION=ALL FREQUENCIES is, in general, the most accurate
option; however, the computational overhead associated with recomputing the projected equations at
every frequency can signicantly reduce the cost benet of the subspace method versus a direct solution.
The SUBSPACE PROJECTION=CONSTANT option is the most inexpensive choice, but it should be
chosen only when the material properties do not depend strongly on frequency. In general, the accuracy
and cost associated with the four SUBSPACE PROJECTION parameter values are strongly problem
dependent. In this example problem the results and computational expense for all four parameter values
for SUBSPACE PROJECTION are discussed.
The results from the various subspace analyses are compared to the results from a *STEADY STATE
DYNAMICS, DIRECT analysis.
Each of the subspace analyses utilizes all 20 modes extracted in the *FREQUENCY step. Figure 3.1.32
shows the frequency response plots of the vertical displacements of the roads reference node for the
direct solution along with the four subspace solutions using each of the SUBSPACE PROJECTION
parameter values discussed above. Similarly, Figure 3.1.33 shows the frequency response plots of the
horizontal displacement of a node on the tires sidewall for the same ve analyses. As illustrated in
Figure 3.1.32 and Figure 3.1.33, all four of the subspace projection methods yield almost identical
solutions; except for small discrepancies in the vertical displacements at 92 and 120 Hz, the subspace
projection solutions closely match the direct solution as well. Timing results shown in Table 3.1.31 show
that the SUBSPACE PROJECTION method results in savings in CPU time versus the direct solution.
Input files
tiredynamic_axi_half.inp
tiredynamic_symmetric.inp
tiredynamic_freqresp.inp
Axisymmetric model, ination analysis.

Partial three-dimensional model, footprint analysis.
Full three-dimensional model, steady-state dynamic
analyses.
3.1.33

Printed on:
DYNAMIC TIRE
tiredynamic_axi_half_ml.inp
tiredynamic_symmetric_ml.inp
tiredynamic_freqresp_ml.inp
Axisymmetric model, ination analysis with Marlow

hyperelastic model.
Partial three-dimensional model, footprint analysis with
Marlow hyperelastic model.
analyses with Marlow hyperelastic model.
Nodal coordinates for axisymmetric model.
Table 3.1.31 Comparison of normalized CPU times (normalized

with respect to the DIRECT analysis) for the frequency sweep from
80 Hz to 130 Hz and the *FREQUENCY step.
Normalized
CPU time
SUBSPACE PROJECTION=ALL FREQUENCIES
SUBSPACE PROJECTION=EIGENFREQUENCY
SUBSPACE PROJECTION=PROPERTY CHANGE
SUBSPACE PROJECTION=CONSTANT
DIRECT
*FREQUENCY
0.89
0.54
0.49
0.36
1.0
0.073
3
1
Figure 3.1.31
Uniform three-dimensional tire mesh.
3.1.34

Printed on:
DYNAMIC TIRE
Direct
All frequencies
Eigenfrequencies
Property change
Constant
Figure 3.1.32 Frequency response of the vertical road displacement due to a vertical
harmonic point load of 200 N applied to the reference node.
Direct
All frequencies
Eigenfrequencies
Property change
Constant
Figure 3.1.33
Frequency response of the horizontal sidewall displacement due to a vertical

harmonic point load of 200 N applied to the reference node.
3.1.35

Printed on:
TIRE SUBSTRUCTURE
3.1.4
STEADY-STATE DYNAMIC ANALYSIS OF A TIRE SUBSTRUCTURE
This example illustrates the use of the substructuring capability in Abaqus (Dening substructures,
Section 10.1.2 of the Abaqus Analysis Users Manual) to create a substructure from a tire under ination and
footprint loading. Use of tire substructures is often seen in vehicle dynamic analyses where substantial cost
savings are made using substructures instead of the whole tire model. Since tires behave very nonlinearly, it
is essential that the change in response due to preloads is built into the substructure. Here the substructure
must be generated in a preloaded state. Some special considerations for creating substructures with preloads
involving contact are also discussed.
A description of the tire model used is given in Symmetric results transfer for a static tire analysis,
Section 3.1.1. In this problem ination and fooptrint preloads are applied in a series of general
analysis steps identical to Symmetric results transfer for a static tire analysis, Section 3.1.1. The
*SYMMETRIC MODEL GENERATION and *SYMMETRIC RESULTS TRANSFER options are
used to exploit the symmetric nature of the structure and loading. Nodes in the bead area are tied to the
rigid body representing the rim.
The substructures retained nodes include the rim node and all nodes in the footprint. To enhance
the dynamic response of the substructure, these interfacial degrees of freedom are augmented with
generalized degrees of freedom associated with the rst 20 xed interface eigenmodes. Depending on
the nature of the loading, it may be necessary to increase the number of generalized degrees of freedom
to cover a sufcient range of frequencies. The extra cost incurred due to the addition of the extra
frequency extraction step is offset by the enhanced dynamic response of the substructure.
Loading
An ination load of 200 kPa is applied in the axisymmetric half-tire model contained in
substructtire_axi_half.inp. This is followed by a footprint load of 1650 N applied to the three-dimensional
half-tire model given in substructtire_symmetric.inp; and, subsequently, results are transferred to
the full tire model with the complete footprint load of 3300 N. All of these steps are run with the
NLGEOM=YES parameter, so all preload effects including stress stiffening are taken into account
when the substructure is generated.
To retain degrees of freedom that are involved in contact constraints at the footprint, it is necessary
to replace the contact constraints with boundary conditions. This is done once the footprint solution is
obtained by xing the retained nodes in the deformed state and using the *MODEL CHANGE option
to remove the contact pair between the footprint patch and the road surface. Without this change,
the contact constraints produce large stiffness terms in the substructure stiffness that can produce
non-physical behavior at the usage level. The mechanical response of the substructure is unchanged
since the *BOUNDARY, FIXED option is used to hold the tire in its deformed state. These boundary
3.1.41

Printed on:
TIRE SUBSTRUCTURE
conditions on the retained degrees of freedom are then released in the *SUBSTRUCTURE GENERATE
step, in which they are replaced with concentrated loads. To carry out these steps, it is necessary to
obtain the list of nodes in contact with the road. Hence, the substructure is generated in a restart analysis
following the analysis with the preloads. This makes it possible to construct the list of nodes that are
involved in contact with the road at the end of the preloading. It is necessary to specify the *MODEL
CHANGE option with the ACTIVATE parameter in the analysis prior to substructure generation to
enable the removal of the contact constraints.
To enhance the dynamic response of the substructure, several restrained eigenmodes are included
as generalized degrees of freedom. These restrained eigenmodes are obtained from a *FREQUENCY
step with all the retained degrees of freedom restrained. In this example the rst 20 eigenmodes,
corresponding to a frequency range of 50 to 134 Hz, are computed. With 67 nodes in the footprint, one
rim node with six degrees of freedom, and 20 generalized degrees of freedom, the substructure has 227
degrees of freedom.
At the usage level the nodes that form the footprint patch in the tire model are restrained to a single
node. The steady-state response of the substructure to harmonic footprint loading is analyzed over a
range of frequencies from 40 to 130 Hz.
A steady-state dynamic analysis of the substructure with *STEADY STATE DYNAMICS, DIRECT is
relatively inexpensive compared to running a similar analysis with the entire tire model. The results for
the frequency sweep are shown in Figure 3.1.41, which compares the response of the substructure to
the response of the entire tire model. All resonances in the tire model are captured by the substructure.
This result shows that, although the static response of the tire is used to condense the stiffness and the
mass for the retained degrees of freedom, relatively few generalized degrees of freedom can adequately
enhance the dynamic response of the substructure. However, the associated expense of calculating the
restrained eigenmodes must be taken into account when considering the total cost.
The tire model used for comparison in this example is the same as the model used in Symmetric
results transfer for a static tire analysis, Section 3.1.1, with one difference. Friction is activated in a
step prior to the *STEADY STATE DYNAMICS step to activate constraints in the contact tangential
direction on nodes in the footprint so that constraints equivalent to those applied on the footprint nodes
in the substructure model are produced.
Input files
substructtire_axi_half.inp
substructtire_symmetric.inp
substructtire_full.inp
substructtire_generate.inp
substructtire_dynamic.inp

Full three-dimensional model, nal equilibrium analysis.
Substructure generation analysis.
Usage level model with steady-state dynamics analysis.
3.1.42

Printed on:
TIRE SUBSTRUCTURE
Figure 3.1.41
Vertical response of the road node due to unit vertical harmonic load.
3.1.43

Printed on:
TIRE-AIR ACOUSTIC ANALYSIS
3.1.5
COUPLED ACOUSTIC-STRUCTURAL ANALYSIS OF A TIRE FILLED WITH AIR
The air cavity resonance in a tire is often a signicant contributor to the vehicle interior noise, particularly
when the resonance of the tire couples with the cavity resonance. The purpose of this example is to study
the acoustic response of a tire and air cavity subjected to an ination pressure and footprint load. This
example further demonstrates how the *ADAPTIVE MESH option can be used to update an acoustic mesh
when structural deformation causes signicant changes to the geometry of the acoustic domain. The effect
of rolling motion is ignored; however, the rolling speed can have a signicant inuence on the coupled
acoustic-structural response. This effect is investigated in detail in Dynamic analysis of an air-lled tire
with rolling transport effects, Section 3.1.9.
The acoustic elements in Abaqus model only small-strain dilatational behavior through pressure degrees
of freedom and, therefore, cannot model the deformation of the uid when the bounding structure undergoes
large deformation. Abaqus solves the problem of computing the current conguration of the acoustic
domain by periodically creating a new acoustic mesh. The new mesh uses the same topology (elements
and connectivity) throughout the simulation, but the nodal locations are adjusted periodically so that the
deformation of the structural-acoustic boundary does not lead to severe distortion of the acoustic elements.
The calculation of the updated nodal locations is based on adaptive mesh smoothing.
Problem description
A detailed description of the tire model is provided in Symmetric results transfer for a static tire
analysis, Section 3.1.1. We model the rubber as an incompressible hyperelastic material and include
damping in the structure by specifying a 1-term Prony series viscoelastic material model with a
relaxation modulus of 0.3 and relaxation time of 0.1 s. We dene the model in the frequency domain
using the *VISCOELASTIC, FREQUENCY=PRONY option since we want to include the damping
effects in a steady-state dynamics simulation.
The air cavity in the model is dened as the space enclosed between the interior surface of the tire
and a cylindrical surface of the same diameter as the diameter of the bead. A segment of the tire is shown
in Figure 3.1.51. The values of the bulk modulus and the density of air are taken to be 426 kPa and
3.6 kg/m3 , respectively, and represent the properties of air at the tire ination pressure.
The simulation assumes that both the road and rim are rigid. We further assume that the contact
between the road and the tire is frictionless during the preloading analyses. However, we use a nonzero
friction coefcient in the subsequent coupled acoustic-structural analyses.
Model definition
We use a tire model that is identical to that used in the simulation described in Symmetric results transfer
for a static tire analysis, Section 3.1.1. The air cavity is discretized using linear acoustic elements and
is coupled to the structural mesh using the *TIE option with the slave surface dened on the acoustic
domain. We model the rigid rim by applying xed boundary conditions to the nodes on the bead of the
3.1.51

Printed on:
tire, while the interaction between the air cavity and rim is modeled by a traction-free surface; i.e., no
boundary conditions are prescribed on the surface.
The *SYMMETRIC MODEL GENERATION and *SYMMETRIC RESULTS TRANSFER
options, together with a *STATIC analysis procedure, are used to generate the preloading solution,
which serves as the base state in the subsequent coupled acoustic-structural analyses.
In the rst coupled analysis we compute the eigenvalues of the tire and air cavity system. This
analysis is followed by a direct and a subspace projection steady-state dynamics analysis in which we
obtain the response of the tire-air system subjected to harmonic excitation of the spindle.
A coupled structural-acoustic substructure analysis is performed as well. Viscoelasticity material
is ignored in this simulation since the only form of damping the substructure can accurately represent
is Raleigh-type damping. The substructure is generated by retaining 50 eigenmodes and the structural
degrees of freedom at only two nodes: the road reference node and the wheel spindle. An equivalent
model without substructures is also included and serves as reference solution.
Loading
results transfer for a static tire analysis, Section 3.1.1. The simulation starts with an axisymmetric
model, which includes the mesh for the air cavity. Only half the cross-section is modeled. The ination
pressure is applied to the structure using a *STATIC analysis. In this example the application of pressure
does not cause signicant changes to the geometry of the air cavity, so it is not necessary to update the
acoustic mesh. However, we perform adaptive mesh smoothing after the pressure is applied to illustrate
that the updated geometry of the acoustic domain is transferred to the three-dimensional model when
symmetric results transfer is used.
The axisymmetric analysis is followed by a partial three-dimensional analysis in which the
footprint solution is obtained. The footprint load is established over several load increments. The
deformation during each load increment causes signicant changes to the geometry of the air cavity. We
update the acoustic mesh by performing 5 mesh sweeps after each converged structural load increment
using the *ADAPTIVE MESH option. At the end of this analysis sequence we activate friction between
the tire and road using the *CHANGE FRICTION option. This footprint solution, which includes the
updated acoustic domain, is transferred to a full three-dimensional model. This model is used to perform
the coupled analysis. In the rst coupled analysis we extract the eigenvalues of the undamped system,
followed by a *STEADY STATE DYNAMICS, DIRECT analysis in which we apply a harmonic
excitation to the reference node of the rigid surface that is used to model the road. We perform two
simulations: one in which the excitation is applied normal to the road surface and one in which the
excitation is applied parallel to the road surface along the rolling (fore-aft) direction.
We want to compute the response of the coupled system in the frequency range in which we expect
the air cavity to contribute to the overall acoustic response. We consider the response near the rst
eigenfrequency of the air cavity only, which has a wavelength that is equal to the circumference of the
tire. Using a value of 344 m/s for the speed of sound and an air cavity radius of 0.240 m (the average
of the minimum and maximum radius of the air cavity), we estimate this frequency to be approximately
230 Hz. We extract the eigenvalues and perform the steady-state dynamic analysis in the frequency
range between 200 Hz to 260 Hz. Stiffness contributions from the frequency-dependent viscoelasticity
3.1.52

Printed on:
material model are evaluated at a frequency of 230 Hz in the eigenvalue analysis. The PROPERTY
EVALUATION parameter on the *FREQUENCY option is used for this purpose.
The same ination and loading steps were used for the substructure analysis. The model is forced
up and down by a *BOUNDARY condition specied at the road reference node in the *STEADY STATE
DYNAMICS, DIRECT step and by an equivalent secondary *BASE MOTION in the *STEADY STATE
DYNAMICS, SUBSPACE PROJECTION step. The INTERVAL=RANGE parameter on the *STEADY
STATE DYNAMICS option is used to avoid computing unbounded responses at eigenfrequencies.
Figure 3.1.51 shows the updated acoustic mesh near the footprint region. The geometric changes
associated with the updated mesh are taken into account in the coupled acoustic-structural analyses.
The eigenvalues of the air cavity, the tire, and the coupled tire-air system are tabulated in
Table 3.1.51. The resonant frequencies of the uncoupled air cavity are computed using the original
conguration. We obtain two acoustic modes at frequencies of 227.98 Hz and 230.17 Hz. These
frequencies correspond to two identical modes rotated 90 with respect to each other, as shown in
Figure 3.1.52 and Figure 3.1.53; the magnitudes of the frequencies are different since we have used a
nonuniform mesh along the circumferential direction. We refer to the two modes as the fore-aft mode
and the vertical mode, respectively. These eigenfrequencies correspond very closely to our original
estimate of 230 Hz. The table shows that these eigenfrequencies occur at almost the same magnitude
in the coupled system, indicating that the coupling has a very small effect on the acoustic resonance.
The difference between the two vertical modes is larger than the difference between the fore-aft modes.
This can be attributed to the geometry changes associated with structural loading. The coupling has a
much stronger inuence on the structural modes than on the acoustic modes, but we expect the coupling
to decrease as we move away from the 230 Hz range.
Figure 3.1.54 to Figure 3.1.57 show the response of the structure to the spindle excitation.
Figure 3.1.54 and Figure 3.1.55 compare the response of the coupled tire-air system to the response
of a tire without the air cavity. Figure 3.1.56 and Figure 3.1.57 show the acoustic pressure measured
in the crown and side of the air. We draw the following conclusions from these gures. The frequencies
at which resonance is predicted by the steady-state dynamic analysis correspond closely to the
eigenfrequencies. However, not all the eigenmodes are excited by the spindle excitations. For example,
the fore-aft mode is not excited by vertical loading. Similarly, the vertical mode is not excited by
fore-aft loading. In addition, only some of the structural modes are excited by the spindle loads, while
others are suppressed by material damping. These gures further show that the air cavity resonance has
a very strong inuence on the behavior of the coupled system and that the structural resonance of the
coupled tire-air system occurs at different frequencies than the resonance of the tire without air. As
expected, this coupling effect decreases as we move further away from the cavity resonance frequency.
The eigenfrequencies obtained in the substructure analysis are identical to the eigenfrequencies
obtained in the equivalent analysis without substructures. The reaction force obtained at the road
reference node is also compared to the reaction force at the same node in the equivalent analysis without
substructures. As shown in Figure 3.1.58 the results for the two *STEADY STATE DYNAMICS steps
in the substructure analysis are virtually identical, and they compare well, in general, with the reaction
force obtained in the nonsubstructure analysis. The observed differences are due to the fact that only a
3.1.53

Printed on:
relatively small number of eigenmodes are used to generate the substructure; and, hence, the dynamics
of the substructure are not fully captured. Moreover, since there is no damping in these models
(viscoelastic material effects are ignored), the response at eigenfrequencies is innite. Consequently,
the reaction forces are less predictable near eigenfrequencies, which explains some of the differences
near the peaks.
Input files
tire_acoustic_axi.inp
tire_acoustic_rev.inp
tire_acoustic_re.inp
tire_acoustic_air.inp
tire_substracous_axi.inp
tire_substracous_rev.inp
tire_substracous_re.inp
tire_substracous_dyn_ref.inp
tire_substracous_gen.inp
tire_substracous_dyn.inp

Full three-dimensional model, coupled structuralacoustic analyses.
Nodal coordinates for the axisymmetric tire mesh.
Mesh data for the axisymmetric acoustic mesh.
Axisymmetric model,
ination analysis,
no
viscoelasticity.
Partial three-dimensional model, footprint analysis, no
viscoelasticity.
Full three-dimensional model, coupled structuralacoustic analyses, no viscoelasticity, static analysis
only.
Full three-dimensional model, coupled structuralacoustic analyses, no viscoelasticity, no substructures,
*FREQUENCY and *STEADY STATE DYNAMICS
analyses.
Full three-dimensional model, coupled structuralacoustic analyses, no viscoelasticity, substructure
generation analysis.
Substructure usage analysis (*FREQUENCY and
*STEADY STATE DYNAMICS analyses).
3.1.54

Printed on:
Table 3.1.51
Mode
Eigenfrequencies in the 200260 Hz frequency range.
Air cavity
1
2
3
4
5
6
7
8
9
10
11
12
Tire
Coupled tire-air
204.93
206.56
215.75
218.52
222.06
201.74
203.74
213.94
216.67
221.30
228.27
230.11
231.91
239.41
239.52
248.29
253.24
228.58
232.34
230.17
240.67
240.80
250.23
254.90
Step: Step-1,
Increment
1: FOOTPRINT (LOAD CONTROLLED)

1: Step Time =
1.000
Deformed Var: U
Figure 3.1.51
Deformation Scale Factor: +1.000e+00
Deformation of tire and air in the footprint region.
3.1.55

Printed on:
POR
+1.000e+00
+8.333e-01
+6.667e-01
+5.000e-01
+3.333e-01
+1.667e-01
+0.000e+00
-1.667e-01
-3.333e-01
-5.000e-01
-6.667e-01
-8.333e-01
-1.000e+00
Y
X
Z
3
2
1
Step: Step-3, 3: EXTRACT EIGENVALUES

Mode
6: Value = 2.05508E+06 Freq =
Primary Var: POR
Figure 3.1.52
228.16
(cycles/time)
Fore-aft acoustic mode.
POR
+1.000e+00
+8.357e-01
+6.715e-01
+5.072e-01
+3.430e-01
+1.787e-01
+1.448e-02
-1.498e-01
-3.140e-01
-4.783e-01
-6.425e-01
-8.068e-01
-9.710e-01
Y
X
Z
3
2
1
Step: Step-3, 3: EXTRACT EIGENVALUES

Mode
8: Value = 2.12655E+06 Freq =
Primary Var: POR
Figure 3.1.53
232.09
Vertical acoustic mode.
3.1.56

Printed on:
(cycles/time)
Acoustic resonance (232.09 Hz)

Coupled tire air system
Tire without air
Figure 3.1.54
Vertical road displacement due to vertical excitation load.
Acoustic resonance (228.16 Hz)

Coupled tire air system
Tire without air
Figure 3.1.55
Horizontal road displacement due to horizontal excitation load.
3.1.57

Printed on:
Crown
Side
Figure 3.1.56
Acoustic pressure due to vertical excitation load.
Crown
Side
Figure 3.1.57
Acoustic pressure due to horizontal excitation load.
3.1.58

Printed on:
no substructures - direct
substructure - direct
substructure - subspace
Figure 3.1.58
Vertical reaction force due to road displacement excitation.
3.1.59

Printed on:
IMPORT OF A STEADY-STATE ROLLING TIRE
3.1.6
Product: Abaqus/Explicit
This example illustrates the use of the results transfer capability in Abaqus (Transferring results between
Abaqus/Explicit and Abaqus/Standard, Section 9.2.2 of the Abaqus Analysis Users Manual) to import
results from an Abaqus/Standard steady-state transfer analysis to Abaqus/Explicit to simulate transient
rolling. Examples of loading during transient rolling include impact of a tire with an obstacle or vehicle
acceleration. This problem analyzes the impact of a tire with a curb. Because of the small stable time
increment necessary for the explicit dynamic procedure and the large time scales involved in simulating
quasi-static and steady-state loading, simulating quasi-static ination loading and steady-state rolling in
Abaqus/Standard provides signicant cost savings over performing these simulations in Abaqus/Explicit.
Moreover, the cost to obtain a steady-state rolling simulation in Abaqus/Explicit increases with rolling speed,
whereas in Abaqus/Standard the cost is independent of the magnitude of the rolling speed.
The model used in this example differs slightly from that used in Symmetric results transfer for a static
tire analysis, Section 3.1.1. Since only elements common to both Abaqus/Standard and Abaqus/Explicit
can be imported, reduced-integration solid elements are used in this example and nodes in the bead area
are attached to rigid elements representing the rim. A plot of the tire before impact with the curb, a
0.025 m high step, is shown in Figure 3.1.61. The large rigid body rotation involved in a transient
rolling analysis necessitates the use of the nondefault second-order-accurate kinematic formulation in the
explicit dynamic analysis. In addition, it is recommended that the enhanced hourglass control scheme
be used for both Abaqus/Standard and Abaqus/Explicit for all import analyses. Use section controls in
the axisymmetric analysis to specify the enhanced hourglass control scheme.
The ination and footprint preloads are applied in a series of general analysis steps identical to
those described in Symmetric results transfer for a static tire analysis, Section 3.1.1. Symmetric model
generation and symmetric results transfer are used to exploit the symmetric nature of the structure and
loading (see Symmetric model generation, Section 10.4.1 of the Abaqus Analysis Users Manual,
and Transferring results from a symmetric mesh or a partial three-dimensional mesh to a full threedimensional mesh, Section 10.4.2 of the Abaqus Analysis Users Manual).
The repeated dynamic impact of tire nodes as they come into contact with the road is an unavoidable
source of the high frequency noise that can be seen in the reaction force at the rim. The viscoelastic
property of the matrix provides sufcient damping to reduce such high frequency noise. No material
damping (mass proportional or stiffness proportional) is necessary in this case. The viscoelastic stress
contributions from the steady-state transport analysis are imported into the Abaqus/Explicit analysis.
Loading
An ination load of 200 kPa is applied to the axisymmetric half-tire model in importrolling_axi_half.inp.
This load is followed by a footprint load of 1650 N applied to the three-dimensional half-tire model
3.1.61

Printed on:
in importrolling_symmetric.inp, and subsequently results are transferred to the full tire model with the
complete footprint load of 3300 N.
The rolling analysis involves rolling the tire up to free rolling conditions. As in Steady-state rolling
analysis of a tire, Section 3.1.2, a translational velocity of 10 km/h is applied to the tire. The free rolling
velocity of the tire is determined in an independent analysis similar to the one described in Steady-state
rolling analysis of a tire, Section 3.1.2. The rolling velocity corresponding to free rolling conditions is
8.9759 rad/s. Inertia loads are accounted for in this example, since the objective is to import results to
a transient dynamic analysis in which inertial effects should be considered during impact with the curb.
Viscoelastic effects in the rubber matrix are also included, since viscolasticity is the primary damping
mechanism present in the tire.
During the transient dynamic analysis, the tire is moved forward with a prescribed velocity of
10 km/h and the vehicle load is applied to the rim reference node. The tire is allowed to rotate freely
about the axle. All other degrees of freedom at the road and rim reference nodes are held xed.
Results are imported and the analysis begins at time t = 3 sec. A plot of the rotational velocity at the rim
in Figure 3.1.62 shows that the oscillations at the beginning of the analysis are well within acceptable
limits. Impact with the curb is initiated after about a quarter rotation of the tire subsequent to import, at
approximately t = 3.13 sec. The vertical response of the rim reference node can be seen in Figure 3.1.63
and shows oscillations in the tire after impact. The contact patch and the stresses in the belts under
steady-state rolling conditions in the transient dynamic solution compare well with the direct steady-state
solution from Abaqus/Standard. The pressure in the footprint can be seen in Figure 3.1.64, and the
stresses in the belts are plotted in Figure 3.1.65. A plot of the shear stress during impact in Figure 3.1.66
shows that, as expected, the maximum stress is set up in the shoulder region.
Input files
importrolling_axi_half.inp
importrolling_symmetric.inp
importrolling_full.inp
importrolling_roll.inp
importrolling_xpl.inp

Half-symmetric three-dimensional model, ination and
footprint analysis.
Full three-dimensional model, ination and footprint
analysis.
Steady-state, free rolling solution.
Import and transient dynamic analysis.
3.1.62

Printed on:
Figure 3.1.61
Tire before impact with the curb.
Figure 3.1.62
Rotational velocity at rim.
3.1.63

Printed on:
Figure 3.1.63
Vertical response of the rim reference node.
Figure 3.1.64
Pressure in the footprint patch.
3.1.64

Printed on:
Figure 3.1.65
Figure 3.1.66
Stress in the belts.
Shear stress in the shoulder region during impact with the curb.
3.1.65

Printed on:
TIRE ANALYSIS WITH MULLINS EFFECT
3.1.7
ANALYSIS OF A SOLID DISC WITH MULLINS EFFECT AND PERMANENT SET
Products: Abaqus/Standard
Abaqus/Explicit
This example illustrates the use of the *MULLINS EFFECT option to model the static and steady-state
rolling interaction between a solid rubber disc and a rigid surface. The *MULLINS EFFECT option
(Mullins effect in rubberlike materials, Section 19.6.1 of the Abaqus Analysis Users Manual) is used with
the *HYPERELASTIC option to model the phenomenon of stress softening upon unloading from a certain
deformation level that is observed in certain lled elastomers. A variation of this example includes modeling
of permanent set (Permanent set in rubberlike materials, Section 20.7.1 of the Abaqus Analysis Users
Manual) as well.
Problem description
This example is divided into three sections. The rst section involves calibration of experimental data
to determine the material coefcients for the Mullins effect. The second section describes the static
response of a solid disc subjected to cyclic deformation induced by contact with a at rigid surface that
represents the road. This kind of test is commonly carried out in the tire industry to investigate the
effect of stress softening on the load-deection behavior of tires. The third section is a continuation
of the problem described in the second section and studies the rolling solution of the deformed disc.
The rolling is modeled using both a Lagrangian approach that involves rotation of the disc mesh and the
steady-state transport capability in Abaqus/Standard (*STEADY STATE TRANSPORT) that obtains the
steady-state solution of the rolling disc using more of an Eulerian approach. This example also illustrates
the enhanced hourglass control capability and modeling of permanent set in Abaqus.
Calibration of material properties
The rst step in using the Mullins effect material model is calibration of test data. Figure 3.1.71 shows
a typical set of uniaxial tension test data for lled rubber. These data are representative of a class of
rubber materials that is used in the tire industry and have been provided by Cooper Tire and Rubber
Company. The bold curve labeled exp_mono represents the primary behavior of the rubber material
that would be obtained from a monotonic test to a certain deformation level. These data are provided
to Abaqus using the *UNIAXIAL TEST DATA option along with the *HYPERELASTIC, TEST
DATA INPUT option. In this example the Yeoh material model is used for calibrating the primary
material behavior. Reduced polynomial models, such as the Yeoh model, provide a better t when
limited test data are available, as is the case here. The unloading-reloading data, needed to calibrate
the Mullins effect coefcients, are provided as stabilized loading/unloading cyclic data from three
different maximum strain levels. Although the model for Mullins effect in Abaqus predicts unloading
and reloading from a given maximum strain level to occur along a single curve, the real material
often shows a cyclic behavior. In fact, the unloading/reloading cycles from a given maximum strain
level also show evidence of progressive damage during the rst few cycles. However, after a few
cycles the behavior stabilizes. The unloading-reloading data can be provided to Abaqus either in the
3.1.71

Printed on:
form of data points for all loading/unloading cycles from different maximum strain levels or in the
form of data points for just the stabilized cycles. In this example the stabilized cycle has been chosen
because the structure considered is expected to undergo repeated cyclic loading. The curves labeled
exp_unload1, exp_unload2, and exp_unload3 represent the stabilized unloading/reloading
cycles for maximum nominal strain levels of 0.099, 0.26, and 0.51, respectively. The above data are
input using the *UNIAXIAL TEST DATA option three times (one for each maximum strain level)
along with the *MULLINS EFFECT, TEST DATA INPUT option.
For models that include permanent set, the data are specied using the *PLASTIC option so that a
softer response (compared to a purely hyperelastic material) is obtained in loading. These models also
include the Mullins effect material model as discussed above.
During design studies the analyst may wish to calibrate the Mullins effect parameters using the
loading part of the stabilized cycle, the unloading part of the stabilized cycle, or an average of the two.
This can be accomplished easily by creating three data les that include the loading, unloading, and
both the loading and unloading parts of the stabilized cyclic data, respectively. Subsequently any one of
these les can be referenced from the input le by using the *INCLUDE option. In this example, which
considers cyclic loading instead of primarily monotonic loading, both loading and unloading data have
been used for calibration; hence, an average behavior is studied.
Figure 3.1.72 shows the calibrated response of Abaqus along with the test data. The continuous
bold line represents the numerical response for cyclic loading and unloading to the three maximum strain
levels for which test data have been provided (the response at the lowest maximum strain level is not
visible in the scale of this gure due to a relatively small amount of damage). As the gure illustrates,
the model in Abaqus approximates the stabilized cycle at each strain level with a single curve. The gure
also illustrates that the model does not capture the progressive damage during the rst few cycles at any
strain level. Thus, while the numerical results for unloading from a given strain level begin from the
primary curve, the maximum stress for the test results for the stabilized cycle may be somewhat below
the measured primary material behavior.
The numerical response can be obtained by carrying out a uniaxial loading/unloading test with a
single element. Alternatively, the numerical response for both the primary and the unloading-reloading
behavior can be obtained by using the *PREPRINT option with MODEL=YES and simply carrying out
a datacheck run. In the latter case the response computed by Abaqus is printed to the data (.dat) le
along with the experimental data. These tabular data can be plotted in Abaqus/CAE for comparison and
evaluation purposes. The primary material behavior can also be evaluated with the automated material
evaluation tools available in Abaqus/CAE.
Static response of a solid disc to cyclic deformation
A solid disc made out of the rubber material described in the earlier section is subjected to cyclic
deformation. The coefcients for the Yeoh model determined during the calibration are used along with
a value of
5.E06 to introduce a small amount of compressibility in the material. This value of
is obtained based on the measured value of the initial bulk modulus of the rubber. The Mullins effect
model in Abaqus assumes that the damage is associated with the deviatoric behavior only. The disc has
an outer diameter of 3 inches, an inner diameter of 1.75 inches, and a thickness of 0.7 inches. The inner
surface of the disc is fully constrained. The outer surface is initially dened to be just touching a at
3.1.72

Printed on:
rigid surface. The coefcient of friction between the disc and the rigid surface is assumed to be zero.
During the analysis the rigid surface is pushed up 0.15 inches, brought back to its original position, then
again pushed up 0.22 inches before being pushed back to its original position. The above deformation
history constitutes two displacement-controlled deformation cycles. The analysis is carried out using
both Abaqus/Standard and Abaqus/Explicit. An axisymmetric model is created to dene the geometry
of the disc. Symmetric model generation (*SYMMETRIC MODEL GENERATION) is used to create
a three-dimensional disc model using rst-order reduced-integration bricks (C3D8R elements) with
enhanced hourglass control. The Abaqus/Explicit model is created by importing the model denition
from Abaqus/Standard. The rst step in the Abaqus/Standard analysis is a do-nothing step, which is
used to import the initial state into Abaqus/Explicit.
Figure 3.1.73 shows the force versus displacement at the rigid body reference node. The two
cycles in this gure correspond to the two deformation cycles discussed earlier. During the rst cycle the
unloading response is softer compared to the loading response due to damage associated with the Mullins
effect. During the second loading cycle the response is identical to the unloading segment of the rst
cycle until the displacement of 0.15 inches is reached. Beyond this point the response is a continuation
of the original loading segment of the rst cycle. Thus, the load-displacement behavior is consistent with
the expected behavior due to the Mullins effect.
Figure 3.1.74 shows the time history of the energy dissipated in the whole model due to damage.
The dissipation increases during the loading segment of the rst cycle because the material undergoes
more and more damage as it is deformed. During the unloading segment of the rst cycle and the loading
segment of the second cycle up to a displacement of 0.15 inches, no additional damage occurs. As a
result, the total dissipation remains constant. For additional displacement beyond 0.15 inches, more
damage occurs. This results in further increase of the total damage energy. During the nal unloading
cycle the damage energy again remains constant.
The loading in the Abaqus/Explicit analysis is carried out using displacement boundary conditions
with an amplitude that uses the smooth step denition (*AMPLITUDE, DEFINITION=SMOOTH
STEP) to reduce the noise in the response. Thus, the time history of the displacement is different
between the Abaqus/Standard and the Abaqus/Explicit simulations, although the total amount of
displacement is identical in both analyses. As a result, the time history of the damage dissipation
between the two analyses shows some differences in the slope of the response. However, the total
dissipation at the end of each stage is identical in the two cases.
Rolling response of a solid disc
The geometry and material of the disc are identical to those described earlier except that the inner
surface of the disc is not totally constrained, as it is for the static problem. Instead, all the nodes on
the inner surface are attached to a node (axle node) located at the center of the disc using kinematic
coupling constraints. This facilitates the application of angular velocity or displacement to the axle
node to simulate rolling in the Lagrangian approach, as well as the measurement of reaction forces and
moments at the axle. The mesh is more rened for the rolling problem compared to the static problem.
In particular, two elements are used through the thickness of the disc. The rst step is a do-nothing step
to facilitate import of the initial state from Abaqus/Standard to Abaqus/Explicit. This step is followed
by a static step in which the rigid surface is pushed against the disc a distance of 0.15 inches. The next
3.1.73

Printed on:
step involves rolling of the loaded disc against the rigid surface, which is accomplished in two ways.
The rst is a Lagrangian analysis in which an angular velocity of 2.5 radians per second is applied to
the axle node of the disc. In this example the structure reaches a steady state after one full revolution.
No additional damage occurs in subsequent revolution cycles. Therefore, the total time is chosen such
that the disc undergoes two revolutions. A variation of the Lagrangian analysis includes modeling of
permanent set using the *PLASTIC option. In the second analysis the rolling is simulated using the
steady-state transport capability in Abaqus/Standard. Frictional and inertial effects are neglected in
both cases.
The steady-state transport capability directly obtains the steady-state rolling solution of the disc on
the rigid surface. Due to Mullins effect the stress state for the rolling solution can be quite different
from the stress state for the static non-rolling solution. As a result, an attempt to obtain a steady-state
rolling solution directly from a static non-rolling solution may lead to convergence problems in the
Newtons scheme that is used to solve the overall nonlinear system of equations. Since the damage and,
hence, the discontinuity in state are independent of the angular rolling speed, a time increment cutback
during the steady-state transport step does not overcome the convergence difculties. Such convergence
difculties can be resolved by introducing the damage gradually over an additional steady-state transport
step preceding the actual analysis. In this example this is accomplished by following the static loading
step with a steady-state transport step with a small rolling angular speed of 0.25 radians per second and
with the MULLINS parameter on the *STEADY STATE TRANSPORT option set equal to RAMP. This
step is followed by another steady-state transport step at an angular speed of 2.5 radians per second with
the MULLINS parameter set equal to STEP, which provides the solution we are interested in obtaining.
A Lagrangian simulation is also carried out in Abaqus/Explicit. The kinetic energy is monitored
to ensure that the problem remains essentially quasi-static. The revolutions of the disc are carried out
by applying a rotational displacement (corresponding to two full revolutions) at the axle node using an
amplitude with a smooth step denition (*AMPLITUDE, DEFINITION=SMOOTH STEP) to reduce
the noise in the response. The compressibility parameter,
, is chosen to be 5 10 5 , an order of
magnitude higher than the actual value, to obtain relatively higher time increments and, thus, a relatively
lower run time.
Figure 3.1.75 compares the time history of the reaction force at the axle node for the Lagrangian
and the steady-state rolling analyses in Abaqus/Standard. The results for the Lagrangian problem without
permanent set (curve labeled lagr) indicate that the reaction force increases during the loading step and
decreases during the rst revolution of the disc. The decrease in the reaction force is a result of lower
overall stresses due to damage in the material. During the second revolution of the disc the reaction
force remains constant as no additional damage occurs. The results of the Lagrangian problem including
permanent set (curve labeled fefp) show a softer behavior compared to the former problem and clearly
indicate the presence of permanent set in the footprint area as the solid disc rolls against the rigid surface.
The steady-state rolling results (labeled sst) show a gradual transition of the reaction force during the
rst steady-state transport step that ramps up the Mullins effect. During the second steady-state transport
step the reaction force remains constant at the value reached at the end of the prior step. This curve
also illustrates that the damage associated with the Mullins effect is independent of the angular speed of
rotation. The reaction force remains the same at angular speeds of both 0.25 and 2.5 radians per second.
3.1.74

Printed on:
If the Mullins effect is not applied gradually over the rst steady-state transport step, the discontinuity
between the rolling and the static states may lead to convergence difculties.
The same results can be observed from a different viewpoint in Figure 3.1.76, which shows the
reaction force as a function of the number of revolutions for both Abaqus/Standard and Abaqus/Explicit.
The reaction force decreases during the rst revolution and remains steady (except for the noise in the
Abaqus/Explicit analysis) during the second revolution.
Figure 3.1.77 shows a comparison of the time histories of the reaction moment at the axle node
between the Lagrangian and the steady-state rolling solutions. The Lagrangian results are labeled mlagr, while the steady-state rolling results are labeled m-sst. If the material were purely hyperelastic
(without damage), the contact forces would be symmetric about a plane normal to the rigid surface and
containing the axle; hence, no torque would be required to rotate the disc. However, as a result of the
damage associated with Mullins effect, the contact forces are not symmetrical because material particles
transition through the contact area during the very rst revolution. This leads to the reaction moment
during the rst revolution, as shown in the results for the Lagrangian analysis. The moment reduces to
zero during the second revolution. The steady-state rolling results do not include the transient solution of
the rst revolution; hence, they show a zero moment at all times. Figure 3.1.78 shows the same results
from a different viewpoint. In this gure the reaction moment is plotted as a function of the number of
revolutions for both Abaqus/Standard and Abaqus/Explicit.
Figure 3.1.79 shows a contour plot of the damage energy dissipated at material points at an instant
of time that corresponds to about three-quarters of the way into the rst revolution of the disc. The gure
indicates damage in the material that has already passed through the contact area and no damage in the
material that is yet to pass through the contact area. This corresponds to damage in about three-quarters of
the disc material. The remaining quarter is still undamaged, as it has not undergone any deformation yet.
The full disc will be damaged at the end of the rst revolution, and the damage state remains unchanged
during the second revolution.
The results are discussed in the individual sections above and clearly demonstrate the different effects
of damage in the material.
Input files
mullins_calibrate.inp
mullins_axi_tire.inp
mullins_full_tire.inp
mullins_full_tire_xpl.inp
mullins_axi_tire_ref.inp
mullins_full_tire_roll_lag.inp
Unit element test to calibrate the material model.

Axisymmetric model for the static non-rolling problem.
Full three-dimensional model for the static non-rolling
problem.
Full three-dimensional model for the static non-rolling
problem (quasi-static simulation using Abaqus/Explicit).
Rened axisymmetric model for the rolling problem.
Full three-dimensional model for the Lagrangian rolling
problem.
3.1.75

Printed on:
mullins_full_tire_roll_lag_xpl.inp
mullins_full_tire_roll_sst.inp
mullins_calibrate_testdata.inp
mullins_hepl_axi_tire_ref.inp
mullins_hepl_full_tire_roll_lag.inp

problem (quasi-static simulation using Abaqus/Explicit).
Full three-dimensional model for the steady-state rolling
problem.
Uniaxial test data for calibrating the Mullins effect
coefcients.
Axisymmetric reference model for rolling problem with
permanent set.
problem with permanent set.
3.1.76

Printed on:
exp_mono
exp_unload
exp_unload2
exp_unload3
Figure 3.1.71
Test data for calibrating the Mullins effect.
exp_mono
exp_unload1
exp_unload2
exp_unload3
resp
Figure 3.1.72
Calibration of Mullins effect.
3.1.77

Printed on:
Explicit
Standard
Figure 3.1.73
Force versus displacement for the static non-rolling solution.
Explicit
Standard
Figure 3.1.74
Whole model damage energy history for the static non-rolling solution.
3.1.78

Printed on:
280.
Reaction Force
240.
200.
160.
120.
lagr
fefp
80.
sst
40.
0.
0.0
2.0
4.0
6.0
8.0
Time
Figure 3.1.75
Time history of reaction force for the rolling solution.
Explicit
Standard
Figure 3.1.76
Reaction force versus number of revolutions.
3.1.79

Printed on:
m-lagr
m-sst
Figure 3.1.77
Time history of reaction moment for the rolling solution.
Explicit
Standard
Figure 3.1.78
Reaction moment versus number of revolutions.
3.1.710

Printed on:
Figure 3.1.79
Contour plot of damage energy.
3.1.711

Printed on:
TREAD WEAR SIMULATION
3.1.8
TREAD WEAR SIMULATION USING ADAPTIVE MESHING IN Abaqus/Standard
This example illustrates the use of adaptive meshing in Abaqus/Standard as part of a technique to model tread
wear in a steady rolling tire. The analysis follows closely the techniques used in Steady-state rolling analysis
of a tire, Section 3.1.2, to establish rst the footprint and then the state of the steady rolling tire. These
steps are then followed by a steady-state transport step in which a wear rate is calculated and extrapolated
over the duration of the step, providing an approximate consideration of the transient process of wear in this
steady-state procedure.
With some exceptions, noted here, the description of the tire and nite element model is the same as
that given in Import of a steady-state rolling tire, Section 3.1.6. Since the focus of this analysis is
tread wear, the tread is modeled in more detail. In addition, a linear elastic material model is used in the
tread region to avoid difculties advecting the hyperelastic material state during the adaptive meshing
procedure.
The axisymmetric half-model of the 175 SR14 tire is shown in Figure 3.1.81. The rubber matrix is
modeled with CGAX4 and CGAX3 elements. The reinforcement is modeled with SFMGAX1 elements
that carry rebar layers. An embedded element constraint is used to embed the reinforcement layers in the
rubber matrix. The tread is modeled with an elastic material of elastic modulus 6 MPa and Poissons ratio
0.49. The rest of the tire is modeled with the hyperelastic material model. The polynomial strain energy
potential is used with coefcients C10 =106 , C01 =0.0, and D1 =2 108 . The rebar layers used to model
the carcass bers are oriented at 0 to the radial direction and have an elastic modulus of 9.87 GPa. The
modulus in compression is set to 1/100th of the modulus in tension. The Marlow hyperelastic model is
used to specify the nominal stress-nominal strain data for such a material denition. The elastic modulus
in tension of the material of the belt bers is 172.2 GPa. The modulus in compression is set to 1/100th
of the modulus in tension. The bers in the belts are oriented at +20 and 20 with respect to the hoop
(circumferential) direction.
The three-dimensional model is created by rst revolving the axisymmetric half-model, using
symmetric model generation, by 360 to generate the partial three-dimensional model shown in
Figure 3.1.82. A focused mesh is applied at the footprint region. The partial three-dimensional
model is then reected about a line to generate the full three-dimensional model. The results are
then transferred from the end of the footprint simulation for the partial three-dimensional model (see
Figure 3.1.82).
Adaptive meshing limitations in tire wear calculations
The use of adaptive meshing necessarily places the following restrictions on the tire model used for this
example:
3.1.81

Printed on:
Cylindrical elements are currently not supported with adaptive meshing and are not used in this
model.
Adaptive meshing generally performs poorly with hyperelastic material models, due to the
inaccurate advection of deformation gradient state variables. The tread, therefore, is modeled with
an elastic material denition.
Embedded elements including rebar layers cannot be used within the adaptive mesh domain.
Adaptive mesh smoothing on free surfaces occurs in directions determined by features of the
element geometry, which may not always be consistent with or easily enable description of wear
directions. As a result, as discussed below, you will generally have to do additional work to
explicitly describe the direction of wear.
Loading
The analysis is conducted in ve stages, beginning with an axisymmetric model and ending with a full
three-dimensional model created using symmetric model generation. The rst four steps follow closely
those used in Steady-state rolling analysis of a tire, Section 3.1.2.
1. Axisymmetric ination: An ination pressure of 200 kPa is applied to the interior of the tire, while
symmetry conditions are imposed at the midplane.
2. Three-dimensional footprint analysis of the partial model: The axisymmetric half-model is revolved
around the axle axis.
3. Three-dimensional footprint analysis of the whole model: The partial three-dimensional model is
reected about a line to generate the full three-dimensional model.
4. Steady-state transport: The generated full model is then subjected to a steady-state transport analysis
at 32 km/h. An angular velocity of 25 rad/s is specied for the tire. These conditions correspond to
a state of braking. Inertia effects and viscoelasticity are taken into account during this step.
5. Tread wear simulation: The tread wear simulation is conducted in the nal step, in which the velocity
of the tire is held constant and the wear is computed from the frictional energy dissipated and applied
around the periphery of the tire. Inertia effects and viscoelasticity are taken into account during this
simulation as well. This step is run for a duration of 3.6 106 s, simulating 32,000 kilometers of
travel of the tire at 32 km/h.
The nal step uses a wear model, which predicts a wear, or surface ablation, rate based on the steady
rolling of the tire. We are interested in predicting tire conguration changes as a result of this wear rate;
hence, we must introduce some modeling assumptions that enable modeling of a transient effect in a
steady-state procedure.
The basic assumption made is that the steady-state transport step time can be interpreted as a realtime duration of rolling at the current angular velocity. We consider that the conguration changes due
to wear have only a minor effect on the rolling tire solution at any time; hence, the results remain valid in
a steady-state sense at each time through the step. With this assumption we can simultaneously consider
effects at two disparate time scales: the shorter tire revolution time scale and the longer tire life time
scale.
3.1.82

Printed on:
Wear model
To illustrate the wear process, a simple wear model is employed based on the assumption that the wear
rate is a linear function of the local contact pressure and slip rate. Although we can calculate these
quantities locally, due to the Eulerian formulation used in steady-state transport they must be applied
over tread streamlines to model the wear of the entire tire perimeter.
Wear rate calculation
The wear constitutive model employed for this simulation is a form of the Archard model,
where is the volumetric material loss, or wear, rate; k is a nondimensional wear coefcient; H is the
material hardness; P is the interface normal pressure; A is the interface area; and is the interface slip
rate. Here we can see that the terms
describe a frictional energy dissipation rate. For tire rubber
we assume a wear coefcient k = 103 and a material hardness H = 2 GPa.
The goal of the following development is an expression for a material recession, or ablation, rate
, which can be applied to a node to simulate wear. First, consider a ribbon around the tire, where the
centerline of this ribbon is dened by a sequence of nodes comprising one of the surface streamlines
on the tires tread. This centerline is then bounded on either side by the tributary region of the surface
associated with each node. The combination of all such stream ribbons then comprises the total surface
of the tire involved in tire-to-road contact interactions. We expect the wear to occur uniformly over this
stream ribbon; hence, we express a wear rate for the entire ribbon,
where t is the time and is the current conguration position. Since we are using the Eulerian steadystate transport procedure, we can now rewrite this expression in a time-independent form,
where s is a position along the streamline and

is the width of the stream ribbon at position s. We
can also express as a function of the local material recession rate,
Equating these two expressions in a discrete form results in the following expression, summed over a
streamline:
3.1.83

Printed on:
where
is a nodal ablation velocity and
is the nodal contact area. This equation implies that is
generally not uniform along the streamline, which follows as a consequence of the stream ribbon width
varying as it enters and leaves the tire footprint. Since, however, we are ablating nodes away from the
footprint region solely to maintain a reasonable general shape of the worn tire conguration, we will
accept the assumption of a uniform nodal ablation velocity. This enables the following expression for :
Using again the assumption that the variation in stream ribbon width can be neglected,
, and
recognizing that the nodal contact area
enables a simpler expression that does not require
the use of contact areas:
Wear process implementation
With this expression for wear rate in the form of a surface ablation velocity, the wear can now be applied in
a steady-state transport analysis. User subroutine UMESHMOTION is used to specify the ablation velocity
vectors at the nodes that are on the exterior surface of the tire. UMESHMOTION denes adaptive mesh
constraint velocities and is used in conjunction with adaptive meshing, a mesh smoothing technique
applied at the end of each converged increment. The ablation velocities specied through the user
subroutine are applied at the tread surface nodes, and adaptive mesh smoothing adjusts nodes in the
interior tread region to maintain a well-shaped mesh.
To accumulate wear quantities around each tread streamline, the nodal numbering scheme along
the streamlines must be recorded in the user subroutine. This record is made in a set of common block
variables. The common block records nodes that belong to node set NADAPT (Figure 3.1.84) and
that lie at the reference cross-section (0) of the full model. The common block variables also include
the node numbering offsets specied with the symmetric model generation REVOLVE and REFLECT
parameters, which, together with the reference cross-section denitions, completely describe the tread
surface node numbering. The following variables need to be dened in the external common block:
nStreamlines: The number of nodes at the reference section (full model) at which wear is
applied.
nGenElem: The number of sectors or element divisions along the streamline in the model.
nRevOffset: The node offset specied under *SYMMETRIC MODEL GENERATION,
REVOLVE.
3.1.84

Printed on:
nReflOffset: The node offset specied under *SYMMETRIC MODEL GENERATION,

REFLECT. (Set this parameter equal to zero if the model is not reected).
jslnodes: The array that contains the required nodal information for all nodes potentially
undergoing wear at the reference section. This is an array of size (2, nStreamlines). For each
streamline the rst component is the node number of the root node (node a in the following
discussion), which is the node on that particular streamline at the reference section. The second
component is the node that provides the wear direction (node b in the following discussion). This
second component is necessary only for the tread corner nodes. Set it equal to the number of the
node at the reference section that denes the wear direction. For nodes that are not located at tread
corners, set the second array component equal to zero. The wear will be applied opposite to the
local 3-direction for these nodes located away from tread corners.
The variables in the expression for wear are accessed from the analysis database using the utility
routines GETVRN and GETVRMAVGATNODE. P is accessed from output variable CSTRESS; from
variable CDISP; and the
are determined from streamline nodal coordinates, accessed from variable
COORD.
Wear motion directions
The wear rate, , is then applied as the components of the mesh constraint vector variable ULOCAL.
This variable is passed into the user subroutine with default mesh smoothing motions dened in a local
coordinate system ALOCAL, which reects a measure of the surface normal at the current node. The
3-direction is dened as the direction of the outward normal and is based on an average of element facet
normals near the node. Under most circumstances it is sufcient to describe wear as resulting in ablation,
or nodal recession, opposite to this direction. At the tread corners, however, this average normal does not
provide an accurate wear direction. The appropriate normal is shown in Figure 3.1.85 and is computed
as follows: Suppose a is a corner node on the tread. It is possible to identify a node b that lies along the
edge of the tread. In this case the wear direction is given by the vector ab. By knowing the coordinates of
a and b, the wear can be calculated in the global coordinate system and rotated into the local coordinate
system (ALOCAL) directions.
The tire model is run for a duration of 3.6 106 s, or 1000 hours, the equivalent of 32,000 kilometers
of operation at 32 km/h. Figure 3.1.86 shows the resulting tread prole including the effects of wear.
Figure 3.1.87 shows the footprint contact pressure distribution both for a new tire and one in the as-worn
conguration.
Input files
treadwear_axi.inp
treadwear_rev.inp
treadwear_re.inp
treadwear_roll.inp

Full three-dimensional model, footprint analysis.
Full three-dimensional model, steady rolling analysis.
3.1.85

Printed on:
treadwear_wear_straight.inp
Full three-dimensional model, steady rolling wear

analysis.
Full three-dimensional model, steady rolling wear
analysis with slip.
UMESHMOTION user subroutine.
treadwear_wear_slip.inp
treadwear.f
Reference
Archard, J. F., Contact and Rubbing of Flat Surfaces, Journal of Applied Physics, vol. 24,
pp. 981988, 1953.
Figure 3.1.81
Axisymmetric cross-section of the tire.
Figure 3.1.82
Partial three-dimensional model.
3.1.86

Printed on:
Figure 3.1.83
Full three-dimensional model.
Figure 3.1.84 Nodes belonging to node set NADAPT at a particular

sector (NADAPT includes all such nodes from all the sectors).
3.1.87

Printed on:
Figure 3.1.85
Figure 3.1.86
Wear directions at tread corners.
Tread prole, in the unworn and worn congurations.
3.1.88

Printed on:
Figure 3.1.87 Footprint contact pressure distribution, in

the unworn and worn congurations.
3.1.89

Printed on:
CONVECTIVE TIRE-AIR ACOUSTIC ANALYSIS
3.1.9
DYNAMIC ANALYSIS OF AN AIR-FILLED TIRE WITH ROLLING TRANSPORT

EFFECTS
This example extends the analyses of Coupled acoustic-structural analysis of a tire lled with air,
Section 3.1.5, to include rolling transport effects in the tire and air. The acoustic cavity is modeled as part of
an axisymmetric model, which is inated, revolved, reected, and deformed to obtain a footprint in a manner
consistent with the aforementioned example.
The purpose of this example is to examine the effect of steady-state rolling transport on the acoustic
response of the tire and air cavity, after it has been subjected to the ination pressure and footprint load. The
air cavity resonance in a tire is often a signicant contributor to the vehicle interior noise, particularly when
the resonance of the tire couples with the cavity resonance. This coupled resonance phenomenon, however,
is affected by the rotating motion in the uid and the solid.
Problem description
A detailed description of the tire model is provided in Symmetric results transfer for a static tire
analysis, Section 3.1.1. We model the rubber as an incompressible hyperelastic material. Viscoelasticity
in the material is ignored in this example.
The air cavity in the model is dened as the space enclosed between the interior surface of the tire
and a cylindrical surface of the same diameter as the diameter of the bead. A cross-section of the tire
model is shown in Figure 3.1.91. The values of the bulk modulus and the density of air are taken to be
426 kPa and 3.6 kg/m3 , respectively, and represent the properties of air at the tire ination pressure.
The simulation assumes that both the road and rim are rigid. We further assume that the contact
between the road and the tire is frictionless during the preloading analyses. However, we use a nonzero
friction coefcient in the subsequent coupled acoustic-structural analyses.
To assess the effect of rolling motion on the dynamics of the coupled tire-air system, we rst generate
dynamic results for the stationary tire. In a subsequent analysis the tire and air are set into rolling motion,
and corresponding dynamic results are obtained.
Model definition
We use a tire cross-section that is identical to that used in the simulation described in Symmetric
results transfer for a static tire analysis, Section 3.1.1. The air cavity is discretized using linear acoustic
elements and is coupled to the structural mesh using the *TIE option with the slave surface dened on
the acoustic domain. We model the rigid rim by applying xed boundary conditions to the nodes on
the bead of the tire, while the interaction between the air cavity and rim is modeled by a traction-free
surface; i.e., no boundary conditions are prescribed on the surface.
We rst create an axisymmetric mesh of half of the cross-section of the tire and air, then revolve
it into half-symmetry. The *SYMMETRIC MODEL GENERATION and *SYMMETRIC RESULTS
3.1.91

Printed on:
TRANSFER options, together with a *STATIC analysis procedure, are used to generate the preloading
solution, which serves as the base state in the subsequent coupled acoustic-structural analyses.
In the rst coupled analysis we reect the revolved tire and air model into a full three-dimensional
conguration. We then compute the real eigenvalues of the stationary tire and air cavity system. During
frequency extraction steps, xed boundary conditions are imposed automatically on the tire-road
interface in the contact normal direction. Fixed boundary conditions are also applied in the tangential
direction for points that are sticking. Points that are slipping are free to move in the tangential direction.
This analysis is followed by a direct steady-state dynamic analysis in which we obtain the response of
the tire-air system subjected to imposed harmonic motion of the road.
In the second coupled analysis the reected model is restarted from the rst step, after the footprint
equilibrium conguration had been established. The *STEADY STATE TRANSPORT procedure is used
to obtain the free-rolling state of the tire at 60 km/h. The corresponding magnitude of the *TRANSPORT
VELOCITY is determined independently in a separate analysis. In this step the *ACOUSTIC FLOW
VELOCITY option signies that the acoustic medium is also in rotational motion. It is assumed that the
air inside the tire rotates with the same angular velocity as the tire. The direct steady-state analysis, with
similar parameters to those used in the stationary case, is repeated.
Additional analyses using a substructure that was generated with the effect of acoustic ow velocity
are also included.
Loading
results transfer for a static tire analysis, Section 3.1.1. The simulation starts with an axisymmetric
model, which includes the mesh for the air cavity. Only half the cross-section is modeled. The ination
pressure is applied to the structure using a *STATIC analysis. In this example the application of pressure
does not cause signicant changes to the geometry of the air cavity, so it is not necessary to update the
acoustic mesh. However, we perform adaptive mesh smoothing after the pressure is applied to illustrate
that the updated geometry of the acoustic domain is transferred to the three-dimensional model when
symmetric results transfer is used.
The axisymmetric analysis is followed by a reectionsymmetric three-dimensional analysis in
which the footprint solution is obtained. The footprint load is established over several load increments.
The deformation during each load increment causes signicant changes to the geometry of the air cavity.
We update the acoustic mesh by performing ve mesh sweeps after each converged structural load
increment using the *ADAPTIVE MESH option. At the end of this analysis sequence we activate
friction between the tire and road using the *CHANGE FRICTION option. This footprint solution, which
includes the updated acoustic domain, is transferred to a full three-dimensional model. This model is
used to perform the coupled analysis. In the rst coupled analysis we extract the eigenvalues of the
undamped system, followed by a *STEADY STATE DYNAMICS, DIRECT analysis in which we apply
a harmonic excitation to the reference node of the rigid surface that is used to model the road.
In the stationary and rolling analyses we compute the response of the coupled system in the same
frequency range used in Coupled acoustic-structural analysis of a tire lled with air, Section 3.1.5
(200 to 260 Hz). This band includes the natural frequencies of the fore-aft and vertical acoustic modes,
at 225.67 Hz and 230.94 Hz, respectively.
3.1.92

Printed on:
The model is excited by a *BOUNDARY condition specied at the road reference node in the
*STEADY STATE DYNAMICS, DIRECT step. A small amount of stiffness proportional damping is
applied to the rubber to avoid computing unbounded response at the eigenfrequencies.
The characteristic frequencies of the coupled tire-air system are affected by the rolling motion. Generally,
we expect a mode observed in the stationary coupled tire-air system to convert to a pair of modes,
corresponding to forward and backward wave travel. This does not always occur in complex systems,
because the stationary modes are not all affected by the rolling motion to a similar degree. However,
the mode split described above can be observed for several of the structural modes and the fundamental
acoustic modes of the cavity. For an observer in a nonrotating reference frame attached to the axle of
the tire, similar to the reference frame used by the *STEADY STATE TRANSPORT procedure, these
modes appear as waves traveling clockwise and counterclockwise along the circumference of the tire.
The mode corresponding to forward wave travel increases in frequency, whereas the mode corresponding
to backward wave travel decreases in frequency. The modes of a stationary tire appear as static vibrations.
The resonant frequencies of the stationary case are computed using the real-valued frequency extraction
procedure (*FREQUENCY). The complex frequency procedure will yield almost identical results for
the stationary case, since the damping used in this example is relatively low. However, for the rolling
analysis the complex frequency procedure must be used to obtain accurate results using all of the element
contributions due to rotation. In Table 3.1.91 an example is shown of a pair of structural and acoustic
modes splitting into a pair of corresponding modes in the rolling case. The structural mode is a radial
mode of circumferential order two. If there is no footprint loading, the stationary case will predict
identical frequencies for these modes; however, the split would still be observed for a rotating tire. The
modes are shown in Figure 3.1.92, Figure 3.1.93, and Figure 3.1.94. Similar behavior is observed
for radial modes of higher order.
Figure 3.1.95, Figure 3.1.96, and Figure 3.1.97 show the response of the structure to the imposed
vertical motion at the spindle. Figure 3.1.95 compares the acoustic response, at the crown, of the coupled
tire-air system for the stationary case and the rolling case at 60 km/h. Figure 3.1.96 shows the fore-aft
reaction force, and Figure 3.1.97 shows the vertical reaction force.
These gures further show that the rolling motion of the solid has a very strong inuence on the
behavior of the coupled system and that the rolling motion of the air exerts a similarly strong effect in
the frequency range observed here. In particular, the resonances affecting the reaction force occur at
different frequencies for the stationary and rolling cases. The resonances observed in the reaction force
frequency response diagram also proliferate as rolling is introduced since waves traveling against and
with the direction of rolling propagate at different speeds with respect to the observer. For a stationary
tire, a vertical excitation to the road produces negligible fore-aft reaction forces. However, the fore-aft
reaction forces due to a vertical excitation are signicant in the case of a rolling tire.
The same effect can be shown when using a substructure generated with the *ACOUSTIC FLOW
VELOCITY option. The tire model with a coarser mesh is used following the same pattern of actions:
axisymmetric model followed by the revolution, reection, and the steady-state transport analysis for a
rolling tire. The substructure is used within the frequency range of 200250 Hz using a direct steady-state
3.1.93

Printed on:
dynamic procedure. The same reaction force split resonance can also be observed for the full nite
element model used in place of the substructure.
Input files
sst_acoustic_axi.inp
sst_acoustic_rev.inp
sst_acoustic_re.inp

Full three-dimensional model, coupled structuralacoustic analyses, no transport effects.
Full three-dimensional model, coupled structuralacoustic analyses, transport effects at moderate speed.
Nodal coordinates for the axisymmetric tire mesh.
Mesh data for the axisymmetric acoustic mesh.
Coarse axisymmetric model, ination analysis.
Coarse partial three-dimensional model.
Coarse full three-dimensional model, coupled structuralacoustic analyses, no transport effects.
Coarse full three-dimensional model, coupled structuralacoustic analyses, transport effects at moderate speed.
Coarse mesh data for the axisymmetric acoustic mesh.
Frequency extraction and substructure generation,
including acoustic ow velocity.
Direct steady-state analysis with substructure.
sst_acoustic_roll100.inp
tire_acoustic_air.inp
sst_acoustic_axi_sm.inp
sst_acoustic_rev_sm.inp
sst_acoustic_re_sm.inp
sst_acoustic_roll_sm.inp
P_tire_acoustic_air.inp
substracous_afv_sm_gen.inp
substracous_afv_sm_use.inp
Table 3.1.91
Mode description
Example of eigenvalue splitting.
Stationary coupled air-tire
Rolling coupled air-tire
Structural
(radial, circumferential
order 2)
95.467 Hz, 99.474 Hz
86.69 Hz, 103.77 Hz
Acoustic
(fundamental modes
of the acoustic cavity)
225.67 Hz, 230.94 Hz
218.48 Hz, 237.31 Hz
3.1.94

Printed on:
Figure 3.1.91
Cross-section of tire and air.
Y
X
Figure 3.1.92
Radial mode, stationary case, at 95.46 Hz.
3.1.95

Printed on:
Z
Y
Figure 3.1.93
Radial mode, rolling case, backward, at 86.7 Hz (value at angle 90).
Z
Y
Figure 3.1.94
Radial mode, rolling case, forward, at 103.8 Hz (value at angle 90).
3.1.96

Printed on:
Figure 3.1.95
Acoustic pressure at crown due to imposed vertical motion.
Figure 3.1.96
Fore-aft reaction force due to road displacement excitation.
3.1.97

Printed on:
Figure 3.1.97
Vertical reaction force due to road displacement excitation.
3.1.98

Printed on:
ACOUSTICS IN A CIRCULAR DUCT WITH FLOW
3.1.10
This example illustrates the use of the symmetric model generation and symmetric results transfer capabilities
to create an acoustic model and the use of the steady-state transport capability in Abaqus to specify a rotational
velocity in acoustic media and to analyze the acoustic eld subject to the effects of rotational ow. A tireshaped acoustic cavity is analyzed.
The model is a simplied version of the acoustic cavity in a vehicle tire. The cross-section of the
cavity is established using axisymmetric elements, and these elements are assigned material properties
corresponding to air. A dummy step is used to establish the various les used in the subsequent threedimensional analysis steps.
The complete toroidal cavity is created using *SYMMETRIC MODEL GENERATION, REVOLVE
to form a circumferentially uniform mesh of 60 segments around the circle. Steady-state dynamic,
eigenfrequency extraction, and complex frequency extraction procedures use the *ACOUSTIC FLOW
VELOCITY option to include the effects of the rotational ow.
The air cavity is analyzed at rest, at 28.0 radians per second (corresponding to 100 kilometers per
hour), and at 320 radians per second (corresponding to Mach 0.3 at the outer edge of the cavity).
The steady-state dynamic results are computed at frequencies between 150 and 370 Hz. For the
complex analyses, modes are requested between 45 and 1000 Hz.
Loading
In the time-harmonic dynamic analysis, a unit imaginary concentrated volumetric acceleration acoustic
load is applied to a pair of nodes at the base of the torus. No boundary conditions are applied.
In the real and complex frequency analyses, no acoustic loads or boundary conditions are applied.
Results from the steady-state dynamic analysis show the effect of the rotating ow on the acoustic
pressure eld. Strong effects in the eld develop as speed increases.
Results from the real frequency analysis show some effect of the rotating ow on the acoustic
pressure modes and the frequencies. However, real frequency analysis ignores the important convective
term in the formulation, which is complex-valued. Consequently, real frequency analysis of rotating
acoustic media should be regarded as a preliminary step for the complex frequency computation, and
the results of the real frequency analysis should not be assigned much physical signicance. Complex
frequency analysis uses the entire, unsymmetric complex acoustic element formulation. The complex
contributions are, however, antisymmetric; therefore, the resulting resonant frequencies computed by
the complex frequency procedure are pure imaginary; that is, there is no real part of the eigenvalues
associated with gain or loss of energy. This is consistent with the physics of acoustics in rotating media.
3.1.101

Printed on:
The effect of the ow on the eigenvalues is evident in the apparent splitting of the resonances as speed
increases.
Input files
exa_acrotowaxi.inp
exa_acrotow3dssdd.inp
Axisymmetric model.
analysis.
Full three-dimensional model, real and complex
frequency analysis.
exa_acrotow3dfreq.inp
POR
+6.500e+01
+5.958e+01
+5.417e+01
+4.875e+01
+4.333e+01
+3.792e+01
+3.250e+01
+2.708e+01
+2.167e+01
+1.625e+01
+1.083e+01
+5.417e+00
+0.000e+00
Step: Step-2, STEADY STATE DYNAMICS, SLOW ROLL

Increment
8: Frequency =
209.2
Primary Var: POR
Complex: Magnitude
Deformed Var: not set
Deformation Scale Factor: not set
Figure 3.1.101
Steady-state dynamic response at 209 Hz, very low ow velocity.
3.1.102

Printed on:
POR
+6.500e+01
+5.958e+01
+5.417e+01
+4.875e+01
+4.333e+01
+3.792e+01
+3.250e+01
+2.708e+01
+2.167e+01
+1.625e+01
+1.083e+01
+5.417e+00
+0.000e+00
2
3
1
Step: Step-6, STEADY STATE DYNAMICS, M = 0.3
Increment
8: Frequency =
209.2
Primary Var: POR
Complex: Magnitude
Deformed Var: not set
Deformation Scale Factor: not set
Figure 3.1.102
Steady-state dynamic response at 209 Hz, ow velocity of Mach 0.3.
3.1.103

Printed on:
100 km/h
Mach 0.3
0 km/h
4.40
4.00
3.60
3.20
Log10 of Acoustic Pressure
2.80
2.40
2.00
1.60
1.20
0.80
0.40
150.00
200.00
250.00
300.00
350.00
Frequency
Figure 3.1.103
Steady-state dynamic response at node 13204, ow velocities from slow to Mach 0.3.
3.1.104

Printed on:
VEHICLE ANALYSES
3.2
Vehicle analyses
Inertia relief in a pick-up truck, Section 3.2.1

Substructure analysis of a pick-up truck model, Section 3.2.2
Display body analysis of a pick-up truck model, Section 3.2.3
Continuum modeling of automotive spot welds, Section 3.2.4
3.21

Printed on:
INERTIA RELIEF TRUCK
3.2.1
INERTIA RELIEF IN A PICK-UP TRUCK
This example illustrates how to perform inertia relief (Inertia relief, Section 11.1.1 of the Abaqus Analysis
Users Manual) in a static analysis in Abaqus/Standard. The problem involves stopping a pick-up truck,
moving with an initial velocity of 50.0 km/h (13.89 m/s), by applying braking loads. Inertia relief is used here
to supply inertia forces in a static analysis that oppose the braking loads specied in the model. The solution
provides the rigid body deceleration and the static stresses in the pick-up truck. For comparison purposes a
dynamic analysis is performed with the same initial velocity and braking loads.
A 1994 Chevrolet C1500 pick-up truck (see Figure 3.2.11) is modeled using approximately 55,000
elements. The model was obtained from the Public Finite Element Model Archive of the National
Crash Analysis Center at George Washington University. The nite element model was converted into
an Abaqus/Standard input le, and several missing constraints were added to carry out the analyses.
The model consists of various partssuch as cabin, truck bed, doors, etc.which are meshed with
shell elements, three-dimensional beam elements, and three-dimensional solid elements. The parts are
attached with connector elements, coupling elements, and multi-point constraints.
The materials used in the truck model are idealized as elastic or elastic-plastic. Suitable adjustments
are made to the material properties to account for unmodeled features of various parts such as the internal
details of the engine, gearbox, etc. A summary of the material properties and the parts for which they
are used is given in Table 3.2.11 and Table 3.2.12.
Rigid body denitions are used for brakes and brake assemblies to take advantage of the high
stiffness of these parts relative to other parts. Connector elements are used to model kinematic constraints
governing relative motions between various parts (see Substructure analysis of a pick-up truck model,
Section 3.2.2, for details).
The nite element model of the truck is oriented such that the positive 1-direction goes from the rear
to the front of the truck, the positive 2-direction goes from the passenger (right-hand) side to the driver
(left-hand) side, and the positive 3-direction is upward. In this system the braking loads are applied at
the respective wheel spindles as concentrated forces in the negative 1-direction.
To simplify the analysis, normal contact between the tires and the road surface is modeled through
spring elements that have one node connected to the wheel spindle and the other node xed against
displacement in the 3-direction and kinematically constrained to the wheel spindle in the other directions.
Friction between the tires and the road surface is assumed to be nonexistent. This allows the truck to
translate freely in the 1- and 2-directions and rotate freely about the 3-direction; the constraints on the
spring nodes prevent translation in the 3-direction and rotation about the 1- and 2-directions.
3.2.11

Printed on:
Loading
A separate static analysis is performed to obtain the correct initial conguration and stress distribution
under the applied gravity load. The details of this analysis are explained in Substructure analysis of a
pick-up truck model, Section 3.2.2. This gives us the base state for the analysis of interest.
The total braking load for the truck moving at 13.89 m/s is computed by assuming the truck to
be a rigid body that comes to rest over a distance of 20 m after the brakes are applied. This gives a
deceleration of 4.82 m/s2 in the 1-direction for the truck. The total mass of the truck as computed from
the nite element analysis is 1.72 103 kg, which gives the total inertial force resisted by the brakes
(or braking load) as 8.30 kN. Assuming that the front brakes provide 75% of the total resistance and
the rear brakes provide the remaining 25%, the braking load for each of the front wheels is 3.11 kN and
the braking load for each of the rear wheels is 1.04 kN. The four braking loads applied to the truck are
balanced in a static analysis with an inertia relief load. The inertia relief load represents the dynamic
effects (not modeled otherwise in a static analysis) of a constant deceleration from the trucks travel
velocity to a complete stop.
Since the truck is free to translate in the 1- and 2-directions and rotate about the 3-direction, inertia
relief is performed in these three directions. The other directions are constrained by boundary conditions
as explained in the previous section.
For comparison purposes a transient dynamic analysis is also performed (after the initial static
equilibrium under gravity load) in which the truck is accelerated from zero velocity to the nal uniform
velocity of 13.89 m/s. This dynamic analysis step is followed by another dynamic analysis step in which
the braking loads are applied to bring the truck to a complete stop. The braking loads are ramped up
smoothly from zero to the maximum value over 0.5 seconds and then kept constant for 2.88 secondsthe
time required to bring the truck to rest from the initial velocity of 13.89 m/s with an average deceleration
of 4.82 m/s2 . To minimize the analysis time, substructures are used in the dynamic analysis for all
deformable parts except the chassis and suspension components, which are modeled as fully deformable
since they are the parts that show signicant stresses.
The results for inertia relief in the pick-up truck model with braking loads show that the truck decelerates
at 4.83 m/s2 in the 1-direction. The truck has an angular acceleration of 0.01 rad/s2 about the 3-direction
at the center of mass due to asymmetry in the distribution of mass. The vertical displacements at the
wheel spindles (the front wheel spindles dip about 0.7 mm, and the rear wheel spindles rise about 0.7 mm
without loss of contact between tires and the road surface) indicate that the truck pitches forward due to
the braking action. A plot of the Mises stress shown in Figure 3.2.12 indicates that the largest stresses
occur in the suspension components and the regions where the suspension components are connected to
the chassis. Plots of the active yield ag and equivalent plastic strains (not shown) indicate that there is
no plastic yielding in any part of the truck.
The results for the transient dynamic analysis indicate that the average deceleration after the full
braking load has been applied is around 4.94 m/s2 in the 1-direction and the average angular acceleration
about the 3-direction is 0.03 rad/s2 . The truck pitches forward with the front wheel spindles dipping
3.2.12

Printed on:
about 0.7 mm and the rear wheel spindles rising about 0.7 mm in the braking load step. The Mises stress
for the dynamic analysis, shown in Figure 3.2.13, shows a distribution similar to that obtained for inertia
relief. There is no plastic yielding in the chassis or suspension components.
Inertia relief relies on the assumption that the body undergoing loading is free to translate and rotate
as a rigid body. Therefore, no external or internal constraints are allowed in the free directions (with the
exception of the case where statically determinant boundary conditions are applied and all available
directions are considered inertia relief directions). In a complex model like the pick-up truck, with
various kinematic constraints and large geometry changes, it is necessary to ensure that the base state
for the step including inertia relief is converged to a tight residual tolerance. If it is not converged to a
tight tolerance, the out-of-balance forces and moments in the base state will act as internal constraints on
rigid body motions. Hence, such unequilibrated forces and moments may prevent a geometrically linear
or nonlinear analysis from converging. In this example the *CONTROLS option is used to tighten the
convergence tolerance in the gravity load step preceding the inertia relief step.
The comparison of results for inertia relief and dynamic analysis of the truck shows that inertia relief
is an inexpensive alternative to dynamic analysis for obtaining the steady-state response of a dynamic
system for certain loading situations. In this braking analysis, for example, the static analysis with inertia
relief runs about 10 times faster than the general deformable transient dynamic simulation.
Input files
irltr_brake_irl.inp
irltr_brake_dyn.inp
Static equilibrium analysis for gravity load and initial

stresses followed by inertia relief with braking loads.
Static equilibrium analysis for gravity load and initial
stresses followed by dynamic acceleration to uniform
velocity and deceleration with braking loads.
The model data are contained in multiple smaller les and referenced as *INCLUDE les in the main
input les. The *INCLUDE le names are given in Substructure analysis of a pick-up truck model,
Section 3.2.2.
3.2.13

Printed on:
Table 3.2.11
Elastic-Plastic
Material Name
Steel
Properties for elastic-plastic materials used in the truck model.

Properties
E (N/m2 )
2.1 1011
(N/m2 )
0.3
2.7 108
(kg/m3 )
Part Name
7.89 103
rail (chassis),
engine oil box,
radiator mounting,
fenders,
wheel housings,
cabin,
bed,
fan center,
fuel tank,
rear rim,
steering support,
battery tray,
seat track,
radiator outer
Steel
2.1 1011
0.3
3.5 108
7.89 103
engine mountings,
radiator mountings,
radiator,
fender mountings,
hood, doors,
cabin hinges
Plastic
2.8 109
0.3
4.5 107
1.2 103
fan cover
Glass
7.6 1010
0.3
1.38 108
2.5 103
windows, windshield
Plastic
Plastic
3.4 10
3.4 10
0.3
0.3
1.0 10
1.0 10
3.2.14

Printed on:
radiator side block
dashboard interior
1.1 10
7.1 10
Table 3.2.12
Elastic Material Name
Properties for elastic materials used in the truck model.

Properties
E (N/m2 )
Part Name
(kg/m3 )
Steel
2.1 1011
0.3
7.89 103
A-arm mountings,
fan, door lock beams,
headrest connector beams,
radiator mounting beams,
oil pan beams,
rear axle,
drive shaft,
steering,
A-arm-rim connectors,
A-arm-rail connectors,
bed-rail connector,
dashboard support,
steering column,
rail connector,
brakes,
gearbox CV joint,
front rim
Steel
1.2 1011
0.3
3.89 103
engine gearbox
Steel
10
2.1 10
0.3
1.82 10
Steel
2.1 1011
0.3
2.5 103
Rubber
Steel
2.461 10
11
door lock
3
tires
8.0598 10
0.3
3.5765 10
rear suspension
Steel
11
2.1 10
0.3
2.089 10
brake assembly
Steel
2.1 1011
0.3
6.911 103
brake assembly
Rubber-Metal Composite
Foam
2.1 10
0.323
engine front
11
2.1 10
2.0 10
0.3
1.96 10
battery
0.3
2.527 10
Foam
2.0 10
0.3
7.55 10
seat top
Foam
2.0 109
0.3
1.69 102
seat headrest
3.2.15

Printed on:
seat bottom
Figure 3.2.11
Finite element model of 1994 Chevrolet C1500 pick-up truck.
3.2.16

Printed on:
S, Mises
Multiple section points
(Ave. Crit.: 75%)
+1.948e+02
+1.785e+02
+1.623e+02
+1.461e+02
+1.298e+02
+1.136e+02
+9.738e+01
+8.115e+01
+6.492e+01
+4.869e+01
+3.246e+01
+1.623e+01
+8.470e-11
Figure 3.2.12 Mises stress in front chassis and Aarm components for inertia relief.
S, Mises
(Ave. Crit.: 75%)
+1.913e+02
+1.753e+02
+1.594e+02
+1.435e+02
+1.275e+02
+1.116e+02
+9.564e+01
+7.970e+01
+6.376e+01
+4.782e+01
+3.188e+01
+1.594e+01
+8.551e-10
Figure 3.2.13 Mises stress in front chassis and A-arm

components for dynamic analysis.
3.2.17

Printed on:
SUBSTRUCTURE ANALYSIS OF A PICK-UP TRUCK
3.2.2
SUBSTRUCTURE ANALYSIS OF A PICK-UP TRUCK MODEL
This example illustrates the use of the substructure capability in Abaqus to efciently simulate the vehicle
dynamics of a detailed pick-up truck model traveling over road bumps. The pick-up truck model geometry
described in Inertia relief in a pick-up truck, Section 3.2.1, is used in this example. The model is organized as
a collection of individual parts that are connected together. Twenty substructures are then created, one for each
part that may undergo large motions but for which it is reasonable to assume small-strain elastic deformation
(e.g., the chassis). Several parts that may deform nonlinearly (e.g., leaf springs for the rear suspension or the
stabilizer bar in the front) are modeled using the usual general nonlinear modeling options. Connection points
are created for each part using the *COUPLING option (in most cases). The parts are then attached together
using appropriate connector elements. A simplied CALSPAN tire model is used (UEL) to model the radial
forces in the tires. The vehicle is loaded statically by gravity, accelerated in a dynamic step on a at road, and
run over bumps. Without stress recovery in the substructures the substructure analysis runs an estimated 120
times faster than an equivalent analysis without substructures.
Geometry and materials
The pick-up truck model (1994 Chevrolet C1500) discussed here is depicted in Figure 3.2.21 riding
over antisymmetric bumps. The model geometry, element connectivity, and material properties are
obtained from the Public Finite Element Model Archive of the National Crash Analysis Center at
George Washington University. The materials used are described in Inertia relief in a pick-up truck,
Section 3.2.1.
The model is organized as a collection of individual parts connected together. Most parts that
undergo only small deformations in addition to a large rigid body motion are dened as substructures.
Substructures are created for the following parts: the chassis, each of the four A-arms for the front
suspension, each of the four wheels, the rear axle, the driveshaft, the engine/transmission, the cabin,
each of the two doors, the hood, the seat, the front bumper, the truck bed, and the fuel tank.
The number of retained nodes for each substructure is determined primarily by its connection points
with neighboring parts as illustrated in Figure 3.2.22 for the cabin substructure. There are twenty points
associated with this substructure that are used to connect the cabin to other parts in the model. There are
six retained nodes on the cabin bottom (connections to the chassis); three retained nodes for the hood
connections (two hinges and the hood lock); three retained nodes for each of the two door connections
(two hinges in the front and the door lock in the back); four retained nodes for the seat connections; and
one node retained at the center of mass of the vehicle used for yaw, pitch, and roll measurement purposes.
Several parts deform too much to be considered substructures and are modeled using regular
elements. The leaf springs in the back and the stabilizer bar in the front are both modeled with beam
elements. The front brake assemblies are modeled as rigid bodies.
The connections between parts are modeled using connector elements. JOIN and REVOLUTE
connectors are used to model the hinges between each of the following parts: the A-arms and chassis,
the doors and cabin, the hood and cabin, the wheels and knuckles, and the leaf springs and chassis.
3.2.21

Printed on:
CARTESIAN and CARDAN connections with appropriately dened *CONNECTOR ELASTICITY,

*CONNECTOR DAMPING, and *CONNECTOR FRICTION options are used to dene some of
the bushing connections (e.g., engine mounts). Two UNIVERSAL connectors are used to model the
driveshaft connections to the transmission in the front and to the differential in the back. BEAM
connectors are used to model rigid connections between parts. The *CONNECTOR MOTION option
applied to an AXIAL connector is used to lock (or open) the doors and the hood. The *CONNECTOR
MOTION option applied to a SLOT connector is used to specify the steering by moving the steering
rack. The struts are modeled using an AXIAL connection by specifying approximate nonlinear elasticity
and damping. Several of the suspension-related parts are shown in Figure 3.2.23.
The radial forces in the tires are modeled approximately using a simplied CALSPAN tire model
(Frik, Leister, and Schwartz, 1993) implemented via user subroutine UEL. A radial stiffness of 600 N/mm
is considered.
Models
For all analyses the gravity-loaded static equilibrium conguration is found rst. Since the given mesh
geometry corresponds to the gravity-loaded equilibrium position and data are not available for the prestress in the suspension springs and tires, the pre-stress forces have to be computed. To achieve this
end, a separate static stress analysis with articial properties for the suspension springs and articial
boundary conditions is rst performed, as follows. The vehicle is supported with boundary conditions in
the vertical direction at the four wheel spindles (where the tire UELs will be connected) and xed at the
center of mass to prevent in-plane rigid body motion (degrees of freedom 1, 2, and 6). The stiffnesses
of the suspension springs are increased articially by a thousand times in this independent analysis to
minimize deformation. The gravity load is then applied to obtain equilibrium stresses in the suspension
components and reaction forces at the wheel spindles.
In the analysis of interest (with realistic properties and boundary conditions), the stresses and
reaction forces obtained from the articial static step are used as initial stresses in the suspension spring
components and as pre-stress forces in the tires, respectively. A *STATIC gravity loading step is run to
obtain an equilibrium conguration. This equilibrium conguration differs only slightly from the given
initial geometry (the wheel spindles move laterally about two millimeters). Thus, the initial stress state
in the suspension springs and tires accurately represents static equilibrium, and the vehicle is ready for
dynamic loading.
The vehicle model is prescribed an initial velocity and then accelerated (0.5 g) to the desired velocity
(5 m/sec or 7 m/sec) in a *DYNAMIC step. Once the cruise velocity is achieved, the truck model is
run over symmetric or antisymmetric bumps (0.2 m high and 5.0 m long).
In Figure 3.2.24 a snapshot of the truck moving forward with a velocity of 7 m/sec (25.2 km/h) and
jumping over symmetric bumps is shown. The wheels loose contact with the ground and then land
again on the road (not shown).
More results are presented for the case when the truck is riding over antisymmetric bumps (see
Figure 3.2.21). Stresses are recovered for the lower left A-arm substructure and shown in Figure 3.2.25
3.2.22

Printed on:
when the front wheels have traveled 3.2 m over the bumps. The radial forces on the tires are shown in
Figure 3.2.26, beginning from the moment when the front tires are about to go over the bumps. A zero
radial force indicates that the tire is out of contact. The yaw, pitch, and roll angles recorded using a
CARDAN connector attached to the model at its center of mass are shown in Figure 3.2.27.
The advantage of using substructures instead of regular deformable elements becomes obvious when
the total times needed to complete these types of analyses are compared. A full analysis using regular
elements has not been performed since approximately ve CPU days would be necessary to complete
either of the two analyses discussed above. This total time was estimated by running a few increments,
estimating the time needed per iteration, and then multiplying the time per iteration by the total number
of iterations needed to complete the analysis. Using these estimates, the substructure analysis is up to
120 times faster than the regular mesh analysis, depending on the amount of recovery performed for each
substructure.
The abaqus substructurecombine execution procedure can combine model and results data from
two substructure output databases into a single output database. For more information, see Combining
output from substructures, Section 3.2.17 of the Abaqus Analysis Users Manual. To combine the
substructure output databases, at least one frame of eld output must be requested.
Input files
tr_entire_truck_in_phase.inp
tr_entire_truck_anti_phase.inp
tr_road_antiphase.inp
tr_road_inphase.inp
tr_all_nodes.inp
tr_parameters.inp
tr_materials.inp
tr_materials_plastic_irl.inp
tr_initial_stress.inp
tr_lock_doors_and_hood.inp
tr_substruct_recovery.inp
tr_brake_front_left.inp
tr_brake_front_right.inp
tr_parameters_inphase.inp
tr_parameters_antiphase.inp
tr_readme.inp
Substructure global analysis of the truck going over inphase bumps.

Substructure global analysis of the truck going over
antiphase bumps.
Denition of antiphase road prole.
Denition of in-phase road prole.
All node denitions.
*PARAMETER denitions.
Material denitions.
Material denitions.
Initial stress denition for the leaf springs.
*CONNECTOR MOTION that keeps the doors and the
hood locked.
Output denitions for the substructures.
*RIGID BODY denition of the front left brake assembly.
*RIGID BODY denition of the front right brake
assembly.
*PARAMETER denitions for the in-phase analysis.
*PARAMETER denitions for the antiphase analysis.
Description of procedure used to run the jobs.
User subroutine
exa_tr_radial_uel.f
UEL used to dene the tire model.
3.2.23

Printed on:
Substructure generation files
tr_chassis_gen.inp
tr_retained_chassis.inp
tr_susp_lower_arm_left_gen.inp
tr_susp_lower_arm_right_gen.inp
tr_susp_upper_arm_left_gen.inp
tr_susp_upper_arm_right_gen.inp
tr_rear_axle_gen.inp
tr_retained_rear_axle.inp
tr_engine_gen.inp
tr_driveshaft_gen.inp
tr_cabin_gen.inp
tr_retained_cabin.inp
tr_hood_gen.inp
tr_door_left_gen.inp
tr_door_right_gen.inp
tr_seat_gen.inp
tr_bed_gen.inp
tr_fuel_tank_gen.inp
tr_bumper_gen.inp
tr_wheel_back_left_gen.inp
tr_wheel_back_right_gen.inp
tr_wheel_front_left_gen.inp
tr_wheel_front_right_gen.inp
Chassis.
Retained nodes for the chassis.
Lower left A-arm.
Lower right A-arm.
Upper left A-arm.
Upper right A-arm.
Rear axle.
Retained nodes for the rear axle.
Engine and transmission.
Driveshaft.
Cabin and front fenders.
Retained nodes for the cabin.
Hood.
Left door.
Right door.
Seat.
Truck bed.
Fuel tank.
Front bumper.
Rear left wheel.
Rear right wheel.
Front left wheel.
Front right wheel.
Element definitions
tr_rear_susp_leaf_springs.inp
tr_stabilizer_elts.inp
tr_steering_rods_elts.inp
tr_chassis_elts.inp
tr_susp_lower_arm_left_elts.inp
tr_susp_lower_arm_right_elts.inp
tr_susp_upper_arm_left_elts.inp
tr_susp_upper_arm_right_elts.inp
tr_rear_axle_elts.inp
tr_engine_elts.inp
tr_driveshaft_elts.inp
tr_cabin_elts.inp
tr_hood_elts.inp
tr_door_left_elts.inp
tr_door_right_elts.inp
tr_seat_elts.inp
Rear leaf spring suspension.

Front stabilizer bar.
Steering rods and steering rack.
Chassis.
Lower left A-arm.
Lower right A-arm.
Upper left A-arm.
Upper right A-arm.
Rear axle.
Driveshaft.
Hood.
Left door.
Right door.
Seat.
3.2.24

Printed on:
tr_bed_elts.inp
tr_fuel_tank_elts.inp
tr_bumper_elts.inp
tr_wheel_back_left_elts.inp
tr_wheel_back_right_elts.inp
tr_wheel_front_left_elts.inp
tr_wheel_front_right_elts.inp
Truck bed.
Fuel tank.
Front bumper.
Rear left wheel.
Rear right wheel.
Front left wheel.
Front right wheel.
*MPC definitions
tr_chassis_mpc.inp
tr_engine_mpc.inp
tr_cabin_mpc.inp
tr_hood_mpc.inp
tr_door_left_mpc.inp
tr_door_right_mpc.inp
tr_seat_mpc.inp
tr_fuel_tank_mpc.inp
tr_bumper_mpc.inp
Chassis.
Hood.
Left door.
Right door.
Seat.
Fuel tank.
Front bumper.
Coupling definitions
tr_chassis_coup.inp
tr_susp_lower_arm_left_coup.inp
tr_susp_lower_arm_right_coup.inp
tr_susp_upper_arm_left_coup.inp
tr_susp_upper_arm_right_coup.inp
tr_rear_axle_coup.inp
tr_engine_coup.inp
tr_driveshaft_coup.inp
tr_cabin_coup.inp
tr_hood_coup.inp
tr_door_left_coup.inp
tr_door_right_coup.inp
tr_seat_coup.inp
tr_bed_coup.inp
tr_fuel_tank_coup.inp
tr_bumper_coup.inp
tr_wheel_back_left_coup.inp
tr_wheel_back_right_coup.inp
tr_wheel_front_left_coup.inp
tr_wheel_front_right_coup.inp
Chassis.
Lower left A-arm.
Lower right A-arm.
Upper left A-arm.
Upper right A-arm.
Rear axle.
Driveshaft.
Hood.
Left door.
Right door.
Seat.
Truck bed.
Fuel tank.
Front bumper.
Rear left wheel.
Rear right wheel.
Front left wheel.
Front right wheel.
Connector definitions
tr_conn_aarms_left.inp
Left A-arms.
3.2.25

Printed on:
tr_conn_aarms_right.inp
tr_conn_brake_left.inp
tr_conn_brake_right.inp
tr_conn_steering_rods.inp
tr_conn_stabilizer.inp
tr_conn_leaf_springs.inp
tr_conn_engine.inp
tr_conn_driveshaft.inp
tr_conn_cabin_to_chassis.inp
tr_conn_hood.inp
tr_conn_door_left.inp
tr_conn_door_right.inp
tr_conn_seat.inp
tr_conn_bed.inp
tr_conn_fuel_tank.inp
tr_conn_bumper.inp
tr_conn_wheels_back.inp
tr_conn_wheels_front.inp
Right A-arms.
Left front brake assembly.
Right front brake assembly.
Steering rods.
Stabilizer bar.
Leaf springs.
Driveshaft.
Cabin to chassis.
Hood.
Left door.
Right door.
Seat.
Truck bed.
Fuel tank.
Front bumper.
Rear wheels.
Front wheels.
Reference
Frik, S., G. Leister, and W. Schwartz, Simulation of the IAVSD Road Vehicle Benchmark
Bombardier Iltis with FASIM, MEDYNA, NEWEUL, and SIMPACK, in Multibody Computer
Codes in Vehicle System Dynamics, Ed. W. Kortum and R. S. Sharp, February 1993.
Figure 3.2.21
Substructure truck model going over antisymmetric (left-right) bumps.
3.2.26

Printed on:
retained nodes for the

hood connection

left door connection
chassis connection
(only two nodes shown)
Figure 3.2.22
Figure 3.2.23
Substructure mesh for the cabin.
Chassis, suspension, and power-train-related parts.
3.2.27

Printed on:

seat connection
(only two nodes shown)
Figure 3.2.24
Substructure truck model going over symmetric (left-right) bumps.
S, Mises
SNEG, (fraction = -1.0)
(Ave. Crit.: 75%)
+2.890e+02
+2.653e+02
+2.415e+02
+2.178e+02
+1.940e+02
+1.703e+02
+1.465e+02
+1.228e+02
+9.903e+01
+7.528e+01
+5.153e+01
+2.778e+01
+4.030e+00
Figure 3.2.25
Recovered stresses in lower left A-arm.
3.2.28

Printed on:
left-back
left-front
right-back
right-front
Figure 3.2.26
Tire radial forces for the truck going over antisymmetric bumps.
pitch
roll
yaw
Figure 3.2.27
Yaw, pitch, and roll for the truck going over antisymmetric bumps.
3.2.29

Printed on:
DISPLAY BODY ANALYSIS OF A PICK-UP TRUCK
3.2.3
DISPLAY BODY ANALYSIS OF A PICK-UP TRUCK MODEL
This example illustrates the use of the display body capability in Abaqus to simulate efciently the vehicle
dynamics of a detailed pick-up truck model driving over a symmetric road bump. The pick-up truck model
geometry is described in Inertia relief in a pick-up truck, Section 3.2.1. The model is organized as a
collection of several parts, twenty-two of which are converted to display bodies. The display bodies are
connected using connectors. Several parts that may deform nonlinearly (e.g., the leaf springs for the rear
suspension and the stabilizer bar in the front) are modeled using beam elements. The mass, center of mass,
and rotary inertia are computed for every part by running a datacheck analysis on separate models, each
containing one part. A mass and a rotary inertia element are then created at the reference point of the display
body. The reference point of each display body is connected to its attachment points using connector elements.
A simplied CALSPAN tire model is used to model the radial forces in the tires via user subroutine UEL. The
vehicle is loaded statically by gravity, accelerated in a dynamic step on a at road, and run over bumps.
The results are then compared to corresponding analyses with substructures as in Substructure analysis of a
pick-up truck model, Section 3.2.2.
The problem discussed here is a model of a pick-up truck (1994 Chevrolet C1500). The model
geometry, element connectivity, and material properties are obtained from the Public Finite Element
Model Archive of the National Crash Analysis Center at George Washington University. The materials
used are described in Inertia relief in a pick-up truck, Section 3.2.1.
The model is organized as a collection of individual parts connected together. Most parts that
undergo only small deformations in addition to a large rigid body motion are dened as display bodies;
their deformation can be neglected since only the overall motion of the vehicle is of interest. A display
body is created for each of the following parts: the chassis, each of the four A-arms of the front
suspension, each of the four wheels, the rear axle, the driveshaft, the engine/transmission, the cabin,
each of the two doors, the hood, the seat, the front bumper, the truck bed, and the fuel tank.
The number of attachment points for each display body is determined by its connection points with
neighboring parts. The display bodies of the front-wheel assembly are shown connected at attachment
nodes in Figure 3.2.31.
Several parts deform nonlinearly and are modeled using regular elements. For example, the leaf
springs in the rear and the stabilizer bar in the front are modeled with beam elements.
To create the display bodies with appropriate attachment points and realistic material properties, the
following steps are taken:
1. A datacheck analysis is run on every part with deformable elements to determine its mass, rotary
inertia, and center of mass.
2. An instance of this part is used to create a display body.
3.2.31

Printed on:
3. A reference node created at the center of mass for this part is also used as the reference node for the
corresponding display body.
4. A mass element and a rotary inertia element are created at the reference node.
5. The reference node is then connected to the attachment points of the part using either BEAM or
CARTESIAN and CARDAN connectors with stiff elastic properties.
The connections between the attachment points of two display bodies are modeled using connector
elements. JOIN and REVOLUTE connectors are used to model the hinges between each of the following
parts: the A-arms and chassis, the doors and cabin, the hood and cabin, the wheels and knuckles, and the
leaf springs and chassis. CARTESIAN and CARDAN connections with connector elasticity, connector
damping, and connector friction behaviors are used to dene some of the bushing connections (e.g.,
the engine mounts). Two UNIVERSAL connectors are used to model the driveshaft connections to the
transmission in the front and to the differential in the back. BEAM connectors are used to model rigid
connections between parts. In cases where the BEAM connectors form a closed loop, the model becomes
overconstrained. In such situations CARTESIAN and CARDAN connectors with elasticity are used
instead of some of the BEAM connectors. A connector motion is applied to an AXIAL connector to
lock (or open) the doors and the hood. For steering the vehicle, a connector motion is applied to a SLOT
connector and is used to specify the steering for the steering rack. The struts are modeled using an
AXIAL connection by specifying approximate nonlinear elasticity and damping.
The radial forces in the tires are modeled approximately using a simplied CALSPAN tire model
(Frik, Leister, and Schwartz, 1993) implemented via user subroutine UEL. A radial stiffness of 600 N/mm
is considered. Friction is not accounted for in this user subroutine.
Models
The analysis is performed using the technique described in Substructure analysis of a pick-up truck
model, Section 3.2.2.
In Figure 3.2.32 a snapshot of the truck moving forward with a velocity of 7 m/sec (25.2 km/h) and
going over a symmetric bump is shown. The wheels lose contact with the ground and then land again
on the road (not shown).
The pitch and roll angles recorded using a CARDAN connector attached to the model at its center
of mass are shown in Figure 3.2.33. The pitch angles are almost the same for the display body model
and the substructure model (see Substructure analysis of a pick-up truck model, Section 3.2.2) of a
truck going over a symmetric bump with a difference of 1 degree. The difference is due to the fact that
the display body model uses several CARTESIAN and CARDAN connectors with elasticity to remove
overconstraints. These connectors introduce some compliance. The roll angles are almost zero for both
models.
The advantage of using substructures or display bodies instead of regular deformable elements
becomes obvious when the total times needed to complete these types of analyses are compared. It
is estimated that a full analysis using regular elements would take approximately 5 days to complete.
3.2.32

Printed on:
It is estimated that the substructure analysis is up to 120 times faster than the regular mesh analysis,
depending on the amount of recovery performed for each substructure. If the total substructure generation
and analysis time for the substructure model is compared to the analysis time for the display body model,
the display body model runs 5 times faster than the substructure model. If only the analysis times are
compared for both models, the display body model runs 2.5 times faster than the substructure model.
Input files
tr_entire_truck_in_phase_db.inp
tr_parameters_db.inp
tr_materials_db.inp
tr_initial_stress_db.inp
tr_lock_doors_and_hood_db.inp
exa_tr_radial_uel_db.f
Display body model of the truck going over in-phase

bumps.
Parameter denitions.
All material denitions.
Initial stress denition for the leaf springs.
Connector motion that keeps the doors and the hood
locked.
UEL used to dene the tire model.
Reference
Frik, S., G. Leister, and W. Schwartz, Simulation of the IAVSD Road Vehicle Benchmark
Bombardier Iltis with FASIM, MEDYNA, NEWEUL, and SIMPACK, in Multibody Computer
Codes in Vehicle System Dynamics, Ed. W. Kortum and R. S. Sharp, February 1993.
Figure 3.2.31
Display bodies of the front-wheel assembly connected at attachment nodes.
3.2.33

Printed on:
Figure 3.2.32
Figure 3.2.33
Display body truck model going over symmetric (left-right) bumps.
Comparison of the pitch and roll angles for the display body and substructure
models of the truck going over symmetric bumps.
3.2.34

Printed on:
CONTINUUM MODELING OF AUTOMOTIVE SPOT WELDS
3.2.4
Objectives
This example demonstrates the following Abaqus features and techniques:
using three-dimensional continuum elements and intricate material models (elastic-plastic and
damage constitutive behavior) to reproduce experimentally observed load-displacement curves
(courtesy of BMW) of tested spot weld specimens; and
demonstrating how virtual experiments allow for the generation of load-displacement data of
structural components from readily available geometric and material data. The load-displacement
curves can be used subsequently in calibrating connector behavior (not discussed in this section)
for efcient use in large-scale models, such as full-vehicle analyses.
Since the material data are highly proprietary, the input les provided below contain ctitious
material data. The material data used in the input les were obtained from the actual material data by
subjecting it to a number of transformations that preserve the trends in the overall shape of the curves
without revealing the exact material behavior. Moreover, the stiffness of the testing machine referenced
in this section is ctitious. However, the results published in this example use the actual data for
comparison with the physical tests. For these reasons, you will not obtain the force-deection curves or
deformed congurations published in this section when you run the associated input les.
Application description
The use of spot welds for the bonding of metal sheets is an extremely common practice in the automobile
industry. The number of such bonds in a typical vehicle is on the order of several thousand. The
use of Abaqus connector elements to model spot welds in full-vehicle analyses leads to efcient nite
element models that are able to capture the structural response of these local features with optimal
computational effort; however, load-displacement curves required for the modeling of spot welds may
not be readily available. Furthermore, the number of experimental tests required for the proper calibration
of a complete set of spot weld pairs in a vehicle can be prohibitive since the mechanical response of these
local mechanisms is dependent on both the geometric data, such as the thickness of the metal plates and
the radius of the spot weld, and the material properties of the plates being welded. Virtual testing can
generate the necessary modeling parameters when experimental data are not available. In this example
we model a range of failure mechanisms typically observed in spot welds. These virtual experiments are
compared with laboratory-obtained data (courtesy of BMW).
Geometry
The geometry of the patented test specimens used (Hahn et al., 1996, and Hahn and Rohde, 2004) is
shown in Figure 3.2.41 and Figure 3.2.42. A single spot weld of radius 2.65 mm connects two steel
plates that are 1.4 mm thick. These steel plates are 50 mm long and are bent over radii of 4.0 mm.
3.2.41

Printed on:
Materials
All specimens used in this study are made of galvanized high-strength steel H340LAD+Z100. This
material behaves in an elastic-plastic manner during the initial loading stages. As the material is further
loaded, it can either display a ductile damage response (caused by growth and coalescence of voids) or
it can display a shear failure mechanism (caused by the formation and growth of cracks within shear
bands). For condentiality reasons, the material data published in the input les associated with this
section are ctitious. The data were obtained by transforming the actual material test data to preserve
the overall trends in the behavior without revealing the true material behavior.
Boundary conditions and loading
A photo of the testing machine (Hahn et al., 2000) is shown in Figure 3.2.43. In the physical tests
the vertical sides of the specimens were longer than those shown in Figure 3.2.41 and Figure 3.2.42,
and they were rigidly clamped in the clamping device (not shown) of the tensile testing machine. In
the analysis these clamping conditions are modeled by xing the bottom edges of the specimens and
imposing a constant velocity of 0.15 m/s along the top edges in the vertical global 3-direction. For
the physical tests and analysis, a hinge is located 400 mm from the top edges of the specimens along the
vertical global 3-direction allowing the xed edges to rotate about the horizontal axis. During the pull and
peeling tests the specimens are aligned so the top edges are initially orthogonal to the vertical direction,
whereas during the shear test the edges are initially parallel to the vertical direction. The combined
stiffness of the loading piston and restraints used in the analysis is 50 kN/mm in pull and peeling tests
and 35 kN/mm in shear tests. For condentiality reasons both values are ctitious, but they approximate
the compliance of the actual testing machine used in the physical tests.
Abaqus modeling approaches and simulation techniques
A total of 18 different simulations were performed corresponding to the three test cases (pull test, shear
test, and peeling test). Each of the simulations was solved with a coarse mesh and a ne mesh using
three different scaling factors for the thermal inuence modeling, as discussed below.
Summary of analysis cases
Case 1
Pull test.
Case 2
Shear test.
Case 3
Peeling test.
The sections that follow discuss analysis considerations that apply to all cases, except where noted
otherwise.
Mesh design
All simulations were performed with 8-node, linear brick, reduced integration elements (C3D8R). The
density of the meshes increases toward the center of the plate where most of the deformation occurs.
3.2.42

Printed on:
Each test was performed with a coarse mesh using four elements through the thickness of each plate and
a ne mesh using six elements through the thickness of each plate. The coarse mesh and ne mesh used
for the pull and shear tests are shown in Figure 3.2.44 and Figure 3.2.45, respectively. Figure 3.2.46
shows the initial undeformed conguration of the coarse mesh used in the peeling test simulation, while
Figure 3.2.47 shows the initial conguration of the ne mesh. The same meshes are used for the pull
and shear tests since they are based on the same model geometry.
Material model
For condentiality reasons, some but not all details of the Abaqus models used for constitutive behavior
and progressive damage analysis are discussed below. For guidelines on obtaining the material
parameters from experimental data, see Progressive failure analysis of thin-wall aluminum extrusion
under quasi-static and dynamic loads, Section 2.1.16.
Elastic-plastic behavior
Werner et al. (2004) have shown that correct modeling of the elastic-plastic deformation of spot welds
is a prerequisite for realistic predictions of subsequent failure mechanisms. The authors showed that
by taking into consideration material property changes in the weld nugget they could obtain different
failure modes involving peeling or shearing failure of the spot weld. Furthermore, they suggested the
use of hardness measurements as a possible indicator for the change in properties of the welded material.
The change in hardness between the center of a spot weld and at a distance far away from the center
depends on the material grades joined, as well as their thicknesses.
In this example we assume both the elastic and the plastic behavior to be isotropic with the yield
surface described by a Mises yield function (see Inelastic behavior, Section 20.1.1 of the Abaqus
Analysis Users Manual). Different hardening curves are considered to encapsulate thermal effects near
the spot weld. For simplicity, the specimen is partitioned into three zones corresponding to the different
thermal exposures observed in the vicinity of the spot weld during the welding process. Different scaling
factors for the stress-strain curves are used in the three zones as derived from hardness measurements.
The geometry of each zone can be prescribed according to the welding process parameters. The scaling
of the yield curve is accomplished with the use of a eld variable dened as constant throughout each
region, and we test three scaling magnitudes for each test. For condentiality reasons the scaling factors
given below are ctitious, but they reect the trends of the elastic-plastic behavior near the spot weld:
A baseline conguration, where the original material properties (no scaling) are assigned to the
specimens in all three zones. While this choice is not realistic, it provides an extreme solution for
comparison purposes.
A second conguration uses a scaling of 1.2 of the yield curve in Zone 1 and a scaling of 1.1 in
Zone 2.
A third conguration uses a scaling of 1.4 of the yield curve in Zone 1 and a scaling of 1.2 in Zone 2.
These three scaling factor congurations help us to understand the effect of the thermal inuence
zone in capturing the correct behavior, as discussed below.
3.2.43

Printed on:
Damage initiation and evolution
The failure of aluminum-alloy sheets and thin-walled extrusions results from one or more of the following
mechanisms (Hooputra et al., 2004): nucleation, growth, and coalescence of voids; shear bands; and
necking. Damage due to initiation, growth, and coalescence of voids leads to ductile failure in metals;
the formation of cracks within shear bands leads to shear failure. Abaqus offers phenomenological
damage initiation criteria for both of these mechanisms. The ductile criterion is specied by providing
the equivalent plastic strain at the onset of ductile damage as a function of stress triaxiality and strain
rate. Similarly, the shear criterion is specied by providing the equivalent plastic strain at the onset
of shear damage as a function of shear stress ratio and strain rate (see Damage initiation for ductile
metals, Section 21.2.2 of the Abaqus Analysis Users Manual). The actual damage initiation criterion
data were provided by BMW but for condentiality reasons, the data were transformed to preserve the
overall trends without revealing the actual material behavior.
Damage evolution occurs once the damage initiation criteria are satised and further loading is
applied. A plastic displacementbased linear damage evolution law is used for each damage initiation
criterion. The value of the plastic displacement at which the damage variable reaches 1.0 (complete
degradation) is taken as 0.1, based on data from independent base material testing. The default maximum
degradation rule is used, and the elements are removed from the mesh once the maximum degradation
has occurred (see Maximum degradation and choice of element removal in Damage evolution and
element removal for ductile metals, Section 21.2.3 of the Abaqus Analysis Users Manual). Damage
initiation and evolution are assumed to be the same in all three thermal inuence zones described above,
a simplifying modeling assumption.
Initial conditions
As discussed in the material model section above, a eld variable is used to scale the yield surface and
capture the thermal effects of the welding process on the yield strength.
Boundary conditions
The specimens are loaded by xing their bottom edges and imposing a constant velocity of 0.15 m/s
along the top edges in the vertical global 3-direction.
Constraints
The top and bottom edges are constrained by kinematic couplings to model the rigid clamping in the
testing devices. A TRANSLATOR connection is used to model the stiffness of the loading apparatus.
Interactions
General frictionless contact is dened between all surfaces, an appropriate approximation since friction
forces are small when compared with the forces in the weld nugget.
3.2.44

Printed on:
Analysis steps
All simulations consist of one explicit dynamic step. All analyses consider geometric nonlinearity and
utilize mass scaling to model quasi-static loading conditions (see Mass scaling, Section 11.7.1 of the
Abaqus Analysis Users Manual).
Output requests
Field output requests include the following quantities: displacement, stress, strain, element status,
and damage initiation criteria variables. The history output request consists of displacement, velocity,
acceleration, and reaction force at the reference points of the kinematic coupling constraints. Energy
output variables are requested for the entire model.
Discussion of results and comparison of cases
The results presented in this example compare the actual material data with the physical test data. For
condentiality reasons, the actual material data are not published in the associated input les. Hence,
you will not obtain the results below when you run the input les provided.
The Mises stress contour and nal deformed shapes of the pull, shear, and peeling tests with
the coarse mesh and baseline material (no scaling of the yield curve) are depicted in Figure 3.2.48,
Figure 3.2.49, and Figure 3.2.410, respectively. The nal deformed shape in pull and peeling tests
show good qualitative agreement with experimental results provided by BMW. Shear test results predict
failure of the spot weld rather than on the surrounding plate, regardless of the amount of scaling that
was applied to the yield curve. This behavior was seen in some but not all experimental results, and it
does not seem to inuence the overall bearing capacity of the structure in shear.
The load-displacement history obtained from the simulations is compared with the experimental
results in Figure 3.2.411 to Figure 3.2.416. A good match is observed for the pull and shear tests with
the yield curve scaled by 1.2 in Zone 1. Results obtained with the baseline material underpredict the
peak load capacity, and results obtained with the 1.4 scaling overpredict the peak load of the structures.
The thermal zone scaling does not have a signicant impact on the qualitative character of the loading
curves. Good mesh convergence is also observed, indicating that acceptable results can be obtained even
with the coarse mesh discretization.
Results obtained for the peeling test do not show such good agreement with the experimental results
provided to SIMULIA by BMW; even the baseline material simulation results overpredict the peak load
capacity. This overstiff behavior is shown even in the purely elastic levels of deformation (in very early
deformation stages), which indicates that modeling of the loading apparatus for this loading conguration
may be inexact (data not available).
In conclusion, the results from both the quasi-static spot weld connector simulations match the
experimental pull and shear data very well. Changes in the material properties induced by the welding
process are not essential to the capture of the peak loads during the spot weld failure.
3.2.45

Printed on:
Input files
Case 1a: Pull test, coarse mesh
spotweld_pu_cm.inp
spotweld_pu_cm_1p2.inp
spotweld_pu_cm_1p4.inp
Input le to create and analyze the model (baseline

material).
Input le to create and analyze the model (spot weld yield
surface scaled by a factor of 1.2).
Case 1b: Pull test, fine mesh
spotweld_pu_fm.inp
spotweld_pu_fm_1p2.inp
spotweld_pu_fm_1p4.inp

material).
Case 2a: Shear test, coarse mesh
spotweld_sh_cm.inp
spotweld_sh_cm_1p2.inp
spotweld_sh_cm_1p4.inp

material).
Case 2b: Shear test, fine mesh
spotweld_sh_fm.inp
spotweld_sh_fm_1p2.inp
spotweld_sh_fm_1p4.inp

material).
Case 3a: Peeling test, coarse mesh
spotweld_pe_cm.inp
spotweld_pe_cm_1p2.inp
spotweld_pe_cm_1p4.inp

material).
3.2.46

Printed on:
Case 3b: Peeling test, fine mesh
spotweld_pe_fm.inp
spotweld_pe_fm_1p2.inp
spotweld_pe_fm_1p4.inp

material).
Auxiliary files
spotweld_material.inp
spotweld_cm_part.inp
spotweld_cm_assembl.inp
spotweld_fm_part.inp
spotweld_fm_assembl.inp
spotweld_pe_cm_part.inp
spotweld_pe_cm_assembl.inp
spotweld_pe_fm_part.inp
spotweld_pe_fm_assembl.inp
Input le with material model data.

Input le with part data for Case 1a and Case 2a.
Input le with assembly data for Case 1a and Case 2a.
Input le with part data for Case 1b and Case 2b.
Input le with assembly data for Case 1b and Case 2b.
Input le with part data for Case 3a.
Input le with assembly data for Case 3a.
Input le with part data for Case 3b.
Input le with assembly data for Case 3b.
References
Abaqus Analysis Users Manual
Mass scaling, Section 11.7.1 of the Abaqus Analysis Users Manual

Progressive damage and failure, Section 21.1.1 of the Abaqus Analysis Users Manual
Abaqus Keywords Reference Manual
*DAMAGE INITIATION
*DAMAGE EVOLUTION
Abaqus Verification Manual
Progressive damage and failure of ductile metals, Section 2.2.20 of the Abaqus Verication
Manual
Hahn, O., J.R. Kurzok, and M. Oeter, Test specication for KS-2 specimen, Laboratory of
Materials and Joint Technology, University of Paderborn, 2000.
Hahn, O., and A. Rohde, Procedures to manufacture specimen and specimen clamping device,
Patent Nr. 19522247 B4, April 15, 2004.
Hahn, O., A. Rohde, and D. Gieske, Specimen and specimen clamping device for use in tensile
testing machines, Patent Nr. 19510366 C1, August 22, 1996.
Other
3.2.47

Printed on:
Hooputra, H., H. Gese, H. Dell, and H. Werner, A Comprehensive Failure Model for
Crashworthiness Simulation of Aluminium Extrusions, International Journal of Crashworthiness,
vol. 9, pp. 449463, 2004.
Werner, H., H. Hooputra, H. Dell, and H. Gese, A Phenomenological Failure Model for Sheet
Metals and Extrusions, Annual Review Meeting and Workshop, Impact and Crashworthiness
Laboratory, Massachusetts Institute of Technology, 2004.
25 mm
25 mm
12.4 mm
2.65 mm
12.4 mm
1
1.4 mm
50 mm
2
1.5 mm
4 mm
18 mm
2
Figure 3.2.41
1.4 mm
Pull and shear test specimen geometry.
3.2.48

Printed on:
25 mm
25 mm
9 mm
2.65 mm
12.4 mm
1
1.4 mm
50 mm
2
1.5 mm
4 mm
18 mm
2
1.4 mm
Figure 3.2.42
Peeling test specimen geometry.
1
2
Figure 3.2.43
Experimental setup.
3.2.49

Printed on:
Thermal Influence Zone 1

3
2
1
Figure 3.2.44
Pull and shear tests: coarse mesh and thermal inuence zones.

3
2
1
Figure 3.2.45
Pull and shear tests: ne mesh and thermal inuence zones.
3.2.410

Printed on:

Figure 3.2.46
Peeling test: coarse mesh and thermal inuence zones.

Figure 3.2.47
Peeling test: ne mesh and thermal inuence zones.
3.2.411

Printed on:
S, Mises
(Avg: 75%)
+6.518e+02
+6.002e+02
+5.485e+02
+4.968e+02
+4.452e+02
+3.935e+02
+3.418e+02
+2.902e+02
+2.385e+02
+1.868e+02
+1.351e+02
+8.348e+01
+3.181e+01
3
2
1
Figure 3.2.48
Pull test, coarse mesh, baseline material (no yield curve scaling): nal
deformed shape of specimen and Mises stress contours.
S, Mises
(Avg: 75%)
+7.213e+02
+6.613e+02
+6.013e+02
+5.413e+02
+4.813e+02
+4.213e+02
+3.613e+02
+3.013e+02
+2.413e+02
+1.813e+02
+1.213e+02
+6.129e+01
+1.290e+00
Figure 3.2.49
Shear test, coarse mesh, baseline material (no yield curve scaling): nal
deformed shape of specimen and Mises stress contours.
3.2.412

Printed on:
S, Mises
(Avg: 75%)
+6.507e+02
+5.970e+02
+5.432e+02
+4.894e+02
+4.357e+02
+3.819e+02
+3.281e+02
+2.743e+02
+2.206e+02
+1.668e+02
+1.130e+02
+5.925e+01
+5.476e+00
3
1
2
Figure 3.2.410 Peeling test, coarse mesh, baseline material (no yield curve scaling):
nal deformed shape of specimen and Mises stress contours.
Baseline Material
Yield scaled by 1.2
Yield scaled by 1.4
Experimental Result
Coarse Mesh
3
[x10 ]
12.00
Reaction Force (N)
10.00
8.00
6.00
4.00
2.00
0.00
0.00
5.00
10.00
15.00
Imposed Displacement (mm)
Figure 3.2.411 Pull test with coarse mesh: comparison of reaction force versus imposed displacement
for different yield surface scaling. Experimental result courtesy of BMW.
3.2.413

Printed on:
Baseline Material
Yield Scaled by 1.2
Yield Scaled by 1.4
Experimental Result
Fine Mesh
3
Reaction Force (N)
[x10 ]
12.00
8.00
4.00
0.00
0.00
5.00
10.00
15.00
Figure 3.2.412 Pull test with ne mesh: comparison of reaction force versus imposed displacement
Baseline Material
Yield scaled by 1.2
Yield scaled by 1.4
Experimental Result
Coarse Mesh
3
Reaction Force (N)
[x10 ]
10.00
5.00
0.00
0.00
4.00
8.00
Figure 3.2.413 Shear test with coarse mesh: comparison of reaction force versus imposed
displacement for different yield surface scaling. Experimental result courtesy of BMW.
3.2.414

Printed on:
Baseline Material
Yield scaled by 1.2
Yield scaled by 1.4
Experimental Result
Fine Mesh
3
Reaction Force (N)
[x10 ]
10.00
5.00
0.00
0.00
4.00
8.00
Figure 3.2.414 Shear test with ne mesh: comparison of reaction force versus imposed displacement
Baseline Material
Yield scaled by 1.2
Yield scaled by 1.4
Experimental Result
Coarse Mesh
3
[x10 ]
Reaction Force (N)
4.00
3.00
2.00
1.00
0.00
1.00
0.00
8.00
16.00
24.00
32.00
Figure 3.2.415 Peeling test with coarse mesh: comparison of reaction force versus imposed
displacement for different yield surface scaling. Experimental result courtesy of BMW.
3.2.415

Printed on:
Baseline Material
Yield scaled by 1.2
Yield scaled by 1.4
Experimental Result
Fine Mesh
3
[x10 ]
Reaction Force (N)
4.00
3.00
2.00
1.00
0.00
1.00
0.00
8.00
16.00
24.00
32.00
Figure 3.2.416 Peeling test with ne mesh: comparison of reaction force versus imposed displacement
3.2.416

Printed on:
OCCUPANT SAFETY ANALYSES
3.3
Occupant safety analyses
Seat belt analysis of a simplied crash dummy, Section 3.3.1

Side curtain airbag impactor test, Section 3.3.2
3.31

Printed on:
SEAT BELT ANALYSIS
3.3.1
SEAT BELT ANALYSIS OF A SIMPLIFIED CRASH DUMMY
Automotive seat belts dramatically reduce the risk of vehicle occupant injury in the case of collision. This
example illustrates the use of Abaqus in analyzing a simplied crash dummy restrained by a complete seat
belt system as a moving vehicle comes to a sudden stop. The seat belt system includes:
the webbing sliding through D-rings attached to the car frame;

one retractor that models spool locking, tension preloading, and spool effects; and
one pretensioner that tightens the seat belt.
Several SLIPRING connectors strung together are used to model the belt webbing passing through the rings.
The RETRACTOR and HINGE connection types are used to model the retractor and pretensioner.
As shown in Figure 3.3.11, the model consists of several distinct entities: the dummy, the seat belt, the
retractor, and the seat. The modeling strategy consists of applying an initial velocity for the dummy of
approximately 45 miles per hour while the seat and the seat belt attachment points to the car frame are
held xed with boundary conditions, thus emulating a vehicle that comes to a sudden stop.
Dummy
A very simple dummy model is used (none of the limbs are modeled). The dummy model has two distinct
parts: the lower torso and the upper torso. The lower torso is modeled with rigid surface elements.
The upper torso is modeled in the same fashion except in the chest area. The human body compliance
is modeled approximately by meshing a region in the chest area using deformable shell elements. In
addition, four CARTESIAN and CARDAN connectors with nonlinear elastic and damping behavior are
inserted between four nodes around the chest area and four nodes belonging to the rigid back of the
upper torso. The upper and lower torsos overlap over a small region around the waist area, and general
contact is used to model the contact interactions between them. The approximate mass of the dummy is
35 kilograms.
Seat
The seat modeling is minimal since the focus of this example is to illustrate the seat belt modeling
technique. Only one solid element with crushable foam material properties is used to model the lower
part of the seat. The back support is not modeled.
Seat belt webbing passing through rings
The seat belt is modeled primarily using several SLIPRING connectors strung together. To model the
contact interactions between the belt and the chest and lap areas accurately, membrane elements are used
to model short portions of the belt in these regions. Figure 3.3.12 shows the seat belt arrangement.
3.3.11

Printed on:
SEAT BELT ANALYSIS
The node numbers associated with the connector elements are also illustrated. The seat belt is dened
starting from the right side of the gure (adjacent to the B-pillar in the car) and moving to the chest area,
the waist-level click-in buckle, the lap area, and nally the attachment to the car bottom oor, as follows:
Three SLIPRING connectors are used between nodes 8300228, 8300237, 8300243, and 8300247.
In this order the four nodes correspond to the exit point from the retractor at the bottom of the
B-pillar, the trim exit point along the B-pillar, the shoulder-level ring at the top of the B-pillar, and
the connection point with the membrane-mesh region of the belt at the top of the chest area.
The three SLIPRING connectors model the belt as it ows and stretches through these points.
As outlined in Connector element library, Section 28.1.4 of the Abaqus Analysis Users Manual,
the SLIPRING connection type activates the nodal material ow degree of freedom (10) at these four
nodes. Material can ow freely between the three elements but cannot ow through the connection
point with the membrane-meshed area at node 8300247. Hence, a boundary condition on this degree
of freedom is specied at this node.
The membrane-meshed area between nodes 8300247 and 8300248 models the contact with the chest
area of the dummy accurately.
Two SLIPRING connectors are used between nodes 8300248, 8300251, and 8300253. In this order
the three nodes correspond to the connection point with the membrane-mesh region of the belt at
the bottom of the chest area, the ring of the waist-level click-in buckle, and the connection point
with the membrane-meshed region of the belt at the left side of the lap area.
Since no material ow occurs at the connection points with the membrane-meshed regions,
boundary conditions on degree of freedom 10 are specied at nodes 8300248 and 8300253.
The membrane-meshed area between nodes 8300253 and 8300254 models the contact with the lap
area of the dummy accurately.
One SLIPRING connector is used between nodes 8300254 and 8301398. These two nodes
correspond to the connection point with the membrane-mesh region of the belt at the right side of
the lap area and the attachment point to the car frame.
Boundary conditions on degree of freedom 10 are specied at both nodes.
The stretching of the belt is governed by the specied nonlinear elastic connector behavior.
Retractor
The retractor device is located at the bottom of the B-pillar. It is modeled using a RETRACTOR
connector in parallel with several HINGE connectors as illustrated in Figure 3.3.13. The connections
are as follows:
The RETRACTOR connector is connected to the rst SLIPRING connector along the B-pillar at
node 8300228. It converts the material ow at this node into a rotation about a local axis oriented
along the longitudinal direction of the car.
Several HINGE connectors model the various mechanisms in the retractor device. Their axes are
all parallel to each other and are oriented along the longitudinal direction of the car.
The preload HINGE applies a small pretension elastic force typically given by a weak torsional
elastic spring. Its purpose is to eliminate the slack in the belt.
3.3.12

Printed on:
SEAT BELT ANALYSIS
The spool lock HINGE uses a connector lock denition to lock the almost-free rotation of the
spools axle if the velocity of the belt material exiting the retractor exceeds a certain threshold.
The spool effect HINGE is attached to the car frame and uses a connector plasticity denition
to model the compression effects of the spool. This is accomplished via a tension-versusspooled-out-material curve.
The spool effect is inactive until the spool lock connector locks; hence, the two HINGE connectors are
placed in series. The preload is applied at all times and, therefore, is placed in parallel with the two other
HINGE connectors.
Pretensioner
The pretensioner device is attached to the car frame in the vicinity of the waist-level click-in buckle. It
is modeled using a SLIPRING connector and a RETRACTOR and a HINGE connector in parallel as
illustrated in Figure 3.3.14. The connections are as follows:
The pretensioner spool (HINGE) connector is attached to the car frame. It uses an amplitudedelayed connector motion denition to specify the rate at which the belt material ows into the
pretensioner device.
The RETRACTOR connector converts the specied motion in the HINGE connector into a material
ow that is expelled from the following SLIPRING connector.
The SLIPRING connector models the stiff cable to which the waist buckle is attached. Its length
shortens as the pretensioner device is being triggered.
A PIN-type MPC is used to connect the node associated with the waist-level click-in ring that slides
over the seat belt (node 8300251) with the buckle it clicks into (node 8300351). By using the
additional node 8300351, the material ows associated with the pretensioner SLIPRING and the
SLIPRINGs of the adjacent seat belt segments are prevented from interacting with each other.
Models
Frictionless as well as frictional seat belt models are analyzed using Abaqus/Explicit.
The undeformed and deformed shapes (t=0.0215 seconds) for the model are shown in Figure 3.3.15 and
Figure 3.3.16, respectively. Results for belt tensions and material ow in and out of the shoulder level
slipring are shown in Figure 3.3.17 and Figure 3.3.18. At this junction SLIPRING connector elements
8888803 and 8888804 share a common node 8300243. For the frictionless analysis the normalized belt
tensions are shown in Figure 3.3.17. As expected, the two tension histories are the same. For the
frictional case, the ratio of the belt tension in adjacent belt segments is shown in Figure 3.3.18. For
the case when the belt is slipping, the ratio of the belt tension is given by
, where and
are the tensions in the adjacent SLIPRING connector elements, is the coefcient of friction, and
is the angle between the two adjacent sliprings. For the seat belt model with friction where =0.1 and
=1.718132 radians, the ratio of the belt tension is
. As shown in Figure 3.3.18, the ratio
agrees well with the analytical result. Near the end of the analysis (19 milliseconds) the ratio of the
3.3.13

Printed on:
SEAT BELT ANALYSIS
belt tension drops from this value. Figure 3.3.18 shows that the normalized accumulated slip remains
constant for the remainder of the analysis; hence, we can conclude that the ratio drops because the belt
starts sticking. Figure 3.3.19 shows that the material ows across node 8300243, which is the second
node of connector element 8888803 and the rst node of connector element 8888804, are identical as
expected.
Input files
seatbelt_xpl.inp
Analysis of a frictionless seat beltrestrained simplied

dummy using Abaqus/Explicit.
Analysis of a frictional seat beltrestrained simplied
dummy using Abaqus/Explicit.
seatbelt_fric_xpl.inp
Figure 3.3.11
Seat belt and dummy arrangement.
3.3.14

Printed on:
SEAT BELT ANALYSIS
8300243
8300237
8300247
8300228
8300248
8300251
8300254
8300253
8301398
Figure 3.3.12
Seat belt arrangement.
3.3.15

Printed on:
SEAT BELT ANALYSIS
S L IP R IN G
8 3 00 2 28
S po ol lock
(H IN G E )
F lo w con verter
(R E T R A C T O R )
+
P reload
(H IN G E )
1 0 01 1
S po ol effects
(H IN G E )
1 0 11 1
Figure 3.3.13
Retractor model.
S L IP R IN G
D-ring
8 3 00 2 51
M P C typ e P IN
8 3 00 3 51
P rete nsion er cab l e
(S LIP R IN G )
8 3 00 3 57
P rete nsion er flo w
converter
(R E T R A C T O R )
+
S po ol (H IN G E )
1 0 11 0
Attached to
the car floor
Figure 3.3.14
Pretensioner model.
3.3.16

Printed on:
SEAT BELT ANALYSIS
Slipring
Slipring
Slipring
Slipring
Slipring
Retractor
and hinge
Figure 3.3.15
Slipring
Dummy and seat belt system in the undeformed conguration.
Slipring
Slipring
Slipring
Slipring
Slipring
Retractor
and hinge
Figure 3.3.16
Slipring
Dummy and seat belt system in the deformed conguration.
3.3.17

Printed on:
SEAT BELT ANALYSIS
Figure 3.3.17
Normalized belt tension of adjacent frictionless sliprings.
Viewport: 2
Graph: XYPlot2
Figure 3.3.18 Ratio of belt tension of adjacent sliprings with

friction and the normalized accumulated slip.
3.3.18

Printed on:
SEAT BELT ANALYSIS
Figure 3.3.19
Normalized material ow across node between adjacent sliprings.
3.3.19

Printed on:
SIDE CURTAIN AIRBAG
3.3.2
SIDE CURTAIN AIRBAG IMPACTOR TEST
This example problem illustrates the use of surface-based uid cavities in a multi-chambered automotive
airbag ination analysis using Abaqus/Explicit. Airbags are supplemental safety devices that minimize the
chance of an occupant striking the interior of a vehicle during a crash. The proper deployment of the airbag
within a fraction of a second after a collision is detected is crucial to ensure the safety of the occupant.
This example describes the modeling and validation of a side curtain airbag impactor test simulation against
experimental results.
Geometry and model
The side curtain airbag model described in this example was provided by Autoliv GmbH. The model setup
is shown in Figure 3.3.21, which shows an impactor hitting the deployed side curtain airbag. The airbag
is divided into 18 chambers to approximate the nonuniform distribution of the gases across the airbag
and the gradients in pressure observed during the complex unfolding of the airbag. The temperature and
pressure are considered to be uniform in each chamber at any point of time. Each airbag chamber is
modeled as a surface-based uid cavity. The surface-based uid cavity capability provides a coupling
between the pressure exerted by the contained gas on the chamber boundary and the deformation of the
gas-lled chamber. The individual chambers are separated by ctitious partitions modeled with surface
elements, which represent the area of uid exchange between the different chambers. The airbag has three
inators located in chambers 1, 6, and 9 that are used to inate the airbag. The inators are activated
at different times using ination time amplitude denitions associated with the *FLUID INFLATOR
ACTIVATION option. The inators inject a mixture of 6 gas species, with the molecular fraction, mass
ow rate, and temperature of the individual species dened as functions of time. The gas species behave
as ideal gases with an adiabatic temperature update. Leakage of the gases to the environment through the
airbag fabric is modeled by specifying a discharge coefcient (ratio of the effective leakage area to the
total airbag surface area) of 0.0003 for uid exchange from each airbag chamber. In addition, the loss of
heat energy to the environment is also considered by providing the heat energy ow rate as a function of
the temperature difference across the airbag fabric.
The external skin of the airbag is represented by 3-node membrane elements (M3D3) and uses
a nonlinear elastic denition of the fabric material model. The fabric material model is calibrated by
specifying the uniaxial behavior of the bers in the warp and weft directions as well as the shear behavior
of the fabric. The behavior of the bers has been dened such that they offer negligible resistance to
compression until a certain compressive strain limit beyond which they begin to offer nite resistance
to compression to prevent excessive deformations of the fabric elements. Bilinear shear behavior is
dened reecting an initial shear modulus of
=9.44 N/mm2 and a modulus of
=47.2 N/mm2 at
angles greater than 60 to prevent severe shear deformation. In addition, a small amount of stiffness
proportional damping is applied. A reference mesh (initial metric) is dened to specify the unfolded
stress-free conguration of the airbag.
3.3.21

Printed on:
SIDE CURTAIN AIRBAG
The unstressed reference conguration of the airbag is dened using the *INITIAL CONDITIONS,
TYPE=REF COORDINATE option. The initial folded conguration and the reference conguration
are used to account for the wrinkles that arise from the airbag folding process. The reference mesh
and the folded airbag conguration are shown in Figure 3.3.22 and Figure 3.3.23, respectively. The
fabric material is dened such that any initial compressive strains along the yarn directions will cause
compressive stresses in the yarn. Alternatively, the initial compressive strains along the yarn directions
can be recovered stress free.
The general contact algorithm is used to consider self-contact between the airbag fabric layers as
well as contact with surrounding structures during the ination of the airbag.
The impactor is modeled as a rigid body with a total mass of 7.2 kg and is given an initial velocity
of 4500 mm/s towards the deploying airbag. The mass and velocity of the impactor reect human head
mass and velocities typically observed in vehicle side impacts, respectively.
The airbag at different stages of deployment is shown in Figure 3.3.24 (8 ms) and Figure 3.3.25 (16 ms).
Figure 3.3.26, Figure 3.3.27, and Figure 3.3.28 show the time history of displacement, velocity, and
acceleration of the impactor as it impacts the airbag, along with comparisons to experimental results.
These plots demonstrate the close correlation between Abaqus/Explicit simulation results and physical
test results.
Input files
sidecurtain_airbag_fabric.inp
sidecurtain_airbag_refmesh.inp
Figure 3.3.21
Side curtain airbag impactor test.

Reference mesh for the side curtain airbag.
Model setup showing the fully deployed airbag (t=80 ms).
3.3.22

Printed on:
SIDE CURTAIN AIRBAG
Figure 3.3.22
Figure 3.3.23
Figure 3.3.24
Airbag reference conguration.
Folded airbag conguration.
Deformed conguration at 8 ms.
3.3.23

Printed on:
SIDE CURTAIN AIRBAG
Figure 3.3.25
Deformed conguration at 16 ms.
0.00
Experiment
ABAQUS
-40.00
Displacement [mm]
-80.00
-120.00
-160.00
-200.00
-240.00
0.00
0.02
0.04
0.06
0.08
0.10
Time [s]
Figure 3.3.26
Time history of displacement of the impactor.
3.3.24

Printed on:
SIDE CURTAIN AIRBAG
3
[x10 ]
2.00
Experiment
ABAQUS
1.00
Velocity [mm/s]
0.00
-1.00
-2.00
-3.00
-4.00
-5.00
0.00
0.02
0.04
0.06
0.08
0.10
Time [s]
Figure 3.3.27
Time history of velocity of the impactor.
3
[x10 ]
175.00
ABAQUS
Experiment
150.00
Acceleration [mm/s^2]
125.00
100.00
75.00
50.00
25.00
0.00
0.00
0.02
0.04
0.06
0.08
0.10
Time [s]
Figure 3.3.28
Time history of acceleration of the impactor.
3.3.25

Printed on:
MECHANISM ANALYSES
4.
Mechanism Analyses
Mechanism analyses, Section 4.1

Printed on:
MECHANISM ANALYSES
4.1
Mechanism analyses
Resolving overconstraints in a multi-body mechanism model, Section 4.1.1

Crank mechanism, Section 4.1.2
Snubber-arm mechanism, Section 4.1.3
Flap mechanism, Section 4.1.4
Tail-skid mechanism, Section 4.1.5
Cylinder-cam mechanism, Section 4.1.6
Driveshaft mechanism, Section 4.1.7
Geneva mechanism, Section 4.1.8
Trailing edge ap mechanism, Section 4.1.9
Substructure analysis of a one-piston engine model, Section 4.1.10
Application of bushing connectors in the analysis of a three-point linkage, Section 4.1.11
Gear assemblies, Section 4.1.12
4.11

Printed on:
RESOLVING OVERCONSTRAINTS
4.1.1
RESOLVING OVERCONSTRAINTS IN A MULTI-BODY MECHANISM MODEL
An overconstraint occurs when multiple consistent or inconsistent kinematic constraints are applied to the
same degree of freedom. Overconstraints may lead to inaccurate solutions or prevent convergence. A
large number of overconstraint situations are detected and eventually resolved automatically either in the
preprocessor or during an Abaqus/Standard analysis (see Overconstraint checks, Section 31.6.1 of the
Abaqus Analysis Users Manual). The vast majority of the overconstraints that are not resolved by the
preprocessor are detected by the equation solver. The following symptoms identify such overconstrained
models in Abaqus/Standard:
Zero-pivot warning messages issued in the message (.msg) le indicating that the system of equations
is rank decient.
Unreasonably large reaction forces.
Very large time average forces in the message le.
A displacement solution that violates the imposed constraints.
By default, overconstraint checks are performed continuously by the equation solver during the analysis.
Abaqus/Standard does not resolve these overconstraints. Instead, detailed messages describing the modeling
features that generated the overconstraint are issued to help the user resolve the problems. The message
rst identies the nodes involved in either a consistent or an inconsistent overconstraint by using zero pivot
information from the Gauss elimination in the solver (see Direct linear equation solver, Section 6.1.4 of the
Abaqus Analysis Users Manual). A detailed message containing constraint information is then issued.
Geometry and model
This example deals with resolving overconstraints in the multi-body mechanism model shown in
Figure 4.1.11. The model consists of nine rigid bodies interconnected with connector elements
(Connectors: overview, Section 28.1.1 of the Abaqus Analysis Users Manual). The bodies named in
the gure are connected as follows:
LINK is connected to both CRANK and DISH using two CVJOINT (JOIN + CONSTANT
VELOCITY) connector elements. Each of these rigid bodies spins about its own axis.
TRANS CONNECT is connected to DISH using a JOIN connector element, which acts like a
pin connection. TRANS CONNECT is also constrained to translate along the direction dened
by TRANS GUIDE using a TRANSLATOR (SLOT + ALIGN) connector element between the
two. In addition, TRANS CONNECT is attached to SLIDE BLOCK using a HINGE (JOIN +
REVOLUTE) connection with the hinge axis oriented along the global Z-direction.
SLIDE BLOCK in turn is constrained to slide along SLIDE GUIDE using a TRANSLATOR
(SLOT + ALIGN) connector element.
TRANS GUIDE and ROD GUIDE are connected using a CYLINDRICAL (SLOT + REVOLUTE)
connector element.
4.1.11

Printed on:
ROD is allowed to slide in ROD GUIDE using a TRANSLATOR (SLOT + ALIGN) connector
element.
These connections enable the deformed conguration shown in Figure 4.1.12. See Crank mechanism,
Section 4.1.2, for another example using this model.
Loading and boundary conditions
Reference nodes 10009 of rigid body ROD and 10006 of rigid body SLIDE GUIDE are xed completely.
In addition, translations and rotations along the global X- and Y-directions are constrained at reference
nodes 10001 of rigid body DISH and 10003 of rigid body CRANK. The mechanism is actuated using
the *BOUNDARY option to prescribe a rotation of 360 about the global Z-direction at reference node
10001 (rigid body DISH) in a *STATIC step.
Understanding overconstraint messages
When Abaqus/Standard attempts to nd a solution for this model, two zero pivots are identied in
the rst increment of the analysis suggesting that there are two overconstraints in the model. These
overconstraints have to be identied and removed to render the model properly constrained. One way
to identify possible overconstraints in the case of simple models is to count the number of degrees of
freedom and constraints. There are nine rigid bodies in the model with a total of 54 degrees of freedom.
There are 21 constraints specied using the *BOUNDARY option. The connector elements enforce
additional constraints: three TRANSLATOR connection types enforce 5 constraints each, two CVJOINT
connection types enforce 4 constraints each, one CYLINDRICAL connection type enforces 4 constraints,
one HINGE connection type enforces 5 constraints, and one JOIN connection type enforces 3 constraints.
Thus, the number of constraints enforced by connector elements is 35. Consequently, there are two (21
+ 35 54) constraints too many in the model, corresponding to the number of zero pivots identied by
the equation solver.
To help the user identify the constraints that should be removed, the following message is produced
in the message le outlining the chains of constraints that generated the rst overconstraint:
***WARNING: SOLVER PROBLEM.
ZERO PIVOT WHEN PROCESSING ELEMENT 20025
INTERNAL NODE 1 D.O.F. 4

OVERCONSTRAINT CHECKS: An overconstraint was detected at one of the Lagrange
multipliers associated with element 20025. There are multiple constraints
applied directly or chained constraints that are applied indirectly to this
element. The following is a list of nodes and chained constraints
between these nodes that most likely lead to the detected overconstraint.
LAGRANGE MULTIPLIER: 2321 <-> 863: connector element 20025 type SLOT ALIGN
constraining 2 translations and 3 rotations
..2321 -> 10007: *RIGID BODY (or *COUPLING - KINEMATIC)
4.1.12

Printed on:
....10007 -> 3159: *RIGID BODY (or *COUPLING - KINEMATIC)

......3159 -> 3031: connector element 20030 type SLOT REVOLUTE constraining
2 translations and 2 rotations
........3031 -> 10008: *RIGID BODY (or *COUPLING - KINEMATIC)
..........10008 -> 3134: *RIGID BODY (or *COUPLING - KINEMATIC)
............3134 -> 2824: connector element 20035 type SLOT ALIGN constraining
..............2824 -> 10009: *RIGID BODY (or *COUPLING - KINEMATIC)
................10009 -> *BOUNDARY in degrees of freedom 1
..863 -> 10004: *RIGID BODY (or *COUPLING - KINEMATIC)

......427 -> 3157: connector element 20010 type JOIN constraining 3 translations
............780 -> 3156: connector element 20005 type
JOIN CONSTANT VELOCITY constraining
................10002 -> 781: *RIGID BODY (or *COUPLING - KINEMATIC)
..................781 -> 3155: connector element 20001 type
JOIN CONSTANT VELOCITY constraining
....................3155 -> 10003: *RIGID BODY (or *COUPLING - KINEMATIC)
......................10003 -> *BOUNDARY in degrees of freedom 1
..........10001 -> *BOUNDARY in degrees of freedom

......3158 -> 1539: connector element 20015 type JOIN REVOLUTE constraining
............1575 -> 2027: connector element 20020 type SLOT ALIGN
Please analyze these constraint loops and remove unnecessary constraints.
The zero pivot warning message identies an internal node (Lagrange multiplier) associated with the
identied zero pivot. A typical line contains information pertaining to one constraint. The following line
from the output:
4.1.13

Printed on:
LAGRANGE MULTIPLIER: 2321 <-> 863: connector element 20025 type SLOT ALIGN
identies that the Lagrange multiplier associated with the zero pivot enforces one of the ve constraints
(SLOT and ALIGN) associated with connector element 20025 between user-dened nodes 2321 and 863.
Each of the subsequent lines conveys information related to one constraint in the chains of constraints
originating at the zero pivot node or in chains adjacent to them. For example, the line
informs the user that there is a rigid body constraint between nodes 10007 and 3159, while the line
states that there is a *BOUNDARY constraint xing degrees of freedom 1 through 6 at node 10009.
Indentation levels are used to help in identifying the links in a chain of constraints. A detailed
explanation of the chains is printed at the rst occurrence of an overconstraint in the message le.
Using this methodology, the following chains of constraints starting from the two nodes involved in the
Lagrange multiplier constraint are identied:
Lagrange multiplier: 2321 --> 10007 --> 3159 --> 3031 --> 10008 --> 3134 --> 2824
--> 10009 --> *BOUNDARY
Lagrange multiplier: 863 --> 10004 --> 427 -> 3157 --> 10001 --> 780 --> 3156
--> 10002 --> 781 --> 3155 --> 10003 --> *BOUNDARY
Lagrange multiplier: 863 --> 10004 --> 427 -> 3157 --> 10001 --> *BOUNDARY
Lagrange multiplier: 863 --> 10004 --> 3158 --> 1539 --> 10005 --> 1575 --> 2027
--> 10006 --> *BOUNDARY
If any of the chains terminates in a free end (meaning the chain does not form a closed loop or end
in a constraint), the chain does not have any contribution in generating the overconstraint. In the
example above, all the identied chains terminate in a constraint and, therefore, may contribute to the
overconstraint.
A second zero pivot is generated by the same Lagrange multiplier associated with internal node 1
of connector element 20025 at degree of freedom 5. The chains associated with the zero pivot caused at
degree of freedom 5 are identical to the ones at degree of freedom 4 and are not repeated in the message
le.
Correcting the overconstrained model
A node set containing all the nodes in the chains of constraints associated with a particular zero pivot is
generated automatically and can be displayed in the Visualization module.
4.1.14

Printed on:
In most overconstrained models there are many ways to resolve the overconstraints. The most
obvious solution in the example above is to eliminate the unnecessary connector constraints. Upon
investigation we see that two rotation constraints associated with the TRANS CONNECT rigid body
are enforced by the SLOT + ALIGN connector element between the TRANS CONNECT and TRANS
GUIDE bodies (nodes 2321 and 863) as well as by the JOIN + REVOLUTE connector element
between the TRANS CONNECT and SLIDE BLOCK bodies (nodes 3158 and 1539), which renders
the model overconstrained. Without affecting the intended kinematic behavior of the system, the JOIN
+ REVOLUTE connector can be replaced by a JOIN connector, enforcing only three displacement
constraints and removing the two extra constraints on these degrees of freedom.
It is important to analyze the chains of constraints carefully and remove constraints properly rather
than relax any two arbitrary constraints. For example, removing any two *BOUNDARY constraints
would neither produce the desired kinematic behavior nor remove the overconstraints.
An alternative solution is to add exibility to some of the rigid bodies or constraints, which can be
achieved by either making the TRANS CONNECT rigid body elastic or using appropriate combinations
of CARTESIAN and CARDAN (EULER or ROTATION as well) connectors together with exible
connections (*CONNECTOR ELASTICITY) to enforce some of the kinematic constraints in an
approximate manner.
As mentioned earlier, an additional method of identifying overconstraints is to plot the reaction forces at
the constrained degrees of freedom. These forces are unreasonably large in the overconstrained model
even though the displacement solution looks acceptable. In this model a plot of connector reaction
forces (CRF) for element 20025 (TRANSLATOR connector between nodes 863 and 2321 of rigid bodies
TRANS GUIDE and TRANS CONNECT, respectively) shows that they are unreasonably large. Once the
overconstraints are resolved and the analysis is rerun, there are no zero pivots generated in the message
le and the connector reaction forces become very small (approximately 106 magnitude) as expected.
Comparisons of the connector reaction forces shown in Figure 4.1.13 and Figure 4.1.14 suggest that
the model is properly constrained.
One other option for resolving the overconstraints is to relax some of the constraints associated with
the TRANSLATOR (SLOT + ALIGN) connector for which zero pivots were identied. A CARDAN
connection type with *CONNECTOR ELASTICITY could be used instead of the ALIGN connector.
Alternatively, the HINGE constraint attached to the SLIDE BLOCK rigid body could be relaxed by
using JOIN and CARDAN connection types with *CONNECTOR ELASTICITY for the CARDAN
components.
Input files
mbmech_overconst.inp
mbmech_overconst_crig.inp
mbmech_resolved.inp
Overconstrained multi-body mechanism model.

Overconstrained multi-body mechanism model using the
*CONNECTOR ELASTICITY, RIGID option.
Multi-body mechanism model with overconstraints
resolved.
4.1.15

Printed on:
mbmech_resolved_crig.inp
mbmech_nodedefs.inp
mbmech_elemdefs.inp
mbmech_rigbdefs.inp
mbmech_conndefs_oc.inp
mbmech_conndefs_res.inp
mbmech_cardan_elas.inp
mbmech_cardan_elas_parts.inp
Multi-body mechanism model with overconstraints

resolved using the *CONNECTOR ELASTICITY,
RIGID option.
Node denitions used in the overconstrained and resolved
models.
Element denitions used in the overconstrained and
resolved models.
Rigid body denitions used in the overconstrained and
resolved models.
Connector element denitions used in the
overconstrained model.
Connector element denitions used in the resolved model.
Multi-body mechanism model with the HINGE connector
element replaced by a JOIN + CARDAN connection and
the *CONNECTOR ELASTICITY option. This model
uses parts and assemblies; the connector elements are
dened at the assembly level.
Node, element, and rigid body denitions for all parts
used in mbmech_cardan_elas.inp.
4.1.16

Printed on:
CRANK
LINK
DISH
ROD GUIDE
ROD
TRANS GUIDE
TRANS CONNECT
SLIDE GUIDE
SLIDE BLOCK
Figure 4.1.11
Figure 4.1.12
Model of the multi-body mechanism.
Displaced conguration of the mechanism.
4.1.17

Printed on:
RESOLVED MODEL
OVERCONSTRAINED MODEL
Figure 4.1.13 Comparison between time histories of CTM1 in

element 20025 for the overconstrained and corrected models.
RESOLVED MODEL
OVERCONSTRAINED MODEL
Figure 4.1.14 Comparison between time histories of CTM2 in

element 20025 for the overconstrained and corrected models.
4.1.18

Printed on:
CRANK MECHANISM
4.1.2
CRANK MECHANISM
Abaqus/Explicit
This example illustrates the use of connector elements to model kinematic constraints between rigid bodies
in a multi-body mechanism.
Problem description
The crank mechanism considered here transmits a rotational motion through two universal joints and
then converts the rotation into translational motion of two slides. The mechanism is modeled using nine
rigid components attached with eight connector elements. The various kinematic constraints modeled
with connector elements include TRANSLATOR, which allows relative translation along a line but no
rotations; HINGE, which allows one relative rotation and xes relative translations; CYLINDRICAL,
which allows relative translation along a line and relative rotation about that line; JOIN, which xes
relative translations but leaves the rotations free; PLANAR, which keeps a point on a plane and
allows only relative rotations about the normal to that plane; and UJOINT, which xes the relative
translations and enforces a universal constraint on the relative rotations. The complete model is shown
in Figure 4.1.21.
The axes of rotation of the small and large disks are parallel but offset. A constant angular velocity
of the small disk is specied about its axis with a velocity boundary condition on its rigid body reference
node. All other degrees of freedom of the rigid body reference node are xed. The rotational motion
of the small disk is transmitted to the large disk through two UJOINT connections and a rigid link.
A UJOINT connection, or a universal rotation constraint with shared translational degrees of freedom,
between two nonaligned shafts will not transmit constant angular velocity. However, two symmetrically
placed universal constraints, as here, will produce constant angular velocity coupling between the two
disks. The large disk is connected to a rigid circular rod with a JOIN connection. A JOIN connection
is equivalent to a ball-and-socket or a spherical joint. The circular rod connects through a sleeve to a
at block. The rod and sleeve constraint is modeled with a CYLINDRICAL connection, which allows
the sleeve to translate along and rotate about the rod. The attachment of the circular rod to the at
block is a HINGE connection, which allows only a single relative rotation about the shared hinge axis.
The at block, in turn, is assumed to slide between two xed parallel plates. This sliding constraint is
modeled with a PLANAR (SLIDE-PLANE and REVOLUTE) connection. The sleeve on the circular
rod is connected to a square-section sleeve on the square rod with a HINGE connection. The square rod
is xed in space. The square-section sleeve slides along the square bar without rotating. This sliding
constraint is modeled with a TRANSLATOR connection.
A Python script is included that reproduces the model using the Scripting Interface in Abaqus/CAE.
The script imports the parts from an ACIS le and creates the constraints and connectors that dene the
dynamics of the mechanism. The script creates both an Abaqus/Standard and an Abaqus/Explicit model
that are ready to be submitted for analysis from the Job module.
Models with frictional interactions, plasticity, and damage in connectors are created by editing the
input les without friction to introduce the friction, plasticity, or damage denitions.
4.1.21

Printed on:
CRANK MECHANISM
A model that includes sensors and actuation via user subroutine VUAMP is also included to illustrate
a means to model control engineering aspects in multibody analyses. Assuming that an electric motor
provides the rotational input to the system, the intent is to shut off this motor (and complete the step) as
the horizontal traveling head completes one full cycle.
Figure 4.1.22 shows the position of the mechanism at various times. By visual inspection it can be
observed that the connector elements are enforcing the correct kinematic constraints.
In this model there are nine rigid bodies with 6 degrees of freedom each, accounting for 54 rigid
body degrees of freedom. The eight connector elements eliminate 33 rigid body degrees of freedom
through kinematic constraints (enforced via Lagrange multipliers) as itemized in Table 4.1.21. Hence,
the model has 21 rigid body degrees of freedom to be specied as boundary conditions or determined
by the solution. In this case all remaining rigid body degrees of freedom are specied as boundary
conditions, with the z-component of angular velocity specied for the small disk and 20 additional xed
boundary conditions used.
Files
rigmultimech_std.inp
rigmultimech_std_sensor.inp
uamp_rigmultimech.f
rigmultimech_exp.inp
rigmultimech_std_fric.inp
rigmultimech_exp_fric.inp
rigmultimech_std_plas.inp
rigmultimech_exp_plas.inp
rigmultimech_exp_dam.inp
rigmultimech_exp_sensor.inp
vuamp_rigmultimech.f
rigmultimech_bulk.inp
rigmultimech.py
rigmultimech.sat
Abaqus/Standard analysis.
Abaqus/Standard analysis with sensors and actuation via
user subroutine UAMP in uamp_rigmultimech.f.
Abaqus/Standard user subroutine UAMP to specify
amplitudes.
Abaqus/Explicit analysis.
Abaqus/Standard analysis with friction.
Abaqus/Explicit analysis with friction.
Abaqus/Standard analysis with plasticity.
Abaqus/Explicit analysis with plasticity.
Abaqus/Explicit analysis with damage.
Abaqus/Explicit analysis with sensors and actuation via
user subroutine VUAMP in vuamp_rigmultimech.f.
Abaqus/Explicit user subroutine VUAMP to specify
amplitudes.
Node and element bulk data for the rigid bodies.
Python script that creates both an Abaqus/Standard
and an Abaqus/Explicit model using Abaqus/CAE.
The script imports the parts from an ACIS le named
rigmultimech.sat.
ACIS le containing the geometry of the model.
4.1.22

Printed on:
CRANK MECHANISM
Table 4.1.21
Rigid body degrees of freedom eliminated by kinematic constraints.
Connection type
Number of kinematic
constraints
Total rigid body dofs

eliminated from model
UJOINT (2)
JOIN
CYLINDRICAL
HINGE (2)
10
SLIDE-PLANE and
REVOLUTE
TRANSLATOR
5
Total eliminated: 33
4.1.23

Printed on:
CRANK MECHANISM
JOIN
CYLINDRICAL
TRANSLATOR
HINGE
2
PLANAR
UJOINT
2
3
1
Figure 4.1.21
Rigid mechanism model.
4.1.24

Printed on:
CRANK MECHANISM
Figure 4.1.22
Step: 1
Frame: 1
Step: 1
Frame: 6
Step: 1
Frame: 11
Step: 1
Frame: 16
Step: 1
Frame: 21
Step: 1
Frame: 26
Time history of the motion of the mechanism during the rst revolution.
4.1.25

Printed on:
SNUBBER-ARM MECHANISM
4.1.3
This example illustrates the use of connector elements to model a snubbing mechanism (i.e., two solids coming
into contact).
Geometry and model
A snubber is used to cushion the closing of a cowling on an aircraft. The cowling opens and closes by
applying a torque about the Z-axis at a xed point (the origin, Q, in Figure 4.1.31). The snubber-arm
assembly is attached to the cowling at point C. The snubber arm is constrained at point A by a spring
and is allowed to rotate relative to the cowling about point C. The other end of the spring is attached to
the cowling at point E. A shoe is attached to the snubber arm at point B and contacts with a block as the
cover closes, as shown in detail in Figure 4.1.32. The shoe is allowed to pivot about point B. The spring
retains the snubber arm against a stop on the cowling with an initial force of 50 lbs when the cover is
open. The shoe contacts the block as the cover closes.
A small rotational spring (10 lb/rad) keeps the shoe from rotating relative to the snubber arm while
the shoe is not in contact with the block. The block is xed to the ground. As the shoe comes into contact,
the two surfaces slide relative to each other. The coefcient of friction between the block and the shoe
is 0.5.
The closing speed of the cowling is slow enough so that inertia effects can be neglected.
Figure 4.1.31 shows the position of the mechanism when the shoe and the block make initial contact.
The cowling rotates ve more degrees about point Q after the block and the shoe make initial contact.
Model interactions
The bodies named in Figure 4.1.31 and Figure 4.1.32 are connected as follows:
ARM is connected to COWLING at point C using JOIN and ROTATION connector elements. The
rotation is constrained using the *CONNECTOR STOP option to prevent ARM from rotating past
the stop.
The spring connecting ARM to COWLING is modeled using an AXIAL connector element dened
between points A and E. The pre-tension in the spring is modeled using the *CONNECTOR
CONSTITUTIVE REFERENCE option.
SHOE is connected to ARM at point B using JOIN and ROTATION connector elements. The small
rotational spring is modeled using the *CONNECTOR ELASTICITY option.
All bodies in the model are modeled using display bodies connected to the relevant connector nodes.
SHOE (deformable elements) and BLOCK (rigid surface) make exception to this rule to allow contact
interaction denition between these bodies. Some models include friction and plasticity effects in the
connectors.
4.1.31

Printed on:
Figure 4.1.33 shows the displaced congurations of the mechanism as the cowling is rotated. As the
cowling opens, the spring pulls on the arm at point C until the arm is stopped. After contact is established,
when the cowling is closing, the rotation of the arm increases the tension in the spring. During a ight
this mechanism allows the absorption of vibrations between the cowling and the aircraft.
Such analyses allow one to study not only the dynamics of the snubber-arm mechanism but also
the contact interaction between the block and the shoe. Figure 4.1.34 shows the contact pressure and
frictional shear stress at a node located at the center of the shoe contact surface. Over time the coefcient
of friction between the shoe and the block will decrease. Knowledge of the contact characteristics
between the block and the shoe is a critical component of the design of the mechanism.
Input files
snubber_model.py
snubber.inp
snubber_surf.inp
snubber_fric.inp
snubber_plas.inp
Python replay le for constructing the snubber-arm

mechanism model in Abaqus/CAE.
Snubber-arm mechanism model.
Snubber-arm mechanism model with surface-to-surface
contact.
Snubber-arm mechanism model with friction.
Snubber-arm mechanism model with connector plasticity.
4.1.32

Printed on:
ARM
closed position
initial contact position

11.44
0.5
Q
W
12.25
7.58
10.14
COWLING
22.53
31.19
36.17
38.37
Figure 4.1.31
Undeformed conguration of the snubber-arm mechanism.
BLOCK
D
COWLING
B
SHOE
ARM
Figure 4.1.32
Undeformed conguration of the shoe and the block.
4.1.33

Printed on:
Figure 4.1.33
Displaced congurations of the mechanism.
CPRESS
CSHEAR1
Figure 4.1.34 Contact pressure and frictional shear stress

at the center of the shoe contact surface.
4.1.34

Printed on:
FLAP MECHANISM
4.1.4
FLAP MECHANISM
Abaqus/Explicit
This example illustrates the use of connector elements to model a three-dimensional trailing edge mechanism.
Geometry and model
The complete model of the ap is shown in Figure 4.1.41. An actuator rotates a bell crank through the
deployment of an actuator arm. The bell crank pushes and pulls a connecting rod that attaches to the arm
of the ap.
The ap is connected to a rigid shaft on the aircraft wing structure at points E and I. An arm on the
shaft is attached to the rod at point F. The other end of the rod attaches to a bell crank at point D. The bell
crank is attached to the airplane so that it can rotate about point B. The axis of rotation of the bell crank
passes through point B and is parallel to the global Z-axis. The rotation of the bell crank is driven by the
deployment of the actuator arm. The actuator system attaches to the bell crank at point C. The other end
of the actuator system is attached to the aircarft structure at point A and is allowed to compensate for the
change of angle caused by the rotation of the bell crank.
A load is applied at the center of gravity of the ap so that it is colinear and oriented along the global
Z-axis.
Model interactions
The bodies named in Figure 4.1.41 are connected as follows:
ACTUATOR is connected to the ground at point A using a HINGE connector element.

ACTUATOR and ACTUATOR ARM are connected using a TRANSLATOR connector element. The
*CONNECTOR MOTION option is used to modify the conguration of the actuator system as a
function of time.
BELL CRANK is physically attached to the ground with a hinge connection at point B. The axis
of rotation of the hinge connection is parallel to the global Z-axis. However, using a HINGE
connector element to attach BELL CRANK to the ground would overconstrain the model. Because
of the connections used between point A and point C, point C is already constrained to travel in
the global XY plane. Because the position of point B has to remain xed in space, the rotation of
BELL CRANK about the
-axis is already constrained. As a result, only three translations and the
rotation of BELL CRANK about the
-axis need to be constrained to realize the hinge connection.
BELL CRANK is, thus, attached to the ground using JOIN and UNIVERSAL connector elements
at point B. The UNIVERSAL connection is used to constrain the relative rotation of BELL CRANK
with respect to the ground about the
-axis.
ROD is connected to BELL CRANK at point D and to ARM at point F using JOIN connector elements.
A CARDAN connector element is added at point F between ROD and ARM to introduce an elastic
behavior to prevent the free rotation of ROD around its axis.
4.1.41

Printed on:
FLAP MECHANISM
FLAP and ARM are connected by a WELD connector element at point E. FLAP is attached to the
ground using a HINGE connection at point I.
All bodies in the model are visualized using display bodies connected to the relevant connector
nodes. Separate models in Abaqus/Standard and Abaqus/Explicit include friction, plasticity, and damage
in the connectors.
The amplitude curve used to drive the deployment of the actuator arm is shown in Figure 4.1.43.
Figure 4.1.42 shows the conguration of the ap mechanism at intermediate instants as it is actuated.
As the actuator system is deployed, the ap rotates around the
-axis to modify the aerodynamics of
the wing.
Input file
ap_model.py
ap.inp
ap_fric.inp
ap_exp_fric.inp
ap_plas.inp
ap_exp_plas.inp
ap_exp_plas_dam.inp
Figure 4.1.41
Python replay le for constructing the ap mechanism

model in Abaqus/CAE.
Abaqus/Standard ap mechanism model.
Abaqus/Standard ap mechanism model with friction.
Abaqus/Explicit ap mechanism model with friction.
Abaqus/Standard ap mechanism model with plasticity.
Abaqus/Explicit ap mechanism model with plasticity.
Abaqus/Explicit ap mechanism model with plasticity
and damage.
Undeformed conguration of the ap mechanism.
4.1.42

Printed on:
FLAP MECHANISM
Figure 4.1.42
Deformed congurations of the ap mechanism.
4.1.43

Printed on:
FLAP MECHANISM
Figure 4.1.43
Translator connector motion.
4.1.44

Printed on:
TAIL-SKID MECHANISM
4.1.5
TAIL-SKID MECHANISM
Abaqus/Explicit
This example illustrates the use of connector elements to model the tail-skid mechanism of an airplane.
Geometry and model
The complete model of the tail-skid mechanism is shown in Figure 4.1.51. It consists of a bell crank,
an actuator, a tail-skid arm, and a large cylinder. The bell crank is connected to the actuator at point
A and to the large cylinder at point B. The tail-skid arm is connected to the actuator at point D and
to the large cylinder at point E. The large cylinder is a compressible single-use cartridge designed to
absorb energy in the unlikely event of a tail strike. As such, it behaves like a compression-only linear
spring with a stiffness K=875600 N/m (K=5000 lb/in) and a damping coefcient C=175100 N/m sec
(C=1000 lb/in sec). The bell crank pivots about point C to move the large cylinder. The whole mechanism
is attached to the aircraft frame through points C and D. All point locations are listed in Table 4.1.51.
Two types of analyses are of interest: the deployment and stowage kinematics analysis and the
tail strike analysis. In the kinematics analysis the actuator contracts a distance of 145.8 mm (5.74 in),
which will rotate the bell crank and move the mechanism from the deployed to the stowage position. The
large cylinder does not compress during stowage or deployment. The tail strike analysis simulates the
over-rotation of an aircraft during takeoff. In this analysis the tail-skid arm strikes the ground at a speed
of 1.839 m/s (6.0 ft/s) at point F when it is at the deployed position. The strike creates a force acting
at an angle of 10 degrees with respect to the 2-axis as a result of friction between the airplane and the
runway. During the strike the large cylinder is compressed a maximum distance of 101.6 mm (4.0 in)
to absorb energy before it bottoms out. The mass and inertia of the tail-skid components are assumed to
be negligible. The body of the airplane is assumed to be rigid, has a mass of 6193 kg (270000 lb), and a
mass moment of inertia of 2220 kg m2 (1.5e8 lb in2 ) relative to the 3-axis. The center of gravity of the
airplane is located at 18.06 m, 1.422 m, 0.0 m (711.0 in, 56 in, 0.0 in); it is not shown in the model
because the dimension of the aircraft is much larger than the mechanism.
Model interaction
The correct behavior of the tail-skid mechanism is modeled by dening appropriate connectors between
the discrete points in Figure 4.1.51. For visualization purposes, display bodies are attached to those
points to model the bell crank, actuator, tail-skid arm, and large cylinder.
The actuator behavior is modeled using an AXIAL connector between points A and C.
The large cylinder behavior is modeled by the combination of an AXIAL connector and a SLIDEPLANE connector between points E and G.
The bell crank is connected to the large cylinder by a HINGE (REVOLUTE+JOIN) connector and
to the actuator by a JOIN connector.
The tail-skid arm is connected to the actuator and the large cylinder at points D and E, respectively,
using two HINGE connectors.
4.1.51

Printed on:
TAIL-SKID MECHANISM
The tail-skid mechanism is then connected to the aircraft frame at points C and D using two HINGE
connectors. The air frame is modeled with two BEAM connectors connecting points C and D to the center
of gravity of the airplane. Since both the actuator and the large cylinder consist of two display bodies,
additional connectors are necessary to constrain the relative motion between them. For this purpose a
SLIDE-PLANE connector element and an ALIGN connector element are dened between points A and
D, and an ALIGN connector element is dened between points D and G.
In the kinematics analysis the position of the aircraft is xed; the contraction of the actuator, which
is realized through prescribed connector displacement of the AXIAL connector, moves the mechanism
from the deployed to the stowage position. In the tail strike analysis the over-rotation of the airplane
is modeled by allowing the airplane to be in free motion and applying an initial rotating velocity to it.
During the strike the direction of the reaction of the ground to the airplane is assumed to remain xed.
As a result the reaction force is modeled by applying a xed boundary condition at point F in a local
coordinate system that will generate a reaction at an angle of 10 degrees with the 2-axis. The actuator is
xed during the strike, while the large cylinder is compressed. The large cylinder will absorb energy until
it stops after being compressed 4 inches. This physical behavior of the large cylinder is modeled using
the *CONNECTOR ELASTICITY, *CONNECTOR DAMPING, and *CONNECTOR STOP options.
Separate models include friction, plasticity, and damage in the connectors.
In Abaqus/Explicit the tail strike analysis is repeated with a user-dened element in addition to the
axial connector between points A and D. The element is modeled to mimic the elastic behavior of an
axial connector; the damping and locking behavior are still dened through the axial connector acting in
parallel with the user-dened element.
A sequence of the deformed tail-skid mechanisms in the kinematics analysis is shown in Figure 4.1.52.
By visual inspection it can be observed that the connector elements are enforcing the correct kinematic
constraints. Some of the numerical results for the tail strike analysis are shown in Figure 4.1.53 and
Figure 4.1.54. Figure 4.1.53 shows the total, elastic, and viscous force of the AXIAL connector that
models the mechanical behavior of the large cylinder. Figure 4.1.54 shows the relative distance between
points G and D. The sudden changes of the connector force and displacement shown in both gures are
due to the large cylinder reaching its maximum contraction (modeled using the *CONNECTOR STOP
option) during tail strike.
Input files
tail_kinematics.inp
tail_strike.inp
tail_strike_fric.inp
tail_strike_exp_fric.inp
tail_strike_exp_fric_plas_dam.inp
tail_kinematics_model.py
Kinematics analysis.
Tail strike analysis.
Tail strike analysis with friction.
Abaqus/Explicit tail strike analysis with friction.
Abaqus/Explicit tail strike analysis with friction,
plasticity, and damage.
Python replay le for constructing the kinematics model
using Abaqus/CAE.
4.1.52

Printed on:
TAIL-SKID MECHANISM
tail_strike_model.py
Python replay le for constructing the tail strike model

using Abaqus/CAE.
Abaqus/Explicit tail strike analysis with a user-dened
element modeling the elastic behavior of one of the axial
connectors. The element is dened in user subroutine
vuel_axial_connector3d.f.
User subroutine VUEL to model the axial connector.
vuel_tail_strike_exp_fric.inp
vuel_axial_connector3d.f
Table 4.1.51
Point locations in the deployed position.
Point
14.95
19.62
0.00
14.15
10.24
0.00
15.29
16.13
0.00
0.00
0.00
0.00
13.69
3.49
0.00
24.72
12.24
0.00
4.1.53

Printed on:
TAIL-SKID MECHANISM

A
C

D
10
reaction force
(tail strike analysis)
o
Figure 4.1.51
Tail-skid mechanism model (the deployed position).
4.1.54

Printed on:
TAIL-SKID MECHANISM
Figure 4.1.52
Intermediate positions of the tail-skid mechanism during tail strike.
Elastic Force
Total Force
Viscous Force
Figure 4.1.53
Elastic and damping forces of the connector modeling the large cylinder.
4.1.55

Printed on:
TAIL-SKID MECHANISM
Figure 4.1.54
Relative displacement of the connector modeling the large cylinder.
4.1.56

Printed on:
CYLINDER-CAM MECHANISM
4.1.6
This example illustrates the use of connector elements to model a cylinder-cam mechanism.
Geometry and model
The cylinder-cam mechanism is shown in Figure 4.1.61. The solid aluminum cylinder is lying on the
Z-axis and has a slot around its circumference. The centerline of the slot is dened by a plane intersecting
the cylinder at 45 degrees. The slot has a radius of 7.62 mm (0.30 in) and a depth of 7.62 mm (0.30 in).
A pin with a spherical end of radius 6.35 mm (0.25 in) is set into the slot. The pin is constrained to be
parallel to the X-axis and remains in the XZ plane. Because of the difference in radii, there is a 1.27 mm
(0.05 in) gap between the slot and the pin head. As the cylinder rotates about the Z-axis, the pin travels
in a track that is parallel to the cylinder. The pin has a weight of 13.34 N (3.0 lbf).
The cylinder rotates with a speed of 30 rpm, which drives the pin back and forth in the track.
Resistance to the motion is caused by the friction between the pin and the track. The tolerance mismatch
due to the pin head being smaller than the slot is considered in the model.
Model interactions
As shown in Figure 4.1.61, the track, the pin, and the cylinder are modeled using display bodies. The
bodies in Figure 4.1.61 are connected as follows:
MASS and ROTARYI elements are attached to each display body through BEAM connector
elements to account for the inertia of each part in the model.
The interaction between PIN and TRACK is modeled using ALIGN+SLOT connectors between
point D and point E. The friction dissipation effects between PIN and TRACK are taken into account
using the *CONNECTOR DAMPING option.
The interaction between PIN and CYLINDER is modeled by dening a CARTESIAN connector
element between point B and point C. The 1.27 mm (0.05 in) gap between the pin and the slot is
modeled using the *CONNECTOR STOP option. The local coordinate system of the CARTESIAN
connector element (used to measure the displacement of point C relative to point B) is attached at
point B to CYLINDER and has its 23 plane lying in the plane of the slot. The connector stop
is dened along component 1 of the connector element. A displacement of 1.27 mm (0.05 in) of
the pin relative to the center of the slot along the Z-axis corresponds to a displacement of point C
relative to point B of 0.898 mm (0.035 in) in the connector local coordinate system. The connector
stop is, thus, dened at a distance of 0.898 mm (0.035 in) in the positive and negative 1-directions
of the CARTESIAN connector element.
4.1.61

Printed on:
Figure 4.1.62 shows the positions of the mechanism as the cylinder is rotated. The time histories of
the displacement of point C in the global coordinate system and the displacement of point C relative
to point B in the CARTESIAN connector local coordinate system are shown in Figure 4.1.63. In the
connector local coordinate system, point C travels a maximum distance of 0.035 in relative to point B.
This corresponds to the value given in the connector stop denition. When PIN and CYLINDER are
in contact, the displacement of point C relative to point B in the connector local coordinate system
remains constant, and the cylinder forces the pin to translate. When contact is lost, the connector relative
displacement varies with time, and the motion of the pin stops as a result of damping.
File
cylcammech.inp
cylcammech.py
Abaqus/Explicit analysis.
Python script that creates an Abaqus/Explicit model using
Abaqus/CAE. The script imports the parts from an ACIS
le named cylcammech.sat.
ACIS le containing the geometry of the model.
cylcammech.sat
TRACK
D(E)
Slot + Align
CYLINDER
PIN
Beam
C
Cartesian
B
Beam
Figure 4.1.61
Cylinder-cam mechanism.
4.1.62

Printed on:
3
Figure 4.1.62
Displaced positions of the mechanism.
4.1.63

Printed on:
Displacement of point C in global coordinate system (U3)

CARTESIAN connector element relative displacement (CU1)
Figure 4.1.63
Time histories of the displacement of the pin head (Point C).
4.1.64

Printed on:
DRIVESHAFT MECHANISM
4.1.7
Abaqus/Explicit
This example illustrates the use of connectors to model a driveshaft mechanism.

Geometry and model
Driveshaft mechanisms are important components of motor vehicles. These mechanisms often contain
universal joints that are used to carry motion from one shaft to another where the two shafts are not
perfectly aligned. A typical universal joint assembly is shown in Figure 4.1.71. The joint consists
of a cross that carries needle roller bearings at the four extremities. The bearing cups are attached to
yokes, two on the input shaft and two on the output shaft. The universal joint is not a constant velocity
joint. If the input and output shafts are not aligned, the uniform rotation of the input shaft will result in a
nonuniform rotation of the output shaft. When the input shaft is rotated at constant velocity, the output
shaft will accelerate and decelerate twice per revolution (a pulsation effect).
The driveshaft mechanism in this example problem consists of three shafts, SHAFT AB (the input
shaft), SHAFT CD (the center shaft), and SHAFT EF (the output shaft). SHAFT AB is connected to
SHAFT CD through a universal joint that connects B and C. SHAFT CD is connected to SHAFT EF
through a universal joint that connects D and E. B31 beam elements are used to model the shafts. SHAFT
AB and SHAFT EF have a length of 7 units each, and SHAFT CD is 10 units long. SHAFT AB and
SHAFT EF are coplanar and parallel to each other. SHAFT CD is also coplanar to SHAFT AB and
SHAFT EF but is inclined to SHAFT AB and SHAFT EF at an angle of 20 degrees. SHAFT AB,
SHAFT CD, and SHAFT EF are circular in cross-section; and the radius of their cross-section is 0.1
units. The universal joints connecting B and C and connecting D and E are modeled using CONN3D2
elements.
The variation of angular motion between the input and output shafts can be avoided. The pulsation
effect between the input shaft and the center shaft can be compensated entirely by the pulsation effect
between the center shaft and the output shaft if the universal joints are oriented properly with respect to
each other. A sufcient condition for the input/output velocity ratio to be constant is that the directions
normal to the crosses of the universal joints be colinear. The conguration of the shafts is illustrated in
Figure 4.1.72. In this gure we can see that the crosses lie in parallel planes. This will guarantee a
constant input/output velocity ratio.
Model interactions
All the degrees of freedom except the rotational degree of freedom about the beam axis are xed at A
and F. A prescribed rotation of 360 degrees is specied at A. No other boundary conditions are specied.
The bodies in Figure 4.1.72 are connected as follows:
SHAFT AB is connected to SHAFT CD with a UJOINT connector element (Joint 1).

SHAFT CD is connected to SHAFT EF with a UJOINT connector element (Joint 2).
4.1.71

Printed on:
Because the rotary motion of one shaft is to be transmitted to another shaft that is not colinear,
separate coordinate orientation systems are created and used to connect the connector nodes for each
UJOINT connector connecting the shafts. The coordinate systems used to dene Joint 2 are the systems
used to dene Joint 1 rotated by 90 degrees about the global 1-axis. This ensures proper orientation of
the joint crosses and a constant input/output velocity ratio. Separate models for Abaqus/Standard and
Abaqus/Explicit include friction in the connectors.
The rotation of SHAFT AB about its axis leads to a rotation of SHAFT EF about its axis. The rotation
of the input and output shafts about their axis is plotted in Figure 4.1.73. We can see that the motion of
these shafts is perfectly synchronized.
Input files
driveshaft_model.py
Python replay le for constructing the driveshaft

Abaqus/Standard driveshaft mechanism model.
Abaqus/Standard driveshaft mechanism model with
friction.
Abaqus/Explicit driveshaft mechanism model with
friction.
driveshaft.inp
driveshaft_fric.inp
driveshaft_exp_fric.inp
input shaft
yokes
output shaft
cross
3
1
2
Figure 4.1.71
Universal joint.
4.1.72

Printed on:

20

E
10

20
7
2
1
3
Figure 4.1.72
Conguration of the driveshaft mechanism.
4.1.73

Printed on:
Input shaft
Output shaft
Figure 4.1.73
Rotation of the input and output shafts about the global 1-axis.
4.1.74

Printed on:
GENEVA MECHANISM
4.1.8
GENEVA MECHANISM
This example illustrates the use of connector elements to model a Geneva mechanism, which converts
continuous rotary motion into intermittent rotary motion.
Geometry and model
The Geneva mechanism is essentially a timing device. It is used in counting instruments and other
applications where a continuous rotary motion needs to be converted to an intermittent rotary motion.
For example, it is used in clocks to limit the number of winding rotations of the clock spring and in
movie lm projectors to move the lm frame by frame. The Geneva mechanism consists of a rotating
body with a protruding pin and another rotating body with slots into which the pin slides.
In this example problem the mechanism consists of three bodies named PIN, SLOT1, and SLOT2,
as illustrated in Figure 4.1.81. SLOT1 is an analytical rigid surface that overlays SLOT2. SLOT2 is
modeled as a display body. SLOT1 and SLOT2 are rigidly joined to each other at point A so as to allow
them to rotate in unison about A. SLOT2 has a radius of 3.0 units and is 1.0 units thick. The pin is
constrained to be rigid. The thickness of the rigid part of PIN is 0.5 units. The rigid portion of PIN has
a reference point B, and PIN is allowed to rotate about B. Distance AB is 4.24264 units in the model.
The protruding portion of PIN is located at a distance of 3.0 units from reference point B, and its length
is 1.0 units. The rigid portion of PIN has a radius of 3.0 units.
Model interactions
The contact between PIN and SLOT1 takes place through a single slave node located at C, the center
of the protruding portion of PIN. As a result, the slots in SLOT1 (used for contact evaluation) are 0.05
units wide, whereas the slots in SLOT2 (used for display purposes only) are 0.25 units wide. The slots
have a length of about 2.0 units to allow for the pin to slide. This contact is considered to be frictionless.
All the degrees of freedom at A and B, except the rotational degrees of freedom about the 3-axis,
are xed. A rotation of 720 degrees about the 3-axis is prescribed at B over three steps.
An EULER connector is constructed connecting A, the reference point of SLOT1, to the ground.
The *CONNECTOR DAMPING and *CONNECTOR FRICTION options are used to introduce damping
and friction in the EULER connector. Damping and friction will prevent rigid body motion of SLOT1
and SLOT2 after the pin has left the slot.
The protruding portion of PIN enters the slots in SLOT1 as PIN is rotated. This interaction between
PIN and SLOT1 results in the rotation of SLOT1 and SLOT2, and they rotate 90 degrees for every
complete 360 degree rotation of PIN. Figure 4.1.82 shows the conguration of the bodies of the Geneva
mechanism at some intermediate instants during the analysis. Figure 4.1.83 shows a plot of the U3
rotation of SLOT1 at A as a function of the U3 rotation of PIN at B.
4.1.81

Printed on:
GENEVA MECHANISM
Input files
geneva_model.py
Python replay le for constructing the Geneva mechanism

model in Abaqus/CAE.
Geneva mechanism model.
geneva.inp
SLOT 1
SLOT 2
PIN
3
2
Figure 4.1.81
Exploded view of the Geneva mechanism.
4.1.82

Printed on:
GENEVA MECHANISM
6
Figure 4.1.82 Congurations of the Geneva mechanism
for increasing values of PIN rotation.
4.1.83

Printed on:
Rotation U3 at A
GENEVA MECHANISM
Rotation U3 at B
Figure 4.1.83 Plot of the U3 rotation of SLOT1 at A as a function
of the U3 rotation of PIN at B. Rotation values are in degrees.
4.1.84

Printed on:
TRAILING EDGE FLAP MECHANISM
4.1.9
This example illustrates the use of connectors to model the deployment of a trailing edge ap mechanism in
an aircraft.
Geometry and model
The trailing edge ap mechanism is a critical component of an aircraft. The shape and positioning of the
trailing edge aps are important determinants of the aircrafts lift and aerodynamic behavior.
The trailing edge ap structure in this example problem consists of three aps, BASE, MIDFLAP,
and ENDFLAP; and these aps are connected to each other by arms, which are rigid links pinned to the
aps at different points. The conguration of the aps is illustrated in Figure 4.1.91. There are nine
arms used in the model: ARM AB, ARM BF, ARM DEFG, ARM EI, ARM HIM, ARM JK, ARM MLJ,
ARM NK, and ARM PO. BASE, MIDFLAP, and ENDFLAP are modeled as display bodies; and the arms
are modeled as rigid truss members. MIDFLAP and ENDFLAP are deployed by rotating ARM AB pinned
on BASE at point A.
Model interactions
The bodies in Figure 4.1.91 are connected as follows:
JOIN connector elements are used to connect the arms at their endpoints to BASE, MIDFLAP, and
ENDFLAP. The endpoints of some arms are connected to other arms instead of being connected
to the aps. ARM AB is connected to ARM BF at B. ARM BF and ARM EI are connected to ARM
DEFG at F and E, respectively. ARM EI and ARM MLJ are connected to ARM HIM at I and M,
respectively. ARM MLJ is connected to ARM JK at J, and ARM MLJ is connected to MIDFLAP at
L, which is not one of the endpoints of ARM MLJ.
LINK connector elements are used to rigidly x the positions of the pivot points on the aps. These
link connectors are called LINK GL, LINK GN, LINK GP, LINK KO, LINK LN, and LINK LP.
These additional link connectors are necessary to support BASE, MIDFLAP, and ENDFLAP because
these aps are modeled as display bodies and not as deformable or rigid bodies. LINK GL, LINK
GN, LINK GP, LINK LN, and LINK LP are used to position points G, L, N, and P on MIDFLAP;
and LINK KO is used to position points K and O on ENDFLAP.
To deploy the ap, we x BASE in space by xing points A, D, and H and rotate ARM AB 90 degrees
about point A. One analysis models plasticity effects in some of the connectors.
The rotation of ARM AB results in the deployment of MIDFLAP and ENDFLAP. Figure 4.1.92 contains
a series of illustrations showing the positions of the aps at the beginning and end of the analysis and
also at some intermediate instants.
4.1.91

Printed on:
Input files
edgeap_model.py
edgeap.inp
edgeap_plas.inp
Figure 4.1.91
Python replay le for constructing the trailing edge ap

Trailing edge ap mechanism model.
Trailing edge ap mechanism model with connector
plasticity.
Undeformed conguration of the trailing edge ap mechanism.
4.1.92

Printed on:
Figure 4.1.92
Deformed congurations of the trailing edge ap mechanism.
4.1.93

Printed on:
ONE-PISTON ENGINE
4.1.10
SUBSTRUCTURE ANALYSIS OF A ONE-PISTON ENGINE MODEL
This example illustrates the use of the substructure capability in Abaqus to model efciently multi-body
systems that undergo large motions but exhibit only small linear deformations. The models also illustrate
how to switch between a full-mesh representation of a part, a substructure representation, and a rigid body
representation of the same part depending on the modeling needs. Both static and dynamic analyses are
performed, and the results are compared between the various models.
Geometry and material
The multi-body system discussed here is the simplied one-piston engine model depicted in
Figure 4.1.101 and in the detail shown in Figure 4.1.102. The model consists of several parts that
are constrained together using connector elements. Each part is modeled either as a fully deformable
regular mesh, a rigid body, or a substructure. The connections between parts are idealized using the
*COUPLING option and connector elements. On each part the *COUPLING option is used to constrain
a few nodes near the connection points so that the connection forces get distributed to a nite area,
much like in the physical model. The *COUPLING reference nodes are then used to dene appropriate
connector elements between neighboring parts that enforce the desired kinematic linkage between
bodies.
The two parts of the crankshaft are rigidly connected using a BEAM connector element. The
connecting rod between the crankshaft and the pistons head has HINGE connections at both ends.
The gear transmission (2:1 gear ratio) between the crankshaft and the valve-cam is modeled using the
*EQUATION option between two rotational degrees of freedom. An analytically dened rigid surface
is used to model the contact surface of the cam. This surface is rigidly attached to the larger gear used to
spin the cam. The contact between the cam followers and the cam is modeled using a single node on each
cam followers tip dened using the *SURFACE, TYPE=NODE option. The connections between cam
followers and push rods and between push rods and valve rockers are modeled using JOIN connectors.
The valve rockers use HINGE connections about xed points in space (the engine block is not modeled).
The valve rockers are pushing against the spring loaded valves. The valve springs are precompressed
and are dened using CARTESIAN connectors together with the *CONNECTOR CONSTITUTIVE
REFERENCE and *CONNECTOR ELASTICITY options. The valve seats and the cylinders are not
modeled. Steel material properties are used for all parts. The model is actuated using the *CONNECTOR
MOTION option to spin the right end of the crankshaft for a little more than two full revolutions. The
*CONNECTOR FRICTION option is used in some of the analyses to model frictional effects in the
connector elements. Plasticity effects are specied in connectors in some of the analyses.
Models
The following three models are considered:
4.1.101

Printed on:
ONE-PISTON ENGINE
1. All parts are modeled using regular meshes (no substructures or rigid bodies). This is the largest
model and could include the most modeling options, including nonlinear material modeling options
(e.g., plasticity), if desired. A large computation time is needed to complete the analysis.
2. All parts are modeled as rigid bodies. While much less computational time is needed to complete
this analysis, this model cannot compute stresses in the various parts. However, it reveals very
useful information about the overall kinematics/dynamics of the model. Reaction forces at support
points and connector output variables (such as connector reaction forces or elastic forces in the valve
springs) are also predicted accurately.
3. Some parts (crankshaft, connection rod, piston head, cam, and cam followers) are modeled as
substructures. The remainder of the parts (push rods, valve rockers, and valves) are modeled as
rigid bodies. This model is very efcient in obtaining both overall kinematic/dynamic information
as well as the stresses and strains in the selected parts.
The input les are organized so that one can very easily switch (just by changing a few lines) between
the different models.
Before the global analysis can be run, the substructure for each part must be generated. To minimize
the computational effort in the global analysis, the minimum necessary number of retained nodal degrees
of freedom is chosen. The reference nodes of all *COUPLING constraints must be retained since the
parts are connected through these nodes. In addition, since the parts will undergo large rotations during
the analysis, at least three nonaligned nodes must be retained at the substructure generation level to
allow for the precise calculation of the rigid body motion of each part (see Dening substructures,
Section 10.1.2 of the Abaqus Analysis Users Manual). Since for most parts only two *COUPLING
nodes are dened (most parts are connected at only two points), an additional node not aligned with the
*COUPLING nodes is also retained for most substructures. Because of the small number of retained
nodes the global analysis is very efcient. For these particular choices it runs faster than the model when
all parts are modeled as rigid.
The reduced mass matrix for each substructure is generated by including the *SELECT
EIGENMODES option during the substructure generation. To improve the representation of the
substructures dynamic behavior in the global analysis, six dynamic modes, which are extracted using
a *FREQUENCY preload step, are included during the substructure generation using the *SELECT
EIGENMODES option. The substructures are loaded by gravity in the vertical direction by using the
GRAVITY LOAD=YES parameter on the *SUBSTRUCTURE GENERATE option and the *DLOAD
option at the substructure usage level.
Both *STATIC and *DYNAMIC analyses are conducted. The vertical displacements of the valves and
the contact forces on the cam followers are monitored.
In Figure 4.1.103 the vertical displacement of the valve on the left is depicted for a *STATIC
analysis. The valve descends (opens) in the beginning of the cycle and then closes for most of
the analysis. Since the crankshaft rotates a little more than two revolutions, the valve starts opening
again in the end. The normal contact force between the left cam follower and the cam is shown in
Figure 4.1.104. A sharp peak is encountered when the valve opens, while the contact force stays
4.1.102

Printed on:
ONE-PISTON ENGINE
constant when the valve is closed. Since the valve seats are not modeled, the closed positions for the
rigid body model and for the substructure model are slightly different due to elastic deformation of the
cam followers in the latter case. The stresses in the left cam follower are depicted in Figure 4.1.105 for
an intermediate position of the crankshaft. The stresses are recovered from the substructure using the
*SUBSTRUCTURE PATH option.
Since the inertia forces of the moving parts are relatively small when compared to the compression
forces in the valve springs, the contact forces and the valve displacements in the *DYNAMIC analyses
are similar to those in the *STATIC analysis. However, if the valve spring stiffness is chosen to be
very small (such as would happen if the spring failed), the inertial effects become very important and
contact between the cam and the cam follower could be lost. In Figure 4.1.106 the valve displacement is
depicted for such a case. Contact is established only intermittently between the two parts (corresponding
to zero valve displacement). For most of the time the two parts are not contacting each other, and the
valve displacement is completely different from the intended behavior.
The abaqus substructurecombine execution procedure combines model and results data from two
substructure output databases into a single output database. For more information, see Combining
output from substructures, Section 3.2.17 of the Abaqus Analysis Users Manual.
Input files
crank_substr_sta.inp
crank_substr_dyn.inp
crank_substr_sta_fric.inp
crank_substr_dyn_fric.inp
crank_substr_dynimp.inp
crank_rb_sta.inp
crank_all_def_sta.inp
crank_all_def_dyn.inp
crank_shaft_gen.inp
crank_shaft2_gen.inp
crank_conrod_gen.inp
crank_head_gen.inp
crank_cam_gen.inp
crank_camfollow1_gen.inp
crank_camfollow2_gen.inp
crank_conn_gr_ref_nodes.inp
crank_small_mass.inp
crank_materials.inp
crank_shaft_nodes.inp
*STATIC substructure global analysis.

*DYNAMIC substructure global analysis.
*STATIC substructure global analysis with frictional
effects.
*DYNAMIC substructure global analysis with frictional
effects.
*DYNAMIC substructure impact analysis (softer valve
springs).
*STATIC analysis with rigid bodies only (no
substructures).
*STATIC regular mesh analysis.
*DYNAMIC regular mesh analysis.
Substructure generation for the right half of the shaft.
Substructure generation for the left half of the shaft.
Substructure generation for the connection rod.
Substructure generation for the pistons head.
Substructure generation for the cam.
Substructure generation for the rst cam follower.
Substructure generation for the second cam follower.
Connector, ground, and reference node denitions.
Small masses used to avoid numerical singularities in
*DYNAMIC analyses.
Material properties.
Node denitions for the right half of the shaft.
4.1.103

Printed on:
ONE-PISTON ENGINE
crank_shaft_elts.inp
crank_shaft_coupling.inp
crank_shaft2_nodes.inp
crank_shaft2_elts.inp
crank_shaft2_coupling.inp
crank_conrod_nodes.inp
crank_conrod_elts.inp
crank_conrod_coupling.inp
crank_head_nodes.inp
crank_head_elts.inp
crank_head_coupling.inp
crank_cam_nodes.inp
crank_cam_elts.inp
crank_cam_coupling.inp
crank_camfollow1_nodes.inp
crank_camfollow1_elts.inp
crank_camfollow1_coupling.inp
crank_camfollow2_nodes.inp
crank_camfollow2_elts.inp
crank_camfollow2_coupling.inp
crank_valves_nodes.inp
crank_valves_elts.inp
crank_valvelifters_nodes.inp
crank_valvelifters_elts.inp
crank_tierods_nodes.inp
crank_tierods_elts.inp
Element denitions for the right half of the shaft.

*COUPLING denitions for the right half of the shaft.
Node denitions for the left half of the shaft.
Element denitions for the left half of the shaft.
*COUPLING denitions for the left half of the shaft.
Connecting rod node denitions.
Connecting rod element denitions.
Connecting rod *COUPLING denitions.
Piston head node denitions.
Piston head element denitions.
Piston head *COUPLING denitions.
Cam node denitions.
Cam element denitions.
Cam *COUPLING denitions.
Node denitions for the rst cam follower.
Element denitions for the rst cam follower.
*COUPLING denitions for the rst cam follower.
Node denitions for the second cam follower.
Element denitions for the second cam follower.
*COUPLING denitions for the second cam follower.
Node denitions for the crank valves.
Element denitions for the crank valves.
Node denitions for the crank valve lifters.
Element denitions for the crank valve lifters.
Node denitions for the tie rods.
Element denitions for the tie rods.
4.1.104

Printed on:
ONE-PISTON ENGINE
Figure 4.1.101
Figure 4.1.102
Mesh used for the complete one-piston model.
Detail of the mesh used for the one-piston model.
4.1.105

Printed on:
ONE-PISTON ENGINE
rigid body analysis

substructure analysis
Figure 4.1.103
Valve displacement in *STATIC analyses.
Figure 4.1.104 Cam follower contact force in a

substructure *STATIC analysis.
4.1.106

Printed on:
ONE-PISTON ENGINE
Figure 4.1.105
Cam follower stresses in a substructure *STATIC analysis.
Figure 4.1.106 Valve displacement in a substructure *DYNAMIC

analysis with very soft valve spring stiffness.
4.1.107

Printed on:
BUSHING MODELING
4.1.11
APPLICATION OF BUSHING CONNECTORS IN THE ANALYSIS OF A THREEPOINT LINKAGE
Abaqus/Explicit
This example illustrates the use of detailed nite element bushing models to calibrate the constitutive
behavior of bushings (when experimental data for a particular bushing design are not available) so that a
very inexpensive 2-node connector element representation of the bushing can be used in subsequent analyses
involving the bushing. This approach is effective in reducing computational costs in assembly models while
accurately capturing the macroscopic response of the bushing. To demonstrate this approach, a three-point
linkage is simulated with connector elements used to model bushings.
The geometry of the three-point linkage (TPL) is shown in Figure 4.1.111. Since the main focus of this
example is to illustrate modeling of the bushings, a simplied representation of the TPL is used. The
TPL is connected to the subframe (assumed xed in space) via the front and rear bushings. A ball joint
connects the TPL to a wheel assembly (not shown). The TPL is a steel shell structure and is modeled as
a linear elastic material with Youngs modulus of 2.1 105 MPa, Poissons ratio of 0.3, and a density of
7.82 109 tonnes/mm3 .
As shown in Figure 4.1.112, the rear bushing is a hollow cylinder with a rubber portion enclosed
between two (inner and outer) thin steel tubes. The rubber portion has two symmetrically placed cutouts
of different sizes, and it is rigidly bonded to the steel tubes. The steel tubes are 2 mm thick and use the
same material properties as the TPL. The inner diameter of this bushing is 28 mm, the outer diameter is
72 mm, and the axial length is 20 mm. The outer steel tube is connected to the TPL using a kinematic
coupling. The inner tube is connected to a xed node that represents a subframe using a distributed
coupling. The rubber is modeled as a second-order Ogden hyperelastic material (Hyperelastic behavior
of rubberlike materials, Section 19.5.1 of the Abaqus Analysis Users Manual), which may undergo
nonlinear, nite deformation. The material parameters are
= 1.671,
= 9.0067,
= 2.154E4,
= 4.86970,
= 1.0, and
= 1.0. The material density is 1.5 109 tonnes/mm3 .
The geometry of the front bushing is shown in Figure 4.1.113; it consists of three noncoaxial steel
tubes with two rubber parts between them. The axial length of the bushing is 40 mm. The outer diameters
of the three tubes are 40 mm, 28 mm, and 16 mm. All three tubes have a thickness of 1 mm. The outer
steel tube is connected to the TPL using a kinematic coupling. The inner tube is connected to a xed
node that represents a U-shaped bracket. The front bushing uses the same rubber and steel materials as
those used for the rear bushing.
Models
To characterize the nonlinear constitutive behavior of the bushings, a series of static analyses of the
front bushing are performed in Abaqus/Standard and a series of quasi-static analyses of the rear bushing
are performed in Abaqus/Explicit. Self-contact occurs in the rear bushing analyses due to severe
4.1.111

Printed on:
BUSHING MODELING
deformation. In all cases the reference node of the kinematic coupling connected to the outer steel tube is
held xed in all degrees of freedom, while the second reference node connected to the inner tube moves.
For the calibration analyses, a BUSHING connector for which no constitutive behavior is dened is
placed between the two reference nodes. The use of a BUSHING connector to drive the deformation in
these models is desirable since this choice ensures appropriate kinematic and kinetic modeling when the
connector is used in other models (such as a full-system analysis). Since the main interest is capturing
the hyperelastic response of the rubber bushings, both unidirectional and coupled multidirectional tests
(simultaneous deformations in up to three directions) are performed as summarized below.
To analyze several loading conditions, some connector components of relative motion can be xed
and others can be prescribed nonzero motion. The reaction forces and relative motions in the connector
are saved as history output. The data from these analyses are then used in a subsequent analysis of the
three-point linkage assembly to dene nonlinear connector elasticity data.
Front bushing calibration tests
For the front bushing, static analyses are performed in Abaqus/Standard as follows:
Test 1_FB:
A translational motion is applied in the local 1-direction with all other relative motions xed. The
motion is applied in the positive and negative directions separately since the front bushing is not
symmetric with respect to the local 1-axis. The magnitude of this motion is 5 mm. The CRF1 vs.
CU1 data generated from this analysis are used to dene nonlinear elasticity data for component 1.
Test 2_FB:
magnitude of this motion is 5 mm. The CRF2 vs. CU2 data generated from this analysis are used
to dene nonlinear elasticity data for component 2.
Test 3_FB:
Test 4_FB:
A 5 bending rotational motion about the local 1-direction is applied with all other relative motions
xed. The CRM1 vs. CUR1 data generated from this analysis are used to dene nonlinear elasticity
data for component 4.
Test 5_FB:
4.1.112

Printed on:
BUSHING MODELING
Test 6_FB:
A 15 twisting rotational motion about the local 3-direction is applied with all other relative
motions xed. The CRM3 vs. CUR3 data generated from this analysis are used to dene nonlinear
elasticity data for component 6.
Test 7_FB:
A 5 mm displacement along the local 3-direction is applied about three preloaded congurations.
The test attempts to capture coupling effects in the bushing after complex deformation is achieved.
The preloaded congurations are:
1. A twisting of 0 about the local 3-direction.
2. A twisting of 7.5 about the local 3-direction.
This test is a collection of three two-step analyses. In each of the analyses a preload motion is applied
in the rst step, and the bending motion is applied in the second step. Assuming a hyperelastic-like
quasi-static response in the bushing, for a given nal coupled deformation, the deformation path is
not relevant. Hence, the prestress and the actual loading steps can be run sequentially. The sets of
CRF3 vs. CU3 data generated from the second steps of these analyses are used to dene elasticity
data for component 3 using independent component 6.
Test 8_FB:
A 5 bending motion about the local 2-direction is applied about three preloaded congurations.
The preloaded congurations are:
2. A twisting of 7.5 about the local 3-direction.
This test is a collection of three two-step analyses. In each of the analyses a preload motion is applied
in the rst step, and the bending motion is applied in the second step. Assuming a hyperelastic-like
quasi-static response in the bushing, the deformation path is not relevant for a given nal coupled
deformation. Hence, the prestress and the actual loading steps can be run sequentially. The sets of
CRM2 vs. CUR2 data generated from the second steps of these analyses are used to dene elasticity
data for component 5 using independent component 6.
Force vs. displacement and moment vs. rotation curves are shown in Figure 4.1.114 for the cases
without preload (Tests 1_FB to 6_FB) and in Figure 4.1.115 for the cases with preload (Tests 7_FB and
8_FB).
Script les were generated to automatically create the input les and run the analyses for the
uncoupled tests and each of the coupled tests. These scripts essentially build up a CROSS design
parametric study for the uncoupled tests and a MESH design for the coupled tests (see Scripting
parametric studies, Section 17.1.1 of the Abaqus Analysis Users Manual).
4.1.113

Printed on:
BUSHING MODELING
Rear bushing calibration tests
For the rear bushing, quasi-static analyses are performed in Abaqus/Explicit as follows:
Test 1_RB:
Test 2_RB:
Test 3_RB:
Test 4_RB:
Test 5_RB:
A 15 bending rotational motion about the local 2-direction is applied with all other relative
motions xed. The CRM2 vs. CUR2 data generated from this analysis are used to dene nonlinear
elasticity data for component 5.
Test 6_RB:
A 5 twisting rotational motion about the local 3-direction is applied with all other relative motions
Test 7_RB:
A 10 mm displacement in the local 1-direction is applied about several preloaded congurations.

The displacement is applied in the positive and negative directions separately since the rear bushing
is not symmetric with respect to the local 1-axis. The test attempts to capture coupling effects in the
bushing after complex deformation is achieved. The preloaded congurations are a combination
of a displacement in the local 2-direction and a rotation about the local 2-direction for which the
following design points were selected:
1. A displacement of 0.0 mm in the local 2-direction.
2. A displacement of 2.5 mm in the local 2-direction.
4.1.114

Printed on:
BUSHING MODELING
3.
4.
5.
6.
A displacement of 5.0 mm in the local 2-direction.

A 0 bending rotational motion about the local 2-direction.
A 7.5 bending rotational motion about the local 2-direction.
This test is a collection of 18 two-step analyses. In each of the analyses two preload motions are
applied in the rst step, and the translational motion in the local 1-direction is applied in the second
step. Assuming a hyperelastic-like quasi-static response in the bushing, the deformation path is not
relevant for a given nal coupled deformation. Hence, the prestress and the actual loading steps
can be run sequentially. The sets of CRF1 vs. CU1 data generated from the second steps of these
analyses are used to dene elasticity data for component 1 using independent components 2 and 5.
Test 8_RB:
A 5 mm displacement in the local 2-direction is applied about several preloaded congurations.

The preloaded congurations are a combination of a displacement in the local 1-direction (positive
and negative loadings are considered separately due to asymmetry) and a rotation about the local
2-direction, for which the following design points were selected:
1.
2.
3.
4.
5.
6.
7.
8.

A 7.5 bending rotational motion about the local 2-direction.
This test is a collection of 15 two-step analyses. In each of the analyses two preload motions are
applied in the rst step, and the translational motion in the local 2-direction is applied in the second
step. Assuming a hyperelastic-like quasi-static response in the bushing, the deformation path is not
relevant for a given nal coupled deformation. Hence, the prestress and the actual loading steps
can be run sequentially. The sets of CRF2 vs.CU2 data generated from the second steps of these
analyses are used to dene elasticity data for component 2 using independent components 1 and 5.
Force vs. displacement and moment vs. rotation curves are shown in Figure 4.1.116 for the
cases without preload (Tests 1_RB to 6_RB). Figure 4.1.117 shows a comparison of the force vs.
displacement with no preload and with the most extreme preload conditions for Tests 7_RB and 8_RB.
Script les were generated to automatically create the input les and run the analyses for the
uncoupled tests and each of the coupled tests. As in the case of the front bushing, these scripts essentially
build up a CROSS design parametric study for the uncoupled tests and a MESH design for the coupled
tests. Furthermore, Python les were created to automatically gather the force vs. displacement data
points and to create corresponding report les.
4.1.115

Printed on:
BUSHING MODELING
Additional comments on calibration tests
Since each bushing has several symmetry planes, the tests above (unless specied otherwise) are
conducted only for positive relative motions when the responses in the opposite directions are
symmetric. Therefore, nonlinear elasticity data are generated only for positive relative motions for
those cases. Nonlinear elasticity is dened for negative relative motions by symmetrizing the elasticity
data with respect to the origin.
The analyses chosen above are deemed appropriate to generate the necessary and complete nonlinear
elasticity data to enable BUSHING connectors to represent the front and rear bushings in subsequent
analyses of the TPL. The quasi-static analyses were run over sufciently large motion ranges to cover the
motion range expected in the analysis where the BUSHING connector is used. Nonlinear elasticity data
are generated in all six relative uncoupled directions in the BUSHING connector. For the front bushing,
elasticity data are generated (Tests 7_FB and 8_FB) for two coupled deformation modes (involving
two components of local motion each) that were thought to be the dominant coupling modes in the
subsequent analysis where the BUSHING connectors are used. For the rear bushing, elasticity data are
generated (Tests 7_RB and 8_RB) for two coupled deformation modes (involving three components of
local motion each) that were thought to be the dominant coupling modes in the subsequent analysis where
the BUSHING connectors are used. In general, the number and complexity of coupled deformation tests
can be increased to match any particular modeling needs.
TPL models
Three different models are created for the analysis of the three-point linkage system: a fully meshed
model, a bushing connector model, and a substructure model. They all model the TPL using the front
and rear bushings to connect to the subframe (not modeled) that is assumed to be xed in space. Loads
are applied to a reference point where the ball joint is attached to model loads that the wheel assembly
would exert on the TPL. The constraint imposed by the xed subframe is modeled in all three cases by
constraining the motion of the inner cylinder distributed coupling reference node in all six degrees of
freedom.
In the fully meshed model the linkage itself is modeled using shell elements, while the front and
rear bushings are modeled using continuum elements. The front bushing is connected to the reference
node of a U-bracket that is modeled as a rigid body (see Figure 4.1.111). The rigid body reference node
of the U-bracket is held xed in all six degrees of freedom.
In the bushing connector model the bushings in the fully meshed model are replaced with BUSHING
connectors. The constitutive data for the connectors are obtained from the series of tests described above.
The front bushing is connected to the reference node of a U-bracket that is modeled as a rigid body (see
Figure 4.1.111). The rigid body reference node of the U-bracket is held xed in all six degrees of
freedom.
In the substructure model the three-point linkage is modeled with a substructure, and the bushings
are modeled with connector elements as dened in the bushing connector model. For clarity, the
U-bracket is not represented in the substructure model since it is considered to be rigid and xed. The
rear bushing inner cylinder is connected to the subframe via a large bolt.
Two load cases are applied in all three models:
4.1.116

Printed on:
BUSHING MODELING
Load case 1:
A geometrically nonlinear, single-step, static analysis is performed where a concentrated force of

250 N is applied at the ball joint reference node in the negative global 1-direction. The analysis
models a horizontal load at the ball joint.
Load case 2:
A geometrically nonlinear, two-step, static analysis is performed. In the rst step the TPL is lifted
by 10 about the global 1-direction by applying a displacement boundary condition of 20 mm to the
ball joint reference node in the negative global 3-direction. In the second step a concentrated force
of 250 N is applied at the ball joint reference node in the negative global 1-direction. The analysis
models a horizontal load on the ball joint as the wheel and, hence, the ball joint go over a bump.
It can be seen from the results of Tests 7_FB and 8_FB (Figure 4.1.115) that the front bushing behavior
in both the local 3-direction and about the local 2-direction is affected signicantly by the amount of
preload about the local 3-direction. For the rear bushing it can be seen from the results of Tests 7_RB
and 8_RB (Figure 4.1.117) that the behavior in the local 1-direction and especially the behavior in the
local 2-direction are affected signicantly by the amount of the combined preload (in the local 2-direction
and about the local 2-direction for Test 7_RB, and in the local 1-direction and about the local 2-direction
for Test 8_RB).
The Mises stresses in the TPL calculated for each of the three modeling approaches are very similar,
as shown in Figure 4.1.118, Figure 4.1.119, and Figure 4.1.1110. In addition, displacement histories
at the ball joint (where the loading was applied) show very good agreement between the three models
(Figure 4.1.1111). As expected, the bushing connector model analysis and the substructure model
analysis produced identical displacements at the ball joint. When compared with the fully meshed model,
the differences for the displacement histories in both directions are 4% toward the end of the curves.
The main motivation to use the BUSHING connector in this and similar applications is to reduce
the complexity of the models and the computation time. In this TPL analysis the fully meshed model
analysis takes approximately 50 times longer than the analysis that uses connectors to model the bushings.
Furthermore, the substructure model analysis completes in about one-eighth of the time required for the
analysis that uses connectors to model the bushings, approximately 400 times faster than the original
fully meshed model.
In summary, this example demonstrates signicant improvement in analysis efciency when
bushings are modeled with pre-calibrated BUSHING connectors or substructures without sacricing
accuracy.
Files
tpl_fb_uncoupled.inp
tpl_fb_uncoupled.psf
Front bushing model template used for generating

constitutive data curves for all tests without preload.
Python script le to automatically generate input les and
run analyses for all tests without preload.
4.1.117

Printed on:
BUSHING MODELING
tpl_fb_CRF3vsCU3.inp
tpl_fb_CRF3vsCU3.psf
tpl_fb_CRM2vsCUR2.inp
tpl_fb_CRM2vsCUR2.psf
tpl_rb_uncoupled.inp
tpl_rb_uncoupled.psf
tpl_rb_CRF1vsCU1.inp
tpl_rb_CRF1vsCU1.psf
tpl_rb_CRF2vsCU2.inp
tpl_rb_CRF2vsCU2.psf
tpl_const_data_ext-rb_CRF1vsCU1.py
tpl_const_data_ext-rb_CRF2vsCU2.py
tpl_full_loadcase1.inp
tpl_full_loadcase2.inp
tpl_connectors_loadcase1.inp
tpl_connectors_loadcase2.inp
tpl_substructure_gen.inp
tpl_substructure_loadcase1.inp
tpl_substructure_loadcase2.inp

constitutive data curves for Test 7_FB.
run analyses for Test 7_FB.
constitutive data curves for Test 8_FB.
run analyses for Test 8_FB.
Rear bushing model template used for generating
constitutive data curves for all tests without preload.
run analyses for tests without preload.
constitutive data curves for Test 7_RB.
run analyses for Test 7_RB.
constitutive data curves for Test 8_RB.
run analyses for Test 8_RB.
Python le to automatically gather force vs. displacement
results for Test 7_RB.
Python le to automatically gather force vs. displacement
results for Test 8_RB.
Fully meshed model analysis for Load case 1.
Fully meshed model analysis for Load case 2.
Bushing connector model analysis for Load case 1.
Bushing connector model analysis for Load case 2.
Substructure generation analysis for the three-point
linkage.
Substructure model analysis for Load case 1.
Substructure model analysis for Load case 2.
4.1.118

Printed on:
BUSHING MODELING
Figure 4.1.111
Three-point linkage (TPL) assembly.
4.1.119

Printed on:
BUSHING MODELING
1
3
Figure 4.1.112
Rear bushing geometry.
1
3
Figure 4.1.113
Front bushing geometry.
4.1.1110

Printed on:
BUSHING MODELING
Figure 4.1.114 Front bushing coupled force vs. displacement

and moment vs. rotation curves without preload.
Figure 4.1.115 Front bushing coupled force vs. displacement

and moment vs. rotation curves with preload.
4.1.1111

Printed on:
BUSHING MODELING
Figure 4.1.116 Rear bushing force vs. displacement and

moment vs. rotation curves without preload.
Figure 4.1.117 Comparison of rear bushing coupled force

vs. displacement for Tests 7_RB and 8_RB.
4.1.1112

Printed on:
BUSHING MODELING
Figure 4.1.118
Figure 4.1.119
Mises stresses in the TPL for the fully meshed model (Load case 2).
Mises stresses in the TPL for the bushing connector model (Load case 2).
4.1.1113

Printed on:
BUSHING MODELING
Figure 4.1.1110
Mises stresses in the TPL for the substructure model (Load case 2).
Bushing connector model

Fully meshed model
Substructure model
Bushing connector model

Fully meshed model
Substructure model
0.0
0.00
0.10
1.0
U2 (mm)
U1 (mm)
0.20
2.0
3.0
0.30
0.40
0.50
4.0
0.60
5.0
0.0
0.70
0.5
1.0
1.5
2.0
Time (s)
0.0
0.5
1.0
Time (s)
Figure 4.1.1111 Comparison of displacements at the ball

joint reference node (Load case 2).
4.1.1114

Printed on:
1.5
2.0
GEAR ASSEMBLIES
4.1.12
GEAR ASSEMBLIES
Objectives
This example illustrates the use of connectors in modeling the kinematics of gear assemblies.
Gears are widely used in a variety of machinery to transfer torque (rotational motion) from one shaft
to another shaft and to act as a means to change the torque (rotational speed) acting on a shaft. Three
types of gear assemblies (spur gear assembly, rack and pinion mechanism, and screw gear assembly) are
studied in this example with a focus on the proper transmission of torque and rotational motion between
the gears. The units used are millimeters-tonnes-seconds.
Geometry
Spur gears
A spur gear assembly consists of a driving gear (labeled Pinion) with input torque acting on it and
a driven gear (labeled Gear) that receives the torque from the pinion, as shown in Figure 4.1.121.
In this example the gear ratio is chosen as 2.5; i.e., the gear has 2.5 times more teeth than the pinion
and, hence, it rotates 2.5 times slower than the pinion. To measure the power transmission in the gear
system, a linear torsional spring is placed on each shaft such that the ratio of their elastic constants
is the square of the ratio of the gear radius on the respective shaft. The whole gear assembly is
mounted on a single rigid body whose reference node is constrained to prevent rigid body motion.
The pitch circle radius of the pinion and the gear is 1.04573 units and 2.61433 units, respectively,
as marked by the position of node 2 in Figure 4.1.121. The elastic constant of the torsional spring
on the pinion shaft is chosen as 1 unit while that on the gear shaft, as calculated from the gear ratio,
is 6.25 units.
Rack and pinion
A rack and pinion mechanism is a special case of the spur gear assembly with one of the gears
having an innite radius. The mechanism consists of a spinning pinion pulling the rack toward
itself, as shown in Figure 4.1.122. A groove is provided to guide the movement of the rack. The
whole rack and pinion assembly is mounted on a rigid body whose reference node is constrained to
prevent any rigid body motion. The pitch circle radius of the pinion is 2.5 units.
Screw gear and housing
A screw gear mechanism is similar to a rack and pinion mechanism except that the translational
motion takes place along the axis of rotation of the gear, as shown in Figure 4.1.123. As in the
case of the rack and pinion arrangement, a slot guides the direction of the motion of the screw. The
screw gear and housing assembly is mounted on a rigid body whose reference node is constrained to
4.1.121

Printed on:
GEAR ASSEMBLIES
prevent any rigid body motion. The motion involves turning the screw one complete revolution as
it runs along the grooves of the screw thread of a xed housing. The screw has a radius of 1.5 units
and a pitch of 2 units.
This example illustrates the use of connectors for modeling gear assemblies in Abaqus/Explicit. The
focus in this example is purely on the kinematics of gears without the consideration of gear-teeth forces.
Case 1: Spur gears
A rotating pinion driving a gear.
Case 2: Rack and pinion
A rotating pinion pulling a rack guided in a slot.
Case 3: Screw gear
A rotating screw moving through a groove in a housing.
Mesh design
All three cases of gear assemblies use CONN3D2 elements with the section behavior dened in detail
in the following sections.
Case 1: Spur gears
This case illustrates using connectors to model the kinematics of a spur gear assembly.
Mesh design
Each spur gear is modeled using one CONN3D2 element with connection type HINGE and one
CONN3D2 element with connection type FLOW-CONVERTER.
Material model
The HINGE of the pinion has a torsional stiffness of 1 unit, and the HINGE of the gear has a torsional
stiffness of 6.25 units.
Boundary conditions
The rst nodes of both HINGEs are xed. The second node of the HINGE of the pinion is driven by
18.85 radians.
Interactions
As shown in Figure 4.1.121, the pinion and gear are each modeled with a HINGE and a FLOWCONVERTER sharing a node (nodes 1 and 3, respectively). Nodes 1 and 3 of HINGEs HG1 and HG2,
respectively, are free to rotate, whereas their second nodes (node 4 and 6, respectively) form a rigid body
along with node 5 (reference node). This rigid body has all of its degrees of freedom xed to prevent
overall rigid body motion of the gear system.
4.1.122

Printed on:
GEAR ASSEMBLIES
The second node of FLOW-CONVERTERs FC1 and FC2 is common (node 2) with all
of its rotations xed. Any relative rotation in HINGE HG1 causes material ow at node 2 of
FLOW-CONVERTER FC1. The gear mechanism formed by HINGE HG2 and FLOW-CONVERTER
FC2 mirrors the pinion mechanism formed by FLOW-CONVERTER FC1 and HINGE HG1. Thus, any
material ow at common node 2 of FLOW-CONVERTER FC2 results in a relative rotational motion
between the two nodes of HINGE HG2.
The gear ratio of 2.5 is maintained by dening the scaling factor (with proper sign) in FLOWCONVERTERs FC1 and FC2; for this case, the scaling factors are 1.0 and 0.4, respectively. The
display bodies of the gear and pinion have reference nodes 10 and 11, respectively, that form a rigid
body with the respective unconstrained nodes of HINGEs HG1 and HG2.
This case illustrates using connectors to model the kinematics of a rack and pinion mechanism.
Mesh design
The pinion is modeled using one CONN3D2 element with connection type HINGE and one CONN3D2
element with connection type FLOW-CONVERTER. The rack is modeled using one CONN3D2 element
with connection type SLIPRING. The groove in which the rack moves is modeled using one CONN3D2
element with connection type TRANSLATOR.
Material model
Both the HINGE and the SLIPRING are dened as rigid elastic.
Boundary conditions
The rst nodes of the HINGE and TRANSLATOR are xed. Degree of freedom 10 of the common node
of the SLIPRING and TRANSLATOR is also xed. The second node of the HINGE of the pinion is
driven by 6.28 radians.
Interactions
The rack and pinion mechanism, as shown in Figure 4.1.122, is modeled similarly to the spur gears
described in Case 1. The pinion is modeled with HINGE HG (nodes 1 and 6) and FLOW-CONVERTER
FC (nodes 1 and 2) that share a common node (node 1). The rack is modeled with a SLIPRING SR
(nodes 2 and 3) and TRANSLATOR TR (nodes 4 and 3) that is used to guide the motion of SLIPRING
SR. SLIPRING SR and FLOW-CONVERTER FC share a common node (node 2).
Nodes 2, 4, and 6 are part of a rigid body with reference node 5 that is constrained to prevent any
rigid body motion of the rack and pinion system. A xed boundary condition on degree of freedom 10
is imposed at node 3 of SLIPRING SR to prevent any material ow past this node. Any relative rotation
between nodes 1 and 6 of HINGE HG results in material ow at node 2 of FLOW-CONVERTER FC.
This material ow at node 2 results in tension in the adjoining SLIPRING SR that drives the rack.
The relation between the translational motion of the rack and the rotational motion of the pinion
is maintained by dening the scaling factor (with proper sign) of FLOW-CONVERTER FC; for this
4.1.123

Printed on:
GEAR ASSEMBLIES
case, the scaling factor is 0.4. As mentioned before, the motion of the rack occurs parallel to the axis
of the TRANSLATOR TR, and it can be measured via the connector displacement of TRANSLATOR
TR. In this example available component 4 in HINGE HG and available component 1 in SLIPRING SR
are both dened as rigid elastic. The display bodies of the rack and pinion have reference nodes 10 and
11, respectively, that form a rigid body with the respective unconstrained nodes of SLIPRING SR and
HINGE HG.
Case 3: Screw gear
This case illustrates using connectors to model the kinematics of a screw gear and housing assembly.
Mesh design
The screw is modeled using one CONN3D2 element with connection type HINGE, one CONN3D2
element with connection type FLOW-CONVERTER, and one CONN3D2 element with connection
type SLIPRING. The groove in which the screw moves is modeled using a CONN3D2 element with
connection type TRANSLATOR.
Material model
Both the HINGE and the SLIPRING are dened as rigid elastic.
Boundary conditions
The rst nodes of the HINGE and TRANSLATOR are xed. Degree of freedom 10 of the common node
of the SLIPRING and TRANSLATOR is also xed. The second node of the HINGE of the screw is
driven by 6.28 radians.
Interactions
The screw gear system, as shown in Figure 4.1.123, is modeled similarly to the rack and pinion system
described in Case 2. The screw is modeled with HINGE HG (nodes 6 and 1), FLOW-CONVERTER FC
(nodes 1 and 2), and SLIPRING SR (nodes 2 and 3). The housing is modeled with TRANSLATOR TR
(nodes 4 and 3). HINGE HG and FLOW-CONVERTER FC share a common node (node 1). SLIPRING
SR and FLOW-CONVERTER FC share node 2.
Nodes 2, 4, and 11 are part of a rigid body with reference node 5 that is constrained to prevent any
rigid body motion of the housing. Material ow at node 3 of the SLIPRING is prevented via a xed
boundary condition on degree of freedom 10. Similar to the rack and pinion mechanism, the relative
rotation between nodes 1 and 6 of HINGE HG results in material ow at node 2 of FLOW-CONVERTER
FC. This material ow at node 2 causes tension in the belt of SLIPRING SR, resulting in a translational
displacement of node 3. Nodes 3 and 6 are tied together as a rigid body and, hence, the translational
motion of node 3 results in the linear motion of the screw. The rotational motion of the screw is identical
to the rotational motion of node 1.
The scaling factor of FLOW-CONVERTER FC is dened to scale the rotational motion of node 1
in relation to its translational motion; for this case, the scaling factor is 3.14. As in the case of the rack
and pinion mechanism, TRANSLATOR TR guides the direction of the motion of the screw. Available
4.1.124

Printed on:
GEAR ASSEMBLIES
component 1 in SLIPRING SR is dened as rigid elastic. The display bodies of the screw and the housing
have reference nodes 10 and 11, respectively, that form a rigid body with the respective unconstrained
nodes of HINGE HG and SLIPRING SR.
Case 1: Spur gears
The pinion shaft is given an angular displacement of 18.85 radians by dening connector motion in
HINGE HG1. The nal angular displacement of the gear shaft is measured to be 7.54 radians, with
the gear ratio holding at 2.5 throughout the analysis duration, as shown in Figure 4.1.124. The
ratio of the input to output power, which is calculated by taking the product of connector moment
CTM1 and connector motion CUR1 of the respective HINGEs, is also found to be constant.
The pinion is rotated one complete revolution by dening connector motion in HINGE HG. The
rack moves 15.7 units, measured by CU1 of TRANSLATOR TR in Figure 4.1.125. This distance
is the same as the circumference of the pitch circle of the pinion.
Case 3: Screw gear
The screw is given an angular rotation of 20 radians by dening connector motion in HINGE HG.
The screw moves 6.37 units, measured by CU1 of TRANSLATOR TR as shown in Figure 4.1.126.
Input files
spur_gear.inp
rack_pinion.inp
screw_gear.inp
Spur gear assembly.

Rack and pinion assembly.
Screw gear assembly.
4.1.125

Printed on:
GEAR ASSEMBLIES
Pinion
Gear
HG1
x FC1
1,4
x
2
FC2
HG2
x
3,6
Y
Z
Figure 4.1.121
Spur gear assembly.
Pinion
1,6 x HG
FC
x
2
Rack
SR
5
Y
Z
3
x TR
Figure 4.1.122
Rack and pinion assembly.
4.1.126

Printed on:
4
x
GEAR ASSEMBLIES
x4
TR
1,6
x HG
Housing
FC
SR
Screw
x
2
Y
Z
Figure 4.1.123
Screw gear assembly.
3.0
Gear Ratio
2.5
Gear Ratio
2.0
1.5
1.0
0.5
0.0
0.00
0.20
0.40
0.60
0.80
1.00
Time
Figure 4.1.124 Ratio of angular displacement of the pinion (CUR1

of HINGE HG1) to that of the gear (CUR1 of HINGE HG2).
4.1.127

Printed on:
GEAR ASSEMBLIES
0.
CU1 of ELSET TR
Displacement
5.
10.
15.
0.00
0.20
0.40
0.60
0.80
1.00
Time
Figure 4.1.125
6.0
Displacement of the rack (CU1 of TRANSLATOR TR).
CU1 of ELSET TR
Displacement
5.0
4.0
3.0
2.0
1.0
0.0
0.00
0.20
0.40
0.60
0.80
1.00
Time
Figure 4.1.126
Displacement of the screw (CU1 of TRANSLATOR TR).
4.1.128

Printed on:
HEAT TRANSFER AND THERMAL-STRESS ANALYSES
5.
Heat Transfer and Thermal-Stress Analyses
Heat transfer and thermal-stress analyses, Section 5.1

Printed on:
HEAT TRANSFER AND THERMAL-STRESS ANALYSES
5.1
Heat transfer and thermal-stress analyses
Thermal-stress analysis of a disc brake, Section 5.1.1
Exhaust manifold assemblage, Section 5.1.3
A sequentially coupled thermal-mechanical analysis of a disc brake with an Eulerian approach,

Section 5.1.2
Coolant manifold cover gasketed joint, Section 5.1.4
Conductive, convective, and radiative heat transfer in an exhaust manifold, Section 5.1.5
Thermal-stress analysis of a reactor pressure vessel bolted closure, Section 5.1.6
5.11

Printed on:
THERMAL-STRESS DISC BRAKE
5.1.1
THERMAL-STRESS ANALYSIS OF A DISC BRAKE
Abaqus/Explicit
Disc brakes operate by pressing a set of composite material brake pads against a rotating steel disc: the
frictional forces cause deceleration. The dissipation of the frictional heat generated is critical for effective
braking performance. Temperature changes of the brake cause axial and radial deformation; and this change
in shape, in turn, affects the contact between the pads and the disc. Thus, the system should be analyzed as a
fully coupled thermo-mechanical system.
In this section two thermally coupled disc brake analysis examples are discussed. The rst example is
an axisymmetric model in which the brake pads and the frictional heat generated by braking are smeared
out over all 360 of the model. This problem is solved using only Abaqus/Standard. The heat generation is
supplied by user subroutine FRIC, and the analysis models a linear decrease in velocity as a result of braking.
The second example is a three-dimensional model of the entire disc with pads touching only part of the
circumference. The disc is rotated so that the heat is generated by friction. This problem is solved using both
Abaqus/Standard and Abaqus/Explicit.
It is also possible to perform uncoupled analysis of a brake system. The heat uxes can be calculated and
applied to a thermal model; then the resulting temperatures can be applied to a stress analysis. However, since
the thermal and stress analyses are uncoupled, this approach does not account for the effect of the thermal
deformation on the contact which, in turn, affects the heat generation.
Another type of geometrical model for a disc brake is used by Gonska and Kolbinger (1993). They
model a vented disc brake (Figure 5.1.11) and take advantage of radial repetition by modeling a pie-slice
segment (Figure 5.1.12). Like the axisymmetric model, this requires the effect of the pads to be smeared,
but it allows the modeling of radial cooling ducts while still reducing the model size relative to a full model.
Geometry and model
Both models analyzed in this example have solid discs, which allows the models to use coarser meshes
than would be required to model the detail of a typical disc brake that has complicated geometrical
features such as cooling ducts and bolt holes. The rst example further simplies the model by
considering the pads to be smeared around the entire 360 so that the system is axisymmetric. The
second example is a full three-dimensional model of the entire annular disc with pads touching only part
of the circumference. However, the geometry of the disc has been simplied by making it symmetrical
about a plane normal to the axis. Therefore, only half of the disc and one brake pad is modeled, and
symmetry boundary conditions are applied.
The dimensions of the axisymmetric model are taken from a typical car disc brake. The disc has
a thicker friction ring connected to a conical section that, in turn, connects to an inner hub. The inner
radius of the friction ring is 100.0 mm, the outer radius is 135.0 mm, and it is 10.0 mm thick. The conical
section is 32.5 mm deep and 5.0 mm thick. The hub has an inner radius of 60.0 mm, an outer radius of
80.0 mm, and is 5.0 mm thick. The pads are 20.0 mm thick and initially cover the entire friction ring
surface.
5.1.11

Printed on:
Two analyses of the axisymmetric model are performed in which the pads and disc are modeled
using fully integrated and reduced-integration linear axisymmetric elements. Reduced integration is
attractive because it decreases the analysis cost and, at the same time, provides more accurate stress
predictions. Frictional contact between the pads and the disc is modeled by contact pairs between surfaces
dened on the element faces in the contact region. Small sliding is assumed. The mesh is shown in
Figure 5.1.13, with the pads drawn in a darker gray than the disc. There are six elements through the
thickness of the friction ring and four elements through the thickness of each of the pads. The mesh is
somewhat coarse but is optimized by using thinner elements near the surfaces of the disc and pads where
contact occurs for better resolution of the thermal gradients in these areas.
The disc for the three-dimensional model has an outer radius of 135.0 mm, an inner radius of
90.0 mm, and a thickness of 10.0 mm (the half-model has a thickness of 5.0 mm). The ring has a thinner
section out to a radius of 100.0 mm, which has a thickness of 6.0 mm (the half-model has a thickness of
3.0 mm). The pad is 10.0 mm thick and covers a little less than one-tenth the circumference. The pad
does not quite reach to the edge of the thicker part of the friction ring.
The pad and disc of the three-dimensional model are modeled with C3D8T elements in
Abaqus/Standard and with C3D8RT elements in Abaqus/Explicit; the contact and friction between the
pad and the disc are modeled by contact pairs between surfaces dened on the element faces in the
contact region. The same mesh is used in both Abaqus/Standard and Abaqus/Explicit. It is shown in
Figure 5.1.14, with the pad drawn in a darker gray than the disc. The disc is a simple annulus with a
thinner inner ring. This mesh is also rather coarse with only three elements through the thickness of
the disc and three elements through the pad. The elements on the contact sides are thinner since they
will be in the areas of higher thermal gradients. There are 36 elements in the circumferential direction
of the disc.
Material properties
The thermal mechanical properties for the axisymmetric model were taken from a paper by Day and
Newcomb (1984) describing the analysis of an annular disc brake. The pad is made of a resin-bonded
composite friction material, and the disc is made of steel. Although Day and Newcomb note that material
changes occur in the pad material because of thermal degradation, the pad in the axisymmetric model has
the properties of the unused pad material. For the axisymmetric model the modulus, density, conductivity,
and coefcient of friction are divided by 18 since the pads actually cover only a 20 section of the disc,
even though they are modeled as being smeared around the entire circumference.
The pad for the three-dimensional model is also a resin-bonded composite friction material
whose thermal mechanical properties are listed in Table 5.1.11 and coefcient of friction is listed in
Table 5.1.12. The properties were taken from a paper by Day (1984). It is noted that above certain
temperatures, approximately 400C, the pad material becomes thermally degraded and is assumed
constant from this point on.
It is assumed that all the frictional energy is dissipated as heat and distributed equally between
the disc and the pad; therefore, the *GAP HEAT GENERATION option is set to 1.0, and the default
distribution is used. The *GAP HEAT GENERATION option allows the user to specify an unequal
distribution, which is particularly important if the heat conduction across the interface is poor. In this
example the conductivity value specied with the *GAP CONDUCTANCE option is quite high; hence,
5.1.12

Printed on:
the results are not very sensitive to changes in distribution. In Abaqus/Explicit arbitrarily high gap
conductivity values may cause the stable time increment associated with the thermal part of the problem
to control the time incrementation, possibly resulting in a very inefcient analysis. In this problem the
gap conductivity value used in the Abaqus/Explicit simulation is 20 times smaller than the one used in
the Abaqus/Standard simulation. This allows the stable time increment associated with the mechanical
part of the problem to control the time incrementation, thus permitting a more efcient solution while
hardly affecting the results.
Loading
The pads of the axisymmetric model are rst pressed against the disc. The magnitude of the load is
divided by 18 since the pads are not actually axisymmetric. The frictional forces are then applied through
user subroutine FRIC to simulate a linear decrease in velocity of the disc relative to the pads. The braking
is done over three steps; then, when the velocity is zero, a nal step shows the continued heat conduction
through the model.
The pad of the three-dimensional model is xed in the nonaxial degrees of freedom and is pressed
against the disc with a distributed load applied to the back of the pad. In Abaqus/Standard the disc is then
rotated by 60 using an applied boundary condition to the center ring. In Abaqus/Explicit this boundary
condition is prescribed using the *AMPLITUDE, TYPE=SMOOTH STEP option to minimize the effects
of centrifugal forces at the beginning and end of the step. Frictional forces between the surfaces generate
heat in the brake.
The initial temperature of both models is 20C.
Solution controls (Abaqus/Standard only)
Since the three-dimensional model has a small loaded area and, thus, rather localized forces and heat
uxes, the default averaged ux values for the convergence criteria produce very tight tolerances and
cause more iteration than is necessary for an accurate solution. To decrease the computational time
required for the analysis, the *CONTROLS option is used to override the automatic calculation of the
average forces and heat uxes. The option is rst used with the FIELD=DISPLACEMENT parameter.
The convergence criterion ratio is set to 1%, and the time-average and average uxes are set to a typical
nodal force (displacement ux):
E-4
where p is the pressure and A is the area of a typical pad element. The option is next used with the
FIELD=TEMPERATURE parameter. The convergence criterion ratio is set to 1%, and the time-average
and average uxes are set to the nodal heat ux (temperature ux) for a typical pad element. The heat
ux density generated by an interface element due to frictional heat generation is
, where is
the gap heat generation factor, is the frictional stress, and v is the velocity. Therefore, the nodal heat
ux is
5.1.13

Printed on:
where A is the contact area of a typical pad element,

is the friction coefcient, and p is the
contact pressure. The angular velocity, , is obtained as the total rotation,
, divided by the
total time, 0.015 sec. The radius, r, is set to 0.120 m, which is the distance from the axis to a point
approximately in the middle of the pad surface. This yields
E-4
Additional solution controls can reduce the solver cost for an increment by improving the initial
solution guess, solving thermal and mechanical equations separately, and reducing the wavefront of
three-dimensional nite-sliding contact analysis. These features are discussed below. The impact of
combining these features is also discussed.
When the default convergence controls are used, it is possible to obtain faster convergence by setting
the EXTRAPOLATION parameter on the *STEP option to PARABOLIC. For the three-dimensional
model the use of this feature yields a 14% enhancement in computational speed per increment.
The coupling between the thermal and mechanical elds in this problem is relatively weak. It is,
therefore, possible to obtain a more efcient solution by specifying separate solutions for the thermal
and mechanical equations each increment. This technique, which is specied by using the *SOLUTION
TECHNIQUE, TYPE=SEPARATED option, results in faster per-iteration solution times at the expense
of poorer convergence when a strong intereld coupling is present. Use of this technique also permits
the use of the symmetric solver and storage scheme, which is invoked by specifying UNSYMM=NO on
the *STEP option. The resulting symmetric approximation of the mechanical equations was also found
to be cost effective for this problem, when combined with a quality initial solution guess obtained by
specifying EXTRAPOLATION=PARABOLIC on the *STEP option. Neither of these approximations
impacts solution accuracy. For the three-dimensional model the use of *SOLUTION TECHNIQUE,
TYPE=SEPARATED and *STEP, UNSYMM=NO, EXTRAPOLATION=PARABOLIC yields a 50%
decrease in the total solution time.
In the three-dimensional model the deformable master surface is dened from a large number of
connecting elements resulting in a large wavefront. By default, Abaqus/Standard employs an automated
contact patch algorithm to reduce the wavefront and solution time. For instance, in the coupled thermalmechanical analysis a substantial savings in solution time (a 30% to 50% decrease) is obtained when the
automatic contact patch algorithm is employed compared to an analysis that uses a xed contact patch
encompassing the entire master surface. The reduction in solution time is system dependent and depends
on several factors, such as CPU type, system memory, and IO speed. This solution time savings is in
addition to any of the other savings discussed in this section. The additional savings is, therefore, realized
when the separated solution scheme and parabolic extrapolation are also specied.
The temperature distribution of the axisymmetric model at an early time increment is shown in
Figure 5.1.15. The temperature is greatest at the interfaces between the disc and pads, and the heat
has just started to conduct into the disc. Figure 5.1.16 shows the temperature distribution at the end
of the analysis when the velocity is zero. The heat has conducted through the friction ring of the disc.
5.1.14

Printed on:
Figure 5.1.17 is a displaced plot of the model at the end of the analysis and shows the characteristic
conical deformation due to thermal expansion. The displacement has been magnied by a factor of 128
to show the deformation more clearly.
The temperature distribution of the disc surface of the three-dimensional model after a rotation of
60 is shown in Figure 5.1.18 (Abaqus/Standard) and Figure 5.1.19 (Abaqus/Explicit). The agreement
between the two results is excellent. The hottest region is the area under the pad, while the heat in the
regions that the pad has passed over has dissipated somewhat. Figure 5.1.110 shows the temperature
distribution of the inside of the brake pad predicted by Abaqus/Standard, while Figure 5.1.111 shows
the same result obtained with Abaqus/Explicit. Again excellent agreement between the two results is
noted. Figure 5.1.112 shows the temperature distribution in the disc predicted by Abaqus/Standard
with the thickness magnied by a factor of 20. The heat has conducted into the disc in the regions that
the pad has passed over.
The stresses predicted by Abaqus/Standard do not account for the effects of centrifugal loads
(*COUPLED TEMPERATURE-DISPLACEMENT is a quasi-static procedure), while the stresses
predicted by Abaqus/Explicit do. These effects can be signicant, especially during the early transient
portion of the simulation when the initially stationary disc is brought up to speed. To compare the
stress results between Abaqus/Standard and Abaqus/Explicit, we gradually initiated and ended the
disc rotation in the Abaqus/Explicit simulation; thus, in Abaqus/Explicit, the centrifugal stresses at
the beginning and end of the step are small compared with the thermal stresses. At points in between,
however, the effects of centrifugal loading are more pronounced and differences between the stress
states predicted by Abaqus/Standard and Abaqus/Explicit are observed. The overall effect on the
thermal response, however, is negligible.
The Abaqus/Explicit analysis did not include mass scaling because its presence would articially
scale the stresses due to the centrifugal loads. It is possible to include mass scaling to make the analysis
more economical, but any results obtained with mass scaling must be interpreted carefully in this
problem.
Input files
Abaqus/Standard input files
discbrake_std_cax3t.inp
discbrake_std_cax3t.f
discbrake_std_cax4t.inp
discbrake_std_cax4t.f
discbrake_std_cax4rt.inp
discbrake_std_cax4rt_surf.inp
discbrake_std_cax4rt.f
discbrake_3d.inp
discbrake_postoutput.inp
Axisymmetric model with CAX3T elements.

User subroutine FRIC used in discbrake_std_cax3t.inp.
Axisymmetric model with CAX4T elements.
User subroutine FRIC used in discbrake_std_cax4t.inp.
Axisymmetric model with CAX4RT elements.
Axisymmetric model with CAX4RT elements using the
surface-to-surface approach.
User subroutine FRIC used in discbrake_std_cax4rt.inp.
Three-dimensional model.
*POST OUTPUT analysis of the three-dimensional
model.
5.1.15

Printed on:
discbrake_3d_extrapara.inp
Three-dimensional model with the second step run with

*STEP, EXTRAPOLATION=PARABOLIC and with the
default *CONTROLS option.
Three-dimensional model with the second step run
with *STEP, EXTRAPOLATION=PARABOLIC. It is
assumed that several revolutions occurred and the initial
temperature for the disc brake and pad is 300C.
Three-dimensional model run using the *SOLUTION
TECHNIQUE, TYPE=SEPARATED option.
discbrake_3d_extrapara_300c.inp
discbrake_3d_separated.inp
Abaqus/Explicit input file
discbrake_3d_xpl.inp
Three-dimensional model.
References
Day, A. J., An Analysis of Speed, Temperature, and Performance Characteristics of Automotive

Drum Brakes, Journal of Tribology, vol. 110, pp. 295305, 1988.
Day, A. J., and T. J. Newcomb, The Dissipation of Frictional Energy from the Interface of an
Annular Disc Brake, Proc. Instn. Mech. Engrs, vol. 198D, no. 11, pp. 201209, 1984.
Gonska, H. W., and H. J. Kolbinger, ABAQUS Application Example: Temperature and

Deformation Calculation of Passenger Car Brake Disks, ABAQUS Users Conference
Proceedings, 1993.
Table 5.1.11
Temperature of property measurement (C)
Thermal-mechanical properties.
20
100
200
300
Youngs modulus, E (N/mm2 )
2200
1300
530
320
Poissons ratio,
0.25
0.25
0.25
0.25
1550
1550
Density,
(kg/m )
1550
1
Thermal expansion coefcient (K )

Thermal conductivity,
Specic heat,
10 10
(w/mK)
(J/kgK)
Table 5.1.12
30 10
1550
6
0.5
0.5
0.5
0.5
1200
1200
1200
1200
Brake lining temperature characteristic.
Temperature of property measurement (C)
100
200
300
400
Friction coefcient,
0.38
0.41
0.42
0.24
5.1.16

Printed on:
Figure 5.1.11
A vented brake disc design.
Segment
Sectional plane
through center of
cooling duct
Figure 5.1.12
Modeling a segment of a brake disc.
5.1.17

Printed on:
Figure 5.1.13
Mesh for the axisymmetric model, Abaqus/Standard.
3
2
Figure 5.1.14
Mesh for the three-dimensional model.
5.1.18

Printed on:
NT11
VALUE
+1.99E+01
+2.18E+01
+2.36E+01
+2.54E+01
+2.72E+01
+2.90E+01
+3.08E+01
+3.27E+01
+3.45E+01
+3.63E+01
+3.81E+01
+3.99E+01
+4.17E+01
+4.36E+01
Figure 5.1.15
NT11
Isotherms of the axisymmetric model at
0.675, Abaqus/Standard.
VALUE
+2.11E+01
+2.35E+01
+2.58E+01
+2.81E+01
+3.04E+01
+3.27E+01
+3.51E+01
+3.74E+01
+3.97E+01
+4.20E+01
+4.43E+01
+4.67E+01
+4.90E+01
+5.13E+01
Figure 5.1.16
Isotherms of the axisymmetric model when braking has ended, Abaqus/Standard.
5.1.19

Printed on:
Figure 5.1.17 Deformation of the axisymmetric disc,

displacement magnied by 128, Abaqus/Standard.
NT11
+5.019e+01
+4.767e+01
+4.515e+01
+4.263e+01
+4.011e+01
+3.759e+01
+3.507e+01
+3.255e+01
+3.003e+01
+2.751e+01
+2.499e+01
+2.247e+01
+1.995e+01
2
3
Figure 5.1.18
Isotherms of the disc surface, Abaqus/Standard.
5.1.110

Printed on:
NT11
+4.712e+01
+4.486e+01
+4.260e+01
+4.034e+01
+3.807e+01
+3.581e+01
+3.355e+01
+3.129e+01
+2.903e+01
+2.676e+01
+2.450e+01
+2.224e+01
+1.998e+01
2
3
Figure 5.1.19
Isotherms of the disc surface, Abaqus/Explicit.
NT11
+4.965e+01
+4.718e+01
+4.471e+01
+4.223e+01
+3.976e+01
+3.729e+01
+3.482e+01
+3.235e+01
+2.988e+01
+2.741e+01
+2.494e+01
+2.247e+01
+2.000e+01
2
1
Figure 5.1.110
Isotherms of the inside of the brake pad, Abaqus/Standard.
5.1.111

Printed on:
NT11
+4.685e+01
+4.461e+01
+4.237e+01
+4.014e+01
+3.790e+01
+3.566e+01
+3.342e+01
+3.119e+01
+2.895e+01
+2.671e+01
+2.447e+01
+2.224e+01
+2.000e+01
2
1
Figure 5.1.111
NT11
Isotherms of the inside of the brake pad, Abaqus/Explicit.
VALUE
+1.99E+01
+2.22E+01
+2.45E+01
+2.68E+01
+2.91E+01
+3.14E+01
+3.37E+01
+3.59E+01
+3.82E+01
+4.05E+01
+4.28E+01
+4.51E+01
+4.74E+01
+4.97E+01
3
1
2
Figure 5.1.112
Isotherms of the disc with the thickness magnied 20 times, Abaqus/Standard.
5.1.112

Printed on:
DISC BRAKE WITH AN EULERIAN APPROACH
5.1.2
A SEQUENTIALLY COUPLED THERMAL-MECHANICAL ANALYSIS OF A DISC

BRAKE WITH AN EULERIAN APPROACH
The prediction of fatigue and failure of a disc brake system is fundamental in assessing product performance.
Disc brakes operate by pressing a set of brake pads against a rotating disc. The friction between the pads and
the disc causes deceleration. The brake system then converts the kinetic energy of vehicle motion into heat.
Severe temperature changes as well as mechanical loadings cause inelastic deformation and circumferential
tensile stress in the disc, which may eventually lead to the failure of the disc.
The traditional way of analyzing this kind of problem is to use a Lagrangian approach in which
the mesh used to discretized the disc rotates relative to the brake assembly. Since many revolutions are
typically required to reach the state of interest to the analyst, this approach is prohibitively expensive and
cumbersome. The steady-state transport analysis capability in Abaqus/Standard (Steady-state transport
analysis, Section 6.4.1 of the Abaqus Analysis Users Manual), which uses the Eulerian method in which
the nite element mesh of the disc does not rotate relative to the brake assembly but the material ows
through the mesh, provides a cost-effective alternative approach. The paths that the material points follow
through the mesh are referred to as streamlines. This kinematic description converts the moving disc brake
problem into a pure spatially dependent simulation. Thus, the mesh has to be rened only in a xed region
where the brake pads are in contact with the disc initially.
Geometry and model
The model analyzed in this example is a solid disc. An axisymmetric model is created to dene the
cross-sectional geometry of the disc, as shown in Figure 5.1.21. The disc has a thicker friction ring
connected to a conical section that, in turn, connects to an inner hub. The inner radius of the friction
ring is 86.5 mm, the outer radius is 133.0 mm, and the ring is 13.0 mm thick. The conical section is
27.2 mm deep. The inner radius of the conical section is 64.75 mm, the outer radius is 86.5 mm, and the
section is 6.4 mm thick. The conical section has a thinner section out to a radius of 71.25 mm, which
has a thickness of 4.5 mm. The hub has an inner radius of 33.0 mm, an outer radius of 71.25 mm, and
is 6.2 mm thick. Symmetric model generation (Symmetric model generation, Section 10.4.1 of the
Abaqus Analysis Users Manual) is used to create a three-dimensional disc, as shown in Figure 5.1.22,
by revolving the two-dimensional cross-section about the symmetry axis and to create the streamlines
needed for the steady-state transport analyses in this example. There are eight elements through the
thickness of the friction ring, four elements through the thickness of the hub, and four elements through
the conical section. There are 40 element sectors in the circumferential direction of the disc, with a more
rened mesh used in the region with higher thermal and stress gradients. The model consists of 9440
rst-order forced convection/diffusion bricks (DCC3D8) in the heat transfer analysis, giving a total of
about 11520 degrees of freedom; and it consists of 9440 rst-order bricks (C3D8) in the subsequent
steady-state transport analysis, giving a total of about 34560 degrees of freedom.
The disc pads are not modeled in the example. Instead, the thermal and mechanical interactions
between the disc and the pads are represented by the application of appropriate distributed heat uxes
5.1.21

Printed on:
in the heat transfer analysis and by the application of appropriate concentrated loads in the steady-state
transport analysis, respectively.
Material properties
The disc is made of metallic material, with a Youngs modulus of 93.5 GPa, a yield stress of 153 MPa,
a Poissons ratio of 0.27, and a coefcient of thermal expansion of 11.7 106 per C at room
temperature. In this example the dissipation of the frictional heat-generated temperature uctuates,
ranging from a minimum value of 40C to a maximum value of 560C over the entire braking cycle.
The temperature distribution when the disc is heated to its peak value is shown in Figure 5.1.23.
Under such operating conditions plastic deformation, as well as creep deformation, is observed. The
two-layer viscoelastic-elastoplastic model, which is best suited for modeling the response of materials
with signicant time-dependent behavior as well as plasticity at elevated temperatures, is used to model
the disc (see Two-layer viscoplasticity, Section 20.2.11 of the Abaqus Analysis Users Manual).
This material model consists of an elastic-plastic network that is in parallel with an elastic-viscous
network. The Mises metal plasticity model with kinematic hardening is used in the elastic-plastic
network, and the power-law creep model with strain hardening is used in the elastic-viscous network.
Because the elastic-viscoplastic response of the material varies greatly over this temperature range,
temperature-dependent material properties are specied.
The thermal properties for the disc are temperature dependent with a conductivity of 51 103 W/mm
per C, a specic heat of 501 J/kg per C, and a density of 7.15 106 kg/mm at room temperature.
Problem description and loading
A simulation of braking a solid disc rotating initially at an angular velocity of 155.7 rad/sec is
performed. The braking time is approximately 5 seconds, followed by a cooling period of 600 seconds.
A sequentially coupled thermal-mechanical analysis is performed on the solid disc using the Eulerian
approach: a forced convection/diffusion heat transfer analysis is followed by a steady-state transport
analysis. Heat uxes with lm condition and prescribed mass ow velocity through user subroutine
UMASFL are applied to the thermal model, which consists of three steps. The rst step, which lasts
0.2 seconds, simulates the response of the disc under constant distributed uxes and a constant angular
velocity. The second step involves 4.8 seconds during which the distributed uxes and the angular
velocity are decreased linearly to small values near zero at the end of the step. The nal step, which
lasts 600 seconds, simulates the continued cooling in the model. The resulting temperatures obtained
during the heat transfer analysis are applied to the subsequent mechanical analysis, which involves ve
steady-state transport analysis steps.
The purpose of the rst step in the mechanical analysis is to obtain a steady-state solution for a disc
under constant concentrated loads due to the application of the brake pads to the disc. There is only one
increment in this step. A constant temperature of 40C is used, and a constant angular transport velocity
of 155.7 rad/sec is specied.
The second step obtains a series of quasi-steady-state transport solutions under different temperature
loading passes through the disc. The pass-by-pass steady-state transport analysis technique is used for
this purpose. This step lasts 0.2 seconds with a constant angular velocity of 155.7 rad/sec throughout
5.1.22

Printed on:
the entire step. Several increments are involved, with each increment corresponding to a complete
temperature loading pass through the disc. The temperature values obtained during the rst step of the
heat transfer analysis are read into this step.
The third step also obtains a series of quasi-steady-state transport solutions under different
temperature loading passes through the disc. However, this step involves 4.8 seconds over which the
angular velocity is decreased linearly from 155.7 rad/sec at the beginning of the step to a small value
close to zero at the end of this step. There are several increments in this step, with each increment
corresponding to a complete temperature loading pass through the disc. The temperature values
obtained during the second step of the heat transfer analysis are read into this step.
The fourth step obtains a steady-state solution for the disc due to the removal of the concentrated
loads. There is only one increment in this step.
The last step obtains a series of quasi-steady-state transport solutions when the disc cools down.
There are several increments over a step period of 600 seconds. The temperatures obtained during the
last step of the heat transfer analysis are read into this step. Since the angular velocity is very small, this
step essentially simulates a long-term elastic-plastic response for the disc.
Solution controls
Since the modied Newtons method is used in a steady-state transport analysis, more numerical
iterations are necessarily required to obtain a converged solution. To decrease the computational time
required for the analysis due to the unnecessary cutback of the increment size, the time incrementation
control parameters are used to override the default values (see Time integration accuracy in transient
problems, Section 7.2.4 of the Abaqus Analysis Users Manual).
One of the considerations in the design of a disc brake system is the stress distribution and deformation in
the region where the brake pads are applied. Circumferential tensile stress, which may cause the fracture
of the disc, will develop, making this region critical in the design. The results shown in Figure 5.1.26
through Figure 5.1.211 are measured in this region (element 7817, integration point 5; see point A in
Figure 5.1.22). The temperature in this region (node 7820) is shown in Figure 5.1.24 as a function of
the time over the entire braking process. Figure 5.1.25 shows the Mises stress distribution just before
the distributed loads are removed and the cooling period starts (Step 3, increment 60).
Figure 5.1.26, Figure 5.1.27, and Figure 5.1.28 show the evolution of the circumferential
stress, circumferential plastic strain, and circumferential viscous strain, respectively, as a function
of the time throughout a complete braking cycle. A tensile stress of 54 MPa is developed after the
disc is cooled down completely. Both plastic strain and viscous strain reach their saturation levels
during the cooling period. The time evolution of the circumferential stress versus the circumferential
plastic strain, shown in Figure 5.1.29, is obtained by combining Figure 5.1.26 with Figure 5.1.27.
Similarly, the time evolution of the circumferential stress versus the circumferential viscous strain,
shown in Figure 5.1.210, is obtained by combining Figure 5.1.26 with Figure 5.1.28. The shapes of
the stress-strain curves represent the plastic and viscous energies dissipated, respectively, over an entire
5.1.23

Printed on:
braking cycle. These dissipated energies, which could be used to predict the fatigue life for the disc, are
shown in Figure 5.1.211 as a function of the time.
Acknowledgments
SIMULIA gratefully acknowledges PSA Peugeot Citron and the Laboratory of Solid Mechanics of the
Ecole Polytechnique (France) for their cooperation in developing the Eulerian algorithm for steady-state
transport analysis and for supplying the geometry and material properties used in this example.
Input files
discbrake_sst_heat_axi.inp
discbrake_sst_heat_symm.inp
exa_discbrake_sst_heat_symm.f
discbrake_sst_axi.inp
discbrake_sst_symm_pbp.inp
Axisymmetric model for the heat transfer analysis.

Three-dimensional model for the heat transfer analysis.
User subroutine UMASFL used in
discbrake_sst_heat_symm.inp.
Axisymmetric model for the mechanical analysis.
Three-dimensional model for the mechanical analysis.
References
Maitournam, M. H., Formulation et Rsolution Numrique des Problmes Thermoviscoplastiques

en Rgime Permanent, Thse de lEcole des Ponts et Chausses, 1989.
Nguyen-Tajan, T. M.L., Modlisation Thermomcanique des Disques de Frein par une Approche
Eulrienne, Thse de lEcole Polytechnique, 2002.
Figure 5.1.21
Mesh for the axisymmetric model.
5.1.24

Printed on:
Figure 5.1.22
Mesh for the three-dimensional model.
NT11
+5.635e+02
+5.196e+02
+4.758e+02
+4.319e+02
+3.880e+02
+3.441e+02
+3.002e+02
+2.563e+02
+2.124e+02
+1.685e+02
+1.246e+02
+8.075e+01
+3.686e+01
Figure 5.1.23
Temperature distribution when the disc is heated to its peak value.
5.1.25

Printed on:
Figure 5.1.24 Temperature at node 7820 as a function of

time during the entire braking period.
S, Mises
(Ave. Crit.: 75%)
+3.321e+02
+3.048e+02
+2.775e+02
+2.502e+02
+2.228e+02
+1.955e+02
+1.682e+02
+1.409e+02
+1.135e+02
+8.623e+01
+5.891e+01
+3.159e+01
+4.266e+00
Figure 5.1.25
Mises stress distribution (Step 3, increment 60) just before the cooling period starts.
5.1.26

Printed on:
Figure 5.1.26
Evolution of the circumferential stress as a function of time.
Figure 5.1.27 Evolution of the circumferential plastic

strain as a function of time.
5.1.27

Printed on:
Figure 5.1.28 Evolution of the circumferential viscous

strain as a function of time.
Figure 5.1.29 Evolution of the circumferential stress

versus the circumferential plastic strain.
5.1.28

Printed on:
Figure 5.1.210 Evolution of the circumferential stress

versus the circumferential viscous strain.
Plastic energy
Viscous energy
Figure 5.1.211 Evolution of the plastic dissipated energy and

the viscous dissipated energy as a function of time.
5.1.29

Printed on:
EXHAUST MANIFOLD ASSEMBLAGE
5.1.3
Engine exhaust manifolds are commonly subject to severe thermal cycles during operation and upon
shutdown. Thermal expansion and contraction of the manifold is constrained by its interaction with the
engine head to which it is bolted. These constraints govern the thermo-mechanical fatigue life of the
manifold.
The initial assembly procedure consists of bolting the anges of the manifold to the engine head with
prescribed bolt forces that produce uniform axial bolt stresses. Under subsequent operating conditions such
as thermal cycling and creep, these bolt forces may increase or relax, possibly changing normal pressures and
resulting in lateral slippage between the engine head and the manifold anges. Thus, the boundary constraints
on the manifold anges are a function of the response of the entire assembly to its operating conditions. As
such, these boundary constraints cannot be prescribed a priori. This example shows how to simulate these
varying boundary constraints with the prescribed assembly load capability of Abaqus.
The problem scenario consists of three steps:
1. Apply prescribed bolt loads to fasten the exhaust manifold to the engine head.
2. Subject the assembly to the steady-state operating temperature distribution.
3. Return the assembly to ambient temperature conditions.
Geometry and model
The exhaust manifold assemblage being analyzed is depicted in Figure 5.1.31. It consists of a four tube
exhaust manifold with three anges, bolted with seven bolts to a small section of the engine head.
The manifold is cast from gray iron with a Youngs modulus of 138 GPa, a Poissons ratio of 0.283,
and a coefcient of thermal expansion of 13.8 106 per C. In this example the region of the manifold
where the hot exhaust gases converge is subject to temperatures ranging from an initial value of 300 K
to an extreme of 980 K. The elastic-plastic response of gray cast iron varies greatly over this range of
temperatures, so the temperature-dependent plasticity curves shown in Figure 5.1.32 are used for the
manifold material. Gray cast iron exhibits different behavior in tension and compression; therefore,
these curves represent the average response. The Mises metal plasticity model with isotropic hardening
is used. The three manifold anges contain a total of seven bolt holes. The 9.0 mm diameter of these
bolt holes is slightly greater than the 8.0 mm diameter of the bolt shanks to allow for some unobstructed
lateral motion of the manifold.
For simplicity, only a portion of engine head directly beneath the manifold anges is modeled.
The head is made from aluminum, with a Youngs modulus of 69 GPa, a Poissons ratio of 0.33, and a
coefcient of thermal expansion of 22.9 106 per C. The head has four exhaust ports leading into the
manifold tubes. It has seven bolt holes used to secure the manifold.
Seven bolts fasten the manifold to the head. The bolts are made from steel, with a Youngs modulus
of 207 GPa, a Poissons ratio of 0.3, and a coefcient of thermal expansion of 13.8 106 per C. The
bolt shanks have a diameter of 8 mm. The bolt head diameters are 16 mm.
5.1.31

Printed on:
Three-dimensional, deformable-to-deformable, small-sliding contact conditions apply to the model.

The bottoms of the bolt heads form contact bearing surfaces, with the top surfaces of the manifold
anges lying directly beneath them. In addition, the bottoms of the manifold anges form contact bearing
surfaces with the top of the engine head. Each of these surfaces is dened in Abaqus with the *SURFACE
option. Respective mating surfaces are paired together with the *CONTACT PAIR option. Normal
pressures will be transmitted through these contact pairs as a result of the bolt tightening forces in
Step 1. The forces carried by the bolts will vary as they respond to the thermal cycling of the assembly
in subsequent steps. These uctuations in bolt loads will result in varying normal pressures transmitted
across the contact pairs. Lateral slip of the mating components will occur if the critical frictional shear
stress limit is surpassed by lateral forces developed in the system. A friction coefcient of 0.2 is used
between all contacting surfaces. Contact conditions are not necessary between the bolt shanks and the
holes in the manifold anges because of the design clearance between them. Contact between the bolt
shanks and the holes in the engine head is not modeled.
All three structural components (manifold, head, and bolts) are modeled with three-dimensional
continuum elements. The model consists of 7450 rst-order brick elements with incompatible
deformation modes, C3D8I, and 282 rst-order prism elements, C3D6. The C3D6 elements are used
only where the complex geometry precludes the use of C3D8I elements. The C3D8I elements are
selected to represent the bending of the manifold walls with only one element through the thickness of
the tube walls.
Loading and boundary constraints
It is assumed that the engine head is securely xed to a stiff and bulky engine block, so the nodes along
the base of the head are secured in the direction normal to the base (the global x-direction) but are free
to move in the two lateral directions to account for thermal expansion. It is also assumed that the bolts
are threaded tightly into the engine head, with the bolt threads beginning directly beneath the section of
engine head modeled. Therefore, the nodes at the bottom of the bolt shanks are shared with the nodes
of the surrounding engine head elements and are also secured in the global x-direction. The manifold
anges are sandwiched between the top of the engine head and the base of the bolt heads using the
*CONTACT PAIR option. The line of action of the bolt forces (bolt shank axes) is along the global x
degree of freedom. Soft springs acting in the global y- and z-directions are attached to the outlet end of
the manifold and to the two ends of the head to suppress rigid body motions of the manifold and head,
respectively. These springs have no inuence on the solution.
In the rst step of the analysis each of the seven bolts is tightened to a uniform bolt force of 20 kN. In
subsequent steps the variation of the bolt loads is monitored as the bolts respond to the thermal loading on
the assembly as a whole. The prescribed assembly load capability of Abaqus is used. For each bolt we
dene a cut, or pre-tension section, and subject the section to a specied tensile load. As a result, the
length of the bolt at the pre-tension section will change by the amount necessary to carry the prescribed
load, while accounting for the compliance of the rest of the system. In the next step the prescribed bolt
loads are replaced by the condition that the length changes calculated in the previous step remain xed.
The remainder of the bolt is free to deform.
The same procedure is used for all seven bolts. First, pre-tension sections are dened as cuts
that are perpendicular to the bolt shank axes by using the *SURFACE option on the faces of a group of
5.1.32

Printed on:
elements within each bolt shank, as shown in Figure 5.1.33. The line of action of the bolt force is in the
direction that is normal to this surface. Next, each bolt is assigned an arbitrary, independent node that
possesses one degree of freedom (dof 1), to which the bolt force will be applied. These nodes are called
the pre-tension nodes (all seven bolt pre-tension nodes are placed into a node set named BOLTS). The
spatial position of a pre-tension node is irrelevant. Finally, each surface is associated with the appropriate
pre-tension node using the *PRE-TENSION SECTION option.
A portion of the Abaqus model denition section dening the pre-tension section is shown below:
*ELSET, ELSET=BCUT1, GENERATE
19288,19307
*SURFACE, NAME=BOLT1
BCUT1,S2
*NODE, NSET=BOLTS
99991,
21.964
, -139.80
, -12.425
99997,
21.964
,
137.38
, -12.226
PRE-TENSION
SECTION,
SURFACE=BOLT1,
NODE=99991
*
In Step 1 of the analysis a concentrated clamping load of 20 kN is applied to each of the pre-tension
nodes in node set BOLTS. In Step 2 the concentrated load from Step 1 is removed and replaced by a
xed boundary condition that will hold the pre-tension section length changes from Step 1 xed.
Over the course of a step in which a load is replaced by a boundary condition, CF1 is ramped down,
while RF1 is ramped up to replace it. Therefore, the total force across the bolt is the sum of the
concentrated force (CF1) and the reaction force (RF1) on the pre-tension node. This total force is
available as TF1. Additionally in this step of the analysis nodal temperatures depicting the steady-state
temperature distribution in the manifold are read from an external le. The temperature distribution
is shown in Figure 5.1.34. These nodal temperatures can be generated by an Abaqus heat transfer
analysis. Each of the nodes in the model has its temperature ramped up from the initial ambient
temperature of 300 K to its nal steady-state temperature. These nodal temperatures are interpolated
to the element integration points so that the correct temperature-dependent plasticity data can be used
in the constitutive calculations. Finally, in Step 3 the nodal temperatures are ramped back down to the
initial ambient temperature of 300 K.
The analysis is performed as a small-displacement analysis. The nonlinearities in the problem are the
result of changing contact conditions, frictional slip and stick, and temperature-dependent plasticity.
Figure 5.1.35 shows the lateral displacement of the bottom surface of the ange at the end of the
heat-up step. As a result of frictional sticking, the ends of the two outer manifold anges have expanded
outward relative to one another by only about 0.75 mm. Plastic yielding conditions result since thermal
expansion of the remainder of the manifold is constrained by this limited lateral ange motion. A separate
thermal-stress analysis of the manifold only, with no bolt constraints included, produced relative lateral
expansions of about 1.1 mm and very little plasticity.
5.1.33

Printed on:
Figure 5.1.36 is a plot of the forces carried by each of the seven bolts throughout the load history.
This plot can be obtained with the XY plotting capabilities in Abaqus/CAE. The curves contain the
values of the total forces (TF1) for the pre-tension nodes in node set BOLTS. The loads carried by the
bolts increase signicantly during the heat-up step. The loads do not return precisely to the original bolt
load specication upon cool down because of the residual stresses, plastic deformation, and frictional
dissipation that developed in the manifold.
Input files
manifold.inp
manifold_node_elem.inp
manifold_nodaltemp.inp
Input data for the analysis.

Node and element denitions.
Nodal temperature data.
5.1.34

Printed on:
1
3
Figure 5.1.31
Manifold assemblage.
TEMP_1173
TEMP_293
TEMP_573
TEMP_773
TEMP_973
Figure 5.1.32
Gray cast iron temperature-dependent plasticity curves.
5.1.35

Printed on:
pre-tension section
1
2
3
Figure 5.1.33
Pre-tension section.
TEMP
(Ave. Crit.: 75%)
+9.801e+02
+9.391e+02
+8.980e+02
+8.570e+02
+8.160e+02
+7.749e+02
+7.339e+02
+6.929e+02
+6.519e+02
+6.108e+02
+5.698e+02
+5.288e+02
+4.877e+02
3
1
2
Figure 5.1.34
Steady-state temperature distribution.
5.1.36

Printed on:
U, U2
+9.789e-01
+8.506e-01
+7.223e-01
+5.940e-01
+4.657e-01
+3.374e-01
+2.091e-01
+8.085e-02
-4.744e-02
-1.757e-01
-3.040e-01
-4.323e-01
-5.606e-01
3
1
Figure 5.1.35
Lateral expansion of manifold footprint.
BOLT-1
BOLT-2
BOLT-3
BOLT-4
BOLT-5
BOLT-6
BOLT-7
Figure 5.1.36
History of bolt forces.
5.1.37

Printed on:
COOLANT MANIFOLD COVER GASKETED JOINT
5.1.4
Engine gaskets are used to seal the mating surfaces of engine components to maintain the integrity of
the closed system throughout a wide range of operating loads and environmental conditions. Inadequate
gasket performance leads to diminished engine pressure and uid leakage, resulting in degradation of
engine performance and potential engine damage. The gasket, the engine component anges, and the
fastenerscollectively referred to as a gasketed jointmust be considered as a unit when determining the
system sealing performance because most gasketed joints do not obtain a uniform contact stress distribution
due to nonuniform bolt spacing and ange distortion during assembly and subsequent operational loading.
Engine gaskets are often complicated geometric constructs of various engineering materials and are
subject to large compressive strains. The compressive response of the gasket is highly nonlinear. Such
complexities make detailed modeling of gaskets with continuum elements difcult and impractical when
analyzing complete assemblies.
Abaqus has a dedicated class of elements, referred to as gasket elements, that simplify the modeling
of such components while maintaining the essential ingredients of the nonlinear response. Typical use
of these gasket elements involves a tabular representation of the pressure versus closure relationship in
the thickness direction of the gasket. The pressure versus closure models available in Abaqus allow the
modeling of very complex gasket behaviors, including nonlinear elasticity, permanent plastic deformation,
and loading/unloading along different paths. These behaviors are usually calibrated directly from test data.
In this manner a complex gasket can be modeled effectively using a single gasket element in the thickness
direction.
In this example a paper foam gasket with a silkscreened silicone bead is compressed between the lower
engine intake manifold and the coolant manifold cover. The coolant manifold cover seals the lower intake
manifold coolant passages so that the coolant can be distributed to the cylinder heads. An exploded view
of the gasketed joint model is shown in Figure 5.1.41. It consists of two steel bolts, an aluminum coolant
manifold cover, a paper foam gasket with a silicone bead, andfor simplicityonly a portion of the lower
intake manifold, which is composed of steel. Symmetry conditions reduce the structure to a half model. The
gasketed joint is subjected to the following mechanical and environmental load conditions:
1.
2.
3.
4.
5.
Simulate the bolt loading sequence to fasten the joint.

Heat the assembly to the maximum operating temperature and apply interior cavity pressure.
Cool the assembly to the minimum operating temperature while maintaining interior cavity pressure.
Return the assembly to ambient conditions with the interior pressure removed.
Disassemble the gasketed joint.
The portion of the lower intake manifold that is modeled has two passages. Coolant ows from one
passage into the manifold cover and back out through the other passage. Two steel bolts secure the cover
to the manifold. The bolt shanks have a diameter of 6.0 mm, and the bolt heads have a diameter of
5.1.41

Printed on:
11.8 mm. The bolts and the lower intake manifold are assigned a Youngs modulus of 2.0 105 MPa, a
Poissons ratio of 0.28, and a coefcient of thermal expansion of 1.6 105 per C. The aluminum coolant
manifold cover has a Youngs modulus of 7.1 104 MPa, a Poissons ratio of 0.33, and a coefcient of
thermal expansion of 2.3 105 per C.
The metal components (bolts, cover, and intake manifold) are modeled with three-dimensional
continuum elements: 1304 rst-order brick elements with incompatible deformation modes (C3D8I) and
208 rst-order prism elements (C3D6). The C3D8I elements are chosen to capture the bending of the
cover, using only one element through its thickness. The C3D6 elements are used only where geometric
constraints preclude the use of C3D8I elements.
The gasket schematic shown in Figure 5.1.42 has two distinct regions. The majority of the gasket
is composed of a 0.79 mm thick, at, crushable paper foam material. To ensure proper sealing pressures
for this joint, a 0.076 mm thick silicone bead has been silkscreened along the top surface of the gasket
encircling the interior cavity. Placing silicone beads on gaskets results in a change in the load transmitting
characteristics of the gasket, which often improves both the recovery properties of the gasket and its
potential to remain sealed for the long term.
The entire gasket, including the bead, is modeled as a at sheet with one gasket element through
the thickness (see Figure 5.1.43). A ne mesh is used for the gasket to capture the in-plane variation
of the gasket sealing pressure. This creates a mismatched mesh across the contacting surfaces, but
Abaqus contact denitions do not require one-to-one matching meshes across contact pairs. The gasket
components (silicone bead region and paper foam region) are modeled with 973 rst-order 8-node area
elements (GK3D8) and 29 rst-order 6-node area elements (GK3D6). The physical thickness of the
entire sheet of gasket elements corresponds to the initial combined height of the paper foam and the
silicone bead, 0.866 mm. The elements in the region of the gasket beneath the silicone bead are assigned
different gasket properties from the rest of the elements in the gasket model. The paper foam region
is initially not in contact with the cover. The initial gap is 0.076 mm. No pressure is generated in this
portion of the gasket until the gap has closed. Gasket region property distinctions, such as initial gaps and
different pressure versus closure relationships, are assigned to corresponding element sets by referring
to different *GASKET SECTION options.
Experimentally determined pressure versus closure curves for the two distinct gasket regions
without the initial gap taken into account are shown in Figure 5.1.44. Tabular representations of
these curves are specied using the *GASKET THICKNESS BEHAVIOR option that is associated
with the respective *GASKET SECTION options. Creep/relaxation properties of the gasket and
temperature-dependent pressure versus closure properties, capturing such effects as the glassy transition
temperature of the silicone bead, are not accounted for in this example. Initially, Abaqus considers
the gasket behavior to be nonlinear elastic, such that loading and unloading occur along the same
user-dened nonlinear path. Abaqus considers yielding to occur once the slope of the pressure versus
closure curve decreases by at least 10%. In addition to the single loading curve, whose closure increases
monotonically, the user can dene any number of unloading curves at different levels of plastic closure.
Yielding occurs at a closure of 0.1118 mm for both regions of the gasket in this example, after which
the gasket stiffness decreases slightly up to a closure of 0.15 mm, the nal point on the loading curve.
Beyond the data of the loading curve dened by the user, Abaqus considers the gasket to behave with a
5.1.42

Printed on:
fully crushed elastic response by linearly extrapolating the last segment of the last specied unloading
curve (alternatively, the user could have specied a piecewise linear form).
A single unloading curve is dened for each of the two gasket regions: the unloading curve for the
silicone bead region is dened at 0.11 mm of plastic closure, and the unloading curve for the paper foam
region is dened at 0.09 mm of plastic closure. Any unloading of the gasket beyond the yield point occurs
along a curve interpolated between the two bounding unloading curves, whichfor this exampleare
the initial, nonlinear elastic curve and the single unloading curve.
Gasket materials often have higher coefcients of thermal expansion than most of the metals
from which the bolts and anges are made. For situations involving wide and rapid temperature
uctuations resultant differences in relative expansion and contraction can have a signicant effect on
the sealing properties of the gasket. The coefcient of thermal expansion for the silicone bead region is
1.2 104 per C, and for the paper foam region it is 3.0 105 per C.
In this case, because of the differences in thermal expansion between the aluminum cover and the
steel intake manifold, it is important to account for the membrane and transverse shear properties of
the gasket and to model frictional effects between mating surfaces. For this analysis the silicone bead
region of the gasket is dened to have a membrane stiffness of 75 MPa and a transverse shear stiffness
of 40 MPa. The base foam material is dened with a value of 105 MPa for the membrane stiffness and a
value of 55 MPa for the transverse shear stiffness. A friction coefcient of 0.2 is used between all mating
surfaces.
A separate analysis is included in this example problem using the thickness-direction only version
of the gasket elements (GK3D8N and GK3D6N). These elements respond only in the thickness direction
and have no membrane or transverse shear stiffness properties. They possess only one degree of freedom
per node. As a result, frictional effects cannot be included at the surfaces of these elements. They
are more economical than more general gasket elements that include membrane and transverse shear
responses and may, thus, be preferable in models where lateral response can be considered negligible.
Loading and boundary constraints
Symmetry boundary constraints are placed along the nodes on the symmetry plane. Furthermore, it is
assumed that the intake manifold is a stiff and bulky component, so nodes along the base of the portion
of the manifold modeled are secured in the normal direction (the global z-direction). Except for a soft
spring constraint to eliminate rigid body motion, these manifold base nodes are free to displace laterally
to allow for thermal expansion. Soft springs are also attached to the cover to eliminate rigid body motion
in the x- and z-directions.
The bottoms of the bolt heads form contact bearing surfaces with the top surface of the cover
ange. In addition, the top of the gasket interacts with the bottom of the cover, while the bottom of
the gasket contacts the top of the manifold. Each of these surfaces is dened with the *SURFACE
option. Mating surfaces are paired together with the *CONTACT PAIR option. Three-dimensional,
deformable-to-deformable, small-sliding contact conditions apply to each of these contact pairs. The
gasket is attached to the manifold base using the *SURFACE BEHAVIOR, NO SEPARATION option,
thus constraining it against rigid body motion in the global z-direction. The gasket membrane is allowed
to stretch, contract, or shear as a result of frictional effects on both sides of the gasket. The bolts are
5.1.43

Printed on:
assumed to be threaded tightly into the base. Therefore, the nodes at the bottom of the bolt shanks are
shared with the intake manifold. Contact between the bolt shanks and the bolt holes is not modeled.
The prescribed assembly load capability is used to dene pre-tension loads in each of the bolts.
For each of the two bolts we dene a cut or pre-tension section and subject the section to a specied
load. As a result, the length of the bolt at the pre-tension section changes by the amount necessary to
carry the prescribed load, while accounting for the compliance of the rest of the joint. Once a bolt has
been pre-tensioned, the applied concentrated bolt load is replaced with a xed boundary condition,
which species that the length change of the bolt at the cut remains xed, while the remainder of the
bolt is free to deform.
The sequence in which the bolts are tightened can have an impact on the distribution of the resultant
contact area stress. A poorly specied bolt sequence can cause excessive distortion of the gasket and
the anges, which may lead to poor sealing performance. In the rst step of the analysis the left bolt
is pre-tensioned to a load of 6000 N using the *PRE-TENSION SECTION option. In the second step
the right bolt is pre-tensioned to 6000 N and the prescribed load on the left bolt is replaced with a xed
boundary condition as described above. Since only half of each bolt is modeled, a total load of 12000 N
is carried by each bolt.
Step 3 is the beginning of the three-step thermo-mechanical operational cycle. In Step 3 the entire
assembly is heated uniformly to its maximum operating temperature of 150C, while simultaneously the
interior cavity is pressurized to 0.689 MPa and the prescribed load on the pre-tension section of the right
bolt is replaced with a xed boundary condition. In Step 4 the system temperature is decreased to the
minimum operating temperature of 40C while maintaining the interior pressure load of 0.689 MPa.
In Step 5 the gasketed joint is returned to the ambient temperature conditions and the internal cavity
pressure is removed.
The sixth and nal step in the analysis simulates disassembly of the gasketed joint by removing the
bolt loads. This process demonstrates the interpolated unloading response for the different regions of the
permanently deformed gasket.
The prime interest in this problem is the variation of bolt forces during the initial assembly and thermomechanical cycle and the resultant distribution and variation of the gasket sealing pressure.
The function of the fasteners in a gasketed joint is to apply and maintain the load required to seal the
joint. The bolt pattern and tension are directly related to the sealing pressure in the clamped gasket. At
the maximum service temperature the bolt loads can be expected to be at their peak as a result of thermal
expansion effects. It is important to ensure that the stress values of the metal engine components remain
below yield and that there is no signicant bending of the anges, which may cause improper sealing of
the gasket. At the minimum operating temperature the bolt loads are expected to reach a minimum as
a result of thermal contraction effects. Hence, it is necessary to assess that adequate sealing pressure is
retained throughout the gasket.
Figure 5.1.45 shows the bolt load variation over the course of the six analysis steps. During the
rst step the pre-tension section node on the right bolt was prescribed a zero change of length constraint,
which implies that the right bolt has just been placed in position but not torqued tightly. Hence, as the
left bolt is tightened during Step 1, a small reaction load is generated in the right bolt. At the end of
5.1.44

Printed on:
the second step during which the right bolt is tightened to carry a force of 6000 N, the force in the left
bolt increases to 6200 N. In Step 3 the deformation of the assembly causes the bolt forces to increase to
maximum values of 6800 N in the left bolt and 6600 N in the right bolt because of thermal expansion
and interior pressurization. When the assembly is cooled to the minimum operating temperature, the bolt
loads reach their minimum values. Due to thermal cycling and interior cavity pressure inducing inelastic
response in the gasket, the bolt forces at the end of the operational cycle reduce to 6050 N in the left bolt
and 5950 N in the right bolt.
The gasket sealing pressure pattern depends on the rigidity of the anges. Hence, it is useful to
predict how the structure will deform due to the applied loading. Figure 5.1.46 shows the deformed
shape of the coolant manifold cover at a displacement magnication factor of 50. Bowing of the cover
from initial assembly and subsequent operational loads will lead to a nonuniform sealing pressure
distribution in the gasket. Figure 5.1.47 illustrates the gasket pressure distribution after initial fastening
of the joint.
Figure 5.1.48 shows the sealing pressure as a function of position along the perimeter of the silicone
bead at the end of each of the analysis steps. The sealing pressure reaches a minimum at the point
equidistant from the bolts, making this the critical point in the gasketed joint design. This gure also
reects the reduction in the sealing pressure near the bolt holes as a result of plastic deformation of the
gasket body during the operational cycle. Figure 5.1.49 is a contour plot of the permanent deformation
in the gasket after completion of the thermo-mechanical cycle.
Figure 5.1.410 follows the pressure/closure history of one point in the gasket during this analysis in
relation to the user-specied loading/unloading test data. The mechanical closure (total closure, E11,
minus thermal closure, THE11) is plotted along the abscissa of this gure. The material point traced
(element 18451, integration point 1) is located along the inside periphery of the silicone bead at the
symmetry plane of the assembly nearest the left bolt. Step 1 shows that this point follows the initial elastic
loading curve up to the closure of 0.1118 mm. After this amount of closure, further loading causes plastic
deformation. In the second step the tightening of the bolt results in a very small amount of unloading for
this material point. For purposes of clarity, this deformation is not shown in the gure. Step 3 involves
heating the system to the maximum operating temperature and pressurizing the interior cavity so that
further yielding of the material point occurs. Step 4 results in the partial unloading of the point due to the
thermal contraction associated with cooling the assembly to the minimum operating temperature. For
this case the unloading path is based on a curve interpolated between the initial, nonlinear elastic curve
and the single unloading curve. The return of the assembly to ambient conditions partially reloads this
point along the same path as the previous unloading; however, no further yielding of this material point
occurs during this step. In the nal step the gasket is unloaded completely.
The analysis using the thickness-direction only gasket elements runs in nearly half the CPU time
of the full three-dimensional gasket element model. Minimum gasket sealing pressures in Step 4 of this
analysis are predicted to be about 20% lower because frictional effects are neglected.
Input files
manifoldgasket.inp
manifoldgasket_mesh.inp
Input data for the analysis.

Node, element, and surface denitions.
5.1.45

Printed on:
manifoldgasket_thick.inp
manifoldgasket_thick_mesh.inp
Thickness-direction only gasket element analysis.

Node, element, and surface denitions for the
thickness-direction only gasket element analysis.
Reference
Czernik, D. E., Gasket Handbook, McGraw-Hill, New York, 1996.
right bolt
left bolt
coolant manifold
cover
gasket
partial model of
lower intake
manifold
3
2
1
fluid flow
Figure 5.1.41
Coolant manifold assemblage.
5.1.46

Printed on:
silicone bead
gasket body
top view
cross-section
0.076 mm
0.866 mm
Figure 5.1.42
Schematic representation of a silicone bead printed on the gasket body.
5.1.47

Printed on:
2
3
1
Figure 5.1.43
Mesh of gasket with silicone bead highlighted.
60.
gasket
gasket w/ bead
50.
pressure (MPa)
40.
30.
20.
10.
0.
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
closure (mm)
Figure 5.1.44
Pressure versus closure behavior for the gasket and the gasket with silicone bead.
5.1.48

Printed on:
7.
[ x10 3 ]
LEFT BOLT
RIGHT BOLT
6.
BOLT FORCE (Newtons)
5.
4.
3.
2.
1.
0.
0.
1.
2.
3.
4.
5.
TOTAL TIME
Figure 5.1.45
History of bolt force.
3
1
Figure 5.1.46 Deformed shape of coolant manifold cover

at a displacement magnication factor of 50.
5.1.49

Printed on:
6.
S11
VALUE
+9.26E-03
+4.09E+00
+8.17E+00
+1.23E+01
+1.63E+01
+2.04E+01
+2.45E+01
+2.86E+01
+3.27E+01
+3.68E+01
+4.08E+01
+4.49E+01
+4.90E+01
+5.31E+01
Figure 5.1.47
Gasket pressure distribution after initial fastening sequence.
45.
STEP
STEP
STEP
STEP
40.
2
3
4
5
35.
Sealing Pressure (MPa)
30.
25.
20.
15.
10.
5.
Center
Line
0.
0.
Center
Line
20.
40.
60.
80.
Distance along Inner Periphery of Silicone Bead (mm)
Figure 5.1.48
Sealing pressure along inside periphery of silicone bead region of gasket.
5.1.410

Printed on:
PE11
VALUE
+0.00E+00
+5.95E-03
+1.19E-02
+1.79E-02
+2.38E-02
+2.98E-02
+3.57E-02
+4.17E-02
+4.76E-02
+5.36E-02
+5.95E-02
+6.55E-02
+7.14E-02
+7.74E-02
Figure 5.1.49
Plastic closure in gasket after operational cycle.
48.
44.
gasket w/ bead
STEP 1
STEP 3
STEP 4
STEP 5
STEP 6
40.
36.
Pressure (MPa)
32.
28.
24.
20.
16.
12.
8.
4.
0.
0.00
0.05
0.10
0.15
Closure (mm)
Figure 5.1.410
Typical pressure-closure diagram for material point in silicone bead region of gasket.
5.1.411

Printed on:
EXHAUST MANIFOLD HEAT TRANSFER
5.1.5
CONDUCTIVE, CONVECTIVE, AND RADIATIVE HEAT TRANSFER IN AN EXHAUST

MANIFOLD
Objectives
This example demonstrates the following Abaqus features and techniques:
computing steady-state heat transfer in an exhaust manifold,

comparing results for radiation heat transfer formulations using approximate and fully implicit
methods, and
using lm conditions to simulate the convective heat transfer from the exhaust gases.
Heat transfer in engine exhaust manifolds is governed by three effects: conduction through the metal,
convection from the hot exhaust gases, and radiative exchange between different parts of the metal
surface. This example illustrates the computation of the equilibrium thermal state of a manifold subject
to these effects. The units of length in this example are millimeters; otherwise, standard metric units are
used.
The procedure consists of a single heat transfer step in which the thermal loading conditions are
ramped up from zero. The boundary constraints on the manifold anges are a simplication of those
experienced under operating conditions: the temperatures at the cylinder head and the outlet are xed.
Convection due to heat transfer from the hot exhaust is applied at the internal surfaces of the manifold
tubes. Radiation is modeled between the internal surfaces of the tubes using two methods: the fully
implicit cavity radiation method and an approximate cavity radiation method.
Geometry
The exhaust manifold part being analyzed is depicted in Figure 5.1.51. It consists of a four tube exhaust
manifold with three anges, as in Exhaust manifold assemblage, Section 5.1.3.
Materials
The manifold is cast from gray iron with a thermal conductivity of 4.5 102 W/mm/C, a density of
7800 109 kg/mm3 , and a specic heat of 460 J/kg/C. The manifold begins the analysis with an initial
temperature of 20C. The part is dimensioned in millimeters, and the temperature is measured in C, so
the Stefan Boltzmann constant is taken as 5.669 1014 W/mm2 /K4 and absolute zero is set at 273.15C
below zero. The surface emissivity of gray iron is taken as a constant value of 0.77.
Initial conditions
The initial temperature of the manifold is set to 20C.
5.1.51

Printed on:
The hot exhaust gases create a heat ux applied to the interior tube surfaces. In this example this effect is
modeled using a surface-based lm condition, with a constant temperature of 816C and a lm condition
of 500 106 W/mm2 /C. A temperature boundary condition of 355C is applied at the ange surfaces
attached to the cylinder head, and a temperature boundary condition of 122C is applied at the ange
surfaces attached to the exhaust.
The radiative transfer between the interior surfaces of the manifold tubes is modeled using two methods
for comparison: the fully implicit cavity radiation method and the approximate cavity radiation method
(see Cavity radiation, Section 37.1.1 of the Abaqus Analysis Users Manual). In the fully implicit
method, geometric viewfactors are computed in Abaqus between each facet of the mesh on the exposed
interior tube surface. These viewfactors quantify the effect of radiative transfer between each facet and
each of the other facets in the user-dened cavity. The viewfactors, in turn, are used to compute a fully
populated interaction matrix to compute the radiation ux between each pair of facets in the model. In
the approximate method, the geometric viewfactors are approximated by assuming that each facet has
an equal view of all other facets; this has the effect of modeling the ux at each facet as equal to that
resulting from a black enclosure, held at the average temperature in the cavity, enclosing the facet. In
the fully implicit method, some of the facets on the interior of the manifold have a view of the exterior,
which is not modeled in this example. The exterior ambient temperature is taken to be the average of the
temperatures used for the cylinder head and exhaust boundary conditions. In the approximate method,
only the temperatures on the surface are considered, so an ambient temperature does not need to be
dened. For simplicity, both the fully implicit method and the approximate method are dened using a
single surface that includes all of the interior facets of the manifold tubes.
Case 1
Steady-state heat transfer with lm and radiation effects; radiation modeled using the
fully implicit method.
Case 2
Steady-state heat transfer with lm and radiation effects; radiation modeled using the
approximate method.
The following sections discuss analysis considerations that are applicable to both cases.
Analysis types
Due to the fourth-order dependence of the radiation ux on the surface temperatures, this example
problem is intrinsically nonlinear. For both cases the steady-state heat transfer procedure is used. This
is a general analysis step in Abaqus, chosen because iteration is required for convergence. An initial
increment is chosen as one-tenth of the nal value.
5.1.52

Printed on:
Mesh design
The manifold is meshed using linear hexahedral and wedge heat transfer elements with linear
interpolation.
Figure 5.1.52 shows the nodal temperature eld for the manifold. On the left, the analysis results using
the fully implicit method are shown; on the right, results from the approximate method are shown. In
this problem we observe good agreement between the two methods, although some differences can be
discerned in the plots.
The peak temperature in the eld is higher when using the fully implicit method. The
effect of radiation heat transfer is to smooth out the temperature eld in the equilibrium solution:
high-temperature zones radiate more heat, which is absorbed by the cooler areas. In the fully implicit
method, this smoothing effect is limited and affected by the geometric viewfactors: the distance and
orientation of the surface facets affects the degree to which radiation exchange can occur. In the
approximate method, this is not the case. Each facet absorbs or emits radiative heat ux based on its
temperature and the averaged cavity temperature only; the localizing effects of viewfactors are ignored.
Therefore, the average method results reect the greater smoothing effect of the radiation model used,
resulting in lower peak values.
Figure 5.1.53 shows the ux magnitude results. The ux eld shows even greater agreement than
the temperature eld.
Because the fully implicit radiation algorithm uses a fully populated matrix operator to model
the interactions of each facet, it is signicantly more computationally expensive than the approximate
radiation method. Table 5.1.51 illustrates the differences between the two methods. In this problem the
cavity surface contained 4505 facetsit consists of the entire interior of the manifold. The savings in
memory are quite signicant and directly reect the cost of the large operator used in the fully implicit
method. The memory and timing results were obtained on a desktop computer using Xeon processors,
but the relative comparisons between run times are more pertinent than the specic run times.
Input files
heattransfermanifold.inp
heattransfermanifold_cavity.inp
Input data for the analysis using the approximate radiation

method.
Input data for the analysis using the fully implicit
radiation method.
References
Steady-state analysis in Uncoupled heat transfer analysis, Section 6.5.2 of the Abaqus Analysis
Users Manual
Cavity radiation, Section 37.1.1 of the Abaqus Analysis Users Manual
5.1.53

Printed on:
*CAVITY DEFINITION
*HEAT TRANSFER
*RADIATION VIEWFACTOR
*SFILM
*SRADIATE
Abaqus Theory Manual
Uncoupled heat transfer analysis, Section 2.11.1 of the Abaqus Theory Manual
Cavity radiation, Section 2.11.4 of the Abaqus Theory Manual
Viewfactor calculation, Section 2.11.5 of the Abaqus Theory Manual
Table 5.1.51
Relative computational costs of the approximate and fully implicit methods.

Approximate
method
Fully implicit
method
Approximate
ratio
Minimum memory required (MB)
31
501
16
Memory to minimize I/O (MB)
52
1052
20
User time (sec)
14.6
621.5
42
System time (sec)
0.3
10.5
35
Total CPU time (sec)
14.9
632
42
Wallclock time (sec)
16
653
41
Increments
Total iterations
10
0.7
Wallclock time per iteration
1.6
93.3
58
5.1.54

Printed on:
Figure 5.1.51
NT11
+8.101e+02
+7.528e+02
+6.954e+02
+6.381e+02
+5.807e+02
+5.234e+02
+4.661e+02
+4.087e+02
+3.514e+02
+2.940e+02
+2.367e+02
+1.793e+02
+1.220e+02
Figure 5.1.52
Manifold mesh.
NT11
+8.101e+02
+7.528e+02
+6.954e+02
+6.381e+02
+5.807e+02
+5.234e+02
+4.661e+02
+4.087e+02
+3.514e+02
+2.940e+02
+2.367e+02
+1.793e+02
+1.220e+02
Equilibrium temperature eld in the manifold using the fully implicit

(left) and approximate (right) radiation methods.
5.1.55

Printed on:
HFL, Magnitude
(Avg: 75%)
+1.050e+00
+9.545e01
+8.591e01
+7.636e01
+6.682e01
+5.727e01
+4.773e01
+3.818e01
+2.864e01
+1.909e01
+9.545e02
+0.000e+00
HFL, Magnitude
(Avg: 75%)
+1.050e+00
+9.545e01
+8.591e01
+7.636e01
+6.682e01
+5.727e01
+4.773e01
+3.818e01
+2.864e01
+1.909e01
+9.545e02
+0.000e+00
Figure 5.1.53 Equilibrium heat ux magnitude eld in the manifold using the fully
implicit (left) and approximate (right) radiation methods.
5.1.56

Printed on:
REACTOR PRESSURE VESSEL BOLTED CLOSURE
5.1.6
THERMAL-STRESS ANALYSIS OF A REACTOR PRESSURE VESSEL BOLTED

CLOSURE
Abaqus/CAE
Objectives
This example demonstrates the following Abaqus features and techniques for heat transfer and static
stress analyses:
specifying adaptive remeshing rules in different regions of a model in a particular analysis step;
using an automated process to remesh the model adaptively based on the remeshing rules specied;
and
viewing error indicator results as a means of assessing mesh quality.
This example examines the thermal and stress behavior of the bolted closure region of a nuclear reactor
vessel assembly. The vessel assembly forms the pressure boundary surrounding the fuel core. This
example considers the strength of sustaining the following loading conditions:
pre-tension load in the stud bolts,

constant internal pressure, and
specied heat-up/cool-down rate.
These loading conditions cover the most basic design requirements of a reactor vessel. A short and rapid
temperature change is one of the most severe loading cases and will be considered in this example. The
International System of units (SI) will be used in the following sections to describe the model. The
analysis itself is performed in English units. The model and analysis are derived from details of the
Shippingport pressurized water reactor (1958).
Geometry
The problem domain comprises a cylindrical vessel shell, a hemispherical bottom head, a dome-shaped
closure head, and the closure and seal assembly, as shown in Figure 5.1.61. The overall height of the
vessel shell including the bottom head is 7650 mm (301 in). The bottom head has an inner radius of
1410 mm (55.5 in) and a thickness of 157 mm (6.18 in). The inner radius of the vessel shell is 1380 mm
(54.5 in), and the thickness is 213 mm (8.40 in). The closure head has a height of 2330 mm (91.8 in),
an inner radius of 1310 mm (51.5 in), and a thickness of 210 mm (8.25 in). The closure head is attached
to the vessel shell by a seal and closure assembly. The assembly includes 40 stud bolts passing through
the bolting anges of the closure head and the vessel shell, each of which is restrained by two cap nuts.
To complete the closure assembly, an omega seal is welded to the under surface of the closure head and
top surface of the vessel shell. The stud bolt is 2290 mm (90 in) in length and has a diameter of 146 mm
(5.75 in). The closure nuts are 304 mm (12 in) long with a thickness of 28.6 mm (1.13 in).
5.1.61

Printed on:
The outside of the vessel is exposed to air that has a constant temperature of 21C (70F). The inside of
the vessel is lled with hot water with an operating temperature of 320C (600F). During a cool-down
process, the internal temperature is reduced by 38C (100F) in two hours. The water inside imposes a
constant pressure of 1.38 107 Pa (2000 psi) on the internal surface of the vessel.
The objective of this analysis is an understanding of stresses near the vessel-to-head interfaces.
Although the assembly contains many features, such as inlet and outlet nozzles, the example ignores
these details since they are far away from the vessel-to-head interface. The rest of the geometry
is cyclically symmetric, which allows the example to model the entire 360 structure at a reduced
computational expense by analyzing only a single repetitive sector of the model. Since there are 40 stud
bolts along the circumference of the reactor vessel, the following analysis is performed on a 9 model
with one sector as shown in Figure 5.1.62.
The example also takes advantage of the fact that the thermal and mechanical responses of the
vessel are only weakly coupled. Based on this fact, a sequentially coupled thermal-stress analysis is
performed on the reactor vessel. The distribution of the temperature eld is obtained rst through a heat
transfer analysis, then the mechanical response of the vessel is obtained by performing a static stress
analysis with the temperature eld specied using the results of the thermal analysis.
Case 1
Steady-state and transient heat transfer analyses, with adaptive remeshing in

Abaqus/CAE.
Case 2
Static stress analysis, with adaptive remeshing in Abaqus/CAE.
The following sections discuss analysis considerations that are applicable to both analyses. More detailed
descriptions are provided later including discussions of results and listings of the les provided. The
models for the two analyses were generated using Abaqus/CAE and imported ACIS-format les.
Analysis types
The thermal analysis includes a steady-state and a transient heat transfer step. The structural analysis is
performed using multiple linear general static steps.
Mesh design
The omega seal is meshed with rst-order brick elements, while the rest of the model is meshed with
second-order tetrahedral elements. The geometry is partitioned to create a ne initial mesh in the area
near the closure assembly, where the geometry is most complex and high stress and heat ux are expected.
The purpose of this mesh design is to obtain an accurate estimate of the error indicators specied in the
remeshing rules, which will result in faster convergence in the adaptivity procedure.
5.1.62

Printed on:
Adaptive remeshing rules
The patch recovery techniques that Abaqus/CAE uses to calculate the error indicator variables can
have a signicant impact on the analysis solution time. In particular, the element energy density is
calculated after each increment and is more costly than the other error indicator variables. To reduce the
computational expense, the omega seal is excluded from the region where remeshing rules are specied.
The closure head and the vessel shell are also partitioned in such a way that regions relatively farther
away from the stud bolt are not included in the remeshing regions. Four separate remeshing rules are
dened in the closure head, the vessel shell, the cap nuts, and the stud bolt. In both the thermal and
structural analyses, the remeshing rules are specied in only the last step.
Additional details of the remeshing rules used in the two analyses are discussed with each example.
Constraints
To simulate the welding constraints, the bottom surfaces of the omega seal are tied to the surfaces of the
bolting anges in both analyses.
Adaptivity process
Each adaptivity process species a maximum number of three remesh iterations.

Heat transfer analysis
The example starts by performing a heat transfer analysis to obtain the temperature distribution in the
pressure vessel under the thermal loading.
Analysis types
The analysis consists of a steady-state heat transfer step, representing the steady operation of the reactor.
This step is followed by a transient heat transfer step, representing a rapid cool-down event. The resulting
temperatures obtained are applied to the subsequent mechanical analysis.
Mesh design
When performing the heat transfer analysis, rst-order hexahedral diffusive heat transfer elements
(DC3D8) are used in the omega seal, and the rest of the geometry is meshed with second-order
tetrahedral diffusive heat transfer elements (DC3D10).
The heat ux error indicator is chosen in all remeshing rules. For each remeshing rule the sizing method
is set to uniform error distribution and the error indicator target is set to automatic target reduction.
Material model
The heat transfer analysis requires specication of thermal conductivity, which is 46.7 W/m/C
(2.25 Btu/h/in/F), and specic heat, which is 460 J/kg/C (0.11 Btu/lb/F). The density of the material
5.1.63

Printed on:
is also specied, which is 7850 kg/m3 (0.284 lb/in3 ). One solid, homogenous section is used to assign
material properties to the elements.
Initial conditions
The initial temperature is set to 21C (70F) in the entire model.

Boundary conditions
No temperature boundary conditions are applied. The thermal response of the model is driven entirely
by thermal loading through lm coefcients.
Interactions
Conductive heat transfer is dened between adjacent/contacting surfaces, and a gap conductance
coefcient is specied, which is 1220 W/m2 /C (1.5 Btu/h/in2 /F). Heat ux on the surfaces is applied
by lm conditions. The outer surfaces are exposed to air, which has a lm coefcient of 28 W/m2 /C
(0.035 Btu/h/in2 /F). The inner surfaces are in contact with water with a lm coefcient of 580 W/m2 /C
(0.70 Btu/h/in2 /F). The outer surfaces are initially associated with a sink temperature of 21C
(70F), and the inner surfaces, 320C (600F). During a subsequent two-hour cooling process, the sink
temperature associated with the inner surfaces is reduced by 38C (100F).
Analysis steps
The heat transfer analysis is performed using a steady-state step followed by a transient step. The purpose
of the rst step is to obtain a steady-state solution of temperature distribution in the whole model. The
second step lasts for 7200 seconds (2 hours), and it simulates the response of the thermal model during
a rapid cooling process.
Convergence
The convergence of the error indicator HFLERI during a three-iteration adaptivity process is presented
in Table 5.1.61. Convergence is observed in each of the remeshing regions.
Run procedure
The model for the heat transfer analysis is generated using Abaqus/CAE to import the geometry, create the
thermal loading, mesh the assembly, create the remeshing rules, and run the adaptivity process. Python
scripts are provided to build the model and to submit the adaptivity process. The scripts can be run
interactively or from the command line.
To create the heat transfer model, select FileRun Script from the Abaqus/CAE main menu and
select adaptReactorVesselHT_model.py.
When you are ready to run the adaptivity process, select FileRun Script and select
adaptReactorVesselHT_job.py. After the Abaqus Scripting Interface scripts have created
the model and run the adaptivity process, you can use Abaqus/CAE to view the model and to explore
variations of the example.
5.1.64

Printed on:
The magnitude of the heat ux (HFL) along a particular path near the head to vessel interface is shown
in Figure 5.1.63. This gure shows the spatial variation of the heat ux as the mesh is rened through
the three remesh iterations. Figure 5.1.64 shows the original mesh and the nal mesh resulting from the
adaptive remeshing process. The rened mesh shows how Abaqus/CAE reacted to the higher temperature
gradients in the bolted ange region.
Structural analysis
The temperature distribution calculated in the heat transfer case will now augment bolting and pressure
loads to dene the structural loading of the vessel assembly.
Analysis types
A series of static steps is performed to simulate the mechanical response of the model under both thermal
and force loading.
Mesh design
The omega seal is meshed with rst-order reduced-integration continuum elements (C3D8R), and the
rest of the geometry is meshed with modied second-order tetrahedral elements (C3D10M).
The response of the model varies from step to step during the analysis; therefore, the time historydependent error indicator ENDENERI is chosen to capture the extreme of the models response to the
load history. For each remeshing rule the sizing method is set to uniform error distribution and the error
indicator target is set to automatic target reduction.
Material model
The linear static structural analysis requires specication of Youngs modulus, which is 2.07 107 N/m2
(3.0 107 lbf/in2 ), and Poissons ratio, which is 0.29. A thermal expansion coefcient is also dened,
which is 6.3 106 . One solid, homogenous section is used to assign material properties to the elements.
Boundary conditions
Symmetry boundary constraints are placed on the two side surfaces of the sector. Since the two
symmetry constraints overlap at the center line and such denitions are not allowed by the analysis
input le processor, partitions are made on the side surfaces so that the center line and a small part of
its surrounding region are excluded from the symmetry boundary constraints. The nodes on the center
line are constrained separately and are free to move only in the axial direction. The center node on the
outer surface of the bottom head is xed to prevent rigid body motion.
5.1.65

Printed on:
Loading
A pre-tension load of 2200 kN (5 106 lbf) is applied to the stud bolt. The inner surfaces of the head
and the vessel shell are subject to a constant pressure of 1.38 107 Pa (2000 psi) from the water.
Predefined fields
When the bolt loading is applied in the pre-assembly step, a constant temperature of 21C (70F) is
applied. The temperature eld is specied using the thermal results from the previous steady-state heat
transfer analysis when the inner pressure is applied. The results of the temperature after each increment
during the transient heat transfer analysis are imported in the last step when no additional loading is
applied.
Interactions
The outer surface of the stud bolt is tied to the inner surfaces of the cap nuts. Small-sliding surfaceto-surface contact interactions are dened between the contact surfaces of the cap nuts and the bolting
ange on the head and vessel shell. A friction coefcient of 0.2 is specied in the contact between the
cap nuts and the bolting anges. The analysis assumes that the contact between the closure head and the
vessel shell is frictionless. The augmented Lagrange method is chosen to enforce the contact constraints.
Analysis steps
The structural analysis is performed by using two static steps. The bolt force and internal pressure are
both applied in the rst step, a predened temperature eld is also specied using the results obtained
in the steady-state heat transfer step. The temperature obtained from the transient heat transfer step is
specied in the last loading step, and the structure expands with the change of temperature.
Output requests
Default eld output requests are specied in the rst step. In the second step, eld output requests are
made at specied time points to match the results of the structural analysis to those of the thermal analysis
at the exact step times.
Convergence
The convergence of error indicators ENDENERI and MISESERI during a three-iteration adaptivity
process is presented in Table 5.1.62.
Run procedure
The model for the static analysis is generated using Abaqus/CAE to import the geometry, create the
structural loading, mesh the assembly, create the remeshing rules, and run the adaptivity process. Python
scripts are provided to build the model and to submit the adaptivity process. The scripts can be run
interactively or from the command line.
To create the structural model, select FileRun Script from the Abaqus/CAE main menu and
select adaptReactorVesselSTR_model.py.
5.1.66

Printed on:
When you are ready to run the adaptivity process, select FileRun Script and select
adaptReactorVesselSTR_job.py. After the Abaqus Scripting Interface scripts have created
the model and run the adaptivity process, you can use Abaqus/CAE to view the model and to explore
variations of the example.
The magnitude of the Mises stress (MISES) along a particular path near the head to vessel interface is
shown in Figure 5.1.65. This gure shows the spatial variation of the Mises stress as the mesh is rened
through the three remesh iterations. Figure 5.1.66 shows the original mesh and the nal mesh resulting
from the adaptive remeshing process. The rened mesh shows how Abaqus/CAE reacted to the higher
stress gradients near the nut-to-bolted ange interfaces and the vessel-to-head interface.
The thermal and structural cases presented in this example are complementary; the structural case
depends on the thermal case. You can compare how adaptive remeshing renes the mesh in each
case. As seen, in Figure 5.1.64 and Table 5.1.61 for the heat transfer case and Figure 5.1.66 and
Table 5.1.62 for the structural case, the mesh renement is signicantly different.
Files
To create the models and to run the adaptivity processes, you can use the Python scripts listed below.
Heat transfer analysis
adaptReactorVesselHT_model.py
adaptReactorVesselHT_job.py
Script to create the model.

Script to analyze the model.
Structural analysis
adaptReactorVesselSTR_model.py
adaptReactorVesselSTR_job.py
Script to create the model.

Script to analyze the model.
References
Adaptive remeshing: overview, Section 12.3.1 of the Abaqus Analysis Users Manual
Abaqus/CAE Users Manual
Understanding adaptive remeshing, Section 17.12 of the Abaqus/CAE Users Manual
5.1.67

Printed on:
Other
Naval Reactors Branch, Division of Reactor Development, United States Atomic Energy
Commission, The Shippingport Pressurized Water Reactor, Reading, Massachusetts: Addison
Wesley Publishing Company, 1958.
Table 5.1.61 The convergence of the thermal error indicator HFLERI

during an adaptivity process with three iterations.
Remeshing
Region
Error Indicator
Result (%)
Element Count
Stud Bolt
11.7
7.0
4.1
1080
1386
2112
Closure Head
4.9
2.5
1.7
2210
4846
9226
Cap Nuts
2.8
1.5
1.4
1500
4210
5317
Vessel Shell
5.4
2.3
1.5
2443
5124
11036
Table 5.1.62 The convergence of the stress-based error indicator

ENDENERI during an adaptivity process with three iterations.
Remeshing
Region
Error Indicator
Result (%)
Element Count
Stud Bolt
1.8
1.7
1.5
1334
1737
2211
Closure Head
8.7
4.7
3.2
2210
5120
13320
Cap Nuts
6.0
5.7
4.8
1500
3315
7534
Vessel Shell
6.7
3.6
2.7
2438
5813
15290
5.1.68

Printed on:
Control rod drive

mechanism penetrations
Closure hardware
Outlet nozzles
Inlet nozzles
Figure 5.1.61
Reactor vessel assembly.
5.1.69

Printed on:
Figure 5.1.62
9 sector model.
5.1.610

Printed on:
HFLiter1
HFLiter2
HFLiter3
HFL Magnitude
30.00
25.00
20.00
15.00
0.00 4.00 8.00 12.00 16.00 20.00
True distance along path Path 1
Figure 5.1.63
Heat ux along a path near the vessel-to-head interface.
Figure 5.1.64
Original and rened mesh for the heat transfer analysis.
5.1.611

Printed on:
Misesiter1
Misesiter2
Misesiter3
[x10 3 ]
5.50
5.00
4.50
S Mises
4.00
3.50
3.00
2.50
2.00
1.50
1.00
0.00 2.00 4.00 6.00 8.00 10.00
True distance along path Path 1
Figure 5.1.65
Mises stress along a path near the vessel-to-head interface.
Figure 5.1.66
Original and rened mesh for the structural analysis.
5.1.612

Printed on:
FLUID DYNAMICS AND FLUID-STRUCTURE INTERACTION
6.
Fluid Dynamics and Fluid-Structure Interaction
Fluid dynamics and uid-structure interaction, Section 6.1

Printed on:
FLUID DYNAMICS AND FLUID-STRUCTURE INTERACTION
6.1
Fluid dynamics and fluid-structure interaction
Conjugate heat transfer analysis of a component-mounted electronic circuit board, Section 6.1.1
6.11

Printed on:
CIRCUIT BOARD CONJUGATE HEAT TRANSFER
6.1.1
CONJUGATE HEAT TRANSFER ANALYSIS OF A COMPONENT-MOUNTED

ELECTRONIC CIRCUIT BOARD
Abaqus/CFD
Abaqus/CAE
Objectives
This example demonstrates the following Abaqus features and techniques for Abaqus/CFD to
Abaqus/Standard co-simulation:
using Abaqus/CFD to Abaqus/Standard co-simulation to analyze a conjugate heat transfer problem,

solving both the laminar ow eld and buoyancy-driven natural convection heat transfer occurring
external to a structure in Abaqus/CFD, and
solving for the heat transfer within the structure using Abaqus/Standard.
This example examines the transient conjugate heat transfer between a single printed circuit board
(PCB)mounted electronic component and the ambient air. The component is subjected to a
passive power dissipation that results in the transfer of heat both within the component and the
PCB due to conduction. Furthermore, the heated surface of the component and the PCB induces a
temperature-dependent density differential in the surrounding air, thereby setting up a buoyancy-driven
natural convection process external to the surface. Heat is thus transferred from the component
and PCB surfaces to the ambient air through this convection process. Understanding the resulting
conduction-convection conjugate heat transfer phenomenon results in more accurate damage estimation
and life prediction for electronic components.
The details of the model are largely derived from Eveloy and Rodgers (2005).
Geometry
Nominal component/PCB geometry dimensions are considered. The PCB dimensions are 7.8 11.6
0.16 cm. The mounted electronic component consists of an encapsulant of dimension 3 3 0.7 cm
that encapsulates a heat slug of dimension 1.8 1.8 0.3 cm mounted atop a die of dimension 0.75 0.75
0.2 cm. The assembled PCB-component package is shown in Figure 6.1.11. The cross-section view
of the assembled package is shown in Figure 6.1.12. The size of the computational domain that encloses
the electronic package is taken to be 27.8 20 12.56 cm for the Abaqus/CFD ow computations.
Materials
The PCB material has a thermal conductivity of 19.25 W/m/K, a density of 8950 kg/m3 , a specic heat of
1300 J/kg/K, and a thermal expansion of 1.6 105 /K. The heat slug material has a thermal conductivity
of 398 W/m/K, a density of 8940 kg/m3 , a specic heat of 385 J/kg/K, and a thermal expansion of
3.3 106 /K. The die material has a thermal conductivity of 130.1 W/m/K, a density of 2330 kg/m3 ,
a specic heat of 712 J/kg/K, and a thermal expansion of 3.3 106 /K. The encapsulant material has
6.1.11

Printed on:
a thermal conductivity of 0.63 W/m/K, a density of 1820 kg/m3 , a specic heat of 882 J/kg/K, and a
thermal expansion of 1.9 105 /K. For the external uid domain, properties of air, namely a density of
1.127 kg/m3 , a thermal conductivity of 2.71 102 W/m/K, a specic heat of 1006.4 J/kg/K, a thermal
expansion of 3.43 103 /K, and a viscosity of 1.983 105 kg/m/s, are assumed.
Initial conditions
The initial temperature of the assembled electronic package and the external uid (ambient air) is set to
293 K. In addition, the uid is initially assumed to be quiescent and, hence, its velocity is set to zero
everywhere.
The electronic component of the assembled package is subjected to a passive power dissipation
corresponding to a specied body heat ux value of 5 107 W/m3 . The bottom surface of the external
computational domain is assumed to be a rigid oor, and an adiabatic wall condition is assumed
there. The electronic package is rmly attached to the bottom of the computational domain, as shown
in Figure 6.1.13. The computational domain boundaries for the external ow and heat transfer
calculations are positioned at sufcient distances from the assembled electronic package such that the
effects of the boundaries on the results are negligible. A no-slip, no-penetration ow wall boundary
condition is enforced at the bottom surface. At the top, an outlet boundary condition is specied with
the uid pressure set to zero. For all other boundaries, free-stream conditions are applied for both the
velocity and temperature.
Gravity loading is applied along the negative y-axis (see Figure 6.1.13).
Interactions
The component-mounted circuit board is subjected to volumetric power dissipation, which generates
heat within the structure. Heat transfer to the surface of the structure takes place via conduction. The
heated surface then generates a buoyancy-driven natural convection in the surrounding ambient air.
The Abaqus/CFD to Abaqus/Standard co-simulation approach is used to solve the conjugate heat transfer
problem. The Abaqus/Standard model of the co-simulation consists of the component-mounted PCB
electronic package (see Figure 6.1.11). The Abaqus/CFD model of the co-simulation consists of the
external region bounding the electronic package (see Figure 6.1.13). Co-simulation regions across
which data will be exchanged during the co-simulation analysis are identied on each model at the
location of the electronic package surface.
Analysis types
The Abaqus/Standard model includes a transient heat transfer step. The Abaqus/CFD model includes
an incompressible laminar ow analysis coupled with the energy equation that governs the temperature
distribution.
6.1.12

Printed on:
Mesh design
The Abaqus/Standard model is meshed with linear hexahedral (DC3D8) elements. The Abaqus/CFD
model is meshed with linear hexahedral (FC3D8) uid elements.
The heat generated within the electronic component gets distributed to its surface and also to the surface
of the PCB substrate by conduction. The heated air near the component surface becomes less dense
and rises upward due to buoyancy. The Grashof number (a parameter describing the ratio of buoyancy
to viscous forces) for this problem, based on the board length, was estimated to be of the order 106 as
reported by Eveloy and Rodgers (2005). Because of incompressibility, the upward rising air entrains
cold air from the surroundings, thereby setting up a natural convection ow. The upward rising hot air
and entrained cold air from the sides are illustrated in Figure 6.1.14, where the velocity vectors (V) are
plotted on the center cross-sectional slice of the domain in the yz plane. In Figure 6.1.14 entrainment
from the out-of-plane direction appears as small dots. The contours of temperature distribution (TEMP)
on the same cross-sectional slice are shown in Figure 6.1.15, where the hot electronic component surface
and the upward rising thermal plume can be clearly seen. The temperature distribution (NT11) along a
particular path in the spanwise direction (x-axis) on the surface of the component-mounted electronic
package is shown in Figure 6.1.16. Heat is seen to spread from the component surface (the area of
near-constant heat distribution) to the surfaces of the PCB substrate on either side. The temperature
distribution (NT11) along a particular path in the streamwise direction (y-axis) on the surface of the
electronic package shown in Figure 6.1.17 illustrates the same phenomenon. These distributions are
qualitatively similar to the experimentally measured temperature proles given in Eveloy and Rodgers
(2005).
Files
circuit_board_cht.py
cfd_mesh.inp
std_heat_transfer.inp
Script to generate the model in Abaqus/CAE

using the orphan meshes from cfd_mesh.inp and
std_heat_transfer.inp.
Orphan mesh for the Abaqus/CFD domain.
Orphan mesh for the Abaqus/Standard domain.
References
Incompressible uid dynamic analysis, Section 6.6.2 of the Abaqus Analysis Users Manual
Abaqus/CFD to Abaqus/Standard or to Abaqus/Explicit co-simulation, Section 14.1.5 of the
6.1.13

Printed on:
Other
Eveloy, V., and P. Rodgers, Prediction of Electronic Component-Board Transient Conjugate

Heat Transfer, IEEE Transactions on Components and Packaging Technologies, vol. 28, no. 4,
pp. 817829, 2005.
PCB
Component
Figure 6.1.11
Component-mounted electronic circuit board.
Heat slug
Die
PCB substrate
Encapsulant
Figure 6.1.12
Cross-sectional view of the electronic package indicating the different

materials forming the component and circuit board.
6.1.14

Printed on:
Figure 6.1.13 Electronic package anchored to the bottom

of the computational domain (enclosing box).
6.1.15

Printed on:
V, Resultant
+1.179e+01
+1.081e+01
+9.823e+00
+8.841e+00
+7.859e+00
+6.876e+00
+5.894e+00
+4.912e+00
+3.929e+00
+2.947e+00
+1.965e+00
+9.823e01
+0.000e+00
Y
X
Figure 6.1.14 Plot of velocity vectors induced due to natural

convection on a cross-sectional slice.
6.1.16

Printed on:
TEMP
+3.106e+02
+3.091e+02
+3.076e+02
+3.062e+02
+3.047e+02
+3.033e+02
+3.018e+02
+3.003e+02
+2.989e+02
+2.974e+02
+2.959e+02
+2.945e+02
+2.930e+02
Y
X
Figure 6.1.15
Plot of temperature contours on a cross-sectional slice.
6.1.17

Printed on:
310.
Temperature
Temperature
305.
300.
295.
0.0
2.0
4.0
6.0
8.0
True spanwise distance along path
Figure 6.1.16 Temperature along a path in the spanwise direction

(x-axis) on the surface of the electronic package.
310.
Temperature
Temperature
305.
300.
295.
0.
4.
8.
12.
True streamwise distance along path
Figure 6.1.17 Temperature along a path in the streamwise direction

(y-axis) on the surface of the electronic package.
6.1.18

Printed on:
ELECTRICAL ANALYSES
7.
Electrical Analyses
Piezoelectric analyses, Section 7.1

Joule heating analyses, Section 7.2

Printed on:
PIEZOELECTRIC ANALYSES
7.1
Piezoelectric analyses
Eigenvalue analysis of a piezoelectric transducer, Section 7.1.1

Transient dynamic nonlinear response of a piezoelectric transducer, Section 7.1.2
7.11

Printed on:
PIEZOELECTRIC TRANSDUCER
7.1.1
EIGENVALUE ANALYSIS OF A PIEZOELECTRIC TRANSDUCER
This problem performs an eigenspectrum analysis of a cylindrical transducer consisting of a piezoelectric

material with brass end caps. Various elements are used in the analysis. The elements range from
axisymmetric elements to three-dimensional elements, using both lower- and higher-order elements. The
basis of the piezoelectric capability in Abaqus is described in Piezoelectric analysis, Section 2.10.1 of the
Abaqus Theory Manual.
This problem is identical to the one discussed in a report by Mercer et al. (1987). The structure is shown
in Figure 7.1.11 and consists of a piezoelectric material PZT4 with brass end caps. The piezoelectric
material is electroded on both the inner and outer surfaces.
The properties for PZT4 in a cylindrical system are:
Elasticity Matrix:
GPa
Piezoelectric Coupling Matrix (Stress Coefcients):
coulomb/m
Dielectric Matrix:
farad/m
The 1-direction is radial, the 2-direction is axial, and the 3-direction is tangential. From these matrices
it is seen that the poling direction is radially outwards from the axis of symmetry. (The order of the
stresses in Abaqus may differ from those typically used in electrical applications. Abaqus uses the
7.1.11

Printed on:
standard mechanical convention, where the stress components are ordered as

.
See Piezoelectric behavior, Section 23.6.2 of the Abaqus Analysis Users Manual.)
The brass is elastic and isotropic with a Youngs modulus of 104 GPa and a Poissons ratio of 0.37.
Models
The transducer is modeled with a variety of elements. It is modeled as an axisymmetric structure utilizing
both the planar, axisymmetric elements and the three-dimensional elements. For the axisymmetric
elements, ve meshes employing 4-node, 6-node, and 8-node elements are used in the nite element
discretization. The rst two meshes use 4-node elements with two levels of renement, the third mesh
uses 6-node elements, and the last two meshes use the 8-node elements with two levels of renement.
Lumped mass matrices are used for the lower-order elements. Consistent mass matrices are used in the
higher-order elements. The meshes used for the 4-node and 6-node axisymmetric elements are shown
in Figure 7.1.12. The meshes used for the 8-node axisymmetric elements are shown in Figure 7.1.13.
The three-dimensional model uses a slice of the structure and applies axisymmetric boundary
conditions; 8-node and 20-node brick elements are used. The discretization used for each model
is shown in Figure 7.1.13. These models use the *ORIENTATION option to maintain the proper
denitions of the material properties. Also, in order to prescribe the axisymmetric boundary conditions,
the nodal degrees of freedom are transformed into a cylindrical coordinate system.
All the models are considered to be open-circuited. The potentials on the inside surface are
restrained to zero. The frequencies correspond to those for anti-resonance.
The solutions obtained with the various Abaqus models are shown in Table 7.1.11. Even for these coarse
models, the results are quite close to the experimental results. In addition, the results from Abaqus for
the lower-order axisymmetric elements with lumped mass and the higher-order axisymmetric elements
with consistent mass matrices in the computation of both the resonant and antiresonant frequencies match
well with the numerical results reported in Mercer et al. The rst four mode shapes for the more rened
model with CAX8RE elements are shown in Figure 7.1.14.
Similar analyses have been performed considering the problem to be closed-circuited to obtain the
resonant frequencies. For this situation, the potentials on both the inner and outer surfaces are set to zero.
The results also compare well with those given in Mercer et al.
Input files
eigenpiezotrans_cax4e_coarse.inp
eigenpiezotrans_cax4e_ne.inp
eigenpiezotrans_cax6e.inp
eigenpiezotrans_cax8re_coarse.inp
eigenpiezotrans_cax8re_ne.inp
eigenpiezotrans_c3d8e.inp
eigenpiezotrans_c3d8e.f
Coarse mesh with 4-node axisymmetric elements.

Rened mesh with 4-node axisymmetric elements.
Mesh with 6-node axisymmetric elements.
Coarse mesh with 8-node axisymmetric elements.
Rened mesh with 8-node axisymmetric elements.
Mesh with 8-node three-dimensional elements.
User subroutine ORIENT used in
eigenpiezotrans_c3d8e.inp.
7.1.12

Printed on:
eigenpiezotrans_c3d20e.inp
eigenpiezotrans_c3d20e.f
Mesh with 20-node three-dimensional elements.

User subroutine ORIENT used in
eigenpiezotrans_c3d20e.f.
Data that test the use of *ELEMENT MATRIX OUTPUT
with piezoelectric elements.
Data for a job that reads in the matrices output in
eigenpiezotrans_elmatrixout.inp and performs an
eigenvalue analysis.
eigenpiezotrans_elmatrixout.inp
eigenpiezotrans_usr_element.inp
Reference
Mercer, C. D., B. D. Reddy, and R. A. Eve, Finite Element Method for Piezoelectric Media,
University of Cape Town/CSIR Applied Mechanics Research Unit Technical Report, no. 92, April
1987.
Table 7.1.11
Piezoelectric transducer eigenvalue estimates.
Model
Frequencies (kHz) for mode number
Type
# of Elements
CAX4E
13
14.1
39.1
56.2
66.1
79.3
CAX4E
320
18.6
40.3
57.8
64.2
88.1
CAX6E
10
20.0
43.2
63.2
70.4
98.8
CAX8RE
19.6
42.8
61.0
66.9
96.3
CAX8RE
80
18.6
40.3
57.6
64.2
87.6
C3D8E
16
19.8
41.8
62.0
68.7
95.2
16
19.7
42.9
60.4
66.5
91.7
18.6
35.4
54.2
63.3
88.8
C3D20E
Experimental
(1)
(1): Experimental results obtained from Mercer et al. (1987).
7.1.13

Printed on:
Axis of symmetry
Top brass cap
3 mm
Piezoelectric
Ceramic PZT4
12.5 mm
Plane of
symmetry
Electroded surfaces
Bottom brass cap
11 mm
12.5 mm
Figure 7.1.11
Piezoelectric transducer.
7.1.14

Printed on:
Figure 7.1.12
Meshes used with 4-node and 6-node axisymmetric elements.
7.1.15

Printed on:
Figure 7.1.13 Meshes used with 8-node axisymmetric and

8-node and 20-node three-dimensional elements.
7.1.16

Printed on:
Mode 1: 18.6 kHz
Mode 2: 40.3 kHz
Mode 3: 57.6 kHz
Mode 4: 64.2 kHz
Figure 7.1.14
Mode shapes for 8-node axisymmetric elements.
7.1.17

Printed on:
TRANSIENT DYNAMIC PIEZOELECTRIC TRANSDUCER
7.1.2
TRANSIENT DYNAMIC NONLINEAR RESPONSE OF A PIEZOELECTRIC

TRANSDUCER
This example demonstrates the Abaqus capability for predicting the nonlinear transient dynamic behavior
of structural systems that incorporate piezoelectric components. It utilizes a piezoelectric bending-type
transducer that is idealized as a simple layered beam structure. The fabrication of the transducer is modeled,
and the eigenfrequencies of the preloaded structure are extracted. Finally, the dynamic response due to
a transient electrical potential pulse is monitored. Piezoelectric transducers are commonly used in the
following application areas:
Ultrasonic imaging systems

Ultrasonic cleaning systems
Ultrasonic welding/bonding systems
Audio systems
Acoustic transducers
Active vibration control systems
The transducer is shown in Figure 7.1.21. It has a composite beam that is 0.0025 m wide and 0.18 m
long. The initially straight beam has a 0.0005 m thick insulating core; 0.000125 m thick PZT5H
piezoelectric strips are bonded to the top and the bottom faces of the core material. The piezoelectric
strips are only 0.060 m long and are centered at the beam mid-span. The piezoelectric strips are polarized
along the beams in the through-thickness direction. The properties for the PZT5H material are as
follows:
Elastic properties:
Engineering
constants
60.61 GPa
48.31 GPa
60.61 GPa
0.512
0.289
0.408
23.0 GPa
7.1.21

Printed on:
Engineering
constants
23.5 GPa
23.0 GPa
Piezoelectric coupling matrix (strain coefcients):
m/volt
Dielectric matrix:
farad/meter
The local 1-direction is in the beam longitudinal direction, and the local 2- and 3-directions are in the
beam cross-section. From these matrices it can be seen that the poling direction is in the local 2-direction
of the piezoelectric strips.
The core material is elastic and isotropic with a Youngs modulus of 6 GPa, a density of 1500 kg/m3 ,
and a Poissons ratio of 0.35.
Model
The beam core material is modeled with 46 C3D20 elements, and the piezoelectric strips are modeled
using 16 C3D20E elements each. A perfect bond between the core and the piezoelectric materials is
dened via the *TIE option, for which the piezoelectric surfaces are retained as the master surfaces. The
electrical potentials for the top and the bottom surfaces of each piezoelectric strip are coupled to the
electrical potentials of the master nodes assigned to each surface using the *EQUATION option. The
electrical potentials and the reaction charges can be monitored at these master nodes. The piezoelectric
surfaces bonded to the core material are assigned a zero electrical potential throughout the analysis.
The rst 8 steps are used to represent the fabrication procedure and to investigate the behavior of
the fabricated transducer. These include linear perturbation steps to investigate the eigenmodes of the
fabricated transducer at various stages during its fabrication. Steps 9 through 11 represent a general
nonlinear transient analysis of the transducer. A square wave electrical potential pulse of 200 volts is
applied using the *BOUNDARY option with a STEP amplitude denition. Immediately after the pulse,
a closed circuit condition (potential gradients are prescribed) is maintained and the reaction charges are
monitored. Subsequently, the circuit is opened (the potential gradient is not prescribed; the potential is
an active degree of freedom that is determined as part of the solution). In the case of an open circuit
condition the resulting voltage can be used to measure the transducers open circuit free vibration. The
analysis steps are as follows:
1. Static shape fabrication: deformation induced via an applied potential of 1000 volts.
7.1.22

Printed on:
2. Static shape fabrication: x support ends, and reduce the applied potential to 0 volts.
3. Closed circuit modal analysis about the base state obtained at the end of Step 2.
4. Static shape fabrication: apply open circuit conditions.
5. Open circuit modal analysis about the base state obtained at the end of Step 4.
6. Static test at an operational load of 200 volts.
7. Closed circuit modal analysis about the base state obtained at the end of Step 6.
8. Reset to zero voltage condition.
9. Transient dynamic response: apply voltage pulse of 200 volts for 0.00265 seconds.
10. Transient dynamic response: free vibration under closed circuit condition with 0 volts.
11. Transient dynamic response: free vibration under open circuit condition.
In Figure 7.1.22 the deformed and superimposed undeformed shapes are shown after applying
1000 volts at both piezoelectric strips. In Figure 7.1.23 the deformed shape is shown at the end of
Step 2, where both ends are xed and an applied potential of 0 volts is prescribed. Subsequently, the
eigenfrequencies are extracted about this preloaded state. In Figure 7.1.24 the third eigenmode with a
frequency of 150.9 cycles/sec is shown for this closed circuit condition. Figure 7.1.25 shows the third
mode shape with a frequency of 154.4 cycles/sec under open circuit conditions. The third mode shape
for the closed circuit condition with a prescribed voltage of 200 volts is again almost identical to the third
mode shape shown in Figure 7.1.24. However, the eigenfrequency has changed to 181.7 cycles/sec.
In Steps 9 and 10 a closed circuit condition is prescribed for the piezoelectric strips, and
subsequently in Step 11 the voltage boundary conditions are removed resulting in open circuit
conditions. Figure 7.1.26 shows the time history of the potential under closed and open circuit
conditions. Typically, the transducer acts as a driver in a closed circuit condition as the potential
gradient is prescribed, thereby driving the structure. On the other hand, the transducer acts as a
receiver in an open circuit condition as the voltage output can be used to measure the mechanical
response. In Figure 7.1.27 the reaction charge at the top piezoelectric strip is shown under the
above conditions. In a closed circuit condition the reaction charge changes in time, while in an open
circuit condition the reaction charge is equal to zero. The displacement of the transducers center is
demonstrated in Figure 7.1.28.
The results are presented to illustrate the general capabilities in Abaqus for predicting the transient
response of piezoelectric structures. If this system represented part of an actual ultrasound system,
additional output related to the design/analysis objective would be created and analyzed.
Input files
dynamictransducer.inp
dynamictransducer_mesh.inp
Transient dynamic nonlinear response of a piezoelectric

transducer.
Assembly denition.
7.1.23

Printed on:
core material
piezoelectric strips
0.125 mm
0.5 mm
60 mm
180 mm
Figure 7.1.21
Geometry of the piezoelectric transducer.
Y
Z X
Figure 7.1.22 Deformed shape with superimposed

undeformed shape at the end of Step 1.
7.1.24

Printed on:
Y
Z X
Figure 7.1.23 Deformed shape with superimposed

undeformed shape at the end of Step 2.
Y
Z X
Figure 7.1.24 Mode shape 3 with superimposed shape

of the base state after Step 2.
7.1.25

Printed on:
Y
Z X
Figure 7.1.25 Mode shape 3 with superimposed shape

of the base state after Step 4.
Figure 7.1.26
Transient response of potential at top piezoelectric strip.
7.1.26

Printed on:
Reaction charge at top
[x1.E6]
8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
1.0
5.00
5.01
5.02
5.03
5.04
5.05
Time
Figure 7.1.27 Transient dynamic response of the reaction
charge at top piezoelectric strip.
Figure 7.1.28
Transient response of center displacement.
7.1.27

Printed on:
JOULE HEATING ANALYSES
7.2
Joule heating analyses
Thermal-electrical modeling of an automotive fuse, Section 7.2.1
7.21

Printed on:
THERMAL-ELECTRICAL ANALYSIS
7.2.1
THERMAL-ELECTRICAL MODELING OF AN AUTOMOTIVE FUSE
Joule heating arises when the energy dissipated by electrical current owing through a conductor is converted
into thermal energy. Abaqus provides a fully coupled thermal-electrical procedure for analyzing this type of
problem.
An overview of the capability is provided in Coupled thermal-electrical analysis, Section 6.7.2 of the
Abaqus Analysis Users Manual. This example illustrates the use of the capability to model the heating of
an automotive electrical fuse due to a steady 30 A electrical current. Fuses are the primary circuit protection
devices in automobiles. They are available in a range of different current ratings and are designed so that
when the operating current exceeds the design current for a period of time, heating due to electrical conduction
causes the metal conductor to melt andhencethe circuit to disconnect.
A description of the original problem, as well as experimental measurements, can be found in Wang
and Hilali (1995). The experimental data and some of the material properties were rened subsequent to
this publication. These properties are used here, and the nite element results are compared with the rened
measurements (Hilali, July 1995).
Problem description
An automotive electrical fuse consists of a metal conductor, such as zinc, embedded within a transparent
plastic housing. The plastic housing, which only protects and supports the thin conductor, is not
represented in the nite element model. Figure 7.2.11 shows front and top sections of the geometry
of the conductor. It consists of two 0.76 mm thick blades, with an S-shaped fuse element supported
between the blades. The blades t tightly into standard electrical terminals that are built into the circuit
and provide the connection between the electrical circuit and the fuse element. The fuse element is
usually much thinner than the fuse blades (in this case 0.28 mm thick) and is designed to melt when the
operating current exceeds the design current for a period of time. The fuse blades are 8 mm wide and
30.4 mm long. The fuse element is approximately 3.6 mm wide.
The model is discretized (see Figure 7.2.11) with 8-node rst-order brick elements (element type
DC3D8E), using one element through the thickness. Two 6-node triangular prism elements (element
type DC3D6E) are used to ll regions where the geometry precludes the use of brick elements. For
comparison rened mesh input les are also included.
The electrical conductivity of zinc varies linearly between 16.75 103
mm at 20C and
12.92 103
mm at 100C. The thermal conductivity varies linearly between 0.1120 W/mmC
at 20C and 0.1103 W/mmC at 100C. The density is 7.14 106 kg/mm3 , and the specic heat is
388.9 J/kgC. The *JOULE HEAT FRACTION option is used to specify the amount of electrical energy
that is converted into thermal energy. We assume that all electrical energy is converted into thermal
energy.
The analysis is done in two steps. In the rst step heating of the conductor due to current ow is
considered. Once steady-state conditions are reached, the current is switched off and the fuse is allowed
to cooldown to the ambient temperature in a second step. During the rst part of the analysis, the coupled
7.2.11

Printed on:
thermal-electrical equations are solved for both temperature and electrical potential at the nodes using
the *COUPLED THERMAL-ELECTRICAL procedure. In the subsequent cooldown period, since there
is no longer any electric current in the fuse, an uncoupled *HEAT TRANSFER analysis (Uncoupled
heat transfer analysis, Section 6.5.2 of the Abaqus Analysis Users Manual) is performed. Input les
illustrating both steady-state and transient analyses are provided. Steady-state analysis is obtained by
specifying the STEADY STATE parameter on the *COUPLED THERMAL-ELECTRICAL procedure.
No data lines are required. Transient analysis is available by omitting the STEADY STATE parameter.
In this analysis the DELTMX parameter is set to 20C so that automatic time incrementation is used.
The END parameter is set to SS so that the analysis terminates when steady-state conditions are reached.
Steady state is dened here as the point at which the temperature rate change is less than 0.1C/s. This
condition is dened on the data line following the *COUPLED THERMAL-ELECTRICAL option. We
specify a total analysis time of 100 s, with an initial time step size of 0.1 s.
The electrical loading is a steady 30 A current. This is applied as a concentrated current on each
of the nodes on the bottom edge of the left-hand-side terminal. The *CECURRENT option is used for
this purpose. The *SECTION FILE option is used to output the total current and the total heat ux in a
section dened through the fuse element. The electrical potential (degree of freedom 9) is constrained
at the bottom edge of the right-hand-side blade by using a *BOUNDARY option (Boundary conditions
in Abaqus/Standard and Abaqus/Explicit, Section 30.3.1 of the Abaqus Analysis Users Manual). This
option is also used to keep the bottom edges of the fuse blades at sink temperatures (degree of freedom 11)
of 29.4C and 30.2C, respectively.
It is assumed that the exposed metal surfaces lose heat through convection to an ambient temperature
of
23.3C. Heat loss from the thin edges is ignored. The lm coefcient varies with temperature
according to the empirical relation
where is surface temperature (C); h is the lm coefcient (W/mm2 C); and

is a constant
that depends on the surface geometry
= 4.747 106 W/mm2 for the blade surfaces, and
= 5.756 106 W/mm2 for the fuse element surfaces. This dependence is entered as a table of
lm property values using the *FILM PROPERTY option and is referred to on the *FILM option (see
Thermal loads, Section 30.4.4 of the Abaqus Analysis Users Manual).
Figure 7.2.12 shows a contour plot of the magnitude of the electrical current density vector at steadystate conditions. Since the dissipated electrical energyand, hence, the thermal energyis a function
of current density, this gure represents contours of the heat generated. The gure indicates that most of
the heat is generated near the inside curves of the S-shaped fuse element and near the center hole. The
dissipated energy in the fuse blades is negligible compared to that in the fuse element.
Figure 7.2.13 shows a contour plot of the temperature distribution at the end of the rst analysis
step. The maximum temperature is reached near the center of the S-shaped fuse element. This area is
expected to fail rst when the operating current exceeds the design current. Figure 7.2.14 compares the
temperatures at the measuring positions (dened in Figure 7.2.11) with the experimental measurements
7.2.12

Printed on:
(Hilali, July 1995). While the results show some discrepancies between the experiment and analysis,
it is clear that the analysis is sufciently representative to provide a useful basis for studying such
systems. Figure 7.2.15 shows the variation of temperature at measuring position 6 during the heating
and subsequent cooldown periods.
The results discussed above are for the coarse mesh model. The rened mesh models yield slightly
different results from the coarse ones. The maximum difference in the magnitude of the electrical current
density vector for the steady-state analysis is approximately 11.4%.
Acknowledgments
Mr. Hilali and Dr. Wang of Delphi Packard Electric Systems supplied the geometry of the fuse, the
material properties, and the experimental results. Delphi Packard assumes no responsibility for the
accuracy of the analysis method or data contained in the analysis.
Input files
thermelectautofuse_steadystate.inp
thermelectautofuse_transient.inp
thermelectautofuse_transient_po.inp
thermelectautofuse_node.inp
thermelectautofuse_element.inp
thermelectautofuse_controls.inp
teaf-steadystate-rened.inp
teaf-transient-rened.inp
Steady-state analysis.
Transient analysis.
*POST OUTPUT analysis.
Nodal coordinates for the model.
Element denitions.
Identical to thermelectautofuse_steadystate.inp, except
that it uses the *CONTROLS option for control of
convergence criteria.
Rened mesh model for the steady-state analysis.
Rened mesh model for the transient analysis.
References
Hilali, S. Y., Private communication, July 1995.
Wang, B. -J., and S. Y. Hilali, Electrical-Thermal Modeling Using ABAQUS, 1995 ABAQUS
Users Conference, Paris, May 1995, pp. 771785.
Hilali, S. Y., and B. -J. Wang, ABAQUS Thermal Modeling for Electrical Assemblies, 1995
ABAQUS Users Conference, Paris, May 1995, pp. 441457.
7.2.13

Printed on:
9
7
10
Fuse Element
11
Blades
1
12
Blade
Figure 7.2.11
Figure 7.2.12
Fuse Element
Geometry and nite element discretization.
Contours of the magnitude of the current density vector (A/mm2 ).
7.2.14

Printed on:
Blade
Figure 7.2.13
Contours of temperature eld (C).
Analysis
Experiment
XMIN
XMAX
YMIN
YMAX
1.000E+00
1.200E+01
2.900E+01
1.033E+02
Figure 7.2.14
Temperature (C) at measuring positions.
7.2.15

Printed on:
NT11
XMIN
XMAX
YMIN
YMAX
0.000E+00
1.595E+02
0.000E+00
1.029E+02
Figure 7.2.15
Variation of temperature (C) at measuring position 6 with time (s).
7.2.16

Printed on:
MASS DIFFUSION ANALYSES
8.
Mass Diffusion Analyses
Mass diffusion analyses, Section 8.1

Printed on:
MASS DIFFUSION ANALYSES
8.1
Mass diffusion analyses
Hydrogen diffusion in a vessel wall section, Section 8.1.1

Diffusion toward an elastic crack tip, Section 8.1.2
8.11

Printed on:
HYDROGEN DIFFUSION
8.1.1
HYDROGEN DIFFUSION IN A VESSEL WALL SECTION
This one-dimensional problem provides a simple verication of the mass diffusion capability in Abaqus. The
uncoupled mass diffusion formulation used in Abaqus is described in Mass diffusion analysis, Section 6.9.1
of the Abaqus Analysis Users Manual, and Mass diffusion analysis, Section 2.13.1 of the Abaqus Theory
Manual.
The physical problem considered here is that of a pressure vessel shell wall fabricated from 2 1/4 Cr1 Mo
steel alloy base metal with an internal weld overlay of Type 347 stainless steel. These vessels are typically used
at high temperatures and under high pressure conditions. Under such service conditions hydrogen dissolves
into the alloys (Fujii et al., 1982) and during cooldown may cause disbonding of the weld overlay from the
base metal and, possibly, crack initiation and growth in the base metal due to hydrogen embrittlement. In this
example we are concerned with the hydrogen diffusion aspect of the problem.
Problem description
The problem is shown in Figure 8.1.11 and consists of a section of the vessel wall made up of a 200-mm
thick base metal and a 5-mm thick weld metal. The problem is one-dimensional, the only gradient being
through the thickness of the wall. The purpose of the analysis is to predict the evolution of hydrogen
concentration through the wall thickness during cooling caused by a shutdown.
Geometry and model
Since the problem is one-dimensional, we use a plane mesh with only one element in the y-direction (see
Figure 8.1.12). The mesh is graded, with more elements near the interface between the two materials
because we expect very high concentration gradients in this vicinity.
The material properties of the two metals given by Fujii et al. (1982) are strongly dependent on
temperature and can be written as follows.
Solubility in weld metal:
Diffusivity in weld metal:
Solubility in base metal:
8.1.11

Printed on:
HYDROGEN DIFFUSION
Diffusivity in base metal:
where is temperature in kelvins. These temperature-dependent properties are entered in Abaqus in

tabulated form, as shown in the input listings.
The wall is initially at a uniform temperature of 727.5 K (454.4C), and during the shutdown
schedule it cools down to 298.15 K (25.0C) at a constant rate over a period of 21.5 hours.
The boundary conditions are as follows. Under the initial steady-state conditions the exterior of the
weld metal has a hydrogen concentration of 35.85 ppm, which corresponds to a normalized concentration
of 0.1225 N1/2 mm1 . Normalized concentration is used as the primary solution variable (continuous over
the discretized domain) and is given as the concentration divided by the solubility. The exterior of the
base metal has a zero hydrogen concentration. As the cooling period begins, the hydrogen concentration
at the exterior of the weld metal is assumed to drop to zero instantaneously.
Time stepping
The problem is run in two parts. The rst part consists of a step in which a single increment of *MASS
DIFFUSION, STEADY STATE analysis is performed with an arbitrary time step to establish the initial
steady-state hydrogen concentration distribution corresponding to the initial temperature.
The hydrogen diffusion during cooling is then analyzed in four subsequent *MASS DIFFUSION
transient analysis steps, using automatic time stepping. This need not be done in four separate steps. We
do it here because the results given by Fujii et al. (1982), with which we compare the Abaqus results,
are presented at four specic times during the transient: 2.7 h (673.15 K, 400.0C), 5.2 h (623.15 K,
350.0C), 10.2 h (523.15 K, 250.0C), and 21.5 h (298.15 K, 25.0C).
The accuracy of the time integration for the *MASS DIFFUSION transient analysis steps, during
which cooling occurs, is controlled by the DCMAX parameter. This parameter species the allowable
normalized concentration change per time step. Even in a linear problem such as this, DCMAX controls
the accuracy of the solution because the time integration operator is not exact (the backward difference
rule is used). In this case DCMAX is chosen as 0.01 N1/2 mm1 , which is a very tight value. This is
necessary to obtain an acceptably accurate integration of the concentration because the solubility of
the materials decreases signicantly (by more than two orders of magnitude in the base metal) as the
temperature decreases and, therefore, the changes in concentration become larger for a given change in
normalized concentration.
An important issue in transient diffusion problems is the choice of initial time step. As in any
transient problem, the spatial element size and the time step are related to the extent that time steps
smaller than a certain size may lead to spurious oscillations in the solution and, therefore, provide no
useful information. This coupling of the spatial and temporal approximations is always most obvious
at the start of diffusion problems, immediately after prescribed changes in the boundary values. For the
mass diffusion case the suggested guideline for choosing the initial time increment (see Mass diffusion
analysis, Section 6.9.1 of the Abaqus Analysis Users Manual) is
8.1.12

Printed on:
HYDROGEN DIFFUSION
where
is a characteristic element size near the disturbance (that is, near the weld metal surface in
our case), and D is the diffusivity of the material. For the weld metal in our model we choose a typical
0.125 mm and we have
0.85 mm2 /h at the initial temperature, which gives
0.003 h.
For the base metal in our model we choose a typical
1.25 mm and we have
4.88 mm2 /h at
the initial temperature, which gives
0.053 h. Based on these calculations an initial time step of
0.1 h is used, which gives an initial solution with no oscillations, as expected.
Figure 8.1.13 shows hydrogen concentration distributions in the weld metal for the initial steady-state
condition and four different times during the cooling period. Figure 8.1.14 shows corresponding
hydrogen concentration distributions in the base metal. These results compare very well with
those presented by Fujii et al. (1982) which are not plotted here since they would appear almost
indistinguishable from the Abaqus results.
It can be observed that, although the primary solution variable (the normalized concentration)
remains continuous across the material interface during the transient, the hydrogen concentration
becomes increasingly discontinuous across the interface. During the cooling process the hydrogen
concentration in the base metal decreases, whereas the hydrogen concentration in the weld metal
increases very signicantly, reaching a peak at the weld metal side of the interface.
Input files
hydrodiffvesselwall_2d.inp
hydrodiffvesselwall_3d.inp
hydrodiffvesselwall_3d_po.inp
hydrodiffvesselwall_ck.inp
hydrodiffvesselwall_nonlinear.inp
hydrodiffvesselwall_heat.inp
hydrodiffvesselwall_massdiff.inp
Two-dimensional analysis.
Three-dimensional analysis.
*POST OUTPUT analysis of
hydrodiffvesselwall_3d.inp.
Two-dimensional analysis using Ficks law.
Nonlinear version (including concentration dependence
on the material properties) of the two-dimensional
analysis.
Heat transfer analysis that writes temperatures to a results
le for use in hydrodiffvesselwall_massdiff.inp.
Two-dimensional mass diffusion analysis that
reads temperatures from the results le written in
hydrodiffvesselwall_heat.inp.
Reference
Fujii, T., T. Nazama, H. Makajima, and R. Horita, A Safety Analysis on Overlay Disbonding of
Pressure Vessels for Hydrogen Service, Journal of the American Society for Metals, pp. 361368,
1982.
8.1.13

Printed on:
HYDROGEN DIFFUSION
base metal
weld metal
Finite element model slice
5 mm
200 mm
Figure 8.1.11
Pressure vessel shell wall section.
Figure 8.1.12
Finite element model of shell wall.
2
3
8.1.14

Printed on:
HYDROGEN DIFFUSION
t
t
t
t
t
Figure 8.1.13
Hydrogen concentration distribution in weld metal.
t
t
t
t
t
Figure 8.1.14
= 0.0 h
= 2.7 h
= 5.2 h
= 10.2 h
= 21.5 h
Hydrogen concentration distribution in base metal.
8.1.15

Printed on:
= 0.0 h
= 2.7 h
= 5.2 h
= 10.2 h
= 21.5 h
DIFFUSION TOWARD A CRACK
8.1.2
DIFFUSION TOWARD AN ELASTIC CRACK TIP
This simple two-dimensional problem veries the sequentially coupled, stress-assisted mass diffusion
capability in Abaqus. The mass diffusion formulation used in Abaqus is described in Mass diffusion
analysis, Section 6.9.1 of the Abaqus Analysis Users Manual, and Mass diffusion analysis, Section 2.13.1
of the Abaqus Theory Manual.
A center-cracked plate fabricated from 2 1/4 Cr1 Mo steel alloy is subjected to end loading in a
hydrogen-rich environment. Hydrogen is drawn to the crack-tip region by high hydrostatic stresses and may
assist in crack growth resulting from hydrogen embrittlement. In this example we are concerned with the
hydrogen diffusion aspect of the problem.
Geometry and model
The problem geometry and boundary conditions are shown in Figure 8.1.21. The specimen is 10-mm
thick, 20-mm wide, and 80-mm high, with a 4-mm crack at its center. The mesh near the crack is focused
at the crack tip, with the element size growing as the square of the distance to the crack tip (*NFILL,
SINGULAR=1). A very ne mesh (see Figure 8.1.22) is used to capture accurately the gradients of
concentration and stress near the crack tip. Four combinations of stress and mass diffusion analyses are
presented:
Stress analysis with quadratic elements and quarter-point spacing at the crack tip, followed by a
mass diffusion analysis with linear elements.
Stress analysis with quadratic elements and quarter-point spacing at the crack tip, followed by a
mass diffusion analysis with quadratic elements and quarter-point spacing at the crack tip.
Stress analysis with quadratic elements (no quarter-point spacing), followed by a mass diffusion
analysis with quadratic elements (no quarter-point spacing).
Stress analysis with linear elements, followed by a mass diffusion analysis with linear elements.
singularity
The quarter-point spacing technique is used in fracture mechanics analyses to enforce a
at the crack tip, where r is the distance from the crack tip.
The sequentially coupled mass diffusion analysis consists of a static stress analysis, followed by a
mass diffusion analysis. Equivalent pressure stresses from the static analysis are written to the results
le as nodal averaged values. Subsequently, these pressures are read in during the course of the mass
diffusion analysis to provide a driving force for mass diffusion.
The material properties for mass diffusion given by Fujii et al. (1982) are as follows.
Solubility:
8.1.21

Printed on:
Diffusivity:
where is the temperature in degrees Celsius and

273 is the absolute zero temperature.
Stress-assisted diffusion is specied by dening the pressure stress factor,
as
where
8.31432 Jmol1 K1 is the universal gas constant,
2.0 103 mm3 mol1 is the partial molar
volume of hydrogen in iron-based metals, and is the normalized concentration. The concentration
dependence of
is entered in Abaqus in tabulated form as shown in the input listings. It is important to
note that although is dened in terms of normalized concentration, , the tabular data must be entered
in terms of concentration,
The following properties are also used in the stress analysis: elastic modulus,
2.0 105 Nmm2 ,
and Poissons ratio,
0.3.
The specimen is maintained at a constant temperature of
325 K throughout the analysis.
Under the initial steady-state conditions the specimen has a uniform concentration of 50 ppm, which
corresponds to a normalized concentration of 265 N1/2 mm1 . Normalized concentration is used as the
primary solution variable (continuous over the discretized domain) and is given as the concentration
divided by the solubility. The exterior of the specimen has a constant hydrogen concentration equal to
the initial concentration. A 1 MPa distributed pressure is applied to the ends of the specimen, ramped
linearly over the length of the step, and the steady-state distribution of hydrogen is obtained.
The analytical solution for normalized concentration, presented by Liu (1970), has the form
where is the normalized concentration obtained in the unstressed state and p is the equivalent pressure
stress. This solution dictates that for a crack-tip problem, the concentration follows the singularity of the
stresses.
Figure 8.1.23 and Figure 8.1.24 show the nal distribution of equivalent pressure stress and
concentration predicted by the Abaqus analysis in the region around the crack tip. The results shown
represent the rst case described above, using a quadratic, quarter-point mesh for stresses and a linear
mesh for mass diffusion. The shapes of the contours show good agreement, since contours of constant
pressure stress should be contours of constant concentration, as indicated by the analytical solution above.
Figure 8.1.25 and Figure 8.1.26 show the pressures (in MPa) and concentrations (in ppm) ahead
of the crack tip for all four combinations of stress and mass diffusion analyses. Results are presented
as functions of the ratio of the distance to the crack tip, r, over the crack length, a. For the region
8.1.22

Printed on:
immediately ahead of the crack, linear elastic fracture mechanics yields the analytical solution for
equivalent pressure stress:
where
is the stress intensity factor for a Mode I crack of length a and is the externally
applied distributed load.
As can be seen from the gures, the nite element results for all four combinations of element
types are identical except at the rst element, where the results are not expected to be valid. The results
show good agreement with the analytically predicted solutions for both equivalent pressure stress and
concentration as the distance to the crack tip, r, approaches zero. Farther from the crack tip, the deviation
between the analytical solution and the nite element solution increases. This deviation is consistent with
the fact that the linear elastic crack-tip solution is valid only as r approaches zero.
No mesh convergence studies were conducted with respect to the number of elements in the crack-tip
region. For comparison with the solutions presented here, an analysis was conducted with equally spaced
elements approaching the crack tip. The results (not shown here) indicate that biasing the elements
toward the crack tip is necessary to capture the gradients of concentration and equivalent pressure stress
adequately. In addition, the equivalent pressure stress results demonstrate that the effect of using quarterpoint positioning of the nodes at the crack tip is insignicant in this problem as long as the mesh is rened
sufciently.
Differences between the nite element and analytically predicted concentrations are a direct result
of the differences between the nite element and analytically predicted values of pressure stress. If the
analytical values of equivalent pressure stress are used to drive the Abaqus concentration solution, the
resulting curve is indistinguishable from the analytical concentration shown.
Input files
difftocrack_quarterpstress.inp
difftocrack_linearmassdiff1.inp
difftocrack_stress.inp
difftocrack_massdiff.inp
difftocrack_quarterpmassdiff.inp
Quadratic stress analysis with quarter-point spacing

at the crack tip.
This analysis writes the results
le used in difftocrack_linearmassdiff1.inp and
difftocrack_quarterpmassdiff.inp.
Linear mass diffusion analysis that reads results le data
from difftocrack_quarterpstress.inp.
Stress analysis with quadratic elements (no quarter-point
spacing). This analysis writes the results le used in
difftocrack_massdiff.inp.
Mass diffusion analysis with quadratic elements that reads
equivalent pressure stresses from the results le written in
difftocrack_stress.inp.
Mass diffusion analysis with quadratic elements and
quarter-point spacing. This analysis reads equivalent
8.1.23

Printed on:
pressure stresses from the results le written in

difftocrack_quarterpstress.inp.
Stress analysis with linear elements. This analysis writes
the results le used in difftocrack_linearmassdiff2.inp.
Mass diffusion analysis with linear elements that reads
equivalent pressure stresses from the results le written
in difftocrack_linearstress.inp.
Node data for all the analyses.
Element data for the analyses using quadratic elements.
Element data for the analyses using linear elements.
difftocrack_linearstress.inp
difftocrack_linearmassdiff2.inp
difftocrack_node.inp
difftocrack_quad_elements.inp
difftocrack_linear_elements.inp
References
Fujii, T., T. Hazama, H. Nakajima, and R. Horita, A Safety Analysis on Overlay Disbonding of
Pressure Vessels for Hydrogen Service, Journal of the American Society for Metals, pp. 361368,
1982.
Liu, H. W., Stress-Corrosion Cracking and the Interaction Between Crack-Tip Stress Field and
Solute Atoms, Transactions of the ASME: Journal of Basic Engineering, vol. 92, pp. 633638,
1970.
2b
model region
2h
sharp
crack
2a
a = 0.2 cm
b = 1 cm
h = 4 cm
= o
Figure 8.1.21
Center crack specimen geometry.
8.1.24

Printed on:
Figure 8.1.22 Finite element model of center crack specimen

(with 1/4 symmetry) with detail of crack-tip mesh.
8.1.25

Printed on:
PRESS
VALUE
-5.74E+01
-5.25E+01
-4.76E+01
-4.27E+01
-3.78E+01
-3.28E+01
-2.79E+01
-2.30E+01
-1.81E+01
-1.32E+01
-8.31E+00
-3.40E+00
Figure 8.1.23
NNC11
Contours of equivalent pressure stress at the crack tip.
VALUE
+2.65E+02
+2.66E+02
+2.67E+02
+2.68E+02
+2.69E+02
+2.70E+02
+2.71E+02
+2.72E+02
+2.73E+02
+2.74E+02
+2.75E+02
+2.76E+02
Figure 8.1.24
Contours of normalized hydrogen concentration at the crack tip.
8.1.26

Printed on:
0.
linear
quad
-2.
quad_sing
Equivalent Pressure Stress [MPa]
analytical
-4.
-6.
-8.
-10.
-12.
0.00
0.05
0.10
0.15
0.20
Distance from Crack Tip (r/a)
Figure 8.1.25
Distribution of pressure stress ahead of the crack tip.

51.0
linear
quad_sing
quad
analytical
Normalized Concentration [ppm]
linear_sing
50.5
50.0
0.00
0.05
0.10
0.15
0.20
Distance from Crack Tip (r/a)
Figure 8.1.26
Hydrogen concentration distribution ahead of the crack tip.
8.1.27

Printed on:
ACOUSTIC AND SHOCK ANALYSES
9.
Acoustic and Shock Analyses
Acoustic and shock analyses, Section 9.1

Printed on:
ACOUSTIC AND SHOCK ANALYSES
9.1
Acoustic and shock analyses
Fully and sequentially coupled acoustic-structural analysis of a mufer, Section 9.1.1

Coupled acoustic-structural analysis of a speaker, Section 9.1.2
Response of a submerged cylinder to an underwater explosion shock wave, Section 9.1.3
Convergence studies for shock analyses using shell elements, Section 9.1.4
UNDEX analysis of a detailed submarine model, Section 9.1.5
Coupled acoustic-structural analysis of a pick-up truck, Section 9.1.6
Long-duration response of a submerged cylinder to an underwater explosion, Section 9.1.7
Deformation of a sandwich plate under CONWEP blast loading, Section 9.1.8
9.11

Printed on:
STRUCTURAL ACOUSTICS OF A MUFFLER
9.1.1
FULLY AND SEQUENTIALLY COUPLED ACOUSTIC-STRUCTURAL ANALYSIS OF

A MUFFLER
Abaqus/Explicit
This example demonstrates the solution of the acoustic eld in the vicinity of a mufer in air caused by the
vibrations of the mufer shell. Steady-state and transient dynamic computations are done using both the fully
coupled (Acoustic, shock, and coupled acoustic-structural analysis, Section 6.10.1 of the Abaqus Analysis
Users Manual) and sequentially coupled acoustic-solid (Node-based submodeling, Section 10.2.2 of the
Abaqus Analysis Users Manual) interaction procedures in Abaqus. In the fully coupled case the solid medium
of the mufer is directly coupled to the enclosed and surrounding air in a single analysis. In the sequentially
coupled case the mufer vibrations are considered to be independent of the loading effects of the surrounding
air, while the acoustic vibrations of the surrounding air are forced by the motion of the mufer. This allows the
mufer vibration and acoustic radiation problems to be solved in sequence, using the submodeling procedure
in Abaqus. The results for the sequentially coupled model are veried by comparing them to the results from
the fully coupled procedure.
Full modeling vs. submodeling in Abaqus
The fully coupled model includes the effect of the acoustic pressure in the surrounding air loading the
mufer body during vibration of the system. When modeling the acoustics of metal structures in air,
such as in this case, such acoustic pressure loading is often negligible in comparison with other forces in
the structure. The submodeling capability (*SUBMODEL) can be used in this situation. The part of the
interacting system that is unaffected by the other is treated as the global model, while the part whose
solution depends strongly on the solution of the other is treated as the submodel. In the case of an
acoustic analysis, of course, this nomenclature refers to the hierarchy of the solutions, not the geometric
sizes of the models.
When sequential coupling is physically appropriate, its use offers a performance advantage
over a fully coupled solution. Two problems, each smaller than the fully coupled problem, are less
computationally expensive. If the applicability of the sequentially coupled solution method is uncertain,
the user should make characteristic test computations in the frequency range of interest. If these
computations show little difference between the fully and sequentially coupled solutions, the less
expensive sequentially coupled method can be used.
Geometry and model
The system considered here consists of a cylindrical mufer and the interacting air. The mufer is a
simple tube 180 mm in diameter and 1 m in length, with inlet and outlet pipes 70 mm in diameter and
100 mm in length. The mufer structure is made from stainless steel sheeting, 0.75 mm in thickness. A
porous packing material, which dampens the acoustic eld, surrounds the inner pipe.
Although this problem is in essence axisymmetric, a narrow three-dimensional wedge (subtending
an angle of 10) of the coupled system is modeled because Abaqus has a limitation on the use
9.1.11

Printed on:
of submodeling with axisymmetric shells. Appropriate boundary conditions are applied to the
three-dimensional model so that the axisymmetric solution is captured. The meshes of the surrounding
air, the exterior mufer shell, and the air inside the mufer are shown in Figure 9.1.11, Figure 9.1.12,
and Figure 9.1.13, respectively.
The air inside the mufer is meshed with AC3D10 elements (second-order tetrahedra) in
Abaqus/Standard and with AC3D4 elements in Abaqus/Explicit. The innermost column of uid
elements models the undamped air. The adjacent annulus models the air in the region of the packing
material. These two regions are highlighted in Figure 9.1.13, where the annulus is shown as the
darker region. The effect of the packing material is modeled using the *ACOUSTIC MEDIUM,
VOLUMETRIC DRAG option. The mufer is meshed with S4R shell elements.
The exterior uid is shown in Figure 9.1.11. Its outer boundary is made up of spherical and
cylindrical segments, on which spherical and cylindrical absorbing boundary conditions are imposed
using *SIMPEDANCE, TYPE=SPHERE and *SIMPEDANCE, TYPE=CIRCULAR, respectively. The
cylindrical and spherical absorbing boundary conditions can be combined in Abaqus, allowing the
external mesh to conform to the geometry of the radiating object more closely. Combinations of different
boundary condition types are most effective when the boundaries are continuous in slope as well as
displacement. Second-order hexahedral acoustic elements (AC3D20) are used in Abaqus/Standard and
reduced-integration acoustic brick elements (AC3D8R) are used in Abaqus/Explicit to ll in the volume
of the exterior uid region.
In Abaqus/Explicit the possibility of using acoustic innite elements to model the effect of the
exterior uid is explored. The use of acoustic innite elements removes the need of *SIMPEDANCEtype absorbing boundary conditions on the outer boundary. Acoustic innite elements are used in two
different ways. In the rst approach the mesh modeling the exterior uid is replaced by a single row
of AC3D8R elements, and acoustic innite elements ACIN3D4 are dened on the outer boundary of
this row. In the second approach ACIN3D4 elements are dened directly on the outer boundary of the
mufer and tied to the mufer surface using the *TIE option.
In the submodeling procedure performed in Abaqus/Standard the interface between the surrounding
air and the mufer is meshed with 8-node acoustic interface elements (ASI8); in the Abaqus/Explicit
submodeling analysis the *TIE option is used to dene this coupling. The choice of mesh density
(element size) is discussed in Acoustic, shock, and coupled acoustic-structural analysis, Section 6.10.1
of the Abaqus Analysis Users Manual. In both cases the inner boundary of the exterior air mesh
conforms to the mufer shell and to rigid bafes, which isolate the exterior eld from the exhaust and
inlet noise. These bafe pipes are the same diameter as the inlet and exhaust pipes but are modeled
simply by imposing no boundary condition on the acoustic elements in this region. This is equivalent to
imposing the condition that the acceleration on this boundary is zero, which is correct for a rigid bafe.
In Abaqus/Standard we are most interested in performing a frequency sweep about the rst resonant
frequency of the fully coupled system. For problems involving air and metal structures, the structure
usually dominates the behavior of the system. Therefore, an estimate of the rst important resonance of
the coupled system is found by performing a frequency sweep in the vicinity of the rst eigenfrequency
of the mufer shell, computed without any interaction with the interior or exterior air. This occurs at
f = 172 Hz. Although the resonant frequencies of the fully coupled system do not coincide with the
resonant frequencies of the mufer shell alone, they are close, especially at lower frequencies.
9.1.12

Printed on:
Using the Abaqus/Standard *STEADY STATE DYNAMICS, DIRECT procedure to search around
172 Hz, we nd that the rst resonant frequency for the fully coupled system occurs at approximately
180 Hz. A frequency sweep of both the fully coupled and the sequentially coupled models from 179.0 Hz
to 181.0 Hz at 0.2 Hz increments is performed. A pressure wave of unit magnitude is applied to the
mufer inlet at each frequency, and a plane wave absorbing boundary condition is applied at the mufer
outlet.
A transient dynamic analysis is performed in Abaqus/Explicit over the period of time that
corresponds to the rst resonant frequency of 180 Hz found in Abaqus/Standard. The pressure boundary
conditions applied at the mufer inlet have a sinusoidal variation over time to simulate the steady-state
dynamic procedure performed in Abaqus/Standard. The absorbing boundary conditions are imposed in
the same way as in the *STEADY STATE DYNAMICS procedure.
The material properties for the air are a bulk modulus
of 0.142 MPa and a density
of
1.2 kg/m3 , yielding a characteristic sound speed of 344 m/s. The volumetric drag, , specied for the
air in the packing material region is 1.2 N s/m. Volumetric drag values are considered small if they
are small compared to 2
, a condition satised by = 1.2 N s/m for the frequency range of interest.
The mufer is made of stainless steel with Youngs modulus E of 190 GPa, Poissons ratio of 0.3, and
density
of 7920 kg/m3 .
Material properties affect the mesh parameters appropriate for wave problems. The characteristic
wavelength of air at
180 Hz,
1131 rad/sec, is
1.91 m, which
is long compared to the overall system geometry. The internodal spacing of roughly 40 mm used in
the surrounding acoustic mesh and 30 mm in the interior acoustic mesh is adequate for this frequency.
The acoustic wavelength must also be considered in selecting the overall size of the exterior domain.
Accuracy of the solution requires placement of the radiating boundary at least one-quarter wavelength
from the acoustic sources; in this problem a standoff distance of approximately 700 mm is selected. The
characteristic exural wavelength
of the steel plating can be computed using the thickness h and the
formula
203 mm. The discretization requirements of the nite element
method in wave problems require at least six nodes per wavelength; here, we use an internodal distance
of approximately 30 mm for the shells.
The fully coupled model consists of all three meshes shown in Figure 9.1.11, Figure 9.1.12, and
Figure 9.1.13, constrained at their abutting surfaces using the *TIE option.
The sequentially coupled analysis is performed in two jobs. The global model job consists of the
meshes shown in Figure 9.1.12 and Figure 9.1.13. The shell displacements, and displacement phases
in Abaqus/Standard, are saved from this analysis and drive the second submodel analysis through
the use of the *BOUNDARY, SUBMODEL option. In Abaqus/Standard the second model consists of
the exterior air mesh (Figure 9.1.11) used in the fully coupled case, with ASI8 elements placed on
the boundary that abuts the shell surface. These elements convert the displacements from the global
analysis to the appropriate boundary conditions for acoustic elements. In this analysis the ASI8 elements
conform to the acoustic submodel mesh but not to the shell mesh of the global model. The nodes of the
ASI8 elements are placed in a node set, specied in the model data by the *SUBMODEL option. The
GLOBAL ELSET parameter must be used in this case to ensure that only the displacements of the ASI8
elements are driven by the shell elements. Without the GLOBAL ELSET parameter, Abaqus may attempt
to drive the acoustic pressure of the ASI8 elements by the interior acoustic elements, since those elements
9.1.13

Printed on:
share the shell nodes in the global model. In Abaqus/Explicit the *TIE option is used in both the global
and submodel analyses to couple the mufer structure with the surrounding acoustic medium.
It is good practice to check the absorbing boundary conditions used on a particular mesh at a desired
frequency by analyzing only the exterior uid mesh with some test forcing on the boundary where
acoustic excitations are expected. If the forcing is at a single point, the pressure phase angles should
show a pattern of concentric circles, minimally distorted by the radiating boundary. While not a rigorous
numerical test, such a result usually coincides with a properly offset radiating boundary. As shown in
Figure 9.1.14, this criterion is met by the mesh used in this analysis.
Figure 9.1.15 is a plot of the radial displacement of the mufer inlet as a function of frequency for
both the fully coupled and the global models. The resonant peak for the fully coupled model at 179.9 Hz
is clearly illustrated. In contrast, the resonant peak for the global model (without the acoustic medium)
occurs at approximately 180.0 Hz. The difference in the two peaks can be accounted for by the fact that
the exterior air on the fully coupled model adds a small amount of damping due to radiation as well as
mass to the system, which results in a lower natural frequency, as well as a slightly lower peak response.
It is clear from Figure 9.1.15 that for the frequency range of interest the coupling between the exterior
air and the mufer is most important at 179.9 Hz.
Figure 9.1.16 and Figure 9.1.17 contain contour plots of the pressure magnitude and phase for
the mufer interior at 181.0 Hz for both the global model and the fully coupled model. In both cases
the results indicate that the modeling assumptions of the sequentially coupled analysis appear to be valid
for the solutions in the mufer interior.
Contour plots of the pressure magnitude and phase for the mufer exterior at 181.0 Hz are shown in
Figure 9.1.18 and Figure 9.1.19. The resulting pressure magnitude in the exterior air is small in both
cases. The differences in the pressure amplitudes and phase as computed by the two analyses are not
considered to be signicant. Two factors that account for the small differences are the different modeling
methods (fully coupled vs. sequentially coupled) and the different techniques used to couple the mufer
to the exterior air (*TIE vs. acoustic interface elements).
Figure 9.1.110 and Figure 9.1.111 contain contour plots of the pressure magnitude and phase for
the mufer interior at 179.9 Hz for both the global model and the fully coupled model. It is clear that
at 181.0 Hz, the modeling assumptions of the sequentially coupled analysis are less valid than they are at
179.9 Hz for the solutions in the mufer interior. This result is anticipated by Figure 9.1.15. However,
the solutions are still reasonably close to one another, indicating that the sequentially coupled analysis is
still a reasonable approximation for this system even at a resonant peak.
Contour plots of the pressure magnitude and phase for the mufer exterior at 179.9 Hz are shown in
Figure 9.1.112 and Figure 9.1.113. Again, the resulting pressure magnitude in the exterior air is small
in both cases. The differences in the pressure amplitudes and phase as computed by the two analyses are
less evident in the exterior than they were in the interior.
The pressure magnitudes along the mufer centerline at both 179.9 Hz and 181.0 Hz are shown
in decibels in Figure 9.1.114. The reference pressure is chosen as one unit for convenience. The plot
illustrates the variation of acoustic pressure in the mufer near resonance.
9.1.14

Printed on:
Table 9.1.11 shows comparative solution times and memory requirements for the fully and
sequentially coupled analyses. The total computational time for the sequentially coupled case is lower,
and the peak memory requirements are signicantly lower. These differences will be greater for larger
models. Optimal speed increases occur when global and submodels have nearly equal numbers of
degrees of freedom. Here, solving the fully coupled system does not impose as much of a speed penalty
as might be expected, because the sparse solver used by Abaqus exploits the extreme sparsity of the
uid-solid coupling term. When the number of system nodes involving uid-solid coupling is a large
percentage of the total number of nodes, the sparsity of the coupling term decreases, favoring the
sequentially coupled procedure. Sequentially coupled analyses are even more advantageous than fully
coupled analyses when many different submodels need to be analyzed, driven by a single set of global
results.
Abaqus issues a series of warning messages in this example, because the narrow wedge domain
results in some three-dimensional acoustic elements with bad aspect ratios. These messages can be
ignored in this study, since the solutions are essentially axisymmetric and the gradient of the solution in
the circumferential direction is nearly zero. Moreover, elements with scalar degrees of freedom, such as
the acoustic elements used in this example, are much less sensitive to geometric distortion than elements
with vector degrees of freedom, such as continuum stress/displacement elements.
The results obtained in Abaqus/Explicit agree well with the Abaqus/Standard results. For the fully
coupled analysis the pressure variation in time at the mufer outlet centerline is shown in Figure 9.1.115
(for a clear comparison the Abaqus/Standard analysis is also performed as a transient simulation). The
Abaqus/Explicit models using acoustic innite elements give results that agree well with the results using
the *SIMPEDANCE-type absorbing boundary. In Figure 9.1.115 we include the results for the test
using acoustic innite elements, where the mesh modeling the exterior uid is replaced by a single row
of AC3D8R elements and acoustic innite elements ACIN3D4 are dened on the outer boundary of this
row. For the Abaqus/Explicit submodeling analysis the inside air pressure in the global model and the
outside air pressure of the submodel compare well with the air pressures obtained in these regions in the
fully coupled problem.
Input files
Abaqus/Standard input files
mufer_full.inp
mufer_globl.inp
mufer_innerair_freq.inp
mufer_innerair_freq_ams.inp
mufer_submo.inp
mufer_shell_nodes.inp
mufer_intair_nodes.inp
mufer_extair_nodes.inp
mufer_shell_elem.inp
mufer_intair_elem.inp
mufer_extair_elem.inp
Three-dimensional fully coupled model.

Mufer and internal air global model.
Internal air eigenanalysis model, Lanczos.
Internal air eigenanalysis model, AMS.
Exterior air submodel.
Nodal coordinates for mufer shell mesh.
Nodal coordinates for interior air mesh.
Nodal coordinates for surrounding air mesh.
Element denitions for mufer shell mesh.
Element denitions for interior air mesh.
Element denitions for surrounding air mesh.
9.1.15

Printed on:
mufer_freq.inp
mufer_bctest.inp
Natural frequency extraction for shell mesh.

Radiating boundary condition test.
Abaqus/Explicit input files
mufer_full_xpl.inp
mufer_full_acoinfxpl.inp
Three-dimensional fully coupled transient analysis.

Three-dimensional fully coupled transient analysis using
acoustic innite elements.
Three-dimensional fully coupled transient analysis using
acoustic innite elements tied to the mufer outer surface.
Mufer and internal air global model, transient analysis.
Mufer and exterior air submodel, transient analysis.
Nodal coordinates for mufer shell mesh.
Element denitions for mufer shell mesh.
Nodal coordinates for interior air mesh.
Element denitions for interior air mesh.
Nodal coordinates for surrounding air mesh.
Element denitions for surrounding air mesh.
Element denitions for surrounding air mesh for model
using acoustic innite elements.
mufer_full_acoinftiexpl.inp
mufer_global_xpl.inp
mufer_submodel_xpl.inp
mufer_shell_nodes.inp
mufer_shell_elem.inp
mufer_intair_nodes_xpl.inp
mufer_intair_elem_xpl.inp
mufer_extair_nodes_xpl.inp
mufer_extair_elem_xpl.inp
mufer_extair_elem_ainxpl.inp
Table 9.1.11 Comparison of relative CPU times (normalized with

respect to the CPU time for the sequential analysis) and approximate
problem size for the frequency sweep excluding preprocessing.
Global model
Submodel
Fully coupled model
Sequential analysis
Memory
DOF
Relative CPU Time
10 Mb
15 Mb
29 Mb
10030
19030
29060
0.325
0.675
1.086
1.000
9.1.16

Printed on:
Figure 9.1.11
Mesh of surrounding air.
Figure 9.1.12
Mesh of mufer.
Figure 9.1.13
Mesh of interior air.
9.1.17

Printed on:
Figure 9.1.14
Radiating boundary condition test at 165 Hz.
Fully coupled model

Global model
Figure 9.1.15
Radial displacement of the mufer inlet as a function of frequency.
9.1.18

Printed on:
POR
+1.300e+00
+1.192e+00
+1.083e+00
+9.750e-01
+8.667e-01
+7.583e-01
+6.500e-01
+5.417e-01
+4.333e-01
+3.250e-01
+2.167e-01
+1.083e-01
+0.000e+00
Figure 9.1.16 Mufer internal pressure magnitudes at 181.0 Hz, mufer inlet at top: fully coupled
solution on left, global model (without the exterior acoustic medium) on right.
PPOR
+1.800e+02
+1.500e+02
+1.200e+02
+9.000e+01
+6.000e+01
+3.000e+01
+0.000e+00
-3.000e+01
-6.000e+01
-9.000e+01
-1.200e+02
-1.500e+02
-1.800e+02
Figure 9.1.17 Mufer internal pressure phase at 181.0 Hz, mufer inlet at top: fully coupled
9.1.19

Printed on:
POR
+1.260e-01
+1.155e-01
+1.050e-01
+9.450e-02
+8.400e-02
+7.350e-02
+6.300e-02
+5.250e-02
+4.200e-02
+3.150e-02
+2.100e-02
+1.050e-02
+0.000e+00
Figure 9.1.18
Mufer external pressure magnitudes at 181.0 Hz, mufer inlet at top:

fully coupled solution on left, submodel on right.
POR
+1.800e+02
+1.500e+02
+1.200e+02
+9.000e+01
+6.000e+01
+3.000e+01
+0.000e+00
3.000e+01
6.000e+01
9.000e+01
1.200e+02
1.500e+02
1.800e+02
Figure 9.1.19 Mufer external pressure phase at 181.0 Hz, mufer inlet at top:
9.1.110

Printed on:
POR
POR
+2.310e+00
+2.118e+00
+1.925e+00
+1.733e+00
+1.540e+00
+1.348e+00
+1.155e+00
+9.626e01
+7.701e01
+5.775e01
+3.850e01
+1.925e01
+0.000e+00
+2.157e+00
+1.977e+00
+1.798e+00
+1.618e+00
+1.438e+00
+1.258e+00
+1.079e+00
+8.988e-01
+7.190e-01
+5.393e-01
+3.595e-01
+1.798e-01
+0.000e+00
Figure 9.1.110 Mufer internal pressure magnitudes at 179.9 Hz, mufer inlet at top: fully coupled
PPOR
+1.800e+02
+1.500e+02
+1.200e+02
+9.000e+01
+6.000e+01
+3.000e+01
+0.000e+00
-3.000e+01
-6.000e+01
-9.000e+01
-1.200e+02
-1.500e+02
-1.800e+02
Figure 9.1.111 Mufer internal pressure phase at 179.9 Hz, mufer inlet at top: fully coupled
9.1.111

Printed on:
POR
+1.580e-01
+1.448e-01
+1.317e-01
+1.185e-01
+1.053e-01
+9.217e-02
+7.900e-02
+6.583e-02
+5.267e-02
+3.950e-02
+2.633e-02
+1.317e-02
+0.000e+00
Figure 9.1.112
Mufer external pressure magnitudes at 179.9 Hz, mufer inlet at top:

POR
+1.800e+02
+1.500e+02
+1.200e+02
+9.000e+01
+6.000e+01
+3.000e+01
+0.000e+00
3.000e+01
6.000e+01
9.000e+01
1.200e+02
1.500e+02
1.800e+02
Figure 9.1.113 Mufer external pressure phase at 179.9 Hz, mufer inlet at top:
9.1.112

Printed on:
179.91 Hz
181.00 Hz
Figure 9.1.114 Mufer internal pressure magnitude at 179.9

and 181.0 Hz: dB along mufer centerline.
Explicit
Explicit Infinite
Standard
Figure 9.1.115
Internal pressure at the mufer outlet for the transient analysis.
9.1.113

Printed on:
STRUCTURAL ACOUSTICS OF A SPEAKER
9.1.2
COUPLED ACOUSTIC-STRUCTURAL ANALYSIS OF A SPEAKER
This example illustrates the effect of coupling between a structure and an acoustic medium. Such coupling
problems arise when the solid-uid interaction is fundamental to the overall vibrational behavior of the body
or of the acoustic uid. Typical examples of such problems include loudspeaker enclosures, uid-lled tanks,
mufer systems, and vehicle cabin enclosures. The basis of the coupled acoustic/structural vibration capability
in Abaqus is described in Coupled acoustic-structural medium analysis, Section 2.9.1 of the Abaqus Theory
Manual.
Geometry and model
The model is shown in Figure 9.1.21. The system considered here consists of a speaker box, a speaker
cone, and the interacting interior. To simplify the problem, the effect of interacting air at the outside of
the speaker box is neglected. The width, depth, and height of the speaker box are 0.5 m, 0.4 m, and
0.6 m, respectively. Its thickness is 0.015 m. The speaker box is made of wood with a Youngs modulus,
E, of 11.6 GPa; a Poissons ratio, , of 0.3; and a density, , of 562 kg/m3 . At the center of the front
speaker box, there is a cone-shaped speaker 0.345 m in diameter, 0.04 m in height, and 0.0001 m in
thickness. No mass or impedance of the speaker is considered. The speaker is made of polyethylene
with a Youngs modulus, E, of 3.4 GPa; a Poissons ratio, , of 0.3; and a density, , of 450 kg/m3 . The
air has a density, , of 1.11 kg/m3 and a bulk modulus,
, of 0.134 MPa. Volumetric drag of the air is
assumed to have a negligible effect in this problem, so it is ignored in this analysis.
First-order hexahedral acoustic elements (AC3D8) and rst-order acoustic triangular prism
elements (AC3D6) are used to ll in the volume of the interior air region. The speaker box and speaker
are meshed with S4R and S3R elements, respectively. No mesh convergence study has been done since
the example is intended only as an illustration. The choice of mesh density (element size) is discussed
in Acoustic, shock, and coupled acoustic-structural analysis, Section 6.10.1 of the Abaqus Analysis
Users Manual.
The surface-based contact approach is used. Surfaces are dened at the inside of the speaker box
and the speaker cone and at the free surface of the air. The *TIE option is used to couple the structure with
the inside air. To constrain the structure, four corner points of the bottom panel are simply supported.
A substructure analysis is perfomed as well. The entire speaker model is turned into a coupled
structural-acoustic substructure using 150 extracted eigenmodes. A substructure load case is generated
to be used in the forced response analyses.
The results for each of the analyses are discussed below.
9.1.21

Printed on:
Natural frequency analyses
If the eigenvalues of the structure alone or the acoustic medium alone are not in the range of interest, it is
not necessary to consider the whole system simultaneously. Thus, it is recommended to understand the
modal characteristics of each part separately before analyzing the whole system. The *FREQUENCY
procedure takes acoustic-structural coupling effects into account by default if an acoustic medium and a
structure are joined by the *TIE option or by ASI-type elements. To ignore this effect, set the ACOUSTIC
COUPLING parameter equal to OFF on the *FREQUENCY option or remove the *TIE interaction.
The results of the natural frequency analysis of the uncoupled system are summarized in
Table 9.1.21. The natural frequencies for both the structure and the air span the same range in this
example, which shows that the two parts can affect each other. Thus, the coupled approach should be
adopted to understand the characteristics of the whole speaker system in this example.
Table 9.1.22 shows the results of the natural frequency analysis of the coupled system. Due to
the coupling effect, the eigenfrequencies shift mode by mode. Each mode shape is also more complex
than those in the uncoupled case, so that each mode has nonzero components on both the structural and
acoustic parts. The substructure analysis yields eigenfrequencies identical to those from the analysis
without substructures.
Coupled forced response analyses
The response of the system is obtained by using mode-based, direct-solution, or subspace-based

steady-state dynamic analysis. If mode-based or subspace-based steady-state dynamic analysis is
used, a *FREQUENCY step should be performed prior to the *STEADY STATE DYNAMICS step.
Coupled analyses are performed as frequency sweeps from 300400 Hz. The system is excited by a
concentrated force at node 83 (at the center point of the speaker cone), whose magnitude is 1.0 N in
phase and 0.06 N out of phase. The *STEADY STATE DYNAMICS, DIRECT and *STEADY STATE
DYNAMICS, SUBSPACE PROJECTION=ALL FREQUENCIES results for the coupled system are
shown in Figure 9.1.22. This gure illustrates the acoustic pressure at the center point of the speaker
cone, plotted as a function of frequency. The results for the mode-based steady-state dynamic analysis
(not shown) are identical to those for the subspace-based analysis. The substructure analysis produces
results that are virtually identical to the results from the equivalent analysis without substructures. The
effect of the air on the structural response is revealed when the optional ACOUSTIC COUPLING
parameter is set equal to OFF on the *FREQUENCY option. As shown in this gure, the effect of
coupling is quite signicant since both the structure and the acoustic medium have several natural
frequencies in this region. The response peaks correspond to the coupled modes 4, 6, and 7 computed in
the *FREQUENCY step. The transient dynamic response of the system to a concentrated force is also
computed using the *DYNAMIC and *MODAL DYNAMIC analyses types. The responses computed
with the two options compare well.
Input files
speaker.inp
Natural frequency extraction for coupled system,

subspace projection, and direct steady-state dynamics.
9.1.22

Printed on:
speaker_uncoup.inp
speaker_dyn.inp
speaker_gen.inp
speaker_use.inp
speaker_nosubstr.inp
speaker_uncoup_ams.inp
speaker_uncoup_ams_restart.inp
speaker_nosubstr_ams.inp
Natural frequency extraction for uncoupled system and

subspace projection steady-state dynamics.
Natural frequency extraction for coupled, implicit
dynamic time integration and modal dynamic time
integration.
Substructure generation for the coupled structuralacoustic model.
*STEADY STATE DYNAMICS analyses using the
substructure.
Equivalent *STEADY STATE DYNAMICS analysis
without substructures.
Natural frequency extraction for uncoupled system
and steady-state dynamics, using Abaqus/AMS for
eigenanalysis.
Steady-state dynamics, restarting from Abaqus/AMS
eigenanalysis.
Equivalent *STEADY STATE DYNAMICS analysis
without substructures and using Abaqus/AMS for
eigenanalysis.
9.1.23

Printed on:
Table 9.1.21
Uncoupled frequency analysis.
Mode
Frequency, Hz
Description
285.13
Air: z-direction
293.88
Structure: y-direction
315.12
Structure: z-direction
340.14
Air: x-direction
346.14
Structure: front and rear

panel
422.52
Air: y-direction
431.81
Air: skew-symmetric
mode in xz plane
450.26
Structure
10
466.58
Structure
Air: constant pressure
Table 9.1.22
Coupled frequency analysis.
Mode
Frequency, Hz
Description
282.03
Coupled mode
287.84
Coupled mode
308.92
Coupled mode
314.73
Coupled mode
343.01
Coupled mode
348.88
Coupled mode
421.92
Coupled mode
430.62
Coupled mode
10
442.52
Coupled mode
Air: constant pressure
9.1.24

Printed on:
Z
Y
X
Figure 9.1.21
Three-dimensional model of speaker system.
SSD-Subspace-Uncoupled
SSD-Direct
SSD-Subspace-Coupled
Figure 9.1.22
Acoustic pressure behind center of speaker cone.
9.1.25

Printed on:
UNDEX SHOCK ANALYSIS
9.1.3
RESPONSE OF A SUBMERGED CYLINDER TO AN UNDERWATER EXPLOSION

SHOCK WAVE
This example demonstrates how Abaqus/Explicit can be used to predict the transient response of submerged
structures that experience loading by an acoustic pressure shock wave resulting from an underwater explosion
(UNDEX). This class of problem is characterized by a strong coupling between the structural motions and
acoustic pressures on the wetted interface between the external uid and the structure. The structural response
in a strongly coupled acoustic-structural system can be described as a combination of the following:
Low-frequency response characterized by structural wavelengths that are signicantly shorter than the
associated acoustic wavelengths. The external uid on the structure adds an effective mass to the structure
on the wetted interface.
High-frequency response characterized by structural wavelengths that are signicantly longer than the
associated acoustic wavelengths. The external uid on the structure acts as a simple damping mechanism,
where energy is transported away from the structure via acoustic radiation.
Intermediate-frequency response characterized by structural wavelengths that are similar in length to the
associated acoustic wavelengths. In this frequency regime the external uid has both an added mass and
a radiation damping inuence on the structure.
The spherical pressure wave associated with an UNDEX shock loading is characterized by a very steep
front where the maximum pressure is attained over an extremely short time duration (rise time). The pressure
then drops off exponentially over a signicantly longer period of time. Therefore, UNDEX shock loads can
be expected to excite submerged structures over a large frequency range that will include low, high, and
intermediate response frequencies. The boundaries of the external uid must be located a sufcient distance
from the structure to ensure proper low-frequency response, while the size of the acoustic elements must
be small enough to accurately represent the propagation of high-frequency acoustic waves away from the
submerged structure.
Problem and geometry description
This example problem is based upon an UNDEX experiment in which a submerged test cylinder is
exposed to a pressure shock wave produced by a 60 lb HBX-1 explosive charge. Kwon and Fox originally
described the experiment along with a set of selected experimental results. The objective of this class of
analysis is to evaluate the behavior and integrity of a structure under UNDEX loading conditions.
The test cylinder is made of T6061-T6 aluminum. It has an overall length of 1.067, an outside
diameter of 0.305, a wall thickness of 6.35 mm, and 24.5 mm thick welded endcaps. The cylinder is
suspended horizontally in a 40 m deep fresh water test quarry. The 60 lb HBX-1 explosive charge and
the cylinder are both placed at a depth of 3.66 m. The charge is centered off the side of the cylinder and
located 7.62 m from the cylinder surface. The suspension depths, charge offset, and duration of the test
are selected such that cavitation of the uid is not signicant and no bubble pulse occurs. Strain gauges
are placed at several locations on the outer surface of the test cylinder, as shown in Figure 9.1.31. The
9.1.31

Printed on:
strain gauge experimental data are ltered at 2000 Hz. The experimental data presented here are obtained
by digitizing the Kwon and Fox strain history curves.
When the acoustic uid behavior is linear (i.e., no cavitation), the total acoustic pressure within
the uid consists of an incident wave and a scattered wave component. For this example the incident
wave is the shock wave produced by the UNDEX charge. The scattered wave is the acoustic eld
generated by the interaction of the incident wave and the submerged structure. The nature of the incident
wave can be determined from either empirical formulas or experimental data. Therefore, the spherical
incident shock wave is applied as a transient load active on both the acoustic and structural meshes
at their common surfaces (the wetted interface), and the external uid pressure degrees of freedom
represent only the unknown scattered component of the total acoustic pressure. The *ACOUSTIC WAVE
FORMULATION option is used to select either a SCATTERED WAVE or a TOTAL WAVE formulation
for the incident wave loading. The scattered wave formulation described above is the default condition
for Abaqus/Explicit analyses. The total wave formulation is used for cases where nonlinear uid response
is expected or where the total acoustic pressure history is prescribed at an acoustic uid boundary.
During the UNDEX test two pressure transducers are positioned 7.62 m from the charge, away
from the cylinder but at the same depth as the cylinder. These transducers provide an experimental
determination for the pressure vs. time history of the spherical incident shock wave as it travels by the
point on the cylinder closest to the charge (strain gauge location B1). Figure 9.1.32 shows a time history
curve of the incident pressure wave recorded by the transducers. The input le shock-pulse.inp contains
this time history curve as an amplitude table used to dene the incident wave loading.
Abaqus/Explicit model
Figure 9.1.33 shows the S4R nite element shell mesh used to represent the test cylinder. The mesh
consists of 2402 nodes (14412 dof) and 2400 elements with 40 circumferential divisions and 53 axial
divisions. The element connectivity is such that each shell normal is directed into the external uid.
The nodes are positioned on the outside surface of the test cylinder; thus, the OFFSET=SPOS parameter
is included as part of the shell section denitions. The S4R elements adjacent to the endcaps are
dummy elements with reduced mass and stiffness used only to provide surfaces that correspond to the
thickness of the endcaps. BEAM type MPCs are used to tie the endcaps to the main cylinder body. The
*ORIENTATION option is used to dene the shell element material axes for postprocessing, such that
the local 1-direction is aligned with the cylinders axis for the main body and is radially directed for
the endcaps. The local 2-direction is in the circumferential (hoop) direction for both the cylinder main
body and the endcaps.
The external uid is meshed with 4-node AC3D4 acoustic tetrahedral elements. The outer boundary
of the external uid is represented by a cylindrical surface with spherical ends. The characteristic radius
of the outer boundary is 0.915 m. The outer boundary must be placed a sufcient distance from the
cylinder so that the added mass associated with the low-frequency beam bending modes of the cylinder
is represented adequately. The beam bending modes correspond to an
sinusoidal translation of
the cylinders cross-section through the uid. For evaluating added mass effects when using a simple
plane wave radiation impedance boundary for the external uid, the outer boundary of the uid can be
considered rigid (nonradiating). Therefore, an analytical solution for the added mass associated with the
translation of an innite cylinder of radius
located within a uid-lled innite cylinder of radius
9.1.32

Printed on:
can be used to determine an appropriate characteristic radius for the external uid. Results for
the analytical solution presented by Blevins are listed in Table 9.1.31. The characteristic radius is
based upon an outer boundary ( ) to cylinder radius ( ) ratio of 6.0, which corresponds to an added
mass error of about 6% for innite cylinders. When using enhanced surface impedance models (i.e.,
*IMPEDANCE PROPERTY, TYPE=CIRCULAR or SPHERE) the outer uid boundary location can
be placed at about half of the distance required when using the plane wave radiation impedance model.
However, for this example the
ratio was maintained at 6.0 even when using the source-based
surface impedance models for the external uid. Comparable results for the structural response can
be obtained when the source-based boundaries were located half as far from the structure. For the
low-frequency beam bending modes, system losses (damping) caused by hydrodynamic drag and/or uid
viscosity are not accounted for by acoustic radiation. Therefore, mass-proportional damping applied to
the test cylinder mesh is used to approximate these types of losses.
Figure 9.1.34 shows the combined external uid and test cylinder meshes. One quarter of the
uid mesh is omitted from this gure to allow an inspection of the acoustic element mesh inside the
external uid domain. The mesh is generated with Abaqus/CAE. The nodal seeding on the uid
outer boundary is set at 0.10 m, corresponding to 9.7 element divisions per acoustic wavelength at a
response frequency of 1500 Hz. The nodal seeding on the uid wetted interface with the test cylinder
is set at 0.04 m, corresponding to 24.4 element divisions per acoustic wavelength at 1500 Hz. The
*IMPEDANCE PROPERTY option is used to dene radiation properties for the uid outer boundary
cylindrical and spherical surfaces (TYPE=CIRCULAR and SPHERE). The *SIMPEDANCE option is
used to activate the radiation boundary condition on the uid outer-boundary surfaces.
Fluid-structure coupling and shock wave loading
The acoustic structural coupling between the uid mesh acoustic pressures and the test cylinder structural
displacements at their common surfaces (the wetted interface) is accomplished with the *TIE constraint
option. Figure 9.1.35 shows the surface mesh at the acoustic-structure wetted interface associated with
the external uid (Figure 9.1.33 shows the test cylinder surface). Since the acoustic mesh is coarser than
the structural mesh, the surface of the external uid at the wetted interface is designated as the master
surface. This pairing creates an internal coupling of the acoustic pressure and structural displacements at
the test cylinder (slave) surface nodes and ties the cylinders acoustic pressures to the uid mesh acoustic
pressures at the wetted interface.
Figure 9.1.35 also illustrates the concept of a source point and a standoff point as they relate to an
incident acoustic wave loading. For this example the source point represents the actual physical location
of the explosive charge relative to the structure. The standoff point represents the location of the incident
wave (shock front) at the start of the analysis (total time = 0.0) and is the point at which the pressure
history of the incident wave is provided. For solution efciency the standoff point should be placed at
the location on the uid-structure interface that is closest to the source point. The standoff point can
be placed away from the structure closer to the source point, but this will only delay the onset of the
transient response. Under no circumstances should the standoff point be located within or behind the
structure being analyzed.
The *INCIDENT WAVE PROPERTY option is used to specify the incident wave as either planar
or spherical. The data lines for this option are used to dene the location of the standoff point and
9.1.33

Printed on:
source point in terms of global Cartesian coordinates. For a spherical shock wave, as in this example, the
relative positions of the standoff point and source point determine how the waves pressure will decay
with distance from the source point. For a planar wave, which does not decay, the relative positions
of the standoff point and source point are used to dene the direction of incident wave travel. The
*INCIDENT WAVE FLUID PROPERTY option denes the acoustic properties for the incident wave
(i.e., wave speed) and must directly follow the *INCIDENT WAVE PROPERTY option. Dening the
incident wave properties independent of the acoustic mesh allows incident wave loading to be used in
the analysis of weakly coupled or uncoupled acoustic-structural systems (i.e., air blast analyses). For
these cases the incident wave loading can be applied to a structure when no acoustic medium is directly
modeled.
The *INCIDENT WAVE option is used to activate incident wave loading. The PROPERTY
parameter on this option is used to designate the name assigned to the appropriate *INCIDENT WAVE
PROPERTY option. The PRESSURE AMPLITUDE parameter is used to designate an amplitude table
that denes the incident wave pressure history at the standoff point. The amplitude data lines specify the
surface name to which the incident wave loading is applied and a reference magnitude for the pressure
curve. For acoustic-structural systems where the uid and structure are both modeled and coupled, the
incident wave loading must be dened to act upon both the uid and structural surfaces at the wetted
interface. Acoustic volumetric acceleration loads corresponding to the incident wave are then applied
to the uid surface, while the incident wave pressures are applied to the structural surface.
The Abaqus/Explicit model for this UNDEX example has a total of 23337 active degrees of freedom
and requires approximately 160 MB of memory. The transient analysis is run for 0.008 seconds with
a 1.69 10-6 critical time increment (~4733 solution increments). Figure 9.1.36 shows the time
history of axial displacement (U3) for the center nodes of the endcaps. These curves clearly show the
periodic response associated with a dominant axially directed mode of the cylinderendcap structure.
Figure 9.1.37 shows the 1-direction translation (U1) of the endcap center nodes. The 1-direction is
also the primary direction of shock wave propagation. The response curves clearly illustrate that there is
a rigid body translation of the cylinder, and the oscillations are representative of the fundamental beam
bending mode of the cylinder. Figure 9.1.38 shows the time history of vertical (U2) displacement
for nodes located at the top and bottom midplane of the test cylinder. These curves suggest that a
dominant N=2 ovalization mode of vibration occurs at about 170 Hz (based on an estimated period
of 0.0059 seconds). The frequency for the rst ovalization mode of the test cylinder in a vacuum is
330 Hz, based upon an Abaqus/Standard eigenvalue extraction analysis. This shift in the N=2 response
mode frequency illustrates the added mass effect of the external uid on the response of the submerged
cylinder.
Figure 9.1.39 through Figure 9.1.311 contain time history plots of the test cylinder strains
obtained from the Abaqus/Explicit analysis with experimental data for locations B1, C1, and A2. The
experimental curves are obtained by digitizing the response plots published by Kwon and Fox. The
digitized curves are shifted to the left by 0.0002 seconds on the time axis to account for an apparent
time differential between the experiment and the Abaqus/Explicit solution. Figure 9.1.39 contains
history plots of the axially directed strains at location B1. The analytical-experimental correlation at an
9.1.34

Printed on:
early time (peak strain prediction) is very good, as is the prediction for the dominant response frequency
of the test cylinder. The predicted strain oscillations at longer times suggest that the modeling of
hydrodynamic drag damping and viscous losses by applying mass damping to the cylinder mesh could
be improved. Figure 9.1.310 contains the history plots for the axially directed strains at location C1.
The initial peak response (high frequency) contained in the Abaqus/Explicit solution is not present in
the experimental data. This may be due to the sampling rate and ltering techniques used to obtain the
data or to high strain gradients being averaged over the effective length of the strain gauge. Otherwise,
the Abaqus/Explicit solution closely tracks the experimental data and provides a conservative estimate
for the peak response. Figure 9.1.311 contains the history plots for the hoop-directed strains at location
A2. As in Figure 9.1.310, the initial peak response (high frequency) contained in the Abaqus/Explicit
solution is not present in the experimental data. Otherwise, the Abaqus/Explicit solution closely tracks
the experimental data. Figure 9.1.39 through Figure 9.1.311 indicate that the overall UNDEX analysis
model provides a conservative estimate of the cylinders peak response and is, therefore, appropriate
for meeting the analysis objective.
Figure 9.1.312 shows a contour plot of accumulated equivalent plastic strain (PEEQ) on the outer
surface of the test cylinder. To obtain the plot, an averaging threshold of 100% is used and the maximum
contour value is specied as 9.16E3. The plot corresponds to the end of the transient analysis, which
is well after the last increment of plastic strain is detected from a plot of the cylinders total plastic
strain energy vs. solution time. The slight degree of solution nonsymmetry exhibited about the cylinders
midplane is due to the nonsymmetric nature of the free tetrahedron acoustic element mesh of the external
uid.
Input files
submerged_cyl_driver.inp
submerged_cyl_cylinder.inp
submerged_cyl_water.inp
submerged_cyl_pulse.inp
Abaqus/Explicit analysis of a submerged cylinder

subjected to an UNDEX shock wave.
The nite element mesh data for the test cylinder,
including element and node set denitions for output
requests.
The nite element mesh data for the external water,
including element and node set denitions for surface
creation and output requests.
The time history of the shock wave pressure at the
standoff point dened by the *AMPLITUDE option.
References
Kwon, Y. W., and P. K. Fox, Underwater Shock Response of a Cylinder Subjected to a Side-On
Explosion, Computers and Structures, Vol. 48, No. 4, 1993.
Blevins, R. D., Formulas for Natural Frequencies and Mode Shapes, Robert E. Fruger Publishing
Co., 1979.
9.1.35

Printed on:
Table 9.1.31 Added mass for N=1 translation mode of an innite

cylinder (uid between concentric cylinders).
Cylinder Radius
Ratio (
)
Added Mass Ratio (External

Boundary/Infinite Domain)
1.5
2.600
2.0
1.667
4.0
1.133
6.0
1.057
8.0
1.032
16.0
1.008
24.0
1.004
9.1.36

Printed on:
Figure 9.1.31 Strain gauge locations (A1, A2, B1, B2, B3,
C1, C2) with B1 closest to the charge.
Figure 9.1.32
Incident pressure wave transient (shock pulse).
9.1.37

Printed on:
Figure 9.1.33
Figure 9.1.34
Test cylinder model.
Test cylinder and external uid acoustic mesh.
9.1.38

Printed on:
Figure 9.1.35
Displacement(Meters)
External uid surface mesh at the acoustic-structure wetted interface.
Time(Seconds)
Figure 9.1.36
Axially directed displacements (U3) at the center of the endcaps.
9.1.39

Printed on:
Time(Seconds)
Displacements at the center of the endcaps.
Figure 9.1.37
Time(Seconds)
Figure 9.1.38
Vertical (U2) displacements at the cylinder midplane (top and bottom).
9.1.310

Printed on:
Strain
(Abaqus/Explicit)
Time(Seconds)
Figure 9.1.39
Axially directed strains at location B1.
Strain
(Abaqus/Explicit)
Time(Seconds)
Figure 9.1.310
Axially directed strains at location C1.
9.1.311

Printed on:
Strain
(Abaqus/Explicit)
Time(Seconds)
Figure 9.1.311
Hoop-directed strains at location A2.
PEEQ
SPOS, (fraction = 1.0)
(Ave. Crit.: 100%)
+9.160e-03
+8.142e-03
+7.124e-03
+6.107e-03
+5.089e-03
+4.071e-03
+3.053e-03
+2.036e-03
+1.018e-03
+0.000e+00
T
Z
Figure 9.1.312
Accumulated equivalent plastic strains (PEEQ).
9.1.312

Printed on:
SHOCK ANALYSIS CONVERGENCE STUDIES
9.1.4
CONVERGENCE STUDIES FOR SHOCK ANALYSES USING SHELL ELEMENTS
Structures subject to shock loads require careful design because of the high-frequency content of the loading
signal and the low-pass ltering property of a nite element mesh. To address these issues, a simplied model
is useful to estimate optimal parameters for an accurate, efcient, and realistic model.
The goal of the present study is to propose a methodology for designing a computationally efcient
structural model, by answering the following questions:
Given a load signal

accurate response?
Given a minimum possible element size, what is the best signal

to
with reasonable accuracy?
, what is the maximum element size for a given structure to obtain a reasonably
to emulate the structural response
Model description
For the purpose of this study, consider a cylinder made of a linear elastic material with E=2.11 1011 N/m2
and =0.3. It has an overall length of 10 m, a diameter of 3.5 m, and a wall thickness of 0.04 m. Quarter
symmetry is employed by applying the appropriate boundary conditions on the corresponding geometry
edges. The model is shown in Figure 9.1.41. Incident wave loading is applied on the outer surface of
the cylinder, using the Kwon & Fox (KF) signal described in Response of a submerged cylinder to an
underwater explosion shock wave, Section 9.1.3. Abaqus/Explicit is used with a xed time increment
of 1 106 s, chosen to provide a suitable integration scheme for the load. The transient response (U, V,
A) is analyzed at a standoff point over a period of 6 ms. Uniform meshes of S4R elements with various
sizes are employed.
First, convergence of the structural response at the standoff point is examined, applying the original KF
loading signal. Errors are reported in a -norm over the entire 6 ms duration of the response. The
converged solution is formed by successively applying ner meshes until a desired relative accuracy is
obtained. From Figure 9.1.42 and Table 9.1.41 we conclude that a mesh with h=0.025 m is a reference
solution. The solution with h=0.05 m can also be considered converged; therefore, it can become the
reference solution for subsequent studies (the acceleration response differs by only 1.1% with respect to
the solution with h=0.025 m).
Figure 9.1.43 and Figure 9.1.44 depict the relative accuracy of the displacement and velocity
responses with respect to the solution with h=0.05 m. For example, Figure 9.1.43 shows that if an
accurate velocity response is of interest, element size restrictions can be relaxed to h=0.25 m (6.3%
relative error versus the reference conguration); if the displacement response is of interest, the element
size restrictions can be relaxed further to
m (Figure 9.1.44).
9.1.41

Printed on:
Figure 9.1.45 is constructed by dening a mesh characteristic frequency,

, used as x-data for
the error curve. For a given number of elements, N, per excitation wavelength, the mesh characteristic
frequency is
where
is the characteristic wave speed (here the shell exural wave speed) and h is the element
diameter. The structural wave speed is dispersive; i.e., it varies with the excitation frequency. In
Figure 9.1.45, however, a constant reference value of =1200 m/s is used. The previously dened
mesh characteristic frequency also depends on the users choice of the number of elements per
wavelength, N. To preserve accuracy and to address the pollution effect (Ihlenburg and Babuska, 1995),
N should increase with the excitation frequency. The choice of h, N, and the wavespeed in the medium,
, determines the value of the mesh characteristic frequency,
, which can be interpreted as a cutoff
frequency for designing the ltered signal.
Figure 9.1.45 shows that by considering a mesh with h=0.05 m and N=15, the mesh characteristic
frequency becomes
=1600 Hz. If the number of elements per wavelength is relaxed to N=5,
will
increase to 4800 Hz. If the latter value is applied as the cutoff frequency, the ltered signal will capture
the true signal more accurately; however, it is likely to produce more noise if the waves are not resolved
properly.
The study presented above may lead to meshes that are impractical for realistic models, especially
if the acceleration response is of primary interest. For this model of 10 m length, an element of 0.05 m
provides sufcient accuracy at the standoff point. Since this may lead to prohibitive computational costs
for realistic models, the next studies address ways to relax the element size restraints by modifying
the shock signal applied to the structure. These following studies examine the impact of ltering and
modied rise time upon the element size restrictions, in an attempt to minimize the noise at coarse meshes.
Filtered signals are applied successively to the structure, in an attempt to quantify the dependence
of the optimal cutoff frequency on the element size. The sine-Butterworth second-order lter is used.
Although lter performance is very important for optimal results, a comparison of various ltering
techniques is beyond the purpose of this study.
The ltered signals are postprocessed via linear scaling with the parameter a such that the nal and
original signals are equivalent in a n-norm:
The norm can be varied to design a nal signal suitable for various responses: since initially the KF is
applied as a load signal, the innity norm is suitable for obtaining a more accurate acceleration response,
while the 1-norm (impulse conservation) is suitable for obtaining a more accurate velocity response.
For this study, ltered signals are designed with scaling parameters suitable for acceleration
response. Sample postprocessed ltered signals are presented in Figure 9.1.46. By applying these as
incident wave loads on the mesh with h=0.05 m, the results shown in Figure 9.1.47, Figure 9.1.48,
and Figure 9.1.49 are obtained for the acceleration, velocity, and displacement, respectively. There is a
9.1.42

Printed on:
clear distinction between Figure 9.1.47, a well-captured acceleration response, and Figure 9.1.48 and
Figure 9.1.49, which are poor estimates of the velocity and displacement responses, since the applied
ltered load has a signicantly higher impulse than the unltered signal.
The results presented in Figure 9.1.410 and Figure 9.1.411 are obtained by applying the same
load signals to coarser meshes. A summary is presented in Table 9.1.41 and Figure 9.1.414. The
expected trend is visible: as the mesh coarsens, the acceleration response when applying ltered signals
is considerably less noisy than the response for unltered signals. As the mesh is rened, the response
converges to the wrong solution given by the lter of choice. Thus, when computational resources
are scarce, there is an obvious benet to applying preltered signals, which attenuate the noise due to
insufcient spatial discretization.
As an alternative to the sine-Butterworth lter, the signal can be idealized in a heuristic manner by
using a linear rise, followed by an exponential decay. By keeping the decay constant, you can study the
sensitivity of the structural element size to the rise time of the shock signal.
The original rise time, , of the KF signal is multiplied by a rise time factor ( = 2, 5, 10, 15, 20)
to obtain new signals, which are then applied to the structure. To account for noise contributions only,
the error of the acceleration response is reported on the decay portion [
], where
is the time
stamp of the peak amplitude of the load signal.
The results using this technique are presented in Figure 9.1.412 and Figure 9.1.413 and
summarized in Table 9.1.42 and Figure 9.1.415. By comparison with Figure 9.1.414, the smoothing
strategy appears considerably more effective for the acceleration response than using a sine-Butterworth
lter. For example, the case with h=0.125 m and
=20 (scaling the upper bound of the spectrum at
~2250 Hz) yields an error of 9%; for the same element size with h=0.125 m, the sine-Butterworth lter
with a cutoff frequency of 3000 Hz yields a 32% error.
Input files
step_data.inp
cyl_h0025.inp
cyl_h005.inp
cyl_h0125.inp
cyl_h025.inp
cyl_h05.inp
sf_unlt.inp
sf_300.inp
sf_600.inp
sf_900.inp
sf_1200.inp
sf_1500.inp
Step data for all models.

Model data for submerged cylinder with an element size
of 0.025 m.
of 0.05 m.
of 0.125 m.
of 0.25 m.
of 0.5 m.
Amplitude data for unltered KF signal.
Amplitude data for KF signal ltered at 300 Hz.
9.1.43

Printed on:
sf_2000.inp
sf_3000.inp
sf_5000.inp
sr_2.inp
sr_5.inp
sr_10.inp
sr_15.inp
sr_20.inp
driver_h0025.inp
driver_h005.inp
driver_h0125.inp
driver_h025.inp
driver_h05.inp

Amplitude data for KF signal with a rise time factor of 2.
Driver le for the mesh with h=0.025 m and all signals.
References
Ihlenburg, F., and I. Babuska, Finite Element Solution of the Helmholtz Equation with High
Wave Numbers. Part 1: The h-version of the FEM, Computers & Mathematics with Applications,
no. 30(9), pp. 937, 1995.
Kwon, K. W., and P. K. Fox, Underwater Shock Response of a Cylinder Subjected to a Side-On
Explosion, Computers and Structures, vol. 48, no. 4, 1993.
Table 9.1.41 Percent relative error in acceleration response for pre-ltered signals using
sine-Butterworth lter. Reference solution is the unltered load signal applied to a mesh with h=0.025 m.
Cutoff (Hz)
h=0.05 m
h=0.125 m h=0.25 m
h=0.5 m
300
142
105
105
104
600
116
102
102
102
900
91
90
95
90
1200
72
75
88
90
1500
64
64
85
101
2000
49
49
47
110
3000
33
32
43
129
5000
16
16
32
116
Unltered
1.2
18
91
103
9.1.44

Printed on:
Table 9.1.42 Percent relative error in acceleration response for smoothed rise time signals.
Reference solution is the unltered load signal applied to a mesh with h=0.025 m. Errors are measured
in -norm over
, where
is the time stamp of the peak amplitude.
Rise time factor
h=0.05 m
h=0.125 m h=0.25 m
h=0.5 m
0.7
12
67
93
2.1
4.0
55
93
10
4.2
4.4
20
85
15
14
70
20
20
44
standoff
point
1
3
Figure 9.1.41
Test problem geometry.
9.1.45

Printed on:
h=0.025
h=0.05
h=0.125
h=0.25
h=0.5
(base)
(error=1.1%)
(error=19.6%)
(error=93.6%)
(error=117%)
Figure 9.1.42 Convergence of athwartship acceleration

at standoff point. Element size h [m].
h=0.05
h=0.125
h=0.25
h=0.5
(base)
(error=2.3%)
(error=6.3%)
(error=11.7%)
Figure 9.1.43 Convergence of athwartship velocity at

standoff point. Element size h [m].
9.1.46

Printed on:
h=0.05
h=0.125
h=0.25
h=0.5
(base)
(error=0.1%)
(error=0.3%)
(error=1.5%)
Figure 9.1.44 Convergence of athwartship displacement

at standoff point. Element size h [m].
acceleration;
acceleration;
velocity;
velocity;
displacement;
displacement;
N=15;
N=5;
N=15;
N=5;
N=15;
N=5;
c=1200
c=1200
c=1200
c=1200
c=1200
c=1200
Figure 9.1.45 Shock cutoff frequency based on structure

response. Wave speed c [m/s].
9.1.47

Printed on:
300 Hz
600 Hz
900 Hz
1200 Hz
1500 Hz
Unfiltered
Figure 9.1.46 Sample signals obtained after applying

sine-Butterworth lter to the original KF signal.
300 Hz
600 Hz
1500 Hz
3000 Hz
5000 Hz
Unfiltered
Figure 9.1.47 Acceleration response at standoff point for

ltered load signals and mesh with h=0.05 m.
9.1.48

Printed on:
300 Hz
600 Hz
1500 Hz
3000 Hz
5000 Hz
Unfiltered
Figure 9.1.48 Velocity response at standoff point for ltered

load signals and mesh with h=0.05 m.
300 Hz
600 Hz
1500 Hz
3000 Hz
5000 Hz
Unfiltered
Figure 9.1.49 Displacement response at standoff point for

9.1.49

Printed on:
300 Hz
600 Hz
1500 Hz
3000 Hz
5000 Hz
Unfiltered

300 Hz
600 Hz
1500 Hz
3000 Hz
5000 Hz
Unfiltered

9.1.410

Printed on:
rtf=2
rtf=5
rtf=10
rtf=15
rtf=20
Original

smoothed load signals and mesh with h=0.125 m.
rtf=2
rtf=5
rtf=10
rtf=15
rtf=20
Original

smoothed load signals and mesh with h=0.25 m.
9.1.411

Printed on:
Figure 9.1.414 Error in acceleration response for sine-Butterworth lter with

various cutoff frequencies. Element size h [m].
Figure 9.1.415
Error in acceleration response for various rise time factors. Element size h [m].
9.1.412

Printed on:
SUBMARINE UNDEX ANALYSIS
9.1.5
UNDEX ANALYSIS OF A DETAILED SUBMARINE MODEL
Modeling large underwater structures subject to shock loads typically leads to computationally intensive
numerical models that require considerable computing resources.
This example illustrates how
Abaqus/Explicit can be used to predict the transient response of such a large and complicated structure
subject to a shock wave loading resulting from an underwater explosion (UNDEX). A publicly available full
submarine model is modied via a modeling approach that minimizes the computational costs and obtains an
accurate response in a particular region of interest. Consequently, particular attention is given to structural
details in the respective region of interest, while simplications meant to reduce analysis costs are assumed
elsewhere.
The structure is loaded by an incident wave with a shock prole amplitude. If structural integrity is
analyzed, a reasonable assumption is that most affected regions will be around the standoff point; thus,
more attention to modeling details should be paid to the front part of the submarine (Figure 9.1.51 and
Figure 9.1.55).
The model presented was created based on specications provided by Forschungsanstalt der Bunderwehr
fr Wasserschall und Geophysik (FWG), Kiel, Germany (Fiedler and Schneider, 2002). This model is
known as the Benchmark Target Strength Simulation (BeTSSi) model and is a reasonably sophisticated
replica of a Kilo-class Soviet submarine. The complexity of the structure is considered suitable for testing
target-strength simulation codes for more realistic problems (Schneider et al., 2003). In this example
problem the BeTSSi model is adapted to test Abaqus/Explicit for a realistic UNDEX simulation.
Structural details include ooded bow compartments, sonar array, torpedo tubes, ooded sail
compartments, sail access tubes, and an aft chamber (Figure 9.1.51 through Figure 9.1.54). The
pressure hull is enveloped by an external hull only along the upper side and is in direct contact with the
uid throughout the bottom side. The intra-hull space is also ooded along the length of the submarine.
The entire structure is hit by a spherical shock wave due to the explosion of a charge located 16.5 m
away from the submarine (Figure 9.1.51). The pressure-time signal corresponds to a 60lb HBX-1
charge, as in Response of a submerged cylinder to an underwater explosion shock wave, Section 9.1.3,
and taken from Kwon & Fox (1993).
Reduced model
Since the Kwon & Fox (KF) load signal carries a large excitation spectrum, the outside water has to
extend to a large distance, corresponding to the lowest end of spectrum, and both the submarine structure
and the outside water have to be discretized with a small element size, corresponding to the wavelength
of the high end of the spectrum. This generates a considerably large computational model, due also to
the signicant difference between the length of the submarine model (62 m) and the low wavelength of
the high end of the spectrum. Consequently, a modeling approach is employed in this example problem
9.1.51

Printed on:
where a region of interest is dened around the shock-wave standoff point. The reduced model still
includes details such as ooded bow compartments, torpedo tubes, a sonar dome, and a portion of the
pressure and exterior hulls, with the uid domain dened in the intra-hull space. In addition, stiffeners
are added for the detailed region of the pressure hull. The rest of the submarine model is simplied using
beam elements, coupled with the region of interest using kinematic coupling constraints. The section
behavior of the beams is dened using meshed cross-sections to approximate the cross-section inertia of
the real structure (Figure 9.1.56, Figure 9.1.57, and Figure 9.1.58). The tapered region at the back
of the submarine is approximated via 13 stepwise-constant circular cross-sections. Finally, the inertia
effects of the outside water are included.
The exterior acoustic domain (water) envelops only the region of interest and extends to a distance
of approximately 1.5 m from the submarine structure, corresponding to half the wavelength inwater at
a frequency of 500 Hz. Impedance boundary conditions are applied on truncation surfaces to reduce
spurious reections.
Applying the above simplications and element sizes of

for the structure and
for
both the inner and outer water domains, the aggregate model size is 1.2 million nodes and 5.1 million
elements, including AC3D4, S4R, S3R, and B31 elements. The analysis is stopped after 6 ms, enough
for the wavefront traveling through the acoustic medium to sweep the region of interest.
The pressure for the inner and outer acoustic domains is depicted in Figure 9.1.59 at the end of the
6 ms response. Observe the high-frequency content of the acoustic pressure inside the bulkheads, due to
the excitation from the walls. The high-frequency content of the waves in the bulkheads walls can be
seen in Figure 9.1.513 and Figure 9.1.514.
Figure 9.1.510, Figure 9.1.511, and Figure 9.1.512 illustrate the history of acceleration, velocity,
and displacement responses at the standoff point, respectively. A kick-off acceleration of almost 8 km/s2
is reached due to the initial shock, followed by a rapid decay. By the end of the shock duration (4 ms),
it is almost stabilized around zero. The kick-off velocity reaches a value of almost 8 m/s, while the
athwartship displacement shows a maximum drift of about 10 mm.
Structural integrity is one of the major concerns of such an analysis. Almost the entire structure is
assumed to be made of perfect elastic-plastic steel plates, with position-dependent thicknesses. The only
exceptions are the stiffeners, dened as beams with T cross-sections. The material is the same as for the
rest of the submarine model, a perfect elastic-plastic steel. High stress concentration regions are analyzed
by plotting Mises stresses (Figure 9.1.515); permanent deformations are monitored via the equivalent
plastic strain (Figure 9.1.516). High stress concentration regions as well as permanent deformations are
observed for regions around the joints. The maximum stresses are observed on the upper panels of the
sonar dome, while the largest permanent deformation is obtained around the joint between the pressure
hull, the exterior hull, and the horizontal bow compartment.
Input files
undex_driver_xpl.inp
undex_parts.inp
Driver input le.

Parts denition input le.
9.1.52

Printed on:
undex_assembly.inp
undex_outwater_h01.inp
undex_innerwater_h01.inp
undex_subbody_h005.inp
undex_tapered_beam_elsets.inp
undex_tapered_beam_sections.inp
undex_ampl.inp
undex_acoustics_s.inp
undex_materials_s.inp
undex_ties.inp
undex_step.inp
undex_boundary_conditions.inp
undex_output_requests.inp
undex_beam_section_front.inp
undex_beam_section_sail.inp
undex_beam_section_back1.inp
Assembly input le.

Outer water mesh data.
Inner water mesh data.
Submarine structure mesh data.
Element set denitions for the tapered region of the
submarine body.
Section data for the element sets dened for the tapered
region of the submarine body.
Amplitude data.
Impedance and incident wave model data.
Material data.
Tie denitions.
Step data.
Boundary condition data.
Output request data.
Section mesh data for generating cross-section beam
properties for the front region of the submarine.
properties for the sail region of the submarine.
properties for the 1st back region.
properties for the 2nd back region.
properties for the 3rd back region.
properties for the 4th back region.
9.1.53

Printed on:

References
Fiedler, Ch., and H. G. Schneider, BeTSSi-SubBenchmark Target Strength Simulation

Submarine, Technical Report, Forschungsanstalt der Bundeswehr fr Wasserschall und
Geophysik, Kiel, 2002.
Kwon, K. W., and P. K. Fox, Underwater Shock Response of a Cylinder Subjected to a Side-On
Explosion, Computers and Structures, vol. 48, no. 4, 1993.
Schneider, H. G. et al., Acoustic Scattering by a Submarine: Results from a Benchmark Target

Strength Simulation Workshop, Proceedings of Tenth International Congress on Sound and
Vibration, Stockholm, Sweden, 2003.
standoff point
*
source
point
Figure 9.1.51
BeTSSi full submarine model.
9.1.54

Printed on:
Figure 9.1.52 Cut view through front part of the model including
details such as bulkheads, torpedo tubes, and the sonar dome.
Figure 9.1.53
Cut view through sail; sail compartments and access tubes are visible.
9.1.55

Printed on:
Figure 9.1.54
Figure 9.1.55
Flooded bow compartments.
Cut view of reduced model, including exterior uid domain.
9.1.56

Printed on:
Figure 9.1.56
Meshed cross-section of the front beams.
Figure 9.1.57
Meshed cross-section of the sail beams.
Figure 9.1.58
Meshed cross-section of the back beams.
9.1.57

Printed on:
POR
+3.506e+06
+2.936e+06
+2.366e+06
+1.796e+06
+1.226e+06
+6.554e+05
+8.537e+04
-4.847e+05
-1.055e+06
-1.625e+06
-2.195e+06
-2.765e+06
-1.293e+07
Figure 9.1.59
Pore pressure contours of the inner and outer water.
Figure 9.1.510
Athwartship acceleration at the standoff point.
9.1.58

Printed on:
Figure 9.1.511
Figure 9.1.512
Athwartship velocity at the standoff point.
Athwartship displacement at the standoff point.
9.1.59

Printed on:
Figure 9.1.513
Cut view of deformed conguration.
Figure 9.1.514 Deformed conguration of pressure hull,

bulkheads, torpedo tubes, and sonar dome: whole view.
9.1.510

Printed on:
S, Mises
(Ave. Crit.: 75%)
+6.024e+08
+5.523e+08
+5.022e+08
+4.522e+08
+4.021e+08
+3.521e+08
+3.020e+08
+2.519e+08
+2.019e+08
+1.518e+08
+1.017e+08
+5.166e+07
+1.593e+06
Figure 9.1.515 Mises stress for bow compartments, pressure

hull, torpedo tubes, and sonar array.
PEEQ
(Ave. Crit.: 75%)
+2.580e-02
+1.000e-03
+9.250e-04
+8.500e-04
+7.750e-04
+7.000e-04
+6.250e-04
+5.500e-04
+4.750e-04
+4.000e-04
+3.250e-04
+2.500e-04
+1.750e-04
+1.000e-04
+0.000e+00
Figure 9.1.516 Equivalent plastic strain for bow compartments,

pressure hull, torpedo tubes, and sonar array.
9.1.511

Printed on:
COUPLED ACOUSTIC-STRUCTURAL ANALYSIS OF A PICK-UP TRUCK
9.1.6
This example illustrates the capability in Abaqus to perform fully coupled acoustic-structural analyses of a
pick-up truck model. This type of analysis has become critically important in the automotive industry and it
provides essential benets toward designing vehicles for ride comfort and quietness. This example uses the
pick-up truck model geometry described in Inertia relief in a pick-up truck, Section 3.2.1. Only a portion of
the pick-up truck is modeled, including the chassis, the cabin, and the air inside the cabin. Structural elements
are used to model the cabin and the chassis, and acoustic elements are used to model the air interior. Connector
elements are used to connect the various structural parts together. The coupling between the structure and
acoustic medium is modeled by applying the surface-based *TIE option. Frequency domain analyses are
performed for both regular models (without substructures) and for models using coupled structural-acoustic
substructures.
The pick-up truck model (1994 Chevrolet C1500) discussed here is depicted in Figure 9.1.61 through
Figure 9.1.63.
The air is modeled only inside the cabin, and the process of constructing the air mesh in this
particular case is worth a brief discussion. Normally, if the solid geometry of the structural part (the
cabin) were available, a Boolean subtraction could be performed in Abaqus/CAE to obtain the solid
geometry of the included space (in this case the air inside the cabin). Unfortunately, the solid geometry
of the cabin is not available since the structural model is based on a public-domain mesh, as discussed
in Inertia relief in a pick-up truck, Section 3.2.1. To overcome this issue, the following strategy is
adopted. An approximate air geometry is created in Abaqus/CAE to follow roughly the contour of the
cabin interior, including the dashboard, the doors, the cabin oor, the cabin top, the back wall, and
the seat. The mesh created from this geometry does not conform exactly to the geometry of the cabin
interior. However, by using the ADJUST parameter on the surface-based *TIE constraint that connects
the air mesh to the structural parts, the nodes belonging to the air surface will be pushed onto the inside
cabin surface or onto the seat surface to conform to the structural mesh (see Figure 9.1.62).
The materials used for the cabin and chassis are described in Inertia relief in a pick-up truck,
Section 3.2.1. The air properties used inside the cabin are: air density of 1.2 kg/m3 and air bulk modulus
of 1.39 105 Pa, which produce a sound speed of 340 m/s.
Models both with and without damping are constructed. In the models where damping is considered,
two forms of damping are modeled, as follows. The *DAMPING option is used to introduce Rayleigh
stiffness proportional damping, governed by the parameter , into the structural materials in the model.
For a given value of applied to all materials in the structure (mostly steel), the damping fraction for
a mode with natural circular frequency
is given by the formula
. The value of in
the model is chosen to give approximately 1% critical damping for the modes whose natural frequencies
are in the middle of the range of excitation (at about 80 Hz). Surface impedance is also specied on
the cabin oor to model the acoustic damping effect of a carpet by using the *SIMPEDANCE and the
9.1.61

Printed on:
*IMPEDANCE PROPERTY options. The impedance properties on this surface are chosen such that
about 80% of a planar wave incident to this surface in the normal direction would be reected.
Models
Four different models are considered.

Model 1
In the rst model only the cabin and the interior air are considered. The structural part of the nite
element model is shown in Figure 9.1.61 with the doors removed for illustration purposes. The cabin
has six connection points with the chassis: two in the front and four under the seat. The outermost four
of these connection points are xed with boundary conditions, and a natural frequency extraction is
performed. Several *STEADY STATE DYNAMICS analyses (mode-based, DIRECT, and SUBSPACE
PROJECTION) follow the *FREQUENCY step. Separate analyses with and without damping are
conducted. There are 43,663 structural elements (mostly shells) and 12,171 acoustic elements in this
model for a total of 207,994 degrees of freedom. The average structural element size is about 90 mm,
and the average acoustic element size is approximately 325 mm. Considering that at least 56 elements
are needed per wavelength for accurate representation of the dynamics, the highest excitation frequency
for which results can be computed accurately is about 175 Hz. Two excitation cases are considered:
harmonic point loading (at the two hook-up points that have not been constrained) and incident wave
loading on the bulkhead below the dashboard. The latter case models engine compartment noise
propagating through an air path. This airborne load is modeled both as an incident plane wave and as a
diffuse eld.
Model 2
In the second model the cabin-air model is reduced to a fully coupled structural-acoustic substructure.
The four connection points where boundary conditions are applied in the rst model are retained using
the *RETAINED NODAL DOFS option. In addition, 200 coupled structural-acoustic eigenmodes are
extracted. The eigenmodes are then retained using the *SELECT EIGENMODES option to better
represent the dynamics of the substructure in the frequency range of interest. Consequently, the
substructure is represented by a total of 224 degrees of freedom to represent the 207,994 degrees of
freedom in the rst model. The substructure is then used in a separate, one-element natural frequency
extraction analysis. The results are recovered from the substructure and compared to the results obtained
from the rst model.
Model 3
In the third model both the cabin and the chassis are considered on the structural side (Figure 9.1.63),
while the air is modeled inside the cabin only. Since no air mesh is used to model the ambient air, this
type of model can be used to study the structural path contributions to the noise inside the cabin. In the
*STEADY STATE DYNAMICS analyses of this model the excitation is provided by point loads applied
to the engine mounts, while the chassis is supported with xed boundary conditions at its ends. There
9.1.62

Printed on:
are 53,897 structural elements in this model, while the number of acoustic elements is the same as in the
rst model.
Model 4
Finally, the cabin-chassis model is reduced to two substructures: a fully coupled structural acoustic cabinair substructure and a structural-only chassis substructure. As in the previous substructure model both the
*RETAINED NODAL DOFS and the *SELECT EIGENMODES options are used to generate the two
substructures. The two substructures are represented by 236 and 284 degrees of freedom, respectively.
The substructures are then used in a separate, two-element natural frequency extraction analysis; and the
results are compared to the results from the third model.
Some of the *STEADY STATE DYNAMICS results for the undamped cabin-air model analyses are
shown in Figure 9.1.64 and Figure 9.1.65. In all analyses 200 sampling points are selected in the
frequency range of interest (35120 Hz) using the TYPE=RANGE option. This frequency range
corresponds to engine-induced vibrations in the range of 21007200 RPM. Of particular interest in
these analyses is the sound pressure level at a location in the vicinity of the drivers ear. The response is
shown in Figure 9.1.64 and is calculated from the acoustic pressure using the following equation
where
and
Pa. Figure 9.1.65 shows the displacement response of
one of the nodes on the cabin oor where the harmonic load is applied. The results from the subspacebased and the mode-based steady-state dynamic analyses are virtually identical (as expected), and they
compare quite well with the results from the direct steady-state dynamics analysis (Figure 9.1.64 and
Figure 9.1.65). The sound pressure level as computed in these analyses is very high since no damping
is considered (neither structural nor impedance-like at the structural-acoustic interface).
Figure 9.1.66 shows the noise level for the cabin-air model when damping is considered. Since the
mode-based steady-state dynamics analysis would not take into account the forms of damping considered
here, only results from the subspace projection and direct analyses are computed. The results compare
quite well. Notably, the subspace projection analysis is approximately 20 times faster than the direct
analysis. While the sound pressure level is signicantly lower in this analysis when compared to the
analysis with no damping, the level is still very high. This suggests that the damping considered in this
model is still quite low. Impedance-type damping is considered only on the cabin oor; thus, 100% of
the acoustic waves would be reected from the cabin walls, roof, doors, and windows to produce a higher
sound pressure level.
Figure 9.1.67 shows the real part of the acoustic pressure in the cabin volume at 120 Hz for both
the plane wave (left) and diffuse eld (right) excitations. The same damping is used as in the previous
casean impedance dened on the cabin oor. The same source location, a point on the radiator, and the
same standoff, a point on the bulkhead, are used for both excitations. These gures illustrate the typical
effect of the diffuse excitation optionit produces a more even distribution of pressure in the cabin. This
occurs because the incident pressure is divided into many waves, striking the bulkhead from different
9.1.63

Printed on:
directions and resulting in a degree of cancellation and averaging of the pressure load on the surface.
The results shown are obtained using the direct steady-state dynamics procedure; using the subspace
projection steady-state dynamics procedure, the results appear nearly identical.
The frequency analysis performed on the cabin-air substructure generates eigenvalues identical to
those from the model without substructures. Moreover, the eigenmodes obtained from the regular nonsubstructure model (Figure 9.1.68) and those recovered from the substructure model (Figure 9.1.69)
compare very well (shown here for the air pressure for the 25th eigenmode).
The frequency response obtained for the cabin-air-chassis model is shown in Figure 9.1.610 and
Figure 9.1.611. Given the size of the model, the direct steady-state dynamics analysis is computationally
less efcient and, thus, is not performed. In addition to the natural frequency extraction procedure for
the whole structure, a frequency analysis is performed on the equivalent cabin-air-chassis model using
two substructures. While the eigenfrequencies are not identical to those obtained from the regular nonsubstructure model, the differences are quite small for the range of interest, as shown in Figure 9.1.612.
Once the substructures are generated, the analysis to extract eigenfrequencies from the two-element
substructure model is hundreds of times faster than the analysis to extract them from the regular nonsubstructure model.
The mode-based and the subspace projection steady-state dynamics procedures in Abaqus
demonstrate signicant improvements in computational efciency when compared to the direct
steady-state dynamics approach. When damping is small or if it can be well approximated using modal
damping coefcients, the mode-based procedures are extremely efcient. When damping is more
complex, the subspace projection method also demonstrates signicant computational advantage in
comparison with the direct-integration approach.
The use of substructures is also demonstrated to produce signicant gains in computational
efciency. The reduction of the acoustic volume and of its bounding structure to a substructure has clear
advantages. The low-dimensional coupled acoustic-structural substructures are very computationally
efcient, and the data for the acoustic response inside the substructure can be recovered when the global
analysis is completed.
Input files
tr_acous_cabin_mode.inp
tr_acous_cabin_mode_ams.inp
tr_acous_cabin_direct.inp
tr_acous_cabin_sp_impedance.inp
tr_acous_cabin_direct_impedance.inp
tr_acous_cabin_subspace_bulkhead.inp
Mode-based and subspace projection steady-state

dynamic analysis of the cabin-air model without damping.
Mode-based steady-state dynamic analysis of the cabinair model without damping and using Abaqus/AMS.
Direct steady-state dynamic analysis of the cabin-air
model without damping.
Subspace projection steady-state dynamic analysis of the
cabin-air model with damping.
model with damping.
Subspace projection steady-state dynamic analysis of
the cabin-air model with damping and incident wave
excitation on the bulkhead.
9.1.64

Printed on:
tr_acous_cabin_direct_bulkhead.inp
tr_acous_cabin_gen.inp
tr_acous_cabin_sub_freq.inp
tr_acous_cabin_chassis_gen.inp
tr_acous_chassis_gen.inp
tr_acous_cabin_chassis_sub.inp
tr_materials_acous.inp
tr_cabin_air_w.inp
tr_acous_chassis_coup.inp
tr_cabin_elements.inp
tr_cabin_elsets.inp
tr_cabin_nodes.inp
tr_cabin_nsets.inp
tr_cabin_sections.inp
tr_parameters_inphase.inp
tr_parameters.inp
tr_all_nodes.inp
tr_cabin_elts.inp
tr_cabin_coup_steer_col.inp
tr_cabin_mpc.inp
tr_seat_elts.inp
tr_seat_coup.inp
tr_conn_seat.inp
tr_door_left_elts.inp
tr_door_left_coup.inp
tr_door_left_mpc.inp
tr_conn_door_left_nobehav.inp
tr_door_right_elts.inp
tr_door_right_coup.inp
tr_door_right_mpc.inp
tr_chassis_elts.inp

model with damping and incident wave excitation on the
bulkhead.
Cabin-air coupled substructure generation for the second
model.
Frequency analysis of the cabin-air model using one
substructure.
Cabin-air substructure generation analysis for the fourth
model.
Chassis substructure generation analysis for the fourth
model.
Frequency analysis of the cabin-air-chassis model using
two substructures.
All material denitions.
Interior air model.
Coupling denitions for the chassis.
Element denitions for the cabin.
Element set denitions for the cabin.
Node denitions for the cabin.
Node set denitions for the cabin.
Section denitions for the cabin.
All node denitions.
Element denitions for the cabin.
Coupling denitions.
*MPC denitions for the cabin.
Element denitions for the seat.
Coupling denitions for the seat.
Connector denitions for the seat.
Element denitions for the left door.
Coupling denitions for the left door.
*MPC denitions for the left door.
Connector denitions for the left door.
Element denitions for the right door.
Coupling denitions for the right door.
*MPC denitions for the right door.
Element denitions for the chassis.
9.1.65

Printed on:
Figure 9.1.61
The cabin model (doors removed for clearer illustration).
Figure 9.1.62
The air mesh inside the cabin.
9.1.66

Printed on:
Figure 9.1.63
The cabin-air-chassis model.
mode based
subspace projection
direct
Figure 9.1.64
Sound pressure level at the ear position for the cabin-air model (no damping).
9.1.67

Printed on:
mode based
subspace projection
direct
Figure 9.1.65
Displacement at one of the harmonically excited cabin oor points (no damping).
subspace projection
direct
Figure 9.1.66
Sound pressure level at the ear position for the cabin-air

model when damping is considered.
9.1.68

Printed on:
Figure 9.1.67 Real part of the acoustic pressure in the cabin-air model when damping
is considered and when the lower bulkhead is excited with an incident pressure eld
(planar wave case, left; diffuse eld case, right).
9.1.69

Printed on:
POR
+2.486e-04
+1.993e-04
+1.501e-04
+1.008e-04
+5.151e-05
+2.236e-06
-4.704e-05
-9.631e-05
-1.456e-04
-1.949e-04
-2.441e-04
-2.934e-04
-3.427e-04
Figure 9.1.68
Air pressure for the 25th eigenmode (35.131 Hz) from the cabin-air model.
POR
+2.486e-04
+1.993e-04
+1.501e-04
+1.008e-04
+5.151e-05
+2.236e-06
-4.704e-05
-9.631e-05
-1.456e-04
-1.949e-04
-2.441e-04
-2.934e-04
-3.427e-04
Figure 9.1.69
Air pressure for the 25th eigenmode (35.131 Hz) from the cabin-air substructure model.
9.1.610

Printed on:
mode based
subspace projection
Figure 9.1.610
Sound pressure level at ear level from the cabin-air-chassis model (no damping).
mode based
subspace projection
Figure 9.1.611
Displacement at one of the cabin oor points for the cabinair-chassis model (no damping).
9.1.611

Printed on:
Figure 9.1.612 Eigenfrequency differences between the cabin-air-chassis model and the
equivalent model with substructures for the range of interest (35120 Hz).
9.1.612

Printed on:
HULL WHIPPING ANALYSIS
9.1.7
LONG-DURATION RESPONSE OF A SUBMERGED CYLINDER TO AN

UNDERWATER EXPLOSION
Abaqus/Explicit
This example demonstrates how Abaqus can be used to predict the long-duration response of submerged
structures that experience loading by a wave resulting from an underwater explosion (UNDEX).
An emphasis on structural dynamic motions naturally leads to the use of beams to model the structure,
rather than solid or shell elements. This class of problem is characterized by structural dynamic motions
at speeds much slower than the acoustic wave speed in the uid, so that the uid can be modeled as an
incompressible medium. The implication of modeling the uid in this manner is to reduce its effect on the
structure to an added mass on the beam.
The spherical pressure waves associated with an UNDEX event are characterized by two distinct phases.
The rst, very short, phase is the initial wave produced by the detonation. It involves a very steep rise to a
characteristic pressure value, followed by a more gradual decay. In the second phase of the loading the gas
produced by the explosive expands to a maximum volume, at which the pressure of the surrounding uid
forces it back upon itself. At some minimum volume the gas and uid system emits another pressure pulse,
and the gas bubble expands again. This process may repeat many times, causing several pressure pulses. As
the gas bubble oscillates, it also acts under the effects of buoyancy, causing an unsteady motion opposite to
the force of gravity.
This problem involves a structural-dynamics model of a submerged submarine under athwartships attack
from an underwater explosion. The ship is 100 m in length and 50 m below the surface; and the detonation
point of the charge is centered along the ships length, 15 m to one side and 15 m below the ship. The
response of the ship to the initial direct and reected shock waves as well as to the rst few bubble pulses
is of interest, so dynamic simulations are carried out to 5 seconds.
Model
The model is made of 100 B31 beam elements, arrayed along a line. Their (uniform) section properties
are dened using the *BEAM GENERAL SECTION option. The structure has an overall length of 50 m.
Point mass elements of 10000 kg are dened at each node to simulate the effect of internal equipment on
the beam structural dynamics. The effect of the entrained uid is simulated using the *BEAM FLUID
INERTIA option, with which a uid mass density of 1025 kg/m3 , an outside radius of 5 m, and a uid
drag coefcient of 1.0 are specied. Structural damping is specied using the *DAMPING option. No
additional acoustic uid elements or absorbing boundary conditions are required.
Fluid-structure coupling and shock wave loading
The loading specication for this problem includes descriptions of the explosive charge, the uid medium
in which the wave propagates to the structure, and the geometry of the charge with respect to the structure.
9.1.71

Printed on:
In Abaqus the time histories of pressure, its derivatives, and the motion of the explosive gas
bubble are dened using the Geers-Hunter model. This model is invoked using the *AMPLITUDE,
DEFINITION=BUBBLE option. Under this option material properties of the explosive, its mass, its
distance from the free surface, and some other control parameters are specied. The data on this option
are used to govern a separate bubble dynamics time integration operation, performed as part of the
preprocessing. Parameters dened on this option do not affect the rest of the analysis. Here a charge of
100 kg is used, with model parameters set to suppress wave loss effects within the bubble simulation.
An initial depth of 65 m is specied: this affects the oscillation of the gas bubble and the duration of the
bubble dynamics, since the solution naturally terminates when the bubble reaches the free surface. In
this analysis, however, the bubble simulation time is cut off at 0.6 seconds. The bubble migration is
dened to be along the z-axis. Default values for the bubble dynamics time integration parameters are
used.
The actual loads on the structure are dened using the *INCIDENT WAVE option and the associated
INCIDENT
WAVE PROPERTY, *INCIDENT WAVE FLUID PROPERTY, and *INCIDENT WAVE
*
REFLECTION options. The *INCIDENT WAVE option denes the distributed time-varying loads
within an analysis step on the structural surface, due to the specied parameters. Only the surface
dened for the beam elements and the reference load magnitude need to be indicated, as is the case for
most distributed loads in Abaqus. The *INCIDENT WAVE REFLECTION option denes any planes
outside the computational domain for the purpose of calculating additional incident wave loads due
to reections. Here a soft (zero total pressure) reecting plane is dened, located 65 m from the
original position of the source and oriented normal to the z-axis. The remaining geometric and physical
parameters dening the load are specied using the *INCIDENT WAVE PROPERTY and *INCIDENT
WAVE FLUID PROPERTY options. The original position of the source point is dened as (50, 15,
15), and the standoff point is dened as (50, 3.536, 3.536). The uid properties, used for propagation
of the wave across the structure, are given as mass density
= 1025 kg/m3 and bulk modulus
=
2.30635 GPa.
The model for this UNDEX example has a total of 606 active degrees of freedom and requires
approximately 15 MB of memory and 267 KB of disk space.
Figure 9.1.71 shows the time history of vertical displacement (V3) for the center node of the
structure using a logarithmic time axis. This curve clearly shows the initial shock-induced velocity
peak, the velocity peak caused by the reected path, and the decaying periodic response associated with
a structural motion after the loading ceases at t = 0.6. The response curve clearly illustrates that there is
a rigid body translation of the cylinder, due to the velocity induced by the initial shocks. The peaks due
to the direct and reected shocks are of the same sign, since the reected wave, with a negative sign, is
traveling in the opposite vertical direction from the direct wave.
Figure 9.1.72 shows the strain along the axis of the beam for the section points oriented along
the 1-direction of the section. The 1-direction is also the y-direction in the global system. The curves
suggest that a dominant mode of vibration occurs at about 2.1 Hz (based on an estimated period of 0.48
seconds). Figure 9.1.73 shows the axial strain for section points oriented in the section 2-direction, or
9.1.72

Printed on:
the global z-direction. Again, the two peaks corresponding to the incident shocks are evident, followed
by the decaying oscillation at roughly 2.1 Hz.
Input files
iw_exa_whip_std.inp
Abaqus/Standard analysis of a submerged cylinder

Abaqus/Explicit analysis of a submerged cylinder
iw_exa_whip_xpl.inp
Reference
Hicks, A. N., The Theory of Explosion Induced Hull Whipping, Naval Construction Research
Establishment, Dunfermline, Fife, Scotland, Report NCRE/R579, March 1972.
V3 N: 51 NSET NOUT
Figure 9.1.71
Vertical velocity at ship midpoint.
9.1.73

Printed on:
E11 E: 51 IP: 1 SP: (1-coordinate = -5.000000 ELSET ELOUT

E11 E: 51 IP: 1 SP: (1-coordinate = 5.000000 ELSET ELOUT
Figure 9.1.72
Strain at horizontal extrema at ship midpoint.
E11 SP: (2-coordinate = -5.000000 ELSET ELOUT)

E11 SP: (2-coordinate = 5.000000 ELSET ELOUT)
Figure 9.1.73
Strain at vertical extrema at ship midpoint.
9.1.74

Printed on:
SANDWICH STRUCTURE UNDER BLAST LOADING
9.1.8
DEFORMATION OF A SANDWICH PLATE UNDER CONWEP BLAST LOADING
Objectives
This example demonstrates the following Abaqus/Explicit features and techniques:
applying CONWEP blast loading,

comparing computational results using Abaqus/Explicit and experimental measurements of
deformation of sandwich structures under blast loads, and
demonstrating a generic example of a nonlinear analysis of a sandwich structure.
Modern structural designs can be used to protect various systems that are vulnerable to large blast
loads. Sandwich structures are a particular class of modern structures that could be explored for such
applications. Dynamic analysis with blast loading using Abaqus/Explicit provides an important analysis
tool that can help engineers and designers build better structures to resist blast loads.
Geometry
Two models are used in this example. One model is the sandwich plate structure described by
Dharmasena et al. (2008) and shown in Figure 9.1.81. The sandwich structure consists of a square
honeycomb core with vertical webs welded to top and bottom plates. The dimensions of the overall
sandwich plate structure are 610 610 61 mm. The sandwich structure lies in the XY plane, while
the blast source is 100 mm vertically above (along the z-direction) the center of the top plate of the
sandwich structure. The top and bottom plates are 5 mm thick, and the square honeycomb core webs
are 0.76 mm thick. The spacing between the honeycomb webs is 30.5 mm measured from the midplane.
The second model is a solid plate, which was chosen for comparison with Dharmasena et al.
Materials
The top and bottom plates and honeycomb core of the sandwich structure and the solid plate are all made
of a high ductility stainless steel alloy (Al-6XN) comprised of 49% Fe, 24% Ni, 21% Cr, and 6% Mo by
weight as described by Nahshon et al. (2007).
Initial conditions
All structures are initially at their undeformed state at 273 K temperature.

All the edges of the sandwich structure and the solid plate are xed. One-quarter of the full plate is
modeled, assuming symmetry of the solution.
9.1.81

Printed on:
Interactions
In the sandwich structure the honeycomb web is welded to the inner surfaces of the plates.
This example demonstrates the usage of CONWEP blast loading using Abaqus/Explicit. This example
was chosen based on experiments reported by Dharmasena et al. (2008) that discuss the deformation
of a particular sandwich structure and an equivalent solid plate subjected to CONWEP blast loads due
to 1, 2, and 3 kg of TNT. The solid plate with equivalent material is presented as a simple example.
The solid plate is modeled using shell elements and three-dimensional continuum elements. Finally,
the main problem of interest, the sandwich structure described by Dharmasena et al., is analyzed. A
tonne-millimeter-second-kelvin unit system was chosen for all simulations.
Case 1
Solid plate modeled with S4R shell elements under 1 kg

TNT blast load.
Case 2

TNT blast load.
Case 3

TNT blast load.
Case 4
Solid plate modeled with C3D8R continuum elements under

1 kg TNT blast load.
Case 5

Case 6

Case 7
Sandwich structure under 1 kg TNT blast load.
Case 8
Case 9
Analysis types
Dynamic analysis using Abaqus/Explicit is performed for all cases. Solutions are computed up to
1.5 milliseconds, where no further permanent deformation is observed for all load values.
Material model
The mechanical properties of the steel alloy as described in Nahshon et al. (2007) are specied as follows:
Youngs modulus of 1.61 105 MPa, Poissons ratio of 0.35, density of 7.85 109 metric tons/mm3 ,
and coefcient of expansion of 452 106 Nmm/metric tons K.
9.1.82

Printed on:
A Johnson-Cook model is used to model the elastic-plastic behavior with the following coefcients
and constants:
400 MPa,
1500 MPa,
0.045,
0.4,
1.2, and
0.001 s1 . The
transition temperature is 293 K, and the melting temperature is 1800 K.
Initial conditions
The temperature is initialized to 273 K for all nodes in each model.

Boundary conditions
Symmetry behavior is assumed. Only one-quarter of the structure is modeled for all cases, with the center
of the plate positioned at the origin of the XY plane. The boundaries at
mm and
mm
are xed for all degrees of motion (ENCASTRE). Symmetry conditions about the x-axis (XSYMM) are
applied at the plane
. Similarly, symmetry conditions about the y-axis (YSYMM) are applied at
the plane
.
Loading
The CONWEP blast load is applied on the top surface of the plate. The source of the blast is at a standoff
distance of 100 mm vertically above the center of the top surface of the plate. The property of the blast
load is specied using the incident wave interaction property and the CONWEP charge property at the
model level and the incident wave interaction at the step level.
Analysis steps
Each analysis consists of a single dynamic explicit step.

Output requests
Translational degrees of freedom (UT) are requested at the center of the plate.
The center displacement after 1.5 milliseconds was monitored to compare each case with experimental
results. Single and double precision job execution gave similar results for all cases.
The animation of the deformed plate over the entire time period of 1.5 milliseconds shows large
deformations at the center. The plate stabilizes after a few oscillations. Comparison of the total work
done history (ALLWK) and the total plastic dissipation history (ALLPD) indicates that most of the work
done by the blast load is dissipated in plastic deformation.
Cases 13: Solid plate with shell elements
The solid plate model is modeled with shell elements and is subjected to CONWEP blast loading using
different charge masses.
9.1.83

Printed on:
Mesh design
The plate surface is discretized using 31 31 S4R elements with nine integration points through the
thickness of each element.
Boundary conditions
All degrees of freedom including the rotational degrees of freedom are xed at the
and
edges boundaries, where
mm. Symmetry boundary conditions (XSYMM) are applied at the
edge, and symmetry boundary conditions (YSYMM) are applied at the
edge.
Loading
1, 2, and 3 kg TNT loads are used in Cases 1, 2, and 3, respectively. The blast source is kept at a standoff
distance of 100 mm from the midsection of the shell elements.
The articial energy history (ALLAE) is signicantly lower than the total internal energy (ALLIE),
indicating that the solution is trustworthy with minimal articial effects.
Cases 46: Solid plate with 3-D continuum elements
The solid plate is modeled with three-dimensional continuum elements and is subjected to CONWEP
blast loading using different charge masses.
Mesh design
The plate surface is discretized using 31 31 C3D8R elements with ve layers of elements through the
thickness of the plate.
Boundary conditions
and
face boundaries, where
mm. X-symmetry boundary conditions are applied at the
and Y-symmetry boundary conditions are applied at the
face.
face,
Loading
1, 2, and 3 kg TNT loads are used in Cases 4, 5, and 6, respectively. The blast source is kept at a standoff
distance of 100 mm from the top surface of the plate.
Output requests
History output of the center deection at the top and the bottom surfaces of the plate is requested at every
increment.
9.1.84

Printed on:
Similar behavior using C3D8R elements was observed as that observed using S4R elements. The top
surface center deection and the bottom surface center deection were quite close to each other. The
mean of the two values was used to compare the midsection deection of the solid plate.
Cases 79: Sandwich plate structure
The sandwich plate structure is modeled using three-dimensional continuum elements for the top and
bottom plates and shell elements for the square honeycomb core and is subjected to CONWEP blast
loading using different charge masses.
Mesh design
The plate surface is discretized using 31 31 C3D8R elements with ve layers of elements through the
thickness of the plate. The honeycomb core is meshed using 30 layers of S4R shell elements along the
height of the core with ve integration points through their thickness.
Boundary conditions
and
face boundaries, where
mm. X-symmetry boundary conditions are applied at the
and Y-symmetry boundary conditions are applied at the
face.
face,
Loading
1, 2, and 3 kg TNT loads are used in Cases 7, 8, and 9, respectively. The blast source was kept at a
standoff distance of 100 mm from the top surface of the plate.
Constraints
The edges of the shell elements at the top and bottom of the core are attached to the inner surfaces of
the top plate and the bottom plate, respectively, using tie constraints. The shell element edges form the
node-based slave surface to the master plate surfaces. The tie constraint is dened without allowing any
adjustments and using the node set of the edges to identify the nodes on the slave surface that will be
tied to the master surface.
Interactions
General contact is specied at the step level, including all exterior surface contact interactions.
Output requests
History output of the center deection at the top and the bottom surfaces of the plate is requested at every
increment.
9.1.85

Printed on:
The analysis demonstrates signicant buckling of the honeycomb webs involving self-contact near the
center of the plate, as shown in Figure 9.1.82 for the three loading cases. Figure 9.1.83 shows the
internal view for the 1 kg TNT blast charge case, revealing the deformation of the honeycomb core. It
also conrms that tie constraints using node-based surfaces capture appropriately the weld between the
webs and the inner surfaces of the plates.
Additional cases
Studies of similar structures using different parameters have been documented in the literature.
Alternative material models and loading cases are mentioned here to provide examples for additional
analyses. Different material models have been used to simulate the material behavior under such loads.
The material model for the following atypical strain-rate hardening (Dharmasena et al., 2008) can be
used for simulation of the above cases with user subroutine VUHARD:
The material properties used by Dharmasena et al. and Rathbun et al. (2006) are as follows: Youngs
modulus of
2.00 105 MPa, Poissons ratio of
0.30, density of
7.85 109 metric
3
3
tons/mm , yield stress of
300.0 MPa, tangent modulus of
2.00 10 MPa,
4916 s1 ,
and
0.154.
In addition, the following approximate loading, which was applied on the sandwich structure
described by Dharmasena et al., can be implemented with user subroutine VDLOAD:
where is the shock wave pressure (360 MPa),

is the ambient pressure (0), is the time when the
shock decays to a value close to zero, is the shock arrival time, is the distance from the center of the
plate (in this example, also the origin), and is a reference distance of 120 mm. All variable values are
chosen based on Dharmasena et al.
The Abaqus/Explicit results for the solid plate using the Johnson-Cook model for the steel alloy
(Cases 16) are shown in Table 9.1.81. These results compare well with the experimental results
reported by Dharmasena et al. (2008), shown in Table 9.1.82, for both the shell and continuum elements
with CONWEP blast loading. The Abaqus/Explicit results for the sandwich structure compare within
reasonable error with the experimental results. The simulation and experimental results are presented
graphically in Figure 9.1.84. The solution differs signicantly for higher loads, which Dharmasena et
9.1.86

Printed on:
al. attribute to boundary conditions. At higher loads, it is likely that the edges of the sandwich panel
used in the test arrangement were actually more exible than the clamped condition used in the Abaqus
model, causing differences between the numerical and experimental results. Differences in the results
could also be due to debonding of the honeycomb core webs from the top and bottom plates in the
experimental setup.
The sandwich structure was modeled for the same CONWEP loading with a different material
model. An isotropic bilinear model with an atypical strain-rate hardening used user subroutine VUHARD
with the material properties as described in Additional cases above. The displacements, shown in
Table 9.1.83, were found to be much higher than the Johnson-Cook model and the experimental results,
probably due to lower yield stress values.
In addition, the solid plate and the sandwich structure were modeled with the Johnson-Cook
material model with an approximate loading, as described in Additional cases above, with user
subroutine VDLOAD. The displacements, shown in Table 9.1.84, were found to be lower than the
CONWEP blast loading and the experimental results, probably because less work was done on the
model by the approximate load over the total time.
Input files
Cases 13 Solid plate with shell elements
exa_aco_conwep_sol_1kg.inp
Solid plate model with S4R elements and 1 kg TNT blast

load.
load.
load.
Cases 46 Solid plate with continuum elements
exa_aco_conwep_spl_1kg.inp
Solid plate model with C3D8R elements and 1 kg TNT

blast load.
blast load.
blast load.
Cases 79 Sandwich structure
exa_aco_conwep_snd_1kg.inp
Sandwich structure model with 1 kg TNT blast load.

Additional cases
exa_aco_conwep_snd_vuhard_1kg.inp
Sandwich structure model with user-dened material

model and 1 kg TNT blast load.
9.1.87

Printed on:
exa_blast_vuhard.f
exa_aco_conwep_sol_vdload_1kg.inp
exa_aco_conwep_snd_vdload_1kg.inp
exa_blast_vdload.f
User subroutine VUHARD with user-dened strain-rate

hardening model.
Solid plate model with user-dened approximate loading
equivalent to 1 kg TNT blast.
Sandwich structure model with user-dened approximate
loading equivalent to 1 kg TNT blast.
User subroutine VDLOAD with the approximate loading
equivalent to 1 kg TNT CONWEP blast load.
References
Acoustic, shock, and coupled acoustic-structural analysis, Section 6.10.1 of the Abaqus Analysis
Users Manual
Acoustic and shock loads, Section 30.4.5 of the Abaqus Analysis Users Manual
*CONWEP CHARGE PROPERTY

*INCIDENT WAVE INTERACTION
*INCIDENT WAVE INTERACTION PROPERTY
Abaqus Verification Manual
CONWEP blast loading pressures, Section 3.7.5 of the Abaqus Verication Manual
Dharmasena, K. P., H. N. G. Wadley, Z. Xue, and J. W. Hutchinson, Mechanical Response of

Metallic Honeycomb Sandwich Panel Structures to High-Intensity Dynamic Loading, Journal of
Impact Engineering, vol. 35, pp. 10631074, 2008.
Nahshon, K., M. G. Pontin, A. G. Evans, J. W. Hutchinson, and F. W. Zok, Dynamic Shear Rupture
of Steel Plates, Journal of Mechanics of Materials and Structures, vol. 210, pp. 20492066,
December 2007.
Rathbun, H. J., D. D. Radford, Z. Xue, M. Y. Hu, J. Yang, V. Deshpande, N. A. Fleck, J. W.

Hutchinson, F. W. Zok, and A. G. Evans, Dynamic Shear Rupture of Steel Plates, International
Journal of Solids and Structures, vol. 43, pp. 17461763, 2006.
Blast loading of a circular plate using the CONWEP model, Section 3.7.6 of the Abaqus
Verication Manual
Other
9.1.88

Printed on:
Table 9.1.81
Center deection computed by Abaqus/Explicit for Cases 19.

Center deflection (mm)
Model
Solid plate
(S4R)
Solid plate
(C3D8R)
Sandwich
structure
Charge mass
Top surface
Midsection
Bottom
surface
1 kg TNT
48.54
2 kg TNT
88.38
3 kg TNT
109.75
1 kg TNT
47.09
2 kg TNT
85.72
3 kg TNT
108.38
1 kg TNT
69.15
26.15
2 kg TNT
110.68
66.14
3 kg TNT
141.37
96.63
Table 9.1.82
Center deection measurements from experiments.

Model
Solid plate
Sandwich
structure
Charge mass
Top surface
Midsection
Bottom
surface
1 kg TNT
37.65
2 kg TNT
71.37
3 kg TNT
132.94
1 kg TNT
47.06
15.29
2 kg TNT
98.82
53.73
3 kg TNT
158.04
127.45
9.1.89

Printed on:
Table 9.1.83
Model
Sandwich
structure with
VUHARD
Center deection computed by Abaqus/Explicit

with user subroutine VUHARD.
Charge mass
Center deflection
of top surface
(mm)
Center deflection
of bottom surface
(mm)
1 kg TNT
96.18
50.74
2 kg TNT
156.76
114.79
3 kg TNT
200.03
155.08
Table 9.1.84
Center deection computed by Abaqus/Explicit

with user subroutine VDLOAD.
Model
Charge mass
Solid plate
(S4R) with
VDLOAD
Sandwich
structure with
VDLOAD
Top surface
Midsection
Bottom
surface
1 kg TNT
32.10
1 kg TNT
36.31
18.89
9.1.810

Printed on:
Z
Y
Figure 9.1.81
The sandwich structure model.
9.1.811

Printed on:
Z
YX
Z
YX
Z
YX
Figure 9.1.82 Deformed sandwich structure under CONWEP

blast loads due to (from top to bottom) 1, 2, and 3 kg TNT.
9.1.812

Printed on:
Z
Y
Figure 9.1.83 Deformation of the honeycomb core for

a CONWEP charge of 1 kg TNT.
9.1.813

Printed on:
AbaqusSandwichBottomCenter
AbaqusSandwichTopCenter
AbaqusSolidPlateC3D8RCenter
AbaqusSolidPlateS4RCenter
ExperimentSandwichBottomCenter
ExperimentSandwichTopCenter
ExperimentSolidPlateCenter
Center Deflection (mm)
150.
100.
50.
0.
1.0
1.5
2.0
2.5
3.0
Blast Charge (kg of TNT)
Figure 9.1.84 Comparison of center deection between

Abaqus/Explicit and experimental results for different blast charges.
9.1.814

Printed on:
SOILS ANALYSES
10.
Soils Analyses
Soils analyses, Section 10.1

Printed on:
SOILS ANALYSES
10.1
Soils analyses
Plane strain consolidation, Section 10.1.1

Calculation of phreatic surface in an earth dam, Section 10.1.2
Axisymmetric simulation of an oil well, Section 10.1.3
Analysis of a pipeline buried in soil, Section 10.1.4
Hydraulically induced fracture in a well bore, Section 10.1.5
Permafrost thawingpipeline interaction, Section 10.1.6
10.11

Printed on:
PLANE STRAIN CONSOLIDATION
10.1.1
Most consolidation problems of practical interest are two- or three-dimensional, so that the one-dimensional
solutions provided by Terzaghi consolidation theory (see The Terzaghi consolidation problem,
Section 1.15.1 of the Abaqus Benchmarks Manual) are useful only as indicators of settlement magnitudes
and rates. This problem examines a linear, two-dimensional consolidation case: the settlement history of
a partially loaded strip of soil. This particular case is chosen to illustrate two-dimensional consolidation
because an exact solution is available (Gibson et al., 1970), thus providing verication of this capability in
Abaqus.
Geometry and model
The discretization of the semi-innite, partially loaded strip of soil is shown in Figure 10.1.11.
The loaded region is half as wide as the depth of the sample. The reduced-integration plane strain
element with pore pressure, CPE8RP, is used in this analysis. Reduced integration is almost always
recommended when second-order elements are used because it usually gives more accurate results
and is less expensive than full integration. No mesh convergence studies have been done, although
the reasonable agreement between the numerical results provided by this model and the solution of
Gibson et al. (1970) suggests that the model used is adequateat least for the overall displacement
response examined. In an effort to reduce analysis cost while at the same time preserve accuracy, the
mesh is graded from six elements through the height, under the load, to one element through the height
at the outer boundary of the model, where a single innite element (type CINPE5R) is used to model
the innite domain. This requires the use of two kinematic constraint features provided by Abaqus.
Consider rst the displacement degrees of freedom along line
in Figure 10.1.11. The 8-node
isoparametric elements used for the analysis allow quadratic variation of displacement along their sides,
so the displacements of nodes a and b in elements x and y may be incompatible with the displacement
variation along side
of element z. To avoid this, nodes a and b must be constrained to lie on the
parabola dened by the displacements of nodes A, B, and
The QUADRATIC MPC (multi-point
constraint) is used to enforce this kinematic constraint: it must be used at each node where this
constraint is required (see planestrainconsolidation.inp). Pore pressure values are obtained by linear
interpolation of values at the corner nodes of an element. When mesh gradation is used, as along line
in this example, an incompatibility in pore pressure values may result for the same reason given for
the displacement incompatibility discussed above. To avoid this, the pore pressure at node B must be
constrained to be interpolated linearly from the pore pressure values at A and
This is done by using
the P LINEAR MPC.
The material properties assumed for this analysis are as follows: the Youngs modulus is chosen
as 690 GPa (108 lb/in2 ); the Poissons ratio is 0; the materials permeability is 5.08 107 m/day
(2.0 105 in/day); and the specic weight of pore uid is chosen as 272.9 kN/m3 (1.0 lb/in3 ).
The applied load has a magnitude of 3.45 MPa (500 lb/in2 ). The strip of soil is assumed to lie on
a smooth, impervious base, so the vertical component of displacement is prescribed to be zero on that
10.1.11

Printed on:
surface. The left-hand side of the mesh is a symmetry line (no horizontal displacement). The innite
element models the other boundary.
Time stepping
As in the one-dimensional Terzaghi consolidation solution (see The Terzaghi consolidation problem,
Section 1.15.1 of the Abaqus Benchmarks Manual), the problem is run in two steps. In the rst *SOILS,
CONSOLIDATION step, the load is applied and no drainage is allowed across the top surface of the mesh.
This one increment step establishes the initial distribution of pore pressures which will be dissipated
during the second *SOILS, CONSOLIDATION step.
During the second step drainage is allowed to occur through the entire surface of the strip. This is
specied by prescribing the pore pressure (degree of freedom 8) at all nodes on this surface (node set
TOP) to be zero. By default in a *SOILS, CONSOLIDATION step such boundary conditions are applied
immediately at the start of the step and then held xed. Thus, the pore pressures at the surface change
suddenly at the start of the second step from their values with no drainage (dened by the rst step) to
0.0.
Consolidation is a typical diffusion process: initially the solution variables change rapidly with time,
while at the later times more gradual changes in stress and pore pressure are seen. Therefore, an automatic
time stepping scheme is needed for any practical analysis, since the total time of interest in consolidation
is typically orders of magnitude larger than the time increments that must be used to obtain reasonable
solutions during the early part of the transient. Abaqus uses a tolerance on the maximum change in pore
pressure allowed in an increment, UTOL, to control the time stepping. When the maximum change of
pore pressure in the soil is consistently less than UTOL the time increment is allowed to increase. If the
pore pressure changes exceed UTOL, the time increment is reduced and the increment is repeated. In
this way the early part of the consolidation can be captured accurately and the later stages are analyzed
with much larger time steps, thereby permitting efcient solution of the problem. For this case UTOL is
chosen as 0.344 MPa (50 lb/in2 ), which is 10% of the applied load. This is a fairly coarse tolerance but
results in an economical and reasonable solution.
The choice of initial time step is important in consolidation analysis. As discussed in The
Terzaghi consolidation problem, Section 1.15.1 of the Abaqus Benchmarks Manual, the initial solution
(immediately following a change in boundary conditions) is a local, skin effect solution. Due to the
coupling of spatial and temporal scales, it follows that no useful information is provided by solutions
generated with time steps smaller than the mesh and material-dependent characteristic time. Time steps
very much smaller than this characteristic time provide spurious oscillatory results (see Figure 3.1.52).
This issue is discussed by Vermeer and Verruijt (1981), who propose the criterion
where
is the distance between nodes of the nite element mesh near the boundary condition change,
E is the elastic modulus of the soil skeleton, k is the soil permeability, and
is the specic weight
of the pore uid. In this problem
is 8.5 mm (0.33 in). Using the material properties shown in
Figure 10.1.11,
10.1.12

Printed on:
days.
We actually use an initial time step of 2 105 days, since the immediate transient just after drainage
begins is not considered important in the solution.
The prediction of the time history of the vertical deection of the central point under the load (point P in
Figure 10.1.11) is plotted in Figure 10.1.12, where it is compared with the exact solution of Gibson et
al. (1970). There is generally good agreement between the theoretical and nite element solutions, even
though the mesh used in this analysis is rather coarse.
Figure 10.1.12 also shows the time increments selected by the automatic scheme, based on the
UTOL tolerance discussed above. The gure shows the effectiveness of the scheme: the time increment
changes by two orders of magnitude over the analysis.
Input file
planestrainconsolidation.inp
Input data for this example.
References
Gibson, R. E., R. L. Schiffman, and S. L. Pu, Plane Strain and Axially Symmetric Consolidation
of a Clay Layer on a Smooth Impervious Base, Quarterly Journal of Mechanics and Applied
Mathematics, vol. 23, pt. 4, pp. 505520, 1970.
Vermeer, P. A., and A. Verruijt, An Accuracy Condition for Consolidation by Finite Elements,
International Journal for Numerical and Analytical Methods in Geomechanics, vol. 5, pp. 114,
1981.
10.1.13

Printed on:
Infinite element
q
C
b
B
x
h
y
A
w
Geometry:
h = 50.8 mm (2.0 in)
b = 25.4 mm (1.0 in)
w = 533.4 mm (21.0 in)
Material:
E = 690 GPa (1.0 x 108 lb/in2)
= 0.0
k = 5.08 x 10-7 m/day (2.0 x 10-5 in/day)
w = 2.729 x 105 N/m3 (1.0 lb/in3)
eo = 1.5
Loading:
q = 3.45 MPa (500.0 lb/in2)
Figure 10.1.11
Plane strain consolidation example: geometry and properties.
10.1.14

Printed on:
10-6 in
10-1
Gibson et al. (1970)

ABAQUS
Vertical displacement
10-2
2.0
Vertical
displacement
10-3
7
10-4
1.5
Time increments
10-5
1.0
10-6
10-6
10-5
10-4
10-3
10-2
10-1
Time, days
Figure 10.1.12
Consolidation history and time step variation history.
10.1.15

Printed on:
Time increments, days
10-4 mm
PHREATIC SURFACE IN DAM
10.1.2
CALCULATION OF PHREATIC SURFACE IN AN EARTH DAM
This example illustrates the use of Abaqus to solve for the ow through a porous medium in which uid ow
is occurring in a gravity eld and only part of the region is fully saturated, so the location of the phreatic
surface is a part of the solution. Such problems are common in hydrology (an example is the well draw-down
problem, where the phreatic surface of an aquifer must be located based on pumping rates at particular well
locations) and in some problems of dam design, as in this example. The basic approach takes advantage of
the Abaqus capability to perform partially and fully saturated analysis: the phreatic surface is located at the
boundary of the fully saturated part of the model. This approach has the advantage that the capillary zone,
just above the phreatic surface, is also identied.
Boundary conditions
A typical dam is shown in Figure 10.1.21. We consider uid ow only: deformation of the dam is
ignored. Thus, although we use the fully coupled pore uid ow-deformation elements, all displacement
degrees of freedom are prescribed to be zero. A more general analysis would include stress and
deformation of the dam.
The upstream face of the dam (surface
in Figure 10.1.21) is exposed to water in the reservoir
behind the dam. Since Abaqus uses a total pore pressure formulation, the pore pressure on this face must
be prescribed to be
, where
is the elevation of the water surface, z is elevation,
g is the gravitational acceleration, and
is the mass density of the water. (
, the weight density of
the water, must be given as the value of the SPECIFIC parameter on the *PERMEABILITY option.)
Likewise, on the downstream face of the dam (surface
in Figure 10.1.21),
The bottom of the dam (surface ) is assumed to rest on an impermeable foundation. Since the
natural boundary condition in the pore uid ow formulation provides no ow of uid across a surface
of the model, no further specication is needed on this surface.
The phreatic surface in the dam, , is found as the locus of points at which the pore uid pressure,
, is zero. Above this surface the pore uid pressure is negative, representing capillary tension causing
the uid to rise against the gravitational force and creating a capillary zone. The saturation associated
with particular values of capillary pressure for absorption and exsorption of uid from the porous medium
is a physical property of the material and is dened in the *SORPTION option.
A special boundary condition is needed if the phreatic surface reaches an open, freely draining
surface, as indicated on surface
in Figure 10.1.21. In such a case the pore uid can drain freely
down the face of the dam, so
= 0 at all points on this surface below its intersection with the phreatic
surface. Above this point
0, with its particular value depending on the solution. This example is
specically chosen to include this effect to illustrate the use of the Abaqus drainage-only ow boundary
condition.
This drainage-only ow condition consists of prescribing the ow velocity on the freely draining
surface in a way that approximately satises the requirement of zero pore pressure on the completely
saturated portion of this surface (Pagano, 1997). The ow velocity is dened as a function of pore
10.1.21

Printed on:
pressure, as shown in Figure 10.1.22. For negative pore pressures (those above the phreatic surface)
the ow velocity is zerothe proper natural boundary condition. For positive pore pressures (those
below the phreatic surface) the ow velocity is proportional to the pore pressure value. When this
proportionality coefcient, , is large compared to
where k is the permeability of the medium,
is the specic weight of the uid, and c is a characteristic length scalethe requirement of zero
pore pressure on the free-drainage surface below the phreatic surface will be satised approximately.
The drainage-only seepage coefcient in this model is specied as
101 m3 /Nsec. This value is
5
roughly 10 times larger than the characteristic value,
, based on the material properties listed
below and an element length scale 101 m. This condition is prescribed using the *FLOW option with
the drainage-only ow type label (QnD) as shown in phreaticsurf_cpe8rp.inp.
Geometry and model
The geometry of the particular earth dam considered is shown in Figure 10.1.23. This case is chosen
because an analytical solution is available for comparison (Harr, 1962). The dam is lled to two-thirds of
its height. Only a part of its base is impermeable. Since the dam is assumed to be long, we use CPE8RP
coupled pore pressure/displacement plane strain elements (the mesh is shown in Figure 10.1.24). In
addition, input les containing element types CPE4P and CPE6MP are included for verication purposes.
Additional input les are included to demonstrate the use of the *CONTACT PAIR, *TIE, and *MAP
SOLUTION options in coupled pore pressure-displacement analyses.
Material
The permeability of the fully saturated earth of which the dam is made is 0.2117 103 m/sec. The default
assumption is used for the partially saturated permeability: that it varies as a cubic function of saturation,
decreasing from the fully saturated value to a value of zero at zero saturation. The specic weight of the
water is 10 kN/m3 . The capillary action in the dam is dened by a single absorption/exsorption curve
that varies linearly between a negative pore pressure of 10 kN/m2 at a saturation of 0.05 and zero pore
pressure at fully saturated conditions. This is not a very realistic model of physical absorption/exsorption
behavior, but this will not affect the results of the steady-state analysis signicantly insofar as the location
of the phreatic surface is concerned. Accurate denition of this behavior would be required if denition
of the capillary zone created by lling and emptying the dam at given rates were needed.
The initial void ratio of the earth material is 1.0. The initial conditions for pore pressure and
saturation are assumed to be those corresponding to the dam being fully saturated to the upstream water
level: the initial saturation is, therefore, 1.0; and the initial pore pressures vary between zero at the water
level and a maximum value of 12.19 kN/m2 at the base of the dam.
Loading and controls
The weight of the water is applied by GRAV loading, and the upstream and downstream pore pressures
are prescribed as discussed above. A steady-state *SOILS analysis is performed in ve increments to
allow Abaqus to resolve the high degree of nonlinearity in the problem.
10.1.22

Printed on:
The steady-state contours of pore pressure are shown in Figure 10.1.25. The upper-right part of the
dam shows negative pore pressures, indicating that it is partly saturated or dry. The phreatic surface
is best shown in Figure 10.1.26, where we have chosen to draw the contours in the vicinity of zero
pore pressure. This phreatic surface compares well with the analytical phreatic surface calculated by
Harr (1962), shown in Figure 10.1.22. Figure 10.1.27 shows contours of saturation that indicate a
region of fully saturated material under the phreatic zone and decreasing saturation in and above the
phreatic zone.
Input files
phreaticsurf_cpe8rp.inp
phreaticsurf_cpe4p.inp
phreaticsurf_cpe6mp.inp
phreaticsurf_cpe4p_contactpair.inp
phreaticsurf_cpe4p_mapsolution.inp
Phreatic surface calculation (element type CPE8RP).

Element type CPE4P.
Element type CPE6MP.
Element type CPE4P using the *CONTACT PAIR option.
*MAP SOLUTION continuation of the
phreaticsurf_cpe4p_contactpair.inp analysis.
Element type CPE4P using the *TIE option.
phreaticsurf_cpe4p_tie.inp
References
Harr, M. E., Groundwater and Seepage, McGraw-Hill, New York, 1962.

Pagano, L., Steady State and Transient Unconned Seepage Analyses for Earthll Dams,
ABAQUS Users Conference, Milan, pp. 557585, 1997.
S4
S5
1 = H1
S1
Permeable
material
H1
z
S3
S2
H2
2 = H2
x
y
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
Impermeable material
Figure 10.1.21
Phreatic surface problem.
10.1.23

Printed on:
flow velocity, vn
ks
pore pressure, uw
Figure 10.1.22
Pore pressureow velocity relationship dened on the drainage-only surface.
1.22 m
Harr (analytical)
Phreatic surface
1.83 m
1.22 m
45o
Impervious
45o
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
3.85 m
4.88 m
Figure 10.1.23
Conguration of earth dam and analytical phreatic surface.
10.1.24

Printed on:
Figure 10.1.24
POR
Finite element mesh.
VALUE
-INFINITY
-2.70E+04
-2.34E+04
-1.99E+04
-1.63E+04
-1.28E+04
-9.23E+03
-5.67E+03
-2.12E+03
+1.44E+03
+4.99E+03
+8.55E+03
+1.21E+04
+1.22E+04
Figure 10.1.25
Pore pressure contours at steady state.
10.1.25

Printed on:
POR
VALUE
-1.11E+04
-1.00E+01
-8.18E+00
-6.36E+00
-4.55E+00
-2.73E+00
-9.09E-01
+9.09E-01
+2.73E+00
+4.55E+00
+6.36E+00
+8.18E+00
+1.00E+01
+1.22E+04
Figure 10.1.26
SAT
Pore pressure contours showing phreatic surface.
VALUE
-4.61E-02
+6.00E-01
+6.27E-01
+6.55E-01
+6.82E-01
+7.09E-01
+7.36E-01
+7.64E-01
+7.91E-01
+8.18E-01
+8.45E-01
+8.73E-01
+9.00E-01
+1.11E+00
Figure 10.1.27
Saturation contours at steady state.
10.1.26

Printed on:
AXISYMMETRIC OIL WELL
10.1.3
AXISYMMETRIC SIMULATION OF AN OIL WELL
This example simulates the settlement of soil near an oil well. It is assumed that the oil in question is too
thick for normal pumping. Therefore, steam is injected in the soil in the vicinity of the well to increase
the temperature and decrease the oils viscosity. As a result creep becomes an important component of the
soil inelastic deformation and in the prediction of the effects of the oil pumping. Five years of oil pumping
are simulated. This coupled displacement/diffusion analysis illustrates the use of Abaqus to solve problems
involving uid ow through a saturated porous medium, inelastic material properties with time-dependent
creep behavior, and thermal loading. No experimental data exist to compare with the numerical results of this
example.
Geometry and model
The example considers an axisymmetric model of an oil well and the surrounding soil, as shown in
Figure 10.1.31. The radius of the well is 81 m (265 ft), and the well extends from a depth of 335 m
(1100 ft) to 732 m (2500 ft). A depth of 1463 m (4800 ft) is modeled with 11 different soil layers.
Reduced-integration axisymmetric elements with pore pressure, CAX8RP, are used to model the
soil in the vicinity of the well. The far-eld region is modeled with axisymmetric innite elements,
CINAX5R, to provide lateral stiffness. Reduced integration is almost always recommended when
second-order elements are used, because it usually gives more accurate results and is less expensive
than full integration. A coarse mesh is selected for the illustrative purpose of this example. No mesh
convergence study has been performed.
Soil layers designated by S1, T1, U1, and L1 are modeled using the Drucker-Prager plasticity model
and are specied on the *DRUCKER PRAGER option. Both the elastic and inelastic material properties
are tabulated in Table 10.1.31. The linear form of the Drucker-Prager model with no intermediate
principal stress effect (
1.0) is used. The model assumes nonassociated ow; consequently, the
material stiffness matrix is not symmetric. The use of UNSYMM=YES on the *STEP option improves
the convergence of the nonlinear solution signicantly. The hardening/softening behavior is specied by
the *DRUCKER PRAGER HARDENING option, and the data are listed in Table 10.1.31. No creep
data are provided for these layers since these are far removed from the loading. These layers are assumed
to be saturated with water. A high permeability is assumed for the two top soil layers S1 and T1, while
a low permeability is assigned to layers U1 and L1.
Layers D1 through D7 are modeled with the modied Drucker-Prager Cap plasticity model. The
material property data are tabulated in Table 10.1.32 and are specied by the *CAP PLASTICITY
option. As required by the creep model, no intermediate principal stress effect is included (i.e.,
1.0), and no transition region on the yield surface is dened (i.e.,
0.0). The materials
volumetric strain-driven hardening/softening behavior is specied with the *CAP HARDENING
option, and the data are listed in Table 10.1.32. The initial cap yield surface position,
, is set to
0.02. Abaqus automatically adjusts the position of the cap yield surface if the stress lies outside the cap
surface. Consolidation creep is modeled with a Singh-Mitchell type creep model. The creep material
10.1.31

Printed on:
data are specied with the *CAP CREEP option and are dependent on temperature. The following
creep data are specied:
A=2.2E7 1/day, =3.05 1/MPa (0.021 1/psi),
A=3.5E4 1/day, =3.05 1/MPa (0.021 1/psi),
=1.0 day, n=1.0 at 10C (50F)

=1.0 day, n=1.0, at 100C (212F)
These layers consist of rich organic matter and are saturated with oil. The temperature-dependent
permeability data are specied by the *PERMEABILITY option.
A uniform thermal expansion coefcient of 5.76E6 1/C (3.2E6 1/F) and a constant weight
density 1.0 metric ton/m3 (64.6 lbs/ft3 ) are assumed for all layers.
For a coupled diffusion/displacement analysis care must be taken when choosing the units of the
problem. The coupled equations may be numerically ill-conditioned if the choice of the units is such that
the numbers generated by the equations of the two different elds differ by many orders of magnitude.
The units chosen for this example are inches, pounds, and days.
Initial conditions
An initial geostatic stress eld is dened through the *INITIAL CONDITIONS option and is based on
the soil weight density integrated over the depth. A coefcient of lateral stress of 0.85 is assumed. An
initial void ratio of 1.5 is used throughout all soil layers with an initial uniform temperature eld of 10C
(50F).
Loading
The problem is run in ve steps. The rst step of the analysis is a *GEOSTATIC step to equilibrate
geostatic loading of the nite element model. This step also establishes the initial distribution of pore
pressure. Since gravity loading is dened with distributed load type BZ and not with gravity load type
GRAV, the pore uid pressure reported by Abaqus is dened as the pore pressure in excess of the
hydrostatic pressure required to support the weight of pore uid above the elevation of the material
point.
The second step is a *SOILS, CONSOLIDATION step to equilibrate any creep effects induced
from the initial geostatic loading step. The choice of the initial time step is important in a consolidation
analysis. Because of the coupling of spatial and temporal scales, no useful information is provided by
solutions generated with time steps that are smaller than the mesh and material-dependent characteristic
time. Time steps that are very much smaller than this characteristic time provide spurious oscillatory
results. For further discussion on calculating the minimum time step, refer to Coupled pore uid
diffusion and stress analysis, Section 6.8.1 of the Abaqus Analysis Users Manual. For this example a
minimum initial time step of one day was selected.
The third step of the analysis models the injection of steam into the well region between a depth
of 366 m to 732 m (1200 ft to 2400 ft). The region is indicated by the shaded area in Figure 10.1.31.
The nodes in this region are heated to 100C (212F) during a *SOILS, CONSOLIDATION analysis.
The NO CREEP parameter is included; therefore, creep effects are not considered. The injection of the
steam increases the permeability of the oil and increases the soil creep behavior.
10.1.32

Printed on:
The fourth step simulates the pumping of oil by prescribing an excess pore pressure of 1.2 MPa
(170 psi) at nodes located at the depth of 427 m to 550 m (1400 ft to 1800 ft) below the surface. The
pressure produces a pumping rate of approximately 172.5 thousand barrels per day at the end of the fth
year.
The nal step consists of a consolidation analysis performed over a ve-year period to investigate
the settlement that results from pumping and creep effects in the vicinity of the well.
The two initial steps show negligible deformations, indicating that the model is in geostatic equilibrium.
Figure 10.1.32 shows a contour plot of the soil settlement resulting from consolidation after the ve-year
period. A settlement of 0.13 m (0.4 ft) is expected at the surface. A maximum soil dislocation of 0.24 m
(0.78 ft) occurs above the pump intake. Figure 10.1.33 shows a contour plot of the excess pore pressure.
The negative pore pressure represents the suction of the pump. During the ve-year period, a total of
313.5 million barrels of oil are pumped (as determined from nodal output variable RVT). Figure 10.1.34
through Figure 10.1.36 show contour plots of the vertical stress components, plastic strains, and creep
strains, respectively. Plastication occurs in soil layers D3 through D5. Signicant creep occurs in the
area in which steam is injected.
Input files
axisymoilwell.inp
axisymoilwell_thermalexp.inp
Finite element analysis.

Same as axisymoilwell.inp except that the thermal
expansion of the pore uid is also included.
10.1.33

Printed on:
Table 10.1.31
Soil layer
S1
Soil data using Drucker-Prager model.
Elastic
properties
124 MPa
0.3
T1
468.8 MPa
0.22
L1
2068 MPa
0.25
U1
Inelastic
properties
2482 MPa
0.29
42.0
1.0
0.075 MPa, 0.0

0.083 MPa, 0.058
0.0
0.075 MPa, 0.116
36.0
1.0
0.48 MPa, 0.0

0.62 MPa, 0.058
0.0
0.48 MPa., 0.116
38.0
1.0
1.97 MPa, 0.0

3.17 MPa, 0.0037
0.0
2.47 MPa, 0.04
38.0
1.0
1.97 MPa, 0.0

3.17 MPa, 0.0037
0.0
2.47 MPa, 0.04
10.1.34

Printed on:
Hardening behavior
Table 10.1.32
Soil layer
Soil data using modied Drucker-Prager cap model.
Elastic
properties
D1
328 MPa
0.17
Inelastic properties
d
R
D2
434 MPa
0.17
d
R
D3
546 MPa
0.19
d
R
D4
411 MPa
0.2
d
R
D5
494 MPa
0.17
d
R
D6
775 MPa
0.17
d
R
D7
1,121 MPa
0.17
d
R
1.38 MPa
36.9
0.33
1.38 MPa
39.4
0.33
1.38 MPa
42.0
0.34
1.2 MPa
40.1
0.3
1.38 MPa
40.4
0.3
17 MPa
50.2
0.23
1.7 MPa
58.5
0.23
10.1.35

Printed on:
Hardening
behavior
0.0
1.0
0.02
0.0
1.0
0.02
0.0
1.0
0.02
0.0
1.0
0.02
0.0
1.0
0.02
0.0
1.0
0.02
0.0
1.0
0.02
2.75 MPa, 0.0

4.14 MPa, 0.02
5.51 MPa, 0.05
6.20 MPa, 0.09
1.38 MPa, 0.0
4.14 MPa, 0.02
6.89 MPa, 0.04
55.1 MPa, 0.1
1.38 MPa, 0.0
3.45 MPa, 0.02
13.8 MPa, 0.04
62.0 MPa, 0.06
1.38 MPa, 0.0
5.03 MPa, 0.02
6.90 MPa, 0.10
62.0 MPa, 0.3
2.75 MPa, 0.0
4.83 MPa, 0.02
5.15 MPa, 0.04
62.0 MPa, 0.08
2.76 MPa, 0.0
4.14 MPa, 0.005
7.58 MPa, 0.02
62.0 MPa, 0.05
3.44 MPa, 0.0
4.14 MPa, 0.006
7.58 MPa, 0.012
67.6 MPa, 0.03
CL
E.L.
E.L.
0
-46 m
E.L.
-213 m
E.L.
E.L.
-335 m
-366 m
E.L.
-427 m
E.L.
-550 m
E.L.
-610 m
E.L.
-732 m
E.L.
-915 m
E.L.
-1097 m
SOIL S1
SOIL T1
SOIL D1
SOIL D2
SOIL D3
SOIL D4
SOIL D5
SOIL D6
SOIL D7
SOIL U1
SOIL L1
E.L.
-1463 m
Figure 10.1.31
U2
Axisymmetric model of oil well and surrounding soil.
VALUE
-9.45E+00
-8.47E+00
-7.49E+00
-6.51E+00
-5.53E+00
-4.55E+00
-3.57E+00
-2.59E+00
-1.61E+00
-6.37E-01
+3.42E-01
+1.32E+00
+2.30E+00
+3.28E+00
2
3
Figure 10.1.32
Soil settlement after ve-year period.
10.1.36

Printed on:
POR
VALUE
-1.70E+02
-1.56E+02
-1.43E+02
-1.30E+02
-1.17E+02
-1.04E+02
-9.15E+01
-7.84E+01
-6.53E+01
-5.23E+01
-3.92E+01
-2.61E+01
-1.30E+01
+1.33E-25
2
3
Figure 10.1.33
S22
Contour plot of the pore pressure.
VALUE
-2.16E+03
-1.99E+03
-1.82E+03
-1.66E+03
-1.49E+03
-1.32E+03
-1.16E+03
-9.97E+02
-8.31E+02
-6.64E+02
-4.98E+02
-3.32E+02
-1.66E+02
+2.09E-01
2
3
Figure 10.1.34
Contour plot of the vertical stress components.
10.1.37

Printed on:
PE22
VALUE
-8.77E-04
-7.92E-04
-7.06E-04
-6.21E-04
-5.35E-04
-4.49E-04
-3.64E-04
-2.78E-04
-1.92E-04
-1.07E-04
-2.16E-05
+6.39E-05
+1.49E-04
+2.35E-04
2
3
Figure 10.1.35
CE22
Contour plot of the vertical plastic strain components.
VALUE
-3.73E-03
-3.40E-03
-3.08E-03
-2.76E-03
-2.44E-03
-2.12E-03
-1.80E-03
-1.48E-03
-1.16E-03
-8.47E-04
-5.26E-04
-2.06E-04
+1.13E-04
+4.34E-04
2
3
Figure 10.1.36
Contour plot of the vertical creep strain components.
10.1.38

Printed on:
BURIED PIPELINE
10.1.4
ANALYSIS OF A PIPELINE BURIED IN SOIL
Oil and gas pipelines are usually buried in the ground to provide protection and support. Buried pipelines
may experience signicant loading as a result of relative displacements of the ground along their length. Such
large ground movement can be caused by faulting, landslides, slope failures, and seismic activity.
Abaqus provides a library of pipe-soil interaction (PSI) elements to model the interaction between a
buried pipeline and the surrounding soil. The pipeline itself is modeled with any of the beam, pipe, or elbow
elements in the Abaqus/Standard element library. The ground behavior and soil-pipe interaction are modeled
with the pipe-soil interaction elements. These elements have only displacement degrees of freedom at their
nodes. One side or edge of the element shares nodes with the underlying beam, pipe, or elbow element that
models the pipeline. The nodes on the other edge represent a far-eld surface, such as the ground surface, and
are used to prescribe the far-eld ground motion. The elements are described in detail in Pipe-soil interaction
elements, Section 29.13.1 of the Abaqus Analysis Users Manual.
The purpose of this example is to determine the stress state along the length of a innitely long buried
pipeline subjected to large fault movement of 1.52 m (5.0 ft), as shown in Figure 10.1.41. The pipeline
intersects the fault at 90.0. The results are compared with results from an independent analysis, as described
below.
Problem description
The problem consists of an innitely long pipeline buried at a depth of 6.1 m (20.0 ft.) below the ground
surface. Only a 610.0 m (2000.0 ft.) long section of the pipeline is modeled. The outside radius of the
pipe is 0.61 m (24.0 in), and the wall thickness is 0.0254 m (1.0 in). The pipeline is modeled with 50
rst-order PIPE21 elements. A nonuniform mesh, with smaller elements focused near the fault, is used.
The pipe-soil interaction behavior is model with PSI24 elements. The PSI elements are dened so
that one edge of the element shares nodes with the underlying pipe element, and the nodes on the other
edge represent a far-eld surface where ground motion is prescribed. The far-eld side and the side that
shares nodes with the pipeline are dened by the element connectivity.
A three-dimensional model that uses PIPE31 and PSI34 elements is also included for verication
purposes.
Material
The pipeline is made of an elasticperfectly plastic metal, with a Youngs modulus of 206.8 GPa
(30 106 lb/in2 ), a Poissons ratio of 0.3, and a yield stress of 413.7 MPa (60000 lb/in2 ).
The pipe-soil interaction behavior is elasticperfectly plastic. The *PIPE-SOIL STIFFNESS,
TYPE=NONLINEAR option is used to dene the interaction model. The behavior in the vertical
direction is assumed to be different from the behavior along the axial direction. It is further assumed
that the pipeline is buried deep below the ground surface so that the response is symmetric about the
origin. Abaqus also allows a nonsymmetric behavior to be dened in any of the directions (this is
10.1.41

Printed on:
BURIED PIPELINE
usually the case in the vertical direction when the pipeline is not buried too deeply). The ultimate
force per unit length in the axial direction is 730.0 N/m (50.0 lb/ft), and in the vertical direction it is
1460.0 N/m (100.0 lb/ft). The ultimate force is reached at 0.0304 m (0.1 ft) in both the horizontal and
vertical directions.
The loading occurs in a plane (axial-vertical), so the properties for the pipe-soil interaction behavior
in the transverse horizontal direction are not important.
Loading
The loading on the pipeline is caused by a relative vertical displacement 1.52 m (5.0 ft) along the fault
line. It is assumed that the effect of the vertical ground motion decreases linearly over a distance of
91.4 m (300.0 ft.) from the origin of fault, as shown in Figure 10.1.41.
This linear distribution of ground motion is prescribed as follows. Rigid (R2D2) elements are
connected to the far-eld edges of the PSI to create two rigid surfaces, one on each side of the fault
line. These surfaces extend a distance of 91.4 m (300.0 ft.) from the origin of the fault. The rigid body
reference nodes are also placed a distance of 91.4 m (300.0 ft.) from the fault on the ground surface. The
fault movement is modeled by prescribing a rotation to each of the rigid body reference nodes so that a
positive vertical displacement of 0.76 m (2.5 ft) is obtained on one side of the fault and a negative vertical
displacement of 0.76 m (2.5 ft) is obtained on the other side of the fault, as shown in Figure 10.1.42.
All degrees of freedom on the remaining far-eld nodes are fully xed. In addition, the two end points of
pipeline are fully xed. Figure 10.1.42 does not show the PSI elements or any of the remaining nodes
on the ground surface.
Reference solution
The reference solution is obtained by using JOINTC elements between the pipeline and ground nodes
to model the pipe-soil interaction. These elements provide an internal stiffness, which is modeled with
linear or nonlinear springs; nonlinear springs are used in this example. The behavior of the nonlinear
spring is elastic in the sense that reversed loading does not result in permanent deformation. This behavior
is different from the behavior provided by the nonlinear PSI elements. However, this is not a limitation
in this example since the loading is monotonic.
Another distinct difference between JOINTC elements and PSI elements is that the spring behavior
associated with JOINTC elements is dened in terms of total force, whereas the constitutive behavior for
PSI elements is dened as a force/unit length. This difference requires us to dene a separate stiffness for
each JOINTC element or to use a uniform mesh with JOINTC elements spaced at unit length intervals
along the pipeline. A unit length mesh is used in this example.
Figure 10.1.43 and Figure 10.1.44 show the axial and vertical forces per unit length applied to the
pipeline due to relative ground motion. The gures show that permanent deformation occurs in the pipesoil interaction model near the fault along the axial and horizontal directions, with purely elastic behavior
further from the fault.
10.1.42

Printed on:
BURIED PIPELINE
Figure 10.1.45 compares the axial stress in the bottom wall of the pipeline with the reference
solution. The gure shows that the pipeline behavior is purely elastic. The gure also shows close
agreement with the reference solution. The small differences between the solutions can be accounted
for by the different mesh densities. The reaction forces at the pipeline edges and the maximum pipeline
displacements are also in close agreement with the reference solution.
Input files
buriedpipeline_2d.inp
buriedpipeline_3d.inp
buriedpipeline_ref.inp
Two-dimensional model using PSI24 elements.

Three-dimensional model using PSI34 elements.
Reference solution using JOINTC elements.
Reference
Audibert, J. M. E., D. J. Nyman, and T. D. ORourke, Differential Ground Movement Effects

on Buried Pipelines, Guidelines for the Seismic Design of Oil and Gas Pipeline Systems, ASCE
publication, pp. 151180, 1984.
91 m
ground surface
6.1 m
pipeline
fault
305 m
Figure 10.1.41
Pipe with fault motion.
10.1.43

Printed on:
BURIED PIPELINE
Rigid Surfaces
Pipeline
Figure 10.1.42
Figure 10.1.43
Displaced shape (magnication factor=10.0).
Axial force/unit length applied along the pipeline.
10.1.44

Printed on:
BURIED PIPELINE
Figure 10.1.44
Vertical force/unit length applied along the pipeline.
Figure 10.1.45
Axial stress along the bottom of the pipeline.
10.1.45

Printed on:
HYDRAULIC FRACTURE
10.1.5
HYDRAULICALLY INDUCED FRACTURE IN A WELL BORE
This example demonstrates the use of pore pressure cohesive elements to model the initiation and opening of
hydraulically induced cracks near an oil-well bore hole. With the technique illustrated in this section you can
assess the quantitative impact of the hydraulic fracture process on well bore productivity.
Problem description
The hydraulic fracture process is commonly used in the production of oil and natural gas reservoirs
as a means of increasing well productivity and extending the production lifetime of the reservoir. The
objectives of a hydraulic fracture treatment are
1. to create more surface area that is exposed to the hydrocarbon bearing rocks, and
2. to provide a highly conductive pathway that allows the hydrocarbons to ow easily to the well bore.
The productivity of a hydraulically fractured oil or gas well is directly related to the extent of the fracture
and to how well the well bore is connected to the fracture. Some rock formations contain natural fracture
systems that can further increase a wells productivity, provided that the generated hydraulic fracture can
grow such that it intersects these natural fractures.
A hydraulic fracture job consists of pumping uids into a well at very high pressures so that the
tractions created on the well-bore face reduce the in-situ (compressive) stress in the rock so much that
the rock fractures. Once a fracture initiates in the rock formation it is possible, given enough hydraulic
uid, to propagate the fracture for a considerable distance, sometimes as far as a hundred meters or more.
Execution of a fracture job is a complex operation. In most cases several different types of uid
(stages) will be pumped during a fracture job.
Pumping stages
The initial stage often involves pumping a rather small amount of polymer laden uid, typically
120 barrels (.15 to 3.2 cubic meters) so that data can be gathered on the pressure needed to fracture the
formation and the rate at which the uid will leakoff from the fracture into the pore space of the rock.
The data gathered are used to plan subsequent stages of the job. The main stage of the job might consist
of anywhere from one hundred to several thousands of barrels of hydraulic uid. The size of this stage
is determined by the target fracture size, the leakoff rate, and the capacity (rate) of the pumps.
During the next stage of the fracture job, solid material, known as proppant, is added to the injected
uid and is carried into the fracture volume. Chemicals, typically polymers, are added to the uid in
each stage of a fracture job to produce the necessary properties in the uid (viscosity, leakoff, density).
In the last stage of the job, chemicals are pumped into the fracture that help breakdown the polymers
used in the previous stages and make it easier to ow uid back through the fracture without disrupting
the proppant material.
10.1.51

Printed on:
HYDRAULIC FRACTURE
Geometry and model
The domain of the problem considered in this example is a 50 m (1969 in) thick circular slice of oilbearing rock with the well-bore hole modeled. The domain has a diameter of 400 m (15,748 in). Three
sections of rock are considered: a region where production oil recovery is targeted, and two surrounding
regions of shale. The domain is shown in Figure 10.1.51.
Due to symmetry only a half of the domain is modeled. Figure 10.1.52 shows the nite element
model. The rock is modeled with C3D8RP elements, and the bore hole casing is modeled with M3D4
elements.
The unopened fracture is modeled along the entire height of the model domain. Cohesive elements
(COH3D8P) are used to model a vertical fracture surface.
Rock constitutive model
A linear Drucker-Prager model with hardening is chosen for the rock, while the casing is linear elastic.
Fracture constitutive model
The fracture model consists of the mechanical behavior of the fracture itself and the behavior of the uid
that enters and leaks through the fracture surfaces.
Fracture mechanical behavior
The elastic properties of the bonded interface are dened using a traction-separation description, with
stiffness values of
=
=
= 8.5 104 MPa. The quadratic traction-interaction failure criterion
is chosen for damage initiation in the cohesive elements; and a mixed-mode, energy-based damage
evolution law is used for damage propagation. The relevant material data are as follows:
=
=
0.32 KPa,
=
=
= 28 N/mm, and = 2.284.
Fluid model
Tangential and normal ow are both modeled in the fracture zone cohesive elements. The following
parameters are specied:
Gap ow is specied as Newtonian with a viscosity of 1 106 kPaS (1 centepoise), roughly the
viscosity of water.
Fluid leakoff is specied as 5.879 1010 m/(kPa s) for the early stages. In the nal stage, when
the polymer is dissolved, the uid leakoff coefcient is increased to 1 103 m/(kPa s). This
step-dependent uid leakoff coefcient is set in user subroutine UFLUIDLEAKOFF.
Initial conditions
An initial geostatic stress eld is dened using the user subroutines SIGINI and UPOREP. A depth
varying initial void ratio is specied using user subroutine VOIDRI. Gravity loading is specied; and
an orthotropic overburden stress state is imposed, with the maximum principle stress in the formation
aligned orthogonal to the cohesive element fracture plane.
10.1.52

Printed on:
HYDRAULIC FRACTURE
Loading and boundary conditions
The analysis consists of four steps:

1. A geostatic step is performed where equilibrium is achieved after applying the initial pore pressure
to the formation and the initial in-situ stresses. The bottom hole shut-in pressure is applied as a
traction to the well-bore face.
2. The next step represents the hydraulic fracture stage where the main volume of uid is being injected
into the well. Flow at a rate of 2.4 m3 (15 barrels) per minute is injected along a targeted 8 m extent
in the target formation in the model, and the cohesive elements adjacent to the well bore along this
length are dened as initially open to permit entry of uid. The duration of this stage is 20 minutes.
3. Following the hydraulic fracture, another transient soils consolidation analysis is conducted. The
injection into the well is terminated, and the built-up pore pressure in the fracture is allowed to bleed
off into the formation. An additional boundary condition is applied at this stage, xing the fracture
surface open to simulate the behavior of the proppant material that was injected into the fracture.
4. In the nal step a drawdown pressure of 20 kPa is applied to the well-bore nodes of the fracture
cohesive elements. This step ends after 100 days of production or when steady-state conditions
prevail, dened as pore pressure transients below 0.05 kPa/sec in the model.
The ow injected during Step 2, the pumping stage, initiated and grew a crack extending outward from
the well bore. Figure 10.1.53 shows the resulting geometry of the fracture at the end of the 20 minute
pumping period. These results show that a fracture initiated within the target formation zone tends to
avoid the lower shale region, where the compressive stresses are higher, but does inltrate the upper
shale region, where it can decrease well yields.
Figure 10.1.54 shows the fracture opening prole at various times during the pumping stage; the
nal prole, shown at 1200 s, is then xed in the subsequent steps, holding the fracture open with the
proppant material. Figure 10.1.55 shows a similar history of the pore pressure across the fracture face
and indicates that the pore ow has stabilized.
Figure 10.1.56 shows the resulting well-bore yield following the hydraulic fracture process. This
is compared to an equivalent model where the hydraulic fracturing did not occur. In this simple example
the hydraulically fractured well bore shows a marked improvement, with a ow rate more than 100 times
the unfractured conguration.
Input files
exa_hydfracture.inp
exa_hydfracture.f
Hydraulic fracture analysis.

User subroutines DISP, DLOAD, SIGINI, UPOREP,
UFLUIDLEAKOFF, and VOIDRI.
10.1.53

Printed on:
HYDRAULIC FRACTURE
Figure 10.1.51 Position of the target formation

and surrounding shale.
Figure 10.1.52
Near well-bore mesh.
10.1.54

Printed on:
HYDRAULIC FRACTURE
Figure 10.1.53 Fracture geometry following the injection

stage, with the deformation magnied 500 times.
Figure 10.1.54
History of the fracture opening prole.
10.1.55

Printed on:
HYDRAULIC FRACTURE
Figure 10.1.55
Figure 10.1.56
History of the fracture pore pressure prole.
Well yield before and after the hydraulic fracture process.
10.1.56

Printed on:
PERMAFROST THAWINGPIPELINE INTERACTION
10.1.6
Objectives
This example illustrates the use of the coupled temperaturepore pressure displacement modeling
capability in Abaqus/Standard to model the effects of thaw settlement that commonly occurs in
relatively warm pipelines buried in permafrost.
Thaw settlement generally occurs in shallow waters and at shore crossings where soil ice-bonded
permafrost underlies the pipeline (Xu et al., 2009). Due to the heat released from the relatively warm
pipeline, the surrounding permafrost may gradually thaw over years of operation and create a permafrost
thaw bulb. The load carrying capacity of the soil is generally believed to be reduced as a result of the
thaw formation, which could in turn (as a result of the weight of the soil above it) lead to excessive
deformation and stress in the pipeline leading to its eventual damage.
Two primary mechanisms control the behavior of the problem: heat transfer from the warm pipeline
to the surrounding permafrost leading to thawing of the permafrost and diffusion of the pore uid and
associated consolidation of the thawed soil mass surrounding the pipeline. The analysis is carried out
utilizing two approaches. The rst approach is a sequential one, in which the permafrost melting process
is modeled as a pure heat transfer analysis in Abaqus, and the resulting temperature is used to drive a
soils consolidation analysis. The second approach is fully coupled, in which the heat transfer and soils
consolidation problems are solved in a fully coupled manner in a single analysis. This approach uses
the coupled temperaturepore pressure elements that solve for the temperature eld as a nodal degree of
freedom in addition to the displacement and the pore pressure elds. The results obtained using these
two approaches are compared, and the advantages of using a fully coupled approach are discussed. The
model presented here has been adapted from Xu et al. (2009), where the rst approach alone was used
to investigate this problem.
Geometry
Figure 10.1.61 shows the geometry of the model consisting of two outer sections of sand separated by
an inner section of clay, which approximates the nonuniform soil properties that are often encountered
in practice along a pipeline. The geometry assumes symmetry about the right vertical plane in the gure,
and the dimensions of the soil mass are as shown. The pipe diameter and thickness are assumed to be
0.4 m (15.76 in) and 0.096 m (3.8 in), respectively, and it is assumed to be buried 2.5 m below the soil
surface.
10.1.61

Printed on:
Materials
The material properties assumed for this analysis are shown in Table 10.1.61, Table 10.1.62,
Table 10.1.63, and Table 10.1.64 for clay, sand, steel, and water (pore uid), respectively. The
symbols , , , , , , , , and represent the Youngs modulus, Poisons ratio, thermal
conductivity, density, specic heat, latent heat, thermal expansion coefcient, permeability, and void
ratio, respectively. The elastic and the pore uid ow properties of the soil are assumed to depend
on whether the soil is frozen (more precisely, whether the water in the soil is frozen) or thawed. The
soil is assumed to be signicantly stiff and impermeable when frozen compared to the corresponding
properties when thawed. The properties in the frozen state and the thawed state are distinguished
from each other using the symbols F and T, respectively, next to a property symbol. This approach
approximately captures the effects of phase change of the pore uid on the mechanical properties of
the soil and can be improved upon with a more detailed modeling of the physics of the phase change
process. In the heat transfer part of the sequential approach, the thermal properties are assumed to be
weighted averages of the properties of the individual constituents (clay or sand, and water). The latent
heat of fusion from ice to water is utilized to dene the phase change of the soil. The pipe is assumed
to be made out of steel, with the mechanical behavior dened using a linear elastic model along with
thermal expansion. Inelastic response of the pipe material, such as plasticity and creep, can be easily
included in the model if needed. Both the soil and the pore uid are also assumed to undergo thermal
expansion associated with the temperature variation. Thermal expansion is an important part of the
overall physics of the problem, with the differential expansion of the pore uid relative to pores in the
soil often determining the initial rise in the pore pressure.
Initial conditions
The initial temperature is assumed to be 5C everywhere in the model. The initial void ratio, pore
pressure, and saturation are assumed to be 1.8, 0.0, and 1.0, respectively, everywhere in the model. Thus,
the pore uid ow is assumed to occur under fully saturated conditions. Initial effective geostatic stress
in the vertical direction is dened as a function of the depth below the soil surface. The stress in the
horizontal direction is assumed to be a fraction of the stress in the vertical direction.
All the outer surfaces of the model (except the top surface) are assumed to be restrained from moving
in a perpendicular direction. Following Xu et al. (2009), the model assumes the ow of heat and uid
across all the outer surfaces (except the top surface) to be zero. The top surface allows free drainage
(zero pore pressure). The axial displacement at both ends of the pipe is restrained.
In the consolidation part of the analysis in the sequential approach, the soil is modeled using pore pressure
elements with displacement and pore pressure degrees of freedom, while the pipe is modeled using
solid continuum elements using displacement degrees of freedom only. In the coupled analysis, the soil
is modeled using the coupled elements with displacement, pore pressure, and temperature degrees of
10.1.62

Printed on:
freedom, while the pipe is modeled using coupled elements with displacement and temperature degrees
of freedom only. The soil and the pipe share nodes at the interface. An alternate approach would model
this interface utilizing contact constraints; however, this approach has not been used because contact is
not supported with the coupled temperaturepore pressure elements.
Analysis types
The sequential approach utilizes a heat-transfer analysis rst to determine the temperature prole. This
temperature prole is subsequently used to drive the mechanical part of the simulation that does not model
heat transfer. The latter consists of two steps: a geostatic analysis, followed by a soils consolidation
analysis. On the other hand, the coupled approach solves the thermal and mechanical problem in a fully
coupled manner and consists of two steps: a geostatic analysis, followed by a soils consolidation analysis.
Mesh design
The heat transfer part of the sequential approach uses rst-order brick (DC3D8) elements, while the
consolidation part of the sequential approach uses rst-order reduced-integration (C3D8RP and C3D8R)
elements for the soil and pipe regions, respectively. The coupled analysis uses rst-order reducedintegration fully coupled (C3D8RPT and C3D8RT) elements for the soil and pipe regions, respectively.
Loading
Inside the pipe the uid is assumed to be owing at 20C. A constant convection coefcient on the inner
pipe wall is assumed. The top surface of the soil is subjected to a sinusoidal temperature variation ranging
between 5C and 5C. The above boundary condition approximately represents the season variation of
temperature at the soil surface. The weight of the pipe is considered as a load and is modeled as a body
force in the vertical direction.
While the physics of this problem will likely span two or more decades, the present analysis is
carried out for a period of one year in order to present the basic framework for an analysis of this type.
Solution controls
The simulations use solution controls to specify a nondefault initial value of the time average pore
uid ux. The default choice may not be appropriate in situations such as those encountered in the
present problem where the uid velocities are, in relative terms, lower compared to typical ux values
encountered for other elds (such as displacements or rotations). Without the above specication, the
increments would be treated as linear from the viewpoint of the continuity equation. In other words,
without using solution controls to specify a nondefault initial value of the time average pore uid ux, the
pore uid part of the incrementation will be treated as linear. Consequently, the continuity equation would
be assumed to have been satised at the rst iteration itself, without performing any further iterations to
compute corrections to pore pressure.
Figure 10.1.62 shows a contour of the temperature eld around the buried pipe after a period of one
year. The temperature prole can be used to determine the thaw bulb around the pipe. The results are
10.1.63

Printed on:
obtained from the heat transfer part of the sequential approach; the coupled analysis produces similar
results. The thaw bulb represents the region where the permafrost has thawed and the temperature is
above freezing. Figure 10.1.62 shows the mesh for the whole model along with the contour plot of the
uid velocities. The zero pore pressure boundary condition on the top surface allows the seepage of pore
uid across this surface. Clay is signicantly less permeable compared to sand; and the effect can be
observed in Figure 10.1.63, the ow across the top surface in the clay section is less compared to the
ow in the sand section.
Figure 10.1.64 shows the temperature history at a point (node number 1591) in the soil in the
vicinity of the pipe. The two curves were obtained from the uncoupled (sequential) and coupled analysis.
The temperature results are, in relative terms, close to each other before thawing. However, upon thawing
the results from the coupled analysis predict a higher temperature. The latter may be explained by
recalling that unlike the coupled analysis, the heat transfer part of the uncoupled (sequential) analysis
does not account for heat transfer associated with pore uid ow. The results from both the sequential
and coupled analyses predict that the temperature does not increase signicantly for a certain period of
time after it reaches 0C. This behavior is due to phase change of the pore uid as a result of thawing;
all the available heat is used up as the latent heat of fusion.
Figure 10.1.65 shows the normalized (by the maximum vertical displacement from the uncoupled
analysis) vertical displacement as a function of length along the span of the pipe for both the uncoupled
(sequential) and coupled analyses. The uncoupled analysis predicts a higher peak displacement compared
to the coupled analysis. This can be attributed to the difference in temperatures at points in the vicinity
of the pipe as predicted by the two analyses and the associated effects on the elastic stiffness of the
surrounding soil mass. It is particularly interesting to note that a fully coupled analysis can be used as a
basis for a less conservative design as the predicted displacements and associated stresses are lower.
Input files
buried_pipe_permafrost_ht.inp
buried_pipe_permafrost_consol.inp
buried_pipe_permafrost_ctup.inp
Heat transfer part of the sequential analysis.

Consolidation part of the sequential analysis.
Fully coupled analysis.
References
Coupled pore uid diffusion and stress analysis, Section 6.8.1 of the Abaqus Analysis Users
Manual
Geostatic stress state, Section 6.8.2 of the Abaqus Analysis Users Manual
*GEOSTATIC
*SOILS
10.1.64

Printed on:
Other
Xu, J., B. Abdalla, A. Eltaher, and P. Jukes, Permafrost Thawing-Pipeline Interaction Advanced
Finite Element Model, Proceedings of the ASME 2009 28th International Conference on Ocean,
Offshore, and Arctic Engineering, OMAE200979554, 2009.
Table 10.1.61
Properties for clay (F-frozen state, T-thawed state).
Property
Value
(F)
1.03421 1011 Pa
(T)
6.89476 105 Pa
0.30
2.0 J/sec mC
1602.0 kg/m3
1381.0 J/kg C
9.67 106 /C
1.0 1014 m/sec
(F)
at
(T)
1.0 108 m/sec
at
(T)
2.0 108 m/sec
Table 10.1.62
Properties for sand (F-frozen state, T-thawed state).
Property
Value
(F)
1.03421 1011 Pa
(T)
1.03421 107 Pa
0.30
2.0 J/sec mC
1602.0 kg/m3
1381.0 J/kg C
5.60 106 /C
(F)
1.0 1013 m/sec
10.1.65

Printed on:
Property
Value
at
(T)
1.0 107 m/sec
at
(T)
2.0 107 m/sec
Table 10.1.63
Property
Properties for steel.

Value
2.0 1011 Pa
0.30
52.0 J/sec mC
7832.0 kg/m3
434.0 J/kg C
12.0 106 /C
Table 10.1.64
Property
Properties for water (pore uid).

Value
0.58 J/sec mC
1000.0 kg/m3
4186.0 J/kg C
(at 5C)
51.0 106 /C
(at 0C)
51.0 106 /C
(at 1C)
16.67 106 /C
(at 4C)
0.0
(at 10C)
29.33 106 /C
10.1.66

Printed on:
Pipe
Sand
2.5 m
Clay
Sand
15 m
30 m
120 m
Figure 10.1.61
Pipeline buried in permafrost (half-symmetry model).
NT11
+1.99e+01
+1.76e+01
+1.53e+01
+1.30e+01
+1.07e+01
+8.44e+00
+6.14e+00
+3.85e+00
+1.55e+00
7.42e01
3.04e+00
5.33e+00
7.63e+00
Figure 10.1.62
Thaw bulb around the relatively warm pipe after one year.
10.1.67

Printed on:
Figure 10.1.63 The meshed buried pipeline model showing the

contour of uid velocities. The relatively impermeable clay region
exhibits low uid velocities (blue-colored contour in top center section).
10.1.68

Printed on:
8.0
Temperature
4.0
0.0
4.0
0.
Coupled analysis
Uncoupled analysis
5.
10.
15.
20.
25.
30.
[x1.E6]
Time
Figure 10.1.64
Temperature history in the soil near the pipe.
Normalized vertical displacement
0.5
Coupled analysis
Uncoupled analysis
0.6
0.7
0.8
0.9
1.0
0.
40.
80.
120.
Distance along length of pipe

Figure 10.1.65 Normalized vertical displacement
along the length of the pipe.
10.1.69

Printed on:
Abaqus/Aqua ANALYSES
11.
Abaqus/Aqua Analyses
Abaqus/Aqua analyses, Section 11.1

Printed on:
Abaqus/Aqua ANALYSES
11.1
Abaqus/Aqua analyses
Jack-up foundation analyses, Section 11.1.1

Riser dynamics, Section 11.1.2
11.11

Printed on:
JACK-UP FOUNDATION ANALYSES
11.1.1
Abaqus/Aqua
This example simulates a jack-up rig on a sand foundation subjected to alternating wind loading.
Geometry and model
The modela simplied planar model for the analysis of a multiple leg, portal frame-type structureis
intended for the analysis of 3-leg jack-up rigs with shallow foundation supports. Figure 11.1.11 is a
schematic of a 3-leg jack-up, as represented by the model. The jack-up hull is assumed to be rigid and
triangular, and the connection between the hull and the legs is also taken to be rigid. The jack-up has
two windward legs and one leeward leg; the model is projected onto the vertical symmetry plane that
passes through the leeward leg and between the windward legs. Elastic beam columns are used to model
both the upper and lower segment of each leg. The soil model is chosen to be macro-yield sand. Three
degrees of freedomvertical, horizontal, and rotationalare assumed at each spud can at the base of
each leg. Mass is assumed to be concentrated at the center of the hull. The horizontal degree of freedom
at the center is assumed to represent the motion of the rig for analysis purposes. Wind loading on the rig
is applied as a horizontal force above the center of gravity of the hull.
The leg segments are modeled using B21 elements, and the *BEAM GENERAL SECTION option
is used to dene the structural properties of the beam. The interaction between the spud can and the
soil is modeled through JOINT2D elements and the *JOINT ELASTICITY and *JOINT PLASTICITY
options. Rigid beam elements, RB2D2, are used to model the rigid hull.
The dimensions of the rig and the material properties of the sand and the spud can are as follows
(force units are in kN, and length units are in meters):
Leg length upper segment
Leg length lower segment
Leg EI upper segment
Leg EI lower segment
Leg AE upper segment
Leg AE lower segment
Leg GA upper segment
Leg GA lower segment
Horizontal distance from platform
center of gravity to leeward leg
Horizontal distance from platform
center of gravity to windward leg
Spud can diameter
11.1.11

Printed on:
49.4
13.5
2.7 108
2.7 109
2.2 108
2.2 109
8.1 107
8.1 108
23.4
11.7
10.9
Spud can cone angle

Foundation preload per spud can
Foundation tensile capacity
Operational vertical load (weight)
Vertical distance from center of gravity
to load application point
Soil submerged unit weight
Soil friction angle
Soil Poissons ratio
Foundation elastic shear moduli,
Constant coefcient,
Constant coefcient,
180
50600
0
62700
7.1
10.0
33
0.2
5.14 104
3.87 103
2.04 103
1.0
0.5
The base nodes of the JOINT2D elements are always xed. The required preload is applied to each
spud can using the *INITIAL CONDITIONS option. In the rst step the weight loading is applied at the
center of gravity of the hull. The rig is then subjected to an alternating horizontal wind loading applied at
the specied location above the center of gravity of the hull. The load is applied by using the *CLOAD
option.
The rig is loaded from zero to 5370 kN, unloaded to zero and then to 6440 kN in the opposite
direction, reloaded to 9130 kN in the initial direction, unloaded and reloaded to 9770 kN in the opposite
direction, and unloaded to zero again. Each of these loadings is done in a separate step and is ramped
from zero to the specied magnitude at the end of the step.
The estimated load path for the leeward spud can foundation is plotted in a graph of equivalent horizontal
load,
, versus
The plot is shown in Figure 11.1.12 and is in good
agreement with the load path predicted by an independent analysis, as detailed in the reference below.
The moment-horizontal load response (i.e.,
versus H) for the leeward spud can foundation, shown
in Figure 11.1.13, compares well with the independent analysis.
Input file
jackup.inp
Input data for this example.
11.1.12

Printed on:
Reference
Wong, P. C. and J. D. Murff, Dynamic Analysis of Jack-Up Rigs Using Advanced Foundation
Models, Proceedings, 13th International Conference on Offshore Mechanics and Arctic
Engineering (OMAE), vol. 2 - Safety and Reliability, Houston, pp. 93109, February 1994.
rigid deck
horizontal load
center of gravity
one leg
two legs
beam columns
with EI, AE, GA
JOINT2D elements
Figure 11.1.11
Schematic representation of jack-up rig.
11.1.13

Printed on:
Figure 11.1.12
Figure 11.1.13
Load path for leeward spud can.
Moment versus horizontal load for leeward spud can.
11.1.14

Printed on:
RISER DYNAMICS
11.1.2
RISER DYNAMICS
Abaqus/Aqua
Pipelines extending from the sea oor to the ocean surface (risers) are subject to many types of load: selfweight, buoyancy, internal and external pressure, tensile forces arising from surface moorings, current drag,
and oscillatory loads resulting from wave motion. The response of a riser to these loads is complex, and the
difculty of such analysis is heightened by the relative length of such pipelines (deep water risers). In this
example a riser is analyzed under conditions specied by the American Petroleum Institute for comparison
of drilling riser analyses (API BULLETIN 2J, 1977), and the results are compared with the results shown in
that publication.
Geometry and model
The riser is shown in Figure 11.1.21. Its length is 463.3 m (1520 ft), and it stands in 448.1 m (1470 ft)
of water. The outer diameter of the riser is 405 mm (1.33 ft), and it has a wall thickness of 15.88 mm
(0.0521 ft). The pipeline is made of steel, with a Youngs modulus of 206.8 GPa (4.32 109 lb/ft2 ) and a
density of 11508.685 kg/m3 (22.332 lb-s2 /ft4 ). The riser is modeled with 10 beam elements of type B21.
No mesh convergence studies have been performed; hence, more elements may be required for accurate
prediction of the stress in the riser.
Loading
The riser has a weight of 2575 N/m (176.36 lb/ft) and is loaded by a top tension of 2.224 MN (5 105 lb).
Drag loading is applied by a steady current owing by the riser with a velocity distribution varying
linearly from 0.257 m/s (0.844 ft/s) at the mean water level to zero at the base of the riser. The coefcients
in Morisons equation are transverse drag coefcient ( ) 0.7, tangential drag coefcient ( ) 0.0, and
transverse inertia coefcient (
) 1.5.
The effective outer diameter for the drag calculations is 0.66 m (2.167 ft). Waves of peak to trough
height 6.1 m (20 ft) travel across the water surface with a period of 9 seconds; these are modeled with
the Airy wave theory provided in the *AQUA option (Abaqus/Aqua analysis, Section 6.11.1 of the
Abaqus Analysis Users Manual). The density of the uid is taken to be 1021 kg/m3 (1.982 lbs2 /ft4 ). In
Abaqus/Aqua, user subroutine UWAVE can be used to specify user-dened wave kinematics. We illustrate
this capability by repeating this analysis with a user-specied Airy wave theory that is identical to the
built-in Airy wave option in Abaqus/Aqua.
Boundary conditions
The base of the riser is gimballed, supporting no moments. The top of the riser has two motions
prescribed: an initial offset of 13.716 m (45 ft) from the vertical position of the riser and a sinusoidal
motion about this static conguration, representing the surge of a vessel attached to the riser, with peakto-peak amplitude of 1.22 m (4 ft) and a period of 9 seconds. The vessel surge angle is 15 out of phase
with the surface waves. The phase angle,
, for the Airy wave denition provides an arbitrary choice
11.1.21

Printed on:
RISER DYNAMICS
of origin in time for the vertical displacement of a uid particle. Based on the initial offset and vessel
surge angle, this angle is set to 54.026, where the negative sign indicates that the wave lags behind the
vessel surge.
Analysis
The analysis is done in two steps. The rst is the static step, in which the top tension is applied and the
riser is moved from the vertical to its offset position by specifying the necessary horizontal displacement
at the top of the pipeline. The top tension is 2.224 MN (5 105 lb).
In the second step, which is a dynamic step, the time increment is chosen as a xed value of
0.125 second. The prescribed displacement at the top of the riser has a 9-second period, so this time
step should provide reasonably accurate time integration once the higher modes are damped out by the
uid drag. The half-increment residual values calculated by Abaqus provide a measure of accuracy of
the solution, and these values are typically of order 4.4 kN (1000 lb). Since these values are smaller than
typical actual forces, they suggest that the time integration is reasonably accurate.
The initial static step, which moves the riser to its offset position and applies the static loads, is completed
in four increments. The rst increment requires more iterations than subsequent increments, which is
typical of this class of problem: the riser is initially unstressed and, therefore, is highly exible. After
some loading is applied, the axial tension stabilizes the system, and convergence is more rapid.
At the end of the static step the top of the riser makes an angle of 1.17 with the vertical. This value
agrees well with the value of 1.20 presented in API BULLETIN 2J (1977). The angle predicted at the
base of the riser is 2.48, which compares to 2.55 reported in the API bulletin. The slight discrepancies
are attributed to the relative coarseness of the model.
The dynamic solution is carried out for 18 seconds of response. Typically one equilibrium iteration
is required in each of the time increments. Half-increment residual values for the rst few increments
are of order 178 MN (4.0 107 lb), and at the end of the run they are of order 4.4 kN (1000 lb). This
result is typical: initially there is much high frequency content in the solution, which is reected in the
larger half-increment residual values. As the analysis proceeds, the uid drag dissipates this noise, the
solution becomes smoother, and the half-increment residual values drop accordingly.
The envelope of pipeline excursions during the second cycle of the dynamic excitation is plotted
in Figure 11.1.22, and the envelope of bending stress is shown in Figure 11.1.23. These results are in
basic agreement with those given in the API bulletin.
As expected, the results obtained by the model with the Airy wave theory implemented in user
subroutine UWAVE are identical to those due to the built-in Airy wave option.
Input files
riserdynamics_airy_disp.inp
Analysis with the Airy wave theory. User subroutine

DISP is used to prescribe the sinusoidal surge motion.
This motion could be prescribed instead through the use
of the *AMPLITUDE option. User subroutine DISP is
11.1.22

Printed on:
RISER DYNAMICS
riserdynamics_airy_disp.f
riserdynamics_wavedata.inp
riserdynamics_stokes_disp.inp
riserdynamics_stokes_disp.f
riserdynamics_airy_disp_uwave.inp
riserdynamics_airy_disp_uwave.f
exa_riserdynamics_stokes_disp_uwave.inp
exa_riserdynamics_stokes_disp_uwave.f
used to illustrate the use of this routine to prescribe a

nonzero boundary condition value.
User subroutine DISP used in
riserdynamics_airy_disp.inp.
Wave data for use in riserdynamics_airy_disp.inp.
Analysis with the Stokes wave theory.
User subroutine DISP used in
riserdynamics_stokes_disp.inp.
Analysis with the Airy wave theory implemented in user
subroutine UWAVE.
User subroutines UWAVE and DISP used in
riserdynamics_airy_disp_uwave.inp.
Analysis with the Stokes wave theory implemented in
user subroutine UWAVE.
User subroutines UWAVE and DISP used in
exa_riserdynamics_stokes_disp_uwave.inp.
Reference
American Petroleum Institute, Comparison of Marine Drilling Riser Analyses, API Bulletin 2J,
Washington, DC, January 1977.
11.1.23

Printed on:
RISER DYNAMICS
A
Geometry:
l = 463.3 m (1520.0 ft)
h = 15.2 m (50.0 ft)
d = 405 mm (1.33 ft)
t = 15.8 mm (0.0521 ft)
h
Wave and
current
direction
Material:
Youngs modulus = 206.8 GPa
9
2
(4.32 x 10 lb/ft )
density = 11508.68 kg/m3 (22.332 lb-s2/ft4)
Boundary conditions:
at O => ux = uy = 0
at A => static:
ux = a
2 t)
dynamic: ux = a + b sin(
9
a = 13.716 m (45.0 ft)
b = 610 mm (2.0 ft)
t
l
Loading:
self weight 2575.0 N/m (176.36 lb/ft)
5
top tension = 2.224 MN (5.0 x 10 lb)
Morison coefficients:
transverse drag coefficient = 0.7
tangential drag coefficient = 0.0
transverse inertia coefficient = 1.5
effective outer diameter = 0.66 m (2.167 ft)
Airy wave definition:

wave height (peak to trough) = 6.1 m (20.0 ft)
wave period = 9.0 s
Fluid velocity:
vx = 0.257 m/s (0.844 ft/s) at mean water level
vx = 0.0 at base of riser
vy = 0.0
Fluid density:
w = 1021.0 kg/m3 (1.982 lb-s2/ft4)
Figure 11.1.21
Riser problem denition.
11.1.24

Printed on:
RISER DYNAMICS
Horizontal displacement, ft
0
10
20
30
40
50
1.0
0.9
0.8
Normalized position along pipe
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
12
15
Horizontal displacement, m
Figure 11.1.22
Horizontal displacement envelope during the second cycle of the dynamic excitation.
11.1.25

Printed on:
RISER DYNAMICS
Bending stress, 103 lb/in2

1.0
-2.0
-1.0
1.0
2.0
0.9
0.8
Normalized position along pipe
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-10
-5
10
Bending stress, kPa
Figure 11.1.23 Bending stress envelope during the second

cycle of the dynamic excitation.
11.1.26

Printed on:
DESIGN SENSITIVITY ANALYSES
12.
Design Sensitivity Analyses
Overview, Section 12.1

Examples, Section 12.2

Printed on:
OVERVIEW
12.1
Overview
Design sensitivity analysis: overview, Section 12.1.1
12.11

Printed on:
DESIGN SENSITIVITY ANALYSIS
12.1.1
DESIGN SENSITIVITY ANALYSIS: OVERVIEW
This chapter contains example problems specically created to demonstrate the design sensitivity analysis
(DSA) capability in Abaqus/Design. In addition, the following example problem includes design sensitivity
analysis:
Indentation of an elastomeric foam specimen with a hemispherical punch, Section 1.1.4
Some of these examples employ a scripting command in Abaqus/CAE to aid in creating the shape
variations. The usage of this command is described below.
Using Abaqus/CAE to compute shape variations
A transitional capability for computing shape variations is provided by the internal Abaqus Scripting
Interface command _computeShapeVariations(). Using the command requires some familiarity
with the Abaqus Scripting Interface; specically, the user must understand the Abaqus object model (see
The Abaqus object model, Section 6.1 of the Abaqus Scripting Users Manual) and know how to access
the rootAssembly and partInstance objects.
The command procedure necessary to dene a shape variation requires the following sequence of
actions:
1. Create and mesh the model in Abaqus/CAE. Output the corresponding input le by selecting
JobCreate and JobWrite Input from the main menu bar in the Job module.
(In the discussion that follows it is assumed that the model is named Model-1, the part is named
Part-1, and the part instance is named Part-1-1).
2. Select ModelCopy Model from the main menu bar to copy Model-1 to Model-2, for
example. Select Model-2 from the Model list located in the context bar. This model will be used
in subsequent steps to compute the shape variation.

3. In the Part module select the part for which a shape variation must be calculated from the Part list
located under the context bar. Select FeatureEdit from the main menu bar to edit the associated
sketch. Select AddDimension and EditDimension to change the design parameter. End the
edit of the sketch, and indicate that the geometry should be regenerated automatically.
Editing the sketch will cause the mesh of Model-2 to be deleted.
4. Use the Abaqus/CAE command line or select FileRun Script to execute the commands listed
below. The _computeShapeVariations() command is accessed from the rootAssembly of

either model and requires as input the original part instance, the modied part instance, and the
name of the le to which the data lines of the shape variation option will be written. The .inp
extension will be appended automatically to the specied le name.
The following sequence of commands is appropriate for the creation of a shape variable associated
with a parameter h:
12.1.11

Printed on:
DESIGN SENSITIVITY ANALYSIS
ra1 = mdb.models['Model-1'].rootAssembly
ra2 = mdb.models['Model-2'].rootAssembly
i1 = ra1.instances['Part-1-1']
i2 = ra2.instances['Part-1-1']
ra1._computeShapeVariations(originalInstance=i1,
modifiedInstance=i2,
fileName='shape_h')
5. To compute the shape variation, the mesh of Model-1 is mapped to the changed geometry of
Model-2 and then smoothed. Examine the mapped mesh of Model-2 to verify that the mesh was
mapped as expected. (In general, any changes to the geometry should be smallaround 1%so
as to avoid difculties with the mapping of the mesh).
The shape variation is calculated simply by subtracting the initial node positions from the node
positions calculated after mapping and smoothing the mesh to the changed geometry.
6. To use the shape variation data, copy to the input le for your analysis the data written by the
_computeShapeVariations() command to the shape_h.inp le.

7. Verify the correctness of the data by running your analysis and using the Visualization module in
Abaqus/CAE to view the shape variation.
12.1.12

Printed on:
EXAMPLES
12.2
Examples
Design sensitivity analysis of a composite centrifuge, Section 12.2.1

Design sensitivities for tire ination, footprint, and natural frequency analysis, Section 12.2.2
Design sensitivity analysis of a windshield wiper, Section 12.2.3
Design sensitivity analysis of a rubber bushing, Section 12.2.4
12.21

Printed on:
DSA CENTRIFUGE
12.2.1
DESIGN SENSITIVITY ANALYSIS OF A COMPOSITE CENTRIFUGE
Abaqus/Design
Many industries use centrifuges to separate out contaminants during purication processes. Efciency of
the purication process is directly related to the speed of rotation. Consequently, the centrifuge chamber is
designed to be stiff to maintain its shape and lightweight to reduce self-stressing due to centrifugal loads. This
example uses the design sensitivity analysis capability in Abaqus/Design to examine how the key structural
responses depend on design parameters such as the thicknesses of composite laminae, the layup angles, the
density of the centrifuge end plates, and geometric imperfections.
Geometry and model
The centrifuge depicted in Figure 12.2.11 consists of a composite centrifuge chamber and aluminum
end plates. The centrifuge chamber is a cylinder 970 mm long and 175 mm in diameter. It spins about
its axis at 10000 rpm. The cylinder is a lament-wound composite (hybrid) laminate with a balanced
layup. The laminate layup is (
), where 0 represents the layup of the bers along the
length of the cylinder and = 45 is the angle of the helical layers. The axial (0) and helical layers have
a thickness of 0.15 mm and 0.5 mm, respectively. Aluminum alloy end plates are bonded at both ends
of the cylinder. The lower plate provides an attachment point for a magnetic bearing, while the top plate
supports the magnet of the inductance drive. In the region where the cylinder meets the end plates an
additional circumferential (90) outer layer of 1 mm thickness is added to the laminate. The centrifuge is
modeled using reduced-integration, 4-node shell (S4R) elements. The mesh is shown in Figure 12.2.12.
All the displacements on the perimeter of the end plate lip are constrained at one end of the cylinder, and
only the radial displacements are constrained at the other end; thus, the centrifuge is allowed to change
length freely in the axial direction under loading. A static analysis with centrifugal loading is performed.
Materials
A T800 carbon ber material is used in the 0 and 90 layers, and a HM400 material is used in the
helical layers. The density of both materials is 1600 kg/m3 . The material properties for the composites
are given in Table 12.2.11. The aluminum alloy has a Youngs modulus of E = 70 GPa, a Poissons
ratio of = 0.33, and a density of = 2800 kg/m3 .
Design parameters and normalization
The design parameters are thickness of the helical layer (THM400), thickness of the axial layer (TT800),
and angle of the helical layer (THETA). Because of the high speed of rotation of the centrifuge,
geometric imperfections can have a signicant effect on the displacements and stresses. To study this
effect, an imperfection in the form of the rst bending mode of the centrifuge is used; the magnitude
of the imperfection, ALPHA, is chosen as a shape design parameter. An imperfection in the form of a
bending mode is chosen because the maximum attainable rotation speed is known to be well-predicted
by the natural frequency of the rst bending mode of the centrifuge. The gradients of the nodal
12.2.11

Printed on:
DSA CENTRIFUGE
coordinates with respect to ALPHA required to carry out the sensitivity analysis are obtained from a
*FREQUENCY analysis. Figure 12.2.13 shows the rst bending mode of the centrifuge. Though
bending predominates in this mode, there is also a small twisting component.
All plotted sensitivity results are normalized to enable comparison between the parameters. The
normalization is carried out by multiplying the response sensitivity by the value of the parameter and
dividing by the maximum value of the response. For example, the sensitivity of the stress component
S11 with respect to the design parameter THM400 is normalized by rst multiplying the sensitivity by
the value of the THM400 parameter and then dividing by the maximum S11 value found in the model.
For the shape design parameter ALPHA an estimated imperfection of 0.1% of the total length of the
cylinder (ALPHA = 1 mm) is used while computing the normalized sensitivities.
Figure 12.2.14 shows the deformed shape of the centrifuge at 10000 rpm. The centrifuge contracts in
the axial direction and bulges outward radially. Because of the lap joint where the composite cylinder
meets the end plate, some bending is seen at both ends.
The normalized sensitivity of the radial displacement along the length of the composite cylinder
(between the two end plates) is plotted in Figure 12.2.15 for all the design parameters except the shape
design parameter. The plot shows that the radial displacement has negative sensitivity to THETA and
TT800. An increase in the layup angle of the helical layer or an increase in the thickness of the axial
layers will stiffen the cylinder and reduce the radial displacement. The positive sensitivity of the radial
displacement to THM400 indicates that the added self-induced centrifugal load due to the increase in the
mass of the helical layers will more than negate any advantage gained in the stiffness.
Figure 12.2.16 plots the sensitivity of the radial displacement to the shape design parameter
ALPHA. Since the sensitivity is not axisymmetric for this design parameter, it is plotted for every
meridian in 45 increments counterclockwise around the circumference beginning at the 12 plane.
The radial displacement sensitivities are obtained from the global Cartesian displacement sensitivities
output by Abaqus via a two-dimensional vector transformation at each meridional position. The largest
sensitivities in Figure 12.2.16 are almost two orders of magnitude higher than those observed in
Figure 12.2.15. This means that the geometric imperfections have a relatively large effect on the radial
displacement and that the centrifuge manufacturing process must have a tight tolerance on axial shape
imperfections.
The dominant section force in the structure is in the hoop direction of the cylinder. Figure 12.2.17
shows the contour plot of section force in the hoop direction, and Figure 12.2.18 shows the normalized
sensitivities of section force in the hoop direction plotted as a function of the position along the length
of the cylinder. As expected, only the design parameters that affect the mass (TT800 and THM400)
have nonzero sensitivities with THM400 being more sensitive because of its larger thickness. As the
centrifuge is allowed to contract freely in the axial direction, the net section force in the axial direction is
zero. However, the stresses in the axial direction are not zero. Figure 12.2.19 shows the contour plot of
the sensitivity of the ber stress (S11) in the helical layer with respect to the shape design parameter. To
understand in detail how the stresses in the laminate are affected by the design parameters, normalized
sensitivities of the ber stress (S11) through the thickness of the cylinder wall at a point midway along
the length of the cylinder are plotted in Figure 12.2.110. S11 through the thickness of the wall is also
12.2.12

Printed on:
DSA CENTRIFUGE
plotted after dividing it by the maximum value of S11 in the model. The plot shows that the axial layers
are under compression and the helical layers are under tension. S11 has positive sensitivity to TT800:
increasing the thickness of the axial layers will reduce the compressive stress in the axial layers and
increase the tensile stress in the helical layers. S11 has negative sensitivity to THM400: increasing the
thickness of the helical layers will reduce the stress in the helical layers and increase the stress in the axial
layers. Increasing the angle of the helical layers will reduce the stress in the axial layers since S11 in the
axial layers has a small positive sensitivity to THETA. The sensitivity of S11 to ALPHA is analogous to
the sensitivity of the radial displacement to ALPHA around the circumference of the cylinder.
The sensitivities can be used to compute the change that would be required in the design parameters
to achieve a particular change in response or to assess the change in the response that would result from
a change in the design parameters. Consider, for example, the following objectives: (a) to reduce the
compressive stress in the axial layer by 10% and (b) to determine the maximum compressive stress in
the axial layer caused by a specied magnitude of the shape imperfection ( ALPHA = 0.6).
a. The compressive ber stress in the axial layer is 26.38 MPa. Figure 12.2.110 indicates that the
most effective way to achieve the desired reduction is to increase the thickness of the axial layer.
The required increase is given by
An analysis of the centrifuge with TT800 = 0.15 + 0.0378 = 0.1878 mm shows the compressive
stress in the axial layer to be 24.05 MPa, which is close to the desired value of 23.74 MPa.
b. The maximum compressive stress in the axial layer caused by the specied shape imperfection can
be obtained from
An analysis using the imperfect geometry shows the maximum compressive stress in the axial layer
to be 34.3 MPa, which is close to the predicted value of 34.2 MPa.
Input files
dsacentrifuge_freq.inp
dsacentrifuge.inp
dsacentrifuge_node_elem.inp
Frequency analysis of the centrifuge.

Sensitivity analysis of the centrifuge subjected to
centrifugal loads.
Node and element denition.
12.2.13

Printed on:
DSA CENTRIFUGE
Table 12.2.11
Material
E1 (MPa)
E2 (MPa)
T800
177000.
14920.
HM400
233967.
14778.
Composite material properties.

G13 (MPa)
G12 (MPa)
G32 (MPa)
.21
5700.
5700.
5630.
.032
5777.
10191.
5634.
29 mm
3.45 mm
14 mm
2 mm
end plate
20 mm
2.45 mm
composite cylinder
1000 mm
20 mm
3 mm
13.5 mm
13.5 mm
Centrifuge Chamber
Figure 12.2.11
Centrifuge chamber.
12.2.14

Printed on:
DSA CENTRIFUGE
Figure 12.2.12
Centrifuge model using S4R elements.
Figure 12.2.13
First bending-dominated mode of the centrifuge.
Figure 12.2.14
Deformed shape of the centrifuge; deformation scaled 400 times.
12.2.15

Printed on:
DSA CENTRIFUGE
d_U2_THETA
d_U2_THM400
d_U2_TT800
Figure 12.2.15 Normalized sensitivity of the radial

displacement along the length of the cylinder.
d_Ur_0_ALPHA
d_Ur_135_ALPHA
d_Ur_180_ALPHA
d_Ur_225_ALPHA
d_Ur_270_ALPHA
d_Ur_315_ALPHA
d_Ur_45_ALPHA
d_Ur_90_ALPHA
Figure 12.2.16 Normalized sensitivity of the radial displacement to the shape design parameter
ALPHA. Sensitivities are plotted around the circumference in 45 increments.
12.2.16

Printed on:
DSA CENTRIFUGE
SF, SF2
(Ave. Crit.: 75%)
+9.887e+01
+8.891e+01
+7.894e+01
+6.897e+01
+5.900e+01
+4.904e+01
+3.907e+01
+2.910e+01
+1.913e+01
+9.164e+00
-8.032e-01
Figure 12.2.17
Section force in the hoop direction.
d_SF2_ALPHA
d_SF2_THETA
d_SF2_THM400
d_SF2_TT800
Figure 12.2.18
Normalized sensitivity of the section force in the hoop direction.
12.2.17

Printed on:
DSA CENTRIFUGE
d_S_ALPHA, S11
(Ave. Crit.: 75%)
+3.752e+00
+3.125e+00
+2.498e+00
+1.871e+00
+1.244e+00
+6.169e-01
-1.013e-02
-6.372e-01
-1.264e+00
-1.891e+00
-2.518e+00
-3.145e+00
-3.772e+00
Figure 12.2.19 Sensitivity of the ber stress, S11, in the helical

layer to the shape design parameter ALPHA.
S11
d_S11_ALPHA
d_S11_THETA
d_S11_THM400
d_S11_TT800
Figure 12.2.110 Normalized sensitivity of the ber stress plotted

through the thickness of the cylinder wall for element 590.
12.2.18

Printed on:
DESIGN SENSITIVITY ANALYSIS OF A TIRE
12.2.2
DESIGN SENSITIVITIES FOR TIRE INFLATION, FOOTPRINT, AND NATURAL

FREQUENCY ANALYSIS
Abaqus/Design
The purpose of this example is to demonstrate the application of design sensitivity analysis (DSA) to tire
problems. The base tire model and analysis is the same as that described in Symmetric results transfer for
a static tire analysis, Section 3.1.1. The design sensitivity analysis shows the effects on responses such as
contact pressure and natural frequencies of three important design parameters: the thickness of the side wall,
the elastic modulus of the belt rebar material, and the elastic modulus of the carcass rebar material. This
example demonstrates that the DSA technique in Abaqus/Design can be used effectively for highly nonlinear
analyses including features such as viscoelasticity, contact, and rebar.
Geometry and model
The geometry and model information for this example is identical to that for Symmetric results transfer
for a static tire analysis, Section 3.1.1. However, since the *SYMMETRIC MODEL GENERATION
and *SYMMETRIC RESULTS TRANSFER options are not available in a design sensitivity analysis,
the full three-dimensional model is constructed and analyzed in one model. In addition, the convergence
tolerance on the residual is tightened to improve the accuracy of the tangent stiffness, thereby providing
more accurate sensitivities (see Design sensitivity analysis, Section 16.1.1 of the Abaqus Analysis
Users Manual). In addition to the ination step and the two footprint analysis steps (displacement
control and load control), a frequency extraction is appended as the last step. A sensitivity analysis is
performed in each step with contact pressure as the design response in the static steps and frequency as
the design response in the nal frequency extraction step.
Three primary design parameters are chosen for this problem. The rst is the nominal thickness,
, of the tire in the region of the sidewall. The specic region affected by is shown in the symmetric
portion of the tire cross-sectional view in Figure 12.2.21. This region consists of one layer of elements,
and the thickness in the discretized model is taken as the distance between an outer node and the
corresponding inner node. The thickness between each of the pairs of nodes in this region is related to
the nominal thickness by
, where f is dened as the ratio
. Thus, a change in the design
parameter causes the thicknesses to change proportionally (the actual value of is not important;
however, for the purposes of normalization of the results, as discussed below, a value of 1 is used). A
constraint on the change in the nodal coordinates due to a change in the design parameter is specied such
that the outer nodes stay xed and the inner nodes move inward along the original vectors connecting
the outer nodes to the inner nodes. Figure 12.2.21 depicts the shape of the tire cross-section that would
result from a 50% change in nominal thickness. The other two design parameters chosen are the elastic
modulus of the belt rebar material, Ebelt , and the elastic modulus of the carcass rebar material, Ecarc .
The density of the rubber material, rubber , is also included as a design parameter for demonstration and
verication purposes.
12.2.21

Printed on:
Perturbation sizes for the finite differencing operations
As documented in Design sensitivity analysis, Section 16.1.1 of the Abaqus Analysis Users Manual,
Abaqus uses the semi-analytic approach to compute sensitivities, and this approach involves nite
differencing computations at the element level. By default, Abaqus automatically determines for each
element the perturbation sizes of the design parameters to be used in the nite differencing computations
through a heuristic perturbation sizing algorithm. The perturbation sizes for the elements with the most
contribution to the sensitivities are written to the message le. Since this algorithm can be expensive,
the *DSA CONTROLS option is provided to allow the user to directly supply perturbation sizes. If
appropriate perturbation sizes are not known in advance, a smaller problem can be run with the default
perturbation sizing algorithm, and the perturbation sizes determined by Abaqus can then be inserted
into the larger problem.
This technique for obtaining perturbation sizes is adopted in this example. A smaller axisymmetric
problem (dsatire_axi_half.inp) with one ination (static) step and one frequency step is run. The
perturbation sizes reported in the message le for each step of the smaller problem are subsequently
specied in the full model (dsatire.inp) using the *DSA CONTROLS option.
Normalization of sensitivity results
In the subsequent plots of results, the sensitivities are normalized so that they can be compared
side-by-side. Except as noted, the contour plots of contact pressures (CPRESS) and corresponding
sensitivities are normalized by dividing by the maximum contact pressure at the end of the last static
step, Cmax . In addition, the sensitivities of CPRESS are multiplied by the value of the design parameter.
For example, the sensitivity of CPRESS with respect to , d_CPRESS_tNominal, is normalized as
d_CPRESS_tNominal/Cmax . The eigenfrequency (EIGFREQ) sensitivities are divided by the rst
eigenfrequency value and, as for CPRESS, multiplied by the value of the appropriate design parameter.
The contact pressure on the tire footprint can be considered an important factor in tire handling and wear
properties. As such, the results of the sensitivity analysis are discussed in terms of the contact pressure
distribution on the tire footprint. Figure 12.2.22 shows the actual values of the contact pressure on the
full tire model looking from below at the end of the static footprint analysis. Figure 12.2.23 shows
the normalized contact pressure distribution on a blown-up region of the full tire. As can be seen from
these plots, the contact pressure is less in the center of the footprint than in the surrounding region. The
maximum contact pressure occurs at node 2645, and the center of the footprint is at node 3055. These
nodes are indicated on the plots, and Figure 12.2.24 shows the time history of CPRESS for these nodes.
The objective is to use the sensitivity results to determine how the design parameters can be modied
to distribute the contact pressure more evenly so that the center of the tire picks up more of the load.
Figure 12.2.25 to Figure 12.2.27 show the distributions of the normalized contact pressure sensitivity
for each of the design parameters. Large variations in the sensitivities away from the center of the
footprint are observed in the sensitivity contour plots. These variations can be attributed to the high
gradients in contact pressure in the region surrounding the center of the footprint (see Figure 12.2.23),
12.2.22

Printed on:
because even small changes in the footprint size (due to small changes in the design parameters) can lead
to relatively large changes in the contact pressure. These sensitivity plots show that to increase the contact
pressure at node 3055 (footprint center), the value of Ecarc should be increased, the value of Ebelt should be
increased, and the value of should be decreased. However, based on the magnitudes of the normalized
sensitivities, Ebelt has the most inuence on the contact pressure at the center node. Accordingly, a new
design is investigated in which the design parameter Ebelt is increased by 5%. Figure 12.2.28 shows
the distribution of contact pressure for the new design, and Figure 12.2.29 shows the time history of
CPRESS at nodes 2645 and 3055 for the new design. These gures indicate that the contact pressure
has increased in the center of the tire without appreciably affecting the surrounding contact pressure
distribution; the actual increase is 3.20%. The predicted increase based on the (rst-order) sensitivities
is 2.32%, which is reasonably close to the actual increase considering the high degree of nonlinearity in
this problem.
Figure 12.2.210 to Figure 12.2.214 show the rst ve modes of the tire. The study of the
frequency sensitivities provides insight into the dynamic behavior of the design. For example, we can
conclude from Figure 12.2.215 and Figure 12.2.217 that the frequency of mode 2 is most sensitive
to the sidewall shape, , and less sensitive to the Youngs modulus of the sidewall reinforcement,
Ebelt . In retrospect, this makes good physical sense because mode 2 is primarily shear of the sidewall
(see Figure 12.2.215), but this conclusion may have been much more difcult to formulate without
the sensitivity information.. Figure 12.2.216, Figure 12.2.217, and Table 12.2.21 show the values
of the eigenfrequency and corresponding normalized sensitivities for the rst ve modes of the tire.
The eigenfrequencies for the new design (5% change in Ebelt as discussed above) are the same (to ve
signicant gures) as for the original model. This behavior is accurately predicted in Figure 12.2.217,
where it can be seen that the eigenfrequencies are essentially independent of this design parameter.
Given the highly nonlinear nature of this problem, the user is cautioned against using the sensitivities
beyond their useful limit. The sensitivities are rst-order derivatives; therefore, using them to predict
large changes in the design parameters is not valid, since higher-order terms are not considered. In
addition, using sensitivity results to predict the outcome of simultaneous changes to the design parameters
assumes that superposition is valid, which again is true only for small changes in the design parameters.
For example, if the design parameters Ecarc and Ebelt are increased by 5% and is decreased by 5%, the
contact pressure at the center of the footprint increases by 4.35%, which is nearly twice the predicted
change of 2.42%. This implies that a simultaneous change of 5% is too large for predicting the net effect
on the contact pressure based on the sensitivities.
Input files
dsatire_axi_half.inp
dsatire_axi_half_node.inp
dsatire_axi_half_psv.inp
dsatire.inp
dsatire_model.inp
dsatire_psv.inp
Axisymmetric model with ination and frequency

analysis for obtaining perturbation sizes.
Nodal coordinates for axisymmetric model.
Parameter shape variation data for axisymmetric model.
Full model including ination, footprint, and frequency
analysis.
Model data for full model.
Parameter shape variation data for full model.
12.2.23

Printed on:
Table 12.2.21
Mode
Eigenfrequency
Eigenfrequency sensitivities for the rst ve modes.

Normalized eigenfrequency sensitivity with respect to:
rubber
Ebelt
Ecarc
51.6
1.56E1
5.00E1
1.84E3
1.39E2
52.9
2.18E1
5.13E1
1.81E5
4.97E3
58.7
1.45E1
5.69E1
2.27E3
1.53E2
60.6
1.34E1
5.88E1
1.75E3
1.54E2
84.1
1.69E1
8.15E1
3.48E3
1.86E2
Figure 12.2.21 Effect of 50% change in design parameter on tire

cross-section geometry (dashed line represents original geometry).
12.2.24

Printed on:
CPRESS
STREAD/SROAD
+3.169e+05
+2.927e+05
+2.685e+05
+2.443e+05
+2.200e+05
+1.958e+05
+1.716e+05
+1.474e+05
+1.232e+05
+9.902e+04
+7.482e+04
+5.061e+04
+2.641e+04
+0.000e+00
Figure 12.2.22
Node 2645
Node 3055
Contact pressure distribution on full tire at end of footprint (static) analysis.
Normalized CPRESS
+1.00e+00
+9.17e-01
+8.33e-01
+7.50e-01
+6.67e-01
+5.83e-01
+5.00e-01
+4.17e-01
+3.33e-01
+2.50e-01
+1.67e-01
+8.33e-02
+0.00e+00
Node 2645
Node 3055
Figure 12.2.23 Normalized contact pressure distribution on

blown-up view of tire footprint at end of footprint analysis.
12.2.25

Printed on:
Normalized CPRESS at Node 2645

Figure 12.2.24 Normalized contact pressure history for center node (3055) and node with
maximum contact pressure at end of footprint analysis (2645).
Normalized d_CPRESS_Ecarc
+5.28e-02
+4.49e-02
+3.70e-02
+2.90e-02
+2.11e-02
+1.31e-02
+5.19e-03
-2.75e-03
-1.07e-02
-1.86e-02
-2.66e-02
-3.45e-02
-4.25e-02
Figure 12.2.25 Distribution of normalized contact pressure

sensitivity with respect to Ecarc at end of footprint analysis.
12.2.26

Printed on:
Normalized d_CPRESS_Ebelt
+2.68e-01
+2.31e-01
+1.94e-01
+1.57e-01
+1.20e-01
+8.37e-02
+4.69e-02
+1.01e-02
-2.66e-02
-6.34e-02
-1.00e-01
-1.37e-01
-1.74e-01

sensitivity with respect to Ebelt at end of footprint analysis.
Normalized d_CPRESS_tNominal
+2.04e-01
+1.74e-01
+1.44e-01
+1.14e-01
+8.45e-02
+5.45e-02
+2.46e-02
-5.41e-03
-3.54e-02
-6.53e-02
-9.53e-02
-1.25e-01
-1.55e-01

sensitivity with respect to at end of footprint analysis.
12.2.27

Printed on:
Normalized CPRESS
+1.00e+00
+9.17e-01
+8.34e-01
+7.51e-01
+6.67e-01
+5.84e-01
+5.00e-01
+4.17e-01
+3.34e-01
+2.50e-01
+1.67e-01
+8.34e-02
+0.00e+00
Figure 12.2.28
Normalized contact pressure distribution for new design (5% increase in Ebelt ).

Figure 12.2.29
Normalized contact pressure history at nodes 3055 and 2645 for new design.
12.2.28

Printed on:
Figure 12.2.210
Mode 1.
Figure 12.2.211
Mode 2.
12.2.29

Printed on:
Figure 12.2.212
Mode 3.
Figure 12.2.213
Mode 4.
12.2.210

Printed on:
Figure 12.2.214
Figure 12.2.215
Mode 2 (wire frame).
12.2.211

Printed on:
Mode 5.
EIGFREQ
Figure 12.2.216
Normalized
Normalized
Normalized
Normalized
Figure 12.2.217
Eigenfrequencies for rst ve modes.
d_EIGFREQ_Ebelt
d_EIGFREQ_Ecarc
d_EIGFREQ_denRub
d_EIGFREQ_tNominal
Normalized eigenfrequency sensitivities for rst ve modes.
12.2.212

Printed on:
DSA WINDSHIELD WIPER
12.2.3
DESIGN SENSITIVITY ANALYSIS OF A WINDSHIELD WIPER
Abaqus/Design
In this example we consider a design sensitivity analysis of a statically loaded windshield wiper blade.
The conventional wiper system is composed of three major subsystems: the blade-arm assembly, a linkage
mechanism, and the electric motor. We restrict our design and analysis to the blade-arm subsystem. Wiping
performance is determined by the dynamic performance of the wiper system, while safety regulations require
that the blade wipe a specic area on the windshield surface. It is a major challenge to achieve systems that
optimize wiping performance and satisfy the area conditions. The interaction between the force/deection
response of the blade (rubber) element and the friction at the blade-glass interface is a critical element in
design. A design sensitivity analysis of the rubber-glass interaction gives insight that can be used to improve
the design before undertaking the more complex dynamic stick-slip analysis. The sensitivity analysis
also identies the design parameters that are most effective in reducing the stress in highly stressed areas.
Self-contact between the anges of the rubber element is not considered in this example.
Geometry, model properties, and design parameters
The geometry of the rubber blade is shown in Figure 12.2.31. The arm is assumed to be stiff relative to
the rubber blade and is modeled by prescribing xed boundary conditions. The rubber blade is modeled
as a plane strain model with 231 rst-order hybrid elements. An incompressible hyperelastic material
with a polynomial strain energy function is used to simulate the rubber material behavior. The windshield
is assumed to be rigid and is modeled as an analytical rigid surface. Surface interaction between the blade
and the windshield is modeled using nite-sliding contact with an isotropic Coulomb friction coefcient
of 0.2. Three shape parameters and the coefcient of friction, , are chosen as design parameters for
the design study. We choose as a design parameter to study the effect of the friction coefcient on
the blade-glass interface forces. The shape parameters that can be modied without requiring geometry
changes to other parts of the assembly are chosen as design parameters. These shape parameters will
help us to study the effect of aspects of wiper geometry on wiper stiffness. The shape parameters are
the thickness, , of the neck between the bottom two anges; the thickness, , of the wiper tip; and
the width, , of the lower ange. The corresponding shape gradients (the derivatives of the nodal
coordinates with respect to the shape design parameters) required as input to the sensitivity analysis are
shown as symbol plots in Figure 12.2.32, Figure 12.2.33, and Figure 12.2.34. These gradients are
input using the *PARAMETER SHAPE VARIATION option and calculated using a custom Python script
derived from the Abaqus/CAE replay le. This approach requires familiarity with the Abaqus Scripting
Interface. Shape variation data can be generated more easily using Abaqus/CAE; details are given in
Design sensitivity analysis: overview, Section 12.1.1.
Loading, boundary conditions, and design responses
An interference t of 2.0 units between the blade and the windshield is used to simulate the static vertical
load between them. The windshield is held xed in the vertical direction, and the surface nodes of
12.2.31

Printed on:
the upper ange are constrained in both directions. The wiping motion is simulated by prescribing a
horizontal displacement of 9.0 units to the reference node of the windshield.
It is desirable to reduce the stress concentration in the rubber element since high stresses will
degrade the life cycle of the blade and affect performance. Consequently, the sensitivity of the Mises
stress is taken as the primary design response in this model. Any design change should take into
account the behavior of the contact pressure between the wiper and the windshield. To this end the
sensitivities of CSTRESS are also selected as design responses. Including friction makes the stiffness
matrix unsymmetric; hence, the analysis is run using the unsymmetric solver. Since we are interested in
the history of the contact pressure at the tip, the incremental DSA formulation is chosen. To quantify
the effect of neglecting the unsymmetric terms and of using total DSA, we also compare the maximum
sensitivities for various combinations of DSA formulation and stiffness matrix symmetrization.
A contour plot of the Mises stress distribution on the deformed conguration of the rubber blade is shown
in Figure 12.2.35. It is evident that the two necks between the anges are the areas of interest. In a
magnied contour plot of the Mises stress distribution in the necks (Figure 12.2.36) we can see that the
stress concentration in the lower neck is higher than in the upper one. The normalized sensitivities of the
maximum Mises stresses in the lower neck are listed in Table 12.2.31. Sensitivities are normalized by
dividing them by the maximum Mises value (67857.7 units in element 131 of the wiper blade) and then
scaling the result by the initial value of the design parameter. We notice that the shape parameters have a
direct effect (increasing the parameter increases stress and vice versa) and the friction coefcient has an
inverse effect on the Mises stress. Reducing the thickness of the neck makes the wiper more exible and
reduces stress concentration. Increasing the friction coefcient increases the shear force at the tip. This
increased shear force increases bending of the upper neck and decreases bending in the lower neck, thus
reducing the stress concentration in the lower neck. We infer from the table that the Mises stress is more
sensitive to and than to the ange width and friction coefcient. To reduce stress concentration,
a design with changes in and is considered. A 20% reduction in and a 10% reduction in is
selected. The predicted percentage reduction in the Mises stress is 6.76 (0.2832 0.2 + 0.1094 0.1).
The proposed design changes will affect the contact pressure at the wiper-windshield interface, and
the sensitivity results are used to quantify this effect. The time history of the normalized sensitivities
of contact pressure at a point at the tip is shown in Figure 12.2.37. Contact pressure at this point is
most sensitive to . This result is somewhat unexpected since this design parameter does not affect the
interface explicitly.
has a directly proportional effect, while
has an inversely proportional effect
on the contact pressures. Thus, decreasing
will decrease the contact pressure while decreasing
will increase the contact pressure. The contact pressure histories for the proposed design and the current
design are shown in Figure 12.2.38. The plots indicate that the reduced stress concentration in the
proposed design comes at the cost of decreased wiper performance. (The proposed design has a lower
contact pressure, which could be suboptimal.) If desired, the contact pressure can be positively inuenced
by considering a positive change in .
All results discussed above are obtained in an incremental DSA analysis using the unsymmetric
solver. If the user is interested only in the results at particular increments (typically, the last increment),
total DSA will be advantageous computationally. Similarly, neglecting the unsymmetric terms will
12.2.32

Printed on:
improve computational efciency. To quantify the error in such approximations, we compare the results
for various combinations to the overall nite difference method (OFD). Table 12.2.32 lists the relative
error in the maximum Mises and contact pressure sensitivities compared to the results obtained using
the OFD method. For the dominant shape parameter, , the sensitivity results of the maximum Mises
stress are in good agreement with the OFD method for all the combinations. However, neglecting the
unsymmetric terms in total DSA gives inaccurate results for the contact pressure sensitivities. The
relative error is large for the less dominant shape parameters, with total DSA giving poor results for
both the Mises and contact pressure sensitivities. Total DSA gives inaccurate results for sensitivities
with respect to the friction coefcient. Although the sensitivity results are problem dependent, we can
infer that total DSA may give poor results if we neglect unsymmetric terms. Further sensitivities of
interaction pressures are more sensitive to approximations than the sensitivities of maximum stresses
in the structures. Less signicant sensitivities are affected more by approximations than the dominant
sensitivities.
The proposed design changes are made and the analysis is rerun. A 9% reduction in the stress
concentration is observed. Predicted and actual contact pressure histories are shown in Figure 12.2.39.
The difference between the predicted and the actual results is due to the nonlinear dependence of the
design response on the design parameter. However, the rerun conrms the prediction that the proposed
design is more robust, with a possibly suboptimal wiping pressure.
Input files
dsawiper.inp
dsawiper_pred.inp
dsawiper_node.inp
dsawiper_elem.inp
dsawiper_t2.inp
dsawiper_t3.inp
dsawiper_h2.inp
Primary wiper blade design sensitivity analysis.

Wiper blade analysis incorporating design changes.
Node description.
Element description.
Parameter shape variation data for thickness .
Parameter shape variation data for thickness .
Parameter shape variation data for width .
Table 12.2.31
Normalized sensitivities of maximum stress values.
Design Parameter
Normalized Sensitivity
0.2832
0.1094
0.0154
0.0026
12.2.33

Printed on:
Table 12.2.32
Incremental DSA;
Unsymmetric
Solver
Incremental DSA;
Symmetric Solver
Total DSA;
Unsymmetric
Solver
% Error
MISES
% Error
CPRESS
% Error
MISES
% Error
CPRESS
% Error
MISES
% Error
CPRESS
% Error
MISES
% Error
CPRESS
0.02
0.01
0.37
1.51
0.00
0.01
0.00
0.88
0.06
0.02
12.35
1.97
0.24
0.12
0.13
0.43
0.02
1.81
1.12
51.6
0.11
61.56
94.41
0.88
2.4
1.70
12.36
28.90
22.15
53.20
93.01
21.21
h 2 = 1.5 h 3 = 1.5 h 4 =1.5 h5 = 1.0 h 6 = 2.0
Design
Parameter
Percentage error in maximum stress and contact pressure sensitivities.
t1= 1.0
h 1 =3.0
t 2 = 0.6
t 3 = 1.0
4.0
Figure 12.2.31
Model of the rubber windshield wiper.
12.2.34

Printed on:
Total DSA;
Symmetric Solver
Figure 12.2.32
Symbol plot of the shape variation with respect to
Figure 12.2.33
12.2.35

Printed on:
Figure 12.2.34
S, Mises
(Ave. Crit.: 75%)
+7.920e+04
+7.260e+04
+6.600e+04
+5.940e+04
+5.280e+04
+4.620e+04
+3.960e+04
+3.300e+04
+2.640e+04
+1.980e+04
+1.320e+04
+6.600e+03
+1.872e-02
Figure 12.2.35
Contour plot of Mises stress.
12.2.36

Printed on:
S, Mises
(Ave. Crit.: 75%)
+7.920e+04
+7.260e+04
+6.600e+04
+5.940e+04
+5.280e+04
+4.620e+04
+3.960e+04
+3.300e+04
+2.640e+04
+1.980e+04
+1.320e+04
+6.600e+03
+1.872e-02
Figure 12.2.36
Magnied contour plot of Mises stress.
12.2.37

Printed on:
d_CPRESS_h2
d_CPRESS_mu
d_CPRESS_t2
d_CPRESS_t3
Figure 12.2.37 Time history of the normalized sensitivity

of contact pressure at the wiper tip.
CPRESS_Current
CPRESS_Predicted
Figure 12.2.38
Current and predicted contact pressure history.
12.2.38

Printed on:
CPRESS_Actual
CPRESS_Predicted
Figure 12.2.39
Figure 12.2.310
Actual and predicted contact pressure history.
Symbol plot of the computed shape variation with respect to
12.2.39

Printed on:
DSA RUBBER BUSHING
12.2.4
DESIGN SENSITIVITY ANALYSIS OF A RUBBER BUSHING
Abaqus/Design
The purpose of this example is to demonstrate how design sensitivity analysis can be used to improve the
design of a rubber bushing. The objective is to alter the bushing geometry to lower the maximum axial
stress, thus increasing the service life. Design sensitivity analysis provides a means of predicting the effect
of a change in the geometry on the stress concentration, thereby aiding in identifying the important design
parameters and determining an appropriate design change.
Geometry, model properties, and design parameters
The bushing consists of inner and outer steel tubes that are bonded to a central rubber cylinder
(Figure 12.2.41). It is assumed that the outer perimeter of the bushing is fully xed. The bushing is
457.2 mm (18.0 in) long, with an outside diameter of 508.0 mm (20.0 in) and an inside diameter of
228.6 mm (9.0 in). The steel is elastic with Youngs modulus = 206.0 GPa (3.0 107 psi) and Poissons
ratio = 0.3. The rubber is modeled as a fully incompressible hyperelastic material that at all strain levels
is relatively soft compared to the steel. The nonlinear elastic behavior of the rubber is described by a
strain energy function that is a second-order polynomial in the strain invariants.
The model is discretized with standard axisymmetric elements since the axial loading results in pure
axisymmetric deformation. CAX4 elements are used for the steel components, and CAX4H elements
are used for the rubber component. Rigid elements (element type RAX2) are attached to the inside of the
bushing in both models to represent the relatively stiff shaft. The use of these elements also simplies the
application of the loading conditions. The axisymmetric nite element mesh is shown in Figure 12.2.41.
An axial force of magnitude 10675.0 N (2400.0 lbs) is applied to the rigid body reference node, while
the outer steel tube is fully xed.
Two design parameters are considered for the design sensitivity analysis: the thickness, t, and the
llet radius, r, of the rubber bushing at the top and bottom ends where it is bonded to the inner steel tube,
which is shown in Figure 12.2.41. These parameters represent typical geometry properties that may be
considered during design evaluation.
Design sensitivity analysis and design parameters
To carry out a design sensitivity analysis with respect to a shape design parameter, the gradients of
the nodal coordinates with respect to the design parameter must be specied with the *PARAMETER
SHAPE VARIATION option. One simple approach to obtaining these gradients is to perturb the shape
design parameters r and t one at a time and to record the perturbed coordinates. The gradients are then
found by numerically differencing the initial and perturbed nodal coordinates. In the current study the
constraint is imposed that a change in the thickness causes the line of nodes connecting the thickness
dimension to the llet radius to rotate about the point of tangency of this line to the llet radius. The
Abaqus Scripting Interface command _computeShapeVariations() provides a semi-automated
facility to compute the shape variations (see Design sensitivity analysis: overview, Section 12.1.1).
12.2.41

Printed on:
DSA RUBBER BUSHING
The deformed mesh is shown in Figure 12.2.42. Figure 12.2.43 shows the contours of axial stress in the
rubber part of the bushing at the end of the axisymmetric analysis. The maximum stress occurs near the
top llet close to the axis. Figure 12.2.44 and Figure 12.2.45 show the contours of the sensitivities of
the axial stress for the shape design variables r and t, respectively. Table 12.2.41 shows the normalized
sensitivities of the maximum axial stress,
0.17 MPa (24.55 psi), with respect to the shape design
variables. The normalization has been carried out by multiplying the sensitivities by a characteristic
dimension (initial llet radius
12.7 mm (0.5 in) and initial thickness
15.24 mm (0.6 in)) and
dividing by the maximum stress. As can be inferred from this table, a change in the llet radius inuences
the maximum stress to a larger extent than a change in the thickness of the rubber. Hence, it is desirable
to change r to modify the stresses. To obtain approximately a 10% reduction in the maximum stress in
the axial direction, the llet radius is increased by
Substituting
and
0.008 MPa/mm (28.75 psi/in) (see Figure 12.2.44)
gives
2.25 mm (0.09049 in). A reanalysis of the problem with the radius changed to
14.99 mm (0.59049 in) yields a reduction of 8.8% in the maximum axial stress, which is slightly less
than the goal of 10%. This is expected because of the nonlinearity of the problem; to achieve the 10%
reduction, this process would have to be repeated, which is essentially an optimization problem.
Input files
bushing_cax4_axi_dsa.inp
bushing_node.inp
bushing_steel.inp
bushing_rubber.inp
bushing_rigid.inp
Design sensitivity analysis for the axisymmetric model.

Node denitions.
Element denitions for the steel.
Element denitions for the rubber.
Element denitions for the rigid body.
Table 12.2.41
Normalized sensitivities of the maximum stress.
Parameter
r
0.58
0.11
12.2.42

Printed on:
DSA RUBBER BUSHING
CL
10"
4.5"
18"
2
3
rubber
1
rigid elements
Figure 12.2.41
Axisymmetric cross-section.
12.2.43

Printed on:
steel
steel
DSA RUBBER BUSHING
2
3
Figure 12.2.42
Deformed mesh after axial loading.
S, S22
(Ave. Crit.: 75%)
+2.455e+01
+1.644e+01
+8.336e+00
+2.291e-01
-7.878e+00
-1.599e+01
-2.409e+01
2
3
Figure 12.2.43
Variation of axial stress in the rubber after axial loading.
12.2.44

Printed on:
DSA RUBBER BUSHING
d_S_R, S22
(Ave. Crit.: 75%)
+5.452e+00
-2.474e-01
-5.947e+00
-1.165e+01
-1.735e+01
-2.305e+01
-2.875e+01
2
3
Figure 12.2.44 Variation of the sensitivity of the axial stress

with respect to an increase in the radius of the llet, r.
d_S_T, S22
(Ave. Crit.: 75%)
+3.289e+00
+1.948e+00
+6.067e-01
-7.343e-01
-2.075e+00
-3.416e+00
-4.757e+00
2
3
Figure 12.2.45 Variation of the sensitivity of the axial stress with

respect to a decrease in the thickness of the rubber, t.
12.2.45

Printed on:
POSTPROCESSING OF Abaqus RESULTS FILES
13.
Postprocessing of Abaqus Results Files
Postprocessing examples, Section 13.1

Printed on:
POSTPROCESSING EXAMPLES
13.1
Postprocessing examples
User postprocessing of Abaqus results les: overview, Section 13.1.1
Output radiation viewfactors and facet areas: FRAD, Section 13.1.5
Joining data from multiple results les and converting le format: FJOIN, Section 13.1.2
Calculation of principal stresses and strains and their directions: FPRIN, Section 13.1.3
Creation of a perturbed mesh from original coordinate data and eigenvectors:
Section 13.1.4
FPERT,
Creation of a data le to facilitate the postprocessing of elbow element results: FELBOW,

Section 13.1.6
13.11

Printed on:
USER POSTPROCESSING PROGRAMS
13.1.1
USER POSTPROCESSING OF Abaqus RESULTS FILES: OVERVIEW
This chapter illustrates how to write and utilize postprocessing programs to manipulate data stored in
the Abaqus results le. The results le, which is identied by the le extension .fil, is created by the
*CONTACT FILE, *EL FILE, *NODE FILE, *ENERGY FILE, *MODAL FILE, and *RADIATION FILE
options and contains results based on user-specied output requests. The standard le format is binary, but it
can be changed by user request for each run with the *FILE FORMAT, ASCII option. Alternatively, it can
be set to a default ASCII format during site installation. Abaqus uses FORTRAN unit 8 to communicate
with the results le.
Sample postprocessing programs that perform commonly exercised tasks are presented in separate
sections in this chapter. These include merging multiple results les and converting the resulting results
le from binary format to ASCII, or vice-versa; computing principal values and directions of stress and
strain; and computing a perturbed mesh for a collapse analysis by incorporating a user-specied geometric
imperfection in the form of the critical buckling mode shape.
Each postprocessing program must be linked using the make parameter when running the Abaqus
execution procedure (see Making user-dened executables and subroutines, Section 3.2.13 of the Abaqus
Analysis Users Manual). To link properly, the postprocessing program cannot contain a FORTRAN
PROGRAM statement. Instead, the program must begin with a FORTRAN SUBROUTINE with the name
ABQMAIN.
General programming concepts, Abaqus FORTRAN interfaces, and data processing concepts are
described below. Refer to Chapter 5, File Output Format, of the Abaqus Analysis Users Manual for
additional information. The program listings in each section provide details on the program ow, how to
interface with various computer platforms that use different operating systems and FORTRAN compilers,
and how to interface with Abaqus subroutines to handle data les and records.
Refer to Chapter 9, Using the Abaqus Scripting Interface to access an output database, of the Abaqus
Scripting Users Manual or Chapter 10, Using C++ to access an output database, of the Abaqus Scripting
Users Manual for information on accessing data stored in the Abaqus output database.
Initialization
Details about the variables that are used in the postprocessing programs are discussed in Accessing the
results le information, Section 5.1.3 of the Abaqus Analysis Users Manual. Abaqus uses a 512-word
buffer named ARRAY for the reading and writing of data on the results le. This is dimensioned as
ARRAY(513). The integer equivalent is JRRAY(513) for a 64-bit computer or JRRAY(2,513) for
a 32-bit computer. The EQUIVALENCE statement is used to equivalence ARRAY and JRRAY to simplify
manipulation of real and integer numbers in the data record stored in the buffer.
The information concerning the FORTRAN unit number and format of the results le that is read is
dened in LRUNIT(2,NRU), where NRU is the number of les to be processed. The FORTRAN unit
number for the nth le is stored in LRUNIT(1,n). The information about the le format is stored in
LRUNIT(2,n), which is initialized to 1 for ASCII format and to 2 for binary format. If a new results
le is to be created by the postprocessing program, the le format of the output le is dened similarly
13.1.11

Printed on:
USER POSTPROCESSING PROGRAMS
via the variable LOUTF, which is also initialized to 1 for ASCII format and 2 for binary format. The
root le name for both input and output results les is dened through the character variable FNAME.
The root le name case will be the same as the case in which FNAME is dened; Abaqus denes the
le extensions to be lowercase letters. See Accessing the results le information, Section 5.1.3 of the
Abaqus Analysis Users Manual, for a discussion of the naming convention for the le extensions.
The nal initialization phase is done internally by calling the Abaqus subroutines INITPF and
DBRNU. The FORTRAN interfaces are
CALL INITPF(FNAME, NRU, LRUNIT, LOUTF)
CALL DBRNU (JUNIT)
where the arguments in the call to INITPF are as described above, and JUNIT is the FORTRAN unit
number connecting the le.
These integer variables must be dened before the subroutines are called.
Data processing
Data manipulation requires knowledge of each data record. Details of these records are found in Results
le output format, Section 5.1.2 of the Abaqus Analysis Users Manual.
The data organization in the results le uses a sequential format. Each record must, therefore, be
retrieved in a sequential manner via a call to DBFILE using the interface
CALL DBFILE(0, ARRAY, JRCD)
This call can be placed inside a DO-loop, and the loop count should exceed the number of records stored
in the le. Alternatively, DBFILE can be called as long as JRCD is equal to 0. The rst argument, 0,
indicates that a record is to be read. Each record that is read is stored in the buffer ARRAY and returned
to the calling program for manipulation. The last argument, JRCD, is a return code that is set to 0 unless
an end-of-le condition or an incomplete record is processed, in which case JRCD is set to 1.
If it is desirable to extract or modify certain records and save them in a new results le with the same
data organization as an Abaqus-generated results le, then the subroutine DBFILW should be called with
the interface
CALL DBFILW(1, ARRAY, JRCD)
The new results le will be written with the le extension .fin. Refer to Utility routines for
accessing the results le, Section 5.1.4 of the Abaqus Analysis Users Manual.
13.1.12

Printed on:
POSTPROCESSING PROGRAM FJOIN
13.1.2
JOINING DATA FROM MULTIPLE RESULTS FILES AND CONVERTING FILE

FORMAT: FJOIN
Abaqus/Explicit
This example illustrates how to use a FORTRAN program to extract specic data from different Abaqus
results les and to join the data into a single results le. This program can also be used to convert the format
of results les.
Postprocessing
Sometimes it is desirable to combine a number of results les into a single le or to create a new results
le by retrieving selected data from different results les. The abaqus append utility joins two results
les by stripping the header information from the second results le and appending the step information to
the end of the rst results le. See Joining results (.fil) les, Section 3.2.10 of the Abaqus Analysis
Users Manual, for more information on this utility.
This example postprocessing program demonstrates how a FORTRAN program can be used to
extract specic information from results les created by separate analyses of the same model. In this
example the stress and strain records in three analyses will be merged to create a new results le.
Programming details
The general discussion on programming concepts and Abaqus FORTRAN interfaces in User
postprocessing of Abaqus results les: overview, Section 13.1.1, should be reviewed before running
or modifying this program. Review of the results le format in Chapter 5, File Output Format, of the
Abaqus Analysis Users Manual is also recommended.
The program FJOIN (named fjoin.f on the Abaqus release media) prompts for the values of
NRU, LRUNIT(1,NRU), LRUNIT(2,NRU), and FNAME. Then subroutines INITPF and DBNRU are
called to complete the necessary initializations and le connections. Data processing starts with a double
DO-loop looping over all of the records to be read, one-by-one, via a call to DBFILE. A record can be
skipped or written to the new results le with or without any modications. Each record is identied
by its record key, which is stored in the second entry of the record (see Results le output format,
Section 5.1.2 of the Abaqus Analysis Users Manual).
Each le contains a number of header records, 1900-series. These records contain general
information about the model. Different analyses using the same model place essentially the same
information in these records. Hence, when combining results les from different analyses of the same
model, all 1900-series records from the rst le that is processed should be kept. Similar records in
subsequent les should be skipped to avoid duplication and confusion. However, the 1910, 1911,
1922, and 1980 records should be kept. They are useful for processing results within a substructure,
output requests, and natural frequency extraction results.
The data for each increment of an analysis begin with the increment start record, which is identied
by record key 2000. Record 2000 is followed by the records that correspond to the data requested
13.1.21

Printed on:
through le output options specied in the Abaqus input le. Record 1, the element header record,
is automatically written to the results le when the *EL FILE option is used in the input le. It is of
interest when postprocessing since it contains important information about the element data, including
the location of data within an element (i.e., whether data are written at the element integration points, the
centroid, nodes, etc.). For this example, records 11 and 21 (the stress and strain records, respectively)
are written to the results le since stress and strain were requested through the *EL FILE option. The
increment end record is identied by record key 2001. When an end-of-le condition is encountered
and the previously processed record is a 2001 record, a FORTRAN CLOSE is executed on the current
FORTRAN unit number so that the processing of the next le can begin.
Program compilation and linking
The abaqus make utility is designed to compile and link this type of postprocessing program. It will
also make the aba_param.inc le available during compilation. The command to compile and link
the FJOIN program is as follows:
abaqus make job=fjoin
This command will have to be repeated if FORTRAN errors are discovered during the compilation or link.
The commands used by the abaqus make utility can be changed if necessary. The Abaqus Installation
and Licensing Guide lists the typical compile and link commands for each computer type.
Program execution
Before program execution, the analysis jobs must be run to generate results les to be read by the
program. In this example three jobs are run. The input les for these analyses are fjoin002.inp,
fjoin003.inp, and fjoin004.inp. The results les from these analyses are output in binary
format and are called fjoin002.fil, fjoin003.fil, and fjoin004.fil.
The FJOIN program will read these les via FORTRAN units 2, 3, and 4. The name of the new le
will be fjoinxxx. Before running the program, the results les must be renamed to fjoinxxx.002,
fjoinxxx.003, and fjoinxxx.004. Note that the root le names are the same (dened using
FNAME), and that the extensions are set to the FORTRAN unit numbers used to open the les.
When the program is executed using the command abaqus fjoin, the rst prompt will be
Enter the number of files to be joined:
Enter 3 to set NRU=3. The second prompt will be
Enter the unit number of input file # 1:
Enter 2 to dene LRUNIT(1,1)=2. At the third prompt,
Enter the format of input file # 1 (1-ASCII, 2-binary):
enter 2. This sets LRUNIT(2,1)=2 and means that the le being read is binary. The second and third
prompts are repeated for each additional le to be processed. The program will then ask whether the new
results le should be written in ASCII or binary format,
Enter the format of the output file (1-ASCII, 2-binary):
13.1.22

Printed on:
Enter 2 to set LOUTF=2, which species that binary format has been chosen for the new results le. The
format of the output le may be different from the format of the input les, so this program can also be
used to convert the format of results les. Finally, when the program issues the prompt
Enter the name of the input files (w/o extension):
enter fjoinxxx to dene FNAME (the input les must have been given the root le name fjoinxxx;
the output le will be created as fjoinxxx.fin).
As soon as the nth le has been processed, the message END OF FILE # n is written to the
terminal. After all les have been processed, the program stops and the new results le is created. The
new results le created by this program contains stress and strain records at all integration points in each
element and at all nodal points.
Analysis description
The structure is a 10 10 square plate with unit thickness. The plate lies in the XY plane such that its
bottom edge coincides with the x-axis and the left edge coincides with the y-axis. The nite element
model employs a 2 2 mesh of CPS8R elements. The material is linear elastic with Youngs modulus =
30 106 and Poissons ratio = 0.3. Three separate analyses are performed with displacement-controlled
load steps.
In the rst analysis (fjoin002.inp), the plate is subjected to biaxial tension by prescribing a vertical
displacement of 0.25 along the top edge, a horizontal displacement of 0.25 along the right edge, and
symmetry boundary conditions on the left and bottom edges.
In the second analysis (fjoin003.inp), the structure is forced to deform in simple shear by applying
a horizontal displacement of 0.25 to the top edge while holding the bottom edge xed and allowing the
horizontal displacement to vary linearly with y along the left and right edges. The vertical displacement
is zero everywhere.
In the third analysis (fjoin004.inp), the plate is subjected to uniaxial tension by applying a
displacement of 0.25 in the y-direction to the nodes along the top edge and symmetry boundary
conditions to the nodes along the x- and y-axes.
Since the state of stress and strain is homogeneous, the integration point and nodal averaged values of
stress and strain are identical everywhere. A typical record obtained at the end of each step is included
below:
Analysis
fjoin002.inp
fjoin003.inp
fjoin004.inp
1.07 106
0.0
0.0
1.07 106
0.0
7.50 105
13.1.23

Printed on:
0.0
2.88 105
0.0
Analysis
fjoin002.inp
fjoin003.inp
fjoin004.inp
2.50 102
0.0
7.50 103
2.50 102
0.0
2.50 102
0.0
2.50 102
0.0
Input files
fjoin002.inp
fjoin003.inp
fjoin004.inp
fjoin.f
First analysis le.

Second analysis le.
Third analysis le.
Postprocessing program.
13.1.24

Printed on:
POSTPROCESSING PROGRAM FPRIN
13.1.3
CALCULATION OF PRINCIPAL STRESSES AND STRAINS AND THEIR

DIRECTIONS: FPRIN
This example illustrates the use of a FORTRAN program to read stress and strain records from an Abaqus
results le and to calculate principal stress and strain values and their directions.
General description
This program shows how to retrieve integration point and nodal averaged stress and strain components
from an Abaqus results le and then compute principal values and directions using the Abaqus utility
routine SPRIND. Usage of this utility routine is documented in the input le provided for this problem,
and further details about the interface to this subroutine are discussed in Obtaining stress invariants,
principal stress/strain values and directions, and rotating tensors in an Abaqus/Standard analysis,
Section 2.1.11 of the Abaqus User Subroutines Reference Manual. The results le created by the FJOIN
program in Joining data from multiple results les and converting le format: FJOIN, Section 13.1.2,
is used here to verify that the records that have been put together are retrievable. The previously
generated results le was named fjoinxxx.fin. To use it as an input le for postprocessing program
FPRIN, the le extension must be changed. This program will assume that the results le has the
default .fil extension, which corresponds to FORTRAN unit 8.
Programming details
The user should rst review the general discussion on programming concepts and Abaqus FORTRAN
interfaces in User postprocessing of Abaqus results les: overview, Section 13.1.1, and the detailed
discussion of postprocessing given in Chapter 5, File Output Format, of the Abaqus Analysis Users
Manual.
When running program FPRIN (this program is named fprin.f on the Abaqus release media),
the user will be prompted for the le name that initializes FNAME. Other variables, such as LOUTF, NRU,
LRUNIT(1,NRU), and LRUNIT(2,NRU), are initialized inside the program. INITPF and DBRNU
are then called to complete the neccesary initializations and le connections. Data processing starts
with a double DO-loop over all the records to be read, one-by-one, via a call to DBFILE. Each record
is identied by its record key, which is stored in the second entry of the record. When records 1922
and 2000 are processed by program FPRIN, the heading and the current step and increment numbers
are written out so as to provide a way to recognize the beginning of data in each analysis. Record type
1 is then examined to determine the output location of stress and strain, the number of direct and shear
stress and strain components, and either the element number or the node number for which the records
are written. The stress and strain records (11 and 21, respectively) will be ltered out for processing
by the Abaqus utility routine SPRIND. When a stress or strain record is passed into SPRIND, principal
stresses or strains and the corresponding principal directions are calculated and returned in an unsorted
order.
13.1.31

Printed on:
Before program execution, the FORTRAN program has to be compiled and linked. Both operations, as
well as the inclusion of the aba_param.inc le, are performed by a single execution of the abaqus
make utility:
abaqus make job=fprin
This may have to be repeated until all FORTRAN errors are corrected. After successful compilation, the
programs object code is automatically linked with the Abaqus object codes stored in the shared program
library and interface library in order to build the executable program. Refer to Chapter 4, Customizing
the Abaqus environment, of the Abaqus Installation and Licensing Guide to see which compile and link
commands are used for a particular computer.
Program execution
Before the program is executed, a results le must have been created. In this example the results le
fjoinxxx.fin created by the FJOIN program discussed in Joining data from multiple results
les and converting le format: FJOIN, Section 13.1.2, is used. This le must be renamed to
fjoinxxx.fil since FORTRAN unit 8 (which is associated with the .fil le extension) is used
in the program to read the le. When the program is executed using the command abaqus fprin,
the prompt
Enter the name of the input file (w/o .fil):
will appear. Enter fjoinxxx to dene FNAME. The program processes the data and produces a
le named pvalue.dat, which contains information about principal stresses and strains and their
directions.
The computed principal stress and strain values and their directions are tabulated below.
Analysis
File
Principal
Component
Stress
105
Strain
103
Dir-1
Dir-2
Dir-3
fjoin002.inp
1
2
3
1
2
3
10.714
10.714
0.0
2.8846
2.8846
0.0
25.0
25.0
0.0
12.5
12.5
0.0
1.0
0.0
0.0
0.707
0.707
0.0
0.0
1.0
0.0
0.707
0.707
0.0
0.0
0.0
1.0
0.0
0.0
1.0
fjoin003.inp
13.1.32

Printed on:
Analysis
File
Principal
Component
Stress
105
Strain
103
Dir-1
Dir-2
Dir-3
fjoin004.inp
1
2
3
0.0
7.5
0.0
7.5
25.0
0.0
1.0
0.0
0.0
0.0
1.0
0.0
0.0
0.0
1.0
Input file
fprin.f
13.1.33

Printed on:
POSTPROCESSING PROGRAM FPERT
13.1.4
CREATION OF A PERTURBED MESH FROM ORIGINAL COORDINATE DATA AND

EIGENVECTORS: FPERT
This example illustrates the use of a FORTRAN program to create a perturbed mesh by superimposing a
small imperfection in the form of the weighted sum of several buckling modes on the initial geometry. The
program retrieves the original nodal coordinates and the desired eigenvectors from an Abaqus results le, then
calculates new nodal coordinates for the perturbed mesh.
General description
Collapse studies of a structures postbuckling load-displacement (Riks) behavior are often conducted
to verify that the critical buckling load and mode predicted by an eigenvalue buckling analysis are
accurate. They are also done to investigate the effect of an initial geometric imperfection on the loaddisplacement response. A typical assumption is that an imperfection made up of a combination of the
eigenmodes associated with the lowest eigenvalues will be the most critical. One method of introducing
an imperfection of this type into the model is by adding
to the original mesh coordinates. In
this case
is the ith eigenmode,
is a scaling factor of the ith eigenmode, and M is the total number
of eigenmodes extracted in the buckling analysis. Since the eigenvector is typically normalized to a
maximum absolute value of one, is usually some fraction of a geometric parameter, such as the shell
thickness. The postprocessing program described below can be used to introduce an imperfection of this
type into a model.
The perturbation procedure is illustrated in Buckling of a cylindrical shell under uniform
axial pressure, Section 1.2.3 of the Abaqus Benchmarks Manual. An eigenvalue buckling analysis,
fpert001, is run rst. This analysis creates the results le, fpert001.fil, which contains the
original nodal coordinates and the eigenvectors for the buckling modes. This results le is then used
to generate a perturbed mesh for the postbuckling load-displacement analysis. The postprocessing
program perturbs the original mesh using the relation
where
is the vector containing the new global coordinates; is the vector of original coordinates; M
is the number of buckling modes; and
is the imperfection factor for the ith eigenvector, . The new
coordinates are written to the le fpert002.015, which is read by the load-displacement analysis
fpert002.
Programming details
The general discussion on programming concepts and Abaqus FORTRAN interfaces in User
postprocessing of Abaqus results les: overview, Section 13.1.1, should be reviewed before running
13.1.41

Printed on:
or modifying this program. Review of the results le format in Chapter 5, File Output Format, of the
Abaqus Analysis Users Manual is also recommended.
The FPERT program (this program is named fpert.f on the Abaqus release media) makes some
assumptions concerning the type of results le it will be reading. Variables NRU, LRUNIT(1,NRU),
and LRUNIT(2,NRU) are initialized within the program to 1, 8, and 2. These values indicate that one
le will be read, the FORTRAN unit used will be 8, and the le type will be binary. See Accessing the
results le information, Section 5.1.3 of the Abaqus Analysis Users Manual, for more information on
opening and initializing postprocessing les.
Once the le specication parameters are set, the INITPF and DBNRU subroutines are called to
open and ready the le, whose name is stored in FNAME, for reading. The le to which the perturbed
coordinates are to be written can be directly opened using a FORTRAN OPEN statement. The Abaqus
le utilities are not necessary since the le is a plain text le.
The records with the original nodal coordinates are read using the DBFILE routine and stored in
the local array COORDS(3,8000). The rst index of the COORDS array indicates the x-, y-, and zcoordinate of the node. The second index indicates the node number. The second dimension should be
increased if there are more than 8000 nodes in a model.
Components of the eigenvector are stored in the local array DISP(6,8000). This array holds up
to 6 displacement terms for each node. The second dimension should be increased if there are more than
8000 nodes in a model. Subroutine NODEGEN, a subroutine local to this postprocessing program, is then
called to compute the new nodal coordinates. Once all the requested mode shapes are computed, the new
nodal coordinates are written to the plain text le opened earlier.
The abaqus make utility is designed to compile and link this type of postprocessing program. It will
also make the aba_param.inc le available during compilation. The command to compile and link
the FPERT program is as follows:
abaqus make job=fpert
This command will have to be repeated if FORTRAN errors are discovered during the compilation or link.
The commands used by the abaqus make utility can be changed if necessary. The Abaqus Installation
and Licensing Guide lists the typical compile and link commands for each computer type.
Program execution
Before the program is executed, an eigenvalue buckling job must have been run with Abaqus. In this
example the input le fpert001.inp is used to generate the results le fpert001.fil. When the
FPERT program is executed using the command abaqus fpert, the rst prompt will be
Enter the name of the results file (w/o .fil):
Enter fpert001 to dene FNAME. The second prompt will be
Enter the mode shape(s) to be used in calculating the perturbed
mesh (zero when finished):
13.1.42

Printed on:
Enter 1 followed by 0, since this is the only eigenvector available in the results le for this example. At
the third prompt,
Enter the imperfection factor to be introduced into the geometry
for this eigenmode:
enter 0.25. This sets = 0.25, the shell thickness for this model. The program then processes the data
and writes the nodal coordinates for the new mesh to fpert002.015.
Analysis description
For a full discussion of the analysis, refer to Buckling of a cylindrical shell under uniform axial
pressure, Section 1.2.3 of the Abaqus Benchmarks Manual. The input le fpert001.inp (same
le as bucklecylshell_s9r5_n3.inp) contains a 2 20 mesh of S9R5 elements and data lines
for a buckling analysis. The input le fpert002.inp contains data lines for a Riks analysis using a
perturbed mesh. The source code for the FPERT program is in fpert.f.
Plots produced by these analyses are shown in Figure 13.1.41 and Figure 13.1.42. Figure 13.1.41
is obtained from the eigenvalue buckling analysis and shows the original (cylindrical) mesh and the
critical buckling mode. Figure 13.1.42 is generated when the load level has reached a local maximum
(increment 8) in the Riks analysis using the perturbed mesh.
Input files
fpert001.inp
fpert002.inp
fpert.f
Eigenvalue buckling analysis.

Riks analysis using a perturbed mesh.
13.1.43

Printed on:
Figure 13.1.41
Undeformed shape and eigenvalue buckling mode.
2
3
1
Figure 13.1.42
Deformed shape at rst peak load in Riks analysis.
13.1.44

Printed on:
POSTPROCESSING PROGRAM FRAD
13.1.5
OUTPUT RADIATION VIEWFACTORS AND FACET AREAS: FRAD
This example illustrates the use of a FORTRAN program to read the radiation viewfactors and the facet areas
from the results le.
General description
The program shows how to retrieve the viewfactors and the facet areas from the results le. The
results le created from the benchmark problem detailed in Axisymmetric elemental cavity radiation
viewfactor calculations, Section 1.6.6 of the Abaqus Benchmarks Manual, is used to verify that the
output records have been read and output correctly. This program will assume that the results le has
the default le extension, .fil, which corresponds to FORTRAN unit 8.
Programming details
Before proceeding, review the general discussion on programming concepts and Abaqus FORTRAN
interfaces in User postprocessing of Abaqus results les: overview, Section 13.1.1, and the detailed
discussion of postprocessing given in Chapter 5, File Output Format, of the Abaqus Analysis Users
Manual.
When running the program FRAD (this program is named frad.f on the Abaqus release media),
the user will be prompted for the le name that initializes FNAME. Other variables, such as LOUTF, NRU,
LRUNIT(1,NRU), and LRUNIT(2,NRU), are initialized inside the program. INITPF and DBNRU
are then called to complete the necessary initializations and le connections. By default, the results le
is processed for all steps and increments in the results le. The user can restrict the output by setting
LSTEPA and LINCA to the required step and increment and uncommenting the simple IF - END IF
block. Data processing starts with a DO-loop over all the records to be read, one-by-one, by means of
a call to DBFILE. Each record is identied by its record key, which is stored in the second entry of the
record. When records 1922 and 2000 are processed by FRAD, the heading and the current step and
increment numbers are written out so as to provide a way to recognize the beginning of data in each
analysis. Record types 1605, 1606, 1607, and 1609 are then read; and the desired output is written
to the output le vfout.
Before it can be executed, the FORTRAN program must be compiled and linked. Both operations, as
well as the inclusion of the aba_param.inc le, are performed by a single execution of the abaqus
make utility:
abaqus make job=frad
This procedure may have to be repeated until all FORTRAN errors are corrected. After successful
compilation, the programs object code is linked automatically with the Abaqus object codes stored in
13.1.51

Printed on:
POSTPROCESSING PROGRAM FRAD
the shared program library and the interface library to build the executable program. Refer to the Abaqus
Installation and Licensing Guide for information about the compile and link commands for a particular
computer.
Program execution
Before the program is executed, a results le must have been created with the desired output being
written to that le. In this example the results le xrvda4n1.fil created by running the input le
xrvda4n1.inp discussed in Axisymmetric elemental cavity radiation viewfactor calculations,
Section 1.6.6 of the Abaqus Benchmarks Manual, is used. When the program is executed using the
command abaqus frad, the prompt
will appear. Enter xrvda4n1 to dene FNAME. The program processes the data and produces a le
named vfout, which contains the required information.
The radiation viewfactors and facet areas are read and output to vfout. The output agrees with the
expected results.
Input file
frad.f
13.1.52

Printed on:
POSTPROCESSING PROGRAM FELBOW
13.1.6
CREATION OF A DATA FILE TO FACILITATE THE POSTPROCESSING OF ELBOW

ELEMENT RESULTS: FELBOW
This example illustrates the use of a FORTRAN program to read selected element integration point records
from an Abaqus results le to facilitate the postprocessing of elbow element results. XY data are created that
are suitable for use with the XY plotting capability in Abaqus/CAE.
General description
This program shows how to retrieve integration point data for elbow elements from an Abaqus results
le to visualize one of the following:
1. Variation of a variable along a line of elbow elements,
2. Variation of a variable around the circumference of a given elbow element, or
3. Ovalization of a given elbow element.
An ASCII le containing XY data is created that can be read into Abaqus/CAE for visualization
purposes.
To execute option 1, the elbow elements must be numbered such that they increase monotonically
within the range of elements considered; all elements in the desired range must be elbow elements.
XY data will be created with the X-data being the distance along the line of elbow elements, measured
along the elbow centerline and the Y-data being the variable value. The user must ensure that the
integration point coordinates (COORD) are written to the results le if either option 2 or 3 is needed.
For option 2 X-data are the distance around the circumference of the elbow element, measured along the
middle surface, and Y-data are the variable value. For option 3 the XY data are the current coordinates of
the middle-surface integration points around the circumference of the elbow element, projected to a local
coordinate system in the plane of the deformed cross-section. The origin of the local system coincides
with the center of the cross-section; the plane of the deformed cross-section is dened as the plane that
contains the center of the cross-section and integration points 1 and 2.
Programming details
The user is prompted for the name of the results le (assumed to be binary) and the postprocessing
option (1, 2, or 3). The user is then prompted for additional information depending on the option that
was chosen; this information includes
The range of element numbers (options 2 and 3 require only a single element number),
The section point number (options 1 and 2 only),
The integration point number (option 1 only),
The element variable (options 1 and 2 only),
13.1.61

Printed on:
POSTPROCESSING PROGRAM FELBOW
The component of the variable (as dened in Results le output format, Section 5.1.2 of the
Abaqus Analysis Users Manual, options 1 and 2 only),
The step number, and

The increment number.
The data are processed in a double DO-loop over all records, via a call to DBFILE. The desired data
are stored in variable VAR; the integration point coordinates are stored in COORDS. The program checks
to make sure the requested data are available in the results le. An error is issued if the user tries to
process data that are not found in the results le.
Before program execution, compile and link the FORTRAN program by using the abaqus make utility:
abaqus make job=felbow
Repeat this command until all FORTRAN errors are corrected. After successful compilation, the
programs object code is linked automatically with the Abaqus object codes stored in the shared program
library and interface library to build the executable program. Refer to Chapter 4, Customizing the
Abaqus environment, of the Abaqus Installation and Licensing Guide to see which compile and link
commands are used for a particular computer.
Program execution
Before executing the program, run an analysis that creates a results le containing the appropriate output.
This analysis includes, for example, output for the elements in a given range and the integration point
coordinates of the elements. When the program is executed using the command abaqus felbow, the
prompt
will appear. Enter the name of the results le to dene FNAME. The user is then prompted for other
information, such as the desired postprocessing option, element number, etc. The program processes the
data and produces a le named output.dat that contains the information required to visualize the
elbow element results.
Elastic-plastic collapse of a thin-walled elbow under in-plane bending and internal pressure,
Section 1.1.2, contains several gures created with the aid of this program. The output agrees with the
expected results.
Input file
felbow.f
13.1.62

Printed on:

Abaqus Example Problems Manual: Abaqus Version 6.6 ID: Printed On

Uploaded by

Copyright:

Available Formats

Abaqus Example Problems Manual: Abaqus Version 6.6 ID: Printed On

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Abaqus Example Problems Manual: Abaqus Version 6.6 ID: Printed On

Uploaded by

Copyright:

Available Formats

Abaqus Example Problems Manual

Abaqus Version 6.6 ID:

Example Problems Manual

Abaqus Version 6.6 ID:

Abaqus Version 6.6 ID:

Technical Support Centers

Fremont, CA, Tel: +1 510 794 5891, simulia.west.support@3ds.com

Complete contact information is available at http://www.simulia.com/locations/locations.html.

Abaqus Version 6.6 ID:

Abaqus Version 6.6 ID:

Abaqus Version 6.6 ID:

Static Stress/Displacement Analyses

Axisymmetric analysis of bolted pipe flange connections

Buckling and collapse analyses

Snap-through buckling analysis of circular arches

Upsetting of a cylindrical billet: quasi-static analysis with mesh-to-mesh solution

Fracture and damage

A plate with a part-through crack: elastic line spring modeling

Springback of two-dimensional draw bending

Dynamic Stress/Displacement Analyses

Nonlinear dynamic analysis of a structure with local inelastic collapse

Rigid projectile impacting eroding plate

Mode-based dynamic analyses

Analysis of a rotating fan using substructures and cyclic symmetry

Closure of an air-filled door seal

Tire and Vehicle Analyses

Symmetric results transfer for a static tire analysis

Inertia relief in a pick-up truck

Occupant safety analyses

Seat belt analysis of a simplified crash dummy

Resolving overconstraints in a multi-body mechanism model

Heat Transfer and Thermal-Stress Analyses

Thermal-stress analysis of a disc brake

A sequentially coupled thermal-mechanical analysis of a disc brake with an Eulerian

Fluid Dynamics and Fluid-Structure Interaction

Conjugate heat transfer analysis of a component-mounted electronic circuit board

Eigenvalue analysis of a piezoelectric transducer

Joule heating analyses

Thermal-electrical modeling of an automotive fuse

Mass Diffusion Analyses

Hydrogen diffusion in a vessel wall section

Acoustic and Shock Analyses

Fully and sequentially coupled acoustic-structural analysis of a muffler

Plane strain consolidation

Jack-up foundation analyses

Design Sensitivity Analyses

Design sensitivity analysis: overview

Design sensitivity analysis of a composite centrifuge

Postprocessing of Abaqus Results Files

User postprocessing of Abaqus results files: overview

TIRE AND VEHICLE ANALYSES

Tire and Vehicle Analyses

Tire analyses, Section 3.1

Abaqus Version 6.6 ID:

Symmetric results transfer for a static tire analysis, Section 3.1.1

Abaqus Version 6.6 ID:

TIRE RESULTS TRANSFER

SYMMETRIC RESULTS TRANSFER FOR A STATIC TIRE ANALYSIS

Abaqus Version 6.6 ID: