Finite Element Analysis of Solids and Structures
Uploaded by
Mario AmellFinite Element Analysis of Solids and Structures
Uploaded by
Mario AmellFinite Element Analysis
of Solids and Structures
Finite Element
Analysis of Solids
and Structures
Sudip S. Bhattacharjee
First edition published 2021
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742
and by CRC Press
2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN
CRC Press is an imprint of Taylor & Francis Group, LLC
© 2021 Taylor & Francis Group, LLC
Reasonable efforts have been made to publish reliable data and information, but the author and
publisher cannot assume responsibility for the validity of all materials or the consequences of
their use. The authors and publishers have attempted to trace the copyright holders of all material
reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged please write and
let us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known
or hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, access www.copyright.com
or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. For works that are not available on CCC please contact mpkbookspermissions@
tandf.co.uk
Trademark notice: Product or corporate names may be trademarks or registered trademarks and are
used only for identification and explanation without intent to infringe.
ISBN: 978-0-367-43705-3 (hbk)
ISBN: 978-1-032-04158-2 (pbk)
ISBN: 978-1-003-02784-3 (ebk)
Typeset in Times
by SPi Global, India
Access the Support Material: https://www.routledge.com/Finite-Element-A nalysis-of-Solids-a nd-
Structures/Bhattacharjee/p/book/9780367437053
Contents
Preface........................................................................................................................xi
Author....................................................................................................................... xv
Chapter 1 Introduction to Stress Analysis of Solids and Structures...................... 1
Summary............................................................................................... 1
1.1 Introduction – A Brief Summary of Key Historical
Developments............................................................................. 2
1.2 Role of Analysis/Simulation in Product Engineering................. 3
1.3 Static Equilibrium of Structures and Analysis of Forces –
Statically Determinate and Indeterminate Systems.................... 5
1.4 Stiffness (Displacement) Method of Structural Analysis........... 7
1.5 Components of Stresses in a 3D Body..................................... 11
1.6 Variation of Stresses and Differential Equations of
Equilibrium............................................................................... 13
1.7 Stress Boundary Conditions..................................................... 15
1.8 Practice Problems on Stress Field Visualization with
CAE Tools................................................................................ 16
Chapter 2 Strain–Displacement Relationship and Elasticity of Materials........... 19
Summary............................................................................................. 19
2.1 Measurement of Deformation Intensity (Strain)...................... 20
2.2 General Description of Deformation State in 3D Solids.......... 22
2.3 Compatibility of Strain (Deformation Field) in a Body........... 23
2.4 Stress-Strain Relationships (Hooke’s Law).............................. 23
2.5 Solution of Elasticity Problems Using Stress Distribution
Functions.................................................................................. 28
2.6 Solution of Elasticity Problems Using Displacement
Variation Functions................................................................... 30
2.7 Stiffness Method (Finite Element Method) of Structural
Analysis.................................................................................... 32
2.8 Stiffness Properties of 1-D Truss Element Providing
Resistance to Axial Deformation Only..................................... 34
2.9 Higher Order Truss Element and Model Refinement............... 39
2.10 Model Preparation for Computer-Aided Analysis of
Structures with Finite Element Simulation Software............... 41
2.11 Practice Problems – Stress Analysis of a Cable with Finite
Element Analysis Software ABAQUS...................................... 43
v
viContents
Chapter 3 Analysis of Solids Represented by 2D Stress Fields........................... 47
Summary............................................................................................. 47
3.1 Plane Strain – A Special Form of Elasticity Problem............... 48
3.2 Plane Stress – A Special Form of Elasticity Problem............... 50
3.3 Stress Functions for 2D Plane Strain and Plane Stress
Elasticity Problems................................................................... 51
3.4 Example Use of Stress Function Approach to Solve
Elasticity Problems................................................................... 54
3.5 Stiffness Method Analysis of Solids Represented by
2D Stress-Deformation Fields.................................................. 60
3.6 Iso-Parametric Definition of 2D
Solid Finite Elements............................................................... 63
3.7 Numerical Calculation of Stiffness Matrix for
Iso-Parametric 2D Solid Finite Elements................................. 67
3.8 Higher Order Plane-Stress/Plane-Strain Elements................... 71
3.9 Constant Stress/Strain Triangular Element............................... 73
3.10 Convergence of Finite Element Model Solutions..................... 76
3.11 Selection of Element Types for 2D Stress Analysis
with FEA Software Packages................................................... 78
3.12 Description of Load Types in General Stress Analysis
Problems................................................................................... 82
3.13 Refined Finite Element Model Preparation with a
Pre-Processor (HyperMesh)..................................................... 83
3.14 Practice Problems – Stress Analysis of a Plate Subjected
to In-Plane Stresses................................................................... 84
Chapter 4 FEA Model Preparation and Quality Checks...................................... 85
Summary............................................................................................. 85
4.1 Adaptation of Finite Element Mesh to Stress Field
Variations.................................................................................. 85
4.2 Element Type Selection for a Given Stress Analysis
Problem..................................................................................... 88
4.3 Initial Geometric Design of Structural Components................ 91
4.4 FEA Model Preparation for Case-Study: Plate with
Hole (by Using HyperMesh).................................................... 92
4.4.1 CAD Data Preparation for FEA Modeling.................. 92
4.4.2 Finite Element Meshing of the CAD Geometry.......... 94
4.4.3 Finite Element Mesh Quality Checks.......................... 95
4.4.4 Material and Part Property Assignment....................... 98
4.4.5 Analysis Parameters – Boundary Constraints,
External Loads and Model Outputs............................. 99
4.4.6 Definition of ABAQUS Analysis Step....................... 101
4.4.7 Exporting the Model Input File for ABAQUS
Analysis..................................................................... 101
Contents vii
4.5 Post-Processing of FEA Results............................................. 103
4.6 Practice Problems – Stress Analysis of Plate with Hole......... 103
Chapter 5 Stress Analysis of Axisymmetric and General 3D Solids................. 107
Summary........................................................................................... 107
5.1 Axisymmetric – A Special Form of 3D Elasticity
Problems................................................................................. 107
5.1.1 Equations of Equilibrium in Polar Coordinates......... 109
5.1.2 Strain–Displacement Relationships........................... 109
5.1.3 Stress–Strain Relationships....................................... 110
5.1.4 Compatibility Condition............................................ 111
5.2 Stress Analysis of Axisymmetric Example – Thick
Wall Cylinder.......................................................................... 111
5.3 Finite Element Analysis of Axisymmetric Problems.............. 115
5.4 Stress Analysis of Three-dimensional Bodies –
Concentrated Loads................................................................ 117
5.4.1 Stresses in a Semi-Infinite Solid Subjected to
Concentrated Normal Force on the Boundary........... 117
5.4.2 Stresses in a Solid Beam Subject to
Concentrated Lateral Forces...................................... 119
5.5 Finite Elements for Stress Analysis of General 3D Solids..... 121
5.6 Three-Dimensional FEA Model Preparation and Element
Quality Checks....................................................................... 123
5.7 Practice Problems – Stress Analysis of Axisymmetric
and 3D Solids......................................................................... 124
Chapter 6 Deformation Analysis of Beams for Axial, Bending, Shear, and
Torsional Loads................................................................................. 127
Summary........................................................................................... 127
6.1 Bending Stresses in a Beam................................................... 128
6.2 Stresses Due to Transverse Shear........................................... 129
6.3 Transverse Normal Stress in a Beam...................................... 132
6.4 Torsional Response of a Beam............................................... 133
6.5 Beam Response to Combined Load Effects........................... 142
6.6 Elastic Bending Deflection of Beams..................................... 144
6.7 Stress Analysis of Curved Beams........................................... 146
6.8 Stiffness Properties of Prismatic Euler–Bernoulli
Beam Elements....................................................................... 148
6.9 Stiffness Properties of Beams Including Shear
Deformation............................................................................ 154
6.10 Analysis of Beams and Frames with FEA Software
Packages................................................................................. 158
6.11 Practice Problems: Load–Deflection Analysis of Beams....... 162
viiiContents
Chapter 7 Analysis of 3D Thin-Wall Structures (Plates and Shells)................. 165
Summary........................................................................................... 165
7.1 Bending Stresses and Strains in a Plate.................................. 166
7.2 Analytical Solutions for Plate Bending Deflections............... 169
7.3 In-plane Membrane Stress Resistance of a Shell................... 172
7.4 Bending Stiffness of Flat Plate Element................................. 174
7.5 Flat Shell Element as a Combination of Plate Bending
and Membrane Elements........................................................ 176
7.6 Shear Deformation in Plates................................................... 179
7.7 Curved Shell Elements........................................................... 181
7.8 Shell Element Mesh Quality and Integration Rules............... 184
7.9 Analysis of Shells with FEA Software................................... 186
7.10 Practice Problems: Load–Deflection Analysis of Shells........ 187
Chapter 8 Multi-Component Model Assembly.................................................. 191
Summary........................................................................................... 191
8.1 Element Compatibility and Convergence of Simulation
Results.................................................................................... 191
8.2 Modeling of Kinematic Joints in
Structural Assemblies............................................................. 194
8.3 Deformable Joint Elements for Part-to-Part Connections...... 198
8.4 Mesh-Independent Fasteners for Part-to-Part Connections.... 201
8.5 Simulation of Part-to-Part Interface Contacts........................203
8.6 Thin-Layer Interface Elements............................................... 207
8.7 Modular Organization of Data in Multi-component
Model Assembly..................................................................... 209
8.8 Result Quality Checks............................................................ 211
8.9 Practice Problems: Analysis with Multi-Component
Model Assemblies.................................................................. 213
Chapter 9 Interpretation of Stress Analysis Results for Strength and
Durability Assessment....................................................................... 217
Summary........................................................................................... 217
9.1 Engineering Properties of Materials....................................... 218
9.2 Stress-Strain Results from Linear Elastic Finite Element
Analysis of Solids................................................................... 219
9.3 Strength Assessment of Solids – Use of Material Failure
Theories.................................................................................. 222
9.4 Post-Processing of Finite Element Stress Analysis Results... 229
9.5 Stress Analysis for Durability (Fatigue Life) Assessment
of Structures............................................................................ 230
Contents ix
9.6 Structural Safety Assessment and Quality of FEA
Stress Results.......................................................................... 234
9.7 Stresses at Points of Discontinuity:
Stress Intensity Factor............................................................ 235
9.8 Practice Problems: Assessment of Structural Strength
and Durability......................................................................... 238
Chapter 10 Vibration Frequency Analysis of Structures with
FEA Model........................................................................................ 241
Summary........................................................................................... 241
10.1 Introduction – Dynamic Response of Structures.................... 242
10.2 Vibration Frequency of a Single Degree of Freedom
Spring-Mass System............................................................... 243
10.3 Forced Vibration Response and Resonance of Structures...... 245
10.4 Frequency Separation and Design Targets for Structures....... 249
10.5 Vibration Mode Shape and Frequency of
SDOF Structures..................................................................... 250
10.6 Vibration Frequencies of MDOF Systems............................. 253
10.7 Calculation of System Mass Matrix for MDOF Systems....... 255
10.8 Numerical Calculation of Vibration Mode Shapes
and Frequencies of MDOF Systems....................................... 259
10.9 Vibration Frequency Analysis with ABAQUS.......................261
10.10 Practice Problems: Vibration Analysis of Structures.............. 262
Chapter 11 Linear Dynamic Response Analysis of Structures............................ 265
Summary........................................................................................... 265
11.1 Linear Elastic Response of SDOF Systems to
Impulsive Loading.................................................................. 265
11.2 Response Spectrum of Linear Dynamic Systems................... 268
11.3 Time-Domain Analysis of Dynamic Structural
Response................................................................................. 270
11.4 Numerical Integration Parameters for Time-Domain
Analysis of Structures............................................................. 271
11.5 Time Domain Analysis of MDOF Systems............................ 273
11.6 Mode Superposition Method for Analysis of MDOF
Systems................................................................................... 276
11.7 Explicit Time-Domain Analysis of MDOF Systems.............. 277
11.8 Linear Dynamic Response Analysis with ABAQUS.............. 282
11.9 Practice Problems: Dynamic Response Analysis of
Structures............................................................................... 286
xContents
Chapter 12 Nonlinear Analysis of Structures....................................................... 289
Summary........................................................................................... 289
12.1 Simulation of Nonlinear Force-Deformation Response
of Structures............................................................................ 290
12.2 Nonlinear Material Models for Finite Element
Simulation of Structures......................................................... 294
12.3 Simulation of Large Deformation Response –
Nonlinear Geometric Problems.............................................. 301
12.4 Nonlinearity Arising from Changes to Inter-Body
Contacts.................................................................................. 307
12.5 Nonlinear Dynamic Response Analysis of Structures............ 308
12.6 Material Failure Simulation in Nonlinear
Finite Element Analysis.......................................................... 310
12.7 Computational Methods for Structural Form
Simulation............................................................................... 311
12.8 Practice Problems: Nonlinear Response Analysis of
Structures................................................................................ 313
References������������������������������������������������������������������������������������������������������������� 315
Index���������������������������������������������������������������������������������������������������������������������� 319
Preface
Theories and methods for analysis of solids and structures have developed over sev-
eral centuries. The academic curricula in most engineering programs cover those
ideas through multiple courses starting from the basic course on statics to the very
advanced courses on nonlinear solid mechanics and dynamics of structures.
Integration of all these ideas, in the form of computer-aided matrix method analysis
of multi-degree-of-freedom systems, has evolved into the finite element method –
which is an indispensable part of current engineering practice for design and analysis
of solids and structures. Many books have been produced over the past half century
with a great deal of reference information on analytical formulations and relevant
numerical implementation details. These reference books definitely serve as good
references for the developers of finite element software packages. The theory manu-
als of actual software implementations tend to mimic similar analytical details of the
reference textbooks. Other software-related documents, namely the users’ manuals
and example manuals, tend to focus on the computer graphics-based techniques for
efficient model preparation and post-processing of results. The successful use of
finite element simulation technology, in actual engineering problem solutions,
requires the mastery of all relevant subjects and tools – starting from the basic under-
standing of solids mechanics principles to the computer user skills for post-processing
the right response results as obtained from finite element simulation models. The
large swath of published materials, available in the form of textbooks and software
manuals, and the easy availability of computing equipment and software products
have made the learning of finite element subjects accessible and difficult at the same
time. Aspiring stress analysts in today’s world face the difficult challenge of taking a
long suit of academic and non-academic courses to build the bridge between theory
and software-oriented engineering practice. It is quite normal that graduates in the
structural analysis field appear with disconnected knowledge bins, often with good
understanding of the mechanics field without the knowledge of how to approach a
practical analysis problem; and occasionally with good user skills of specific soft-
ware features without the knowledge of what goes inside those tools.
From the synthesis of 30+ years of hands-on work in research, programming,
teaching, and practical use of the finite element simulation method, the author has
developed this book as an introductory reference to the key ideas of the technique.
The attempt is by no means a substitute for existing books and reference materials. It
is rather a complimentary guidebook for navigating through the complex learning
path of finite element subject, for both academic students and self-learning practicing
engineers. Comprehensive theory-to-practice coverage of the essential subject mat-
ters also makes the book an excellent reference for use in corporate training pro-
grams for engineering graduates who lack the formal rigor of academic preparation
in stress analysis domain. The analytical formulations have been presented by using
matrix and vector notations, unlike the tensor notation as often used in many other
reference books and software manuals. This is intended to make an explicit link
between theory and actual analysis projects that obviously involve input and output
xi
xiiPreface
of data in matrix and vector forms. The analytical and numerical implementation
details of finite element methods have been discussed side-by-side with user options
available in commercial software packages. The example demonstration of software
user features has been discussed with specific references to the analysis capabilities
of the ABAQUS package. The discussions on software-aided model preparation and
quality checks have been presented with example references to the HyperMesh soft-
ware package. These software-specific feature descriptions will need to be adapted if
different software products are used to solve the practice problems.
This book is structured to progressively build the basic expertise in finite element
analysis methods – following the same sequence that the author has used in teaching
of one-semester graduate courses over the past many years. Chapters 1 and 2 present
the theory of elasticity topics that form the foundation of linear elastic finite element
methods. Practice problems are presented, with pre-built analysis model files, to test
the basic stress–strain analysis theories by using finite element analysis software as a
virtual experimentation tool. Chapter 3 introduces the basics of finite element formu-
lations, including the numerical details as implemented in software tools, for analysis
of solids that can be represented by 2D stress fields. Practice problems are presented
to use software analysis tools with manual preparation of finite element model input
files. The objective is to achieve clear knowledge of the input data structure required
for an error-free analysis model preparation. This is an essential skill to debug model
errors that often appear while preparing more complex analysis models by using
computer-graphics-based model preparation tools. Chapter 4 focuses on how to pro-
duce good quality finite element models of two-dimensional solids by using general-
purpose model pre-processing software. Although specific references are made to
HyperMesh software features for model build operations, and to ABAQUS software
for actual FEA solutions, the discussions on key aspects of quality model preparation
are equally applicable to other software products for model preparation and valida-
tion of results. Three-dimensional elasticity problems, having an axis of symmetry,
are discussed in Chapter 5, followed by the introduction of 3D finite element formu-
lations that are required for analysis of general 3D solids. Chapter 6 is dedicated to
the elastic deflection and stress–strain analysis of beams for bending, transverse
shear and torsional load effects. The size of this chapter, with comprehensive cover-
age of both theoretical and finite element implementation aspects, is understandably
large for a single class learning session. A reduced presentation can be formulated by
focusing on finite element implementation methods, with selected references to the
analytical methods, if desired. Chapter 7 presents the plate and shell element formu-
lations for the analysis of 3D thin-walled structures. Analysis problems, discussed in
Chapters 1–7, represent components that can be modeled with single finite element
types. Chapter 8 introduces special numerical techniques that are required to simu-
late the behavior of joints and interface contacts in multi-component model assem-
blies. The basic purpose of finite element simulation, i.e. the interpretation of stress
analysis results for engineering decisions, is discussed in Chapter 9, with special
focus on strength, durability, and integrity assessment of solid products. Chapter 10
presents a comprehensive review of the analytical and numerical methods for vibra-
tion frequency analysis which is an important topic in design and analysis of struc-
tures for cyclic load effects. Chapter 11 is dedicated to the analysis of structural
Preface xiii
response for noncyclic dynamic load events. The use of finite element simulation
models to predict design response spectra, frequency response function, and time-
domain response histories of structures is specifically discussed in this chapter.
Finally, Chapter 12 presents a review of key analysis techniques relevant for predict-
ing the nonlinear response of structures. At the end of each chapter, practice prob-
lems have been presented for solving with finite element analysis models, and for
verification of the results by using simplified analytical prediction models. These
studies are intended to reinforce the importance of conducting minimum quality
checks of the finite element simulation results. For comprehensive learning experi-
ence, some practice problems will require the use of electronic data files that can be
downloaded from the site: https://www.routledge.com/Finite-Element-Analysis-of-
Solids-and-Structures/Bhattacharjee/p/book/ 9780367437053. PowerPoint slides of
the materials covered in this book, and the solution manual of practice problems, are
available upon request from academic instructors.
Author
Dr. Sudip S. Bhattacharjee is a supervisor for vehicle
crashworthiness engineering in Ford Motor Company,
USA. He is also an occasional graduate course instructor
at the University of Windsor, Canada, and the University
of Michigan, Dearborn. Prior to joining Ford in 2000,
Dr. Bhattacharjee was a faculty member at the University
of Windsor, a consulting engineer with SNC-Lavalin in
Montreal, and a postdoctoral research fellow at Ecole
Polytechnique of Montreal. Dr. Bhattacharjee received his
Ph.D. from McGill University, Montreal in 1993. Prior to
that, he was a lecturer of Civil Engineering at Bangladesh
University of Engineering and Technology (BUET),
where he also received his Bachelor and Masters degrees
in Civil Engineering.
xv
1 Introduction to Stress
Analysis of Solids and
Structures
SUMMARY
Theories of stress analysis for solids and structures have evolved over many
centuries – leading to the development of finite element method in the current time
of abundant digital computing power. A short list of the key developments in solid
mechanics, by no means a comprehensive review of the subject matter, is presented
in Section 1.1 as a form of small tribute to the history. Section 1.2 briefly reviews the
present-day product development process where finite element model-based virtual
simulation plays a key role in optimization and validation of the product design.
Before the appearance of digital computing technology, analysis of structures, how-
ever, largely relied on hand calculation of external force effects based on the solution
of static equilibrium equations. Two primary alternative methods, namely the force
and displacement methods, have initially contributed to the development of matrix
method of structural frame analysis. It is the displacement method (or alternatively
known as the stiffness method) that has eventually evolved into the present form of
finite element analysis (FEA) method for solids and structures. Following a brief
review of the flexibility method in Section 1.3, a more detailed description of the
stiffness-based matrix equilibrium equations is presented in Section 1.4. The impor-
tant properties of the stiffness-based matrix analysis method are reviewed in this
section with an easy-to-understand example of elementary springs. The finite ele-
ment discretization technique of solids is briefly introduced in this section to high-
light the underlying similarity between the direct stiffness-based matrix method and
the more advanced finite element method of structural analysis.
Section 1.5 is devoted to the definition of internal body stress components – as
direct extension of describing the force effects on 3D solids. The state of stress-
equilibrium and the stress-boundary conditions, important analytical foundation
blocks of advanced solid mechanics principles, are presented in Sections 1.6 and 1.7.
Finally, two practice problems are presented in Section 1.8 primarily to help with the
learning of powerful graphical processing tools for visualization of stress analysis
results. Pre-prepared finite element simulation models are presented for analysis with
ABAQUS software (Dassault Systems 2020a). However, similar practice models can
be prepared (or the presented model files can be converted) for analysis with other
commercially available software packages. The objective at this initial step is not to
introduce the technique of finite element model building process; it is to use the pre-
built models as virtual experimental tools with visualization of the stress fields and
boundary conditions.
1
2 Finite Element Analysis of Solids and Structures
1.1 INTRODUCTION – A BRIEF SUMMARY OF KEY HISTORICAL
DEVELOPMENTS
Structural analysis techniques to determine the effects of forces have evolved simul-
taneously with the progression of human civilization through millennia. Surviving
ancient structures and historical records provide evidence of thoughtful human
endeavor to build safe structures based on the understanding of effects of natural
forces. The following are examples of few early historical milestones:
3000 BC Use of beams and columns as structural elements in Pyramids Egyptians
582 BC Center of gravity of bodies, and rules of statics Archimedes (Greek)
100 BC Use of arch concept in structures Romans
Subsequent advancements in the understanding of effects of forces on materials
and bodies can be attributed to the following individuals:
15th century Behavior of members in tension, compression, and bending Da Vinci
17th century Strength of materials Gelileo Galilei
17th century Relation between force and deformation Robert Hooke
17th century Laws of motion and calculus Isaac Newton
Industrial revolution in Europe and the advent of rail-road drove the development
of generic analysis techniques to design structures for industrial applications. Notable
developments of that era can be summarized as follows:
17th century Deflection of elastically bent bar James Bernoulli
18th century Buckling of columns under axial load Leonhard Euler
18th century Theory of friction C.A. Coulomb
18th century Generalized forces for the solution of complex structures J.L. Lagrange
19th and 20th centuries experienced a flurry of advanced theoretical develop-
ments forming the discipline of “mechanics of materials”:
19th century Elastic modulus of materials Thomas Young
19th century Theory of bent plates and shells L.M. Navier
19th century Theory of elasticity and plasticity Poisson, Cauchy, Lame,
Kelvin, St. Venant
19th century Deflection of plates Kirchoff
19th century Failure theory – maximum shear stress yield criterion Tresca
19th century Theorem of least work Castigliano
20th century Plasticity theory and material failure von Mises
20th century Elastic solutions of problems in solid mechanics Timoshenko
Introduction to Stress Analysis of Solids and Structures 3
Modern era of structural mechanics has been characterized by the development of
numerical methods to solve problems in solids and structures. Some of the key con-
tributors in this still-evolving field can be listed as follows:
20th century Approximate solution of continuum mechanics Ritz
20th century Concepts of analysis of frames and continuum Argyris, Turner, Clough, and others
20th century Concepts of finite element methods Clough, Wilson, Bathe,
Zienkiewicz, Hughes, and others
Historical developments listed in the above are few from numerous other indi-
viduals and groups that have contributed to the development of discipline what we
call today mechanics of solids or structural mechanics. A comprehensive review of
historical developments is beyond the scope of this chapter. However, key ideas and
individuals are listed here as a tribute to the centuries of theoretical developments
that form the foundation of today’s finite element analysis method. Modern-day
finite element analysis method is the embodiment of centuries of theoretical develop-
ment in mechanics – making the structural analysis techniques readily available to
21st-century engineers through digital computing software products.
1.2 ROLE OF ANALYSIS/SIMULATION IN PRODUCT
ENGINEERING
Externally applied forces cause internal stress and deformation (strain) to materials
that have limited resistance capacity. Figure 1.1(a) gives an example of ductile mate-
rial resistance properties defined by yield strength Sy and failure strain εf. Resistance
capacity of a structural member may also be exhausted before reaching the material
Engineering stress:
σ = F / A0 Force, F
sy
L0
A0
Change in
Engineering strain: length, ∆L
ε = ∆L / L0 εf
(a) (b)
FIGURE 1.1 (a) Resistance capacity of a ductile material (yield strength = Sy and failure
strain = εf); (b) buckling resistance of a thin wall member subjected to axial compressive stress.
4 Finite Element Analysis of Solids and Structures
Ini al design based on basic
understanding of force effects on Virtual stress analysis and Performance valida on of
structural members (using “design op miza on of system design product hardware
codes” or “design best prac ces”)
Bumper
impact
= M*c/I
(a) (b) (c)
FIGURE 1.2 Example of present-day automotive product development process. Simulation
model and crash test pictures are adapted from 2010 Toyota Yaris analysis report (NHTSA
2020).
strength, for example, due to buckling of thin wall members under axial compressive
stresses as shown in Figure 1.1(b). Checking internal stress is, thus, an important step
in product engineering process to determine the safety of a product.
Product design codes and best practices generally provide guidelines to define
initial geometric proportions of members assuming simple stress distributions inside
a member when subjected to idealized external forces and boundary conditions.
Preliminary member designs, based on design best practice rules, are not necessarily
fully optimized design of structures for complex loading and boundary conditions.
Advanced techniques of structural analysis, such as finite element method, provide
the ability to predict stress–deformation response of structures before manufacturing
the hardware products. Virtual assessment and design iterations help to avoid highly
inefficient design-test-correct iteration loops. Figure 1.2 shows an example of pres-
ent-day automotive product development process, starting with: (a) preliminary
design of components based on basic principles of engineering mechanics; followed
by, (b) detail finite element analysis of structures for upfront performance assessment
and design optimization; and (c) final testing of the prototype product for perfor-
mance validation.
Simulation models are not always intended to replace the actual hardware testing.
Results of simulation models often need to be assessed and used within the limita-
tions of assumptions made in the model building process. For example, analysis effi-
ciency may require the use of linear elastic material behavior assumption. So, the
results of such analysis model should be assessed with due consideration to that
assumption. Figure 1.3 lists some of the limitations of simulation models that engi-
neers need to be specifically aware of.
Introduction to Stress Analysis of Solids and Structures 5
Simplificaons/ Assumpons FEA simulaon Outcome & Limitaons
• Locaons of stress
concentraons are
Material: linear elasc reproduced
• Local stress/strain
responses, near
disconnuies, are
Models: Idealized geometry, approximated
Loading: steady / stac
loading and
boundary condions
• Dynamic /fluctuaon
characteriscs are
simplified with analycal
load definion
Geometry: linear edges/
• Design robustness is
planar surfaces esmated based on
approximate stress/strain
results and idealized
material resistance
informaon
FIGURE 1.3 Potential limitations of simulation models and results.
1.3 STATIC EQUILIBRIUM OF STRUCTURES AND ANALYSIS OF
FORCES – STATICALLY DETERMINATE AND INDETERMINATE
SYSTEMS
First step in stress analysis of a structure is the determination of external forces and
boundary conditions. A statically determinate system is one where external forces
and reactions acting on the structure can be determined from the analysis of static
equilibrium condition. Figure 1.4 shows the bumper beam of an automotive body
Simplified model
of bumper beam
FIGURE 1.4 External force “P” acting on an automotive bumper beam. Image of automo-
tive body structure is prepared from Toyota Yaris vehicle model available in NHTSA (2020).
6 Finite Element Analysis of Solids and Structures
y Applied force P = 3 kN
x
Simplified 2D model (sta cally determinate system)
z Calculate unknown reac ons @ boundary constraints
600 mm
= 7338
4 = 0 = 0 = 0
Calculate internal forces, moments and stresses
L/2 = ( ) Mmax = 3 * 600 / 4 = 450 kN-mm
30 mm
σmax = M*c / I = 2.
=7338 mm4 20 mm
= 450*10/7338 = 0.613 GPa
1 mm thick walls
Calculate structural deflec ons (using moment-area method)
=
^
= /
= = 8.8 mm
FIGURE 1.5 Response analysis of the simplified model of an automotive bumper beam.
Sta cally in-determinate system
Not all reac on forces and moments @ boundary
constraints are known from sta c equilibrium
calcula ons
FIGURE 1.6 Example of a statically indeterminate beam model.
structure subjected to an external force, P. End constraints (boundary conditions) can
be idealized to make the system statically determinate as shown in the figure.
Elementary bending theory leads to complete solution including maximum stress
and maximum deflection values for the example analysis problem (Figure 1.5).
However, when the end conditions of bumper beam are assumed to be moment-
resistant rigid connections (as shown in the example illustration in Figure 1.6), equa-
tions of static equilibrium alone will not be sufficient to determine the unknown
reaction forces. Several analysis techniques exist to calculate unknown boundary
forces in statically indeterminate beams (Kennedy and Madugula 1990). Two widely
used analysis techniques are (i) force method (also known as flexibility method) and
(ii) displacement method (also known as stiffness method). Force (or flexibility)
method of structural analysis involves the following key steps:
• Make the structure statically determinate by eliminating sufficient number of
unknown constraints
• Calculate responses of determinate system
• Calculate the response of “determinate” structure for unit value of “eliminated”
force constraint
Introduction to Stress Analysis of Solids and Structures 7
• Calculate the magnitude of “eliminated” force by using principle of superposi-
tion that makes the structural displacement compatible with actual boundary
constraint.
Flexibility method of structural analysis is a very systematic step-by-step procedure,
but it is generally not used in computer-aided analysis of complex structural systems.
Stiffness method (as described in the following section) has become the analysis
method of choice in most structural analysis software products.
1.4 STIFFNESS (DISPLACEMENT) METHOD OF STRUCTURAL
ANALYSIS
Stiffness of a structural element, defined as the resistance to unit deformation, is used
to determine the displacement response from equilibrium condition. An elementary
structural example is represented by the spring element in Figure 1.7(a) where one
end of the spring is attached to a rigid boundary, while the other end is subjected to
an external force P. Displacement response at the load application point is identified
by u. Figure 1.7(b) shows an example force–deformation response of the spring ele-
ment. Assuming linear elastic behavior for the spring, slope of the force–deformation
line in Figure 1.7(b) represents the spring stiffness k. Physically, stiffness k repre-
sents the resistance of the spring against unit axial deformation (u = 1). At the load
application point in Figure 1.7(a), total spring resistance is given by k × u, where u is
the unknown displacement caused by the externally applied load P. Equilibrium of
forces at load application point provides the following simple but very powerful
equation for structural response analysis:
k.u = P (1.1)
For this example of single-degree-of-freedom (DOF) spring system (Figure
1.7(a)), displacement response u can be predicted by using the equilibrium equation
(1.1) for any given magnitude of force P, provided that the stiffness property k is
known. Stiffness property described above can now be used to derive the general 2 ×
2 stiffness matrix of a spring element that has axial stiffness value k1 and two degrees
of freedom (u1, u2) as shown in Figure 1.8. In this 2-DOF system, stiffness term kij is
Example: axial deforma on ONLY S ffness: resistance to “unit”
deforma on – property of a system
P
Load S ffness, k
Displacement, u Force, P (kN) (unit: kN/mm)
(unit: mm) (unit: kN)
Accelera on, ü = 0
u Displacement (mm)
(a) (b)
FIGURE 1.7 (a) Single-degree-of-freedom (DOF) spring and (b) linear-elastic force–
deformation response.
8 Finite Element Analysis of Solids and Structures
Spring-1
S ffness matrix of 2 DOF spring
u1 u2
DOF
(Degrees of
Freedom) k1
u1=1 u2=0
Forces K11=k1 K21= –k1
u1=0 u2=1
Forces K12= –k1 K22=k1
FIGURE 1.8 Stiffness matrix of a 2-DOF spring element (spring-1).
Spring-2
S ffness matrix of 2 DOF spring u2 u3
k2
u2=1 u3=0
K22=k2 K32= –k2
u2=0 u3=1
K23= –k2 K33=k2
FIGURE 1.9 Stiffness matrix of a 2-DOF spring element (spring-2).
defined as the spring resistance at ith degree of freedom for unit deformation at jth
degree of freedom. Similarly, 2 × 2 stiffness matrix of another spring element
(spring-2), having stiffness property equal to k2 and connecting to two degrees of
freedom (u2, u3), is derived in Figure 1.9. A structural assembly can be formed by
joining the two spring elements, 1 and 2, at common DOF u2. Stiffness matrix of the
example 3-DOF system assembly is shown in Figure 1.10.
Stiffness contributions of element-1 are assigned to kij terms (i = 1, 2 and j = 1, 2)
in the 3 × 3 stiffness matrix; and those of element-2 are assigned to kij terms (i = 2, 3
and j = 2, 3). Stiffness contributions of both elements are added to define the stiffness
parameter k22 corresponding to the common degree of freedom u2. Single-DOF sys-
tem equilibrium, given by equation (1.1), can be expanded to define the equilibrium
of 3-DOF spring assembly as follows:
k1 k1 0 u1 P1
k1 k1 k2 k2 u2 P2 (1.2)
0 k2 k2 u3 P3
Introduction to Stress Analysis of Solids and Structures 9
k1 k2
Spring-1 u2 u3 Spring-2
u1
S ffness matrix of 3 DOF system u2 u3
u1 u2
DOF
(Degrees of 0 k2
Freedom) k1
+ u2=1 u3=0
u1=1 u2=0
0
Forces K11=k1 K21= –k1 K22=k2 K32= –k2
u1=0 u2=1 u2=0 u3=1
Forces K12= –k1 K22=k1 K23= –k2 K33=k2
FIGURE 1.10 Stiffness matrix of a 2-spring assembly.
Solution of matrix equilibrium equations (1.2) gives a unique displacement
response, {u}, provided that stiffness matrix [k] is positive definite (Crandall 1956).
Positive definiteness of the stiffness matrix implies that for any displacement vector
{u} we have (Bathe 1996):
u . k . u 0
T
(1.3)
For any displacement vector {u}, equation (1.3) implies that the strain energy,
which is one-half of the quantity on the left side term of the equation, is also positive.
Note that an unrestrained structural element has rigid body motions implying that the
stiffness matrix of such unrestrained system will be semi-definite. The stiffness
matrix of a structure is rendered positive definite by eliminating rows and columns
that correspond to the restrained degrees of freedom, i.e. by eliminating the possibil-
ity of the structure to undergo rigid body motions. An example of modification to
equilibrium equations (1.2), with the introduction of boundary conditions u1 = u3 = 0,
is represented by the system of equations (1.4) where the modified stiffness matrix is
positive definite:
0 0 0 0 P1
0 k1 k2 0 u2 P2 (1.4)
0 0 0 0 P3
After inserting the boundary constraints in matrix equilibrium equations, as
explained in the above, the remaining set of equations can be solved to determine the
unknown displacement ui for an applied set of forces Pi. Key analysis steps for the solu-
tion of multi-DOF system example, presented above, can be summarized as follows:
• Calculate the stiffness matrix, [k]
• Identify known forces and displacement constraints
10 Finite Element Analysis of Solids and Structures
• Modify stiffness matrix to take account of known boundary constraints
• Solve equations of equilibrium to determine unknown displacements: [k] × {u}
= {P}
• Calculate member internal forces: Fi = ki × (uj − ui)
• Calculate average internal stress: σ = Fi/Ai
Solution procedure described above considers member stiffness (and internal
material stress) for resistance to axial deformation only. Internal stresses for bending,
shear, and torsional loading effects can be calculated following the solid mechanics
principles. Figure 1.11 shows beam internal stresses under the actions of axial force
P, moment about z-axis Mz, shear force Vy, and torsion T. Additional stresses can also
be calculated similarly for moment about y-axis, My, and shear force Vz. Beam ele-
ment stiffness calculations for resistance to bending, shear, and torsional effects will
be discussed in Chapter 5.
Relationships among member internal stresses and external forces are straightfor-
ward for skeletal structural members such as the beam example shown in Figure
1.11. With simplified stress distribution assumptions, mechanics of solids provides
analytical expressions to calculate internal stresses when resultant forces are known
from structural analysis. General shaped solid bodies, such as the example shown in
Figure 1.12, require special analytical techniques to relate internal stresses with
external loads and boundary conditions. Finite element analysis method, to be intro-
duced in Chapters 2–8, builds that bridge between the matrix equilibrium equations
of load–displacement structural response, and the theory of elasticity principles
describing the internal stress fields at differential elements of general solid bodies. In
that process, a general 3D structural body is imagined to be a virtual assembly of
standard geometric shape smaller elements that are interconnected through common
nodal DOF (Figure 1.13). Special numerical techniques are used to calculate stiff-
ness properties corresponding to various forms of element response mechanisms.
y
Stresses on beam cross-secon:
z x
Vy : =
Mz
.
: =
P
.
T : =
.
.
, =
Here, A=cross-sectional area, Izz=moment of inertia about neutral axis zz, Qzz=1st moment about neutral axis ‘zz’ for
the area beyond the point where shear stress τxy is calculated, ‘b’ is the width of beam section at the point of shear stress
calculation, ‘r’ is the distance of torsional stress point from beam axis, and J is the polar moment of inertia of beam
section.
FIGURE 1.11 Stresses on beam section under the actions of axial force, P, bending moment
about z-axis Mz, shear force Vy, and torsion T.
Introduction to Stress Analysis of Solids and Structures 11
External forces
Displacement constraints
(with unknown reac on
forces)
Internal stresses caused by
external forces and boundary
constraints
FIGURE 1.12 General 3D body subjected to external forces and boundary constraints.
Forces
Unknown displacements
@ each point (“node”)
Displacement constraints
(boundary condi ons)
FIGURE 1.13 A solid body imagined to be an assembly of finite size elements of standard
geometric shape.
Like the multi-spring system, equilibrium of the “finite element model” is expressed
in the following familiar form (equation 1.5):
k11 k1n u1 P1
.. .. (1.5)
kn1 knn un Pn
where n is the number of DOF, {u} is the vector of nodal displacements, and {P} is the
vector of applied nodal forces. Some of the variables in vectors {u} and {P} are known,
while others are calculated by solving the system equilibrium equations. Once the nodal
displacements, {u}, are known from the solution of equations (1.5), the calculation of
internal stresses follows the element-specific internal formulations. Details of the finite
element properties will be gradually introduced in next chapters. The remainder of
Chapter 1 presents generic descriptions of the internal stress components and their inter-
relations. Chapter 2 will describe the internal deformations (strains) in differential solid
elements, and the inter-relations among stress and strain variables – leading eventually to
the derivation of stiffness matrix and equilibrium equations for general solids.
1.5 COMPONENTS OF STRESSES IN A 3D BODY
Figure 1.14 shows an infinitesimally small area, ΔA, with its normal direction parallel
to the x-axis. Normal and tangential forces acting on that area are represented by ΔFx,
12 Finite Element Analysis of Solids and Structures
Internal forces acng
= lim =
Fy
at point area: ∆A
= lim =
Fx
Fz
= lim =
FIGURE 1.14 Components of internal forces (and stresses) acting on an infinitesimal area
ΔA.
ΔFy, and ΔFz. Intensity of force, at the limit condition of ΔA approaching 0, is defined
as the “stress” acting at the measurement point. Normal and shear stress components,
corresponding to normal and shear forces acting on the plane ΔA, are defined as σx,
τxy, and τxz in Figure 1.14.
The stress measurement point inside a 3D body, as shown in Figure 1.14, can be
considered to reside inside an infinitesimally small volume, ΔV, that is bounded by
three mutually perpendicular planes – each plane being perpendicular to an axis of
orthogonal x–y–z system (Figure 1.15). Expanding the definition of three stress com-
ponents acting on 2D x-plane of Figure 1.14, nine stress components acting on three
mutually perpendicular planes (x, y, and z) are identified in Figure 1.15.
Normal stresses, acting on opposite surfaces of “infinitesimally” small volume,
are assumed equal in magnitude, but opposite in direction (variation of normal stress
is ignored) (Figure 1.16). The relationship between shear stresses (τxy and τyx), acting
on orthogonal faces, can be derived by considering the rotational equilibrium of the
entire element about z-axis:
M z xy .dy.dz .dx yx .dx.dz .dy 0 (1.6)
Equation (1.6) shows that τxy = τyx. Similarly, considering rotational equilibrium
about x- and y-axes, we get, respectively, τyz = τzy and τxz = τzx. Shear stresses on
mutually perpendicular planes are of equal magnitude. This means that stresses at a
σy
τyx
y τyz
τxy
τzy σx
τxz
x σz τzx
z
Stresses on x-plane σx τxy τxz
9 components of
stresses at a Stresses on y-plane τyx σy τyz
point in 3D body
Stresses on z-plane τzx τzy σz
FIGURE 1.15 Components of stresses acting on an infinitesimally small 3D volume.
Introduction to Stress Analysis of Solids and Structures 13
σy σz σx τxy τxz
τyz τyx
τxy τyx σy τyz
σx τzy σx
τxz
σz τzx
τzx τzy σz
σy
y
x
z
FIGURE 1.16 Vector of 6 independent stress components at a point in 3D body.
point in 3D body can be represented by a vector of six independent components (σx,
σy, σz, τxy = τyx, τxz = τzx, τyz = τzy) (Figure 1.16).
1.6 VARIATION OF STRESSES AND DIFFERENTIAL EQUATIONS
OF EQUILIBRIUM
Assuming no variation of stresses in thickness direction, variation of stresses in 2D
x–y plane of an infinitesimal element (dx.dy) is expressed by truncated Taylor’s
expressions in Figure 1.17. Fx and Fy denote body forces per unit volume acting in
directions x and y, respectively. Considering unit thickness of the differential element
in normal direction to x–y plane, equilibrium of forces in the x-direction gives:
x
Fx x .dx .dy x .dy xy xy .dy .dx
x y (1.7)
xy .dx Fx .dx.dy 0
+ .
+ .
+ .
dy
+ .
dx
FIGURE 1.17 Variations of stresses in a differential element in x–y plane.
14 Finite Element Analysis of Solids and Structures
Upon simplification of above equation, and considering dx.dy ≠ 0, differential
equation of stress equilibrium in the x-direction is given by
x xy
Fx 0 (1.8)
x y
Similarly, by considering the equilibrium of forces in the y-coordinate direction,
the following additional differential equilibrium is obtained:
y xy
Fy 0 (1.9)
y x
Expanding the two-dimensional stress-equilibrium state of differential element
(Figure 1.17) to three-dimensional space, and adding additional stress components
and body force component Fz, the following general equations of 3D stress equilib-
rium is obtained (Timoshenko and Goodier 1982, Ugural and Fenster 2012):
x xy xz
Fx 0
x y z
y xy yz
Fy 0 (1.10)
y x z
z xz yz
Fz 0
z x y
Three stress-equilibrium equations (1.10) contain six independent stress vari-
ables. So, additional relationships will be required to determine the internal stress
field in a statically indeterminate general 3D body such as the one shown in
Figure 1.12. Chapter 2 will present those additional relationships to solve the stress
analysis problems both analytically and numerically. Stress-equilibrium equations
(1.10) can also be written in the following compact form:
B . Fi 0
T
(1.11)
where {σ} and [B] represent the following:
x
y
z (1.12)
xy
yz
zx
Introduction to Stress Analysis of Solids and Structures 15
x 0 0
0
0
y
0
0
z
B (1.13)
0
y x
0
z y
0
z x
Vector and matrix notations of equations (1.12) and (1.13) will appear repeatedly in
next chapters on the description of finite element analysis methods. In absence of
body forces, or when body forces are negligible, equation (1.11) implies uniform
stress state inside the differential element of Figure 1.17, i.e. σx is constant in the
x-direction, σy in the y-direction, and τxy is constant in the x–y plane. Stresses inside
a large solid body or along boundaries do not need to be uniform or constant.
Integration of the differential element stresses can equilibrate the resultant effects of
bending, shear and torsional loading conditions on member boundaries – as shown
by elementary beam example in Figure 1.11. A more general formulation of stress-
boundary condition is presented in Section 1.7.
1.7 STRESS BOUNDARY CONDITIONS
Stress components of equation (1.12), measured at any point inside a body, not only
will satisfy the differential equations of equilibrium (1.10), but also will satisfy the
external boundary conditions. Let us consider N be the normal to a boundary surface
point (Figure 1.18), and p the external stress acting at that point with components px, py,
σy σz y
τyx
τyz N
τxy m
σx zy σx x
τxz n
σz zx z l
σy
Internal stresses at a point Normal vector ‘N’
adjacent to the boundary point at a boundary point
FIGURE 1.18 Normal at boundary point and internal stresses at a point adjacent to the
boundary.
16 Finite Element Analysis of Solids and Structures
and pz in Cartesian coordinate directions x, y, and z, respectively. Direction cosines of
the vector N in the x-, y-, z-coordinate systems are given by l, m, and n (projections of
unit area in directions x, y, and z). Internal stresses at a point adjacent to the boundary
point are (σx, σy, σz, τxy, τyz, and τxz). Considering external stresses (px, py, pz) as continu-
ation of internal stresses, we get the stress-boundary conditions of equations (1.14):
px x .l xy .m xz .n
py xy .l y .m yz .n (1.14)
pz xz .l yz .m z .n
Equations (1.14) give the general stress-boundary conditions that must be always
validated by the calculated stress response of a given structural analysis problem.
Software tools for graphical post-processing of finite element model results can be
used effectively to visualize the calculated stresses and verify the boundary condi-
tions. While detailed descriptions of advanced stress analysis methods are introduced
in next chapters, Section 1.8 in the following introduces practice examples solely for
the learning objective of using software graphics tools for stress field visualization.
Interesting insights can be gained from the comparison of elementary beam theory
results with those obtained from detail 3D finite element analysis models.
1.8 PRACTICE PROBLEMS ON STRESS FIELD VISUALIZATION
WITH CAE TOOLS
Computer-aided-engineering (CAE) tools used in design and analysis of solid prod-
ucts can be generally classified in four groups (Figure 1.19): (1) computer-aided-
design (CAD) tools for geometric design, (2) finite element model preparation tools
(model pre-processors), (3) finite element analysis tools (solvers), and (4) result visu-
alization tools (post-processors). Many of the listed tools also provide cross-func-
tional capabilities. For example, CAD software tools also come bundled with
modelling, analysis and post-processing tools, while general-purpose finite element
(1) (2) (3) (4)
CAD tools Model pre-processor FEA solvers Post-processing tools
Caa ANSA ABAQUS HyperView
h p://www.3ds.com/products- h p://www.ansa-usa.com/products/ansa h p://www.3ds.com/products- h p://www.altairhyperworks.com/
services/caa/ services/simulia/products/abaqus/ Product,11,HyperView.aspx
Unigraphix Nx HyperMesh ANSYS Meta
h ps://www.plm.automaon.siem h ps://www.altair.com/hypermesh/ h ps://www.ansys.com/products h ps://www.beta-cae.com
ens.com/global/en/products/nx/ /structures /meta.htm
AutoCAD PRIMER NASTRAN ..
h ps://www.oasys-soware.com/ h ps://www.mscsoware.com/
h ps://www.autodesk.com/
dyna/soware/primer/ product/msc-nastran
products/autocad/overview
Solidworks FEMAP LS-DYNA ..
h ps://www.plm.automaon. h ps://www.lstc.com/
h ps://www.solidworks.com/
siemens.com/global/en/products/
simcenter/femap.html
Meshing and model building Soluons: [K]{u}={P} Results visualizaon
FIGURE 1.19 Partial list of CAE tools for design and analysis of solids.
Introduction to Stress Analysis of Solids and Structures 17
solvers, listed in group (3), also come bundled with solver-specific pre- and post-
processing tools. In the large-scale industrial product engineering setup, execution
efficiency is derived from the use of specialized tools for modularized work flow of
design, modelling, analysis and post-processing. The following practice problems
are presented based on that idea of modular process execution.
Practice Problem-1: A statically indeterminate solid beam, subjected to ideal-
ized load and boundary conditions, is shown in Figure 1.20. Skilled students
of the subject may attempt to prepare a finite element analysis model of the
problem. Alternatively, the ABAQUS finite element model (SOLIDBEAM_
COARSE_MESH.inp), provided in the data download site as listed in the
preface of this book, can be used to practice the post-processing part of this
example. A coarse mesh model, with number of nodes less than 1000, can
be run with student version of ABAQUS (Dassault Systems 2020a).
Successful execution of the ABAQUS model creates a binary results data-
base file, SOLIDBEAM_COARSE_MESH.odb, that can be post-processed
with ABQUS/CAE (Dassault Systems 2020a) or with general-purpose post-
processor HyperView (Altair University 2020). Questions to answer:
a. Identify nonzero stress components on the free-body diagram of a differen-
tial element adjacent to the bottom surface of the solid body.
b. Post-process the simulation model results to plot the shear stress distribu-
tion τxz (identified as S13 in ABAQUS results) in the solid body. Verify that
the simulation model results are consistent with the zero stress-boundary
conditions assumed in (a).
c. Cut a x-section at x = 100 and clip the model on one side; plot the contour
of normal stresses σx (identified as S11 in ABAQUS results)
d. Hand calculate normal stresses at the top and bottom of mid-span section by
using the beam bending theory.
e. How do hand calculated stresses compare with simulation model results?
30 kN point load in Z
direcon @ mid-span
All moons
constrained @ the end All moons
constrained @ the end
y x
FIGURE 1.20 A statically indeterminate solid beam subjected to idealized load and bound-
ary conditions.
18 Finite Element Analysis of Solids and Structures
10 kN point load in Z
direcon @ mid-span
All moons
constrained @ the end All moons
constrained @ the end
y x
FIGURE 1.21 A statically indeterminate hollow tube beam subjected to idealized load and
boundary conditions.
Practice Problem-2: A statically indeterminate hollow tube beam, subjected
to idealized load and boundary conditions, is shown in Figure 1.21. Once
again, skilled students of the subject may attempt to prepare a finite element
analysis model of this problem. Finite element model for ABAQUS analysis
is provided in file TUBE_COARSE_MESH.inp. Run this model in student
version of ABAQUS (Dassault Systems 2020a), and post-process the results
file (TUBE_COARSE_MESH.odb) to answer the following questions:
a. Plot the contour of deformed shape of the tube (z-displacement)
b. Cut a x-section at x = 100 and clip the model on one side; plot the contour
of normal stresses σx (identified as S11 in ABAQUS results) on the deformed
plot of tube
c. Hand calculate normal stresses at the top and bottom of mid-span section,
and the beam deflection at mid-span, by using the beam bending theory.
d. How do hand calculated stress and deflection results compare with simula-
tion model results?
2 Strain–Displacement
Relationship and
Elasticity of Materials
SUMMARY
Solid mechanics principles that form the foundation of advanced stress analysis
methods are reviewed in this chapter. The basic definitions of normal and shear
strains, for infinitesimally small deformation problems, are introduced in Section
2.1. Section 2.2 develops very important strain–displacement relationships that are
critical for the development of finite element formulations in the next steps. Inter-
relationship among strain and displacement components, defining the well-known
compatibility condition, is presented in Section 2.3. Any derived solution for strain/
displacement field must satisfy the deformation compatibility condition – a funda-
mental requirement of the solid mechanics principles. Section 2.4 presents yet
another important formulation – the generalized stress–strain relationships based on
Hookes’ law. All these inter-relationships among stress, strain, and displacement
components are finally expressed through redefinition of the compatibility condition
in terms of stress components. The stress-based compatibility definition, with stress
equilibrium and boundary conditions defined in Chapter 1, presents the possibility of
solving the unknown stress field problems by using analytical stress functions.
Elementary example solution, based on stress function approach, is presented in
Section 2.5. The alternative solution method, based on the use of displacement func-
tion approach, is presented in Section 2.6. Although the stress function-based formu-
lation technique is not the preferred implementation method in standard finite
element software packages, the method however remains potent for specialty prob-
lem solutions that will be revisited in Section 12.7. Section 2.7 presents the core
formulation of finite element stiffness properties, using the strain–displacement and
stress–strain relationships derived in earlier sections of this chapter. This core formu-
lation will be used repeatedly in subsequent chapters to derive the stiffness properties
of 2D and 3D solid elements. Detailed steps of the stiffness formulation technique,
based on linear displacement variation functions, are demonstrated with simple two-
node truss element in Section 2.8. Section 2.9 introduces higher order truss/cable
element – capable of reproducing quadratic variation of displacement response over
the element domain. Key building blocks of a finite element simulation model are
described in Section 2.10 – with ABAQUS specific description of data input syntax.
Other solver software products have their own syntax rules that a user should be
aware of while preparing a similar analysis model. Two practice problems are pre-
sented in Section 2.11 with the learning objective of manually preparing simple finite
element model input files. Understanding the syntax rules of finite element data files
19
20 Finite Element Analysis of Solids and Structures
is an essential skill for diagnosis of syntax errors that often emerge during the soft-
ware-driven preparation of complex finite element models in real engineering appli-
cations. Absence of this understanding often forces the unnecessary reformulation of
complex models that could be easily fixed by simply reviewing the text input files.
2.1 MEASUREMENT OF DEFORMATION INTENSITY (STRAIN)
Material deformation behavior is often tested under uniaxial loading condition.
Figure 2.1 shows a material specimen subjected to a uniaxial load, F, applied at a
very slow rate (termed as “quasi-static” testing). Initial gage length of the specimen,
Δx (before loading), changes by incremental quantity Δu under loading action. Rate
of change of deformation in direction x, x being normal to the cross-section plane A0,
is defined as “normal” strain in the material:
u u (2.1)
x lim
x 0 x x
The above definition of strain can be expanded to describe bi-directional normal
strains inside a differential element in x–y plane (Figure 2.2(a)). Material may also
undergo distortional deformation, i.e. initial position of segment OA (Figure 2.2(b))
may rotate by angle ∂v/∂x, and segment OB by ∂u/∂y. Total angular distortion to ini-
tial right angle AOB, referred to as shear strain of the material in x–y plane, is given
by
u v (2.2)
xy yx
y x
By definition, shear deformation in the material occurs without accompanying
length changes, i.e. without affecting normal strains in the material. Shear strains,
unlike the normal strains, are not generally measured directly in test specimens.
Instead, normal strains are measured by using electric-resistance strain gages – short
Force, F
∆x + ∆u
A0
FIGURE 2.1 Measurement of strain in one-dimensional loading condition.
Strain–Displacement Relationship and Elasticity of Materials 21
+ .
B
=
+ .
dy dy
=
O A
dx dx
(a) (b)
FIGURE 2.2 Measurement of strains in bi-directional loading condition: (a) normal strains;
and (b) shear strain.
length of insulated wires glued to a target surface. When stretching occurs, electric
resistance of the wire increases indicating the deformation intensity in the direction
of measurement. Generally, 3 gages, placed 45° or 60° apart, are used in a cluster,
called “strain rosette” (Figure 2.3). Relationships among measured normal strains
(εa, εb, εc) and the strains εx, εy, γxy (in the general reference coordinate system x,y) are
given by equations (2.3) that can be solved to determine the normal and shear strain
values in (x–y) reference system:
a x cos2 a y sin 2 a xy sin a cos a
b x cos2 b y sin 2 b xy sin b cos b (2.3)
c x cos2 c y sin 2 c xy sin c cos c
Three strain components (εx, εy, and γxy) describe the complete deformation state
in the 2D plane of solids.
b y
c
a
θc θb
θa
x
FIGURE 2.3 Normal strain measurement with an array of strain gages (“strain rosette”).
22 Finite Element Analysis of Solids and Structures
2.2 GENERAL DESCRIPTION OF DEFORMATION STATE IN
3D SOLIDS
Strain components defined in Section 2.1 for 2D reference plane (x–y) can be
expanded to define strains in a 3D differential element as in the following equation
(2.4):
u v w
x , y , z
x y z
u v
xy yx
y x (2.4)
v w
yz zy
z y
u w
xz zx
z x
Similar to the definition of stress components, symmetry applies to shear strain
components as well (i.e. γxy = γyx, γyz = γzy, and γxz = γzx). Equations (2.4) can be rewrit-
ten in the following matrix and vector forms:
x 0 0
0
0
x y
y 0 0
u
z z
. v (2.5)
xy
0 w
yz y x
zx 0
z y
0
z x
Differential operator matrix on the right side of equation (2.5), defining the strain–
displacement relationships, is same as the one shown in equation (1.13) for stress
equilibrium condition. Re-writing equation (2.5) using matrix and vector notations
gives the following:
B . u (2.6)
where {ε} is the vector of six strain components (εx, εy, εz, γxy, γyz, γzx) and {u} is the
vector of three displacement components (u, v, w). Equation (2.6) presents a general
form of strain–displacement relationship that will appear repeatedly throughout the
subsequent discussions on finite element formulations.
Strain–Displacement Relationship and Elasticity of Materials 23
2.3 COMPATIBILITY OF STRAIN (DEFORMATION FIELD)
IN A BODY
Kinematic relations (equations 2.5) relate six components of strains to three defor-
mations u, v, and w. All six components of strain are, therefore, not completely inde-
pendent of one another. Differentiating εx in equation (2.4) twice with respect to y, εy
twice with respect to x, and γxy with respect to x and y, we get
2 x 3u 2 y 3v 2 xy 3u 3v (2.7)
, 2 , 2
y 2
x y 2 x 2
x y x y x y 2
x y
Combining the three parts in equation (2.7), we get
2 x 2 y 2 xy (2.8)
2
y 2 x x y
Similarly, on (y,z) and (x,z) planes, we get
2 y 2 z 2 yz (2.9)
2
z 2 y y z
2 z 2 x 2 xz (2.10)
2
x 2 z x z
Similar to previous steps, additional relationship can be derived by differentiating
εx with respect to y and z, εy with respect to z and x, and εz with respect to x and y:
2 x (2.11)
2 yz xz xy
y z x x y z
2 y (2.12)
2 yz xz xy
z x y x y z
2 z (2.13)
2 yz xz xy
x y z x y z
Inter-relationships among strain components, shown in equations (2.8)–(2.13),
define compatibility conditions among strains inside a body. Strain components, cal-
culated from deformation variables (u, v, w), satisfy all six compatibility equations in
a 3D problem – an essential requirement that the finite element formulations, to be
presented later, must meet.
2.4 STRESS–STRAIN RELATIONSHIPS (HOOKE’S LAW)
Uniaxial test specimens of ductile materials (Figure 2.4) experience changes in
cross-section area and gage length during the loading phase. However, it is general
24 Finite Element Analysis of Solids and Structures
Engineering stress:
s = F / A0 Force, F
L0
A0
Engineering strain: Change in length, ∆L
e = ∆L / L0
FIGURE 2.4 Engineering stress–strain response of a ductile material.
engineering practice to convert the measured force vs deformation response (F vs
ΔL) to engineering stress vs strain relationship by using initial cross-section area (A0)
and initial gage length (L0) (Figure 2.4). True stress at any given moment of loading
can be defined by F/A; and true strain by ΔL/L, where A and L are, respectively,
instantaneous area and gage length. Assuming no overall volume changes in the
material during deformation, i.e. A0L0 = AL, true stresses and strains can be calcu-
lated from the engineering measurements as shown in Figure 2.5. The stress level Sy
defines a significant change in the material behavior. When applied stress σ < Sy,
normal stress σ is linearly proportional to normal strain ε. This linear elasticity rela-
tionship between stress and strain is known as Hooke’s law (Robert Hooke 1635–
1703). Proportionality constant, E, is known as elastic modulus or Young’s modulus
(Thomas Young 1773–1829). E is unique for each material – generally same in
True stress:
σ = F / A = s * (1+e)
Force, F
sy
variable
variable, A
L
E
Change in length, ∆L
True strain:
= ∆L / L = ln (1+e)
FIGURE 2.5 True stress–strain response of the material.
Strain–Displacement Relationship and Elasticity of Materials 25
tension and compression. Engineering unit of E is same as that of stress σ (for exam-
ple, GPa or kN/mm2). Denoting the normal direction to cross-sectional area as x
(same as loading axis), Hooke’s law defining the normal stress–strain relationship is
given by equation (2.14):
x (2.14)
x
E
Experimental data show that elastic axial strain (in tension or compression) is
accompanied by lateral strain (contraction or expansion), and the two are related by
a constant of proportionality, ν (known as Poisson’s ratio after S.D. Poisson
1781–1840):
x (2.15)
y z
E
For most metals: ν = 0.25–0.35. Stress–strain relationships (2.14) and (2.15),
defined for x-directional loading, can also be applied to y- and z-directional loading
effects as follows:
y
y
E (2.16)
x z y
E
z
x y
E (2.17)
z
z
E
Assuming that the deformations are very small (without any effect on external
loads), principle of superposition provides the following generalized Hooke’s law for
normal stresses and strains in 3D homogenous isotropic material:
1
x x y z
E
1
y y x z (2.18)
E
1
z z x y
E
To derive the shear stress–strain relationship of elastic materials, a special case of
biaxial loading on a rectangular element (σy = σ, σx = –σ, σz = 0) is considered in
Figure 2.6. A square element “abcd” is cutout from the center where sides ab, bc,
etc., are at 45° angle to reference axes x and y. Considering the equilibrium of trian-
gular element “obc”, we get the relationship in equation (2.19):
26 Finite Element Analysis of Solids and Structures
=0
y b
O c b
b
4 2
450 c
a O x a O c
d
d
FIGURE 2.6 Internal shear deformation in a differential element under tension–compres-
sion biaxial loading.
2 sin 45 1 (2.19)
Under the tension–compression combination of stresses, diagonal “bd” elongates
and “ac” shortens – causing a change in angle “abc” by “γ” giving the following
equation (2.20):
oc (2.20)
tan
4 2 ob
Expanding the trigonometric terms on the left-hand side of equation (2.20), and
assuming small angular distortion, tan (γ/2) ≈ γ/2, equation (2.20) can be re-written
as
1
2 oc 1 x (2.21)
ob 1 y
1
2
Inserting (σy = σ, σx= –σ, σz = 0) in equations (2.18) and combining that with equa-
tion (2.21), we get
1
1 1
2 E
1 (2.22)
1 1
2 E
Strain–Displacement Relationship and Elasticity of Materials 27
Substituting (σ = τ) from equation (2.19), equation (2.22) leads to the following
relationship between shear stress and shear strain:
1 E
. G. (2.23)
2 E 2 1
G is called shear modulus or modulus of rigidity. Shear stress–strain relationship
of equation (2.23) derived for element in 2D plane can be expanded to define the 3D
relationships as follows:
xy
xy
G
(2.24)
yz yz
G
zx
zx
G
Normal stress–strain relations of (2.18) and shear stress–strain relationships of
(2.24) define the complete stress–strain relationships for 3D problems. Inverting
equations (2.18) and (2.14), we get the following generalized Hooke’s law for 3D
stresses and strains:
1 0 0 0
x 1 0 0 0
x
1 0 0 0
y
y
z E
1 2
0
z (2.25)
0 0 0 0
xy 1 1 2
2 xy
yz 1 2
0 yz
0 0 0 0
2 zx
zx
0 0 0 0 0
1 2
2
Re-writing equations (2.25) with vector and matrix notations gives the following:
{ ) C (2.26)
where {σ} is the vector of six stress components in 3D element, {ε} is the vector of
corresponding strain components, and [C] is the “elasticity” property matrix defining
the stress–strain relationships for 3D solid elements. Like the general strain–
displacement relationship of equation (2.6), general stress–strain relationship of
28 Finite Element Analysis of Solids and Structures
equation (2.26) will also appear repeatedly throughout the subsequent discussions on
finite element formulations. The component details of [C], however, will vary
depending on the response mechanism of different finite element formulations.
Stress–strain relationships of equation (2.25) can also be written in another well-
known form as follows (Timoshenko and Goodier 1982):
x e 2G x
y e 2G y (2.27)
z e 2G z
where e is the volumetric strain given by e = εx + εy + εz; and λ and G are called
Lame’s constants. Shear modulus G has been defined in equation (2.23) and λ is
defined as follows:
E (2.28)
1 1 2
For special case of uniform hydrostatic pressure, σx= σy = σz= –p, relationship
between pressure p and volumetric strain e is given by equation (2.29), where K is
called the bulk modulus of elasticity:
E (2.29)
p .e K .e
3 1 2
For incompressible materials, e approaches 0 implying that Poisson’s ratio
approaches a value of 0.5. Stress–strain relations, defined in terms of λ and G, are
often used in finite element literature related to the modeling of incompressible mate-
rial behavior.
2.5 SOLUTION OF ELASTICITY PROBLEMS USING STRESS
DISTRIBUTION FUNCTIONS
Three equations of stress equilibrium (equations 1.10), six equations of stress–strain
relations (equation 2.25), and six equations of strain compatibility conditions (equa-
tions 2.8–2.13) give a total of 15 equations that are sufficient to determine 15 response
variables (six stress components, six strain components, and three displacement vari-
ables). However, compatibility equations (2.8–2.13), expressed in terms of strain
derivatives, can be re-written in terms of stresses by using the Hooke’s law (2.25).
Combining those with the stress equilibrium equations (1.10), the following set of
equations for stress compatibility conditions can be obtained for the special case of
no body forces, i.e. Fx=Fy=Fz = 0 (Timoshenko and Goodier 1982):
Strain–Displacement Relationship and Elasticity of Materials 29
2 2
1 2 x x 2 0 1 2 xy xy 0
2 2
1 2 y y2 0 1 2 yz yz 0 (2.30)
2 2
1 2 z z 2 0 1 2 zx zx 0
2 2 2
where
2
and x y z
x 2 y 2 z 2
Stress distributions in an isotropic body must satisfy all three conditions: compati-
bility equations (2.30), equilibrium equations (1.10), and stress boundary conditions
(1.14). These 12 equations are generally sufficient to determine the stress field without
ambiguity. Equations of compatibility contain only second derivatives of the stress
components. Hence, if the external forces are such that the equations of equilibrium
and the stress boundary conditions are satisfied by taking stress components either as
constants or as linear functions of the coordinates, the equations of compatibility will
be automatically satisfied – thus giving correct solution of the stress distribution.
EXAMPLE: STRETCHING OF A PRISMATIC BAR BY ITS OWN
WEIGHT
Figure 2.7 shows a bar stretched by its own weight downward from the upper
support at z = ℓ. Assuming ρ is the density of the material, and g is the accelera-
tion due to gravity, body forces per unit volume are given by
Fx Fy 0; Fz g (2.31)
z
FIGURE 2.7 Stretching of a prismatic bar by its own weight.
30 Finite Element Analysis of Solids and Structures
From observation, stress distribution functions inside the body can be defined as
follows:
z gz; x y xy yz zx 0 (2.32)
Stress components defined in equation (2.32) readily satisfy the differential equa-
tions of stress equilibrium (1.10). Stress boundary conditions of zero value on the
surface, zero at the free end (z = 0), and a uniformly distributed reaction of σz = ρgℓ
at the upper end are also satisfied by equations (2.32). Compatibility conditions
(2.30) are also satisfied. So, equations (2.32) give correct description of the stress
functions in the body under the action of self-weight only. Strains in the body can be
found by inserting the stress functions (equations 2.32) in Hooke’s laws (2.18 and
2.24). Integration of the strain–displacement relationships (2.4) provides the follow-
ing equations for displacement responses (Timoshenko and Goodier 1982):
gxz
u
E
gyz (2.33)
v
E
gz 2 g 2 2 gl 2
w
2E
2E
x y
2E
Evidently, because of the Poisson’s effects, lateral displacements, u and v, have
nonzero values except along the axis of the member (x = y = 0). Solution technique,
based on stress function approach, as demonstrated with the elementary example of
prismatic bar, is useful when the stress distribution inside a body can be described
analytically based on a priori knowledge of the system response. Alternative analyti-
cal solution technique, based on displacement field assumption, is described in
Section 2.6.
2.6 SOLUTION OF ELASTICITY PROBLEMS USING
DISPLACEMENT VARIATION FUNCTIONS
In this analysis technique, displacement variation functions, u(x, y, z), v(x, y, z), w(x,
y, z), are introduced to satisfy all necessary conditions, i.e. boundary conditions,
equilibrium equations, and compatibility conditions.
EXAMPLE: STRETCHING OF A PRISMATIC BAR UNDER
AXIAL TENSION
We consider a prismatic bar subjected to an axial tensile force, P, at end (2),
and is restrained at end (1) (Figure 2.8). Considering a linear variation of defor-
mation over the element length, the following displacement response functions
are assumed for the element (Poisson’s effect ignored):
= =
u ax v 0 w = 0 (2.34)
Strain–Displacement Relationship and Elasticity of Materials 31
y Cross sec on: A
Elas c modulus: E
1 2 P x, u
FIGURE 2.8 Stretching of a prismatic bar under axial tension.
Using the strain–displacement relationships of equation (2.4), we get the follow-
ing expressions for strains inside the body:
u v w
x a y 0 z 0 xy yz zx 0 (2.35)
x y z
Strain components in equation (2.35) satisfy the compatibility conditions defined
in equations (2.8)–(2.13). Displacement boundary condition of u = 0 at x = 0 is also
satisfied by the displacement functions (2.34). Now considering the force boundary
condition at x = L (node 2):
Internal resistance A.E. x A.E.a Extenral force P (2.36)
x L
Equation of equilibrium (2.36) yields, a = P/AE. Substituting it in equations (2.34)
and (2.35), we get the following functions for displacements, strains, and stresses:
Px
u v 0 w 0
AE
P (2.37)
x y z 0 xy yz zx 0
AE
P
x y z 0 xy yz zx 0
A
Equations (2.37) represent the correct solution for the problem in Figure 2.8 as all
necessary conditions (compatibility, equilibrium, and boundary conditions) have
been satisfied. Solution method outlined above proceeded through a systematic pro-
cedure of displacement function definition, verification of necessary conditions, cal-
culation of unknown coefficients in displacement functions from the boundary
conditions, followed by calculation of strains and stresses.
Stress function method in Section 2.5 started with the assumption of stress distri-
bution definition; and followed similar systematic procedure as the displacement
method. However, a priori definition of either of the field variable, displacement or
stress, is often not evident for general shaped bodies with complex loading and
boundary conditions, such as the example shown schematically in Figure 1.12. The
definition of displacement variation, however, provides a convenient tool to define
the response mechanism of general shape smaller elements that can be
32 Finite Element Analysis of Solids and Structures
inter-connected to define the overall response of a complex element assembly (Figure
1.13). This concept of discretizing the displacement field of a continuous solid body,
to an assembly of interconnected elements, forms the basis of finite element simula-
tion models. A basic requirement of finite element formulation is the derivation of
element stiffness properties that can be used to solve the system equilibrium equa-
tions involving nodal DOF (equations 1.5). An objective formulation of the element
stiffness property is desired to find a unique solution for the overall displacement
response without ambiguity. Several theoretical alternatives, eventually leading to
the same element stiffness definition, are often discussed in books on finite element
methods. Among different techniques (virtual work, minimization of potential
energy, weighted residual methods, etc.), the virtual work method is presented in
Section 2.7 as it leads to the target formulation in a short easily interpretable way.
2.7 STIFFNESS METHOD (FINITE ELEMENT METHOD)
OF STRUCTURAL ANALYSIS
System equilibrium equations based on stiffness properties of discrete spring ele-
ments have been discussed in Section 1.4. Following derivations present a general-
ized procedure for stiffness property calculations of a solid element (Figure 2.9),
hereafter referred to as “finite element”. Element is connected to a set of nodes that
have displacement degrees of freedom, {u}. External forces acting on the nodes are
denoted by vector {P}. Internal element resistances, producing equilibrium with
external forces, generate a set of internal stresses {σ}. The corresponding internal
strains are represented by the vector {ε}. Let us impose a virtual displacement, Ū, on
the system (Ū is consistent with the prescribed boundary conditions). Using the
strain–displacement relationship of equation (2.6), virtual strains corresponding to
the virtual displacement field Ū are given by
B U (2.38)
Internal virtual work done by the strains :
Internal virtual work dV U T B dV
T T
(2.39)
External virtual work done on the forces {P}:
Variables @ nodes:
Nodal forces: {P}
Displacements: {u}
Internal responses:
strains: { }
Stresses: { }
FIGURE 2.9 Definition of a finite element.
Strain–Displacement Relationship and Elasticity of Materials 33
External virtual work U . P
T
(2.40)
Equating the internal virtual work with the external virtual work, we get
U T B dV U . P
T T
(2.41)
As the relationship (2.41) is valid for any value of virtual displacement, the equal-
ity of multipliers must exist, i.e.:
B . .dV P
T
(2.42)
Equation (2.42) represents the equilibrium between internal stresses and external
forces. Substituting the relationships (2.26) and (2.6) into equation (2.42), and after
re-arranging the terms, we get
B T . C . B .dV . u P
(2.43)
Comparing equation (2.43) with the familiar system equilibrium equation (1.5),
we get the following general definition of the finite element stiffness matrix:
k B . C . B .dV
T
(2.44)
Integral in equation (2.44) is calculated with numerical integration – accuracy of
which depends on element types and integration order chosen in specific element formu-
lations. As discussed earlier, finite element analysis procedure starts with breaking down
of a large complex system into an assembly of smaller “elements” – interconnected at
“node” points (Figure 1.13). Deformation response of the body is represented by the
vector of displacement values at discrete nodes. Analytical assumptions are used to
describe local deformation inside finite elements, and to determine the derivatives of
displacements that define the strain–displacement relationship matrix [B] in equation
(2.6). Element-specific geometric property matrix [B], and the material stress–strain
relationship matrix [C] (equation 2.26), are used to calculate the individual element stiff-
ness matrix (equation 2.44). Calculated element stiffness matrix is mapped into the ele-
ment-specific degrees of freedom in the NxN stiffness matrix in equation (1.5). Overall
stiffness matrix of the larger body is assembled from the individual contributions of
finite elements that make up the entire structure (Figure 1.13).
Load vector, {P}, on the right side of equation (2.43) is calculated as follows:
P BT 0dV f BdV f s dS Rc (2.45)
where σo is initial stress, f B is internal body force per unit volume, f S is surface force
per unit surface area, and Rc represents concentrated forces applied directly at model
34 Finite Element Analysis of Solids and Structures
nodes. Following the assembly of system stiffness matrix and load vector from the
contributions of individual elements, the following steps are executed to determine
the finite element model responses:
• Identify known forces and displacement constraints
• Modify equations of equilibrium for known boundary conditions and external
loads
• Solve equations of equilibrium to determine unknown displacements: [k] * {u}
= {P}
• Calculate strains inside all elements: {ε} = [B]{u}
• Calculate internal stresses: {σ} = [C]{ε}
• Total strain energy stored can be calculated from the summation of element
1 T
contributions: U 0 . .dV
2
A critical step in developing the finite element model of a structure involves the
assumption of displacement field inside individual elements – that in turn define the
strain–displacement relationship matrix [B] in the above calculations. The assump-
tion of known displacement variation inside elements is analogous to the displace-
ment method solution technique discussed in Section 2.6. However, finite element
method does not impose an analytical displacement field equation over the entire
geometry of solution domain. Piece-by-piece analytical description of discrete finite
element responses reproduces the complex displacement response of a general struc-
tural body.
The finite element stiffness formulation, represented by equation (2.44), applies to
deformation response field of any complexity (axial, shear, bending, torsion, etc.).
The remainder of this chapter will focus on the demonstration of stiffness formula-
tion method for the simplest element form that provides resistance to axial deforma-
tion only (truss, cable, etc.). Formulations for more complex 2D and 3D finite
elements will be introduced in subsequent chapters.
2.8 STIFFNESS PROPERTIES OF 1-D TRUSS ELEMENT PROVIDING
RESISTANCE TO AXIAL DEFORMATION ONLY
Figure 2.10 shows two example structural systems, (a) a cable supported bridge and
(b) a truss tower, that are assemblies of members providing resistance to axial defor-
mation only. Displacement degrees of freedom, at nodes 1 and 2 of element in Figure
2.11, are identified as u1 and u2. Positions of the two end nodes of element are defined
by global coordinates x1 and x2. A local coordinate definition r is introduced (Figure
2.11) to define any point inside the element in the range of r = –1 to r = +1. For linear
variation of coordinate over the element length, interpolation functions for internal
coordinate locations are given by, H1 = (1 – r)/2 and H2 = (1 + r)/2. Coordinate posi-
tion, x, at any point inside the element is defined as follows:
x H1. x1 H 2 . x2 (2.46)
Strain–Displacement Relationship and Elasticity of Materials 35
Cables
Downward load
(a) Cable supported bridge (b) Truss structure
FIGURE 2.10 Structural assemblies of elements providing resistance to axial deformation
only.
u1 u2
Displacement DOF
L
x x=x2 Nodal coordinates
x=x1
r
Internal reference coordinate
r= –1 r=0 r=+1
1
= (1 )
2 Interpola on
1 func ons
= (1 + )
2
FIGURE 2.11 Interpolation functions for the nodal variables of a truss element.
After inserting expressions for H1 and H2 in equation (2.46), and re-arranging the
terms:
1 x x
r x 1 2
L / 2
(2.47)
2
36 Finite Element Analysis of Solids and Structures
where L is the length of the element (L = x2 – x1). Taking partial derivative of both
sides of equation (2.47) with respect to x:
r 2 (2.48)
x L
Assuming that the deformation varies linearly over the element length, inter-
nal deformation at any point inside the element can be expressed in terms of the
nodal displacements by using the same coordinate interpolation functions of
Figure 2.11:
1− r 1+ r (2.49)
u = H1.u1 + H 2 .u2 = .u1 + .u2
2 2
Internal strain caused by the nodal displacements, u1 and u2 (Figure 2.11), can be
defined as follows:
u u r (2.50)
.
x r x
Substituting the expressions (2.48) and (2.49) into equation (2.50), we get the fol-
lowing definition for internal strain:
1
e=
L
u2 u1 (2.51)
Relationship between internal strain and nodal displacements, given by equation
(2.51), can be re-written in the following form:
1 1 u1
= L
L u2
(2.52)
Comparing equation (2.52) with (2.6), strain–displacement relationship matrix
[B] for two-node truss element of Figure 2.11 is defined as follows:
1 1 (2.53)
B
L L
Stress–strain relation for one-dimensional axial deformation response of the truss
element is simply defined by the elastic modulus, E, thus giving the following defini-
tion of elasticity matrix [C]:
C E (2.54)
Substituting the definitions of [B] and [C] from equations (2.53) and (2.54) into
equation (2.44), stiffness matrix of two-node truss element is calculated as follows:
Strain–Displacement Relationship and Elasticity of Materials 37
1/L 1 1
k
1/L
E L
L
dV
(2.55)
Element geometry and material parameters inside equation (2.55) are constants
that can be taken out of integral sign, thus, reducing the integral operation in equation
(2.55) to a simple integral of the volume of the truss element. Writing dV = A.dx (A
being the cross-sectional area of prismatic bar), we get the following modified form
of equation (2.55):
E E
L2 L2 . x2
k =
E E
x1
A.dx (2.56)
L2 L2
Integral in equation (2.56) is simply the volume of the element [= A(x2 – x1) = A.L].
Stiffness matrix of the 2-DOF truss element, aligned with the coordinate direction x
(Figure 2.11), is thus given by the following equation:
AE AE
L L
k (2.57)
AE AE
L L
Stiffness matrix in equation (2.57) has been derived based on axial resistance of truss
element against the displacement DOF u1 and u2 that are aligned with member axis direc-
tion, x (Figure 2.11). A truss element in actual structural assembly can be oriented in any
direction in 3D space. Figure 2.12(a) shows 3 global displacement DOF at each end node,
i and j, of an arbitrarily oriented two-node truss element. Figure 2.12(b) shows local
member axis, x, and relevant local DOF, u1 and u2, for internal element response calcula-
tions. Relationships between global and local DOF are defined as follows:
Ui
Vi
u1 lij mij nij 0 0 0 Wi
u T U (2.58)
u2 0 0 0 lij mij nij U j
Vj
W j
where lij, mij, and nij are direction cosines of member (i,j) with reference to the global
coordinate system (X, Y, Z). Stiffness matrix of truss element, calculated with refer-
ence to local member axis direction (equation 2.57), can be transformed by using the
coordinate transformation matrix [T] (equation 2.58), to get the stiffness matrix of
truss element having 6-DOF in global coordinate directions (Figure 2.12(a)), given
by equation (2.59):
38 Finite Element Analysis of Solids and Structures
Vj
x u2 x
Y Wj
Uj
Vi
ε,σ
u1
X
Ui
Z Wi
1
: = : =
2
(a) (b)
FIGURE 2.12 (a) 6-DOF truss element in 3D space with 3-DOF at each node; (b) local
reference direction for axial response calculation of two-node truss element.
S ffness matrix of element E1 S ffness matrix of element E2
11 12 13 14 15 16 44 45 46 47 48 49
21 22 23 24 25 26 54 55 56 57 58 59
31 32 33 34 35 36 64 65 66 67 68 69
41 42 43 44 45 46 74 75 76 77 78 79
51 51 53 54 55 56 84 85 86 87 88 89
61 62 63 64 65 66 94 95 96 97 98 99
U2 U8
U1
1 U5 3
U7
U3 E1 E2 U9
2
U4
U6
FIGURE 2.13 Stiffness matrices of two truss elements each having 6 displacement DOF.
T
K 6 x 6 T 6 x 2 k 2 x 2 T 2 x 6 (2.59)
where [k]2×2 is the local stiffness matrix defined in equation (2.57) and [T] is the
coordinate transformation matrix defined in equation (2.58).
Figure 2.13 shows examples of 6 × 6 stiffness matrices for two truss elements
having arbitrary orientation in 2D space. Following the matrix assembly procedure
Strain–Displacement Relationship and Elasticity of Materials 39
U2 U8
where s ffness
U1 @ shared DOF:
1 U5 3
U7 Kij = sij + qij
U3 E1 E2 U9
2
U4
U6
FIGURE 2.14 Overall stiffness matrix of two truss element assembly.
described in Section 1.4, stiffness matrix of a structure can be built from the contribu-
tions of as many elements as needed to completely describe the structure. Figure 2.14
shows an assembled stiffness matrix for two truss element systems. After introducing
nodal loads and boundary conditions, system responses can be calculated following
the same procedure described in Section 1.4 for two-spring system. The same proce-
dure can also be used for structural systems comprising of a very large number of
elements.
Analysis steps described in the above present the simplest form of finite element
analysis of a structural assembly that is made up of members capable of providing
resistance to axial deformation only. More complex structural response predictions
require more advanced formulations for the calculation of element stiffness proper-
ties that will be introduced in Chapters 3–8.
2.9 HIGHER ORDER TRUSS ELEMENT AND MODEL
REFINEMENT
Two-node truss element described in Section 2.8 allows linear variation of deforma-
tion over the element length (Figure 2.11). However, truss element can be re-formulated
with one additional node in the middle to allow higher order displacement variation
within the element (Figure 2.15). Interpolation functions H1, H2, and H3 (Figure 2.15)
can be used to derive the stiffness matrix of this three-node truss element by follow-
ing the same procedure described for two-node element in Section 2.8. Three-node
truss element is specifically suitable to model the curved profile of a cable as shown
in Figure 2.16. A trivial attempt may lead to a model of the cable with multiple two-
node truss elements – describing the curved geometric profile with piece-wise linear
approximations. However, a cable made with multiple two-node elements will have
40 Finite Element Analysis of Solids and Structures
u1 u2 u3
Displacement DOF
L
x x=x x=x3 Nodal coordinates
x=x1 2
r
r=0 r=+1
Internal reference coordinate
r= –1
1 1 2
1= 1− − (1 − )
2 2
2
2= (1 − )
1 1 2
3= 1+ − (1 − )
2 2
Interpola on
func ons
FIGURE 2.15 Interpolation functions for the nodal variables of a three-node truss element.
1 3
2
FIGURE 2.16 Cable modeled with a three-node truss element.
1 5
4
2
3
1 5
4
2
3
FIGURE 2.17 Instability in multi-element model of a cable.
inherent instability (as shown in Figure 2.17), thereby not meeting the essential
requirement of a positive definite stiffness matrix formulation (Section 1.4). While a
failed analysis is an indicator of structural instability, it is better to use upfront engi-
neering judgment to determine the stability of a structural analysis model before
embarking on the analysis task.
Strain–Displacement Relationship and Elasticity of Materials 41
2.10 MODEL PREPARATION FOR COMPUTER-AIDED ANALYSIS
OF STRUCTURES WITH FINITE ELEMENT SIMULATION
SOFTWARE
Sections 2.7 and 2.8 have provided detailed descriptions of stiffness method analysis
of structures – built as assembly of truss elements providing resistance to axial defor-
mation only. In computer-aided analysis exercise, the actual model description tasks
are undertaken by the engineers while all calculation tasks are left for finite element
analysis software packages (group-3 in Figure 1.19). Each of those analysis software
products has its own syntax rules for model description inputs although basic model
input parameters (nodes, elements, loads, etc.) are similar. Example demonstrations
in this book are presented for ABAQUS models (Dassault Systems 2020b). General-
purpose model preprocessors, such as HyperMesh (Altair University 2020), possess
the capability to import model input files of one finite element solver, and to export
out the input data file for another solver. Computer-aided finite element model prepa-
ration process will be discussed in Chapters 3 and 4. Discussions in this chapter will
remain limited to manual input data preparation for small-scale structural analysis
problems. Descriptions of key data for simple truss analysis problem of Figure 2.18
are presented in the following. Data input syntax for ABAQUS software is presented
101 103
12000 mm
102
Downward load: 0.0016 kN/mm
60000 mm
60000 mm
Simplified model
of suspension
cable
y
x
101 103
12000 mm
1001 1002
102
60000 mm 60000 mm
P=96 kN
FIGURE 2.18 Simplified two truss element model of a suspension cable.
42 Finite Element Analysis of Solids and Structures
as a specific example. Specific syntax rules will need to be followed if a different
analysis software is used for the practice problems.
Nodes: Each node in a model has a unique identification (ID) number and
Cartesian coordinates. A set of nodes can be identified by assigning a name
with the parameter “NSET”. The following data block describes a set of
three nodes with ID numbers, 101, 102, and 103, and their respective x-, y-,
z-coordinates.
*NODE, NSET=setn1
101,0.0,0.0,0.0
102,60000.0, −12000.0,0.0
103,120000.0,0.0,0.0
Elements: Each element in a model has a unique ID number and a list of nodes
that it attaches to. A set of elements can be grouped together with a name
assigned by the parameter “ELSET”. Element description also requires a
“TYPE” deceleration with a standard type name selected from the ABAQUS
element library. Based on selected type, ABAQUS chooses the specific cal-
culation procedures for stresses, strains, and stiffness properties. For exam-
ple, two-node truss element discussed in Section 2.8, is identified in
ABAQUS library as “T2D2” – meaning a truss element in 2D space attached
to 2 end nodes. Following is an example description of two truss elements
in a set named “cable-1”:
*ELEMENT, ELSET=cable-1, TYPE=T2D2
1001, 101, 102
1002, 102, 103
Element properties: Elements require type-specific property definitions. Truss
element type discussed in Section 2.8 requires element cross-sectional area
for stiffness calculation. Following example describes the member cross-
section area (A = 4560.4 mm2) for element set cable-1:
*SOLID SECTION, ELSET=cable-1, MATERIAL=mat1
4560.4
Material Properties: Material name identified in member property data must be
defined with data block identifier *MATERIAL in ABAQUS input file. For
temperature-independent elastic stress–strain response analysis, Young’s
modulus and Poisson’s ratio are the only required material parameters.
Example material data block for material named “mat1” is shown in the
following assigning Young’s modulus=210.0 GPa and Poisson’s ratio=0.3:
*MATERIAL, NAME=mat1
*ELASTIC
210.0,0.3
Strain–Displacement Relationship and Elasticity of Materials 43
Boundary conditions: External boundary conditions on nodes are defined by
data block identifier *BOUNDARY. Boundary constraints can be imposed
on selected DOF (1 to 6) to a node ID or to a set of nodes (identified by a set
name). The following example imposes zero displacement constraint to two
DOF (1–2) of nodes 101 and 103:
*BOUNDARY
101,1,2,0.0
103,1,2,0.0
Loads: *CLOAD identifies data block for concreted nodal loads. Each nodal
load is described with node ID followed by associated DOF and the load
magnitude. The following example specifies a load of 96 kN in negative y
coordinate direction at node ID 102:
*CLOAD
102,2,-96.0
Analysis step: Data block for analysis step in ABAQUS starts with identifier
*STEP; and it ends with identifier *END STEP. In between those two com-
mand lines, analysis types, loads, and output requests can be included. A
small-displacement linear elastic analysis type is specified by including
“PERTURBATION” key word with *STEP. Static load–deflection analysis
is specified by *STATIC command. Following is an example analysis step
including analysis type, applied load, and output requests:
*STEP, PERTURBATION
Load–deflection analysis for a concentrated load of –96 kN
*STATIC
*CLOAD
102, 2, –96.0
*NODE PRINT
U
RF
*EL PRINT
S
*END STEP
2.11 PRACTICE PROBLEMS – STRESS ANALYSIS OF A CABLE WITH
FINITE ELEMENT ANALYSIS SOFTWARE ABAQUS
PROBLEM-1
Figure 2.18 shows a simplified two-element model of a bridge suspension cable.
Cable is assumed to be solid circular section of diameter 76.2 mm (cross-sectional
area = 4560.4 mm2). Cable material is steel with properties: E = 210 GPa and
ν=0.3. Distributed load from the bridge is represented by a “lumped” load of 96
44 Finite Element Analysis of Solids and Structures
kN at node #102. Input data for ABAQUS analysis are presented in the
following:
*HEADING
Cable analysis with two truss elements
*NODE, NSET=setn1
101,0.0,0.0,0.0
102,60000,-12000
103,120000,0.0,0.0
*ELEMENT,TYPE=T2D2,ELSET=cable-1
1001,101,102
1002,102,103
*SOLID SECTION, MATERIAL=mat-1, ELSET=cable-1
4560.4
*MATERIAL, NAME=mat-1
*ELASTIC
210,0.3
*BOUNDARY
101,1,2,0.0
103,1,2,0.0
*STEP, PERTURBATION
96 kN vertical downward load at node 2
*STATIC
*CLOAD
102,2,-96.0
*NODE PRINT
U
RF
*EL PRINT
S
*END STEP
Problem to solve:
1. Type in the above input data in a text data file that can be named anything with
an extension .inp (for example, cable.inp)
2. Run the model in ABAQUS and check the results in text output file (cable.dat)
3. Analyze the two-element truss assembly model by hand
a. What is the stress in the cable? (hint: find member forces from Σforces=0 at
node #102)
b. Calculate the vertical deflection at node #102. Hint: external work
(1/2*P*u102) = total internal energy Σ(1/2*σ*ε*vol)
c. How do hand calculation results compare with the ABAQUS results for
two-element assembly model?
Strain–Displacement Relationship and Elasticity of Materials 45
101 103
12000 mm
102
Downward load: 0.0016 kN/mm
60000 mm
60000 mm
Simplified model
of suspension
cable
y
x
101 103
12000 mm
102
60000 mm 60000 mm
P2=128kN
FIGURE 2.19 Simplified three-node cable element model of suspension cable.
PROBLEM-2
Figure 2.19 shows a three-node truss element model of suspension cable. Concentrated
load at node #102 is calculated from distributed load by using the parabolic influence
function H2 from Figure 2.15.
P102 1 r 2 0.0016 dx
1
L
1 r 0.0016 2 dr
1
2
1
L
2
0.0016 1 r dr
1
2
120000 4
0.0016
2 3
128 kN
Modify the ABAQUS input file of Problem-1 to represent the three-node truss ele-
ment model. Check the ABAQUS results of three-node truss model and compare that
with those of Problem-1.
3 Analysis of Solids
Represented by 2D
Stress Fields
SUMMARY
Stress/deformation response of three-dimensional solids, in certain special cases, can
be predicted with the analysis of simplified two-dimensional models. Theory of elas-
ticity-based description of plane-strain and plane-stress problems falls in this cate-
gory. Stress–strain relationships, specific to plane-strain state, are derived in Section
3.1. Additional essential conditions of elasticity problem solutions, compatibility
condition, stress equilibrium equation, and stress boundary conditions, are also
reviewed in this section. Section 3.2 presents key derivations for the special case of
plane-stress elasticity problem. The fundamental elasticity formulations, discussed
in Sections 3.1 and 3.2 for 2D planar solids, are combined in Section 3.3 to define a
bi-harmonic compatibility condition that must be satisfied by any stress function
solution of the 2D elasticity problems. Section 3.4 presents detailed analytical devel-
opments for the stress function-based solution of bending and shear deformation
responses of a cantilever beam. Finding the desirable polynomial function for an
engineering analysis problem, however, takes systematic execution of lengthy ana-
lytical process. The method has not seen numerical implementations for matrix
method analysis of large structural systems. Although not an analytical ingredient,
for the standard displacement-based finite element formulation technique, the stress
function solution, however, provides a direct insight into the deformation behavior of
elementary solids, particularly for the shear deformation behavior that is often diffi-
cult to simulate well with piecewise linearized displacement response of finite ele-
ment models. The stress function-based solutions, presented for simple solid
mechanics case studies, serve as important references for assessing the deformation
behavior of solid and beam finite element formulations.
The stiffness formulation, for finite element simulation of general 2D solids, is
introduced in Section 3.5 where element orientation is assumed parallel to the
Cartesian x–y coordinate system. A more general-purpose formulation, based on iso-
parametric element definition, is introduced in Section 3.6 – followed by a discussion
on numerical integration of element property matrices in Section 3.7. Higher order
formulations, quadrilateral elements with mid-side nodes, are introduced in Section
3.8, which provide improved in-plane bending response compared to the simpler
four-node elements. Relative risk-benefits of higher order versus lower order incom-
patible elements are also discussed in this section. Triangular elements are often
used, in conjunction with four-node plane solid elements, in geometric discretization
of curved boundaries. Numerical details for representing the constant stress–strain
47
48 Finite Element Analysis of Solids and Structures
response of triangular elements, created by collapsing one side of four-node quadri-
lateral solid elements, are discussed in Section 3.9. Convergence characteristics of
displacement and stress responses, predicted by finite element analysis technique,
are discussed in Section 3.10 – with simulation results of an example case study.
Section 3.11 presents ABAQUS-specific element types that are available for
building a simulation model for plane-stress or plane-strain conditions. Commonly
used options to specify externally applied loads are described in Section 3.12. The
primary focus of this chapter is to learn about the details of element formulations and
their effects on predicted results. The study of convergence characteristics of finite
element model, however, requires preparation of simulation models with large num-
ber of nodes and elements. Section 3.13 discusses about the modeling of a simple
plane-stress square plate problem by using the general-purpose finite element model
pre-processor: HyperMesh (Altair University 2020). More discussions on software-
aided model pre-processing work will follow in Chapter 4. Finally, practice problems
on 2D stress analysis of solids are presented in Section 3.14.
3.1 PLANE STRAIN – A SPECIAL FORM OF ELASTICITY PROBLEM
A long prismatic solid, with constant geometric and material properties along the mem-
ber axis, subjected to a uniform lateral loading along the length, is shown in Figure 3.1.
Member is held between two fixed-end conditions resulting in no deformation in the
axis direction (εz = γyz = γzx = 0). End constraints, however, will lead to nonzero axial
stress (σz ≠ 0) on the cross-sectional plane. State of material deformation in the body
can be described in a two-dimensional cross-sectional plane as shown in Figure 3.1 –
with no variation occurring in axis direction. Strain–deformation relationships for this
plane-strain condition are defined in the following equations (3.1 and 3.2):
u v u v
x , y , xy yx (3.1)
x y y x
x
x Out-of-plane strains are ‘zero’
z
FIGURE 3.1 Plane-strain condition with material deformation occurring in 2D cross-
sectional plane (no variation along the member axis).
Analysis of Solids Represented by 2D Stress Fields 49
w v w u w (3.2)
z 0, yz zy 0, xz zx 0
z z y z x
Three expressions for plane-strain condition (equation 3.1) are functions of only
two in-plane displacement components (u, v). Compatibility condition among the
strain components is represented by equation (2.8) alone. Now from the Hooke’s law
(equations 2.18), strain component in member axis direction gives the following rela-
tionship expressing the out-of-plane normal stress as a function of in-plane normal
stresses:
z z x y 0 z x y (3.3)
E E E
Combining equations (2.18) and (3.3), two in-plane normal strain components are
given by the following equations:
x 1 2 E E
x 2 y
(3.4)
y
y 1 2
E
2 x
E
Shear stress–strain relationship for plane-strain condition is given by the following :
xy 2 1 (3.5)
xy xy
G E
Inverting equations (3.4) and (3.5), stress–strain relationships for plane-strain
condition can be written as follows:
x 1 0 x
E (3.6)
y 1 0 y
1 1 2 1 2
xy
xy 0 0
2
Material property matrix on the right-hand side of equation (3.6) represents the
elasticity matrix [C] to be used in finite element stiffness calculations (equation
2.44). Stress equilibrium equations for plane-strain state can be obtained from equa-
tions (1.10) by considering the stress components in x,y plane only:
x xy
Fx 0
x y
(3.7)
y xy
Fy 0
y x
50 Finite Element Analysis of Solids and Structures
Using the stress–strain relationships (3.4 and 3.5), and equilibrium equations (3.7),
compatibility equation (2.8) can be written in terms of stresses as in the following:
2 2 1 Fx Fy
2 2 x y (3.8)
x y 1 x y
Stress distribution function in a plane-strain condition must satisfy the compati-
bility condition (3.8) and the stress boundary conditions for a given problem. Three-
dimensional stress boundary conditions of equations (1.14) can be simplified to
describe the two-dimensional boundary conditions as in the following:
px xl xy m (3.9)
py xyl y m
3.2 PLANE STRESS – A SPECIAL FORM OF ELASTICITY PROBLEM
Figure 3.2 shows a three-dimensional fabric form supported by a mast and cables.
Fabric material provides resistance at any point only within the plane of differential
element surrounding that point (as shown in the figure). Stress in the out-of-plane
direction at any point is assumed zero. Strain–displacement relationships for this
plane-stress condition are similar to those of equations (3.1) and (3.2) except the out
of plane normal strain, εz, which is not zero:
w
z 0 (3.10)
z
Stress–strain relationships for plane-stress state are given by the following
(equation 3.11):
y
σy
τxy
σx
x
Out-of-plane stresses are ‘zero’
σz = τyz = τzx = 0
FIGURE 3.2 Cable-supported fabric – plane-stress condition in 3D space with material pro-
viding resistance to in-plane deformation only.
Analysis of Solids Represented by 2D Stress Fields 51
x
x v y
E E
(3.11)
y y x
E E
z x y x y
E 1
xy 2 1
xy xy
G E
Inverting equations (3.11), stress–strain relationships for plane-stress state are
given by the following:
x 1 0
x
E
y 1 0
y (3.12)
1 2
xy
xy 1
0 0
2
Material property matrix on the right-hand side of equation (3.12) represents the
elasticity matrix [C] that can be used in finite element stiffness calculations (equa-
tion 2.44). Following the same steps of plane-strain problem description, compatibil-
ity equation for plane-stress problem can be written as follows:
2 2 Fx Fy
2 2 x y 1 (3.13)
x y x y
Stress distribution function in a plane-stress condition must satisfy the compati-
bility condition (3.13) and the stress boundary conditions similar to those given by
equations (3.9).
3.3 STRESS FUNCTIONS FOR 2D PLANE STRAIN AND PLANE
STRESS ELASTICITY PROBLEMS
When body forces are negligible, Fx = Fy ≈ 0, stress equilibrium equations (3.7),
and compatibility equations (3.8) and (3.13) for plane-strain and plane-stress condi-
tions, take the following forms:
52 Finite Element Analysis of Solids and Structures
x xy
0
x y
y xy
0 (3.14)
y x
2 2
2 2 x y 0
x y
Above essential conditions for 2D stress-field solutions are readily satisfied when
a stress function ϕ(x, y) is defined such that stress components are defined by the fol-
lowing derivative forms:
2 2 2 (3.15)
x , y , xy
y 2 x 2 x y
Substituting the stress component definitions from equation (3.15) into compati-
bility equation (3.13) yields:
4 4 4 (3.16)
2 2 2 4 4 0
x 4
x y y
In absence of body force effects, solution of a 2D plane-stress or plane-strain
problem involves the solution of bi-harmonic equation (3.16) subject to boundary
conditions of the problem given by equations (3.9). Despite the simplification of 3D
problem to 2D condition, direct solution of equation (3.16) is still a challenging task.
Common practice is to use indirect method (also known as inverse or semi-inverse
method). In that approach, based on examination of body configuration and exter-
nally applied loads and constraints, a stress function ϕ is defined by a polynomial of
known and unknown coefficients. Unknown coefficients are later determined by sat-
isfying the stress boundary conditions (3.9). For example, a second-degree polyno-
mial (equation 3.17) will clearly satisfy the bi-harmonic compatibility equation
(3.16), and it will also provide non-zero stress values (equation 3.15):
ax 2 bxy cy 2 . (3.17)
A higher order polynomial can also be used provided that compatibility condition
∇4ϕ = 0 is satisfied with the adjustment of non-zero coefficients. Few elementary
stress functions are presented in the following.
The second-degree polynomial of equation (3.18) satisfies the compatibility con-
dition (equation 3.16). Associated stress components, defined by relations (3.15), are
shown in equation (3.19):
a2 2 c
2 x b2 xy 2 y 2 (3.18)
2 2
Analysis of Solids Represented by 2D Stress Fields 53
x c2 , y a2 , xy b2 (3.19)
Figure 3.3 shows graphically the constant stress components (σx, σy and τxy) repre-
sented by function (3.18). One or more of the polynomial coefficients can be adjusted
to represent special stress states of uniaxial (only c2 or a2 is non-zero), biaxial (both
c2 and a2 are non-zero; b2 = 0), or pure shear condition (only b2 is non-zero).
The third-degree polynomial of equation (3.20) satisfies the compatibility condi-
tion (equation 3.16); the associated stress components are given by equation (3.21):
a3 3 b3 2 c d
3 x x y 3 xy 2 3 y3 (3.20)
6 2 2 6
x c3 x d3 y, y a3 x b3 y, xy b3 x c3 y (3.21)
For the special case of a3 = b3 = c3 = 0, stress definitions of equation (3.21) reduce
to the following form (equation 3.22) – representing the special case of in-plane pure
bending as shown with the example of a differential element in Figure 3.4.
x d3 y, y xy 0 (3.22)
The fourth-degree polynomial of equation (3.23) satisfies the compatibility condi-
tion (3.16) when e4 = –(2c4 + a4). Stress components associated with this stress func-
tion are defined by equations (3.24):
y
σY=a2
τxy=b2
σx=c2
FIGURE 3.3 Stress field described by second-degree polynomial function
a2 2 c
2 x b2 xy 2 y 2
2 2
σx=–d3y
FIGURE 3.4 Stress field of pure bending case described by a third-degree polynomial func-
d
tion 3 6 y .
3 3
54 Finite Element Analysis of Solids and Structures
a4 4 b4 3 c d e (3.23)
4 x x y 4 x 2 y 2 4 xy3 4 y 4
12 6 2 6 12
x c4 x 2 d4 xy 2c4 a4 y 2
y a4 x 2 b4 xy c4 y 2 (3.24)
b d
xy 4 x 2 2c4 xy 4 y 2
2 2
The fifth-degree polynomial of equation (3.25) satisfies the compatibility condi-
tion (3.16) when (3a5 + 2c5 + e5).x + (b5 + 2d5 + 3f5).y = 0. Since this condition needs
to be satisfied for any coordinate position (x, y), polynomial coefficients are further
reduced by e5 = –(3a5 + 2c5) and b5 = –(2d5 + 3f5).
a5 5 b5 4 c d e f (3.25)
5 x x y 5 x 3 y 2 5 x 2 y3 5 xy 4 5 y 5
20 12 6 6 12 20
Stress components corresponding to the stress function of equation (3.25) are
defined by the following equations (3.26):
c5 3
x x d5 x 2 y 3a5 2c5 xy 2 f5 y3
3
d
y a5 x 3 3 f5 2d5 x 2 y c5 xy 2 5 y3m (3.26)
3
1 1
xy 3 f5 2d5 x 3 c5 x 2 y d5 xy 2 3d5 2c5 y3
3 3
Polynomial stress functions described above can be combined to define a stress
field that is relatively more complex in nature. The following example demonstrates
the use of analytical stress functions to solve problems of elementary solid mechan-
ics. It can be mentioned here that stress function method is not directly implemented
in the standard finite element analysis software products that are coded based on
displacement field assumptions. Nonetheless, an example problem analysis, pre-
sented in Section 3.4, highlights the capability of stress function-based analysis tech-
nique, and it also highlights certain aspects of stress-deformation response that must
be respected by the finite element simulation models.
3.4 EXAMPLE USE OF STRESS FUNCTION APPROACH TO SOLVE
ELASTICITY PROBLEMS
A thin cantilever beam, thickness t << beam depth 2h, is subjected to a shear load P
at the free end (Figure 3.5). Magnitude of load P is considered large enough to make
the effects of body weight negligible (Fx=Fy ≈ 0 in equations 3.7). Stress state in this
thin body is of plane-stress type (σz=τyz=τxz=0). The top and bottom edges of the
Analysis of Solids Represented by 2D Stress Fields 55
σx τxy
P
x
2h
(a) (b)
FIGURE 3.5 (a) A cantilever beam in x–y plane (with thickness = t in the z-direction) and
subjected to a shear load P at the free end; and (b) internal beam stresses as given by elemen-
tary solid mechanics principles.
beam are stress free. The left end of cantilever beam is free of normal stresses, and
the resultant of shear stresses on that edge must be equal to the externally applied
force P. Equations (3.27) summarize the stress boundary conditions on the member.
xy y h 0, y y h 0, x x 0 0
h (3.27)
t
xy dy P
h
Given that the bending moment on any section in the beam varies linearly with
coordinate distance x, and normal stress at any point on the section varies linearly
with y coordinate, the normal stress σx, which is a second derivative of yet-to-be-
defined stress function ϕ in equation (3.15), can be expressed by the following equa-
tion (3.28), where c1 is a constant:
2 (3.28)
x c1 xy
y 2
Double integration of equation (3.28) with respect to y gives the following general
expression for the stress function ϕ where f1(x) and f2(x) are functions of x to be
determined:
1 (3.29)
c1 xy3 yf1 x f2 x
6
Substitution of stress function ϕ in compatibility condition (3.16) gives the fol-
lowing expression:
4 f1 4 f2
y 0 (3.30)
x 4 x 4
56 Finite Element Analysis of Solids and Structures
Since the condition of equation (3.30) must hold for any value of y, two fourth-order
derivative terms must be independently “zero”, i.e.
4 f1 4 f2 (3.31)
0 and 0
x 4 x 4
Integration of the expressions in equation (3.31) four times with respect to x leads
to the following, where c2, c3, etc., are constants of integration:
f1 x c2 x 3 c3 x 2 c4 x c5
(3.32)
f2 x c6 x 3 c7 x 2 c8 x c9
Inserting the function definitions of equations (3.32) into equation (3.29), the gen-
eral definition of the stress functions is obtained as follows:
1
6
c1 xy3 c2 x 3 c3 x 2 c4 x c5 y c6 x 3 c7 x 2 c8 x c9
(3.33)
Stress components corresponding to the stress function of equation (3.33) follow
from equation (3.15):
2
y 6 c2 y c6 x 2 c3 y c7 (3.34)
x 2
2 1 (3.35)
xy c1 y 2 3c2 x 2 2c3 x c4
x y 2
Using the stress boundary condition equations (3.27) into equations (3.34) and
(3.35), the following values of the polynomial constants are obtained, where
I = 2/3(th3) is the moment of inertia of rectangular beam cross-section about the neu-
tral axis:
c2 c3 c6 c7 0
1 (3.36)
c4 c1h2
2
3P P
c1
2th 3
I
Using the polynomial coefficient values in equations (3.28), (3.34), and (3.35), the
following expressions are obtained for the stress components inside the cantilever
beam:
Analysis of Solids Represented by 2D Stress Fields 57
x
Pxy
I
, y 0, xy
P 2
2I
h y2
(3.37)
Stress responses given by equations (3.37) are exactly same as the ones given by
beam bending theory in solid mechanics. Displacement response of the beam can
now be calculated by using the strain–displacement and stress–strain relationships:
u x Pxy v Pxy (3.38)
x , y x
x E EI y E EI
u v xy
xy
y x G
P
2GI
h2 y 2
(3.39)
Integrating equations (3.38), we get the following expressions (equations 3.40)
for displacements u and v, where f3(y) and f4(x) are unknown functions of y and x,
respectively.
Px 2 y Pxy 2
u f3 y v f4 x (3.40)
2 EI 2 EI
Now substituting the expressions for u and v from equation (3.40) into equa-
tion (3.39), and after rearranging the terms we get the following relationship:
Px 2 f4 Py 2 Py 2 f3 Ph2
(3.41)
2 EI x 2 EI 2GI y 2GI
The first part inside the parenthesis on the left side of equation (3.41) is a function
of x, the second part is a function of y, and the right-hand side is a constant for a given
prismatic beam element. In order for the equality condition to hold true for any coor-
dinate position (x, y), two parts inside the parenthesis on the left-hand side must be
separately constant leading to the following relationships (where d and e are
constants):
Px 2 f4
a constant d ,
2 EI x (3.42)
Py 2 Py 2 f3
and a constant e
2 EI 2GI y
Constants d and e, however, are related to each other by equation (3.43):
58 Finite Element Analysis of Solids and Structures
Ph2 (3.43)
d e
2GI
Re-arrangement of the terms in equation (3.42) leads to the following:
f4 Px 2 f3 Py 2 Py 2 (3.44)
d, and e
x 2 EI y 2 EI 2GI
Integration of both sides of equations (3.44) gives the following expressions for
unknown functions f3 and f4 where p and q are integration constants:
Py3 Py3 Px 3
f3 y ey p, f4 x dx q (3.45)
6 EI 6GI 6 EI
Combining equations (3.45) and (3.40), we get the following equations for dis-
placement response of the beam element:
Px 2 y Py3 Py3 (3.46)
u ey p
2 EI 6 EI 6GI
Pxy 2 Px 3 (3.47)
v dx q
2 EI 6 EI
Constants d, e, p, and q can be determined by using the boundary constraints and
the relationship (3.43). The lateral deflection equation for the beam axis can be
obtained from equation (3.47) by substituting y = 0:
Px 3 (3.48)
vy 0 dx q
6 EI
Now taking the derivative of expression (3.48) with respect to x, and setting it to
zero at x = L (“zero” rotation condition at the fixed end of beam), we get the follow-
ing value for constant d:
PL2
d (3.49)
2 EI
Combining equations (3.43) and (3.49), we get the following value for constant e:
PL2 Ph2
e (3.50)
2 EI 2GI
Inserting zero lateral displacement boundary condition at the fixed end, v = 0 at x
= L, in equation (3.47), and using the expression for constant e from equation (3.50),
we get the following value for constant q:
Analysis of Solids Represented by 2D Stress Fields 59
PL3 (3.51)
q=
3EI
Inserting zero axial deformation boundary constraint at the fixed end, u = 0 at x =
L and y = 0 in equation (3.46) yields p = 0. Finally, inserting the values of constants
d, e, p, and q in equations (3.46) and (3.47), we get the following equations for dis-
placement response of the cantilever beam when subjected to a lateral load P at the
free end:
Px 2 y Py3 Py3 PL2 Ph2 (3.52)
u y
2 EI 6 EI 6GI 2 EI 2GI
Pxy 2 Px 3 PL2 PL3
v x (3.53)
2 EI 6 EI 2 EI 3EI
And the expression for deflection of the beam axis, equation (3.48), takes the fol-
lowing form:
Px 3 PL2 PL3
vy 0 x (3.54)
6 EI 2 EI 3EI
From equation (3.54), deflection of the cantilever beam at the free end (x=0)
comes out to be PL3/3EI which is exactly equal to the value we get from beam theory
analysis in solid mechanics. Stress function method, however, provides additional
insights into the deformation response of the cantilever beam. We can take the deriva-
tive of displacement function u in equation (3.46) with respect to y, and set it to zero
at x = L (zero rotation at the fixed end of the beam), thus giving the following expres-
sion for constants (Timoshenko and Goodier 1982):
PL2 PL2 Ph2
e ,d (3.55)
2 EI 2 EI 2GI
Substituting these in equation (3.47), we get the following expression for deflec-
tion of the beam axis:
Px 3 PL2 PL3 Ph2
vy 0
6 EI 2 EI
x
3EI 2GI
L x (3.56)
Compared to equation (3.54), equation (3.56) gives an additional beam axis
deflection term of Ph2(L – x)/2GI which is attributed to the shear deformation of the
solid (not accounted for in the Euler Bernoulli beam theory analysis).
The cantilever beam example, analyzed by using the stress function-based analy-
sis technique, has simple geometry, loading, and boundary conditions. Two more
plane-stress examples of similar degree of analysis complexity are shown in
Figure 3.6; solutions for these problems are available in Timoshenko and Goodier
60 Finite Element Analysis of Solids and Structures
y y w
2h x 2h x
L L L
(a) (b)
FIGURE 3.6 (a) Thin beam subjected to a uniformly distributed load w per unit length; (b)
thin beam subjected to a linearly varying load over the member length.
(1982). As evident from the detailed analysis presentation of the cantilever beam
example, finding the desirable polynomial function for an engineering analysis
problem takes systematic execution of lengthy analytical process. Moreover, the
method has not seen numerical implementations for matrix method analysis of large
structural systems. The stress function method, however, provides insightful results
about the deformation behavior of elementary solids, particularly for the shear
deformation behavior that is often difficult to simulate well with finite element simu-
lation models that rely on linear displacement field assumptions within finite ele-
ment domains. The stress function-based analysis results, derived in the above, will
be used as reference for quality assessment of displacement-based finite element
simulation models. The stiffness formulation, for finite element simulation models
of general 2D solids, is introduced in Section 3.5 where element orientation is
assumed parallel to the Cartesian x–y coordinate system. A more general-purpose
formulation, based on iso-parametric element definition, will be introduced in
Section 3.6 – followed by a discussion on numerical integration of element property
matrices in Section 3.7.
3.5 STIFFNESS METHOD ANALYSIS OF SOLIDS REPRESENTED
BY 2D STRESS-DEFORMATION FIELDS
Figure 3.7(a) shows a solid element in 2D (x–y) plane connected to four corner nodes
(n1...n4) with each node having two in-plane translational DOF (u,v). Figure 3.7(b)
shows assumed displacement variation over the element domain for unit magnitude
at node n1 and zero values at the other three nodes. Analytical function of that dis-
placement field, often referred to as shape function or interpolation function, is
defined by equation (3.57):
Analysis of Solids Represented by 2D Stress Fields 61
v2
y
v1
u2 u1
n2 n1 magnitude=1
2b v x
3
v4
n3 n4 u4
u3
2a
(a) (b)
FIGURE 3.7 (a) A solid element in 2D x–y plane, connected to four corner nodes (n1...n4),
with each node having 2 displacement DOF (ui, vi); (b) interpolation function over the 2D
plane for a unit magnitude at node n1.
1 x y
H1 1 1 (3.57)
4 a b
Similarly, shape functions for other nodes can be defined as follows:
1 x y
H2 1 1
4 a b
1 x y
H3 1 1 (3.58)
4 a b
1 x y
H4 1 1
4 a b
Displacement responses at any point (x,y) inside the element can be related to the
nodal displacements (ui,vi), with the use of interpolation functions (3.57 and 3.58), as
follows:
u x,y H1u1 H 2u2 H3u3 H 4u4
(3.59)
v x,y H1v1 H 2 v2 H3v3 H 4 v4
Using the above displacement functions in the strain–displacement relationships of
equations (2.4), we get the following expressions for in-plane strains in the 2D solid:
62 Finite Element Analysis of Solids and Structures
u 1 y 1 y
x 1 .u1 1 .u2
x 4a b 4a b
1 y 1 y
1 .uu3 1 .u4
4a b 4a b
v 1 x 1 x
y 1 .v1 1 .v2
y 4b a 4b a
1 x 1 x
1 .v3 1 .v4
4b a 4b a
(3.60)
u v 1 x 1 x
xy 1 .u1 1 .u2
y x 4b a 4b a
1 x 1 x
1 .u3 1 .u4
4b a 4b a
1 y 1 y
1 .v1 1 .v2
4a b 4a b
1 y 1 y
1 .v3 1 .v4
4a b 4a b
Compatibility condition in equation (2.8) will be satisfied by the above strain
functions – a basic requirement of finite element shape function definitions.
Re-writing the equations (3.60) in vector and matrix forms, we get the following
equation (3.61), where matrix [B] is the strain–displacement relationship matrix
defined by equation (3.62):
u1
u2
u3
x
u4 (3.61)
B
y
v1
xy v2
v3
v
4
b y b y b y b y 0 0 0 0
1
B 0
4ab
0 0 0 a x a x a x a x (3.62)
a x a x a x x
a b y b y b y b y
Analysis of Solids Represented by 2D Stress Fields 63
In addition to the strain–displacement relationship matrix, [B], equation (2.44) for
element stiffness calculation also requires the material stress–strain relationship
matrix [C]. From equations (3.6) and (3.12), [C] matrix can be defined by equations
(3.63) and (3.64), respectively, for plane-strain and plane-stress conditions in a solid:
1 0
E (3.63)
C plane strain 1 0
1 1 2 1 2
0 0
2
1 0
E
C plane stress 1 0 (3.64)
1 2
1
0 0
2
Assuming that material properties are constant over an element domain, matrix
[C] defined by equation (3.63) or (3.64), as applicable, is a constant in the right side
integral of equation (2.44). Stiffness property of the element, thus becomes, a sec-
ond-order polynomial function based on the definition of [B] matrix given by equa-
tion (3.62). Integral of stiffness function can be determined analytically for the
simple rectangular solid element that aligns with the x–y coordinate directions. Rest
of the analysis method will follow the same general steps outlined in Section 2.7 for
stiffness-based matrix method analysis of structures.
Unlike the simple rectangular finite element definition of Figure 3.7, not all finite
elements in a general structural analysis model will be aligned perfectly with the
user-defined global coordinate system. Strain–displacement relationship functions
(equations 3.60), therefore, need to be modified to represent an arbitrary orientation
of the element in the user-defined x–y coordinate plane. Sections 3.6 and 3.7 present
the general formulations for a solid element having arbitrary orientation in the 2D
global reference coordinate system.
3.6 ISO-PARAMETRIC DEFINITION OF 2D SOLID
FINITE ELEMENTS
Figure 3.8(a) shows a solid element having arbitrary orientation with respect to
the global (x,y) coordinate system in 2D plane. Nodal coordinates and displace-
ment DOF are defined in global directions, while a local coordinate system (r, s)
is introduced bounding the element domain in the ranges of r = –1 to r = +1, and
64 Finite Element Analysis of Solids and Structures
dy
dx
y s
v1
1 (x1,y1) magnitude=1
v2 s=+1 u1
r
(x2,y2) 2 r=+1
u2 v4
r=–1 v3 u4
u3 s=–1 4 (x4,y4)
3 (x3,y3)
x
(a) (b)
FIGURE 3.8 Iso-parametric formulation for 2D solid element: (a) global coordinates (xi,yi),
and local coordinate system (r,s); (b) element shape function for node (1).
s = –1 to s = +1. Figure 3.8(b) shows the element shape function for unit magni-
tude at node #1. Similar shape functions for the other three connecting nodes of
the element can also be drawn (not shown graphically). Analytical definitions of
shape functions (Hi, i =1…4), in terms of local (r, s) coordinates, are given by
equations (3.65):
1
Hi
4
1 r 1 s , where
1
H1 1 r 1 s
4
1
H 2 1 r 1 s (3.65)
4
1
H3 1 r 1 s
4
1
H 4 1 r 1 s
4
Interpolating the nodal coordinates, global coordinates of a point inside the ele-
ment can be defined by the following functions:
x H1 x1 H 2 x2 H3 x3 H 4 x4
(3.66)
y H1 y1 H 2 y2 H3 y3 H 4 y4
Taking partial derivates of x and y with respect to r and s:
Analysis of Solids Represented by 2D Stress Fields 65
x H x H
i . xi ; i . xi
r r s s (3.67)
y Hi y Hi
. yi ; . yi i 1 4
r r s s
Using the shape function definitions from equations (3.65), partial derivative
terms in equations (3.67) are found to be:
Hi 1 Hi 1
1 s ; 1 r (3.68)
r 4 s 4
Following the rules of partial differentiation:
Hi Hi x Hi y
. .
r x r y r
(3.69)
Hi Hi x Hi y
. .
s x s y s
Writing the relations (3.69) in vector and matrix forms,
Hi Hi x y
r x r r (3.70)
J , where J
Hi Hi x y
s y s s
Matrix [J] in equation (3.70), known as Jacobian, expresses the relationship between
local element coordinates (r,s) and the global model coordinates (x, y). Terms inside
the Jacobian matrix are given by equations (3.67) and (3.68). Physical interpretation
of Jacobian can be provided by considering a small differential element of dimen-
sions dx*dy inside the finite element in Figure 3.8. Differential terms dx and dy can
be expressed in terms of r and s variables as in the following:
x x y y (3.71)
dx .dr .ds; dy .dr .ds
r s r s
Area of the differential element is defined as follows:
x y x y
dA dx dy . . dr dr . . dr ds
r r r s
x y x y
. . dr ds . . ds ds (3.72)
s r s s
66 Finite Element Analysis of Solids and Structures
Setting the vector product terms (dr*dr) and (ds*ds) to “zero”, equation (3.72)
takes the form:
x y x y
dA . . dr ds . . ds dr (3.73)
r s s r
Substituting ds*dr = –dr*ds in equation (3.73):
x y x y
dA . . . dr ds (3.74)
r s s r
Using the definition of Jacobian from equation (3.70), equation (3.74) can be
re-written as follows:
dA dx dy det J . dr ds (3.75)
A positive definite value of “J” defines a one-to-one relationship between global
(x–y) and local (r–s) coordinate systems. Using the equations (3.65), (3.66) and
(3.67), we get the following relations for the special case of rectangular element in
Figure 3.7(a);
x y x y (3.76)
a, 0, 0, b
r r s s
Jacobian of the rectangular element is thus defined by the following:
a 0
J (3.77)
0 b
Determinant of Jacobian given in equation (3.77) represents a quarter of the 2a x
2b rectangular element of Figure 3.7(a). This positive definite value thereby defines
a one-to-one coordinate transformation relationship in equation (3.75). Ratio between
the two diagonal terms of Jacobian, J11 and J22, defines the aspect ratio of element.
Evidently, for a square shape element, two diagonal terms will have equal values. A
distorted element shape, like the one in Figure 3.9, leads to a variable Jacobian defi-
nition over the element domain (equation 3.78):
y (1,1.25)
(-1,0.25)
(-1,-0.75) (1,-0.75)
FIGURE 3.9 A distorted 2D solid element.
Analysis of Solids Represented by 2D Stress Fields 67
1 4 1 s (3.78)
J
4 0 3 r
For this distorted element example, coordinate transformation relationship of
equation (3.75) becomes a function of (r, s) (because of the location-dependent defi-
nition of J in equation 3.78). In finite element modeling, it is preferable to have
square or rectangular shape elements that have constant Jacobian value within a
given element domain, thus, facilitating a robust calculation procedure for the stiff-
ness properties as discussed in Section 3.7.
3.7 NUMERICAL CALCULATION OF STIFFNESS MATRIX FOR
ISO-PARAMETRIC 2D SOLID FINITE ELEMENTS
Similar to the coordinate interpolation exercise of equation (3.66), displacement
responses inside a 2D solid element (u, v) can be related to the nodal displacements
(ui,vi) with the use of shape functions Hi given in equations (3.65):
4 4
u
i 1
Hiui v H v
i 1
i i (3.79)
Strains corresponding to the displacement responses (u, v) are given by the fol-
lowing expressions:
u H
x i .ui
x x
v H
y i .vi (3.80)
y y
u v H H
xy i .ui i .vi
y x y x
Writing the equations (3.80) with vector and matrix forms, we get the following
strain–displacement relationships:
u1
H1 H 2 H3 H 4 v1
0 0 0 0 u
x x x x x 2
H1 H 2 H3 H 4 v2
y 0 0 0 0 (3.81)
y y y y u3
xy H H1 H 2 H 2 H3 H3 H 4 H 4 v3
1
y x y x y x y x u4
v
4
68 Finite Element Analysis of Solids and Structures
Strain–displacement relationship matrix [B], as used in stiffness matrix calcula-
tion of equation (2.44), is defined from equations (3.81) as follows:
H1 H 2 H3 H 4
0 0 0 0
x x x x
H1 H 2 H3 H 4 (3.82)
B 0 0 0 0
y y y y
H1 H1 H 2 H 2 H3 H3 H 4 H 4
y x y x y x y x
Terms inside the [B] matrix, comprising of x and y derivatives of shape functions
Hi, can be determined by inverting the relationships (3.70) as in the following:
Hi Hi
x 1
r
J (3.83)
Hi Hi
y s
The inverse of Jacobian [J] exists when it is positive definite. As discussed earlier,
for rectangular and square shape elements, [J] is a diagonal matrix of constant values
over an element domain. The right-side terms in equation (3.82), defined by equa-
tions (3.68), are linear functions of r and s only, thus making the [B] matrix in equa-
tion (3.82) a linear function of r and s. Considering the above definition of [B] matrix,
and the constant stress–strain relationship matrix [C], given by equation (3.63) or
(3.64) for a 2D problem, stiffness terms inside the integral of equation (2.44) will be
a second-order function of local element coordinates r and s. These stiffness terms
for a given finite element are generally calculated by using numerical integration
methods.
Figure 3.10 shows numerical integration examples for area calculations under
one-dimensional functions that are defined with single variable x. For linear function
of Figure 3.10(a), area under the line is precisely calculated with single-point calcu-
lation – function value at the domain center (f0) multiplied by the domain length (α0).
For a second-order function in one dimension, area under the curve can be accurately
calculated with judicious selection of two integration points as outlined in Figure
3.10(b). Extending this numerical integration procedure of one-dimensional second-
order function to a two-dimensional domain, volume under a second-order surface
can be calculated by evaluating the function values at 2 × 2 integration points (Figure
3.11(a)). Exact value of volume under the surface can be calculated by using the
Gauss integration rule (Press et al. 2007) that specifies integration points at (±0.57735,
±0.57735); and weight factor values as αij=1.0 (Figure 3.11(b)). The second-order
variation of stiffness function for a four-node iso-parametric finite element, as
Analysis of Solids Represented by 2D Stress Fields 69
(a) 1storder varia on (straight line) (b) 2ndorder varia on (curve)
f0 f1 f2
Area under the line Area under the curve
FIGURE 3.10 Numerical integration: (a) one-point integration for linear variation; (b) two-
point integration for second-order variation.
(a) 2nd order variaon (surface) (b) Gauss integraon parameters
z=f (x2,y2,xy,x,y,..) (αij = 1.0)
s +0.57735, +0.57735
s r –0.57735,
x +0.57735 s=+1
x r
x x r=+1
x
x x
r=–1 x
s=-1 +0.57735, –0.57735
Volume under the surface
–0.57735, –0.57735
FIGURE 3.11 (a) Numerical integration of volume under a second-order surface over (x, y)
plane, and (b) corresponding Gauss integration parameters.
discussed earlier, can be schematically compared with the second-order general
function over the two-dimensional plane of Figure 3.11 (a). Equation (3.84) shows
the numerical expressions for the stiffness matrix terms of four-node quadrilateral
element, where fij are the stiffness function values at Gauss quadrature points (i = 1,2
and j = 1,2), and αij are weight factors (= 1.0):
70 Finite Element Analysis of Solids and Structures
ij fij .. .. .. .. .. .. .. ..
.. .. .. .. .. .. .. ..
.. .. .. .. .. .. .. ..
.. .. .. .. .. .. .. ..
k B . C . B .dV
T
(3.84)
.. .. .. .. .. .. .. ..
.. .. .. ... .. .. .. ..
.. .. .. .. .. .. .. ..
.. .. .. .. .. .. .. ..
2 × 2 Gauss integration rule provides exact calculation of stiffness function in equa-
tion (3.84) for four-node square or rectangular shape element where Jacobian J
remains constant over the element domain. Jacobian of a distorted element (example
in Figure 3.9) becomes a function of element local coordinates (r, s) (equation 3.78),
thus, introducing higher order terms in the strain–displacement relationship matrix
defined by equations (3.82) and (3.83). Stiffness matrix terms in equation (3.84),
thus, also contain higher order terms. 2 × 2 Gauss quadrature rule, used in many
standard finite element software products for stiffness calculation of four-node quad-
rilateral elements (equation 3.84), provides accurate values for second-order func-
tions of stiffness variation. Higher order stiffness variation terms for distorted
elements will be approximately evaluated by the default 2 × 2 Gauss integration rule,
thus leading to some calculation errors. It is therefore important to keep four-node
element shapes close to square or rectangular for accuracy of calculations that use
default 2 × 2 Gauss integration rule. Figure 3.12 demonstrates the negative effect of
element shape distortion on predicted stress response of an elementary example.
An element is called “fully integrated” when the stiffness function is accurately/
fully integrated with the appropriate order of integration over the element domain.
For computational speed, finite element software packages may adopt fewer number
MODEL–2
40 10
MODEL–1 1 kN 1 kN
El–1
10 El–1 El–2 10
El–2
1 kN 1 kN
25 25 10 40
Nodal Stresses from FEA Model
–0.0885
Es mated stress: 0.2 0.1882 Big degrada on in
stress predic on !
0.0478
FIGURE 3.12 Negative effects of distorted element shape (MODEL-2) on predicted stress
response.
Analysis of Solids Represented by 2D Stress Fields 71
In-plane bending
FIGURE 3.13 Single-point reduced integration (x) of four-node quadrilateral element –
producing zero resistance to in-plane bending.
of integration points compared to what is required for exact integration of the stiff-
ness function. Such reduced integration of element properties may affect the defor-
mation and resistance response mechanisms of finite elements. Single center-point
calculations for a four-node quadrilateral element (Figure 3.13) capture no strains
and stresses under in-plane bending action. This lack of stress resistance may lead to
an unrealistic zero energy hourglass deformation mode in an assembly of similar
under-integrated finite elements.
Linear shape functions for four-node quadrilateral elements (Figures 3.7 and 3.8)
imply piecewise linear approximations of curved deformation profile. This piece-by-
piece linear approximations of curved deformation profile often make the system
response stiff. An alternative to piecewise linearization of displacement field is to use
higher order finite elements that are capable of producing higher order deformation
response over an element domain (Section 3.8).
3.8 HIGHER ORDER PLANE-STRESS/PLANE-STRAIN ELEMENTS
Figure 3.14 shows an eight-node solid element for analysis of 2D plane-stress or
plane-strain response field. Shape functions for corner node (1), and for mid-side
node (2) are written in terms of local (r, s) coordinate system as follows:
1
H1
4
1 r 1 s r s 1
(3.85)
1
H 2 1 r 2 1 s
2
Similar third-degree shape functions for other nodal DOF of the element can be
derived by considering one DOF at a time. These higher order shape functions can be
used to calculate the element stiffness matrix by following the same steps presented
in Section 3.7 for four-node quadrilateral element. Third-degree polynomial equa-
tions for eight-node element shape functions, however, will lead to second-degree
functions for strain–displacement relationships in equations (3.82), and to fourth-
degree functions in element stiffness terms of equation (3.84). Accurate calculation
72 Finite Element Analysis of Solids and Structures
s=+1
3 2
1
x x x x x x
s
r=–1 r=+1
r
4 x x x 8 x x x
x x x x x x
5 6 7
s=–1
(a) (b)
FIGURE 3.14 (a) Eight-node element for 2D stress-field analysis – with shape function
shown for one corner node, and (b) shape function for a mid-side node.
of these stiffness terms can be achieved with 3 × 3 numerical integration points (x) as
marked in Figure 3.14. Reduced 2 × 2 integration rule for eight-node elements is
used occasionally for higher computation efficiency. Elements having mid-side
boundary nodes, such as the 8-node element, are called “serendipity” elements.
Addition of an internal node, possibly at the center (r = s = 0) makes the element a
“Lagrangian” quadratic element (Figure 3.15). The name “Lagrangian” is used
because the element shape functions can be obtained by taking the products of one-
dimensional Lagrange interpolants. The element behavior is at its best when the inte-
rior node is located at the element center. The geometry of Lagrangian element is
completely defined by the coordinates of eight boundary nodes in 2D space; the ninth
node (if considered in finite element formulations) is added during element calcula-
tions. The shape function associated with the ninth node of quadratic Lagrangian
element (Figure 3.15) is defined by equation (3.86):
s=+1
3 2
1
x x x
s
r=–1 r=+1
r
4 x x9 x 8
x x x
5 6 7
s=–1
FIGURE 3.15 Nine-node Lagrangian element for 2D stress-field analysis – with shape func-
tion shown for ninth node at the element center.
Analysis of Solids Represented by 2D Stress Fields 73
H9 1 r 2 1 s 2 (3.86)
All iso-parametric elements lose accuracy when distorted from rectangular shape.
The nine-node element is much less sensitive than the eight-node element to non-
rectangularity, to curvature of sides, and to placing of side nodes away from mid-
points of the sides. Compared to the response of four-node quadrilateral elements,
eight- and nine-node elements provide much better simulation of in-plane bending
behavior because of the inherent flexibility introduced by the mid-side nodes. In the
past, special form of four-node element has been developed by using higher order
deformation function in between the corner nodes to replicate the edge bending
behavior (Cook et al. 1989). Formulations of higher order deformation mode, with-
out the use of mid-side nodes, commonly known as in-compatible finite elements,
have been implemented in some finite element software packages. These elements
use higher order polynomial functions to define parabolic deformation modes
between nodes, thus, making the element behavior more flexible compared to that of
a standard four-node element. However, absence of the mid-side nodes provides the
opportunity for in-compatible deformation between adjacent elements (discussed
further in Chapter 4). With the present availability of abundant computing power, it
is better to use eight- or nine-node elements that provide computationally stable
response without the risk of unstable numerical response that can emerge from the
use of lower order incompatible elements.
3.9 CONSTANT STRESS/STRAIN TRIANGULAR ELEMENT
Triangular elements are often used, in conjunction with four-node plane solid ele-
ments, in geometric discretization of curved boundaries. A triangular solid element
can be constructed by collapsing one side of a four-node solid element. Figure 3.16(a)
shows a four-node square element, and Figure 3.16(b) shows a triangular element
created by collapsing the edge 1–2 of the four-node element. Combining equations
y y
1 6 3
2 1 21
x x
s 4 x 5
2 2
r
x x 2
3 4 3 4
2 2
(a) (b) (c)
FIGURE 3.16 (a) Four-node element, and (b) triangular element created by collapsing the
side (1–2) with coincident nodes (1) and (2); and (c) six-node triangular element with three
integration points (x).
74 Finite Element Analysis of Solids and Structures
(3.65) and (3.66), and assigning same coordinate values for coincident nodes 1 and
2, i.e. x1 = x2 and y1 = y2, the following equations are obtained for coordinate defini-
tions at any point inside the triangular element:
1 1 1
x
2
1 s . x2 4 1 r 1 s . x3 4 1 r 1 s . x4
(3.87)
1 1 1
y 1 s . y2 1 r 1 s . y3 1 r 1 s . y4
2 4 4
For the example case in Figure 3.16(b), using the actual nodal coordinates into
equations (3.87), we get the following expressions for coordinates (x,y):
1
x
2
1 r 1 s
(3.88)
y 1 s
Now using these expressions for x and y, the Jacobian definition of equation (3.70)
takes the following form:
x y 1 s
r 0
r 2
J (3.89)
x y 1 r 1
s s 2
Inverting the Jacobian matrix of equation (3.89), we get:
2
1 s 0
1
J (3.90)
1 r 1
1 s
Analogous to the coordinate definitions of equations (3.87), displacement
responses inside the triangular element can be written in terms of the nodal displace-
ments as follows:
1 1 1
u
2
1 s .u2 1 r 1 s .u3 1 r 1 s .u4
4 4
(3.91)
1 1 1
v 1 s .v2 1 r 1 s .v3 1 r 1 s .v4
2 4 4
Analysis of Solids Represented by 2D Stress Fields 75
Taking partial derivatives of u and v with respect to element local variables r and s:
u 1 1
r
4
1 s .u3 4 1 s u4
u 1 1 1
u2 1 r .u3 1 r u4
s 2 4 4
(3.92)
v 1 1
1 s .v3 1 s v4
r 4 4
v 1 1 1
v2 1 r .v3 1 r v4
s 2 4 4
Now using the coordinate transformation rule of equation (3.83),
u2
u u 2 1 s 1 s v2
x 0 0 0 0 0
1
r 1 s 4 4 u3
J . .
u u 1 r 1
1
0
1 r
0
1 r v
0 3
y s 1 s 2 4 4 u4
v4
which eventually leads to,
u2
u 1 1 v2
x 0 0 0 0
2 2 u3
(3.93)
u 1 0
1
0 0 0 3
v
y 2 2 u4
v4
Similarly,
u2
v 1 1 v2
x 0 0 0 0
2 2 u3
(3.94)
v 0 1
0
1
0 0 3
v
y 2 2 u4
v4
76 Finite Element Analysis of Solids and Structures
From equations (3.93) and (3.94), strains inside the triangular element at any
point (x, y) are given by:
u 1 1 u2 u2
0 0 0 0 v
x 2 2
2 v2
v 1 1
u3 u3
y 0 2
0
2
0 0 . B
. (3.95)
v3
v3
u v 1 1 1 1 u4 u4
2 0 0
y x 2 2 2
v
4 v4
Evidently, strain–displacement relationship matrix [B] in equation (3.95) is inde-
pendent of the element internal coordinates (r, s). This means that strain values are
constant throughout the element. Triangular elements therefore represent constant
strain (and constant stress) state internally. Constant strain triangular element in the
above has been derived by degenerating a four-node quadrilateral element to a trian-
gular one. Property matrices of a triangular element can also be derived directly by
considering the three-node geometric configuration. Nonetheless, the stress/strain
will remain constant irrespective of how the element stiffness properties are derived.
Some finite element software packages use explicit triangular element formulations
while some other packages use degenerated formulation from four-node element
description. Either way, actual use of constant-strain elements should be kept to min-
imum when a significant gradient is expected in the stress field (Figure 3.4).
Alternatively, higher order triangular element (six-node element in Figure 3.16(c))
can be used to model stress variations in a field. Further discussion on the selection
of element shape functions and integration order is presented in Chapter 4.
3.10 CONVERGENCE OF FINITE ELEMENT MODEL SOLUTIONS
Convergence of finite element simulation models depends on the quality and accu-
racy of finite element formulations that are used for piecewise discretization of a
solid. Developers of finite element software packages often provide analysis exam-
ples including a commonly used form of convergence test known as “patch test”
(Irons and Ahmad 1980). The basic idea of patch test is to demonstrate that finite
element model solutions converge to the exact theoretical solution as the mesh is
refined. It has been argued that patch test is neither sufficient nor necessary for con-
vergence (Bathe 1996). Nonetheless, the study of finite element mesh refinement
effect on predicted results is important to engineers for deciding what degree of finite
element mesh refinement is sufficient to produce good quality results. Figure 3.17
shows a 50 × 50 mm plate of 1 mm thickness-subjected to in-plane stresses and
boundary conditions. Four finite element mesh models with different degrees of
mesh refinement, and the contours of predicted vertical displacements are shown in
Figure 3.18(a). An important characteristic of finite element displacement solution is
its continuity across element boundaries. As expected from the problem description
given in Figure 3.17, vertical displacement response is zero at the left boundary, and
it increases gradually towards the right with the maximum response occurring at the
Analysis of Solids Represented by 2D Stress Fields 77
Trac on: 1 kN/mm2
y
Fully
constrained
boundary
condi on 50
E=210 GPa
mm
v=0.3
x
50 mm thickness=1 mm
FIGURE 3.17 A 50 mm × 50 mm plate of unit thickness – subjected to plane-stress loading
and boundary conditions.
(#1) 2 x 2 mesh (#2) 12 x 12 mesh (#3) 18 x 18 mesh (#4) 72 x 72 mesh
(a)
Ver cal
displacement
@ upper right
corner
(b) Mesh #
FIGURE 3.18 (a) Smooth contours of vertical displacements (in the y-direction) predicted
by finite element mesh models of four different refinements; (b) convergence of predicted
displacement response at upper right corner.
upper right corner. Figure 3.18(b) plots the maximum displacement responses
obtained from 4 different mesh models. The coarse (2 × 2) mesh model, being the
stiffest, predicts the lowest displacement value. The predicted maximum displace-
ment response, however, does not change with increasing refinement beyond model
#2 (12 × 12). In general, the overall displacement response in finite element analysis
tends to converge quickly with a reasonable degree of mesh refinement. In the prob-
lem presented above, mesh refinement with 12 elements was good enough for piece-
by-piece linearized simulation of displacement variation along the x-direction. In
practical finite element analysis, two models can be prepared initially – one model
78 Finite Element Analysis of Solids and Structures
0.9 1.0 1.0 1.0
(a)
2 x 2 mesh 12 x 12 mesh 18 x 18 mesh 72 x 72 mesh
0.84 3.0 4.3 5.2
(b)
FIGURE 3.19 (a) Smooth contours of stresses in vertical (y) direction predicted by finite
element mesh models of four different refinements; (b) contours of stresses in horizontal (x)
direction – with non-convergence at the bottom left corner
having double the refinement level compared to the other one. If the displacement
response of the refined model is substantially different from that of the course model,
further refinement of the model can be undertaken until a satisfactory state of conver-
gence is achieved.
Figure 3.19(a) shows contours of vertical direction stresses (σyy) obtained from 4
finite element mesh models discussed earlier. Coarse mesh model (2 × 2) provides a
poor prediction of the stress boundary condition on the upper edge. Predicted stress
profiles become smoother as the mesh refinement progresses beyond 12 × 12 level.
More refined finite element models provide better predictions of stress profile in the
body. Similar general conclusion can also be drawn for the contours of horizontal
direction stresses (σxx) shown in Figure 3.19(b). However, the predicted stress value
at the bottom left corner becomes increasingly higher as the mesh refinement becomes
higher – implying that the predicted local stress response, near the geometric discon-
tinuity at the left bottom corner, fails to converge with mesh refinement. This is a
classical problem inherited from theory of elasticity solutions for stress responses at
the points of stress singularity. Further discussions on this issue will follow in Chapter 9.
Stress variations near smooth geometric profiles, however, can be well predicted by
selective refinement of finite element mesh as discussed in Chapter 4.
3.11 SELECTION OF ELEMENT TYPES FOR 2D STRESS ANALYSIS
WITH FEA SOFTWARE PACKAGES
As described in Section 2.10, a critical input in finite element analysis models is the
element type – appropriate for the analysis of a given problem. Finite element
simulation software packages provide coded names to be used by the analysts for
selecting desired element formulation types. For example, four-node plane-stress
element, with 2 × 2 Gauss integration rule, is identified in ABAQUS library as
Analysis of Solids Represented by 2D Stress Fields 79
y Trac on: 1 kN/mm2
(0,50) 7 8 9 (50,50)
3 4
4 5 6
1 2
x
1 2 3
(0,0) (50,0)
FIGURE 3.20 A four-element model of 50 mm × 50 mm plate– subjected to plane-stress
loading and boundary conditions.
“CPS4” – meaning a continuum (“solid”) element in 2D space attached to four cor-
ner nodes. Considering the four-element model of example problem discussed in
Section 3.10 (Figure 3.20), plane-stress elements in ABAQUS input file will look as
follows (where the element corner nodes are listed following a counter-clockwise
sequence (as per the data input syntax for 2D elements in ABAQUS):
*ELEMENT, ELSET=Plate50×50, TYPE=CPS4
1, 1, 2, 5, 4
2, 2, 3, 6, 5
3, 4, 5, 8, 7
4, 5, 6, 9, 8
Coordinates of nodes (1,2,…9), used to describe the elements in the above model,
can be defined by ABAQUS data block under “*NODE” keyword as described in
Section 2.10. Reduced one-point integration rule for four-node quadrilateral element
(Figure 3.13) can be selected by choosing “TYPE=CPS4R” in the above element
description. Element type “CPS4” can be used to create “degenerated” triangular
plane-stress elements by assigning the same node ID number twice in the description
of a quadrilateral element. Example of degenerated triangular element in Figure
3.16(b) can be described in ABAQUS as follows:
*ELEMENT, TYPE=CPS4, ELSET=set2
element ID, 3, 4, 2,2
Above element can also be described by choosing the standard three-node plane-
stress element type “CPS3” from ABAQUS element library:
*ELEMENT, TYPE=CPS3, ELSET=set3
element ID, 3, 4, 2
Fully integrated six-node triangular element of Figure 3.16(c) can be described in
ABAQUS input file, starting with the corner nodes first, followed by the listing of
mid-side nodes as follows:
80 Finite Element Analysis of Solids and Structures
*ELEMENT, TYPE=CPS6, ELSET=set4
element ID, 1,2, 3, 4,5,6
Reduced one-point integrated version of six-node triangular element can be selected
by specifying “TYPE=CPS6R” in the above data block. Higher order eight-node
plane-stress element, using 3 × 3 full Gauss integration rule, is identified as
“TYPE=CPS8” in ABAQUS element library. Element example shown in Figure
3.14(a) can be defined in ABAQUS input file by defining the corner nodes first, fol-
lowed by the identification of mid-side nodes, as in the following:
*ELEMENT, TYPE=CPS8, ELSET=set5
element ID, 1, 3, 5,7,2,4,6,8
Eight-node plane-stress element, using reduced 2 × 2 Gauss integration rule, can be
selected by specifying “TYPE=CPS8R” in the above data description. ABAQUS ele-
ments for modeling plane-strain problems follow the exact same data format
described in the above for plane-stress elements, but with unique element type code
names, such as CPE4 instead of CPS4, CPE3 instead of CPS3, etc.
As discussed earlier, higher order eight- and six-node elements are relatively
more flexible compared to the corresponding lower order four- and three-node ele-
ments of the same geometric shape. Lower order elements obviously facilitate
faster computational efficiency, but with stiffer response prediction. An artificial
improvement to the flexibility of four-node quadrilateral element, as also dis-
cussed earlier, is achieved by using the in-compatible elements. ABAQUS formu-
lation for that element is defined by “TYPE=CPS4I” and “TYPE=CPE4I”,
respectively, for plane-stress and plane-strain conditions. In ideal model setup
with regular element shapes, incompatible elements provide results that are com-
parable to those of higher order elements, but with fewer degrees of freedom.
Parabolic deformation profile over the element boundary, however, can naturally
lead to incompatible deformation profiles across element boundaries (discussed
further in Chapter 4). Response of incompatible elements tends to be very sensi-
tive to geometric distortion. Eight-node elements provide a much better and robust
alternative to these incompatible elements. Nine-node Lagrangian elements, with
a center node added to the eight-node element, provide more flexible and robust
behavior when element shape is distorted from regular square or rectangular shape.
However, not all finite element software packages include Lagrangian element
option.
Each set of ABAQUS solid elements requires a property data block with material
name and part thickness information as described in the following:
*SOLID SECTION, ELSET= Plate50×50, MATERIAL=mat1
1.0
where the value “1.0” is an example for part thickness (Figure 3.17). Material proper-
ties, for the given material name “mat1” in the above, are described in the ABAQUS
data block under “*MATERIAL” keyword as described in Section 2.10. These
Analysis of Solids Represented by 2D Stress Fields 81
material properties are used by ABAQUS to calculate stress–strain relationship
matrix [C], using equation (3.63) or (3.64), depending on the stress-field assumption
of plane strain or plane stress, respectively. Nodal coordinates, element connectivity
to nodes, element type selection, part thickness information, and material property
data provide essential and sufficient information to ABAQUS for the calculation of
stiffness matrix using equation (3.84). Assignment of boundary conditions on nodes
follows the same procedure outlined in Section 2.10. In addition to concentrated
nodal loads (described in Section 2.10), solid elements can be subjected to distrib-
uted loads along the edges. Figure 3.20 shows distributed traction applied on the
upper edges of elements 3 and 4. ABAQUS syntax rule is to define pressure data by
identifying element ID followed by the element face or edge, and the magnitude of
applied pressure (+ve sign is used for pressure on the edge). Element face ID of an
element is defined by following the order of element connectivity definitions. Face
between the first and second nodes, listed with the element ID in *ELEMENT defini-
tion, is referred to as face ID “1”; face ID “2” is defined as that defined by second and
third nodes of an element, and so on. ABAQUS input data, for the applied distributed
load on example elements of Figure 3.20, is given in the following:
*DLOAD
3, P3, –1.0
4, P3, –1.0
In the above, “3” and “4” are element ID numbers, “P3” refers to pressure load on
face #3 of the elements, and the negative value of 1.0 indicates a distributed traction
load (a pressure load is associated with +ve sign in ABAQUS). Analysis step defini-
tion in ABAQUS for the given load follows the same format with “*STEP” and
“*END STEP” keywords as discussed in Section 2.10. A complete ABAQUS input
description of the four-element model, shown in Figure 3.20, is provided in the
following:
*HEADING
Plane-stress analysis model using four-node quadrilateral elements
*NODE, NSET=setn1
1,0.0,0.0,0.0
2,25,0
3,50,0
4,0,25
5,25,25
6,50,25
7,0,50
8,25,50
9,50,50
*ELEMENTS, ELSET=Plate50×50, TYPE=CPS4
1, 1, 2, 5, 4
2, 2, 3, 6, 5
3, 4, 5, 8, 7
82 Finite Element Analysis of Solids and Structures
4, 5, 6, 9, 8
*SOLID SECTION, MATERIAL=mat-1, ELSET= Plate50×50
1.0
*MATERIAL, NAME=mat-1
*ELASTIC
210,0.3
*BOUNDARY
1,1,2,0.0
4,1,2,0.0
7,1,2,0.0
*STEP, PERTURBATION
Distributed pressure load on elements
*STATIC
*DLOAD
3, P3, –1.0
4, P3, –1.0
*NODE PRINT
U
RF
*EL PRINT
S
*END STEP
3.12 DESCRIPTION OF LOAD TYPES IN GENERAL STRESS
ANALYSIS PROBLEMS
In actual solution of matrix equilibrium equations, loads are represented at the nodes
aligned with the associated DOF. General-purpose FEA software packages accept
input data describing surface and or body loads (e.g. *DLOAD specification in
ABAQUS). These non-nodal loads are internally substituted by the software pack-
ages with equivalent nodal loads for actual analysis purpose. Element shape func-
tions are generally used to determine the influence functions in nodal load calculations.
For example, influence function for node #1 in eight-node solid element of Figure
3.14(a) is defined by substituting s = 1 in the 1st expression of equation (3.85):
1
H1 r. 1 r (3.96)
2
Equivalent load on node #1, for a distributed load p(r) acting on the edge 1–2–3
of that element, will be calculated as follows:
1
P1
H1. p r .dr
1
(3.97)
Similarly, equivalent loads on nodes 2 and 3 can be calculated by using the influ-
ence functions for nodes 2 and 3, respectively. Similar procedure is also applied to
Analysis of Solids Represented by 2D Stress Fields 83
substitute surface pressure and body loads with equivalent nodal loads in 3D finite
element simulation models. Distributed loads on element edge, surface, and body
may also be simply lumped to the affected nodes during model build process without
using the software provided option of distributed load description.
Pre-existing stresses and strains can be considered in finite element analysis of
structures. General-purpose finite element software packages allow the initial stress
inputs at finite element integration points. Equivalent nodal loads are calculated by
the software packages internally:
P B 0 .dV
T
(3.98)
where {σ0} is the vector of initial stresses at element integration points. Pore fluid pres-
sure, in porous solids, can also be represented by {σ0} in equation (3.98). Temperature
change effects are represented by equivalent thermal strains in the material:
T . T T0 (3.99)
where T is current temperature at element integration point, T0 is the stress-free refer-
ence temperature, and α is the thermal expansion coefficient of the material. Thermal
strains from equation (3.99) are converted to equivalent thermal stresses by using
Hooke’s laws, and the equivalent nodal loads for temperature change effects are cal-
culated by using equation (3.98). Theoretical manuals of software packages gener-
ally contain descriptions of how distributed surface and body loads are substituted
with equivalent nodal loads for internal calculations.
3.13 REFINED FINITE ELEMENT MODEL PREPARATION WITH A
PRE-PROCESSOR (HYPERMESH)
Manual model preparation, as described in Section 3.11 for four-element 2D finite
element model, is not suitable to prepare refined models with large number of ele-
ments and nodes (such as the 12 × 12, 18 × 18, and 72 × 72 models shown in Figure
3.19). Altair’s HyperMesh (Altair University 2020) is a general-purpose finite ele-
ment model preparation software with in-built templates to prepare models for several
finite element analysis software packages including ABAQUS. A step-by-step video
demonstration of preparing a fine mesh model of the plane-stress problem (of Figure
3.17) is provided in the support data website (File name: HyperMesh_ Preparation_
of_ABAQUS_InputFile_for_a_2D_Plane_Stress Problem.mp4). Key data blocks
saved by HyperMesh in the ABAQUS input format are listed in the following:
*NODE
5,0.0,0.0,0.0
…..
…
*ELEMENTS, ELSET=auto1, TYPE=CPS4
…
…
84 Finite Element Analysis of Solids and Structures
*SOLID SECTION, MATERIAL=mat-1, ELSET= auto1
1.0
*MATERIAL, NAME=mat1
*ELASTIC
210,0.3
*STEP, PERTURBATION
*STATIC
*BOUNDARY
…..
…….
*CLOAD
…
……
*END STEP
ABQUS model input file (Plate 25x25.inp), saved at the end of HyperMesh ses-
sion, can be run by ABAQUS and the results of that run can be post-processed either
in ABAQUS/CAE graphical post-processing tool or be processed with Altair’s
HyperView post-processing tool. Similar model building exercise can be undertaken
with other pre-processor tools if desired.
3.14 PRACTICE PROBLEMS – STRESS ANALYSIS OF A PLATE
SUBJECTED TO IN-PLANE STRESSES
PROBLEM 1
Four-element model of the 50 mm × 50 mm plane-stress plate shown in Figure 3.20
underestimated the displacement response (Figure 3.18), and produced crude profiles
of stress distributions in the plate (Figure 3.19). Re-analyze the problem using four
eight-node plane-stress elements (CPS8) and compare the results with those pre-
sented in Figure 3.19.
PROBLEM 2
An approximate stress function for the problem in Figure 3.17 is given by the follow-
ing equation (where c = 25 mm, and p = 1 kN/mm):
p
40c3
10.c3 .(2c x )2 15.c 2 .(2c x )2 .( y c) 2.c 2 .( y c)3 5.(2c x )2 .( y c)3 ( y c)5
(3.100)
Check that the stress function satisfies the compatibility equation and stress bound-
ary conditions. Predict the vertical displacement at upper right corner of the plate;
and compare the result with those reported in Figure 3.18 from finite element analy-
sis models.
PROBLEM 3
Prepare a refined finite element model for the problem in Figure 3.17, using eight-
node plane-stress elements, and compare the results with those obtained in Problem-1.
4 FEA Model Preparation
and Quality Checks
SUMMARY
This chapter focuses on how to produce good-quality finite element models of two-
dimensional solids by using general-purpose model pre-processing software. Although
specific references are made to HyperMesh software features for model build opera-
tions, and to ABAQUS software for actual FEA solutions, the discussions on key
aspects of quality model preparation are equally applicable to other software products
for model preparation and validation of results. Introductory ideas for adaptation of
finite element grid refinement to local variations of stress field are presented in Section
4.1. The important step of choosing finite element formulation type for appropriate
representation of the stress field problem is discussed in Section 4.2. A brief discus-
sion on the processing of geometric design data for stress analysis exercise is pre-
sented in Section 4.3. Section 4.4 presents (a)-to-(g) critical details of finite element
model preparation and quality checks. Section 4.5 focuses on basic quality checks of
analysis results that are essential for validating a model prior to drawing engineering
conclusions from the model predictions. Practice problems on model preparation and
quality checks are presented in Section 4.6. Discussions and examples in this chapter
have remained primarily focused on analysis of 2D stress field problems. However,
many of the same metrics for model and result quality checks will apply to finite ele-
ment simulation of 3D problems that will be introduced in next chapters.
4.1 ADAPTATION OF FINITE ELEMENT MESH TO STRESS FIELD
VARIATIONS
Finite element mesh refinement examples shown in Figures 3.18 and 3.19 have used
uniform grid patterns. However, mesh grid refinement may need to be varied within
the same analysis model depending on the complexity of variation in a stress field.
Stress variation within a homogeneous solid body, in presence of body forces, is
described by equilibrium equations (1.10). In absence of body force, one-dimen-
sional stress field in a homogeneous body (Figure 4.1) shows no variation of stress:
x
0 (4.1)
x
Fx
FIGURE 4.1 Stress variation in a one-dimensional field.
85
86 Finite Element Analysis of Solids and Structures
σl
σn σn
2b
2a
FIGURE 4.2 Stress flow around a hole in an axially loaded flat plate.
However, a geometric discontinuity, such as a hole in a plate (Figure 4.2), can cause
a perturbation of the stress flow. Maximum local tangential stress on the upper edge
of the hole can be defined as follows:
l kl . n (4.2)
where σn is the nominal axial stress acting on the left and right boundary edges of the
plate; and kℓ is the stress concentration factor defining the amplification of stress at
the local measurement point. For a general elliptical hole of dimensions 2a x 2b at
the center of a very large plate, the maximum stress concentration on the upper edge
of hole is defined by equation (4.3) (Ugural and Fenster 2012):
b
kl 1 2. (4.3)
a
For a circular shape hole (a = b), equation (4.3) gives a stress concentration factor
kℓ = 3, thus defining the local stress in equation (4.2) as three times the magnitude of
applied nominal stress (σn). Axially loaded plate with hole (Figure 4.2) is a good
example with known theoretical solution that is studied in this chapter to verify the
convergence property of finite element stress analysis models. It also provides the
opportunity to verify Saint Venant’s principle (Saint-Venant 1797–1886) that the
local disturbance in a stress field disappears at a distance away from the point of
geometric imperfection.
Figure 4.3 shows approximate subdivisions based on the gradient of variation in
the stress field. Finite element modeling can be adapted to this variation by using
more refined mesh in the higher gradient area, and less refined mesh in the relatively
smoother stressed areas (Figure 4.4). Adaptation of mesh refinement with high num-
ber of lower order elements (i.e. linear solid elements in this particular example) is
known as “h-adaptivity” where “h” refers to the element size that is reduced in higher
stress gradient area. An alternative to h-adaptivity is to use higher order polynomial
functions (shape functions), with fewer number of elements, to capture the gradient
of stress variation in a model. Figure 4.5 shows a model of the plate with hole – using
eight-node quadratic elements around the curved perimeter of the hole. This mesh
adaptation is known as “p-adaptivity”, where p refers to the degree of element shape
FEA Model Preparation and Quality Checks 87
σl
σn σn
Lower stress Zone of high Lower stress
gradient stress gradient gradient
FIGURE 4.3 Subdivisions based on the gradient of stress variation in the axially loaded
plate with a circular hole at the center.
Tria Quad
σn σn
Refined mesh to
Less refined capture higher
mesh stress gradient
FIGURE 4.4 Finite element model with fine mesh of lower order solid elements (Tria and
Quad) in the area of high stress gradient.
8-node quadra c
element
σn σn
Higher order
elements in the
zone of high
stress gradient
FIGURE 4.5 Finite element model of higher order elements (eight-node quadratic) in the
area of high stress gradient.
88 Finite Element Analysis of Solids and Structures
function. Quadratic solid elements in Figure 4.5 use mid-side nodes approximately
in-between two corner nodes of solid elements. Special element formulations by
moving middle nodes closer to one corner can be used to reproduce high gradient of
stress–strain variations near points of stress singularity (Zienkiewicz and Taylor
1989). Choice of mesh adaptation technique, higher number of lower order elements
versus higher order elements, depends on the analyst’s experience with specific anal-
ysis problems. Specific nature of the deformation field of some problems may make
the use of higher order elements more suitable than the lower order ones as discussed
in the following section.
4.2 ELEMENT TYPE SELECTION FOR A GIVEN STRESS ANALYSIS
PROBLEM
Figure 4.6(a) shows bending deformation profile of beam type structure where planes
AB and CD, initially perpendicular to beam longitudinal axis, remain plane after
bending, and rotate with the beam axis. Edges AC and BD, shown in Figure 4.6(b),
follow the curved profiles of free boundaries. Planar response of sections AB and CD
implies linear variation of normal strains along the beam section. Any finite element,
used to model the beam bending response, should ideally be capable of representing
this linear variation of strain in direction y (on cross-section perpendicular to the
beam axis). Triangular elements (Section 3.9), which can represent constant stress/
strain response within element domain, make the overall response stiff. Bi-linear
shape functions (equations 3.65), representing the deformation profile of quadrilat-
eral solid elements, can naturally reproduce the linear rotation behavior of beam
sections AB and CD. However, use of the linear function, to represent node-to-node
deformation profile along element boundary, does not naturally reproduce the
A C
B D
(a)
y
M A C M M1 A C M1
θ θ
s
2b
r x
B B D
D 2a
(b) (c)
FIGURE 4.6 (a) Bending deformation of beam, (b) idealized deformation of an element
under in-plane bending, and (c) linear deformation profiles along the edges of a four-node 2D
element.
FEA Model Preparation and Quality Checks 89
bending-induced curved profiles on edges AC and BD in Figure 4.6 (c). Using para-
metric coordinate system definitions from Section (3.6), relationships between local
(r,s) and global (x,y) coordinates for the rectangular element of Figure 4.6(c) are
defined by
x y
=r = ;s (4.4)
a b
Strain components inside the 2D solid element are given by the following equa-
tions (4.5):
u u r u s 1 u
x
x r x s x a r
v v r v s 1 v
y (4.5)
y r y s y b s
u v 1 u 1 v
xy
y x b s a r
Pure bending-induced element boundary rotation, “θ”, produces element nodal dis-
placements, ±ū, in the x-coordinate direction (Figure 4.6(c)). Displacement field
inside the element can, thus, be defined by the following equations (4.6):
u r.s u, v 0 (4.6)
Substituting the displacement functions from equation (4.6) into equation (4.5), we
get the following expressions for internal strains inside the quadrilateral element of
Figure 4.6(c):
s r
x . u, y 0, xy . u (4.7)
a b
Linear variation of normal strain on beam section, caused by applied bending
moment, is re-produced by the first part of equation (4.7). The third part in equation
represents an artificial shear strain (often called “parasitic” shear) as there should be
no shear deformation under pure bending load. This artificial shear strain is intro-
duced by the assumption of linear deformation profile along element boundaries –
preventing the reproduction of curvilinear deformation profiles along edges AC and
BD (Figure 4.6(c)). This parasitic shear strain response produces parasitic shear
stress inside the material, thus producing a different bending resistance (M1) com-
pared to the bending resistance (M) of the ideal element behavior shown in Figure
4.6(b). Displacement field for true bending deformation of element in Figure 4.6(b)
is described by the following equation (4.8):
b
u r.s u,
v 1 r2
a
2b
. 1 s2 .u
2a
(4.8)
90 Finite Element Analysis of Solids and Structures
Element strain responses corresponding to that displacement field are given by
equation (4.9), where natural condition of zero shear strain for pure bending effect is
reproduced:
s s
x . u, y . u, xy 0 (4.9)
a a
By comparing the strain energies corresponding to the two deformation conditions of
Figures 4.6(b) and 4.6(c), the following expression can be derived for the ratio
between artificial element bending resistance and true beam bending resistance
(Cook et al. 1989):
M1 1 1 1 a
2
(4.10)
M 1 1 2 b
For square shape elements (a = b), and material Poisson’s ratio ν = 0.3, equation
(4.10) gives M1 = 1.48 × M. In-plane bending resistance (M1) of quadrilateral solid
element is, thus, very high compared to “true” theoretical bending resistance (M).
And this artificial resistance becomes too big as the element aspect ratio a/b increases.
The use of fully integrated 4-node quadrilateral solid elements should be minimized
in the analysis of in-plane bending problems.
A numerical remedy for parasitic shear problem of quadrilateral elements is to
conduct internal calculations at element centroid (r = s = 0), thus getting zero
value for parasitic shear strain (γxy = 0) in equation (4.7). This action, however,
also produces zero normal strain (εx = 0) at element centroid, making the element
vulnerable to zero energy hourglass deformation mode as discussed in Figure 3.13
and Section 3.7. Software implementation of reduced integrated solid elements
often includes numerical countermeasures to reduce the risk of developing
uncontrolled hourglass deformation mode. ABAQUS element library, for exam-
ple, includes reduced integration forms of quadrilateral solid elements CPS4R and
CPE4R, respectively, for plane-stress and plane-strain conditions. An artificial
internal resistance is added to element formulation for producing some resistance
to hourglass deformation mode. Use of reduced integration elements works well
when many elements are used to capture the bending profile of a solid member.
Higher order solid elements, discussed in Section 3.8, provide more stable and
better choice for simulating the bending deformation problems. Selection of
desired element type is a critical step in preparation of finite element analysis
models by using model pre-processing software products. Section 4.3 in the fol-
lowing introduces a plane stress analysis problem having geometric discontinuity.
Section 4.4 describes the critical model preparation steps, with selective mesh
refinement (h-adaptivity) in areas of expected stress concentration, by using the
general-purpose finite element model pre-processor HyperMesh (Altair University
2020).
FEA Model Preparation and Quality Checks 91
4.3 INITIAL GEOMETRIC DESIGN OF STRUCTURAL
COMPONENTS
Initial geometric attributes of component design can be defined by using point, line,
surface, and volume attribute definition features available inside a finite element
model preparation software package (group 2 in Figure 1.19). However, in complex
engineering projects, it is a standard practice for one team to develop the CAD model
by using a specialized design software such as the ones listed in group 1 of Figure
1.19, while another team performs analysis work on the same design by using differ-
ent software packages from groups 2 and 3. And at the same time, a third team
responsible for manufacturing the product may use yet another different software to
process the initial design data. In-built database formats of CAD design software
packages are generally not compatible with those of the analysis software packages
used by the downstream users and suppliers. Transfer of CAD data from initial
design software to downstream user platform is achieved through the use of standard
data formats. Two of the widely used data exchange formats are IGES and STEP.
IGES (Initial Graphics Exchange Specification), last published by the American
National Standards Institute (ANSI) in 1996, was developed primarily for the
exchange of pure geometric data between computer-aided engineering platforms.
This data exchange format is supported by many general-purpose CAE software
packages. However, STEP (STandard for the Exchange of Product model data),
developed later based on ISO 10303-11 standard (ISO 2020), is currently the default
choice for engineering data exchanges. ISO 10303 employs the neutral ASCII format
as the medium for data exchanges between different CAE platforms. Data file
“Plate_with_hole.step”, provided as support electronic data with this book, contains
the geometry data of the plate with hole shown in Figure 4.7. Data inside a STEP file
is described by using the EXPRESS modeling language specified in Part 11 of the
standard ISO 10303-11. EXPRESS is not a programming language; it is a standard
for the computer-interpretable representation and exchange of product data through
100 D = 20
x
100
250 250
FIGURE 4.7 Geometric description of a plate with hole.
92 Finite Element Analysis of Solids and Structures
schemas and constraints. Example syntax of data description in STEP file is shown
in the following:
ISO-10303-21;
HEADER;
…
DATA;
….
#161=CARTESIAN_POINT(' ',(2.5E+02,10.E+01,0.));
…
……………
#176=LINE(“,#161,#171);.
……………
END-ISO-10303-21;
General-purpose finite element model processing software products, listed in group-2
of Figure 1.19, can be used to import the geometry data file (“Plate_with_hole.step”),
and a finite element model can be prepared for analysis with one of the target FEA
software packages listed in group 3 of Figure 1.19. Section 4.4 in the following
describes the key steps for developing an ABAQUS finite element analysis model of
the plate with hole example by using Altair’s HyperMesh software (Altair University
2020).
4.4 FEA MODEL PREPARATION FOR CASE-STUDY: PLATE WITH
HOLE (BY USING HYPERMESH)
HyperMesh can prepare finite element model input file for several different FEA
packages. Choosing the target FEA software at the launch of HyperMesh session,
through a dialog box as shown in Figure 4.8, helps to build an error-free model in
subsequent steps. Key model preparation steps are described in this section. A video
session file of this specific HyperMesh model preparation example is provided in
data download site as specified in the preface of this book (HyperMesh_Session_for_
ABAQUS_Model_Preparation_of_2D_Plane_Stress_Plate_with_Hole.mp4). A blended
discussion of quality model preparation steps, with HyperMesh specific menu
options, is presented in the following. Users planning to try out HyperMesh model
pre-processor may want to review the following text simultaneously with the above-
mentioned video session file. Key model preparation steps described here are equally
applicable if a user prefers to use a different FEA model preparation software (from
group 2 in Figure 1.19).
4.4.1 CAD Data Preparation for FEA Modeling
First step in the model preparation task is to import the CAD data into the pre-pro-
cessor database, for example by using “File-Import-Geometry” steps from
HyperMesh menu. Upon successful completion of the data import step, model pre-
processing software will show the imported geometry data in its display window.
FEA Model Preparation and Quality Checks 93
FIGURE 4.8 Target FEA solver pre-selection in a model pre-processor.
FIGURE 4.9 Geometry data of plate with hole.
Figure 4.9 shows a shaded view of imported geometry data from file Plate-with-hole.
step (solid surface area is shown filled by selecting “Shaded Geometry and Surface
Edges” from HyperMesh Panel menu). At this stage of model pre-processing, sub-
stantial effort can be required for cleaning of the imported CAD data -specifically for
94 Finite Element Analysis of Solids and Structures
complex part geometries. Current example case study of plate with hole presents
very clear CAD data (as shown in Figure 4.9) without the need for further data clean-
ing steps.
4.4.2 Finite Element Meshing of the CAD Geometry
As discussed in Section 4.2, a critical step before actual mesh modeling of part
geometry is to identify the desired element types for appropriate simulation of the
target stress field. Menu panel “2D” in HyperMesh provides a pick option of “elem
types”. Activation of that task bar brings up ABAQUS element type choices for 2D
problems discussed in Section 3.11. Figure 4.10 shows the selection of CPS3 and
CPS4 for tria3 and quad4 element types, respectively. Higher order CPS6 and CPS8
element types can be selected against tria6 and quad8 categories if these are desired
to be used in subsequent meshing step. These element type names are specific to
ABAQUS element library available for plane stress modeling of part geometry lying
in a 2D plane. Element type names will vary if a different FEA software name is
chosen for target model preparation. Pre-selection of target FEA software, with
appropriate element type names, minimizes the downstream manual work for model
corrections. The next important step in software-aided finite element model prepara-
tion from CAD geometry data is to choose a suitable algorithm for discretization of
geometry to finite elements and nodes. HyperMesh, for example, provides many dif-
ferent options under “2D” menu panel for meshing of both two- and three-dimen-
sional surfaces. Choosing the “automesh” option brings up the panel menu for
user-guided interactive meshing of the CAD data. Parametric controls for interactive
mesh generation operation can be defined at this step such as minimum or maximum
element size, element internal angle, degree of polynomial (linear vs quadratic), etc.
A preliminary mesh of first-order quad and tria elements, with approximate element
size not exceeding 10 mm, shows a refined enough mesh at outer boundaries of 200
mm × 500 mm plate, but a very crude discretization around the hole of 20 mm diam-
eter. Problem-specific prior knowledge is critical at this stage to accept the mesh or
reject it if the discretization is not expected to meet target engineering analysis
results. For example, plate with hole is expected to have high stress-gradient in the
immediate vicinity of the hole boundary. HyperMesh provides the desired opportu-
nity for adapting mesh refinement to specific areas by immediately executing the
“reject” option, which leads to interactive window showing geometric boundary lines
with expected number of grid divisions in the display window (Figure 4.11). Number
of grid division on each CAD feature line can be increased or decreased with mouse
click on the respective displayed number. After adjusting the grid density to desired
FIGURE 4.10 Pre-selection of element types to be used in finite element model.
FEA Model Preparation and Quality Checks 95
60
30
30
30
60
FIGURE 4.11 Discretization of geometric edges for selective adjustment of finite element
grid refinement.
FIGURE 4.12 Automatic mesh generation by a model pre-processor based on user inputs
for mesh grid refinement.
number of divisions on each feature line, a new mesh is generated with the execution
of “mesh” command. Figure 4.12 shows a mesh generated with 30 grid points on the
perimeter of hole, giving an element size of approximately 2.1 mm around the hole,
while element sizes along the boundaries of the plate remain at about 6–8 mm. A
revised version of the mesh can be generated again by rejecting the current displayed
version and adjusting the number of grid points on geometric features lines as desired.
Evidently, a priori knowledge of the analysis problem is essential to guide the inter-
active meshing exercise, with visual assessment of the finite element mesh to meet
the eventual analysis objectives.
4.4.3 Finite Element Mesh Quality Checks
As evident from Figure 4.12, element shapes do not remain at ideal square or equi-
lateral triangular shapes as the mesh transitions from refined density at the vicinity of
the circular hole to coarser density along boundaries of the rectangular plate. Negative
effects of element distortion on predicted result quality have been discussed with an
elementary stress analysis example in Section 3.7 and Figure 3.12. It is important to
check element quality after developing the mesh model of a component. HyperMesh
96 Finite Element Analysis of Solids and Structures
provides several element quality metrics like skew, warpage, aspect ratio, Jacobian,
interior angles, etc., to report the quality of a generated mesh. Metrics for mesh qual-
ity checks vary from software to software. Few of the commonly used element qual-
ity measurements are discussed in the following.
Warpage: A quad element is temporarily split into two adjacent triangles (for
quality check only), and warp angle is defined as the angle between the
normals to two triangular planes. A quad element can be split into triangles
in two different ways by using either of the two diagonal lines. HyperMesh
reports the maximum value of the two possible angles. Ideal value of warp-
age is “zero” while an acceptable limit may be set at ≤10°. Warpage check
is important in 3D surface models. Here in the current example of 2D plane
stress problem, warpage angle is zero.
Aspect ratio: Aspect ratio of an element is defined as the ratio between maxi-
mum element edge length and minimum edge length. Ideal value of aspect
ratio is “1” while an acceptable limit can be considered up to <5 for the
current analysis problem subjected to in-plane traction mode of deforma-
tion only.
Skew: Interior angle at each element corner node of a quad element is calcu-
lated, and the skewness of element is defined as 90° minus minimum inter-
nal angle of the quad element. Ideal value for skewness is “0” while an
acceptable limit is generally considered to be ≤45°.
Jacobian: The determinant of Jacobian matrix (equation 3.70) measures the
numerical relationship between element local and global coordinate sys-
tems. HyperMesh evaluates the determinant of the Jacobian matrix at each
integration point, and it reports the ratio between the smallest and the larg-
est values. An ideal square shape element has that ratio at 1. As the element
becomes distorted, that ratio approaches a zero value, and a concave shape
element reaches a value of -1. Acceptable limit of Jacobian ratio is set at
≥0.7.
Min/max angle: Minimum and maximum values of internal angles of triangu-
lar and quad elements are reported to check the general distortion of ele-
ments. Ideal and acceptable values for quad elements are, respectively, 90°
and ≥45°. Corresponding values for triangular elements are 60° and ≥20°.
HyperMesh reports of element quality checks can be generated by going through
“Tool” menu panel and choosing “check elements” submenu item. Figure 4.13 shows
a graphical report generated by HyperMesh displaying elements with Jacobian ratio
values less than 0.7. Quality issues of computer-generated meshes may need to be
corrected manually by modifying the nodes and elements. A completely new mesh
may need to be created occasionally when there are too many mesh quality issues in
the vicinity of critical engineering interest. Other than element shape quality issues
discussed above, computer-aided finite element mesh generation may also experi-
ence other quality glitches. For example, free edges of finite elements should always
match with free boundaries of CAD geometry. Discovery of internal free edges along
FEA Model Preparation and Quality Checks 97
FIGURE 4.13 Graphical display highlighting the elements that fail to meet specific quality
check criterion.
inter-element boundaries indicates absence of inter-element connectivity along those
free edges. Under the “Tool” menu panel, HyperMesh provides the submenu option
of “edges” that provides command option of “find edges” in selected components.
After deleting the free edges, nodes within immediate vicinity of one another can be
displayed by using the command “preview equiv” with a very small value of toler-
ance (0.1 mm for example). Displayed nodes from “preview equiv” step can be
merged together by choosing the command “equivalence” at that stage. Extra caution
must be exercised to check the physical validity of such merger operation as it may
lead to unacceptable representation of the physical geometric features.
Mistakes made during meshing operations can result in duplication or multiplica-
tion of nodes and elements overlaid on the same geometric entity. Once again, mod-
ern software packages come handy with built-in tools to correct such meshing
errors. Under HyperMesh menu item “Tool”, “check elems” submenu lists a com-
mand named “duplicates” to activate duplicate element check. Unnecessary dupli-
cate elements, found through that check process, can be deleted at that point.
Multi-step model creation and cleaning operations may also leave extra nodes float-
ing in the database without connectivity to element mesh. Under the “Geom” menu
panel, HyperMesh lists “temp nodes” submenu option that can be used to visualize
all nodes; and to eliminate all floating nodes by using the command button “clear
all”. As discussed in Section 3.11, ABAQUS expects that 2D solid elements be
defined by listing the corner nodes following counter-clockwise sequencing, which
means that normal to the elements in the x–y plane will be directed in the z-direction
toward the observer facing the model on the x–y plane. Automatic mesh generation
may occasionally lead to opposite orientation of element normal, which must be
checked and corrected if needed. In the “Tool” menu panel of HyperMesh, “nor-
mals” submenu provides the capability to display element normals by putting a posi-
tive numerical value under “vector display”, and by executing the menu mutton
“display”. Finite element model may need to be rotated in 3D space for proper
visualization of the element normal direction. If the normals appear in the opposite
direction of ABAQUS specification, one click “reverse” button will set the element
98 Finite Element Analysis of Solids and Structures
normals to opposite direction. Normal definition, assumed correct for ABAQUS,
may not be correct for other finite element analysis packages as there is no universal
standard for how nodal connectivity should be sequenced in element definition.
Users should consult the manuals of specific FEA solver to identify the correct ele-
ment definition format.
Above discussions include common mesh quality checks while preparing a finite
element analysis model by using the general-purpose model preparation software
HyperMesh. An audio-visual presentation on mesh quality checks with HyperMesh
is available online at Altiar University (2014). Some of these same items may apply
to model quality checks with other pre-processors as well. Upon verification of the
mesh quality, user may proceed with the following key steps, 4.4.4 through 4.4.7, for
the completion of an ABSQUS finite element analysis model by using HyperMesh
pre-processing software.
4.4.4 Material and Part Property Assignment
Essential material property data for the calculation of elasticity matrix (equation
3.64) are provided through *MATERIAL data block in ABAQUS input files (dis-
cussed in Section 2.10). Through the “Materials” menu panel in HyperMesh, a user
can proceed to specify a material name (for example, Mat-1), and can choose
“ABAQUS_MATERIAL” for “card image” option. Execution of “create/edit” task
bar opens the menu for preparation of the ABAQUS material model inputs (equiva-
lent of manual input description under *MATERIAL data block). Checking of
“Elastic” option will open input cells to define isotropic material properties of
Young’s modulus (E) and Poisson’s ratio (ν). For the case study example of plate
with hole, E = 70 and ν = 0.333 are provided as inputs for the elastic behavior of
aluminum material. Execution of “return” button saves the data in HyperMesh data-
base, and returns the user to main menu panel. Selection of “Property” attribute icon
from main meu panel will open the dialogue subpanel for creating property data for
a part. Required inputs are property name (for example Prop-1), “SOLID SECTION”
as input for “card image”, and selection of previously defined “Mat-1” under the
selection window of “material=”. Execution of “create/edit” will lead to new sub-
menu for defining the section properties of the plane-stress plate. Checking the
“DataLine” option provides an input box under “Attribute_Value” where user can
specify part thickness value (“1.0” for the example plane-stress plate problem).
Execution of the “return” button will take the user back to property creation sub-
panel. Selection of “assign” button on that subpanel leads the user to choose ele-
ments “all” or “displayed”. Previously created property data card (Prop-1) can be
selected in the “property” input box. Execution of the “assign” option at that point
assigns the selected property data (including the associated material data with it) to
all selected elements. An essential step of model quality check is to select “Card
Edit” from main menu panel, and then verify that the material and property data have
been correctly assigned to the finite element model in display window. Assignment
of material and section property data to the elements completes the physical descrip-
tion of the part to be analyzed.
FEA Model Preparation and Quality Checks 99
4.4.5 Analysis Parameters – Boundary Constraints, External Loads
and Model Outputs
Specification of proper boundary conditions is an essential step for building a stable
analysis model. Figure 4.14 shows a schematic description of loads and boundary
constraints to be applied on the example problem of plate with hole. Problem-specific
descriptions of constraints and loads can be described through graphical interfaces of
HyperMesh without the need for downstream manual modifications of the final
model input file. “Analysis” option under HyperMesh menu panel brings up selection
menu for assigning boundary conditions and loads on a finite element model. The
“constraints” selection bar from the menu subpanel leads to the main operation stage
for defining nodal boundary constraints. Several selection options are available under
the “nodes” selection panel. A tedious but visually verifiable way is to click on the
“nodes” selection button, and choose the nodes directly on the model display win-
dow for boundary condition assignment. Figure 4.15 shows selected nodes on the
left-side boundary where constraints on x-displacement can be assigned by checking
the “dof1=0.0” in the menu selection panel, and subsequently pressing the button
“create”. Center node skipped in the selection on Figure 4.15 can be selected sepa-
rately in the next step, and boundary constraints in both x and y directions can be
assigned by checking the options “dof1=0.0” and “dof2=0.0” in the selection panel.
Pressing of “create” task bar will add this boundary constraint definition to the model
database. Boundary conditions defined in above two steps prevents any rigid body
motion possibility of the model in x–y plane. No constraint definitions are needed in
z-direction since this particular example problem has been setup as a 2D analysis
model in x–y plane.
Loading conditions on structures can be described as discrete or distributed
forces. Figure 4.14 shows a uniformly distributed traction force of 0.1 GPa acting on
the right-side boundary of the plate-with-hole. However, in actual finite element
analysis, all forces are represented as nodal forces in the solution of equilibrium
equations (1.5). Assuming that the nodes along the right-side boundary are
Aluminum: E=70 GPa, v=0.33
0.1 GPa
100 D = 20
x
100
250 250
FIGURE 4.14 External loading and boundary constraints acting on the plate in the x–y plane.
100 Finite Element Analysis of Solids and Structures
FIGURE 4.15 Node selections to specify x-constraints on the left-boundary of plate.
FIGURE 4.16 Discrete forces on the right-side boundary nodes of plate with hole.
uniformly spaced @ 6.67 mm, effective load on each boundary node of the unit-
thickness plate will be 0.667 kN except at the two corner nodes that will see an
effective load of 0.333 kN (Figure 4.16). ABAQUS expects discrete nodal loads to
be defined with “*CLOAD” keyword as discussed in Section 2.10. Similar to the
definition process for nodal constraints, HyperMesh provides the option to generate
nodal loads with the item “Forces” listed under “Analysis” menu panel. Target nodes
subjected to 0.667 kN discrete loads can be picked interactively from the display
FEA Model Preparation and Quality Checks 101
window and the relevant load amplitude can be applied in the x-direction. A similar
second step of 0.333 kN load application can be defined by choosing the two corner
nodes on the right side boundary. Boundary constraints and loads, described in
HyperMesh in the above two steps, need to be included in the definition of ABAQUS
analysis step as described in the following Section 4.4.6. Solution of system equilib-
rium equations (1.5) produces results for nodal displacements and reaction forces.
Strains and stresses inside finite elements are calculated by using equations (2.6)
and (2.26). ABAQUS stores all calculated nodal and element response results in a
binary database file (*.ODB). Selective results for nodes, elements, and global
model (such as energy balance report) can be requested to be saved in the text format
results file named with an extension “.DAT”. Specific requests for such results can
be defined through “output bock” that is listed under the “Analysis” menu panel in
HyperMesh.
4.4.6 Definition of ABAQUS Analysis Step
Linear elastic load–deflection analysis of plate with hole, subjected to boundary con-
straints and loads described above, can be executed by specifying static perturbation
analysis option of ABAQUS as discussed in Section 2.10. The “load steps” task bar
listed in “Analysis” menu panel of HyperMesh provides the relevant menu selection
items. The “edit” task bar opens the selection menu to choose “PERTURBATION”
analysis with “Static” load type. Boundary conditions and loads described earlier
need to be selected under “loadcols” selection button. Output requests defined earlier
can be included under the “outputblocks” option.
4.4.7 Exporting the Model Input File for ABAQUS Analysis
Finite element model entities described thus far can be saved as internal database of
HyperMesh. However, for ABAQUS analysis, these data need to be exported out of
HyperMesh by using the “File-Export” menu choice. This can be achieved by choos-
ing ABAQUS under “solver deck” with “standard 2D” option, and by opting to
export all data to an external data file named with extension “.INP”. Model data for
ABAQUS is saved in the ASCII format, and the saved file can be edited with a stan-
dard text editor if needed.
Experienced model builders can generally produce a usable finite element model
by going through the multi-step model preparation process – described in the above,
and also summarized with (a)-to-(g) steps in Figure 4.17. However, less-than-perfect
execution steps often produce incomplete description of the model, thus leading to
failed analysis by the downstream FEA software package. So, it is important to visu-
ally check the exported model data file before sending the model to analysis execu-
tion. Obviously, checking the mesh quality of a large complex model is beyond the
scope of manual data check at that point. However, visual checks often identify com-
mon mistakes that occur in key model assumptions such as element types, loads,
boundary conditions and analysis type selection. Essential key words in ABQUS
input file of plate with hole problem are shown in the following, where dot point
clusters refer to bulk datasets that are omitted from this presentation:
102 Finite Element Analysis of Solids and Structures
*HEADING
…………
*NODE
……..
*ELEMENT, TYPE=CPS3, ELSET=Plate-with-hole
…………
*ELEMENT, TYPE=CPS4, ELSET=Prate-with-hole
…………
*SOLID SECTION, MATERIAL=Mat-1, ELSET=Plate-with-hole
1.0
*MATERIAL, NAME=Mat-1
*ELASTIC
70.0, 0.33
*STEP, PERTURBATION
*STATIC
*BOUNDARY
……………
*CLOAD
…………….
*NODE PRINT
RF
…..
*ENRGY PRINT
*END STEP
Submission of analysis model file to ABAQUS, and extraction of result files follow
the same procedures described for practice problems presented in Chapters 1–3.
CAD file: Develop FEA Mesh
b
Plate_with_hole.STEP
Import it to
HyperMesh a HyperMesh
Assign material and
d Verify model quality c
sec on proper es
Define loads, BCs and Define f Export model for FEA
output variables e analysis step analysis g
FIGURE 4.17 Key-process steps (a)–to-(g) in computer-aided model preparation.
FEA Model Preparation and Quality Checks 103
4.5 POST-PROCESSING OF FEA RESULTS
The very first step in post-processing of finite element analysis results is the basic
quality check of results to validate the reliability of analysis model. Few of the rou-
tine result quality checks include: (i) energy balance check, (ii) verification of system
equilibrium, and (iii) compatibility checks of displacement and stress analysis results.
In static load analysis cases, velocity and acceleration responses at nodal degrees-of-
freedom are “zero”. Internal energy stored in the system must be equal to the external
work done, as defined in the following equation (4.11):
1 1
Internal strain energy, T dV External work done, . P . u (4.11)
T
2 2
Energy error in finite element analysis of linear elastic static problems should be:
Energy error External work done Internal strain energy 0 (4.12)
Energy balance report from ABAQUS can be checked in “.DAT” output file when
“*ENERGY PRINT” command is included in model preparation step (e) of Figure
4.17. Verification of energy balance in an example analysis is shown in Figure 4.18.
Verification of the system equilibrium is another check that should be performed at
the beginning of result post-processing step. In the plate with hole analysis example,
applied load on the right-side boundary is 200*1*0.1 = 20 kN in the x-direction and
“zero” is the y-direction. Summation of reaction forces saved in the “*.DAT” file
produces a value of “20.0” in the x-direction, and “0.0” in the y-direction, thus pass-
ing the equilibrium quality check of results (Figure 4.19). Other simple checks of
result quality can be achieved by verifying the displacement and stress boundary
conditions, as shown in Figures 3.18 and 3.19 for plane-stress analysis of a plate. A
special check of the stress analysis results, obtained for plate with hole, can be
achieved through the contour plot of the x-directional stresses in the plate (Figure
4.20). Maximum stress value predicted at the crest of hole by the finite element
analysis model is found to be 0.2974 GPa, which is very close to the theoretically
expected value of 0.3 GPa (three times the applied boundary stress as per equation
4.3). Model refinement around the hole at the center of plate analysis example has,
thus, adequately captured the high stress concentration at the vicinity of hole in the
stressed plate.
4.6 PRACTICE PROBLEMS – STRESS ANALYSIS OF PLATE WITH
HOLE
PROBLEM 1
Re-analyze the plate with hole example problem described in Figure 4.14 by using
six- and eight-node quadratic plane-stress elements. Compare the results with those
obtained and discussed in this chapter by using linear tria and quad elements. Prepare
104 Finite Element Analysis of Solids and Structures
FIGURE 4.18 Example of energy balance check with ABAQUS output results.
Node ID Rx Ry
Reac on forces @ y
constrained DOF
0.1 GPa
Applied force:
Rx = 0.1*200*1 = 20 kN
-20.0 0.0
FIGURE 4.19 Equilibrium check with simulation model results.
FEA Model Preparation and Quality Checks 105
Theore cal result: maximum stress at the p of
circular hole = 3 * nominal stress = 0.3 GPa
x=0.2974 GPa
FIGURE 4.20 Contour plot of the x-directional stress in plate with hole.
a report with the information on: (i) FEA mesh picture with boundary conditions and
applied loads; (ii) results of mesh quality checks; (iii) verification of FEA model
results; and (iv) contour of stresses σxx. Determine at what distance the Saint Venant’s
principle is applicable to this stress analysis problem of plate with hole, i.e. the local
disturbance in stress field disappears at a distance away from the point of geometric
imperfection.
PROBLEM 2
Figure 4.21 shows a plane-stress plate similar to that of Figure 4.14, with one excep-
tion related to the shape of hole that is hexagonal in Figure 4.21. Conduct finite ele-
ment stress analysis of this problem with two meshes of different refinement at the
vicinity of the hole. Compare the results and comment on the convergence character-
istic of finite element stress analysis results for this example problem.
Aluminum: E=70 GPa, =0.33
0.1 GPa
100 20
x
100
250 250
FIGURE 4.21 Plate with a hexagonal hole at the center.
5 Stress Analysis of
Axisymmetric and
General 3D Solids
SUMMARY
Stress analysis techniques, both analytical and numerical (finite element), have been
presented in Chapter 3 for elastic solids that are amenable to two-dimensional ideal-
ization of plane strain or plane stress states. Three-dimensional elasticity problems,
having an axis of symmetry, can also be simplified to 2D analysis models on the
plane of axial symmetry. Section 5.1 presents the basic elasticity formulations for
axisymmetric stress field by using the polar coordinate system that appears specifi-
cally suitable for this class of problems. Detailed analytical steps deriving the stress
and deformation responses of uniformly pressurized thick cylinder example are pre-
sented in Section 5.2. Understandably, these detailed analytical steps are hard to
replicate for solutions of practical engineering problems. The key ingredients of ana-
lytical stress–strain–displacement relationships, as well the analytical solutions
derived for standard examples, serve as references for the finite element solution
technique. Section 5.3 introduces the finite element method as general-purpose stress
analysis tool for more complex axisymmetric problems that cannot be readily solved
by using the analytical tools. An extension of the polar coordinate analytical frame-
work is used in Section 5.4 to predict stresses in solids for the very specific case study
of concentrated load effects. Once again, more complex 3D stress analysis cases
require the use of finite element method that is introduced in Section 5.5 by simply
adding a third dimension to the two-dimensional formulations of Chapter 3. Section
5.6 specifically focuses on the topic of software-aided model preparation and quality
checks for 3D simulations. Finally, practice problems for both axisymmetric and
general 3D solids are presented in Section 5.7.
5.1 AXISYMMETRIC – A SPECIAL FORM OF 3D ELASTICITY
PROBLEMS
Figure 5.1 shows a cylindrical body having an axis of symmetry with respect to load-
ing, geometry, material properties, boundary conditions, etc. Theory of elasticity-
based analytical solutions for such problems is efficiently described by using polar
coordinates (r, θ) (Figure 5.2). Relationships between polar and Cartesian coordi-
nates in 2D cross-sectional plane of the member are given in the following:
y
x r.cos ; y r.sin ; r 2 x 2 y 2 ; tan 1 (5.1)
x
107
108 Finite Element Analysis of Solids and Structures
Axis of
symmetry
z y
y
r
q x
pi
x p0
a
b
FIGURE 5.1 Three-dimensional problem with axis of symmetry (no variation in circumfer-
ential direction).
y
d
Fr
Fθ c
a
b
r
FIGURE 5.2 Expanded view of stress element from Figure 5.1 – shown in polar coordinates
(r, θ).
Partial derivatives of equations (5.1) yield:
r x r y
cos , sin
x r y r
(5.2)
r x sin x cos
, 2
x r r y r r
Using the chain rule of partial differentiation, the following relations are obtained
between partial derivatives of Cartesian and polar coordinate systems:
r sin
. . cos . .
x x r x r r (5.3)
r cos
. . sin . .
y y r y r r
Stress Analysis of Axisymmetric and General 3D Solids 109
5.1.1 Equations of Equilibrium in Polar Coordinates
Considering unit thickness for the differential element in Figure 5.2, the equilibrium
of forces in radial direction gives the following expression:
r
r r .dr r dr d r .r.d
d d
.d .dr.sin r .dr.sin
2 2 (5.4)
d d
r r .d .dr.cos r .dr.cos Fr .r.dr.d 0
2 2
where Fr is the body force per unit volume. Equation (5.4) can be simplified by sub-
stituting dθ/2 and 1, respectively, for sin(dθ/2) and cos(dθ/2) when (dθ) is very small;
and also by ignoring the higher order terms of the differential quantities dr and dθ. A
similar relationship can also be derived by considering the equilibrium of forces in
tangential (θ) direction. Simplified expressions for equilibrium of stresses in r and θ
are thus obtained as follows:
r 1 r r
Fr 0
r r r (5.5)
1 r 2 r
. F 0
r r r
In absence of body forces (Fr = Fθ = 0), equilibrium equations (5.5) are satisfied by a
stress function ϕ(r, θ) when the stress components in radial and tangential directions
are defined in terms of the stress function as follows:
1 1 2
r .
r r r 2 2
2 (5.6)
2
r
1 1 2 1
r 2 . .
r r r r r
5.1.2 Strain–Displacement Relationships
Denoting u for the displacement in radial direction of the differential element “abcd”
in Figure 5.2, the radial strain can be defined as follows:
u
r (5.7)
r
110 Finite Element Analysis of Solids and Structures
The tangential strain owing to radial deformation u can be derived as follows:
2 r u 2 r u
(5.8)
2 r r
Denoting the tangential deformation in element “abcd” of Figure 5.2 by v, the cor-
responding tangential strain can be expressed as follows:
v / .d 1 v
. (5.9)
r.d r
Combining the contributions of deformations u and v, total tangential strain is
given by
1 v u (5.10)
.
r r
Shear strain caused by the radial deformation u can be expressed as follows:
r
u / .d 1 u
. (5.11)
r.d r
And the shear strain caused by the tangential deformation v is given by:
v v (5.12)
r
r r
Combining equations (5.11) and (5.12), resultant shear strain is given by the follow-
ing equation:
v 1 u v (5.13)
r .
r r r
Equations (5.11), (5.12), and (5.13) provide the strain–displacement relations in
polar coordinate system for axisymmetric solid.
5.1.3 Stress–Strain Relationships
Substituting r and θ for x and y in equations (3.4) and (3.5), Hooke’s laws for axisym-
metric plane strain condition are given by
r 1 2 E E
r 2
1 2 E E
2 r
(5.14)
r 2 1
r r
G E
Stress Analysis of Axisymmetric and General 3D Solids 111
Similarly, Hooke’s laws for plane stress axisymmetric condition are obtained from
equations (3.11) as follows:
r
r
E E
(5.15)
r
E E
r r
2 1
r
G E
5.1.4 Compatibility Condition
Following the steps used for Cartesian coordinate system in Section 2.3, strain com-
patibility condition in the polar coordinate system can be obtained from equations
(5.7), (5.10), and (5.13):
2 1 2 r 2 1 r 1 2 r 1 (5.16)
2. 2 . 2 . r
r 2
r r r r r r r. r
To describe the compatibility condition in term of stress function ϕ, the Laplacian
operator of Cartesian coordinate system (equation 2.30) can be transformed into the
polar coordinate system by using equations (5.2) and (5.3):
2 2 2 1 1 2 (5.17)
2 . .
x 2 y 2 r 2 r r r 2 2
Equation of compatibility (3.16) can thus be written in the polar coordinate sys-
tem as follows:
2 1 1 2
4 2 . 2 . 2 2 0
r r r
(5.18)
r
Theory of elasticity developments in polar coordinate systems, described in the
above, provide efficient solutions for simple axisymmetric engineering problems.
Section 5.2 provides application examples of thick-wall cylinder problems. However,
similar to the discussions in earlier chapters, on relative efficiency and limitations of
analytical versus numerical analysis methods, solutions of general axisymmetric
problems tend to become more manageable with the finite element technique which
will be introduced in Section 5.3.
5.2 STRESS ANALYSIS OF AXISYMMETRIC EXAMPLE – THICK WALL
CYLINDER
Stress and deformation responses inside the thick wall cylinder of Figure 5.1, sub-
jected to uniform internal and external pressure, can be solved by using the basic
112 Finite Element Analysis of Solids and Structures
equations of Section 5.1 – without using the stress-function-based semi-inverse
method. Symmetric deformation response of the cylinder about z-axis implies that
the shearing stress is zero, τrθ = 0. Assuming that the ends of the cylinder are open
and unconstrained, stress in the z-direction is, σz = 0. Assuming z independence of the
unit length ring of Figure 5.1, and assuming no body forces are acting on the system,
the stress equilibrium equations (5.5) reduce to the following form:
r r (5.19)
0
r r
Radial and tangential stresses, σr and σθ, are related to radial and tangential strains,
εr and εθ, that in turn are related to in-plane radial and tangential deformations u and
v. However, symmetric deformation field about z- axis leads to the fact that tangential
deformation v = 0. Strain–displacement relations of equations (5.7), (5.10), and
(5.14), thus, take the following forms:
u u (5.20)
r , , r 0
r r
Combining the first two expressions of equation (5.20), we get a simple definition of
compatibility condition as follows:
u r
r r r r 0 (5.21)
r r r
Combining the Hooke’s laws for plane-stress condition (equations 5.15) and the
strain–displacement relations (equation 5.20), we get the following set of equations
for axisymmetric ring of Figure 5.1:
u 1
r r
r E (5.22)
u 1
r
r E
Re-arranging the terms in equations (5.22):
E E u u
2 r
r 2
1 1 r r
(5.23)
u
r 1 2 r r
E E u
1 2
Substituting equations (5.23) into equilibrium equation (5.19) yields the following
differential equation for radial displacement u:
2u 1 u u
0 (5.24)
r 2 r r r 2
Stress Analysis of Axisymmetric and General 3D Solids 113
The solution of equation (5.24) is given by the following general expression with
constants c1 and c2:
c2 (5.25)
u c1r
r
Substituting equation (5.25) into equation (5.23), we get the following expressions
for radial and tangential stresses:
E 1
r c1 1 c2 r 2
1 2
(5.26)
E 1
c1 1 c2 2
1 2 r
Constants c1 and c2 in equations (5.26) can be determined by using the relevant
stress boundary conditions of a given problem. For the cylinder in Figure 5.1, stress
boundary conditions on inner and outer surfaces can be described by equation (5.27),
where negative sign indicates compressive stress on a surface:
r r a pi , r r b p0 (5.27)
Using the stress boundary conditions of equation (5.27) into the first expression of
equations (5.26), we get the following expressions for constants c1 and c2:
1 a b pi p0
2 2
1 a 2 pi b2 p0
c1 . , c2 . (5.28)
E b2 a 2 E b2 a 2
Using the definitions of constants c1 and c2 from equations (5.28), stresses and defor-
mation inside the thick cylinder of Figure 5.1 are finally given by the following equa-
tions (5.29) (known as Lame’s equations):
a 2 pi b2 p0 a b pi p0
2 2
r
b2 a 2
b2 a 2 r 2
a pi b p0 a b pi p0
2 2 2 2
(5.29)
b2 a 2
b2 a 2 r 2
u .
.
1 a pi b p0 r 1 a b pi p0
2 2 2 2
E b2 a 2 E b2 a 2 r
For the special case of internal pressure only (external pressure: p0 = 0), expressions
in equations (5.29) take the following form:
114 Finite Element Analysis of Solids and Structures
a 2 pi b2
r 1
b2 a 2 r 2
a 2 pi b2
1 (5.30)
b a2 r 2
2
a 2 pi .r b2
u 1 1
E b2 a 2 r2
Numerical value of maximum radial stress occurs on inner surface of the cylinder
(r = a). Since b2/r2 > 1, σr is negative (compressive) for all values of r except at r = b,
where σr = 0. Tangential stress σθ is positive (tensile) for all values of r for this special
case of internally pressured cylinder. When the cylinder wall is thin, (b – a) < a/10,
expression for tangential stress in equations (5.30) takes the following familiar form,
where t = wall thickness = (b – a):
pi .a (5.31)
t
Next, for the special case of external pressure only (i.e. internal pressure: pi = 0),
equations (5.29) take the following form describing the internal stresses and defor-
mation in the thick cylinder:
b2 po a 2
r 1
b2 a 2 r 2
b2 po a 2 (5.32)
1 2
b2 a 2 r
b po .rr
2
a2
u 1 1
E b2 a 2 r2
Numerical value of maximum radial stress occurs at r = b, and it is negative (com-
pressive) throughout the thick wall. Tangential stress σθ is negative (compressive)
when the cylinder is subjected to external pressure only. Applications of these ana-
lytical stress field solutions in designing shrink-fit multilayer compound cylinders
and flywheels are available in Ugural and Fenster (2012). Evidently, foregoing ana-
lytical developments provide useful insights into the stress and deformation responses
of axisymmetric problems having constant attributes (geometry, material, loading,
boundary conditions, etc.) along the member axis. Application of this solution
becomes challenging when one or more of these attributes vary along the member
axis “z” such as the axisymmetric problem in Figure 5.3 where geometry, loading,
and boundary conditions change along the axis of symmetry “z”. Section 5.3 in the
following presents stiffness-based finite element analysis method for solving general
axisymmetric problems.
Stress Analysis of Axisymmetric and General 3D Solids 115
FIGURE 5.3 A partially filled axisymmetric cylinder of variable cross-sectional property
along axis of symmetry.
5.3 FINITE ELEMENT ANALYSIS OF AXISYMMETRIC PROBLEMS
Figure 5.4 shows x–z cross-sectional view of the 3D solid from Figure 5.3. Member
geometry, boundary conditions, and loading shown in the cross-sectional view are
symmetric about the member axis “z”. Axial symmetry of member deformation
implies that any material point on the x–z cross section can have two translational
degrees-of-freedom (u,w), but no deformation in out-of-plane direction (v = 0).
Three-dimensional problem of Figure 5.3, thus, essentially reduces to a two-dimen-
sional analysis problem in x–z plane (Figure 5.4). Following the analytical develop-
ments for axisymmetric problem in Section 5.2, strain–displacement relations in
two-dimensional x–z Cartesian coordinate system can be described as follows:
u
x
x
w
z
z
(5.33)
u w
xz
z x
u
x
z
FIGURE 5.4 Finite element discretization of x–z cross-sectional plane of axisymmetric 3D
body from Figure 5.3.
116 Finite Element Analysis of Solids and Structures
Strain components in Cartesian system are, thus, fully described by in-plane defor-
mations (u, w). Following the iso-parametric 2D element formulations of Sections
3.6 and 3.7, strain–deformation relations for four-node axisymmetric finite element
can be written by expanding the definitions from equations (3.81):
u
H1 H 2 H3 H 4 1
x 0 0 0 0 w1
x x x
x H1 H 2 H3 H 4 u2
0 0 0 0
z z z z z w2 (5.34)
.
xz H1 H1 H 2 H 2 H3 H3 H
H4 H 4 u3
z x z x z x z x w3
H1 0
H2
0
H3
0
H4
0 u4
x x x x
w4
The matrix on the right-hand side of equation (5.34) describes the standard strain–
displacement relationship matrix [B]. Terms inside [B] matrix can be calculated by
following the same procedure described in Section 3.7. Stress–strain relationship
matrix for the axisymmetric case can be written as follows:
1 1
0
1
x
x
E 1
y 1 0 (5.35)
1 1 2
1 1
y
xy
xy
0 0 1 2
0
2 1
1 1 0 1
Strain–displacement relationship matrix [B], defined in equation (5.34), and stress–
strain relationship matrix [C], defined in equation (5.35), are used in numerical inte-
gration scheme of equation (3.84) to calculate the element stiffness matrix [k].
Discussions presented in Sections 3.7 and 4.2, regarding the influences of element
shape and integration rule on the accuracy of numerically calculated stiffness matrix,
equally apply to the axisymmetric formulations as well. Higher order 2D element
formulations, discussed in Section 3.8, and lower order constant strain triangular ele-
ment formulation, discussed in Section 3.9, can also be used for axisymmetric prob-
lem analysis. Convergence properties of finite element models, as discussed in
Section 3.10, are also applicable to the axisymmetric problem analysis.
Model preparation, for axisymmetric problem analysis with general-purpose
finite element software packages, is very similar to that of plane-strain and plane-
stress analysis models described in Chapters 3 and 4. ABAQUS, for example, requires
that cross-sectional plane of axisymmetric geometry be defined by using global x–y
coordinates, where x is the radial direction and y is the member axis direction. Finite
element analysis model of an axisymmetric body can be developed from 2D
Stress Analysis of Axisymmetric and General 3D Solids 117
cross-sectional geometry by following the (a)–(g) model preparation steps described
in Section 4.4 for two-dimensional problems. Axisymmetric model description will
require selection of appropriate element types, such as CAX3 for triangular shape
constant stress/strain elements, and CAX4 for quadrilateral element of bilinear shape
function in ABAQUS models. Higher order elements with quadratic shape types, six-
node CAX6 and eight-node CAX8 element types, can also be selected if desired.
Model quality checks described in Section 4.4 apply to axisymmetric models as well.
Element type selection, for example, “TYPE=CAX4” in ABAQUS, is the key deter-
minant for analysis software to choose appropriate element formulations from equa-
tions (5.34) and (5.35). Nodal loads, corresponding to distributed pressure load on
the axisymmetric inner boundary surface in Figure 5.3, are calculated from the effec-
tive circumferential area around the axis of symmetry.
5.4 STRESS ANALYSIS OF THREE-DIMENSIONAL BODIES –
CONCENTRATED LOADS
5.4.1 Stresses in a Semi-Infinite Solid Subjected to Concentrated
Normal Force on the Boundary
Polar coordinate description of stress field, as presented in Section 5.1, can be used
to analyze the stress response of three-dimensional bodies subjected to concentrated
loads with an axis of symmetry. Figure 5.5 shows a concentrated load “P” acting on
the vertex of a 3D wedge having a vertical axis of symmetry “x”. Following analyti-
cal function, for elastic stress distribution inside the body, is applicable to points (r,θ)
away from the load application point:
cPr sin (5.36)
Concentrated line load
per unit length: P
σr
2
x
FIGURE 5.5 Concentrated load acting on the vertex of 3D wedge.
118 Finite Element Analysis of Solids and Structures
Compatibility condition of equation (5.18) is satisfied by the stress function of equa-
tion (5.36), where c is an unknown constant. Substituting the stress function from
equation (5.36) into equations (5.6), we get the following definitions for the stress
components in polar coordinate system:
cos
r 2cP , 0, r 0 (5.37)
r
The equilibrium condition between applied force P and the resultant vertical force
on a cylindrical surface at a distance r from the load application point can be written
as follows:
2
0
r .cos r.d P (5.38)
Substitution of the expression for σr from equation (5.37) into equation (5.38) leads
to the definition of constant c as follows:
1 (5.39)
c
2 sin 2
Stresses inside the wedge, defined in equations (5.37), can be re-written as follows:
Pcos
r , 0, r 0 (5.40)
1
r sin 2
2
Stresses in a semi-infinite solid, subject to a normal load on its horizontal surface
(Figure 5.6), can be obtained by substituting α = π/2 in equations (5.40):
Concentrated line load
per unit length: P
d r
A
σr
FIGURE 5.6 Radial stress inside a semi-infinite solid subject to concentrated normal load on
the horizontal surface.
Stress Analysis of Axisymmetric and General 3D Solids 119
2 P cos (5.41)
r . , 0, r 0
r
Considering a circle of diameter d, with center on the line of applied normal load on
the horizontal surface of semi-infinite solid (Figure 5.6), location of point A on the
perimeter is given by r = d. cos (θ). Substituting it in equation (5.41), radial stress on
the boundary of circle in the semi-infinite solid is thus given by the following well-
known equation:
2P (5.42)
r
d
Equation (5.42) implies that radial stress is same at all points on the perimeter of
circle in Figure 5.6, except at the load application point, and the magnitude of stress
decreases inversely with increasing distance d.
5.4.2 Stresses in a Solid Beam Subjected to Concentrated Lateral Forces
Stress derivations presented for semi-infinite solid can be effectively used to estimate
stresses inside a solid beam under concentrated lateral loading condition. Figure 5.7
shows a simply supported beam of depth h, length L, and section width b (in normal
direction of the planar view). Using the member reference coordinate system x–y, as
shown in the figure, the bending stress distribution on beam section at the mid-span
is given by the following equation:
M . y PL / 4 . y 3PL
x y (5.43)
I b.h3 /12 bh3
Concentrated: P
a A c
b r
x
h O
B
y
FIGURE 5.7 Stress field inside a simply supported beam at the vicinity of applied concen-
trated vertical load.
120 Finite Element Analysis of Solids and Structures
Using the definition of radial stress, σr, from equation (5.41), the resultant horizontal
force acting on one quadrant of the cylindrical surface “abc” of Figure 5.7 can be
defined as in the following:
/2 /2
2P P
0
r .sin r.d
0
cos sin d
(5.44)
Effects of this horizontal force on beam section at mid-span can be represented by an
axial stress (equation 5.45) and a bending stress (equation 5.46) as defined in the
following:
P (5.45)
x
bh
P h
. 2 y
x 6P y (5.46)
bh 3
12
bh2
Combining the stress components defined by equations (5.43), (5.45), and (5.46),
resultant normal stress on beam section at mid-span is given by the following:
3P 2h P (5.47)
x L y
bh3 bh
Substituting y = h/2 in equation (5.47), normal stress at point B on beam section
(Figure 5.7) is given by
3PL 4 h (5.48)
x 1 .
2bh 3 L
2
Term (3PL/2bh2) in equation (5.48) represents the stress value given by elastic
bending stress of beam, and the term in parenthesis on the right-hand side of this
equation represents a correction factor introduced by the stress effects of concen-
trated load application at the mid-span of beam. This correction factor value is sig-
nificant for beams with large depth and short span (h/L ratio – not small). Discussions
on more accurate prediction of beam internal stresses under concentrated applied
loads can be found in Ugural and Fenster (2012).
Stress analysis of solids, using the stress function approach discussed so far in
Chapters 3 and 5, has attempted to simplify the 3D stress fields to manageable two-
dimensional field problems. Analytical solution of a general three-dimensional elas-
ticity problem, involving equations (1.10, 2.5, 2.8–2.13, and 2.25), is usually not
attempted. Stiffness-based finite element method provides a very attractive and effec-
tive analysis technique by simply expanding the two-dimensional formulations to
three-dimensional space as discussed in the following Section 5.5.
Stress Analysis of Axisymmetric and General 3D Solids 121
5.5 FINITE ELEMENTS FOR STRESS ANALYSIS OF GENERAL
3D SOLIDS
Figure 5.8 shows a general solid subjected to arbitrary load and boundary conditions
in three- dimensional space. Each node of hexahedral solid finite element possesses
3 translational degrees of freedom. Assuming simple element geometry aligned with
coordinate directions x–y–z, shape functions for nodal deformations are given by the
following equations:
1 x y z
H1 1 1 1
8 a b c
1 x y z
H2 1 1 1
8 a b c (5.49)
1 x y z
H8 1 1 1
8 a b c
Internal deformation response of the element is interpolated from nodal displace-
ments by using the shape functions of equations (5.49):
u x,y H1.u1 H 2 .u2 . Hu .u8 Hi .ui
v x,y H1.v1 H 2 .v2 . Hu .v8 Hi .vi (5.50)
w x,y H1.w1 H 2 .w2 . Hu .w8 Hi .wi
Strain–displacement relationship matrix [B] in equation (2.6) is a 6 × 24 matrix of
shape function derivatives as defined in the following:
b y c z b y c z .. .. .. .. .. ..
1
B .. .. .. .. .. .. .. ..
8abc
.. .. .. .. .. .. .. ..
(5.51)
w1
v1
2 1
u1
2b
2a
3 4
5
6
2c
7 8
FIGURE 5.8 Displacement degrees of freedom in finite element model of three-dimensional
solids.
122 Finite Element Analysis of Solids and Structures
For general orientation of solid element in 3D space, shape functions in equations
(5.49) and the derivative terms in [B] matrix definition (equation 5.51) are re-written
by using iso-parametric coordinates (r,s,t) – similar to the two-dimensional descrip-
tion presented in Section 3.6. Relationship between Cartesian model coordinate sys-
tem (x, y, z), and the iso-parametric element definition in (r, s, t), is defined by
expanding the Jacobian from equation (3.70) into 3D system:
x y z
r r r
x y z
J (5.52)
s s s
x y z
t t t
where coordinates (x, y, z) inside an element are defined in terms of the nodal coor-
dinates by using the 3D interpolation functions written in terms of (r, s, t). Effects of
element shape on accuracy of stiffness matrix calculations, discussed in Sections 3.6
and 3.7, are equally applicable to 3D solid elements as well. It is important to main-
tain regular cubic shape of solid elements in 3D finite element models.
Stress–strain relationship matrix [C] for 3D stress state is defined by equation
(2.25). Stiffness matrix in equation (2.44) has 24 terms involving fourth-order coor-
dinate terms arising from pre- and post-multiplication of the [B] matrix terms from
equation (5.51). Numerical calculations of the stiffness terms in equation (3.84) can
produce exact integration of the stiffness values by using 2 × 2 × 2 Gauss integration
rule for the eight-node hexahedral finite element in Figure 5.9(a). ABAQUS input
data for this eight-node solid element example will be as follows:
*ELEMENT, ELSET=setname1, TYPE=C3D8
1, 7,8, 5, 6, 3, 4, 1, 2
Like the behavior of linear 2D solid elements discussed in Section 4.2, fully inte-
grated 3D solid elements, using node-to-node linear shape functions, tend to provide
artificially stiff response under bending and shear. Reduced one-point integration of
the eight-node linear solid element can be specified by choosing element type
“C3D8R” instead of fully integrated “C3D8”, but with the risk of experiencing zero
energy hourglass response mechanism with reduced integration formulation. Similar
to the discussion presented in Section 3.8 for improving the element deformation
behavior, higher order quadratic 3D solid brick element can be formulated by intro-
ducing mid-side nodes as shown in Figure 5.9(b). Stiffness properties of this 20-node
higher element can be calculated by using full 3 × 3 × 3 or reduced 2 × 2 × 2 Gauss
integration rule (ABAQUS element types C3D20 and C3D20R, respectively). Two-
dimensional formulations for constant stress–strain triangular element, presented in
Section 3.9, can also be expanded to define constant stress–strain 3D element of tet-
rahedral shape as shown in Figure 5.9(c) (ABAQUS element type “C3D4”). Stiffness
properties of this element can be calculated by using one-point integration rule while
Stress Analysis of Axisymmetric and General 3D Solids 123
2 1
3 4
5
6
7 8
(a) (b)
(c) (d) (e)
FIGURE 5.9 (a) Linear hexahedral solid element (C3D8), (b) Higher order 20-node brick
element (C3D20), (c) linear tetrahedral element (C3D4), (d) linear pyramid (C3D5), and (e)
higher order 10-node tetrahedron element (C3D10).
the higher order 10-node tetrahedron element of Figure 5.9(e) may use four points
for full integration or one point for faster reduced integration (ABAQUS element
types C3D10 and C3D10R, respectively).
5.6 THREE-DIMENSIONAL FEA MODEL PREPARATION AND
ELEMENT QUALITY CHECKS
Finite element analysis model of a three-dimensional solid body can be developed
from 3D CAD data by following the (a)–(g) model preparation steps described in
Section 4.4. For error-free 3D model preparation, HyperMesh session at the begin-
ning should be launched by selecting ABAQUS-standard 3D template if ABAQUS is
expected to be used in subsequent FEA analysis. The “3D” menu panel of HyperMesh
presents all necessary mesh generation task buttons, including the “elem types” for
pre-selecting the type of 3D elements (Figure 5.9) to be used during geometry dis-
cretization. Interactive mesh generation technique, described in Section 4.4 for 2D
solids, is also available to adapt the refinement of mesh grid in potential areas of high
stress concentration in 3D analysis problems. Quality check of 3D meshes requires
rigorous scrutiny of both external and internal sectional views of the solid bodies.
Appropriate element type selection, through “TYPE” declaration in ABAQUS input
file, will set the internal calculations in motion based on numerical process described
in Section 5.5. Part property definition, through “*SOLID SECTION” key word in
ABAQUS, will include associated material name. Loads can be applied directly as
concentrated loads at nodes along the global coordinate directions. Surface pressure
loads can also be applied on 3D faces of solid elements. ABAQUS will internally
convert the surface pressure loads to equivalent nodal loads for solving the matrix
equilibrium equation (1.5). Data structure for ABAQUS model of 3D solids looks
very similar to the example shown for 2D example in Section 4.4.7.
124 Finite Element Analysis of Solids and Structures
5.7 PRACTICE PROBLEMS – STRESS ANALYSIS OF AXISYMMETRIC
AND 3D SOLIDS
PROBLEM 1
Section 1.8 presented a stress analysis model of statically indeterminate solid beam,
subjected to a concentrated lateral load at mid-span (Figure 1.20). Finite element
analysis result, using the ABAQUS model file, “SOLIDBEAM_COARSE_MESH.
inp” (data file available at a website noted in the preface of this book), was compared
with elementary beam theory prediction in Practice Problem-1 of Section 1.8.
Calculate the normal stress, σx, at the bottom most point of beam mid-span (x = 100)
by using the analytical solution technique presented in Section 5.4. Compare this
result with those obtained in Section 1.8.
ROBLEM 2
P
Figure 5.10 shows the cut segment of a long steel pipeline subjected to an internal
fluid pressure of 0.01 GPa. Ignoring the external boundary conditions along the
length of the pipe, predict stress distribution through the thickness of pipe by using a
plane-strain finite element model on the cross-sectional plane perpendicular to the
member axis direction. Assume E = 210 GPa, ν = 0.3. Compare the finite element
analysis results with analytical predictions obtained by using the axisymmetric stress
analysis formulations presented in Section 5.2.
PROBLEM 3
Figure 5.11 shows a modified design of pipe segment from Figure 5.10, with an
added circumferential steel reinforcement of 10 mm thickness – integrally joined to
the main pipe body over a length of 100 mm. Using an axisymmetric finite element
model on the cross-sectional plane along the member axis direction, predict the effect
of added reinforcement on the stress distribution inside the pipe wall for internal
pressure loading of 0.01 GPa. Compare the predicted stress results with those
obtained in Problem-2.
Axis of
symmetry
y y
30
0.01
GPa
1100 1100
600
z 600
(a) Isometric view (b) cross-sec onal view
FIGURE 5.10 Cut segment of a steel pipeline (dimensions are in mm), subjected to uniform
internal fluid pressure of 0.01 GPa.
Stress Analysis of Axisymmetric and General 3D Solids 125
Axis of
symmetry
y y
30
0.01
500 GPa
10 mm thick circumferen al
100 reinforcement made of steel
1100
500
x
600
z 600
(a) Isometric view (b) cross-sec onal view
FIGURE 5.11 Locally reinforced pipe segment subjected to uniform internal fluid pressure
of 0.01 GPa.
PROBLEM 4
Figure 5.12 shows simplified geometry of a storage tank made of concrete (E = 40
GPa, ν = 0.2) sitting freely on frictionless base support. Assuming the tank is filled
with water, calculate the stress distribution inside the tank wall using an axisymmet-
ric finite element analysis model. What simplified hand calculations can be done to
verify the finite element analysis results?
Axis of
symmetry
y y
300
6000 6000
300
3000
z 3000
(a) Isometric view (b) cross-sec onal view
FIGURE 5.12 Simplified geometry of a free-standing open cylindrical tank of uniform wall
thickness (dimensions in mm).
6 Deformation Analysis
of Beams for Axial,
Bending, Shear, and
Torsional Loads
SUMMARY
Beams are long slender members that carry transverse loads by producing resistance
to bending deflection of member axis. The standard solid mechanics description of
bending stress distribution in beams is reviewed in Section 6.1. The shear stress dis-
tribution, associated with the bending response of beams, is discussed in Section 6.2.
Section 6.3 analyzes the transverse normal stress response that is generally ignored
in the analysis of long slender beams. Section 6.4 presents a detailed review of the
torsional stress responses of prismatic members. The membrane analogy technique
for calculating the torsional response properties of both open and closed section
beams has been discussed in detail. The combined stress response of beams for axial,
bending, shear, and torsional load effects is calculated from the simple superposition
of the component values (discussed in Section 6.5). Section 6.6 presents the well-
known Euler–Bernoulli beam theory that relates the lateral deflection of beam axis to
bending deformation mode without considering the effect of transverse shear defor-
mation of the material. Limitation of the beam bending deflection analysis, based on
Euler–Bernoulli beam theory, is also discussed in this section. Stress analysis of
curved beam profiles is presented in Section 6.7 for the completeness of the solid
mechanics-based beam analysis technique.
A key component of this chapter, i.e. finite element stiffness formulation for beam
resistance mechanisms related to axial, bending, and torsional deformation modes, is
described in Section 6.8. The axial and torsional stiffness properties of prismatic
beam elements are derived by using independent linear interpolations of the associ-
ated nodal DOF. The internal transverse deformation of beam is calculated with
cubic interpolations of nodal DOF associated with transverse and rotation deforma-
tion modes. The relationship between internal strain and transverse deformation is
defined directly based on the Euler–Bernoulli beam theory. The resulting beam ele-
ment stiffness properties, thereby, turn out to be the same values available from direct
stiffness analysis of straight-profile beam structures. The stiffness matrix of two-
node prismatic beam element is, thus, directly calculated based on section properties
and element length – without having to use the numerical integration technique pre-
sented in equation (3.84). This beam element formulation, often referred to as Euler–
Bernoulli beam element, provides accurate results for slender beams (with
127
128 Finite Element Analysis of Solids and Structures
length-to-depth ratio greater than 20). The contribution of shear deformation in the
material becomes significant for smaller span-to-depth ratios. Beam stiffness formu-
lations, including shear deformation in the material, are presented in Section 6.9
based on linear interpolation of nodal response variables. The use of linear shape
function makes the structural response artificially stiff. Corrective actions to reduce
shear locking have also been discussed in Section 6.9. Modeling options of different
beam element types, with specific references to the ABAQUS element library, have
been discussed in Section 6.10. Finally, Section 6.11 presents practice problems on
stress analysis of beam-type structural members.
6.1 BENDING STRESSES IN A BEAM
Figure 6.1 shows the deformation profile of a beam under pure bending load.
Following the Euler–Bernoulli theory of beam deflections (Popov 1978), normal
stress on the beam cross-section, made of single homogeneous material, can be
expressed as a linear function of distance y from beam’s neutral axis while all other
stress components are zero. The stress field inside homogeneous beam for pure bend-
ing load can be described by the following equation:
x y, y z xy xz yz 0 (6.1)
where μ is a constant. In absence of body forces (Fx=Fy= Fz=0), stress functions of
equation (6.1) readily satisfy the stress equilibrium equations (1.10). These stress
functions, with the stress–strain relationships of equations (2.25), also satisfy the
deformation compatibility equations (2.8)–(2.13). Stress functions (equation 6.1),
with the following boundary conditions (equations 6.2), thus represent the exact
solution for pure bending problem of Figure 6.1:
.dA . y.dA 0, y. .dA y .dA M
2
x x (6.2)
Integral in the second expression of equation (6.2) represents the moment of inertia
(I) of beam cross-section about the axis of bending rotation. Using the expression for
constant μ from equation (6.2), bending stress on beam cross-section (equation 6.1)
can be re-written in the following form (equation 6.3):
x
z
FIGURE 6.1 Bending stresses on a beam section.
Deformation Analysis of Beams 129
Strain Stress distribu on
x =E1. x
M
E1
x =E2. x
E2
x
FIGURE 6.2 Bending stress distribution on a beam section made up of two dissimilar mate-
rials with elastic moduli of E1 and E2.
M.y
x (6.3)
I
This is the familiar bending stress formula for a straight beam subjected to pure
bending load condition. Assumption of linear bending stress variation over the depth
of beam section holds true for beams made of single homogeneous material.
Composite beams, constructed by continuous bonding of dissimilar materials, do not
experience linear stress variation through beam depth. In elastic bending response
analysis of composite beams, plane sections of beams are assumed to remain plane
during bending deformation resulting in linear variation of strain through beam depth
(Figure 6.2). Using Hooke’s law, bending stresses in the dissimilar materials are
calculated by using the material-specific elastic modulus as shown in Figure 6.2. For
analysis simplicity, composite material construction is replaced with a single mate-
rial section by scaling the lateral section dimension of a substituted material area
with factor (E2/E1), where E2 is the elastic modulus of substituted material and E1 is
the modulus of substituting material. Moment of inertia I is calculated from the
hypothetical cross-sectional dimensions of equivalent single material geometry, and
the bending stress values are calculated by using the familiar equation (6.3).
6.2 STRESSES DUE TO TRANSVERSE SHEAR
When a beam is bent by transverse loads, usually both a bending moment (M) and a
shear force (V) act on beam cross-section. Figure 6.3 shows a beam segment of length
dx subjected to bending moments M and M + dM, respectively, on the left and right
sides of the beam element. Normal stresses caused by these bending moments can be
defined by using the bending stress formulation given in equation (6.3). Considering
the bending stresses caused by the right-side bending moment M + dM, the resultant
normal force acting on the area Ā, representing part of the beam section above a dis-
tance y from the neutral axis, can be expressed as in the following equation (6.4):
130 Finite Element Analysis of Solids and Structures
M M+dM
yx
y
x
dx
FIGURE 6.3 Beam subjected to bending moment variation over a segment of length dx.
M dM
A
x .dA
A I y.dA
(6.4)
Similarly, considering the bending stresses caused by the left-side bending moment
M, the resultant normal force on the area Ā on the left-side of the beam segment is as
follows:
M
.dA
A
x y.dA
A I
(6.5)
Difference between the normal forces from two sides (equations 6.4 and 6.5) will
cause internal shear stresses (τyx) over the area (b.dx), where b is the width of the
beam section at distance y from the neutral axis. Assuming a uniform distribution of
shear stresses over the area, the overall equilibrium of the part of beam segment
above the position y can be expressed by the following equation:
M M dM
yx .b.dx
y.dA
A I
A I y.dA 0
(6.6)
Re-arranging the terms in equation (6.6) gives:
dM 1
yx .
dx Ib y.dA
A
(6.7)
The variation of bending moment over beam segment length dx represents the shear
force V acting on the beam section (Popov 1978). And the integral term on the right
side of equation (6.7) represents the first moment of the beam section area Ā about
the neutral axis. Writing Q for that integral term, equation (6.7) can be re-written in
the following form:
Deformation Analysis of Beams 131
VQ (6.8)
yx xy
Ib
where the equality condition, τyx = τxy, comes from the property of shear stress distri-
bution in three-dimensional bodies as discussed in Section 1.5. Equation (6.8), thus,
explicitly describes the relationship between shear force V acting on a beam section
and the shear stress τxy measured at a distance y from the neutral axis. For a rectangu-
lar beam section of width b and depth 2h, the shear stress at a distance y from the
neutral axis can be obtained from equation (6.8) by substituting I = b(2h)3/12, and Q =
b(h2 – y2)/2:
xy
3 V
.
4 bh3
h2 y 2
(6.9)
Equation (6.9) indicates a parabolic distribution of shear stress over the beam depth,
with maximum value occurring at the neutral axis (y = 0) given by:
3 V
max . (6.10)
2 2bh
where 2bh is the cross-sectional area of the rectangular beam section. Maximum
shear stress is, thus, 1.5 times the average shear stress for a rectangular beam section.
Equation (6.8) can be used to predict shear stress distribution on beam cross-sections
of general shape. Figure 6.4(a) shows an I-section beam. Shear stress at a distance y
from the neutral axis, for an applied shear force V, can be expressed as follows:
V b 2 2
xy
It 2
t
h h1 h12 y 2
2
(6.11)
h1
y
2h
t h1
b
avg
(a) (b)
FIGURE 6.4 (a) Cross-section of an I-section beam; (b) shear stress distribution over beam
cross-section.
132 Finite Element Analysis of Solids and Structures
where t is the width of the beam section where shear stress is measured. For the
I-section beam, t will be replaced by b when stress is measured inside the flange –
resulting in a very small shear stress value in the wide flange area. Figure 6.4(b)
shows the distribution of shear stress over the beam section following the parabolic
function of equation (6.11). In typical engineering practice, small shear stresses in
the flanges are ignored, and the average shear stress is calculated by assuming a uni-
form distribution over the extended depth of beam web:
V (6.12)
avg
2th
Average shear stress distribution over the beam web, shown by the dotted line in
Figure 6.4(b), provides a reasonable estimate for actual engineering decisions instead
of using the more rigorous estimate given by equation (6.11).
6.3 TRANSVERSE NORMAL STRESS IN A BEAM
Figure 6.5 shows a cantilever beam subjected to a uniformly distributed transverse
load of p per unit length. Assuming the beam section to be rectangular shape with
dimension b in the normal direction of view plane, the equilibrium state of a differ-
ential element in coordinate y-direction (Figure 6.5(c)) can be expressed as follows:
h
xy
y .b.dx
x .dx .b.dy
y
(6.13)
Writing the shear force in the section of cantilever beam at a distance x from the left
end as, V = p.x, shear stress in the beam can be obtained from equation (6.9) as
follows:
xy
3 p. x 2
.
4 bh3
h y2
(6.14)
y
p: load per unit length p
y
x 2h + .
L dx dx
(a) (b) (c)
FIGURE 6.5 (a) A cantilever beam subjected to a uniform transverse load of p per unit
length; (b) transverse loads on segment of length dx; and (c) stresses in y-direction on a dif-
ferential element of length dx and height dy.
Deformation Analysis of Beams 133
Substituting equation (6.14) into equation (6.13), and upon integration of the right-
hand side, the following expression is obtained for the transverse normal stress
(Ugural and Fenster 2012):
p 1 3 y 1 y
3
y (6.15)
b 2 4 h 4 h
The highest value of σy is obtained at the upper surface (y = h):
p (6.16)
ymax
b
Maximum value of bending stress on the beam cross-section, given by equation
(6.3), is expressed as
xmax
p.L.L / 2 .h 3 . p . L 2 (6.17)
b. 2h /12
3
4 b h
Ratio between the maximum transverse normal stress (equation 6.16) and the maxi-
mum bending stress (equation 6.17) is given by the following:
2
ymax 4 h (6.18)
.
xmax 3 L
For typical slender beams with a proportion of L > 20h, equation (6.18) indicates a
very small ratio between transverse normal stress and bending stress. It is, thus, cus-
tomary to assume σy ≈ 0 in slender beams subjected to transverse loading.
6.4 TORSIONAL RESPONSE OF A BEAM
Deformation response of a prismatic circular member under torsional load is gener-
ally assumed to possess three characteristics: (i) plane sections, perpendicular to the
longitudinal axis of the member, remain plane after the application of torque; (ii)
shearing strain varies linearly from zero at the center to maximum on the outer sur-
face; and (iii) material is homogeneous and obeys Hooke’s law (τ = G.γ). Based on
the second and third assumptions, the torsional stress, at a distance r from the center
point of the circular section in Figure 6.6, can be expressed as follows:
r (6.19)
. max
where ρ is the radius of the circular section, and τmax is the maximum torsional stress
on the outer surface. Considering the equilibrium between applied torque T and the
resultant of torsional stresses τ, we obtain:
134 Finite Element Analysis of Solids and Structures
A
max
r
max dA
B
T
L B’
FIGURE 6.6 Torsional stress and angular rotation in a prismatic circular section member.
T r. .dA (6.20)
where dA is a small area at a distance r from the center of the circular cross-sectional
area. Combining equations (6.19) and (6.20) gives:
max 2 (6.21)
T . r .dA
The integral on the right-hand side represents a property of member cross-section
– commonly known as the polar moment of inertia. Substituting J for the integral
expression, and after rearranging the terms, the following well-known expression is
obtained for torsional stress in a circular section prismatic member:
T (6.22)
max
J
Using Hooke’s law, the maximum shear strain can be obtained from equation
(6.22):
max T (6.23)
max
G JG
Relating the shear strain with the angle of twist Θ (Figure 6.6):
(6.24)
max .
L
Combining equations (6.23) and (6.24), we obtain:
T JG (6.25)
L
Deformation Analysis of Beams 135
y
x
FIGURE 6.7 Torsional deformation of a rectangular beam section.
The term JG is commonly referred to as the torsional rigidity of a member.
Relationships (6.19)–(6.25) are valid for circular section members (both solid and
hollow). However, plane sections of noncircular members, e.g. the rectangular sec-
tion in Figure 6.7, do not remain plane during torsional deformation (Timoshenko
and Goodier 1982). Equation (6.25), defining the relationship between applied torque
and angular twist response, can be generally used for noncircular section members,
after taking account of warpage effects in the definition of section property J. Stress
function-based semi-inverse method, used by Saint Venant (Timoshenko and Goodier
1982), provides an effective solution to this problem by assuming that the general
torsional deformation is a superposition of section rotation and section warpage
effects. Considering the coordinate origin at the left end of beam in Figure 6.7, rota-
tion (θ) of cross-section at a distance z from coordinate origin can be related to the
deformations in x–y plane as follows:
u y. v x. (6.26)
Warpage-induced deformation of the member section is expressed by a
function ψ:
w . x,y (6.27)
Displacement field equations (6.26 and 6.27) lead to the following expressions for
strains in the beam:
x y z xy 0
w u (6.28)
xz . y
x z x
w v
yz . x
y z y
Assumptions for warpage-induced displacement field, thus, lead to no distortion
in x–y plane representing member cross-section. Using Hooke’s law, stress compo-
nents corresponding to the strain values of equations (6.28) are obtained as follows:
136 Finite Element Analysis of Solids and Structures
x y z xy 0
(6.29)
xz G . y
x
yz G . x
y
Equations (6.29) represent a pure shear problem defined by τxz and τyz. Substituting
the expressions from equations (6.29) into the stress equilibrium equations (1.10),
and ignoring the body force terms (Fx = Fy = Fz = 0):
xz yz xz yz (6.30)
0 0 0s
z z x y
The first two conditions in equation (6.30) are satisfied since τxz and τyz are inde-
pendent of z (equations 6.29). The third condition of equation (6.30) can be satisfied
by defining the stress components τxz and τyz in terms of a yet-to-be-determined stress
function ϕ:
(6.31)
xz yz
y x
Substituting the expressions from equations (6.31) into (6.29), we obtain:
G . y
y x (6.32)
G . x
x y
Differentiating the first expression of equation (6.32) with respect to y, the second
with respect to x, and adding the 2nd to the first, the following differential equation
is obtained (where F = –2Gθ):
2 2 (6.33)
F
x 2 y 2
Stress function ϕ assumed for a given torsional problem must satisfy the compat-
ibility condition (equation 6.33) as well as the stress boundary conditions (equations
1.14). The first and second expressions in equations (1.14) are readily satisfied by the
stress components (equations 6.29). For zero applied boundary stress in z-direction
(pz = 0), the third expression of equations (1.14) reduces to:
xz .l yz .m 0 (6.34)
Deformation Analysis of Beams 137
. .
yz
dy ds
xz
dx
x
FIGURE 6.8 Boundary condition for torsional stress distribution on a member section.
Using the stress component definitions from equations (6.31), and the direction
cosine definitions at a point on the boundary, ℓ = dy/ds and m = –dx/ds (Figure 6.8),
equation (6.34) is re-written as
dy dx d (6.35)
. . 0
y ds x ds ds
Equation (6.35) shows that the derivative of stress function ϕ on the boundary is zero
– implying that the torsional stress must follow the tangential direction on the bound-
ary. Now considering the overall equilibrium between applied torque T and the tor-
sional stresses on a member cross-section, we have:
T
x. yz y. xz dx.dy
x. x y. y dx.dy (6.36)
Integrating equation (6.36) by parts, and imposing the condition ϕ = 0 at the
boundary (Timoshenko and Goodier 1982), gives:
T 2
.dx.dy (6.37)
Equation (6.37) implies that the magnitude of torque T is equal to twice the vol-
ume under stress function ϕ. Analytical solution of equations (6.33), (6.35), and
(6.36) is tedious. An alternative technique is to rely on the similarity between the
torsion problem and membrane deflection problem (Figure 6.9), and to use the mem-
brane deflection solution as a surrogate for the torsional response. Under the normal
internal pressure p, the responses of membrane in Figure 6.9 are represented by
deflection z and a boundary traction of S per unit length. Considering the membrane
deflection in the x–z plane, slopes at the two ends of a differential element are
described by equation (6.38):
138 Finite Element Analysis of Solids and Structures
y
s
s
x
(a)
z
Membrane
p
x
(b)
FIGURE 6.9 (a) Cross-sectional area of a solid member in the x–y plane with an inflated
membrane over the same area and (b) x–z section view of the membrane under internal pres-
sure p.
z z 2 z (6.38)
, d .dx
x x x 2
Similarly, considering a section of inflated membrane in the y–z plane gives:
z z 2 z (6.39)
, d .dy
y y y 2
Using the equations (6.38 and 6.39) for membrane slope, the equilibrium state in
normal direction of differential membrane element can be written as
z z 2 z z
S.dy . S.dy 2 .dx S.dx .
x x x y
z z2
S.dx 2 .dy p.dx.dy 0 (6.40)
y y
Upon simplification of equation (6.40), membrane deflection equation is obtained as
follows:
2z 2z p (6.41)
2
x 2
y S
Membrane deflection equation (6.41) is similar to the compatibility equation for
torsional problem (equation 6.33), with z analogous to ϕ, 1/S to G, and p to 2θ.
Solution of membrane deflection problem (experimental or analytical) is commonly
used to determine the solution for torsional problem. A very effective use of the
membrane analogy is in the torsion problem solution of thin-walled members
Deformation Analysis of Beams 139
y
t
(a) x
x
(b) h = membrane deflec on
x
Deflected volume of membrane = Am *h
Am
average perimeter of the thin-wall sec on
FIGURE 6.10 (a) Thin-wall closed section of a member subjected to torsion and (b) mem-
brane analogy model of thin-wall section.
(Figure 6.10). Assuming uniform wall thickness for the arbitrary section geometry in
the figure, average torsional stress flowing through the wall is given by the average
slope of membrane over the thickness of thin-walled member:
z h (6.42)
membrane slope,
x t
Membrane deflection, h, in equation (6.42) is defined by
deflected volume of membrane T /2 (6.43)
= h =
membrane surface area Am
where T is the magnitude of applied torsion equal to twice the volume of deflected
membrane. Combining equations (6.42) and (6.43), average torsional stress flowing
through the thickness of thin-walled closed section member in Figure 6.10 is given
by the following equation (6.44):
T (6.44)
2 Am t
Equation (6.44) presents an elegant solution for the torsional stress in a member
of arbitrary cross-section. The powerfulness of membrane analogy becomes more
evident when a thin-walled open section, shown in Figure 6.11, is considered.
Ignoring the curvature in the y-direction, the membrane deflection equation (6.41)
can be re-written in the following form for narrow member section:
2z p (6.45)
x 2
S
140 Finite Element Analysis of Solids and Structures
s s
b x p x
t
t
(a) (b)
FIGURE 6.11 (a) Torsional stresses in a thin-wall member and (b) membrane analogy for
the torsional response.
Integrating equation (6.45) twice, and inserting the boundary conditions of dz/dx
= 0 at x = 0 and z = 0 at x = t/2, membrane deflection equation is obtained as
follows:
1 P t
2
z . . x 2 (6.46)
2 S 2
Volume under the membrane surface is given by
1 p
V
z.dx.dy 12 . S .bt (6.47)
3
Replacing the membrane analogous terms in equation (6.47), torsional resistance of
member (equation 6.37) can be obtained as
1
T 2.V .bt 3 .G (6.48)
3
Writing the total angle of twist over the member length as, Θ = θ*L, equation (6.48)
can be re-written in the following-form:
T 1 3 G J e .G
.bt . (6.49)
3 L L
Deformation Analysis of Beams 141
b3
t3
r
r
b2 t2
t
t1
b1
FIGURE 6.12 Thin-walled open section members.
Comparing equation (6.49) with (6.25), the term Je represents the effective polar
moment of inertia of a narrow rectangular section that is analogous to the polar
moment inertia of a circular cross-section. The value of Je for thin-walled general
open section members (Figure 6.12) can be obtained by extending the definition from
equation (6.49) to the following general form:
1 (6.50)
J e bt 3
3
Angular twist of member under torsion can be obtained from equation (6.49) by
using the effective polar moment of inertia definition from equation (6.50). Average
torsional stress flowing through thin-wall section is given by equation (6.44) where
effective area is given by: Σbt. For the open circular arc section of Figure 6.12(a),
effective polar moment of inertia is Je = (2/3)πrt3. Using the integral definition in
equation (6.21), J for a closed circular section is: 2πr3t. Ratio between these two
expressions is 3.(r/t)2. For a thin-wall tube section, with r/t = 20, torsional rigidity of
the closed section turns out to be 1200 times the value of an open section. The expres-
sion for Je in equation (6.50), derived based on the membrane analogy of Figure 6.11,
is applicable to thin narrow sections. Analysis method can be extended to consider
bi-directional curvature of a membrane deflection simulating the torsional response
of a general rectangular section (Timoshenko and Goodier 1982). Figure 6.13 shows
that the expression for polar moment of inertia of a general rectangular section
approaches that of equation (6.50) as the proportion of rectangular section approaches
that of a thin narrow section. Finite element calculation of stiffness properties
requires the evaluation of member section property Je. Membrane analogy-based
definition of Je, explained in the above, is directly used to define the torsional stiff-
ness of a member of arbitrary cross-section by using equation (6.49).
142 Finite Element Analysis of Solids and Structures
Polar moment of iner a of
solid rectangular sec on:
= .
b/a 1.0 0.667 0.5 0.1 0
0.141 0.196 0.229 0.312 0.333
FIGURE 6.13 Polar moment of inertia of a solid rectangular section.
6.5 BEAM RESPONSE TO COMBINED LOAD EFFECTS
Beam response to individual actions of bending, transverse shear, and torsion has been
described in Sections 6.1–6.4. Stress response of a beam under axial load is similar to that
of a bar or truss member – as described by equation (2.36). Resistance mechanisms
against axial and bending modes of deformation are assumed to be uncoupled (indepen-
dent of each other). Normal stresses acting on a beam cross-section, calculated separately
for axial and bending load effects, are superposed to get the resultant stress values. Figure
6.14 shows an eccentric normal load P1 acting on a beam cross-section. Effects of this
force on beam axis can be expressed by the following three force components:
Fx P1 M y P1.z1 M z P1. y1 (6.51)
Normal stresses caused by bending moments My and Mz can be calculated separately
by using equation (6.3), and be combined with that caused by axial force Fx to get the
-z2
-z 1 P1
y1
z P2 x
FIGURE 6.14 Eccentric normal load acting on a beam section.
Deformation Analysis of Beams 143
resultant normal stress at any point on the beam cross-section. Effects of transverse
load P2 acting on beam cross-section of Figure 6.14 can be expressed by the follow-
ing load components:
Vy P2 Tx P2 .z2 M z P2 . x (6.52)
Bending stress acting on a beam cross-section, at a distance x from the loading
plane of cantilever beam, can be obtained from equation (6.3). Shear stresses caused
by transverse shear force Vy can be calculated by using equation (6.8), and those
caused by torsion Tx can be calculated by using equations (6.19) and (6.22). Torsion
value Tx in equation (6.52) is calculated by multiplying the transverse load P2 with z2
– the offset distance of load action from the vertical axis of symmetry y. Effective
torsion calculation, for member section without a vertical axis of symmetry, requires
the calculation of “shear center” of a beam section (Popov 1978). Figure 6.15 shows
a transverse load P acting through the centroid of a non-symmetric beam section, and
the associated bending moment Mz acting on the section about z-axis. Bending stress
(σx) and the internal shear stress (τzx) on a segment of the beam flange are shown on
a separate free-body diagram. The flow of complementary shear stress (τxz) through
beam flange, as well as the flow of stress (τxy) through beam web, is shown on the
beam cross-section. As evident from the directions of stress flow through the
C-section beam, shear stresses produce a twisting effect about the member axis pass-
ing through the centroid. In order to prevent twisting effect on the beam, the trans-
verse force P should be applied through the point S so that net torsional effect on the
beam section will be “zero”. Point S is the “shear center” of beam section – a point
away from the centroid of non-symmetric section such that transverse force applied
through that point will produce zero twist about the beam axis. Considering the equi-
librium between external and internal effects, the distance to shear center is given by
the following equation (6.53) where parameters refer to the dimensions of C-section
beam in Figure 6.15:
x
zx
xz
xy Mz
P
z S
e
2h
t2
t1 x
b1
FIGURE 6.15 Transverse load and bending moment on a non-symmetric beam section and
the resulting shear flow through the beam section.
144 Finite Element Analysis of Solids and Structures
3 b12t1
e . (6.53)
2 ht2 3b1t1
Shear center can, thus, be located by using the dimensions of beam cross-section.
In finite element models, beam element properties generally refer to the centroid of
section. Effects of surface pressure loads are represented by equivalent loads, moment
and torsion referring to the centroidal axis of member. The concept of shear center is
not directly used in finite element formulations. However, the concept is reviewed
here because of its importance in beam design to minimize the potential torsional
effects produced by transverse load applications.
6.6 ELASTIC BENDING DEFLECTION OF BEAMS
Combining equations (6.1) and (6.3) with the generalized Hooke’s law for stress–
strain relations (equations 2.18 and 2.24), the following expressions are obtained for
the strain responses of the beam under pure bending load:
M.y M.y (6.54)
x , y z . , xy xz yz 0
EI z EI z
where ϑ is Poisson’s ratio and EIz is the flexural rigidity. Overall axial deformation of the
beam under bending load is assumed zero. From the beam deflection profile of Figure
6.16, relative rotation of beam section can be related to normal strain as follows:
x dx
d (6.55)
y
Combining equation (6.55) with the first part of equation (6.54), the following equa-
tion is obtained between beam axis rotation and applied moment M:
d M
(6.56)
dx EI z
y
.dx
x
d
x
dx
FIGURE 6.16 Pure bending deformation of a beam element.
Deformation Analysis of Beams 145
Slope of the deflected beam axis at any point x can be expressed as derivative of
the transverse deflection v as follows:
dv (6.57)
dx
Combining equations (6.56) and (6.57), the following well-known expression is
obtained between the beam deflection v and applied bending moment M, where sym-
bol I is used instead of Iz for the sake of simplicity:
M d 2v (6.58)
=
EI dx 2
Double integration of equation (6.58) with respect to coordinate variable x gives the
following general expression for beam deflection under pure bending load effect:
EI .v x M .dx .dx c1. x c2 (6.59)
Here, c1 and c2 are integration constants. Equation (6.59) can be used to determine
deflection response from known moment distribution function and related bound-
ary conditions. For the cantilever beam example of Figure 3.5, moment M in
Equation (6.59) can be replaced by P.x, where P is the applied load at the left end
of cantilever beam. Equation (6.59) for that specific example, thus, takes the fol-
lowing form:
P. x 3
EI .v x c1. x c2 (6.60)
6
Boundary conditions of zero displacement and zero rotation at the right-end sup-
port (x = L) of beam in Figure 3.5 lead to the following expressions for the constants
c1 and c2:
v P.L2 P.L2 (6.61)
EI . c1 0 c1
x 2 2
P.L3 P.L3 (6.62)
EI .v c1.L c2 0 c2
6 3
Substitution of the expressions for constants c1 and c2, from equations (6.61) and
(6.62) into equation (6.60), gives the following expression for deflection of axis (y =
0) of the cantilever beam in Figure 3.5:
P. x 3 P.L2 . x P.L3
v x (6.63)
6 EI 2 EI 3EI
146 Finite Element Analysis of Solids and Structures
Deflection equation (6.63) for beam axis, derived from bending deformation only,
matches exactly with first three terms of equation (3.56) that has been derived from
plane-stress functions for cantilever beam problem. The last term in equation (3.56),
representing the beam axis deflection due to shear deformation of the material, is obvi-
ously not captured by the bending deflection equation (6.63). In the cantilever beam
example of Figure 3.5, shear deformation in the material will accumulate over the
length of cantilever, thus, resulting into a higher deflection value compared to that given
by equation (6.63). Deflection at the left end (x = 0) of cantilever beam axis, considering
both bending and shear deformation in the material, is obtained from equation (3.56):
PL3 Ph2 L (6.64)
v x 0,y 0
3EI 2GI
Ratio between the two terms on the right-hand side of equation (6.64), second term
representing the shear deformation and the first one representing the bending defor-
mation in material, leads to the following metric for the measurement of a beam’s
slenderness:
2 2 2
Ph2 L / 2GI 3 E h 3 2h 2h
. . 1 . (6.65)
PL3 / 3EI 2 G L 4 L L
where 2h is the depth of rectangular beam section and L is the span of beam in Figure
3.5. Evidently, for long slender beams, L > 10*(2h), equation (6.65) indicates a very
low contribution of shear deformation (<1%) to the overall beam deflection. Euler–
Bernoulli beam theory, considering bending deformation only, provides an accurate
estimate for the overall deflection for long slender beams. However, for short-span
deep beams, ratio in equation (6.65) is not negligible – meaning that material shear
deformation needs to be included in beam deflection calculations. Idea of shear cor-
rection factor application to the Euler–Bernoulli’s beam deflection estimate is attrib-
uted to early 20th-century work by Timoshenko and Ehrenfest – commonly known as
“Timoshenko Beam Theory” (Wikipedia.org 2020). Stiffness properties of beam ele-
ments, with and without material shear deformation effects, are discussed in Sections
6.8 and 6.9.
6.7 STRESS ANALYSIS OF CURVED BEAMS
Analysis techniques described thus far assumed straight profiles of beam axis.
However, the basic assumptions of beam response mechanisms can be extended to
consider initially curved profile of beam with the following strict assumptions:
1. Beam axis follows single-curvature profile both before and after bending
2. Beam cross-section at any point possesses an axis of symmetry pointing
towards the center of curvature of the beam axis
3. Beam cross-sections originally normal to the beam axis remain so after bend-
ing. Deformation due to transverse shear is not considered in curved beam
analysis.
Deformation Analysis of Beams 147
ro ri R
e Cross-section with
axis of symmetry
FIGURE 6.17 Pure bending deformation of a curved beam element.
Figure 6.17 shows an initially curved beam profile with radius of centroidal axis
ř, radius of outer fiber ro, and that of inner fiber ri. Different initial lengths of inner
and outer fibers of curved beam profile will produce different strain amplitudes at the
extreme fibers – implying that neutral axis will not pass through the centroid of beam
section. Distance of the neutral axis from center of curvature is given by the follow-
ing equation (Ugural and Fenster 2012):
A (6.66)
R
dA
r
where dA is an infinitesimal area on beam cross-section at a distance r from the cen-
ter of curvature O, and A is the area of beam cross-section. The distance between
centroidal axis and neutral axis is equal to:
(6.67)
e=r –R
The tangential stress at distance r from the center of curvature is given by the fol-
lowing equation:
M R r
(6.68)
Aer
Curved beam formula (equation 6.68) for the calculation of bending stress is com-
monly known as Winker’s formula (developed by E Winkler 1835–1888). Radius of
neutral axis R (equation 6.66) for standard beam section shapes is often available in
reference literature (Ugural and Fenster 2012). Equation (6.68) provides a simple but
reasonably accurate method for the calculation of bending stress in a curved beam.
Application of the method is, however, limited to simple example cases meeting the
restrictive assumptions listed earlier in this section. Review of curved beam analysis
is included here for the sake of completeness of theory of elasticity approaches for
beam response analysis. General structural geometries subjected to combined axial,
148 Finite Element Analysis of Solids and Structures
transverse and bending load effects can be accurately modelled and analyzed by
using 3D finite elements described in Section 5.5.
6.8 STIFFNESS PROPERTIES OF PRISMATIC EULER–BERNOULLI
BEAM ELEMENTS
A “prismatic” member is one in which cross-section properties (shape, dimensions,
material, etc.) do not change along the member length. For analysis efficiency, a
prismatic beam can be represented by a skeletal (line) element (Figure 6.18). It is
assumed that the cross-sectional dimensions are smaller than the dimension along
member length. A general beam element in 3D space will have 12 degrees of free-
dom (6 at each end). Figure 6.18(b) shows 6 DOF at each end – aligned with global
coordinate directions x–y–z. For member resistance calculations, deformations in
beam local coordinate directions (Figure 6.18(c)) need to be derived from the global
displacement DOF. Coordinate transformation relationships between the global and
local DOF can be expressed as follows where (l,m,n) are direction cosines defined in
Figure 6.18:
u1 l1 m1 n1 0 0 0 0 0 0 0 0 0 U1
u2 l2 m2 n2 0 0 0 0 0 0 0 0 0 U2
u3 l3 m3 n3 0 0 0 0 0 0 0 0 0 U3
u4 0 0 0 l1 m1 n1 0 0 0 0 0 0 U4
u5 0 0 0 l2 m2 n2 0 0 0 0 0 0 U5
u6 0 0 0 l3 m3 n3 0 0 0 0 0 0 U6
(6.69)
u7 0 0 0 0 0 0 l1 m1 n1 0 0 0 U7
u8 0 0 0 0 0 0 l2 m2 n2 0 0 0 U8
u9 0 0 0 0 0 0 l3 m3 n3 0 0 0 U9
u 0 0 0 0 0 0 0 0 0 l1 m1 n1 U10
10
u11 0 0 0 0 0 0 0 0 0 l2 m2 n2 U11
u12 0 0 0 0 0 0 0 0 0 l3 m3 n3 U12
Writing the vectors and matrix of equation (6.69) in compact form gives:
ui T Ui (6.70)
Matrix [T] is orthogonal, i.e. its inverse is equal to its transpose (Cook et al. 1989).
Upon calculation of the element stiffness matrix [k] in local coordinate system,
matrix [T] can be used to transform it to global coordinate system:
T
K T k T (6.71)
Deformation Analysis of Beams 149
U8 U11
U7
Y U12
(a) U10
U9
Z
L
U2
U5 (b)
U1
U6
U4
U3
Direction cosines of axes:
(c) X Y Z
x
x l1 m1 n1
y l2 m2 n2
z l3 m3 n3
FIGURE 6.18 (a) A general beam element in X–Y–Z global reference system; (b) skeletal
line element representation of the beam with DOF in global directions; and (c) DOF in element
local reference system (x, y, z) and the direction cosines of axes.
x
L
(a)
(b)
FIGURE 6.19 Local deformation of a beam element: (a) DOF associated with axial defor-
mation mode and (b) DOF associated with torsional deformation mode.
Local stiffness matrix [k] is assembled from the superposition of beam resistances for
axial, shear, bending, and torsional deformation modes that have been discussed ear-
lier. Axial deformation of a beam element can be calculated from DOF u1 and u7
(Figure 6.19(a)) by using linear interpolation functions shown in Figure 2.11.
Resistance to that axial deformation is defined by the member cross-sectional area
(A) and elastic modulus of material (E). Stiffness matrix for beam resistance to axial
deformation is identical to that of a truss or cable element defined by equation (2.57):
150 Finite Element Analysis of Solids and Structures
AE AE
L
k axial L (6.72)
AE AE
L L
where L is the element length. Linear interpolation functions of Figure 2.11 can also
be used to calculate the torsional deformation of beam caused by angular twists u4
and u10 at the two ends (Figure 6.19(b)). Following the analytical steps, described in
Section 2.8 for axial deformation mode, the stiffness matrix of beam element associ-
ated with torsional resistance mechanism can be defined by extending the definition
from equation (6.49):
JG JG
L
k torsional L (6.73)
JG JG
L L
where G is the shear modulus of material and J is the effective polar moment of iner-
tia of beam cross-section (suffix e is dropped for the sake of simplicity). Shear modu-
lus for homogeneous isotropic material is calculated from Young’s modulus and
Poisson’s ratio by using equation (2.23). Polar moment of inertia of beam section is
calculated by using the case-specific formulations described in Section 6.4.
Bending deformation of a 3D beam under transverse loading can occur in two
coordinate planes – (x,y) and (x,z). Figure 6.20(a) shows the general deformation
profile in (x,y) plane. The transverse displacement (v) at any point on the beam axis
can be expressed in terms of the nodal DOF (u2, u6) and (u8, u12) as follows:
v H 2 .u2 H6 .u6 H8 .u8 H12 .u12 (6.74)
(a) x
(b)
z
FIGURE 6.20 (a) DOF associated with transverse deformation of beam in x–y plane and (b)
those associated with deformation in x–z plane.
Deformation Analysis of Beams 151
where H2, H6, H8, and H12 are the interpolation functions (also known as shape func-
tions) that can be described with third-degree polynomials (for four nodal displace-
ment DOF – u2, u6, u8, and u12). Beam deflection shape functions, with appropriate
boundary conditions as shown in Figure 6.21, are given by the following cubic
functions:
3x2 2 x3
H2 1 3
L2 L
2 x2 x3
H6 x 2
L L (6.75)
3x2 2 x3
H8 2 3
L L
x2 x3
H12 2
L L
Using the Euler–Bernoulli’s beam deformation theory (discussed in Section 6.6),
normal deformation at any point on a beam cross-section can be defined by the fol-
lowing equation:
v
u y. (6.76)
x
Corresponding normal strain at the same point is given by
u 2v
y. 2 (6.77)
x x
FIGURE 6.21 Shape functions for transverse deformation of beam in x–y plane.
152 Finite Element Analysis of Solids and Structures
Combining equations (6.74), (6.75), and (6.77), normal strain on beam cross-section
can be expressed in terms of the nodal DOF (u2, u6, u8, and u12) as follows:
u2
y u6
x 12 x 6 L L 6 x 4 L 12 x 6 L L 6 x 2 L (6.78)
L3 u8
u12
Re-writing equation (6.78) in the familiar finite element terms:
u2
u6 (6.79)
x B
u8
u12
where strain–displacement relationship matrix [B] is a function of x and y:
y
B 3 12 x 6 L L 6 x 4L 12 x 6 L L 6 x 2 L (6.80)
L
Stress–strain relationship for bending deformation mode of beam is given by the
Hooke’s law:
x E. x (6.81)
The familiar stress–strain relationship matrix for bending deformation, thus, takes
the following form:
C E (6.82)
Substituting expressions (6.80) and (6.82) into equation (2.44), and inserting differ-
ential volume dV = A.dx (where A is the cross-sectional area of prismatic beam ele-
ment), stiffness matrix of beam element for bending mode of deformation in (x,y)
plane is obtained as follows:
12 EI z 6 EI z 12 EI z 6 EI z
L3 L2 L3 L2
6 EI z 4 EI z 6 EI z 2 EI z
L2 L2 L
T L
k xy B . C . B .dV (6.83)
12 EI z 6 EI z 12 EI z 6 EI z
L3 L2 L3 L2
6 EI 2 EI z 6 EI z 4 EI z
2
z
L L L2 L
Deformation Analysis of Beams 153
where Iz is the moment of inertia of beam cross-section defined as
Iz
y .dA
2
(6.84)
Following a similar approach, stiffness matrix of beam for transverse deformation
mode in (x–z) plane (Figure 6.20(b)) can be obtained as follows:
12 EI y 6 EI y 12 EI y 6 EI y
L3 L2 L3 L2
6 EI y 4 EI y 6 EI y 2 EI y
L2 L2 L (6.85)
k xz B . C . B .dV
T L
12 EI y 6 EI y 12 EI y 6 EI y
L3 L2 L3 L2
6 EI 2 EI y 6 EI y 4 EI y
y
L2 L L2 L
where Iy is the moment of inertia of beam cross-section defined as
I y z 2 .dA (6.86)
Combining the stiffness contributions from four resistance mechanisms, equation
(6.72) for axial deformation, equation (6.73) for torsion, equation (6.83) for bending
in (x, y) plane, and equation (6.85) for bending in (x,z) plane, overall stiffness matrix
of the beam with reference to its 12 local DOF (Figure 6.18(c)) is defined by the fol-
lowing equation (6.87):
AE AE
L 0 0 0 0 0 0 0 0 0 0
L
12 EI z 6 EI z 12 EI z 6 EI z
0 0 0 0 0 3 0 0 0
L3 L2 L L2
0 1 2 EI y 6 EI y 12 EI y 6 EI y
0 0 0 0 0 0 0
L3 L2 L3 L2
JG JG
0 0 0 0 0 0 0 0 0 0
L L
6 EI y 4 EI y 6 EI y 2 EI y
0 0 0 0 0 0 2 0 0
L2 L L L
0 6 EI z 4 EI z 6 EI z 2 EI z
0 0 0 0 2 0 0 0
[k] L2 L L L
AE AE
0 0 0 0 0 0 0 0 0 0
L L
0 12 EI z 6 EI z 12 EI z 6 EI z
3 0 0 0 2 0 0 0 0 2
L L L3 L
12 EI y 6 EI y 12 EI y 6 EI y
0 0 0 0 0 0 0 0
L3 L2 L3 L2
JG JG
0 0 0 0 0 0 0 0 0 0
L L
0 6 EI y 2 EI y 6 EI y 4 EI y
0 0 0 0 0 2 0 n2
L2 L L L
6 EI z 2 EI z 6 EI z 4 EI z
0 0 0 0 0 2 0 0 0
L2 L L L
(6.87)
154 Finite Element Analysis of Solids and Structures
Stiffness matrix of prismatic Euler–Bernoulli beam element, with a straight geo-
metric profile, is thus explicitly calculated by using the material and geometric prop-
erty information in equation (6.87). Equation (6.71) is then used to transform the
local stiffness [k], calculated from equation (6.87), to global coordinate reference
system (Figure 6.18(b)). Stiffness matrix of each element, calculated by following
the above steps, is assembled to the global stiffness matrix of a multi-element model
for load–deflection analysis. In the calculation of beam stiffness for transverse load
effects, consideration has been given to lateral deformation caused by bending defor-
mation only (equations 6.83 and 6.85). Element formulation based on this assump-
tion is generally referred to as Euler–Bernoulli beam element. As discussed in
Section 6.6, this assumption (of ignoring the contribution of shear deformation) pro-
vides accurate results for slender beams (equation 6.65).
6.9 STIFFNESS PROPERTIES OF BEAMS INCLUDING SHEAR
DEFORMATION
Equation (6.74) has expressed the transverse deformation of internal points of a
beam for coupled effects of transverse and rotational responses at the nodes. Using
the linear interpolation functions from Figure 2.11, and considering the transverse
deformation response in x–y plane for the beam in Figure 6.20(a), internal deforma-
tion responses can be calculated by separately interpolating the nodal responses for
transverse and rotational DOF:
u2
u8 (6.88)
v H 2 H8 0 0
u6
u12
u2
u8 (6.89)
0 0 H6 H12
u6
u12
where v is transverse deformation at any point inside the beam, and β is the rotation
of beam section at that point. Defining the curvature of beam axis as the first deriva-
tive of β with respect to coordinate x:
u2
H6 H12 u8
0 0 B ui (6.90)
x x x u6
u12
Deformation Analysis of Beams 155
where {ui} is the vector of nodal displacements and rotations, and [Bβ] is the relation-
ship matrix defining curvature at selected beam internal point. Defining the flexural
rigidity of prismatic beam by EI (from equation 6.56), and using the curvature-to-
nodal displacement relationship from equation (6.90), stiffness of the beam for bend-
ing mode of deformation can be calculated by using the standard definition from
equation (2.44):
T
k EI B EI B dx (6.91)
A key assumption in Euler–Bernoulli beam theory is that beam section normal to
neutral axis remains normal during bending deformation. However, shear deforma-
tion in the material, when taken into consideration (Figure 6.22), leads to deviation
from that normality condition. Considering that shear deformation of the material
contributes partially to the total rotation response of the beam section (Figure 6.22),
shear strain can be related to v and β as:
v (6.92)
–
x
Using interpolation relationships from equations (6.88) and (6.89):
u2
v H 2 H8 u8
0 0 Bv ui (6.93)
x x x u6
u12
u2
u8
0 0 H6 H12 H ui (6.94)
u6
u12
dx
(a) (b)
FIGURE 6.22 (a) General rotational response of a beam section; (b) rotation due to shear
deformation.
156 Finite Element Analysis of Solids and Structures
Substituting the relationships from equations (6.93) and (6.94) into equation (6.92),
relationship between internal shear strain and nodal response variables is defined by
the following equation:
Bv H ui (6.95)
Using the standard definition of stiffness matrix from equation (2.44), beam stiffness
corresponding to the shear deformation of material (equation 6.95) can be defined by
the following equation:
T
k GA Bv H GA Bv H dx (6.96)
where G is the shear modulus of material, A is the cross-sectional area of beam, and
ξ is a correction factor for assuming uniform shear stress distribution over beam
cross-section (Figure 6.4(b)), in lieu of more rigorous stress distribution defined by
equation (6.8). For a rectangular beam cross-section example, shear strain distribu-
tion is given by the parabolic function (equation 6.9). Equating the shear strain
energy corresponding to the parabolic shear stress distribution, to that of average
shear stress distribution (τa = V/A), the shear stiffness correction factor is found to be
ξ = 5/6 (Bathe 1996). Correction factors for different beam cross-sections were
derived by Cowper (1966). Combining equations (6.91) and (6.96), stiffness matrix
corresponding to bending and shear deformations in (x,y) plane (Figure 6.20(a)) is
given by equation
k xy k EI k GA (6.97)
Following the approach described above, a similar stiffness matrix of the beam,
considering both bending and shear deformations of beam in (x,z) plane, can also be
derived. Similar to the description of Euler–Bernoulli beam element presented in
Section 6.8, overall stiffness matrix of a beam element in its local reference axis
system can be assembled by combing the contributions from bending and shear
deformations with that of axial deformation (equation 6.72) and that of torsional
deformation equation (6.73):
k k AE k JG k EI k GA (6.98)
The standard coordinate transformation rule (equation 6.71) is then used to trans-
form the local stiffness [k] to global coordinate reference system (Figure 6.18(b)).
Shear deformation in beam response was first introduced by Timoshenko and
Ehrenfest (Wikipedia.org 2020); but beam formulations considering shear deforma-
tion in the material is commonly referred to as “Timoshenko” beam element.
Stiffness matrices in equations (6.91) and (6.96) have been derived based on lin-
ear interpolation of translational and rotational responses at nodes, while the stiffness
matrix formulation presented in equation (6.87) has been derived based on cubic
interpolation of both transverse translation and rotational responses at the nodes. Use
Deformation Analysis of Beams 157
L
x
u1=0 2
u2
1=0
1
r=-1 r=+1
r
Interpolation function
x= . =
2
FIGURE 6.23 Linear interpolation of pure bending response of a cantilever beam.
of linear interpolation functions in stiffness calculations has its drawbacks.
Considering the cantilever beam in Figure 6.23, subjected to pure bending moment
M, the translational and rotational responses at the free end (u2 and Λ2) are calculated
by using equation (6.58):
2 M/EI L; u2 ML2 /2EI (6.99)
Using the linear interpolation function shown in the figure, internal transverse
displacement and rotational responses at any point x tun out to be:
1 r 1 r (6.100)
v u2 2
2 2
Inserting the expressions from equation (6.100) into equation (6.92), and using
the coordination relationship shown in Figure 6.23, internal shear strain in the beam
element is given by
u2 1 r (6.101)
2
L 2
Combining equations (6.99) and (6.101), shear strain inside the beam element of
Figure 6.23 is found to be:
ML (6.102)
r
2 EI
Per the principle of solid mechanics, shear strain (and stress) in a beam under pure
bending load should be “zero”. Use of the linear interpolation function in the analysis
of cantilever beam in Figure 6.23, however, produces a shear strain inside the beam
158 Finite Element Analysis of Solids and Structures
except at r = 0. Non-zero shear strain, introduced by the linear interpolation function,
is commonly referred to as parasitic shear or “shear locking” of beam. Element stiff-
ness formulation based on linear interpolation functions (equation 6.96) produces
artificially stiff structural response – similar to the phenomenon discussed in Section
4.2 for 2D solids elements. A potential remedy for the shear locking behavior of
linear beam elements is to calculate shear strain at the element center (r = 0); and
assume it to remain constant through the element length. This action enforces “zero”
shear condition for a pure bending condition without sacrificing actual shear strain
that may appear due to other general loading conditions. Alternative beam element
formulations, with shear flexibility contribution, apply a scaling factor to the shear
rigidity term (ξGA) in equation (6.96) to limit the shear stiffness value as the beam
slenderness increases. Example scaling factor, used in ABAQUS beam element for-
mulation, is given by the following equation (Dassault Systems 2020b):
1 (6.103)
scale fcator
A.L2
1 0.25
12 I
where L is the beam span, A is the cross-sectional area, and I is the moment of inertia.
Yet another approach for considering shear flexibility contribution in beam stiffness
formulation is to apply a shear correction factor to the stiffness terms in equations
(6.83 and 6.85) that have been derived by using cubic interpolation functions:
12 6L 12 6L
2
EI 6L 4 L2 6 L 2 L (6.104)
k 3
L 1 12 6 L 12 6 L
6 L 2 L2 6 L 4 L2
The shear correction factor in equation (6.104) is defined by equation (6.105)
(Logan 2012):
12 EI (6.105)
GAL2
Comparing the terms in equation (6.104) with those in equations (6.83 and 6.85), it
is evident that the shear deformation in beam reduces the magnitude of stiffness
terms associated with transverse displacement and rotational response modes.
Contribution of shear deformation can be neglected for slender beam: L > 10*(2h) in
equation (6.65).
6.10 ANALYSIS OF BEAMS AND FRAMES WITH FEA SOFTWARE
PACKAGES
Dedicated software packages such SAP2000 (CSIAmerica.com) and STAAD Pro
(Bentley.com) are widely used in design and analysis of structural frames.
Deformation Analysis of Beams 159
beam axis
N2
x z
N3 N1
FIGURE 6.24 Local reference system of a beam element in ABAQUS software.
General-purpose finite element analysis packages (e.g. ABAQUS, ANSYS, etc.) also
provide equally capable element library and modeling tools for analysis of skeletal
structural frames. Figure 6.24 shows a straight beam element with ABAQUS-specific
local reference coordinate definitions. Element geometry in three-dimensional space
is described by the following syntax for ABAQUS analysis model description:
*ELEMENT, TYPE=B31, ELSET=setname
Element_ID_no, N1, N2, N3
where nodes N1, N2, and N3, defined in global coordinate system (X,Y,Z), also define
the local orientation of the beam element – nodes N1 and N2 define the element axis
direction (z), and node N3 the local direction x as shown in Figure 6.24. The element
type selection “B31” refers to the beam element formulation, described in Section
6.9, that uses linear interpolation functions with material shear deformation included.
Shear locking behavior in the element is reduced by scaling the shear rigidity term
with scaling factor defined in equation (6.103). Rectangular beam section example,
shown in Figure 6.24, is defined in ABAQUS input file by using the following
syntax:
*BEAM SECTION, ELSET= setname, MATERIAL= matname,
SECTION=RECT
a,b
where dimensions a and b refer to the specific local orientation of rectangular section
identified in Figure 6.24. Section properties specific to axial, shear, bending, and
torsional response modes are internally calculated by ABAQUS based on the section
160 Finite Element Analysis of Solids and Structures
y t3 (x4,y4)
(x3,y3)
t2 N1
x
x t3 t2 N3 (0,0)
N3 N 1 t1 b
t4
a
(x1,y1)
( x 2,y 2) t1
(a) (b)
FIGURE 6.25 Examples of beam section dimensions in local reference system for ABAQUS
beam elements: (a) thin-wall closed box section, and (b) thin-wall open section.
definition provided with the element input data. ABAQUS provides a wide variety of
options to define beam section dimensions including thin-walled closed and open
sections. Section dimensions for the closed box section example, shown in Figure
6.25(a), are defined by using the following ABAQUS input syntax:
*BEAM SECTION, ELSET= setname, MATERIAL= matname, SECTION=BOX
a,b,t1,t2,t3,t4
where a and b are side dimensions of rectangular box section and t1,..t4 are the
thickness values of side walls as identified in Figure 6.25(a). Following data input
syntax describes the dimensions of three-sided arbitrary thin-walled section example
shown in Figure 6.25(b):
*BEAM SECTION, ELSET= setname, MATERIAL= matname, SECTION=ARBITRARY
3, x1, y1, x2, y2, t1
x3, y3, t2
x4, y4, t3
where the number of segments in the arbitrary section is identified at the beginning of
the first data input line, followed by local coordinates (x1,y1) of the first corner point,
coordinates (x2,y2) for the second point, and the thickness t1 of first segment. Each
additional segment of arbitrary section is defined in additional data lines listing the
coordinates (xi,yi) followed by thickness ti of that segment. Polar moment of inertia of
thin-walled open section is calculated by using membrane analogy (equation 6.50).
Two-node Euler–Bernoulli beam element, ignoring the shear deformation in
material, can be selected by choosing “TYPE=B33” in input data for element descrip-
tion. ABAQUS beam element library also includes a three-node element (B32) that
uses quadratic shape functions (Figure 2.15), and it includes the material shear defor-
mation in stiffness formulation following the same developments presented in
Section 6.9. Geometry of three-node element is described as follows:
*ELEMENT, TYPE=B32, ELSET=setname
Element_ID_no, N1, N2, N3, N4
Deformation Analysis of Beams 161
Beam 1 Beam 2
(a) Bolt connec on
n3 Beam 2 n4
n1 Beam 1 n2
(b)
FIGURE 6.26 (a) Assembly of two-beam elements with moment-free bolted joint, and (b)
elements defined with separate nodes n2 and n3 at the joint location.
where N1 and N3 are end nodes, N2 is the mid-side node, and N4 refers to the local
coordinate direction x defining section orientation in 3D space. Extra node
description, N3 in two-node element and N4 in three-node element, is not required
in two-dimensional analysis of frames where local reference direction x is assumed
normal to the model description plane. Element types for two-dimensional model
simulation are “B21” in lieu of “B31” for linearly interpolated two-node beam
element with shear deformation; “B23” in lieu of “B33” for Euler–Bernoulli beam
element; and “B22” in lieu of “B32” for three-node element with shear
deformation.
In frame assemblies, degrees of freedom at shared nodes experience resistances
(stiffness contributions) from all connecting elements, thus, producing fully coupled
motion among the joining members. However, not all joints in beam assemblies are
designed to produce fully coupled motions among the elements. For example, single
bolt connection between two beam parts in Figure 6.26 keeps translational motions
coupled while allowing relative rotation between the two joining ends. Coupling of
selected DOF between connecting members are modeled in ABAQUS by assigning
coincident separate nodes to the joining elements, and then by constraining only the
selected DOF. Input parameters, for example problem of Figure 6.26, can be
described as follows:
*ELEMENT, TYPE=B21, ELSET=set1
Beam-1, n1, n2
Beam-2, n3, n4
*ELEMENT, TYPE=CONN2D2, ELSET=set2
El-id, n2, n3
*CONNECTOR SECTION, ELSET=set2
JOIN
162 Finite Element Analysis of Solids and Structures
where n2 and n3 are coincident nodes attached to beam elements Beam-1 and Beam-
2, respectively. A special joint element is defined, of TYPE=CONN2D2, for connec-
tivity between two nodes (n2 and n3) on a two-dimensional plane. Property of the
connector element (El-id) is defined by selecting “JOIN” keyword in the connector
element property description which implies to ABAQUS that the nodes connected to
element (El-id) are kinematically coupled for translational DOF. Rotational DOF at
nodes n2 and n3 remain un-coupled, meaning that the nodes can rotate independent
of each other. ABAQUS joint element library includes a long list of various other
options to define selective coupling between adjoining elements. More discussion on
the modelling of joints in structural analysis is presented in Chapter 8.
6.11 PRACTICE PROBLEMS: LOAD–DEFLECTION ANALYSIS
OF BEAMS
PROBLEM 1
Figure 6.27 shows a solid rectangular section cantilever beam subjected to an in-
plane transverse load of P = 10 kN at the free end. Determine vertical deflection at
the load application end by using the following analysis models:
a. Analytical solution with plane stress idealization as presented in Section 3.4
b. Beam element model with and without shear deformation consideration
c. Plane-stress finite element model of the solid beam structure
omment on the result differences, and on the reliability of analysis results (b) and
C
(c) relative to the analytical solution given by (a).
P
x
100
10
beam cross- 1000
sec on: 10x100
FIGURE 6.27 A rectangular steel beam (E = 210 GPa, ν = 0.3) is subjected to an in-plane
transverse load of P = 10 kN at the left end while the right end is fully constrained (all dimen-
sions in mm).
Deformation Analysis of Beams 163
8mm thick rigidly aached bracket –upper
surface (40x8 area) is subject to uniform
downward pressure of 0.01 kN/mm2
60
100 uniform wall
thickness: 4 mm
600
z
y x
600
Assume fixed/fixed
at both ends
FIGURE 6.28 Thin-wall C-section beam loaded with a shear bracket at mid-span.
PROBLEM 2
Figure 6.28 shows a 1200 mm long C-section aluminum beam subjected to a uniform
pressure load applied on a loading bracket seamlessly attached to the side of main
beam. Geometry data of the beam (including loading bracket) is given in the file
“3D-C-beam-with-loading-bracket.stp” (support data is available at a website men-
tioned in the preface of this book). Assume that both main beam and loading bracket
are made of aluminum (E = 70 GPa, ν = 0.33). Determine the shear stress flow
through the flanges of the beam at mid-span section using the following analysis
models:
a. Analytical solution for the combined effects of shear and torsion as dis-
cussed in Section 6.5
b. Beam element model of the problem
c. 3D solid element model of the beam and bracket assembly
Comment on the result differences caused by the different modeling assumptions.
PROBLEM 3
Figure 6.29 shows a solid steel hook of circular cross-section – subjected to a force
of P = 0.5 kN. Geometry data (Solid_Hook_Geometry.stp) is available in the support
data website. Determine the normal stress distribution on the section A-B using finite
element simulation technique. Calculate the tangential stress at point A using hand
calculations for curved beam theory; and compare the result with that of finite ele-
ment simulation model.
164 Finite Element Analysis of Solids and Structures
50
32
60
B A
FIGURE 6.29 Stress analysis of a solid circular section steel hook (dimensions in mm).
Assume elastic material properties: E = 210 GPa, ν = 0.3
PROBLEM 4
Geometry of a long beam, including section dimensions, material property, loading
and boundary conditions are described in Figure 6.30. Beam possesses a moment-
free joint at point C. Predict the vertical deflection at joint C by analyzing the struc-
ture using beam and joint elements discussed in Section 6.10.
0.01 kN/mm Beam section & material
A B 100x200 box sec on
D Wall thickness: 5 mm
C
1000 200
1000 Material: steel
2000 100
E=210 GPa
Ends A and D are fully constrained
Support @ B provides constraint against ver cal deflec on
Internal hinge @ C releases moment resistance at that point
Distributed ver cal load of 0.01 kN/mm is applied over the segment CD
Assume constant box x-sec on and constant material proper es.
FIGURE 6.30 Statically indeterminate beam with an internal moment-free joint at point C.
7 Analysis of 3D Thin-Wall
Structures (Plates and
Shells)
SUMMARY
Plates and shells are three-dimensional solid material components that have thick-
ness values much smaller in comparison to the surface dimensions. Finite element
simulation of such structures with 3D solid elements requires too many elements to
be used. Moreover, bending response of lower order solid elements generally tends
to be stiff, thus under-predicting the displacement response of the structures. The
alternative to 3D solid element use is to discretize the mid-surface of thin-walled
parts to surface-based elements that are capable of producing bending and shear
resistance to out-of-plane loads. Fundamentals of bending stress–strain and deflec-
tion responses of flat plates are reviewed in Sections 7.1 and 7.2. Plate bending the-
ory, however, explains only one part of the shell resistance mechanism. Traditional
structural shells have been conceived and used over centuries to provide high resis-
tance in the tangential mid-plane direction. Section 7.3 presents stress analysis of a
general shell – highlighting the dominant role of membrane stresses in the shell
response mechanism. Evidently, the in-plane membrane resistance mechanism can
be simulated by using the plane-stress formulations presented in Chapter 3. Stiffness
calculations for the plate bending response are developed in Section 7.4. The flat
shell element formulation, developed from superposition of the in-plane membrane
resistance and the out-of-plane plate bending resistance mechanisms, is presented in
Section 7.5. Like the deep beam behavior, discussed in Chapter 6, material shear
deformation in thick shells can accumulate to a significant part of the deflection
response of structures. The stiffness formulation of flat shell elements, with added
consideration of material shear deformation, is presented in Section 7.6. Many soft-
ware implementations provide flat shell elements to be used for simulation of both
planar and curved shell structures. It is generally expected that refined finite element
mesh, with quadrilateral and triangular shaped flat elements, can adequately simulate
the response of curved shell structures. However, huge volume of research work,
partly driven by desire for theoretical purity, and partly for improved accuracy, has
eventually led to the development and implementation of iso-parametric curved shell
element formulations in some software packages. The salient features of curved shell
elements are presented in Section 7.7. General topics of finite element mesh quality
checks and integration rules of shell elements are discussed in Section 7.8. A review
of ABAQUS-specific curved shell element options is presented in Section 7.9.
Practice problems for shell structural analyses are presented in the final Section 7.10.
165
166 Finite Element Analysis of Solids and Structures
7.1 BENDING STRESSES AND STRAINS IN A PLATE
Figure 7.1 shows the roof panel of a vehicle subjected to a hypothetical distributed
load in the normal direction of surface. Action of this applied load is resisted by
material resistances to bending and shear – similar to the behavior of beams dis-
cussed in Chapter 6. However, the roof panel in this example cannot be modeled as a
beam-like discrete line element because of its large planar dimensions. Neither can it
be modeled as a three-dimensional solid body providing resistances to translational
degrees of freedom only. Special analytical techniques are required to analyze the
bending and shear deformation responses of plates and shells (Timoshenko and
Woinowsky-Krieger 1970). Considering the plate element of uniform thickness in
local z-direction, deformation degrees of freedom at a material point on the mid-
surface of plate can be represented by translation in the normal direction of mid-
surface and two rotations as shown in Figure 7.2. Rotational degree of freedom about
local z-axis is ignored assuming rigid rotational response in the x–y plane of the
plate. Extending the Euler–Bernoulli’s bending deformation theory of beams,
FIGURE 7.1 Thin sheet metal roof panel of a vehicle – capable of providing resistance to
bending under normal load on the surface (vehicle FEA model: courtesy of NHTSA 2020).
Analysis of 3D Thin-Wall Structures (Plates and Shells) 167
w y
a y
x
b x
z y
x
w v
u
FIGURE 7.2 A thin plate (t < b/20) with mid-surface in the x–y plane – generated by collaps-
ing the 3D solid in the z-direction.
out-of-plane pure bending response of a plate can be idealized with the following
assumptions – commonly known as Kirchoff’s plate bending theory:
• Deflection of mid-surface is small compared to thickness
• Lines normal to mid-surface before deformation remain normal after deforma-
tion (shear deformation in the material is ignored)
• No mid-surface straining occurs due to plate bending
• Stress component normal to mid-surface is zero
Based on these assumptions, the general strain–displacement relationships of equa-
tions (2.4) can be re-written for plate bending case as follows:
u v u v
x , y , xy ,
x y y x
w w v w u
z 0, yz 0, xz 0 (7.1)
z y z x z
Considering the simplified case of one-directional bending of a plate (Figure 7.3),
rotation of plate section in y–z plane, x-axis being normal to that plane, is related to
the transverse deflection of mid-surface w by equation (7.2):
w (7.2)
x
Assumption of plate section normal to the mid-surface remaining normal through
the deformation phase leads to a linear variation of deformation through thickness of
plate:
w
u z. (7.3)
x
168 Finite Element Analysis of Solids and Structures
y =
z
w
FIGURE 7.3 Bending rotation of plate-section in y–z plane.
Similarly, considering the simplified case of uni-modal bending of plate caused by
rotation of section x–z (y-axis being normal to that plane), deformation inside the
plate in the y-direction is defined by the following equation:
w (7.4)
v z.
y
Substituting expressions from equations (7.3) and (7.4) into equations (7.1), non-
zero strain terms are given by the following equations:
u 2w
x z. 2
x x
v 2w (7.5)
y z. 2
y y
u v 2w
xy 2 z.
y x x y
Using the Hooke’s law, stress–strain relationships for bi-directional plate bending
response can be defined as follows (equations 7.6):
E
x
1 v2
x v y
E
y
1 v2
y v x (7.6)
xy G xy
where strains εx, εy, and γxy vary in the z-direction through the plate thickness (equa-
tions 7.5). Similar to the beam response, bending stresses on a plate section (made of
homogeneous isotropic material) vary linearly in the z-direction from the mid-sur-
face, and the shear stress distribution takes a parabolic shape (Figure 7.4).
Analysis of 3D Thin-Wall Structures (Plates and Shells) 169
Transverse
shear stress yz
Bending
y stress y
z z
y
z
t x
Bending Transverse
1 stress x shear stress xz
x
FIGURE 7.4 Bending and shear stresses through the thickness of plate.
7.2 ANALYTICAL SOLUTIONS FOR PLATE BENDING
DEFLECTIONS
Integrating the bending the stresses (σx) over y–z plane of plate section, moment (Mx)
per unit strip length (Figure 7.4) can be defined by
t /2 t /2
E
Mx
t / 2
z. x .dz
1 v2 t / 2
z. x v y .dz
(7.7)
Substituting the expressions for εx and εy from equations (7.5), and after conducting
partial integration with respect to z, equation (7.7) takes the following form:
Et 3 2w 2w 2w 2w
Mx . 2 v 2 D. 2 v 2 (7.8)
12 1 v 2 x y x y
where D [= Et3/12(1–ν2)] is known as flexural rigidity of a plate made with homoge-
neous isotropic material. Inside the right-hand side expression of equation (7.8), the
second-order differential terms represent the change in curvature of the plate in the
x- and y-direction, respectively. Similarly, by integrating the contributions of other
two stress components (σy and τxy), we get the following expressions for moments per
unit strip length (equations 7.9 and 7.10):
2w 2w
M y D. 2 v 2 (7.9)
y x
170 Finite Element Analysis of Solids and Structures
+
y
z +
+
+
p
+
dy
x
dx
FIGURE 7.5 Shear forces and bending moments on the edges of a uniformly loaded plate
element.
2w
M xy D. 1 . (7.10)
x y
Considering the equilibrium of vertical forces acting on a differential plate ele-
ment of dimensions (dx, dy) (Figure 7.5), we obtain:
Vx V (7.11)
.dx.dy y .dx.dy p.dx.dy 0
x y
Factoring out the non-zero multiplication term (dx.dy), equation (7.11) takes the fol-
lowing reduced form:
Vx Vy (7.12)
p 0
x y
Similarly, considering the equilibrium of moments about x and y axes separately, we
arrive at the following equations:
M xy M y (7.13)
Vy 0
x y
M xy M x (7.14)
Vx 0
y x
Combining equations (7.8), (7.9), (7.10), (7.12), (7.13), and (7.14), bending deflec-
tion response of a plate, subjected to a uniform distributed load, is given by a single
differential equation:
4w 4w 4w p
2 (7.15)
x 4 x 2 . y 2 y 4 D
Analysis of 3D Thin-Wall Structures (Plates and Shells) 171
y
y
a
b
b
x
(a)
x
(b)
(c)
FIGURE 7.6 (a) Narrow rectangular plate simply supported on long sides; (b) rectangular
plate simply supported on all sides; and (c) circular plate with rigid support on the perimeter.
Integration of the differential equation (7.15), with the use of appropriate bound-
ary conditions, provides the plate deflection response w(x,y). Once the deflection
function w(x,y) is known, strains and stresses inside the plate can be determined by
using the relationships described earlier in this section. For example, the deflection
of narrow rectangular plate of width b in Figure 7.6(a), with simply supported long
edges and a distributed load of p = po. sin (πy/b), is obtained by integrating equation
(7.15) as follows:
4
b p y (7.16)
w . 0 .sin
D b
where po refers to the peak amplitude of non-uniform load along the plate centerline
at y = b/2. Additional solutions for elementary plate deflection examples can be found
in Timoshenko and Woinowsky-Krieger (1970) and Ugural and Fenster (2012).
Deflection of the simply supported rectangular plate in Figure 7.6(b), subjected to a
uniformly distributed pressure p, is given by the following equation (Ugural and
Fenster 2012):
sin m x /a .sin n y /b
mn. m /a n/b
16. p0 (7.17)
w , m, n 1, 3, 5 .
6D m n
2 2
2
For a uniformly loaded square plate (a = b), with simple supports at the edges, maxi-
mum deflection at mid-span is obtained from equation (7.17) as follows:
a4
wmax 0.0443. po . 3 (7.18)
Et
172 Finite Element Analysis of Solids and Structures
r0
N r
N
+
FIGURE 7.7 Segment of a spherical shell subjected to uniform external pressure (p).
And the maximum bending at the mid-span is given by
M x ,max M y,max 0.0472. po .a 2 (7.19)
Similarly, maximum bending responses of a circular plate (Figure 7.7(c)) of radius a,
with moment–resistant rigid support on the perimeter, are given by
po .a 4
Maximum deflection @ plate center : wmax (7.20)
64 D
po .a 2
Maximum bending moment @ the boundary : Mr (7.21)
8
where po is the uniform pressure applied on the circular plate. Analytical solutions
for elementary plate bending examples are useful in the early designs of structural
members that can be simplified to be represented by one of the known examples.
These solutions can also be used to approximately verify the finite element analysis
results of more complex structural problems.
7.3 IN-PLANE MEMBRANE STRESS RESISTANCE OF A SHELL
Figure 7.7 shows a portion of a spherical shell of radius r and thickness t, subjected
to a uniform pressure p. The uniform external pressure generates an internal “mem-
brane” force per unit length (N), acting in the tangential direction of mid-surface
(normal direction of radial cross-sectional plane). From the consideration of overall
system equilibrium in vertical direction:
2 ro .N .Sin p. .ro2 N 2.Sino
pr pr
(7.22)
2
Analysis of 3D Thin-Wall Structures (Plates and Shells) 173
Membrane stress inside the spherical shell of thickness t, caused by the in-plane
tangential force (N), is found to be:
N pr (7.23)
n
t 2t
where –ve sign implies compressive stress on the shell section. Ignoring the normal
stress on the mid-surface of spherical shell, membrane strain caused by the bi-
directional membrane stresses is given by
n 1 pr (7.24)
n . n .
E E E 2t
Reduced circumference of spherical shell, because of the compressive normal
strain (εn), is defined by
2 r r. n 2 .r 1 n 2 .r (7.25)
where r′ is the new radius of the spherical shell. Change in curvature of the shell can
be calculated from:
1 1 1 1 n 1 n
n n
r r r 1 1 r 1 r 1 n n ..
2
(7.26)
Ignoring the higher order terms of εn in equation (7.26), and using the expression
from equation (7.24), change in curvature of the spherical shell under uniform exter-
nal pressure is defined by the following equation:
1 1 n 1 p (7.27)
r r r E . 2t
Using the expression for the change in curvature from equation (7.27) into the
equation for plate bending moment (equation 7.8), bending moment inside the spher-
ical shell is defined as follows:
Et 3 2w 2w Et 3 1 1 p pt 2
Mr . v . v. .
12 1 v 2 x
2
y 2 12 1 v 2 E E 2t 24 ( 7.28)
Maximum bending stress caused by the bending moment Mr, acting on a rectangular
segment of unit width and thickness t, is given by equation (7.29):
6.Mr p
b (7.29)
t2 4
174 Finite Element Analysis of Solids and Structures
Taking the ratio between membrane stress σn (equation 7.23) and bending stress
σb (equation 7.29) for spherical shell under uniform external pressure, we obtain:
n 2r (7.30)
b t
For typical shell dimensions of r >> t, equation (7.30) indicates that membrane stress
dominates the internal resistance mechanism of the shell under external surface pres-
sure. Stiffness properties of shell, thus, depend on combined resistance mechanisms
of plate bending (Section 7.4) and plane-stress effects (Section 3.5).
7.4 BENDING STIFFNESS OF FLAT PLATE ELEMENT
As discussed in Section 7.1 and illustrated in Figure 7.2, deformation degrees of
freedom for bending response of a plate involve transverse deflection (w) and two
rotations θx and θy. Numerous formulations have been proposed in the literature for
deformation response of plate elements (Logan 2012). In one of the most basic plate
bending formulations, transverse displacement response of four-node plate element,
having a total of 12 DOF, is defined by a polynomial function with 12 coefficients:
w a1 a2 x a3 y a4 x 2 a5 xy a6 y 2
a7 x 3 a8 x 2 y a9 xy 2 a10 y3 a11 x 3 y a12 xy3 (7.31)
This function satisfies the first basic requirement of meeting the compatibility
condition defined by equation (7.15) for zero external normal load (p = 0). Equation
(7.31), however, is complete up to third order (with first 10 terms). The fourth-order
terms x3y and xy3 are chosen to ensure displacement continuity along inter-element
boundaries. This function does not ensure slope continuity along inter-element
boundaries. Full compatibility is impossible to obtain with simple polynomial
expressions involving only 3 DOF at the corner nodes (Zienkiewicz and Taylor
1991). Despite this limitation, convergence properties of many such non-conforming
formulations have been successfully proven in the literature.
Taking partial derivatives of expression in equation (7.31), with respect to coordi-
nate variables x and y, deformation responses of the reference mid-surface of a plate
can be expressed by following relations:
a1
w
1 x y x2 xy y2 x3 x2 y xy 2 y3 x3 y xy3 a2
w 0 2 (7.32)
0 1 0 x 2y 0 x2 2 xy 3y2 x3 3 xy ..
y
0 1 0 2x y 0 3x 2 2 xy y2 0 3x 2 y 3
y ...
w
x a12
Writing the matrix equations (7.32) at four node locations, we get the following
expressions for deformation responses at the nodes:
Analysis of 3D Thin-Wall Structures (Plates and Shells) 175
wi 1 xi yi xi2 xi yi yi2 xi3 xi2 yi xi yi3 yi3 xi3 yi xi yi3 a1
xi 0 0 1 0 xi 2 yi 0 xi2 2 xi yi 3 yi2 xi3 3 xi yi2 a2
yi .. a
3
w j .. ..
.. ..
12 12 matix ..
.. .. .. .. .. .. .. .. .. .. .. .. .. a12
(7.33)
where (xi, yi) are the coordinates of node i. Denoting the (12 × 12) multiplication
matrix on the right-hand side of equation (7.33) as [P], equations (7.33) for the nodal
displacement responses are re-written as follows:
wi a1
xi a2
yi a3 (7.34)
P
w j ..
.. ..
.. a12
Polynomial coefficients, a1 … a12, can be calculated by inverting the relationship
matrix [P]:
(7.35)
a P ui
1
where {ui} is vector of nodal displacements and rotations (wi and θi). Using the dis-
placement function (equation 7.31), curvatures of plate bending can be defined as
follows:
2w
2 a1
x 0 0 0 2 0 0 6 x 2 y 0 0 6 xy
0 a2
2 w
2 0 0 0 0 0 2 0 0 2 x 6y 0 6 xy ..
y 0 0 0 0 2 0 0 4 x 4 y 0 12 x 2 12 y 2 ..
2. w
2
a12
x y
(7.36)
Writing the multiplication matrix on the right-hand side of equation (7.36) in con-
densed form as [Q], curvatures of plate bending can be defined by
2w
2
x
2 w
2 Q a (7.37)
y
2w
2.
x y
176 Finite Element Analysis of Solids and Structures
Substituting the relationship from equation (7.35) into equation (7.37), we get the
following relationship between plate curvature and nodal displacement vector:
2w
2
x
2 w
2 Q P ui B ui
1
(7.38)
y
2w
2.
x y
where [B] = [Q].[P]–1 is the familiar finite element matrix relating nodal displacement
DOF to internal deformation responses. Moment curvature relationships by combin-
ing equations (7.8), (7.9), and (7.10) can be written in the following matrix form:
2w 2w
2 2
x x
Mx 1 v 0
Et 3
w
2 2 w
My (7.39)
v 1 0 . y 2 C . y 2
xy
M 12 1 v
2
1 v
0 0 2w
2 2. w
2
2.
x y x y
where [C] is the moment–curvature relationship matrix defined by elastic material
properties (E and ν) and plate thickness (t). Using the standard stiffness matrix for-
mulation from equation (2.44), bending stiffness of a plate is calculated by integrat-
ing the properties over x–y plane of element:
B
T
K C B dx.dy (7.40)
where [B] and [C] are element property matrices defined in equations (7.38) and
(7.39). Numerical integration with iso-parametric element definition, discussed in
Section (3.7), is used to calculate the stiffness in finite element analysis. Eventually,
standard matrix method analysis of structure, with consideration of applied external
loads and boundary conditions, provides the nodal responses of a finite element
model of plates. Stiffness formulation details for triangular plate bending elements
are not discussed in this book; those are available in Cook et al. (1989) and
Zienkiewicz and Taylor (1991).
7.5 FLAT SHELL ELEMENT AS A COMBINATION OF PLATE
BENDING AND MEMBRANE ELEMENTS
Curved shell surfaces are often represented by assembly of piece-wise linear flat ele-
ments (Figure 7.8). For the ideal case of 4 in-plane nodes, local directions x1 and x2
Analysis of 3D Thin-Wall Structures (Plates and Shells) 177
plane shell elements
4 3
x3 x2
x1
1 2
FIGURE 7.8 Curved shell surface is modeled by using four-node planar shell elements.
are tangential to the shell’s mid-surface, and x3 is normal to that mid-surface.
Approximate assumptions are often made to define the planar surface of a shell. For
example, in LS-DYNA software implementation (LSTC.COM 2020), shell mid-sur-
face is assumed to be on the plane formed by two diagonals (connecting the opposite
corner nodes), and the shell normal is assumed in the direction normal to that plane.
Warped initial geometry, when idealized with a planar surface, leads to significant
calculation errors. Stiffness properties of flat shell elements are generated by super-
posing the independent actions of in-plane stress resistance and plate bending resis-
tance (Figure 7.9). Stiffness matrix of in-plane membrane resistance mechanism is
formulated by using the iso-parametric element formulation procedure described in
Sections (3.6) and (3.7). For four-node element configuration (Figure 7.9(a)),
In-plane membrane acon (plane-stress)
s
x x
r x x
(a)
Transverse displacement & bending of plate
x x
x x
(b)
FIGURE 7.9 Resistance mechanisms in a shell: (a) membrane action and (b) plate bending.
178 Finite Element Analysis of Solids and Structures
membrane stiffness matrix is calculated by numerically adding the contributions of 4
Gauss integration points (equation 3.84):
B
T
kmembrane C B dx.dy (7.41)
Effects of element geometry on result quality, discussed for plane-stress elements
in Chapters 3 and 4, equally apply to shell analyses. Like the four-node plane-stress
solids, full 2 × 2 integration of membrane stiffness matrix also leads to high in-plane
stiffness for shell elements. Reduced integration in the element plane can circumvent
that artificial stiffness effect, but it comes with the risk of producing zero-energy
hourglass deformation mode in the element (Figure 3.13). Same planar integration
rule (full- or reduced), used for the calculation of membrane stiffness matrix, can be
used to calculate the bending stiffness matrix of constant-thickness plate element by
using the integral formulation shown in equation (7.40). This bending stiffness equa-
tion, developed from moment–curvature relationships of equation (7.39), assumes
that shell is made of single-layer homogeneous isotropic material. Stiffness matrix of
composite shell, constructed with multiple layers of different materials, can be for-
mulated by going back to the basic mechanics of plate bending response discussed in
Section (7.1). Combining equations (7.5) and (7.38), bending-induced strains at a
distance z from shell mid-surface are defined by
2w
z. 2
x
x
2 w
y z. 2 z. Q P ui
Bz ui
1
(7.42)
y
xy
2w
2 z.
x y
where [Q] and [P] are element geometric property matrices, defined in equations
(7.33) and (7.36); and [Bz] is the familiar matrix notation relating the internal mate-
rial strains to the nodal displacement and rotational response variables. Stress–strain
relationships for plate bending response, given in equations (7.6), can be re-written
in the following matrix form:
E E
0
x 1 1 2
2
x x (7.43)
E E
y 0
y Cz
y
1 1 2
2
xy 0
0 G xy xy
where [Cz] represents the stress–strain relationship matrix corresponding to the spe-
cific material layer at distance z from the shell mid-surface. Taking the
Analysis of 3D Thin-Wall Structures (Plates and Shells) 179
strain–displacement relationship matrix [Bz] from equation (7.42), and stress–strain
relationship [Cz] from equation (7.43), bending stiffness matrix of composite shell is
calculated by
B
T
k plate bending z Cz Bz dx.dy.dz (7.44)
In addition to the numerical calculation points in the x–y plane (mid-surface) of
element, multiple calculation points are added in thickness direction to take account
of material property variation in composite shell construction (Figure 7.9(b)). Overall
stiffness of shell element, for combined actions of membrane and plate bending
resistance, is obtained by adding the contributions [kmembrane] and [kplate bending], respec-
tively, from equations (7.41) and (7.44). Shell element formulation, presented above,
is basically following the Kirchoff’s plate bending theory without considering shear
deformation in thickness direction. Similar to the Euler–Bernoulli’s beam deflection
theory (discussed in Chapter 6), Kirchoff’s plate bending formulation provides accu-
rate estimation of deflection response for thin shells (t < b/20 in Figure 7.2).
7.6 SHEAR DEFORMATION IN PLATES
Plate response theory with material shear deformation in the thickness direction,
commonly known as Mindlin’s plate deformation theory, was introduced by Mindlin
(1951) with a reference to earlier Russian publication by Uflyand (1948). A similar
theory was proposed earlier by Reissner (1945). Both theories are intended for thick
plates in which normal to the mid-surface remains straight but not necessarily per-
pendicular to the mid-surface. Mindlin’s theory ignores normal stress through the
thickness, thus, implying a plane-stress state in the plane of plate, while Reissner's
theory does not invoke the plane-stress condition. It assumes the shear stress is qua-
dratic through the thickness of plate, and the plate thickness may change during
deformation. Stiffness formulation for Mindlin’s plate element can be derived by
following the same procedure presented in Section (6.9) for the derivation of beam
stiffness properties including material shear deformation. A line that is straight and
normal to the mid-surface before deformation is assumed o remain straight but not
normal to the surface after deformation (Cook et al. 1989). Strains at a point not on
the mid-surface are defined by rewriting equation (7.42) in the following form:
½
° wE °
z. x
Hx ½ ° wx °
° ° °° wE y °°
®Hy ¾ ® z. ¾ ª¬ BE º¼ u ^ui ` (7.45)
°J ° ° wy °
¯ xy ¿ ° § wE wE ·°
° z. ¨ x y ¸°
°¯ © w y wx ¹ °¿
180 Finite Element Analysis of Solids and Structures
where βx and βy are rotations of the reference line obtained from nodal rotational
variables by using element-specific interpolation functions (similar to equation 6.94
presented for shear deformation analysis of beams). Stiffness matrix for plate curva-
ture is calculated by inserting [Bβ] in equation (7.44):
B
T
k plate bending Cz B dx.dy.dz (7.46)
Shear strains through thickness of plate are defined extending the definition from
beam problem (equation 6.92):
w w (7.47)
yz y ; zx xs
y x
where w is the transverse displacement of point (x, y) on the initial mid-surface.
Expressing the transverse displacement w, and rotations βx and βy, in terms of nodal
response variables {ui}, equation (7.47) can be re-written in the following form:
1 0 w 1 0
yz y y
y . Hi .
ui B .
ui (7.48)
zx 0
1 x 0
1
x x
where [Hi] are element interpolation functions. Stress–strain relationships for
through-thickness shear deformation are defined by
yz Gyz 0 yz yz (7.49)
C .
zx 0 Gzx zx zx
For homogeneous isotropic elastic material, shear modulus, Gyz = Gzx = G, is defined
by equation (2.23) in terms of Young’s modulus (E) and Poisson’s ratio (ν). Stiffness
matrix for material resistance to through-thickness shear deformation is obtained by
using [Bγ] and [Cγ] in equation (2.44):
B
T
kshear C B dx.dy.dz (7.50)
Overall stiffness matrix of a flat shell, is finally, obtained by adding the contribu-
tions of all three resistance mechanisms:
kshell k Equation 7.41 k Equation 7.46 k Equation 7.50 (7.51)
For thin shells (t < b/20 in Figure 7.2), third term in equation (7.51) can be ignored,
thus, leaving only the membrane and plate bending resistance terms in the element
Analysis of 3D Thin-Wall Structures (Plates and Shells) 181
stiffness formulation. The derivation of shell stiffness from superposition of mem-
brane and plate bending actions, with or without through thickness shear deforma-
tion, assumes that the behavior of a continuously curved shell surface can be
adequately represented by an assembly of piece-wise flat quadrilateral or triangular
elements. Solutions obtained from such approximations tend to converge to theoreti-
cal analysis results as the size of elements in shell model decreases (Zienkiewicz and
Taylor 1991). Many commercially available finite element simulation packages offer
the option of flat shell elements only. Formulation for curved shell elements, although
relatively more tedious, has also been implemented in several software packages.
7.7 CURVED SHELL ELEMENTS
Displacement-based iso-parametric formulation developed by Ahmad et al. (1970) is
the standard choice in finite element literature for curved shell elements (Cook et al.
1989, Zienkiewicz and Taylor 1991, Bathe 1996). Three-dimensional geometry of a
shell is represented by its mid-surface following the 3D curvature (Figure 7.10), and
the straight lines in thickness direction are not constrained to remain normal to mid-
surface during deformation, thus, allowing for shear deformation in the material.
Coordinates (x, y, z) of any point inside the element at a distance t from mid-surface
(Figure 7.10) are defined by
x xi
ti
y Hi yi Hi .t. .v3i (7.52)
z z 2
i
where, xi, yi, and zi are Cartesian coordinates of nodes at mid-surface, ti are shell
thickness values at the nodes, v3i is the direction cosine of normal to the mid-surface
at the point under consideration, and Hi are coordinate interpolation functions spe-
cific to the nodal composition of an element geometry. For four-node quadrilateral
v3i i v2i
v1i
i
r
mid-surface
FIGURE 7.10 Curved shell element.
182 Finite Element Analysis of Solids and Structures
element shape, Hi are defined by equations (3.65), and for eight-node elements by
equations (3.85). Similar formulations are also available in finite element literature
for 3 and 6 node triangular elements. Using the same interpolation functions, dis-
placements at a point inside the curved shell (u,v,w) are defined by
u ui
ti i (7.53)
v Hi vi Hi .t. .v3i . v1i v2i
w w 2 i
i
where {ui, vi, wi} are translational deformations of nodes on mid-surface, αi and βi are
rotations of reference straight lines in thickness direction; and the direction cosines
of orthogonal tangents to the mid-surface are denoted by v1i and v2i. Ignoring the
strain in normal direction of mid-surface, strains in (x, y, z) coordinate system are
defined from equations (2.4):
u
x
v (7.54)
x
y
y u v
xy
y x
xz u w
yz
z x
v w
z y
Substituting the displacement expressions from equations (7.53) into (7.54), and
using the coordinate transformation matrix J from equation (5.52), strain-to-nodal
displacement relationships can be written in the following standard form for finite
element calculations:
x
y B ui
(7.55)
xy
xz
yz
where {ui} is the vector of nodal displacement responses (ui, vi, wi, αi, βi). Stress–
strain relationships for curved shell elements are defined by Hookes’ law:
Analysis of 3D Thin-Wall Structures (Plates and Shells) 183
1 0 0 0
1 0 0 0
1
x 0 0 0 0 x
E 2 y C
(7.56)
y 1 v 2 1
xy 0 0 0 0 xy
2
xz 1 xz
yz 0 0 0 0
2
yz
where ξ is a shear correction factor for shear strain energy equivalence between
actual parabolic distribution through thickness and the uniform distribution assumed
in finite element calculations (analogous to the shear stress distribution in beams
represented in equation 6.96). Relationship matrices [B] and [C], from equations
(7.55) and (7.56), are used in equation (2.44) to determine the stiffness matrix of
curved shell element. Numerical integration of equation (2.44) involves calculations
on the element mid-surface as well as through the thickness direction – similar to the
scheme shown in Figure 7.9. Single point integration in the thickness direction does
not capture the bending resistance of a shell, thus, reducing it to a purely membrane
element that can be used to analyze fabric roof systems such as the example shown
in Figure 7.11.
FIGURE 7.11 Membrane roof system in 3D space (Bhattacharjee and Chebl 1997).
184 Finite Element Analysis of Solids and Structures
7.8 SHELL ELEMENT MESH QUALITY AND INTEGRATION RULES
Finite element model preparation steps (a)-to-(g), described in Section 4.4 for 2D
solids, are applicable to shell element model preparation as well. As evident from the
above discussions in this chapter, a shell element is formulated from the superposi-
tion of numerical formulations related to in-plane membrane stress and out-of-plane
bending resistance mechanisms. Element shape quality and integration rule effects,
discussed in Chapters 3 and 4 for 2D solid elements, equally apply to the membrane
stress resistance mechanism of shell element formulation. Quadrilateral shell ele-
ments, flat or curved, should be close to square shape in plan view. Excessive distor-
tion of in-plane geometry significantly affects the quality of membrane stress
response (similar to the effect highlighted for 2D plane-stress elements in Figure
3.12). Element shape quality metrics (Jacobian, aspect ratio, etc.) should be used to
check the quality of shell element mesh before embarking on the journey to actual
analysis task. An additional shape quality metric for shell elements is the warpage of
shell plane as shown in Figure 7.12. When using flat shell elements in modeling
curved surfaces, element warpage should be limited to no more than 10 degrees.
Finer mesh grid with selective use of triangular elements tends to keep the geometric
warpage under control while modeling complex 3D surfaces. The use of triangular
shell element should be kept at minimum as it will make the in-plane bending
response very stiff like its 2D counterpart (discussed in Section 3.9).
Fully integrated four-node quadrilateral shell element (Figure 7.9(b)) will also
generate artificial high resistance to in-plane bending deformation – an inherent side
effect of the linear assumption for in-plane deformation modes (Section 4.2, Figure
4.6). Reduced integration shell elements, one-point in-plane integration with multi-
point through-thickness calculations (Figure 7.13(a)), have been successfully used in
simulations of complex structural responses (Belytschko et al. 1984). Once again, the
draw-back of in-plane reduced integration calculation is the hourglass zero energy
Warped element - 4 nodes not on same plate
FIGURE 7.12 Warpage of shell element geometry.
Analysis of 3D Thin-Wall Structures (Plates and Shells) 185
f4 f3
4 3
x
x
x
1 2
f1
f2
(a) (b)
FIGURE 7.13 (a) Flat shell element with reduced in-plane integration rule; and (b) artificial
numerical resistance addition to counteract zero energy deformation mode.
mechanism developing for in-plane bending mode (Figure 3.13). General-purpose
finite element software packages usually contain internal numerical controls to mini-
mize the risk of developing hourglass response mechanism. For example, the zero-
energy in-plane bending response mechanism can be partially circumvented by
adding artificial hourglass resistance forces (Figure 7.13(b)):
f 1 1
f 2 hc E x 1 (7.57)
f 3 1
f 4 1
where {fi} is the vector of artificially added hourglass resistance forces depending on
the hourglass deformation response at the nodes (Δx), hc is a dimensionless penalty
number (usually between 0 and 0.1), and E is the elasticity of material. Different
variations of the hourglass formulation (equation 7.57) have been implemented in
software packages. For example, LS-DYNA implementation (LSTC.COM 2021)
derives hourglass forces based on nodal velocity response and material viscosity
parameter, in lieu of relative displacement and elastic modulus, in dynamic response
analysis of structures. Irrespective of the formulation details, hourglass resistance
forces, acting at the nodal displacement DOF, are non-physical forces. These forces
do absorb energy during element deformation. Reliable simulation software pack-
ages generally track the amount of artificial energy absorbed by the hourglass resis-
tance mechanism. The reliability of a simulation model result, reporting more than
5% energy loss through hourglass mechanism, should be always questioned. A com-
bination of multi-point integration rule for in-plane and out-of-plane bending mecha-
nisms, with reduced integration rule for in-plane shear response, tends to minimize
the negative effects of full and reduced integration techniques. Theory manuals of
specific software packages generally contain the detailed information on what inter-
nal formulation techniques have been implemented to produce accurate simulation
results.
186 Finite Element Analysis of Solids and Structures
7.9 ANALYSIS OF SHELLS WITH FEA SOFTWARE
Standard finite element simulation models for shell structures generally refer to the
mid-surface geometry that can be easily extracted from 3D CAD data by using gen-
eral-purpose model pre-processing software packages (such as HyperMesh and
ANSA). The extracted mid-surface can be discretized into nodes and elements by
using the surface meshing techniques of the pre-processing software. ABAQUS input
model, for example, requires the standard 3D Cartesian coordinate data for nodes,
with optional inputs for direction cosines of normal to the mid-surface at the node
locations:
*NODES, NSET=setname
N1, x1, y1, z1, l1, m1, n1
N2, x2, y2, z2, l2, m2, n2
…
where xi,yi,zi are Cartesian coordinates of node i, and li,mi,ni are direction cosines of
the normal to mid-surface at that node. If the normals are not explicitly defined as part
of the node definition, ABAQUS internally calculates the normal direction on element
surfaces; and it defines the normal at a node as the average of the direction cosines of
normal to adjoining element surfaces. ABAQUS element library includes curved shell
elements of different nodal configurations: S4 for four-node shells, S3-for three-node
shells, S8 for eight-node shells, etc. Example shell element in Figure 7.14 will be
described in ABAQUS model file by following the standard input syntax:
*ELEMENT, TYPE=S4, ELSET=setname
Element_ID_no, N1, N2, N3, N4
where N1, N2, etc., are the element corner nodes. Number of nodes to describe ele-
ment geometry depends on the shell element types (for example, three nodes for type
+ve normal direc on
N3
4 3
N2
1 2
N4 N1
FIGURE 7.14 Four-node curved shell element with mid-surface normals at the corner nodes.
Analysis of 3D Thin-Wall Structures (Plates and Shells) 187
“S3” and eight nodes for element type “S8”). Reduced integration formulations for
ABAQUS shell elements are selected by using the relevant element type name (for
example, S4R for reduced integration of four-node elements; and S8R for eight-node
elements). Uniform shell thickness and number of integration points through the
thickness direction are defined by the following description:
*SHELL SECTION, MATERIAL=matname, ELSET=setname
thickness value, number of integration points through thickness
Variable thickness in the shell elements can be defined by using the “NODAL
THICKNESS” option in shell section description:
*SHELL SECTION, MATERIAL=matname, NODAL THICKNESS, ELSET=setname
thickness value, number of through-thickness integration
Thickness values at the nodes are described by the *NODAL THICKNESS data
block:
*NODAL THICKNESS
Node ID or Set ID, thickness value
Node ID or Set ID, thickness value
….
The sequence of node numbering in element connectivity description defines the
direction of normal to an element surface (Figure 7.14). A surface load acting in the
direction of normal is given a positive sign in distributed load description with
*DLOAD in ABAQUS models:
*DLOAD
element ID or Set ID, P, surface pressure value
element ID or Set ID, P, surface pressure value
….
7.10 PRACTICE PROBLEMS: LOAD–DEFLECTION ANALYSIS
OF SHELLS
PROBLEM 1
3D geometry of an extruded aluminum C-section component is shown in Figure
7.15. Exterior dimensions of the C-section are 60 mm × 100 mm, with uniform wall
thickness of 4 mm; and the end-to-end length of the member is 1200 mm. Geometry
data of the component is available in the file 3D-C-Beam.stp. External loading,
boundary conditions and material properties have been described in the Figure 7.15.
Conduct finite element analysis of the component using shell elements; and plot the
contour of normal stresses on the x–z section at the mid-span of the member. Estimate,
with hand calculations, the maximum bending stress at mid-span of member, and
comment on the reason for differences between FEA and hand calculation results
188 Finite Element Analysis of Solids and Structures
60 Each end is a ached to s ff
support structures with 4 bolts
100
Extruded aluminum beam of
uniform wall thickness: 4 mm
1000 E=70 Gpa
z 1200 =0.33
y x
Member is subjected to a normal pressure load of
0.02 kN/mm2 (in global x direc on) over an area of
approximately 20x40mm in the mid-span
FIGURE 7.15 Practice Problem-1: stress analysis of a thin-wall C-section member.
PROBLEM 2
Re-analyze thin-walled C-section beam structure in Figure 6.28 with shell elements.
Compare the shell model results with those obtained in Practice problem-2 in
Section 6.11.
PROBLEM 3
Dent resistance of an automotive body panel is measured by a force resistance of
>0.15 kN per 0.1 mm deformation when a normal pressure load is applied uniformly
over a circular area of 51 mm diameter. Conduct dent resistance analysis at the mid-
span of vehicle roof panel shown in Figure 7.1. Assume that 0.9 mm thick aluminum
panel (E = 70 GPa, ν = 0.33) is riveted along the edges to the perimeter frames at
every 60 mm. Curved shell surface of roof panel, with a rise of about 70 mm above
the two far ends, covers an opening area of approximately 1460 × 1050 mm. 3D
geometric profile data of roof panel’s mid-surface is available in support data site
(Roof_Panel.step).
PROBLEM 4
Figure 7.16 shows an automotive bumper beam with its material properties, bound-
ary constraints and an arbitrary external loading condition. Geometric shape data of
the component is available in data file (Bumper_Beam.step).
Develop a finite element analysis model of the structural component using triangular
and quadratic shell elements. Verify the element qualities for shape distortion and
warpage. Conduct analysis of the structure for applied loads and boundary
Analysis of 3D Thin-Wall Structures (Plates and Shells) 189
Shell thickness=2.0 mm, E=206 GPa, =0.32
Constrain all DOF at
z the perimeter of this
hole (approxima on)
y
Constrain all DOF at the perimeter of x
this hole (approxima on)
Apply a uniform normal pressure load of 0.01
kN/mm2over an approximate area of 800 mm2
FIGURE 7.16 Practice Problem-4: Automotive bumper beam (Component design extracted
from vehicle FEA model: courtesy of NHTSA 2020).
conditions; and plot a contour of global directional stresses σyy on the outermost layer
(front surface) of the bumper beam shell. Compare the maximum stress value,
obtained from contour plot, with hand calculated maximum bending stress value at
mid-span assuming a straight beam profile of uniform cross-section.
8 Multi-Component
Model Assembly
SUMMARY
Essential attributes of single-component finite element models have been discussed
in Chapters 1–7. This chapter focuses on specific topics relevant for simulation mod-
els of multi-component assemblies. Element-to-element deformation compatibility,
a special topic of concern during finite element modeling of parts having different
behavioral characteristics, is discussed in Section 8.1. For interface deformation
compatibility, the use of higher order elements is recommended in beam-solid and
shell-solid assemblies. The important topic of modeling discrete inter-body connec-
tions is discussed in Section 8.2. Two commonly used inter-part rigid connection
techniques, master-slave option using the elimination of slave DOF and the kine-
matic formulation based on Lagrangian multiplier approach, are reviewed in that
Section. Discrete finite element definition of connector entity, based on deformable
beam stiffness formulations, is introduced in Section 8.3. Section 8.4 extends this
discrete connection modeling technique to mesh-independent implementation tech-
nique – a highly desirable feature in the preparation of large multi-part complex
models. General part-to-part contact formulations, based on kinematic and penalty
formulations, are introduced in Section 8.5. The contact formulation method is also
a topic of high importance in nonlinear analysis of structures – discussed in Chapter
12. The alternative technique of modeling part-to-part thin-layer interfaces, with spe-
cially formulated solid elements, is presented in Section 8.6. Modular organization
method for database management of large multi-part assemblies is discussed in
Section 8.7. Result quality checks for multi-component finite element simulation
models are briefly discussed in Section 8.8. Finally, a set of practice problems for the
topics discussed in this chapter are presented in Section 8.9.
8.1 ELEMENT COMPATIBILITY AND
CONVERGENCE OF SIMULATION RESULTS
Finite element formulations capable of providing resistance to various deformation
modes have been presented in Chapters 2–7. Iso-parametric formulations inherently
satisfy completeness requirement – that the displacement functions of an element
must represent the rigid body displacements and constant strain states (Bathe 1996).
Another fundamental requirement for monotonic convergence of finite element solu-
tions is that the elements must be compatible. Compatibility is satisfied if the ele-
ments have the same nodes on the common interface, and same interpolation
functions are used in adjoining elements. Structural component analysis examples in
Chapters 1–7 have been modeled with single element types. Single element type
191
192 Finite Element Analysis of Solids and Structures
usage guarantees that the elements in the model undergo compatible deformation
responses. Compatible element selection, with error-free description of nodes, prop-
erties, and boundary conditions, virtually guarantees solution convergence towards
theoretical solution provided that all active DOF in a model experience resistance
from connecting elements and external boundary constraints. Number of active
degrees of freedom in a simulation model depends on the resistance mechanism of
selected element type. Models constructed with elements for 2D response simulation
will have active DOF in the analysis plane only. For example, structural cable in
Figure 2.18, providing resistance to axial deformation only, has been modeled with
two truss elements in two-dimensional space. ABAQUS simulation, for example,
keeps only translational DOF active at the nodes of a truss element model. Two-
dimensional model using beam elements, e.g. the beam analysis model in Figure
6.30, has translational as well as rotational DOF active at the nodes. Automotive
bumper beam problem in Figure 7.16, modeled with general 3D curved shell ele-
ments, has all 6 DOF active at each node. It is imperative that no active nodal DOF
can undergo motion without resistance from connected elements. Some software
packages by default keep all 6 DOF active at the nodes; and a user is required to
explicitly define constraints (“zero” displacement boundary condition) on unused
DOF. Model assembly as a whole unit is also required to have enough constraints
against rigid body motion (zero energy mode). Figure 2.17 illustrates an example of
rigid motion mechanism in an unstable analysis model.
Element-to-element deformation compatibility is not guaranteed when elements
with different deformation response mechanisms are used to define an analysis
model. Figure 8.1 shows an example where a structural assembly is defined by a
combination of beam and cable components. Common joint (c) between beam and
cable will have both translational and rotational motion. Cable element (modeled
with single truss element in 2D space) will not provide any resistance to the rota-
tional DOF at joint (C). However, bending resistance of the beam will provide resis-
tance to that DOF. Input data for models comprising of more than one element type
can be constructed, by including as many different element sets and properties as
needed to describe the complete model. Partial ABAQUS input data for the problem
in Figure 8.1 are listed in the following:
Continuous beam structure- ABC
D 400x200 box section 200
y Wall thickness: 12.5 mm
P=20kN
1000 400
A B C Steel: E=210 GPa, =0.3
x
Cable member- CD
2000 1000
Dia: 76.2 mm
Steel, E=200 GPa
FIGURE 8.1 Mixed element-type assembly of beam and cable
Multicomponent Model Assembly 193
*NODE, NSET=All-nodes
101,0.0,0.0,0.0
102,500.0,0.0,0.0
…..
108,3000.0,1000.0,0.0
*ELEMENT,TYPE=B21,ELSET=beams
1001,101,102
......
1006,106,107
*BEAM SECTION, MATERIAL=mat-beam, ELSET=beams, SECTION=BOX
400,200,12.5,12.5,12.5,12.5
*MATERIAL, NAME=mat-beam
*ELASTIC
210,0.3
*ELEMENT,TYPE=T2D2,ELSET=cable
9001,107,108
*SOLID SECTION, MATERIAL=mat-cable, ELSET=cable
4560.4
*MATERIAL, NAME=mat-cable
*ELASTIC
200,0.3
*BOUNDARY
101,1,6,0.0
108,1,2,0.0
*STEP, PERTURBATION
20 kN vertical downward load on beam
*STATIC
*CLOAD
105,2,-20.0
*END STEP
Beam element in the above analysis model provides resistance to rotational DOF at
node #107 (corresponding to point C in Figure 8.1), while the truss element does not.
This in-compatibility between discrete beam and truss elements does not pose any
theoretical or computational challenge. It implies that element formulations do not
need to be compatible in discrete structural element assembly provided that all DOF
experience resistance to deformation. However, inter-element compatibility is a criti-
cal issue in modeling of continuum solids. It is tempting in engineering analysis to
use beam and solid elements together (e.g. to model streel reinforced concrete struc-
ture), and to use shell and solid elements to model adhesively bonded thin sheet
metal assemblies. Such incompatible element assemblies do not simulate continuous
bonding at the element interfaces as intended when beam or shell undergoes bending
deformation independent of the surrounding solid elements (Figure 8.2(a)). Increasing
refinement of finite element mesh can reduce the effect of element-to-element incom-
patibility along interface boundaries. An effective alternative is to use higher order
elements with mid-side nodes for maintaining interface deformation compatibility
between beam and solid, or between shell and solid (Figure 8.2(b)). Use of specially
194 Finite Element Analysis of Solids and Structures
x
x
Linear solid Bending response of 3-node beam element with
element Euler-Bernoulli beam 8-node solid
(a) (b)
FIGURE 8.2 (a) deformation incompatibility between Euler–Bernoulli beam and four-node
solid sharing common nodes; (b) improved compatibility with higher order elements.
formulated transition elements has been proposed in the literature to connect ele-
ments of in-compatible deformation modes (Bathe 1996).
8.2 MODELING OF KINEMATIC JOINTS IN
STRUCTURAL ASSEMBLIES
Figure 2.10(b) has shown a structural assembly where discrete truss elements are
inter-connected at common nodes. Such structural joints, with fully coupled motion
among the connecting members, can be simply modeled with common nodes with-
out the need for any other special consideration. However, connecting member ends
cannot be assigned to same node when some of the DOF at member ends are coupled
while the others remain uncoupled. Beam assembly in Figure 6.26 represents an
example where the joint between two beams keeps the translational motions coupled;
and the rotational motions at the merging ends of two beams remain uncoupled.
Some structural frame analysis software packages allow the definition of moment
“release” corresponding to local rotational degree of freedom at end connections of
beam elements. General-purpose finite element simulation software packages often
allow the elements, merging at common structural joints, be modeled as independent
entities – meaning that component elements will have their independent nodes; and
the nodes merging to a joint will be coupled by using a kinematic joint entity.
Coincident nodes n2 and n3 in Figure 6.26 can have same coordinates; and be part of
independent elements merging to a common joint entity. Motions of these two nodes
can be partially or fully coupled by defining kinematic constraint. A kinematic con-
straint basically enforces same motion on two or more DOF without allowing relative
motion. Equation (8.1) in the following shows an example constraint formulation
defining zero relative motion between two arbitrary DOF (uj and uk) in a model:
u j
u j uk 0 1 1 0 (8.1)
uk
Multicomponent Model Assembly 195
Multiple kinematic constraints in a large model can be defined by using matrix
and vectors:
e pn . un 0 (8.2)
where {u} is the vector of n number of active DOF in a model and [e] is p × n matrix
defining p number of constraint relationships (similar to that of equation 8.1) involv-
ing n number of DOF in a model. Finite element system equilibrium equation (1.5)
and kinematic constraint relationships (8.2) can be combined to the following system
of equations by using p number of Lagrangian multipliers (λ):
K n n enT p un Pn
(8.3)
e p n 0 p 0
As long as constraint equations are linearly independent and p < n, standard matrix
equation solvers can be used to solve equations (8.3) for the unknown displacements
and Lagrangian multipliers (Bathe 1996). Lagrangian multipliers are in fact the inter-
nal force values required to enforce kinematic constraint conditions described by
equation (8.2).
As discussed in Section 6.10, with ABAQUS input description for the example
beam joint in Figure 6.26, nodes n2 and n3 are inter-connected with a kinematic con-
nector element CONN2D2 with property type “JOIN”.
*ELEMENT, TYPE=CONN2D2, ELSET=set2
El-id, n2, n3
*CONNECTOR SECTION, ELSET=set2
JOIN
This kinematic connector element (“JOIN”) enforces same translational motion
among the two joining nodes, while the rotational degrees of freedom at those nodes
remain independent. Several other connector element types exist in ABAQUS library
for defining node-to-node kinematic relationships. Kinematic connector elements
can be considered as virtual elements that apply internal forces (represented by
Lagrangian multipliers in equation 8.3) to eliminate relative motion among selected
pair of DOF. These internal forces in connector elements can be post-processed for
joint integrity assessment. Not all finite element analysis packages implement con-
straint formulation based on the Lagrangian multiplier approach.
Kinematic constraint definition, without involving connector elements, has a dif-
ferent implementation mechanism. In lieu of using the Lagrangian multiplier tech-
nique, general kinematic constraint formulations eliminate the slave degrees of
freedom in most finite element implementations. Once a displacement degree of free-
dom is eliminated because of its “slave” relationship to a master reference DOF,
additional displacement constraints (such as boundary conditions or other kinematic
coupling definitions) cannot be applied to that again. Constraint definition based on
elimination process of slave DOF permits the definition of constraints involving mul-
tiple slave DOF – commonly known as multi-point-constraint (MPC) in finite
196 Finite Element Analysis of Solids and Structures
element literature. An MPC limits the motion of a group of nodes (often referred to
as “slave” nodes) to that of a “master” or “reference” node. The reference (master)
node has both translational and rotational degrees of freedom, and it can be subjected
to boundary constraints if needed. ABAQUS input example for easy-to-define MPC
type is presented in the following:
*KINEMATIC COUPLING, REF NODE=ref_node_id
ni, nj,…. etc. or name of slave node set
In the above example, all available DOF in the slave nodes are constrained to the
DOF of the reference node. A reduced subset of the DOF can also be constrained by
specifying the first and last DOF of the desired range:
*KINEMATIC COUPLING, REF NODE=ref_node_id
slave nodes, first dof, last DOF
Kinematic constraint can also be applied between a reference node and a set of ele-
ments on a surface:
*COUPLING, CONSTRAINT NAME=a-name, REF NODE=ref_node,
SURFACE=s-name
*KINEMATIC
first dof, last DOF
*SURFACE, NAME=s-name, TYPE=ELEMENT
element ID or element set name
A group of nodes located on a surface can also be constrained to a reference node.
Two surfaces can also be tied together. Each node on the first surface (the slave sur-
face) will have the same values for its degrees of freedom as the point on the second
surface (the master surface) to which it is closest. Discrete bonding of two or more
surfaces, such as spot welding between sheet metals, can also be modeled using
kinematic constraints; but it is more useful to model spot welds by using deformable
connector elements (as discussed in Section 8.3).
Multi-point constraint definition in some software packages is identified as “rigid
body” that binds a set of elements or a set of nodes to a reference node whose motion
governs the motion of the entire rigid body. Rigid bodies can be used to model very
stiff components, either fixed or undergoing large motions. The principal advantage
of representing portions of a model with rigid bodies, rather than with deformable
finite elements, is computational efficiency. Although some computational effort is
required to update the motion of the nodes of the rigid body, and to assemble the
concentrated and distributed loads, the motion of the rigid body is determined com-
pletely by a maximum of six degrees of freedom at the reference node. Element-level
calculations are not performed for elements that are part of a rigid body, thus making
the rigid-body definition very appealing for achieving computational efficiency in
simulations of multi-body dynamic problems. For example, in complex models, ele-
ments far away from the particular region of interest could be included as part of a
Multicomponent Model Assembly 197
rigid body, resulting in faster run times. Although the motion of the rigid body is
governed by the six degrees of freedom at the reference node, rigid bodies allow
accurate representation of the geometry, mass, and rotary inertia of the parts.
ABAQUS input syntax for a rigid body definition is shown in the following:
*RIGID BODY, REF NODE=ref_node_id
slave element set or node set
A reader can review users’ guide of specific finite element software package, such as
that of ABAQUS (Dassault Systems 2020b), to identify input syntax for different
types of MPC definition. A general-purpose MPC definition enforces rigid body
motion by eliminating the DOF at the slave nodes. As described earlier, connector
elements in ABAQUS achieve the same objective without eliminating the slave DOF.
Those elements provide the added advantage of calculating internal connector forces
that can be monitored to assess the integrity of a joint. General-purpose MPC or
RIGID BODY definitions, relying on the process of slave DOF elimination, do not
provide the capability to monitor forces transmitted through the joint. The preference
of joint element use, either kinematic formulation based on Lagrangian multiplier
approach or the rigid body MPC approach, depends on the analysis objective. Large
assemblies of relatively flexible structural components, such as the vehicle body
structure shown in Figure 8.3, possess several thousand discrete spot welds that can
FIGURE 8.3 Example of an automotive body structure joint with spot-welds (Figure
prepared from FEA model of vehicles – courtesy of NHTSA 2020).
198 Finite Element Analysis of Solids and Structures
(a)
(b)
FIGURE 8.4 (a) Schematic view of joint details in a structural frame; and (b) idealized rep-
resentation of a deformable joint element connected to primary frame members.
be modeled with rigid kinematic constraints to assess the overall system stiffness.
However, structural integrity assessment at critical joint locations requires a more
detailed stress–strain analysis of the joints themselves. Inherent flexibility of struc-
tural construction joints, such as the ones used in building frames (Figure 8.4), also
requires an explicit representation of the non-rigid joint properties. Simulation of
these structural joints, using kinematic constraint-type rigid elements, often leads to
over-estimation of system stiffness. In practical finite element simulations, flexible
joints are modeled by using special joint elements to approximately represent the
deformable characteristics of physical joints. Section 8.3 in the following presents an
enhanced form of connector element formulations using deformable finite element
properties.
8.3 DEFORMABLE JOINT ELEMENTS FOR
PART-TO-PART CONNECTIONS
Figure 8.5 shows a weld joint element “ab” connecting two plates “ABCD” and
“PQRS”. Local coordinate directions at end “a” are defined by the unit vectors
e1,e2,e3 shown in the figure, where direction e3 is normal to the plane ABCD. Similarly,
local directions at end “b” can be defined by another independent set of unit vectors
with local direction 3 being normal to the plane PQRS. End “b” can be assumed to
have same local coordinate directions of end “a” for the special case when two planes
(ABCD and PQRS) happen to be parallel. Each end point of connector element
(a and b) can have up to six degrees of freedom – three translations and three
Multicomponent Model Assembly 199
S R
D Q C
P
e3 e2
a e1
A B
FIGURE 8.5 A weld element (ab) joining two plates ABCD and PQRS.
rotations. Displacements of points a and b are generally calculated by interpolating
the responses of master surface nodes ABCD and PQRS, respectively. Stiffness prop-
erties of joint element ab can be formulated by considering 1–6 relative deformation
modes between ends “a” and “b”. Measuring the relative position of point “b”, with
respect to point “a”, in local projection directions e1, e2, e3, translational deformation
inside the joint element is defined by the following equation:
u1 x x0 ; u2 y y0 ; u3 z z0 (8.4)
where x0, y0, z0 are the initial coordinates of node “b” relative to node “a” along the
initial directions (e1, e2, e3), and x, y, z are the relative position of point “b” after defor-
mation. Similarly, relative rotational deformation of point “b” with respect to point
“a” can be defined as
u4 1 01; u5 2 02 ; u6 0 (8.5)
where α1, α2, and β are flexural and torsional deformation angles; and α 0 , α 0 , β0 are
1 2
the corresponding initial values. Input data syntax for describing 3D general joint
element behavior in ABAQUS (ABAQUS Connector Elements, Dassault 2020b),
considering both translational and rotational deformation modes, is given in the
following:
*CONNECTOR SECTION, BEHAVIOR=ab_elastic, ELSET= setname
PROJECTION CARTESIAN, PROJECTION FLEXION-TORSION
orientation_a
orientation_b
where the option “PROJECTION CARTESIAN” specifies translational deformation
calculations in the projected local vector directions following equation (8.4); and
option “PROJECTION FLEXION-TORSION” specifies rotational deformation cal-
culations following equation (8.5). Vector orientations at points “a”, named
“orientation_a”, can be defined by listing 3 nodes as in the following:
200 Finite Element Analysis of Solids and Structures
*ORIENTATION, DEFINITION=NODES, NAME=orientation_a
N1,N2,Node_ID_a
where Node_ID_a specifies the node number of joint end point “a” as the origin of
local coordinate system, node ID N1 defines the local vector direction e1, and N2
defines the vector direction e2 (Figure 8.5). Direction vectors for end connection
point “b” can be defined independently, or it can follow the same definition of point
“a”. The orientation directions co-rotate with the rotation of the node to which they
are attached. In element response calculations, the local directions are “centered”
between the two local definitions at points “a” and “b”.
In commonly used two-node joint element formulations, stiffness properties for
the 6 internal joint deformation modes, described by equations (8.4) and (8.5), are
defined by uncoupled stiffness terms as shown in the following equation:
k11 0 0 0 0 0
0 k22 0 0 0 0
0 0 k33 0 0 0
k ab (8.6)
0 0 0 k44 0 0
0 0 0 0 k55 0
0 0 0 0 0 k66
where the first three diagonal terms represent stiffness values against three transla-
tional deformation modes of the joint element, and the remaining three correspond to
rotational deformation modes. Spring-like uncoupled stiffness property definition
(equation 8.6) makes the joint element quite adaptable to special kinematic situa-
tions. For example, a moment-free deformable hinge connection can be simulated by
assigning zero value to stiffness term corresponding to the relative twist between two
end nodes. In equation (8.5), sixth deformation mode (u6) represents that relative
twist, meaning that k66 in equation (8.6) needs to be assigned zero value to create a
torsion-free connection. In some software implementations, first local DOF is defined
in the direction defined by end nodes “a” and “b”, and the fourth diagonal stiffness
term in equation (8.6) is associated with the relative twist mode of connector ele-
ment. Joint element details are generally documented in software-specific users’
manuals. ABAQUS input syntax for describing the stiffness terms of equation (8.6)
is given in the following:
*CONNECTOR BEHAVIOR, NAME= ab_elastic
*CONNECTOR ELASTICITY, COMPONENT=1
k11
*CONNECTOR ELASTICITY, COMPONENT=2
k22
*CONNECTOR ELASTICITY, COMPONENT=3
K33
*CONNECTOR ELASTICITY, COMPONENT=4
Multicomponent Model Assembly 201
K44
*CONNECTOR ELASTICITY, COMPONENT=5
K55
*CONNECTOR ELASTICITY, COMPONENT=6
K66
Stiffness properties in the above example have been assigned to an element set named
“setname” which can contain one or more connector elements. Number of stiffness
terms to be used depends on the formulation of specific joint elements. Connector
behavior, described earlier with “PROJECTION CARTESIAN” and “PROJECTION
FLEXION-TORSION” options, represents the most general connector element
requiring all 6 stiffness terms. The description of connector elements in ABAQUS
follows the standard element data input syntax:
*ELEMENT, TYPE=CONN3D2, ELSET=setname
Joint_Element_id_no, node_id_a, node_id_b
….
When no connector section data card (*CONNECTOR SECTION) is defined for an
element set, ABAQUS uses the Lagrangian multiplier-based kinematic constraint
formulation for the connector elements. DOF at end point “a” of connector element
can be coupled to the nodes of base part ABCD by using the kinematic coupling
formulation (*KINEMATIC COUPLING) described in Section 8.2:
*KINEMATIC COUPLING, REF NODE= node_id_a
NA, NB,…. etc. or name of node set on part ABCD
Similarly, DOF of node “b” can be coupled to the nodes of other base part PQRS.
Stiffness terms defined in equation (8.6), multiplied by the relative deformation mea-
surements from equations (8.4) and (8.5), provide the values of forces transferred
through the connector element.
8.4 MESH-INDEPENDENT FASTENERS FOR PART-TO-PART
CONNECTIONS
Many applications require the modeling of point-to-point connections between parts.
These connections may be in the form of spot welds, rivets, screws, bolts, or other
types of fastening mechanisms. There may be hundreds or even thousands of these
connections in a large system model such as an automotive or airplane body struc-
ture. The fastener can be located anywhere between the parts that are to be connected
regardless of the mesh. In other words, the location of the fastener can be indepen-
dent of the location of the nodes on the surfaces to be connected. For example, con-
nector element end points “a” and “b” in Figure 8.5 may need to be defined
independent of the nodal points in base parts ABCD and PQRS. Meticulous manipu-
lation of nodal positions in connected parts, to define a properly aligned pair of nodes
202 Finite Element Analysis of Solids and Structures
on base parts, is not practical in complex structural simulation models involving
several thousand inter-part connector elements. Use of kinematic coupling option to
couple the DOF of end points of connector element “ab” (Figure 8.5) with the base
part nodes, described in Section 8.3, provides a basic form of mesh-independent con-
nector element description – meaning that connector element “ab” is defined at
desired location independent of the mesh patterns of connected parts ABCD and
PQRS. Mesh-independent description of connections is a basic feature required in
finite element software packages for large-scale complex structural simulations.
Many software packages provide very similar features for automatic generation of
connection elements at user-specified locations. The number of individual connec-
tion elements is equal to the number of reference points specified by the user. Built-in
functions in software packages identify the local nearest element surfaces to be con-
nected including the connector orientation direction and kinematic coupling of DOF
between connector element nodes and connected part nodes. In general, a reference
point should be located as close to the surfaces being connected as possible. The
reference node specifying the reference point can be one of the nodes on the con-
nected surfaces or can be defined separately. ABAQUS determines the actual points
where the fastener layers attach to the surfaces that are being connected by first pro-
jecting the reference point onto the closest surface. By default, ABAQUS projects
each fastener reference point onto the closest surface along a directed line segment
normal to the surface. Alternatively, a user can specify the projection direction if
desired. Once the first connection point on the closest surface has been identified, the
points on the other connection surface are determined by projecting the first connec-
tion point onto the other surface along the normal direction from the first connection
point. The surfaces to be fastened can be specified using two different approaches –
either by directly specifying the surfaces to be connected, or by letting the software
package automatically identify the surfaces within a pre-defined search distance
from the reference nodes. For surface-to-edge or edge-to-edge connections between
parts, special optional attachment methods may need to be invoked.
Software packages allow either rigid or deformable connection element types to
be included in automatic generation process. For example, deformable connector
element “ab” in Figure 8.5 can be automatically generated in ABAQUS by specify-
ing the following set of commands:
*NSET, NSET=rfn
Node_ID_a
*CONNECTOR SECTION, BEHAVIOR=ab_elastic, ELSET= ss1
PROJECTION CARTESIAN, PROJECTION FLEXION-TORSION
orientation_a
** BEHAVIOR= ab_elastic refers to the connector stiffness values defined earlier
in section 8.3
*FASTENER, INTERACTION NAME=nm1, PROPERTY=pp1, ELSET=ss1,
REFERENCE NODE SET=rfn
**blank line for default software choice of the projection direction
surface_1_name, surface_2_name
*FASTENER PROPERTY, NAME= pp1
Multicomponent Model Assembly 203
radius of connector element (a numerical value)
**blank line to indicate all DOF of connector node to be tied to base part
NSET=rfn refers to a set of nodes (one or more) to be used as reference nodes for
automatic generation of connector elements. ELSET=ss1 refers to an empty set of
connector elements to be generated by ABAQUS at the reference node locations.
Element surfaces to be connected are optionally identified by “surface_1_name” and
“surface_2_name” in the above example, where these surfaces are predefined by
*SURFACE command as discussed in Section 8.2. Alternatively, a user can specify
a search radius in the *FASTERNER definition (SEARCH RADIUS=R) to let
ABAQUS search for element surfaces that fall within a sphere of user-specified
radius R with its center at the reference point. Projection of reference point onto the
closest element surface is taken as the first point of connector element, and the local
orientation direction e3 (Figure 8.5) of connector is defined in the local normal direc-
tion of that closest surface. A user can override the default local system by specifying
a local coordinate system name with “ORIENTATION=orientation_name” in the
*FASTERNER definition card. Generally, the user-defined orientation should be
such that the local e3 direction of the orientation is approximately normal to the sur-
faces that are being connected; and the local e1 and e2 directions are approximately
tangent to the surfaces that are being connected. By default, ABAQUS adjusts the
user-defined orientation such that the local e3 direction for each fastener is normal to
the surface that is closest to the reference node for the fastener.
In automatically generated connector elements, the DOF of connection points are
coupled to the average translational DOF of the connecting part nodes within a radius
of influence. Rotational DOF of the connecting part nodes can also be included in
coupling definition by using optional parameter “COUPLING=STRUCTURAL” in
*FASTENER definition. The radius of influence, for the connected surface nodes to
contribute to the motion of connection point, can be defined by specifying a numeri-
cal value with optional parameter “RADIUS OF INFLUENCE=” in *FASTENER
definition. In absence of this parameter, ABAQUS computes a default value of the
radius of influence internally, based on the fastener diameter (defined in *FASTENER
PROPERTY) and the characteristic lengths of elements on connected part. The con-
nector elements are given internally generated element numbers and assigned to the
named user-specified element set (ELSET=ss1 in the above example). Multiple con-
nector element sets may need to be defined to specify unique stiffness values (equa-
tion 8.6) depending on the mechanical properties of the actual fasteners (bolts, rivets,
spot welds, clinched joints, etc.).
8.5 SIMULATION OF PART-TO-PART INTERFACE CONTACTS
Initially separate bodies, rigid or deformable, may come into contact during analysis
of load–deformation response. Changes to the interface contact condition in finite
element simulation models can be simulated by tracking the relative motion between
bodies. It is customary, in simulation software packages, to designate body parts as
slaves and master (Figure 8.6), where the motions of slave part nodes are tracked
with respect to the positions of surrounding master element surfaces. Lagrange
204 Finite Element Analysis of Solids and Structures
Slave
nodes
Master
surface
Relative motion
between master
and slave parts
Contact forces preventing
Relative motion slave node penetration into
between master the master surface
and slave parts
Slave
nodes
FIGURE 8.6 Master-slave concept in contact simulation of solid bodies.
multiplier method, discussed in Section 8.2, can be used to constrain the penetration
of slave nodes into the master surface in the normal direction. Potential contacting
bodies are defined upfront in ABAQUS simulation models by using the following
data input syntax:
*CONTACT PAIR, INTERACTION=interaction_property_name
TYPE=SURFACE TO SURFACE
slave_surface_name, master_surface_name
In contact pair formulation, the Lagrange multiplier method is used for contact
enforcement. In alternative contact formulations, commonly known as the penalty
method, the relative penetration of slave node to master surface (deformation DOF u3
in equation 8.4) is minimized by using a large contact stiffness value (k33 in equation
8.6). With this method, the contact force is proportional to the penetration distance;
so some degree of penetration will occur. Penalty force on slave node is calculated
as:
Fpenalty k33 .u33 (8.7)
An equal and opposite force is distributed on the nodes of master element surface.
The penalty stiffness value 𝑘33 for a shell contact surface can be defined in terms of
the bulk modulus 𝐾i, and the face area 𝐴i of the contact element (LS-DYNA Theory
Manual, LSTC.COM 2021):
K i . Ai
k33 fi (8.8)
max shell diagonal
Multicomponent Model Assembly 205
where fi is a scale factor for the interface penalty stiffness. ABAQUS/Standard, by
default, sets the penalty stiffness to 10 times the representative underlying element
stiffness. A low penalty stiffness typically results in better convergence of the solu-
tion, while the higher stiffness keeps the overclosure at an acceptable level as the
contact pressure builds up. In LS-DYNA, the default option is to use the lower bulk
modulus value of master and slave parts. This formulation may lead to a very low
penalty stiffness value with undesirable inter-part penetrations during simulation of
contacts between hard and very soft parts. Adjustment of penalty stiffness values to
simulate contact between hard and soft bodies varies from software to software.
Nonlinear contact stiffness formulations, as a function of the gap closure, can be used
for improved stability of simulation models. In the nonlinear penalty method, the
penalty stiffness increases linearly between regions of constant low initial stiffness
and constant high final stiffness, resulting in a nonlinear pressure-overclosure rela-
tionship. Advantages of the penalty method, over Lagrange multiplier method,
include solver efficiency and mitigation of over-constraint. ABAQUS input data syn-
tax for selecting penalty method with linear contact stiffness formulation is described
in the following:
*SURFACE INTERACTION, NAME=any_name
*SURFACE BEHAVIOR, PENALTY=LINEAR
In the linear penalty method, the so-called penalty stiffness is constant, so the pres-
sure-overclosure relationship is linear. Relative sliding between contacting bodies
can be considered, including frictional force transfer between the contacting bodies,
by defining a friction coefficient in contact property definition:
*SURFACE INTERACTION, NAME= any_name
*FRICTION
0.4
where 0.4 is an example friction coefficient value assigned to “any_name”.
In addition to the paired surface contact definition, described earlier, ABAQUS
software also offers a general contact option to model surface-to-surface contact,
edge-to-surface contact, edge-to-edge contact, and vertex-to-surface. General con-
tact can also deal with self-contact as shown graphically in Figure 8.7. General con-
tact uses the penalty method to enforce the contact constraints by default. The
simplest way to define an all-inclusive general contact in ABAQUS is described in
the following:
*CONTACT
*CONTACT INCLUSIONS, ALL EXTERIOR
where “ALL EXTERIOR” option includes all exterior element faces and all analyti-
cal rigid surfaces in the general contact definition. Commercially available software
packages also offer an alternative convenient method of specifying the contact enti-
ties by using a pre-defined initial domain. Such domains, defined with a box, identify
206 Finite Element Analysis of Solids and Structures
“penalty” forces are introduced
within a part itself if nodes tend to
penetrate though the master
surface of part
FIGURE 8.7 Self-contact in a body undergoing large deformation.
all parts that can potentially come into contact with one another during deformation.
Master-slave relations are automatically defined by ABAQUS in general contact for-
mulations. In ABAQUS/Standard, traditional pair-wise specifications of contact
interactions generally result in more efficient analyses as compared to an all-inclu-
sive self-contact approach of defining all potential contacts. Therefore, there is often
a trade-off between ease of defining contact and analysis performance.
In model preparations for contact simulation, initial penetration between slave
and master parts should be minimized, and physical intersection between contacting
parts must be avoided with no exceptions (Figure 8.8). Initial positions of part sur-
faces must be placed with adequate offset, taking account of the part thickness val-
ues, to minimize penetration problems during model preparation. Software packages
generally offer options to consider the reduced thickness of parts in contact interfer-
ence calculations, thus, preventing numerical issues arising from artificial penetra-
tion and intersection issues. Refined mesh density is generally helpful for modeling
parts with high curvature.
2
A B A B
2
1 3
1 3
(a) Penetration of node 2 into (b) Intersection between
physical thickness space of part AB parts 1-2-3 and AB
FIGURE 8.8 Initial penetration and intersection in finite element models.
Multicomponent Model Assembly 207
8.6 THIN-LAYER INTERFACE ELEMENTS
Simulation of body contacts based on gap closure criterion, described in Section 8.5,
involves numerical treatments internal to the software package in use. However,
engineering analysis of bodies with embedded pre-existing joints requires an explicit
definition of joint elements in the finite element simulation models. Interface joint
regions (Figure 8.9) can be modeled with standard solid elements with unique mate-
rial properties specific to the known behavior of joints. Simulation of joints with
standard finite element formulations produces reasonably good results for mostly
elastic systems with very small deformations occurring at the interface joints.
However, standard finite element formulations, with coupled deformation response
involving all nodal DOF, produce artificially stiff response when separation and slid-
ing deformations inside joints dominate the system response. Special finite element
formulations have been proposed in the literature for realistic simulation of joint
separation and sliding that are encountered in soil-structure interfaces, rock joints,
and construction joints in concrete dams (Desai et al. 1984, Tinawi et al. 1994).
General-purpose finite element software packages often contain special element
formulations to simulate joint behavior. ABAQUS, for example, offers a library of
cohesive joint elements to model the behavior of adhesive joints, interfaces in com-
posites, and other situations where the integrity of interface joints may be of interest.
The connectivity of cohesive elements is like that of continuum elements, but it is
useful to think of cohesive elements as being composed of two faces separated by a
thickness (Figure 8.10). The relative motion of the bottom and top faces measured
along the thickness direction (local 3-direction for three-dimensional elements) rep-
resents opening or closing of the interface. The relative change in position of the
bottom and top faces measured in the plane orthogonal to the thickness direction
quantifies the transverse shear behavior of the cohesive element. In-plane membrane
stresses are generally not considered in interface joint element formulations. In
Body-1
Shell-1
Shell-2
Body-2 Thin-layer joints
(a) (b)
FIGURE 8.9 Pre-existing thin-layer joints: (a) between solid bodies, and (b) between thin
shells.
208 Finite Element Analysis of Solids and Structures
Thickness
direc on Top surface
n8 n7
n4 n3
n5 n6
n1 n2
3 Bo om surface
2
1
FIGURE 8.10 Finite element definition of a thin-layer interface element.
three-dimensional problems, the continuum-based constitutive model assumes one
normal (through-thickness) strain, and two transverse shear strains at a material
point. Element example, shown in Figure 8.10, can be described in ABAQUS as
follows:
*ELEMENT, TYPE=COH3D8, ELSET=set_1
Eelement_ID, n1, n2, n3, n4, n5, n6, n7, n8
where nodes n1, n2, n3, n4 define the bottom surface and nodes n5, n6, n7, n8 define the
top surface of interface element. The thickness direction is automatically determined
by ABAQUS in normal direction from bottom surface to the top. A user can also
define the local direction by using the optional ABAQUS command *ORIENTATION
if desired. The interface zone is discretized with a single layer of cohesive elements
through the thickness. Both top and bottom surface nodes of an interface element
layer can be tied to the neighboring base material parts by using common node defi-
nitions. More generally, when the mesh in the interface zone does not match with the
mesh of adjacent components, nodes of interface elements can be tied to adjacent
base material nodes by using kinematic constraint formations described in Section
8.2. The constitutive behavior of interface cohesive elements in ABAQUS is defined
by using the following property definition:
*COHESIVE SECTION, ELSET=set_1, MATERIAL=mat_1,
RESPONSE=CONTINUUM
The choice of “RESPONSE=CONTINUUM” specifies to ABAQUS that interface
element will have one normal strain (in thickness direction) and two transverse shear
strains in a plane perpendicular to the thickness direction. The elasticity matrix defin-
ing the stress–strain relationship at material points inside the joint element can be
defined by diagonal terms only – meaning uncoupled normal and shear deformation
modes. By default, cohesive elements retain their resistance to compression even if
their resistance to other deformation modes is completely degraded. As a result, the
cohesive elements resist inter-penetration of the surrounding components even after
the cohesive element has completely degraded in tension and/or shear. This approach
works best when the top and the bottom faces of the cohesive element experience
small relative sliding.
Multicomponent Model Assembly 209
8.7 MODULAR ORGANIZATION OF DATA IN MULTI-
COMPONENT MODEL ASSEMBLY
Single-component analysis models in Chapters 1–7 have been presented by describ-
ing all relevant model data in single files. A complex structural system (e.g. the vehi-
cle model shown in Figure 7.1) may involve may different components, with many
different element types and inter-connections. Description of several component
model entities in one single file makes the management of model database very chal-
lenging, particularly when frequent model updates are required because of changes
happening in a handful of parts. This challenge is mostly circumvented by the gen-
eral-purpose finite element simulation software packages by allowing the assembly
of large complex models from much smaller modular construction units. Basic geo-
metric characteristics of components and subsystems may reside in separate model
files while a master file will assemble all those into a single analysis model with
relevant specifications for loads and boundary conditions. Component model files
must adhere to certain modeling rules and restrictions to avoid conflicts; and also to
facilitate easy execution of analysis exercises for multiple attributes. For example,
finite element model of the vehicle roof panel in Figure 7.1 can be described in a data
file (named Roof_Panel.inp) following ABAQUS input syntax:
*NODE, NSET=Roof_panel_nodes
100001,0.0,0.0,1000.0
100002,5.0,5.0,1000.0
.....
*NSET, NSET=Roof_panel_boundary_nodes
800001, 800002, ….
…..
*ELEMENT, TYPE=S4, ELSET=Roof_panel_shell_elements
100001,100051,100099, 110026, 120345
......
*ELSET, ELSET=Roof_panel_mid_span_elements
731001, 731049, …..
……………..
*SHELL SECTION, MATERIAL=Mat_roof_panel, ELSET=
Roof_panel_shell_elements
1.0, 5
*MATERIAL, NAME= Mat_roof_panel
*ELASTIC
210,0.3
*DENSITY
7.8e-06
Above data descriptions include nodes, elements, and material properties of the
roof panel as well as entity sets of Roof_panel_boundary_nodes (list of all nodes
corresponding to the attachment points of roof panel to the vehicle body structure)
and Roof_panel_mid_span_ elements (list of a subset of shell elements at the
210 Finite Element Analysis of Solids and Structures
mid-span of roof). Entity sets can be used to define loads, boundary conditions, con-
nector elements, and contacts as needed in subsequent analysis models. Additional
entity sets of nodes and elements can be defined that may or may not be used in
subsequent analysis models. A simple load–deflection analysis model of the vehicle
roof panel can be constructed in a separate ABAQUS input file by using the finite
element model file Roof_Panel.inp:
** Main analysis file for load–deflection analysis of vehicle roof panel
*INCLUDE, INPUT= Roof_Panel.inp
*BOUNDARY
Roof_panel_boundary_nodes, 1, 6,
*STEP, PERTURBATION
*STATIC
*DLOAD
Roof_panel_mid_span_elements, P, -0.01
*END STEP
Above analysis model includes the component model file (Roof_Panel.inp), and
an analysis model is constructed by using the node and element entity sets defined in
the component model file. Any future changes to the design and finite element model
of the component model will be contained within the component input data file only,
thus keeping the main analysis file un-affected. Model assembly procedure, using
*INCLUDE option, can be expanded to include multiple component models in a
more complex model assembly:
** Main analysis file for load–deflection analysis of a multi-part assembly
*INCLUDE, INPUT= Part_1.inp
*INCLUDE, INPUT= Part_2.inp
…….
*BOUNDARY
Node_set_1, 1, 6,
Node_set_2, 1, 6,
….
*STEP, PERTURBATION
*STATIC
*DLOAD
Element_set_1, P, -0.01
Element_set_2, P, +0.03
….
*END STEP
Model entities (nodes, elements, etc.) are defined by using pre-defined unique ID
ranges in each component model to preclude conflicting IDs in the main assembly
model. When a component model is changed during design iterations, changes are
made to the relevant component input data file only, thus, keeping other components
as well as the main analysis model practically unchanged.
Multicomponent Model Assembly 211
8.8 RESULT QUALITY CHECKS
Result quality checks, discussed in Sections 3.10 and 4.5 for single-component finite
element analysis models, are equally applicable to multi-component model assem-
blies as well. Multi-component models, however, present additional challenges due
to discontinuous nature of stress–strain distributions at the interfaces and joints.
Patch-test-like standard numerical simulation tests are not readily useful for quality
checks of finite element models involving contacts and joints. Engineers can, how-
ever, make use of the powerful graphical visualization features of post-processing
software tools that are often bundled with actual FEA solver products (such as
LS-Pre-Post with LS_DYNA solver and ABAQUS-CAE with ABAQUS solver).
Result database files, produced by the commonly used FEA solvers, can also be visu-
alized by using general-purpose post-processing software packages, such as
HyperView (Altair.com), META (Beta-CAE.com), and FEMAP (Siemens.com). For
example, Figure 8.11 shows a snapshot of part-to-part penetrations occurring during
simulation of load–deformation response – indicating potentially missing contact
definition during the model preparation.
As discussed in Sections 8.2 and 8.3, simulation of inter-part connectivity by
using deformable joint elements, or by using Lagrange-multiplier-based connector
formulations, can directly produce output results of interaction forces flowing
through these elements. However, commonly used alternative MPC formulations
enforce slave-master constraint relationships by eliminating the slave DOF, thus
missing the direct measurement of interaction forces. This shortcoming is overcome
by defining virtual cut sections through structural assembly models for monitoring
the internal forces on cut-sections. Ignoring the body force effects in the internal
stress equilibrium condition (Fi = 0 in equations 1.10), resultant force effects on a
plane cutting through the internal stress field of multi-element assembly (Figure 8.12)
can be calculated by equation (8.9):
Missing contact Vehicle bumper
defini on between system impact
bumper beam and
body structure part
FIGURE 8.11 Part-to-part penetration during simulation of load–deformation response.
212 Finite Element Analysis of Solids and Structures
Internal stresses caused by
external forces and boundary
constraints
FIGURE 8.12 Internal stresses on the cut-plane of a finite element model.
P B . .dV
T
(8.9)
where {σ} is the vector of internal stresses at element integration points, [B] the
strain-displacement relationship matrix (equation 2.6), and {P} the vector of inter-
nals forces and moments on the cut-plane – calculated by summing the contributions
of all finite elements traversing the user-defined virtual cut-plane. Graphical post-
processors often include features to define virtual cut-planes for calculating internal
forces from the finite element stress results saved in database files of FEA solvers.
Commonly used FEA solvers also usually allow pre-defined virtual cut-planes in the
model input files. Internal forces calculated on those pre-defined virtual planes can
be optionally saved on the result database files, and those can be subsequently post-
processed by the graphical visualization software packages.
ABAQUS provides a simple method to create such an interior surface over the
element facets, edges, or ends by cutting through a region of the model with a plane.
The region can be identified using one or more element sets. The virtual cutting plane
is defined by first specifying a point on the plane and a vector normal to the plane.
ABAQUS then automatically forms a surface close to the specified cutting plane by
selecting the element facets, edges, or ends of the continuum solid, shell, membrane,
surface, beam, truss, or rigid elements in the selected region. The surface generated
in this manner is an approximation for the cutting plane. ABAQUS input data syntax
for defining a cut plane is described in the following:
*SURFACE, NAME=surface_name, TYPE=CUTTING SURFACE
x1, y1, z1, l,m, n
elset-1, elset-2, ….
where coordinates (x1, y1, z1) define the position of cut-plane and (l, m, n) define the
direction cosines of normal to the cut-plane. A virtual internal surface is generated by
cutting through the element sets named elset-1 and elset-2, etc. A blank data line
(without any element set name) generates a surface by cutting the whole model. Only
the element nodal forces that lie on the positive side as defined by the normal to the
Multicomponent Model Assembly 213
cutting plane are included in the calculations. Integrated force and moments values
on the virtual cut-plane can be requested in ABQUS by using the following output
request command:
*INTEGRATED OUTPUT, NAME=surface_name
The integrated surface can also be defined at the interface between parts, and the
output forces can be used to assess the force transmitted through contact between the
parts.
The integrated output section definition does not impose any constraint on the
component nodes. The average motion of nodes on the defined surface can be moni-
tored by including a reference node definition.
8.9 PRACTICE PROBLEMS: ANALYSIS WITH
MULTI-COMPONENT MODEL ASSEMBLIES
PROBLEM 1
Re-analyze Problem-2 (Figure 6.28), described in Section 6.11, using thin-layer
interface elements (Section 8.6) between the main C-section beam and loading
bracket. Graphically show the distribution of shear stress flowing through the inter-
face zone. Determine the total shear flow through the interface by using a virtual
section at the interface between parts.
PROBLEM 2
Re-analyze Problem-3 in Section 6.11 with a cut-section defined at location AB of
Figure 6.29. Verify that resultant force and moment on cut-section match with the
applied external loading on the hook.
spot welds @
spacing of 50-80 mm
Section view
E=206 GPa, =0.31
t=2.0 mm
FIGURE 8.13 Spot-welded assembly of a reinforced automotive bumper beam.
214 Finite Element Analysis of Solids and Structures
PROBLEM 3
Automotive bumper beam in Problem-4 of Section 7.10 (Figure 7.16) is reinforced
by adding a 1.4 mm steel reinforcement plate as shown in Figure 8.13. The back plate
(Bumper_Back_Plate.step) is connected to the upper and lower flanges of the main
beam (Bumper_Beam.STEP) with spot welds spaced approximately at 50–80 mm
interval. Build a finite element simulation model of the spot-welded two-part assem-
bly; and conduct load–deflection analysis of the assembly for the same loading and
boundary conditions described in Problem-4 of Section 7.10. Show the distribution
of normal stresses on a section through the mid-span of bumper beam. Compare the
simulation results with hand calculations based on beam bending theory.
PROBLEM 4
Figure 8.14 shows an automotive connecting rod made of steel. Mesh model of the
part is provided in Connecting_rod.inp. Model the compressive force distribution on
the surface A–A–A as a uniform normal pressure of 0.01 kN/mm2. The support con-
dition on the rod can be described by normal displacement constraints on surface
B–B–B. Conduct stress analysis of the part assuming linear elastic material response.
Define a cross-section C–C–C through mid-height of the connecting and plot the
distribution of normal stresses on the section.
B B
B
C C C
A
A A
FIGURE 8.14 Automotive connecting rod made of steel (E = 200 GPa, ν = 0.3) (FEA model
created from the 3D CAD data available @ GRABCAD.COM 2020).
Multicomponent Model Assembly 215
0.00002 kN/mm2
D Soil Ground surface
5mm thick 5mm thick
D=200 steel pipe interface soil layer
4D
FIGURE 8.15 Cross-sectional view of a buried steel pipe (Assume: steel properties – E =
210 GPa, ν = 0.3; soil properties – E = 0.02 GPa, ν = 0.4).
PROBLEM 5
Figure 8.15 shows a cross-sectional view of an underground steel pipe subjected to a
uniform distributed surface load of 0.00002 kN/mm2. Assuming plane-strain loading
condition, develop a finite element analysis model of the soil-pipe system with thin-
layer interface elements at the interface between soil and pipe. Conduct stress analy-
sis of the system assuming linear elastic material response, assuming elastic
properties of interface material to be same as that of surrounding soil. Produce a plot
of shear stress distribution along the interface between soil and steel pipe.
9 Interpretation of Stress
Analysis Results for
Strength and Durability
Assessment
SUMMARY
The basic purpose of finite element simulation is, obviously, to use the results for
engineering decisions on functionality, strength and durability of products and struc-
tures. Books on finite element methods, as well as the software manuals and docu-
ments, primarily focus on element formulations and/or model preparation techniques.
The correct interpretation of analysis results, based on advanced solid mechanics
principles, is as important as the preparation of a good quality simulation model. This
chapter presents few selected topics relevant for interpretation of the finite element
stress analysis results. Section 9.1 starts with a brief review of the elastic material
properties that are determined from standardized uniaxial material stress tests and are
used directly in simulation models and in subsequent interpretation of results.
Understandably, finite element simulation models deal with more than simple uni-
axial stress field cases. Section 9.2 summarizes the stress component measurements
that are predicted by different finite element formulations presented in earlier chap-
ters. Material failure theories that form the analytical foundation for transformation
of multiaxial stress field measurements to equivalent uniaxial predictor values are
reviewed in Section 9.3. Commonly used stress field measurements, relevant for duc-
tile and brittle materials, are specifically summarized in that section. Section 9.4
discusses the graphical post-processing technique for visualization of complex stress
field responses. Fatigue life assessment of solids, based on finite element stress anal-
ysis results, is discussed in Section 9.5. It needs no mention that the reliability of
finite element stress prediction, when used in conjunction with material strength cri-
teria, is highly relevant for making good engineering conclusion. The sensitivity of
finite element stress results to geometric discontinuities has been discussed in
Chapter 3; and is revisited in Section 9.6 with additional case study results. Section 9.7
demonstrates the capability of standard finite element analysis model for predicting
the stability of cracked solids with the use of linear elastic fracture mechanics metric,
i.e. with the use of derived stress intensity factor from local stress analysis results.
Finally, Section 9.8 presents practice problems to apply the special techniques cov-
ered in earlier sections of this chapter.
217
218 Finite Element Analysis of Solids and Structures
9.1 ENGINEERING PROPERTIES OF MATERIALS
Material specimens of standard size and shape are generally tested under uniaxial
tensile loading condition (Figure 9.1) to determine the mechanical properties such as
elastic modulus, yield strength, ultimate strength, failure strain, etc. ASTM Standard
E8/E8M, for example, describes standards for specimen preparation and testing of
metallic materials. Information generated from such tensile test is primarily used in
material quality controls and in comparative study of different material grades.
Material mechanical properties used in finite element models are also derived from
the material specimen test data. Tensile force applied during testing is divided by the
initial cross-sectional area of testing specimen (A0) to determine the applied engi-
neering stress “s” (Figure 9.2). Similarly, change in length measured over initial
reference gage length (L0) is used to calculate the engineering strain (“e”) inside
material (Figure 9.2). Stress–strain response, thus measured, with respect to initial
geometric dimensions of material specimen, is plotted to produce engineering stress-
strain diagram (Figure 9.2). Assuming no net volume change during load–deformation
L0=4D0
D0
FIGURE 9.1 Uniaxial tensile testing of a material specimen.
Engineering stress: Force, F
s = F / A0
A0 L0
Change in length, L
Engineering strain:
e = L / L0
FIGURE 9.2 Engineering stress–strain response of a material test specimen.
Interpretation of Stress Analysis Results 219
True stress:
= F / A = s * (1+e) Force, F
Sy
L
variable, A
E
True strain: Change in length, L
= L / L = ln(1+e)
FIGURE 9.3 True stress–strain response calculated from engineering stress–strain response
of a ductile material specimen.
response of isotropic ductile material specimen, true instantaneous stress (σ) and true
strain (ε) in the material can be calculated from engineering stress–strain response by
using the following expressions (Figure 9.3):
L
ln 1 e (9.1)
L
F
s. 1 e (9.2)
A
Initial slope of the true stress–strain diagram defines the elastic modulus (E) of the
material (Figure 9.3). Yield stress point Sy defines the departure from linear elastic
stress–strain behavior to nonlinear material deformation (indicating accumulation of
permanent plastic deformation in ductile metals). Uniaxial deformation response of
material, for stress values below yield strength (σ < Sy), is defined by the elastic
modulus (E). When uniaxial force is applied to a solid, it deforms in the direction of
the applied force; and it also expands or contracts laterally depending on whether the
force is tensile or compressive. If the solid is homogeneous and isotropic, and the
material remains elastic under the action of the applied force, the lateral strain bears
a constant relationship to the axial strain. This constant, called Poisson’s ratio, is an
intrinsic material property just like Young’s modulus and shear modulus. ASTM
Standard E132-17 is generally used to determine the Poisson’s ratio from tensile load
testing of material specimens having rectangular cross-section. As discussed in
Chapters 2–7, elastic stress–strain responses of two- and three-dimensional stress
fields can be defined by using E and ν with Hooke’s law (discussed in Section 2.4).
9.2 STRESS–STRAIN RESULTS FROM LINEAR ELASTIC FINITE
ELEMENT ANALYSIS OF SOLIDS
Stiffness properties of a finite element, defined by equation (2.44), involve the mate-
rial stress–strain relationship matrix [C] and the strain–displacement relationship
220 Finite Element Analysis of Solids and Structures
matrix [B]. Same relationship matrices, [C] and [B], are used to calculate the stress
and strain values from the nodal displacement response values. Actual number of the
stress and strain values, calculated at material points, depends on the type of finite
element used in a simulation model. Formulations for truss or bar elements, described
in Chapter 2 for analysis of truss-type structures, produce only axial strain and stress
inside the element (Figure 9.4(a)). Strength assessment of a truss structural member
is generally conducted by simply comparing the model-predicted axial stress value
with uniaxial tensile strength, determined from material tensile tests, or with com-
pressive strength limit defined based on buckling strength capability of the compres-
sively loaded member. The assessment of member tensile strength capability, based
on uniaxial material test result, is a preliminary indication of structural safety. The
average tensile strength of members tends to get lower, particularly for brittle and
quasi-brittle materials, as the member size gets larger – a phenomenon commonly
referred to as “size effect” on material strength (Bažant 2005). Size-adjusted material
stress–strain properties can be considered in nonlinear finite element simulation
models for mesh-independent strength assessment of structures (Bhattacharjee and
Leger 1994).
Simple strength criteria, with or without size effect adjustment, serve the purpose
of safety assessment of members that provide resistance to external loads primarily
through axial deformation mechanism (e.g. trusses, bars, wire strands, etc.). More
general-type structural members may not have a dominant axial stress direction; the
internal stress field can be quite complex with multiple stress components acting in
different directions at the same stress point. Beam section in Figure 9.4(b), for exam-
ple, has both normal and shear stresses acting at a point – and both stress components
vary over the section of the beam. Figure 9.5 shows a relatively more complex stress
field of a plane-stress solid. As discussed with examples in Chapters 3 and 4, three
stress components (σx, σy, and τxy) are required to describe the stress state inside this
u2 x y
x
z
,
u1
Stress Components
Normal stress:
.
= ±
Shear stress: .
= ±
(a) (b)
FIGURE 9.4 (a) Uniaxial stress–strain state in a truss element; (b) variable normal and shear
stresses on a beam section.
Interpretation of Stress Analysis Results 221
x
x
x
x
Strains
y Stress Components
y xy
1 0
x 1 0
=
0 0
x
=0 ; 0
FIGURE 9.5 Strains and stresses at a material point inside a plane-stress solid.
w v
u
Strains
= .
Stress components
1 0 0 0
1 0 0 0
1 0 0 0
( )
= 0 0
( )( ) 0 0 0
( )
0 0 0 0 0
0 0 0 ( )
0 0
FIGURE 9.6 Strains and stresses at a material point inside a 3D solid.
plane-stress solid. Additional stress components appear in finite element calculations
of relatively more complex structural systems. Stress field inside a general 3D solid,
without a strong directional bias for geometry or load or boundary conditions
(Figure 9.6), produces all six stress components (σx , σy, σz, τxy, τyz, and τzx) inside the
material points. A structural shell element, when defined at its mid-plane (Figure 9.7),
produces five stress components (σx, σy, τxy, τyz, and τzx), while the strain and stress in
the normal direction of shell’s mid-plane are considered negligible. Evidently, stress
222 Finite Element Analysis of Solids and Structures
Strains
x
x
x
=
Stress components
1 0 0 0
1 0 0 0
1
0 0
2
= 0 0 1 ×
1 0 0 0 0
0 0
1
0 0
FIGURE 9.7 Strains and stresses at a material point inside a shell element.
fields inside general solids and structures produce multiple nonzero stress compo-
nents at measurement points. Obviously, safety assessment of a complex body can-
not proceed with the comparison of a single stress component against the uniaxial
material tensile or compressive strength value. Strength assessment of a multi-axial
stress field requires the use of advanced solid mechanics principles – based on mate-
rial failure theories discussed in the following Section (9.3).
9.3 STRENGTH ASSESSMENT OF SOLIDS – USE OF MATERIAL
FAILURE THEORIES
Figure 9.8 summarizes how material property parameters, extracted from tensile test
data, are used in different phases of the finite element simulation exercise. Elastic
modulus (E) and Poison’s ratio (ν), generated from uniaxial and bi-axial testing of
material specimens, are used as input parameters to define material stress–strain rela-
tionship matrix, [C], based on Hooke’s law. Multi-axial stress components, calculated
inside the finite element, are transformed to representative index value (σ ) that, in
turn, is compared with the uniaxial material strength (Sy or St) to assess the safety of
body. The transformation of multi-axial stress values to a representative index value
requires the use of advanced mechanics theories based on idealized failure character-
istics of materials. Environmental and history factors (such as temperature, humidity,
stress history, loading rate, etc.) may also need to be considered for a comprehensive
assessment of a stressed body. For engineering analysis of simple monotonic static
load cases, material failures are usually classified into either brittle failure (fracture)
or ductile failure (yielding) type. The most commonly used theories for determining
Interpretation of Stress Analysis Results 223
zz FEA results: multi-axial stresses
zx
xy xx , yy zz xy , yz , zx
yy xx
Transform to a “representative” response
(by using material failure theories)
True stress:
= F / A = s * (1+e) Force, F Compare
with material strength
Sy
= , from uniaxial tensile test
A
L
variable, A
E
Change in length, L
True strain:
= L / L = ln(1+e)
FIGURE 9.8 Process flow for interpretation of finite element stress–strain results.
the failure characteristics of monotonically loaded engineering materials are listed in
the following (Budynas 1999):
1. “Principal” stress theory
2. “Principal” strain theory
3. Maximum strain energy theory
4. Maximum shear stress theory
5. Maximum distortion energy theory
1. Principal stress theory: Finite element analysis produces stresses in user coor-
dinate reference system – such as σx, σy, and τxy in the 2D element shown in
Figure 9.9. Inside the element, stresses on an arbitrary plane can be defined by
two components – a normal stress and a shear stress. Interrelation between the
normal and shear stresses, for any arbitrary orientation of the internal plane, is
defined by a circle – referred to as Mohr’s circle of stress in solid mechanics
(Popov 1978). The maximum and minimum values of normal stress occur on
two orthogonal planes, with zero associated shear stress (Figure 9.9), and are
defined by
2
y x x (9.3)
1,2 x 2 xy
2
2
where σ1 is the maximum value of normal stress, referred to as major principal
stress; and σ2 is the minimum value known as minor principal stress. In the 3D
stress field, principal stresses are defined by the following equations:
224 Finite Element Analysis of Solids and Structures
Transformation of 2D stresses Principal stress: maximum normal stress
occurring on a plane of zero shear stress
y
1 , 2
xy
1
x x x
xy xy St
y y
Material tensile
strength test data
Transformation of 3D stresses
1< St : Safe
zz
zx
xy
yy
xx Principal stresses: 1 , 2 , 3
FIGURE 9.9 Principal stress theory for interpretation of stress analysis results (brittle and
quasi-brittle materials).
1
I1 2
3 3
I12 3I 2 .cos
3I .cos
I 2 2 (9.4)
2 1 I12
3
2
3 3
3I .cos
I 2 4
3 1
I12
3
2
3 3
where I1, I2, and I3 are called stress invariants, defined by
I11 xx yy zz
I 22 xx . yy xx . zz yy . zz xy2
yz2 zx2 (9.5)
I 33 xx . yy . zz xx . yz yy . zx zz . xy
2 2 2
2. xy . yz . zx
and,
3
1 2 I 9I1I 2 27I 3
cos 1 1 (9.6)
1.5
3 2 I12 3I 2
Strength of brittle and quasi-brittle materials is generally defined by the maxi-
mum tensile resistance occurring in a uniaxial tensile stest (St in Figure 9.9). In
multi-axial stress field, the major principal stress is used as the representative
indicator of stress state; and the safety status is assessed by comparing it with
the material tensile strength (St) which may or may not be adjusted for size
effect.
Interpretation of Stress Analysis Results 225
FEA stress-strain results
σy
xy σ2 σ1
σx σx
Strain max
xy
σy
St
strength assessment:
Material tensile
strength test data
FIGURE 9.10 Principal strain theory for interpretation of stress–strain analysis results (brit-
tle and quasi-brittle materials).
2. Principal strain theory: Similar to the principal stress theory, principal strain
theory postulates that failure in brittle and quasi-brittle materials initiates when
the major principal strain becomes equal to a pre-determined material tensile
resistance defined in term of the material tensile strength (equation 9.7)
(Figure 9.10):
St (9.7)
1
E
In a multi-axial elastic stress field, major principal strain (ε1) can be calculated
from strain components, or from the principal stress values (if those are readily
available):
1
1 2 3 (9.8)
E E
3. Maximum strain energy theory: In this theory, strength of ductile material is
defined by the strain energy per unit volume of uniaxial tensile test specimen
that is equal to the area under the stress–strain diagram up to the material yield
strength (Sy) (Figure 9.3). Material is assumed to have failed when the strain
energy, per unit volume in a finite element calculation point, exceeds the allow-
able strain energy (as determined from the simple uniaxial tension test):
1 S2 (9.9)
ijT . ij y
2 2E
4. Maximum shear stress theory: This theory postulates that failure in ductile
materials initiates when the maximum shear stress exceeds the critical material
shear strength determined from uniaxial tensile test. Mohr’s circles for 3D
226 Finite Element Analysis of Solids and Structures
FEA results : mul -axial stresses Mohr’s circles of stresses
principal stresses max
zz
zx
xy
yy
xx
Uniaxial test data
Smax= Sy /2
Sy Force, F
strength assessment:
Sy
L 2 2
E
Change in length, L
FIGURE 9.11 Maximum shear stress theory for interpretation of stress analysis results (duc-
tile materials).
stress field (Figure 9.11) shows that the maximum shear stress is defined by the
following equation (9.10):
1 3 (9.10)
max
2
where σ1 is the major principal stress and σ3 is the minor principal stress. From
the Mohr circle of stress for one-dimensional tensile test of a ductile material
specimen, the shear strength of material corresponding to the state of axial
stress reaching the yield strength is defined by equation (9.11):
Sy (9.11)
Smax =
2
Combining equations (9.10) and (9.11), the criterion for failure initiation in the
multi-axial stress field of a ductile material is defined by equation (9.12):
1 3 Sy (9.12)
max
2 2
5. Maximum distortion energy theory: General stress state in a three-dimensional
body, defined by stress components σxx, σyy, etc., can be represented by three
principal stresses (σ1, σ2, σ3). The matrix representing the principal stress val-
ues can be considered to be made of two matrices – one representing the
Interpretation of Stress Analysis Results 227
average stress state in the body and the other representing the differential
stresses in three principal directions as defined in the following equation:
xx xy xz 1 0 0
xy yy yz 0 2 0
xz yz zz 0 0 3
avg 0 0 1 avg 0 0
(9.13)
0 avg 0 0 2 avg 0
0 0 avg 0 0 3 avg
where σavg is the average stress in the body, defined by σavg = (σ1 + σ2 + σ3)/3.
The differential stresses in three principal directions, represented by the second
matrix on the right side of equation (9.13), are generally referred to as distor-
tional stresses [σ′]:
1 avg 0 0
(9.14)
0 2 avg 0
0 0 3 avg
The main assumption in distortion energy theory is that the hydrostatic tension or
compression, [σavg], does not cause failure in a ductile material; actual material fail-
ure is assumed to initiate when the distortional stresses [σ′] reach a critical state.
Denoting the strains caused by distortional stresses as ε′, the strain energy density
corresponding to the distortional stress–strain state can be derived as follows
(Budynas 1999):
1 1
. . 1 2 2 3 3 1
T 2 2 2
Ud (9.15)
2 6E
Principal stresses in a uniaxial tensile test specimen, at the impending state of yield-
ing, can be defined as follows:
1 Sy , 2 3 0 (9.16)
Substituting the values from equation (9.16) into equation (9.15), critical distortion
energy causing material yielding in uniaxial tensile test specimen is given by
1 2
(9.17)
U dcritical . S y
3E
Maximum distortion energy theory postulates that yielding failure in ductile material
initiates when the distortion energy in a 3D stress field, defined by equation (9.15),
228 Finite Element Analysis of Solids and Structures
FEA results : mul -axial stresses
principal stresses
Distor on energy in the material,Ud
zz 1, 2, 3
zx
xy 1+
yy = [( ) +( ) +( ) ]
xx
Uniaxial test data
From the equivalence
Force, F of distor on energy
Sy
strength assessment:
L
E 1
[( ) +( ) +( ) ]
2
Change in length, L von Mises stress, vm
FIGURE 9.12 Maximum distortion energy theory for interpretation of stress analysis results
(ductile materials).
reaches the critical material strength defined by equation (9.17), leading to the fol-
lowing criterion for stress assessment (Figure 9.12):
1
1 2 2 3 3 1 Sy
2 2 2
(9.18)
2
The stress function on the left side of equation (9.18) is the well-known von Mises
stress (σvm) that is widely used in strength assessment of ductile materials:
vm
1
2
[ 1 2 2 3 3 1 )2 S y
2 2
(9.19)
Each of the five failure theories, described in the above, has its special use depending
on the physical failure characteristics of materials under consideration. For prelimi-
nary strength assessment, based on linear elastic finite element analysis results, prin-
cipal stress and von Mises stress parameters are widely used for brittle and ductile
materials, respectively. Several other material failure theories are also available in the
literature for strength assessment of special material cases. Constitutive models for
nonlinear material response simulations involve many more theories and material
parameters that will be partially discussed in Chapter 12. Material theory manuals of
commercially available finite element software products generally provide important
details of the implemented formulations that are suitable for simulating specific
material failure characteristics.
Interpretation of Stress Analysis Results 229
9.4 POST-PROCESSING OF FINITE ELEMENT STRESS
ANALYSIS RESULTS
Solution of matrix equilibrium equations (1.5) produces displacements at nodal
degrees of freedom. Strain–displacement relationships, defined by equation (2.6),
give strains at internal calculation (integration) points of an element based on the
known nodal displacement values. Stress–strain relationships, defined by equation
(2.26), provide stresses at the element integration points based on the calculated
strain responses. Finite element software packages may save nodal displacements,
and element stresses and strains, in a binary database file by default or when requested
by the analyst. ABAQUS, for example, saves the nodal and element responses in a
database file named with an extension “.ODB”. Saved response values generally
refer to the user-defined model coordinate system (x–y–z). These finite element stress
and strain results can be post-processed to derive the material-specific response
parameters, as discussed in Section 9.3. General-purpose post-processing software
packages, such as HyperView (Altair.COM) and FEMAP (Siemens.COM), can help
with the post-processing of result files saved by many commonly used FEA software
packages. For example, ABAQUS result file (.ODB) can be imported into HyperView
where an analyst can choose to create contour plots of principal stress, principal
strain, or von Mises stress depending on the material strength theory relevant for a
specific case study. Structural shell elements, as discussed in Chapter 7 (also sum-
marized in Figure 9.7), produce stress variations through the thickness direction. In
graphical mid-plane representation of shell geometry, a user can choose to plot the
stress results from one particular layer of through-thickness integration points, or the
maximum response occurring among integration points in the thickness direction.
User chosen material-failure-theory-specific stress response values, calculated from
finite element stress analysis results, can be used in contour plots to identify strength
critical areas of a general structural analysis case-study (Figure 9.13). Identification
of highly stressed areas, based on linear elastic finite element analysis results, does
FIGURE 9.13 Contour plot of maximum through-thickness von Mises stress – plotted on
the mid-surface of shell element model of an automotive bumper beam (FEA model extracted
from a vehicle model – courtesy of NHTSA 2020).
230 Finite Element Analysis of Solids and Structures
not conclusively indicate safety or failure possibility of a structure. These results
provide preliminary indications of critical areas that require further investigations
with more rigorous analysis techniques – including the simulation of nonlinear
response mechanisms if appropriate.
9.5 STRESS ANALYSIS FOR DURABILITY (FATIGUE LIFE)
ASSESSMENT OF STRUCTURES
In monotonic load–displacement response simulations, a simple comparison between
the calculated maximum stress response parameter, and the material strength value,
is a preliminary indicator of the structural safety margin. When the applied stress
exceeds elastic response limit, material can experience partial damage due to yield-
ing or micro-cracking – leaving a permanent distortion in geometry after the external
load is removed. A few applications of repeated loading-unloading cycles can accu-
mulate further internal damage; and the material can eventually fail although ulti-
mate material strength is not exceeded in one load cycle (Figure 9.14(a)). This
phenomenon is generally referred to as low-cycle fatigue damage of materials.
Nonlinear material stress–strain laws are required to predict material damage evolu-
tion with finite element simulation models of high-intensity loading events. Most
engineering products, however, experience low amplitude fluctuations of elastic
stresses and strains during normal operating conditions. An automotive body or chas-
sis component, for example, may experience thousands of low amplitude stress–
strain cycles during every-day driving operations. Cumulative damage in material
under a high number of loading cycles, for stress–strain fluctuations well within the
elastic range of material behavior (Figure 9.14(b)), is generally referred to as high
cycle fatigue damage of materials. The terms “durability” and “fatigue life” are often
used interchangeably to describe the ability of a component to last for the useful life
of the product (for example, 15 years of service or 250,000 km of driving life of a
car). Elastic finite element stress–strain analysis results, with high-cycle material
fatigue life estimation theories, are routinely used to verify the durability of automo-
tive products (World Auto Steel 2015).
Coupon tests
Material stress-strain curve under cyclic loads Material stress-strain curve
S S
Sy Low Cycle Fatigue: Sy
cumulative material
damage under ‘high strain’
cyclical load effects High Cycle Fatigue:
cumulative material damage under
‘low strain’ cyclical load effects
(a) (b)
FIGURE 9.14 (a) Material fatigue under high amplitude cyclic stress–strain loading; (b)
material fatigue under low-amplitude cyclic stress–strain loading.
Interpretation of Stress Analysis Results 231
High cycle fatigue testing to specimen failure Stress range (S) vs Fatigue life (N)
log (S)
S time
Stress
range S-N Curve
Endurance limit:
Stress range: S <
105 106 107
No. of cycles to failure: N Sy
log (N)
No of cycles to failure
(a) (b)
FIGURE 9.15 (a) Constant-amplitude cyclic load testing up to specimen failure; and (b)
S–N curve for fatigue life estimation.
Fatigue life of material is usually measured by testing standard specimens under
constant amplitude uniaxial load cycles until failure (ASTM E466-15 2021). A series
of tests are conducted to determine fatigue lives (number of cycles to failure) for dif-
ferent amplitudes of stress fluctuation. A log–log plot of the stress range (the differ-
ence between maximum and minimum values) versus the number of cycles to failure,
commonly referred to as S–N curve (Figure 9.15), is considered a property of the
specific tested material. When the stress amplitude at a critical location is known
from finite element simulation model, the expected fatigue life is determined from
the material-specific S–N curve. Like uniaxial material tensile strength, fatigue resis-
tance curve of material is also derived from uniaxial stress testing, although the
actual structural components are generally subjected to fluctuations of multi-axial
stress state. Consistent with the material failure theories for monotonic loading con-
dition, major principal stress value for brittle and quasi-brittle materials, and von
Mises stress value for ductile materials can be used to estimate fatigue life from cor-
responding material-specific S–N curve.
Stress fluctuation histories in real engineering applications often comprise of vari-
able amplitude random fluctuations. Several empirical and analytical methods are
available in the literature for reducing a variable amplitude complex stress history to
equivalent blocks of constant amplitude stress cycles (ASTM E1049-85 2021).
Partial fatigue damage caused by each constant amplitude stress block is estimated as
the ratio between number of applied stress cycles (ni) to the fatigue life (Ni), corre-
sponding to the specific fluctuation range of stress variation (Figure 9.16). Cumulative
fatigue damage of all applied stress fluctuations is calculated from simple arithmetic
summation of all fractional contributions:
ni (9.20)
1
Ni
Linear superposition rule of cumulative fatigue damage calculation (equation
9.20), known as Miner’s rule (Miner 1945), can be used in durability assessment of
mechanical components when variable amplitude stress history is known from past-
historical measurements. In the early assessment of mechanical designs, external
232 Finite Element Analysis of Solids and Structures
s4
s3
s1 s1
s2
Equivalent constant- s1 s2 s3
amplitude stress cycles n1 n2 n3
t t
Stress No of Fague Fague
S-N curve range cycles life damage
S1 n1 N1 n1/N1
Si Ni
S2 n2 N2 n2/N2
.. .. .. ..
105 Ni 106 107
cumulave fague damage: 1
FIGURE 9.16 Durability assessment for variable amplitude stress cycles.
loading events are often represented by few selected critical load cases. For example,
fatigue life targets of automotive body structures for normal driving conditions can
be pre-defined, based on historical evidence, as follows (World Auto Steel 2015):
Fatigue critical event Equivalent static loading Target fatigue life
Pothole impact 3g vertical acceleration of un-sprung mass 200,000 cycles
Vehicle turning 0.7g lateral acceleration of vehicle mass 100,000 cycles
Forward braking 0.8g longitudinal deceleration of vehicle mass 100,000 cycles
Linear-elastic finite element simulation model of automotive body structure is
used to identify the critical locations and the corresponding stress values for each of
these load cases. Fatigue life at an identified critical location is estimated from the
material-specific S–N curve for the stress fluctuation range caused by the relevant
loading event. Cumulative fatigue damage for combined load effects at a critical
location is estimated from equation (9.20) where “ni” refers to applicable fatigue load
cycles, and “Ni” refers to estimated fatigue life values from material-specific S–N
curve. When different critical stress locations arise for each of the external loading
conditions, durability assessment reduces to a simple exercise of checking the esti-
mated fatigue life at the critical locations against the pre-defined fatigue life targets
set from past product design experiences. Dedicated fatigue life analysis software
(such as nCode), coupling the finite element stress analysis results with material-
specific S–N curve, can be used for durability assessment of mechanical designs.
Interpretation of Stress Analysis Results 233
(a) (b) (c)
FIGURE 9.17 (a) Hole with undesirable sharp corners; (b) hole with smooth corners to
reduce stress concentration effects; and (c) hole in low stressed zone of a structural member.
When a design does not meet pre-defined fatigue life target, engineering solution
may consider design alternatives of (i) load input modifications (if possible), or (ii)
material modification (to improve fatigue life), or (iii) structural design modifica-
tions to reduce internal stresses (by changes to part geometry and thickness). A good
rule of thumb is that 10% decrease in stress will double the fatigue life of a compo-
nent. In other words, if the estimated fatigue life is at 50% of desired fatigue life,
decreasing the stress level by 10% would allow the part to achieve full 100% of the
durability life. In design for durability, the use of good engineering practice is per-
haps more critical than the prediction of stress response with reliable finite element
simulation technique. When engineered properly, a component can usually meet the
target fatigue life at no cost or weight penalty. The basic principle in design for
durability is to eliminate or minimize the effects of stress raisers – without waiting
to discover local stress concentration issues during late-stage finite element simula-
tion exercises. As discussed in Section (4.1), and illustrated in Figure 4.2, a simple
circular hole in a uniform axial stress field can raise the local stress concentration
value to 3-times the nominal stress value. As per equation (4.3), local stress concen-
tration value increases as the hole shape in Figure 4.2 becomes narrower (a < b),
thereby leading to a lower fatigue life for the component. If a hole is essential in a
stressed component design for special functional reasons, a smooth circular hole is
much more preferable to a square or rectangular hole. However, when a non-circular
hole is required for specific reasons, the corners of the hole must be produced with
smooth radii (Figure 9.17(a)) to reduce stress concentration values around the
corners. When there is an option, cutouts or holes in a member must be located in
low-stressed areas to reduce the impact on durability life of structural component.
In addition to geometric discontinuity in mechanical component design, discrete
joints in multi-component product assemblies also act as stress raisers leading to
fatigue life concerns for engineered products. Spot welds, for example, show poor
fatigue resistance when subjected to tensile loading mode (Figure 9.18(a)); better
fatigue resistance is achieved when the weld joint is engineered to transfer shear
loading (Figure 9.18(b)). Smooth stress transfer at welded joints is further improved
by using structural adhesives in between the mating surfaces of structural sheet met-
als (Dow Automotive Systems 2021).
234 Finite Element Analysis of Solids and Structures
(a) (b)
FIGURE 9.18 (a) Tensile peeling load on a welded joint; and (b) shear loading on a
welded joint.
9.6 STRUCTURAL SAFETY ASSESSMENT AND QUALITY OF FEA
STRESS RESULTS
As evident from the discussions in Sections 9.2–9.5, structural performance assess-
ment for strength and durability, in early design phases, relies on the stress analysis
results from finite element simulation models. Quality of the stress analysis mod-
els, thus, has a direct impact on the quality of engineering decisions made with
respect to strength and durability performance of a proposed structural design con-
figuration. Finite element stress analysis results need to be used with caution while
predicting the strength and fatigue life of engineered products. In general, finite
element simulation models can estimate the locations of high stressed areas in a
structure, but the magnitude of predicted stress can be affected by mesh quality and
severity of geometric discontinuity in the model. The simulation model of a uni-
formly loaded plate, with a smooth circular hole at the center (Figure 4.2), has
500
y
x 0.1 GPa
200 2b=20
Stress concentration factor:
kl
FIGURE 9.19 Indeterminate stress concentration at the tip of a sharp geometric discontinu-
ity in the uniformly stressed plate.
Interpretation of Stress Analysis Results 235
0.1 GPa 0.1 GPa
Element size : 10 mm Element size : 2.5 mm
xx
@ the crack p: 0.13 GPa xx
@ the crack tip: 0.27 GPa
(a) (b)
FIGURE 9.20 Mesh-dependent prediction of stress response at the crack tip of a uniformly
stressed plate.
predicted the internal stress distribution accurately – producing the local stress
concertation effects at the crest of the circular hole consistent with the result of the
analytical solution (equation 4.3). When the circular hole is replaced by a narrow
crack at the center of plate (Figure 9.19), analytical stress prediction at the tip of
sharp geometric discontinuity becomes indeterminate. Finite element simulation
models, however, always predict some stress value at the points of discontinuity
because of the homogenization of the stress singularity over a finite size element
(Figure 9.20). As the finite element mesh is refined, the predicted stress response
keeps rising without convergence to a stable predictable response (Figure 9.20). A
similar mesh-dependent stress result has been obtained in the vicinity of the rigid
boundary constraint of the in-plane bending problem in Figure 3.19(b). It is, there-
fore, important to verify that the stress response predicted by a finite element simu-
lation model is reliable for use in strength-based criteria of structural integrity
assessment. The general rule of practice is to conduct stress analyses with two
finite element models – one having double the mesh density compared to the other.
If the stress results from two models are very close (within 10%), the predicted
stress result can possibly be used with some confidence for structural safety assess-
ment. Close predictions of stress results with two different mesh density models,
however, do not mean that the predicted results are accurate. The magnitude of
stress results may be affected by the inherent limitations of the finite element for-
mulations (as discussed in Section 3.10). Local stress responses, predicted by finite
element simulation models, often need to be supplemented with past experiences
of the analyst to reach meaningful conclusions on the safety assessment of
structures.
9.7 STRESSES AT POINTS OF DISCONTINUITY:
STRESS INTENSITY FACTOR
As evident from the discussions presented in Section 9.6, linear elastic stress analysis
results are not objective at the vicinity of singular stress points. Theory of fracture
mechanics provides an alternative method for the integrity assessment of solids with
cracks. Considering a local coordinate system x–y with origin at the tip of a sharp
236 Finite Element Analysis of Solids and Structures
mode-I mode-II mode-III
(opening) (shearing) (tearing)
FIGURE 9.21 Fracture modes in a general 3D solid.
crack in Figure 9.19, the indeterminate state of elastic stress at the crack tip can be
defined by the following expression:
KI
x (9.21)
2 y
where the term KI defines the intensity of stress at the vicinity of crack tip – com-
monly known as the stress intensity factor (Irwin 1957). As per the theory of fracture
mechanics, the strength of a cracked body is determined by the magnitude of stress
intensity factor. Crack traversing a normal stress field will have unstable growth
when KI reaches a critical material strength value called Fracture Toughness (KIc).
Toughness value changes depending on the mode of crack tip deformation
(Figure 9.21). ASTM Standard (ASTM E1820-20ae1 2021) can be followed to deter-
mine the fracture toughness of material specimens for mode-I (opening) crack.
Alternatively, reference values for preliminary assessment of material fracture resis-
tance can be obtained from ASM Handbook (ASM International 1997).
Stress intensity factor at the tip of a crack inside a general solid depends on crack
length, geometric configuration of cracked body, the external load, and boundary
conditions. Analytical and empirical expressions for calculation of stress intensity
factors in simple geometric and loading configurations are available in Tada et al.
(2000). More complex structural design cases require the use of finite element simu-
lation models to calculate the stress intensity factor. Several alternative numerical
techniques are available for the extraction of stress intensity factor values from finite
element simulation models, such as crack-tip singular element formulations
(Zienkiewicz and Taylor 1991), energy release rate method (Zehnder 2012), and
J-integral method (Rice 1968). ABAQUS software includes a special routine for the
extraction of stress intensity factors from J-integral calculations (Dassault Systems
2020b). A simpler approximate method for estimation of stress intensity factor
involves the use of equation (9.21) directly with the finite element stress analysis
results. For the center crack problem of uniformly stressed plate (Figure 9.19), using
the stress value σx at a distance of 2.5 mm from the crack-tip of finite element model
(Figure 9.20(b)), equation (9.21) gives the following estimated value of KI:
K I x . 2 y 0.1578 2 2.5 0.625 GPa mm (9.22)
Interpretation of Stress Analysis Results 237
Analytical solution for stress intensity factor at the tip of central crack of uni-
formly loaded large plate gives (Broek 2012):
K theoretical nomial . b 0.10 10 0.560 GPa mm (9.23)
The difference between stress intensity factor values in equations (9.22) and
(9.23) is less than 12%. The calculation procedure, based on the local finite element
stress response value (equation 9.22), provides an approximate value for the stress
intensity factor since the stress values, extrapolated at element nodes, are not very
accurate. Alternatively, the stress intensity factor can also be calculated directly from
standard finite element models based on the elastic energy release rate concept.
Elastic energy stored in the plate with 20 mm long crack at the center is four times
the energy stored in the quarter model of plate shown in Figure 9.20(b)):
U 0 4 0.5985 2.3941 kN mm (9.24)
Re-analyzing the model with an extended crack length of 25 mm (2.5 mm exten-
sion of each tip), strain energy stored in the model is given by
U1 4 0.600532 2.4021 kN mm (9.25)
Change in strain energy of the system per unit area of crack extension is given by
U 2.4021 2.3941 (9.26)
R 0.0016 kN/mm
a 1 5 1
Stress intensity factor for the rate of strain energy change, R, is given by (Broek 2012):
K I E R 200 0.00161 0.566 GPa mm (9.27)
The predicted value from the strain energy release rate method (equation 9.27) is
very close to the theoretical value (equation 9.23) – with 1% difference between the
two. Standard finite element simulation models, thus, provide reasonable estimates
of the stress-intensity factor at a crack tip subjected to the tensile opening mode of
loading. Similar calculation procedures can be used to predict stress intensity factor
values for other crack opening modes as well (Figure 9.21). Like the stress-based
criterion for cumulative fatigue damage calculations, fatigue crack growth in a solid
can also be calculated based on the fluctuation of the stress intensity factor:
da
C. K
m
(9.28)
dN
Equation (9.28) is known as Paris law (Paris and Erdogan 1963), where “a” is crack
length, “N” the number of load cycles, “ΔK” the range of stress intensity factor varia-
tion, and C,m are material constants determined from experiments. Finite element
model results for ΔK can be used with equation (9.28) to determine the fatigue crack
growth rate in a body.
238 Finite Element Analysis of Solids and Structures
9.8 PRACTICE PROBLEMS: ASSESSMENT OF STRUCTURAL
STRENGTH AND DURABILITY
PROBLEM 1
From the load-deflection analysis results of the automotive roof problem in
Problem-3, Section 7.10, generate a contour plot of maximum von Mises stress val-
ues corresponding to the applied pressure load of 0.15 kN. Assuming material yield
strength of 0.12 GPa, determine the factor of safety against yielding of material at the
vicinity of load application area.
PROBLEM 2
Vehicle drive over a hypothetical durability test track generates the variable ampli-
tude stress history of Figure 9.22(b) – measured at a critical body structural compo-
nent. For example calculations, Table in Figure 9.22(c) shows the constant amplitude
stress ranges and number of cycles equivalent to one week of vehicle driving condi-
tion. Using the S–N curve from Figure 9.22(d), estimate the fatigue life of the rele-
vant structural component.
PROBLEM 3
Figure 9.23 shows uniform in-plane stress and boundary constraints on a plate with
hexagonal hole at the center. Calculate the stress intensity factor at point “A” of the
hexagonal hole.
PROBLEM 4
Figure 9.24 shows a plate with hole – subjected to in-plane uniform traction (0.01
GPa) on the upper edge. Calculate the maximum stress occurring on the perimeter of
the circular hole.
(a) Vehicle road test driving (b) Stress response
(c) Stress ranges and corresponding number of cycles
(d) S-N Curve
Stress Stress # of cycles
in each
range ranges durability
60 Mpa test, ni
20 MPa 32,000
40 Mpa
40 MPa 12,800
20 Mpa
105 106 107 60 MPa 8,000
No of cycles to failure, N
FIGURE 9.22 Cumulative fatigue damage calculation for variable amplitude loading.
Interpretation of Stress Analysis Results 239
Aluminum: E=70 GPa, =0.33
100 0.1 GPa
20
x
100 A
250 250
FIGURE 9.23 Uniform stress and boundary constraints on a plate with a hexagonal hole at
the center.
Circular hole of dimeter: d = L/10 y
y =0.01 GPa
L=100 mm
x
L=100 mm
FIGURE 9.24 Hole at the center of a unit thickness steel plate subjected to uniform in-plane
traction on the upper boundary (assume E = 200 GPa, ν = 0.3).
10 Vibration Frequency
Analysis of Structures
with FEA Model
SUMMARY
Linear elastic finite element analysis, discussed so far in Chapters 1–9, has consid-
ered time-independent load effects – represented by static equilibrium state between
applied load and stiffness-based deformation resistance of structures. Vibration
response of a structure refers to dynamic oscillations of system responses (stresses,
displacements, etc.) under external perturbation effects. The metrics for structural
vibration, represented by cycle per unit time or the time period taken for one com-
plete cycle of response variation, depend on the stiffness and mass properties of a
given structure. The vibration frequency (or period of structural vibration), represent-
ing the dynamic property of a structure, and its relative relationship with the dynamic
characteristics of applied load, define the amplitude of linear dynamic response of a
structure. The study of vibration frequency is, thus, a very important part of the engi-
neering development process for civil, mechanical, and aerospace engineering prod-
ucts and structures. The basic dynamic equilibrium state, between applied load and
corresponding system resistances (representing the deformation and inertia charac-
teristics of a deformable spring-mass system), is established in Section 10.1 based on
Newton’s second law of motion. The free-vibration response of the single-degree-of-
freedom (SDOF) spring-mass system, induced by an initial perturbation to the static
rest state, is analyzed in Section 10.2 – eventually leading to the important funda-
mental relationship among stiffness, mass, and vibration frequency properties of the
system. Section 10.3 is devoted to the analytical descriptions of forced vibration
response and resonance behavior of SDOF linear elastic system. The effect of inter-
nal energy loss mechanism, represented by the addition of a damping term in the
description of SDOF system, is also analyzed in this section. The relative relation-
ship between structural vibration property and the dynamic load characteristic, defin-
ing the relevance of static versus dynamic response analysis techniques, and the
effect of damping on the overall dynamic amplification of system response, are
graphically demonstrated based on SDOF system solutions. The use of frequency
separation concept, to define targets for subsystem designs, is discussed with refer-
ence to a hypothetical automotive system example in Section 10.4. Analytical tech-
niques to estimate the vibration frequency and mode shapes of relatively more
complex systems, having uniformly distributed system and mass properties, are
developed in Section 10.5.
The basic definitions of SDOF vibration characteristics are extended to multi-
degree-of-freedom (MDOF) system property definitions in Section 10.6
241
242 Finite Element Analysis of Solids and Structures
– developing the analytical formulations to calculate the vibration frequencies from
stiffness and mass property matrices of MDOF systems. Mass matrix calculation of
finite elements, not discussed in earlier chapters, is discussed in Section 10.7.
Numerical techniques for the calculations of MDOF mode shapes and frequencies,
as implemented in finite element software packages, are reviewed in Section 10.8.
Relative efficiencies of different numerical techniques, for the analysis of large
MDOF systems, are also discussed in that section. Software-based analysis of modal
frequencies and mode shapes of finite element models is discussed with ABAQUS-
specific options in Section 10.9. Finally, practice problems for modal frequency
analysis of MDOF finite element models are presented in Section 10.10.
10.1 INTRODUCTION – DYNAMIC RESPONSE OF STRUCTURES
Newton’s second law of motion states that force (F) acting on a free rigid body
causes a change in momentum that is proportional to the force applied, and the equa-
tion of motion for a given mass value (m) is as follows:
u
F m (10.1)
t
Writing the rate of momentum change as acceleration of the body, equation (10.1) is
re-written as follows:
2u
F m m u (10.2)
t 2
Figure 10.1 shows the direct equilibrium between externally applied force (F) and
When the body mass is attached to a
inertia resistance of a free rigid body (m*u).
flexible structural system, represented by a spring in Figure 10.2, the dynamic motion
Displacement, u
Velocity, ů
Accelera on, ü F(t)
(a) (b)
FIGURE 10.1 (a) Dynamic response of a unit mass under externally applied load (F(t)); and
(b) force time histories (external force = inertia resistance).
Vibration Frequency Analysis of Structures with FEA Model 243
Flexible Spring
Displacement, u
Velocity, ů
Acceleration, ü F(t)
FIGURE 10.2 Dynamic response of a flexible spring-mass system under externally applied
load F(t).
of mass leads to spring deformation that is resisted by the stiffness property of struc-
ture. System resistance to mechanical deformation can be defined by using the struc-
tural stiffness properties as described in Section 1.4 (Figure 1.7(b) and equation 1.1).
State of equilibrium between the externally applied force and the combined system
resistances derived from inertia and stiffness properties of the flexible spring-mass
system of Figure 10.2 can be expressed by
F t k.u m u (10.3)
where u is spring deformation and u is the acceleration of mass – both time-depen-
dent responses of the spring-mass system. If the external force F(t) is removed at a
must equilibrate each other leading to
time, spring force (k.u) and inertia force (m⋅u)
free-vibration response of the system – discussed in the following Section 10.2.
Forced vibration response of the system, including the effect of system energy losses
due to damping, will be discussed in Section 10.3.
10.2 VIBRATION FREQUENCY OF A SINGLE DEGREE OF
FREEDOM SPRING-MASS SYSTEM
Figure 10.3(a) shows a mass (m), attached to an elastic spring of stiffness (k). If the
mass is pulled from its rest position by a unit displacement (u = −1) and released, the
unbalanced spring force on the mass will cause an upward acceleration of the mass.
As the mass returns towards the original rest position (u = 0), velocity of the mass
gradually increases, and the spring force gradually reduces – eventually becoming
zero when the mass returns to its rest position. Assuming no energy losses in the ideal
elastic spring-mass system (Figure 10.3(b)), the elastic deformation energy, stored in
the spring at stretched state (u = −1), gradually converts to kinetic energy of the mass
as it returns towards the initial un-stressed position. The velocity of mass, gained
during the retraction phase of spring from deformation state of u = −1 to initial
stress-free state u = 0, will keep driving the mass further upward and that motion will
cause progressive compression in the spring until the spring compresses to a state of
u = +1, and the mass reaches zero velocity state (u = 0) due to the spring resistance
opposing its upward motion. The compressed spring from that position will start the
downward return motion of mass with gradual conversion of spring elastic energy to
kinetic energy of mass. This repeated cycle of energy conversion between spring
deformation and mass kinetic motion will continue un-interrupted resulting in a
244 Finite Element Analysis of Solids and Structures
Force on the
spring:ku
u
stiffness, k u = sin( t)
u=+1
u=0
mass, m Deformation, u time, t
u=-1
(a) (b) (c)
FIGURE 10.3 (a) Single degree of freedom (SDOF) spring-mass system; (b) linear elastic
resistance-to-deformation response of the spring; and (c) harmonic motion of the mass.
cyclic harmonic motion of the system (Figure 10.3(c)) – commonly known as the
free vibration response of a system:
u t sin t (10.4)
where ω (= 2πf) is the cyclic motion of the mass and f is the frequency of vibration
per unit time. The self-equilibrating state of free vibration response, between the
spring resistance force and mass inertia force, can be described by re-writing equa-
tion (10.3) as follows:
k.u m u 0 (10.5)
Substituting the expression (10.4) into equation (10.5) gives:
k m sin t 0
2 (10.6)
Equation (10.6) will be satisfied for any value of ω and t. This means that the multi-
plication term in the parenthesis (k–ω2m) must be zero – leading to the following
frequently used relationships among stiffness (k), and mass (m), and vibration char-
acteristics (frequency f, period of vibration, T, and angular velocity ω) of the spring-
mass system:
k 2 k
2 2 f (10.7)
m T m
The terms f and T are commonly referred to as natural vibration frequency and fun-
damental period of vibration of the SDOF spring-mass system. The fundamental
vibration property, represented by ω, f or T, defines the dynamic response character-
istic of a system when subjected to a time-dependent dynamic force function. The
relative relationship, between time-domain characteristics of force function and the
Vibration Frequency Analysis of Structures with FEA Model 245
responding spring-mass system, is a critical factor defining the intensity of a system’s
dynamic response to the applied external force. Section 10.3 in the following dis-
cusses the forced vibration response – which is an important aspect of structural
design for dynamic load effects.
10.3 FORCED VIBRATION RESPONSE AND RESONANCE
OF STRUCTURES
Figure 10.4(a) shows a general spring-mass system subjected to a time-dependent
force F(t). An example time history of external force is described by the sine har-
monic function in Figure 10.4(b). Assuming linear elastic response of spring without
any energy losses due to damping effects, the dynamic equilibrium equation (10.3)
can be rewritten for the sine harmonic load case as follows:
2u
m k.u t Fo sin t (10.8)
t 2
where Fo is the amplitude, and ω the angular velocity of harmonic load function
( 2 f ; f being the frequency of applied load). Dividing both sides of equation
(10.8) by the spring stiffness term, k:
m 2u F
2 u t o sin t (10.9)
k t k
Substituting ω2 for the ratio between spring stiffness k and mass m, equation (10.9)
is re-written in the following form:
1 2u F
u t o sin t (10.10)
2 t 2 k
A particular solution of second-order partial differential equation (10.10) can
be expressed in term of the applied loading function multiplied by response ampli-
tude G (equation 10.11):
F F(t)=F0.sin ( t)
stiffness, k F0
Displacement, u Time, t
mass, m
Velocity, ů
Acceleration, ü F(t)
(a) (b)
FIGURE 10.4 (a) Dynamic response parameters of a spring-mass system under time-
dependent applied load F(t); and (b) example of a time-dependent loading history.
246 Finite Element Analysis of Solids and Structures
u p G sin t (10.11)
Equation (10.11) represents one part of the displacement response that is in-phase
with the loading function. Substituting expression (10.11) into equation (10.10), and
after re-arranging the terms, the response amplitude G is obtained as follows:
F0 1 F0 1 (10.12)
G
k
2
k 1 2
1
where β is the ratio between ω and ω. A second solution of the second-order partial
differential equation (10.10), known as the complimentary solution, is given by the
following function:
uc A sin t B cos t (10.13)
Equation (10.13) represents the free vibration response of the spring-mass system
without the presence of external force function. Combining equations (10.11),
(10.12), and (10.13), total dynamic response of the spring-mass system is given by
the following equation:
F0 1
u t A sin t B cos t sin t (10.14)
k 1 2
= 0, unknown terms A and B in
Considering a special initial condition of u(0) = u(0)
equation (10.14) are obtained as follows:
F0 (10.15)
A , B 0
k 1 2
Substituting the expressions from equation (10.15) into equation (10.14), the time-
domain response of spring-mass system, for the harmonic force function shown in
Figure 10.4(b), is given by the following equation:
F0 1
u t sin t sin t (10.16)
k 1 2
The first multiplication term (F0/k) in equation (10.16) represents the static displace-
ment response corresponding to load amplitude Fo; second term (1/(1 − β2)) is a
dynamic amplification factor applied to that static response; the first term inside the
parenthesis (sin ω t) represents the response function in-sync with the external force
function described in Figure 10.4(b), and the second term inside the parenthesis (β.
sin ωt) represents the free vibration response of the system with an angular velocity
of ω defined by equation (10.7). The parameter β, as defined earlier, is the ratio
Vibration Frequency Analysis of Structures with FEA Model 247
between frequency of applied loading ( f ) and the natural vibration frequency of
spring-mass system (f). As evident from equation (10.16), when the frequency of
applied external force approaches the natural vibration frequency of system, the
dynamic amplification factor (1/(1−β2)) approaches infinity – implying that the
response of elastic system amplifies indefinitely under the action of external cyclic
loading. This phenomenon of dynamic structural response is commonly known as
the “resonance” of structure when f = f . Most built-up structures, however, experi-
ence energy losses due to hysteretic load-deformation response as shown in Figure
10.5. This response mechanism is generally represented by adding a velocity-
dependent resistance component to the linearized deformation resistance function –
leading to the following modified form of the dynamic equilibrium equation (10.8):
2u u
m c. k.u t Fo sin t (10.17)
t 2 t
The term “c” in equation (10.17) is commonly known as the damping resistance of
the system. Diving both sides of equation (10.17) by the spring stiffness term “k”:
m 2u c u F
2 . u t o sin t (10.18)
k t k t k
Introducing ω2 = k/m, and ξ = c /2 k.m , equation (10.18) can be re-written in the
following form:
1 2u 2 u F
. u t o sin t (10.19)
2 t 2 t k
The term ξ is generally referred to as the damping ratio. Free vibration response of a
damped system diminishes with time due to the presence of damping. Ignoring the
free-vibration response part, the dynamic response of damped system can be defined
by the following load-dependent time function (Clough and Penzien 1975):
F0 2
u t .D.sin t tan 1 (10.20)
k 1 2
where “D” is the dynamic amplification factor for damped spring-mass system
defined in equation (10.21):
1 (10.21)
D
1 2
2 2
2
Figure 10.6 shows graphical representation of the factor D for different damping
ratio values. When the structural vibration frequency is very high compared to the
frequency of applied loading function (β ≈ 0), or in other words, when the
248 Finite Element Analysis of Solids and Structures
External force, F
stiffness, k
Displacement, u damping, c
Velocity, ů mass, m
Deformation, u
Acceleration, ü F(t)
(a) (b)
FIGURE 10.5 (a) General dynamic characteristics of a spring-mass-damper system; and
(b) hysteresis in force–deformation response of the spring.
time-domain variation of loading is very slow compared to the vibration period of
structure, the dynamic amplification factor value is close to “1”. This means that the
system response is well predicted by the static load–deflection analysis irrespective
of the damping behavior of structure. On the other end, when (β > 2), load frequency
is high and the structure is relatively flexible with a low frequency of vibration, the
dynamic amplification factor is well below “1” irrespective of the damping value. In
that situation, inertia of the system dominates the response to dynamically applied
load. At resonant frequency (β = 1), the value of “D” becomes infinite for zero damp-
ing value (ξ = 0.0) – as expected for an undamped system. Figure 10.6 also shows
that for highly damped systems (ξ > 0.5), the value of D approaches “1” for fre-
quency ratio in the range of 0–1, implying that the system response can be predicted
by simple static load–deflection analysis. However, for lightly damped common
structural systems (buildings, towers, bridges, automotive body structures, etc.), the
damping ratio tends to be well below 10% (ξ < 0.1) which implies that “D” will reach
a very high value (>5) when the applied load frequency happens to be in the vicinity
=0
=0.1
=0.2
=0.5
FIGURE 10.6 Dynamic amplification factor for damped systems.
Vibration Frequency Analysis of Structures with FEA Model 249
of natural vibration frequency of the structure. A good part of the design process of
many structures, thus, involves management of the structural response through sepa-
ration of the structural vibration frequency from the input load frequency. Section 10.4
presents a discussion of this topic.
10.4 FREQUENCY SEPARATION AND DESIGN TARGETS
FOR STRUCTURES
Design of structures, when subjected to cyclic dynamic loads, is generally concerned
with the resonant mode of structures, i.e. vibration modes that amplify input forces.
As demonstrated with the dynamic amplification curves in Figure 10.6, alignment
between the structural vibration frequency and the input load frequency leads to large
amplification of the response for lightly damped systems. From a structural design
point of view, the issue is how to design a structure by avoiding resonance with input
excitations. In automotive product engineering, for example, the target design fre-
quencies for suspension and body structure are set apart from the known input fre-
quencies of powertrain operation and road loads (Figure 10.7). Design targets for
natural vibration frequencies of the body structure, in bending and torsional defor-
mation modes, are set in the range of 25–35 Hz, largely to avoid resonance with
suspension motions, and also to stay away from powertrain operation and road input
frequencies in the range of 100–500 Hz. Dynamic load effects on the structure are
significantly reduced by keeping the natural vibration frequencies far apart from the
applied loading frequencies (Figure 10.6). This same idea of frequency separation is
also used in the earthquake resistant design of civil engineering structures (BSSC
2020), where earthquake load is defined as a function of the structural vibration
period; and the design process is driven with the objective of minimizing the dynamic
load effect on the structure. Calculation of the structural vibration frequencies, and
the adjustment of mass and stiffness properties to tune the frequency values, are of
critical importance in all structural engineering disciplines. Sections 10.5 and 10.6 in
the following present commonly used analytical techniques for calculating the struc-
tural vibration frequencies. More powerful finite element model-based calculation
Design elements Load inputs
Suspensions
Engine mounts
Powertrain
Road inputs
Acous c cavity
Body structure
1 10 100 500
Frequency, Hz
FIGURE 10.7 Example of frequency separation: automotive system design.
250 Finite Element Analysis of Solids and Structures
procedures, for large and complex multi-degree-of-freedom (MDOF) structural sys-
tems, are presented next in Section 10.7.
10.5 VIBRATION MODE SHAPE AND FREQUENCY OF
SDOF STRUCTURES
Implicit assumption for the SDOF in Figure 10.3 is that the spring-mass system can
vibrate in only one mode – producing axial expansion and contraction in the spring.
The vibration frequency (and period of vibration) is calculated from equation (10.7)
by using discrete mass and spring stiffness property values. This analysis method can
be used to determine the vibration characteristics of elementary structural members
that can be reduced to an equivalent SDOF system. Figure 10.8(a) shows a beam
member supporting a discrete mass (M) at its mid-span. Any transient perturbation to
the mass element is expected to cause up-and-down vibrational motion of the beam-
mass system. Assuming that the mass of beam element is negligible compared to the
lump mass (M), the vibration frequency of beam-mass system can be calculated from
equation (10.7) by using the effective mass value (Meff = M) and the bending stiffness
of beam that resists the vertical vibrational motion of mass element. The relevant
bending stiffness of simply supported beam, of length L and flexural stiffness EI, can
be calculated by measuring the lateral deflection “δ” for an arbitrary lateral load “P”
at midspan (Figure 10.8(b)):
P 48EI
K eff 3 (10.22)
L
Frequency of vibration of the idealized SDOF beam-mass system can be readily
calculated by using the effective mass value (Meff) and the bending stiffness of beam
“Keff” from equation (10.22).
This above calculation procedure for SDOF vibrating system can be extended to
calculate the vibration frequency of a simply supported beam of uniform flexural
stiffness “EI” and distributed mass per unit length “m” (Figure 10.9(a)). The vibra-
tion mode shape of prismatic beam is described by the following sine harmonic
function:
x
v x sin (10.23)
L
P
M
(a) (b)
FIGURE 10.8 (a) Lateral vibration of a simply-supported beam-mass system; and (b) bend-
ing deformation shape of the beam under a lateral load at mid-span.
Vibration Frequency Analysis of Structures with FEA Model 251
(a) (b)
FIGURE 10.9 (a) Lateral vibration of a simply-supported beam of uniform flexural rigidity
(EI) and mass per unit length (m); (b) higher vibration mode shape of the beam.
Following the Euler–Bernoulli’s beam deflection theory (equation 6.58), the inter-
nal bending moment (BM) in the beam corresponding to the vibration mode shape of
equation (10.23) is given by
d 2v 2 x
BM x EI 2
2
.EI .sin (10.24)
dx L L
Bending stiffness of the beam, defined by the lateral load resistance correspond-
ing to the bending response of equation (10.24), is calculated as follows:
L L d 2 BM x 4 L
x 3
K eff
0
p x dx
0 dx 2
dx
L4
.EI .
0
sin
L
dx 2 .
L3
.EI
(10.25)
Effective total mass of beam participating in the sine harmonic vibration mode of
equation (10.23) is calculated as follows:
L
x 2
M eff m
sin L dx .mL
0
(10.26)
Substituting the effective stiffness and mass values from equations (10.25) and
(10.26) into equation (10.7), the vibration frequency response for a simply supported
prismatic beam is given by the following equation:
2
2 EI (10.27)
2 f
T L m
For example calculations, substituting EI = 1540980 kN-mm2, L = 600 mm and
m = 0.0083 kg/mm, equation (10.27) gives ω = 0.3728 rad/ms. If the effective stiff-
ness of beam bending is approximately calculated by equation (10.22), simulating
252 Finite Element Analysis of Solids and Structures
the bending effect of a concentrated vertical load at the midspan, Keff = 48EI/L3 =
0.34244 kN/mm, the bending frequency response of beam, with the effective mass
value from equation (10.26), can be calculated from equation (10.7) as, ω = 0.3182
rad/ms. This approximate value is within 7% of the theoretical value of 0.3728 rad/
ms. Approximation of the vibration frequency value, by using simplified assump-
tions for stiffness and mass values, provides valuable insights during the preliminary
design phase of structures.
Extending the half-sine harmonic description of beam vibration mode
(Figure 10.9(a)) to higher degree vibration mode shapes (Figure 10.9(b)), a general-
ized definition for the vibration modes of a simply supported prismatic beam can be
given by modifying equation (10.23) as follows:
n x
v x sin ; where n 1, 2, 3.etc. (10.28)
L
The vibration properties of simply supported beam corresponding to higher
degree mode shapes are defined by modifying equation (10.27) as follows:
2
2 n EI
n 2 fn ; where n 1, 2, 3.etc. (10.29)
Tn L m
Evidently, the lowest vibration frequency of system is given by n = 1 (corresponding
to the shape shown in Figure 10.9(a). All other frequency values (for n > 1), corre-
sponding to relatively more complex mode shapes, will be higher than the first fre-
quency of vibration. The lowest frequency value is generally referred to as the
fundamental frequency of vibration of an MDOF system. And the vibration mode
shape, corresponding to the lowest frequency value, represents the lowest energy
mode of the structure. Equation (10.29), thus, presents frequency solutions for the
prismatic beam structure by reducing the distributed stiffness and mass properties to
effective stiffness and mass values of SDOF systems representing different vibration
mode shapes. The methodology, described thus far, represents a very basic but pow-
erful method for calculating the vibration frequencies of structures, having distrib-
uted mass and stiffness properties, by using appropriate functions for vibration mode
shapes. Considering the simply supported flat rectangular plate of dimensions a × b
(Figure 10.10), and distributed mass per unit area = m (m = material density * plate
thickness), the fundamental period of plate vibration is given by
FIGURE 10.10 Vibration mode shape of a simply supported rectangular plate.
Vibration Frequency Analysis of Structures with FEA Model 253
1 E.t 3 1 1
2 2 2 (10.30)
m 12 1 2 a b
Equations (10.7), (10.27), and (10.30), defining the vibration characteristics of
elementary structural examples, provide basic analytical tools to check the results of
more complex finite element models (to be discussed later in this chapter). For exam-
ple, assuming the overall bending stiffness target of automotive body structure to be
10 kN/mm (Figure 10.11), and the effective mass in bending vibration mode given by
equation (10.26) as Meff = 2/π*body structure mass (364 kg) = 231.5 kg, the vibration
frequency target in bending mode of structural vibration is obtained from equation
(10.7) as follows:
1 10 cycles (10.31)
f 1000 33 Hz
2 231.5 ms
Bending frequency, calculated in equation (10.31), serves as a reference value to
guide the development of body structure design with finite element analysis
models.
10.6 VIBRATION FREQUENCIES OF MDOF SYSTEMS
Figure 10.12 shows two vibration mode shapes (ϕ1 and ϕ2) of an arbitrary MDOF
structure, and the time–domain function for the vibration mode shapes are defined as:
ui i sin it (10.32)
where {ϕi} describes the vibration mode shape of system (i = 1,2 …etc.), and ωi is
the angular velocity of vibration (ωi = 2πfi, where fi is cyclic frequency) correspond-
ing to mode shape {ϕi}. Re-writing equation (10.5), equilibrium state during free-
vibration response of an MDOF system can be defined with the following matrix and
vector terms:
FIGURE 10.11 Schematic representation of an automotive body structure subjected to
bending mode of deformation.
254 Finite Element Analysis of Solids and Structures
={ui) ={ui)
+u1 +u1 u
{ui} = { i
sin( it)
+u2 -u2
+u3 -u3 me, t
(a) (b) (c)
FIGURE 10.12 (a) Vibration mode-1; (b) vibration mode-2; and (c) time function of the
vibration mode response.
k . u m .{u} 0 (10.33)
where [k] is the stiffness matrix, [m] the mass matrix, {u} the displacement response
vector and {u} the vector of accelerations of nodal DOF. Substituting expression
from equation (10.32) into (10.33):
k . i i2 m . i sin i t 0 (10.34)
Since equation (10.34) is true for any value of time function sin(ωit), the following
condition must be satisfied by the stiffness and mass property matrices of the MDOF
system:
k . i i2 m . i 0 (10.35)
Equation (10.35) resembles the standard eigenvalue problems encountered in many
engineering mechanics problems, where λi (= ωi2) is the eigenvalue and ϕi is the
eigenvector. Properties of eigenvector dictate that mode shapes be orthogonal to one-
another, i.e.:
i . j ij
T
(10.36)
where δij is the Kronecker delta (δij = 1 when i = j and δij = 0 when i ≠ j). Structural
vibration frequency analysis problem of equation (10.35) can be analytically
described by the following polynomial of degree “n” where “n” is the number of
DOF in the system:
determinant k i2 m 0 (10.37)
Solutions of equations (10.37) can be analytically calculated for 2 or 3 DOF.
Special numerical techniques are required for large systems. Commonly used
Vibration Frequency Analysis of Structures with FEA Model 255
solution techniques focus on determining the mode shapes (i.e. eigenvectors), ϕi, that
are subsequently used to calculate the frequency values. Pre-multiplying equation
(10.35) with {ϕi}T, and after rearranging the terms:
i k . i
T
i
2
i (10.38)
i m . i
T
The terms defined in equation (10.38) are known as Rayleigh’s quotient having
the following properties (Bathe 1996):
1 2 . n (10.39)
With positive definite[k] and [m] matrices, equation (10.38) readily provides the
vibration frequency values of MDOF structure provided that the mode shape vectors
{ϕi} are known. The minimum value of λi providing the lowest vibration frequency
is known as the fundamental vibration frequency of structure. The pair of eigenvalue
λi and corresponding eigenvector ϕi is commonly known as eigenpair. Eigenvectors
are occasionally normalized to represent the following condition:
i . m . i I
T
(10.40)
where [I] is a diagonal matrix of unit values. The numerator on the right side of equa-
tion (10.38) represents the effective stiffness value of structure (structural resistance)
against mode shape ϕi, and the denominator represents the effective mass value par-
ticipating in that mode of vibration. Stiffness matrix of the structure, [k], is calculated
by using the standard finite element formulation given in equation (2.44). Calculation
of mass matrix [m] from the finite element model of a structure is described in the
following Section 10.7. Section 10.8 will introduce the unique numerical techniques
required for the calculation of vibration mode shapes {ϕi} and corresponding fre-
quencies ωi.
10.7 CALCULATION OF SYSTEM MASS MATRIX
FOR MDOF SYSTEMS
Discrete lump mass elements at nodal DOF can be directly assembled into the system
mass matrix. Mass matrix calculation for distributed mass systems requires special
numerical procedures. Using the same displacement interpolation functions (alterna-
tively known “shape functions”), described earlier for calculation of deformation
inside finite elements, the acceleration response at a material point inside a finite
can be calculated from the nodal acceleration responses (ui) as follows:
element (u)
u Hi ui (10.41)
256 Finite Element Analysis of Solids and Structures
Exact definition of the matrix [Hi] in equation (10.41) will depend on the type of
finite elements used to model the structure. For example, considering the 2D solid
element of Figure 3.8(a), the acceleration responses in two orthogonal directions at a
material point inside the element can be calculated from the nodal acceleration val-
ues as follows:
u1
v1
u2
(10.42)
u H1 0 H2 0 H3 0 H4 0 v2
v 0 H1 0 H2 0 H3 0 H 4 u3
v3
u4
v4
where Hi (i = 1, 2, etc.) are the iso-parametric shape functions defined in Equations
(3.65). Adding the inertia resistance term, the virtual work of equation (2.39) can be
re-written as:
dV u u dV U B dV H u dV (10.43)
T T T T
T
i
where ρ is the density of material. Substituting equation (10.41) into (10.43), the
equilibrium state between internal resistances and external forces can be expressed
by re-writing equation (2.43) as follows:
B T . C . B .dV
. u . H T . H .dV
i i i
. u P
i
(10.44)
where {ui} and {ui} are displacements and accelerations at the nodal DOF. The sec-
ond integral term on the left side of equation (10.44) represents the mass matrix of
finite element:
m . Hi . Hi .dV
T
(10.45)
Equation (10.45) produces a positive definite mass matrix formulation that is essen-
tial for the calculation of frequency values from equation (10.38). Mass matrix
derived from numerical integration of equation (10.45) is commonly referred to as
consistent mass matrix since the same displacement interpolation functions are used
for the interpolation of acceleration response. For the simple case of two DOF truss
element in Figure 2.11, equation (10.45) provides the following consistent mass
matrix definition:
Vibration Frequency Analysis of Structures with FEA Model 257
AL 2 1
(10.46)
m
6 1
2
where “ρ” is material density, “A” the cross-sectional area of truss element, and “L”
the length of member. Taking summation of the terms in each row, and lumping the
value at diagonal position will lead to the following definition of diagonal mass
matrix:
AL 1 0
(10.47)
m
2 0
1
Lumped mass matrix definition in equation (10.47) shows half of the element
mass effective at each of the two translational DOF. Consistent mass matrix for
square 2D solid element of Figure 3.16(a), corresponding to 4 translational DOF in
the x-direction only, is obtained from equation (10.45) with the use of shape func-
tions from Equations (3.65):
4 2 1 2
At 2 4 2 1 (10.48)
m
36 1 2 4 2
2 1 2 4
where “ρ” is material density, “A” the plan view area of 2D solid, and “t” the thick-
ness of element. By using the row summation technique, the diagonal lumped mass
matrix is obtained as:
1 0 0 0
At 0 1 0 0
(10.49)
m
4 0 0 1 0
0 0 0 1
Effective mass at each DOF of 4-node square element turns out to be one-quarter
of the element mass. Lumped mass matrix formulation with equal distribution of ele-
ment mass at corner nodes, however, is not applicable for higher order elements and
for elements with rotational degrees of freedom. For the transverse shear and bending
deformation modes of a 2D beam element (Figure 6.21, Equations 6.75), the consis-
tent mass matrix from equation (10.44) is calculated as:
156 22 L 54 13L
AL 22 L 4 L2 13L 3L2
m (10.50)
420 54 13L 156 22 L
13L 3L2 22 L 4 L2
258 Finite Element Analysis of Solids and Structures
where “ρ” is material density, “A” the cross-sectional area of beam element, and “L”
the length of member. The diagonal lumped mass matrix of this beam element is
defined, based on engineering intuition of beam deformation behavior, as follows:
1 0
0 0
2
0 L
0 0
AL 12
m (10.51)
2 1 0
0 0
L2
0 0 0
12
where 1st and 3rd diagonal terms represent half of beam mass effective at each of the
transverse DOF; and the second and fourth diagonal terms represent the mass moment
of inertia for spinning motion of half-length of beam about each end:
L /2
AL L2
0
x 2 . A .dx .
2 12
(10.52)
Various techniques of lumped mass matrix formulations have been extensively
documented by Cook et al. (1989) and Zienkiewicz and Taylor (1991). Both mass
matrix formulations, consistent and lumped, have their merits and demerits. There is
no standard procedure to assess the accuracy of results obtained by one or the other
method. The lumped mass matrix is a convenient numerical tool at the expense of
some accuracy. Consistent mass matrices tend to be accurate for flexural problems
when modeled with beam and shell elements. In nonlinear impact simulations, e.g.
crashworthiness analysis of automotive structures, lumped mass matrices are com-
monly used for computational efficiency as well as for greater numerical stability
against spurious oscillations. Finite element software packages may selectively use
lumped mass matrix formulations for computational efficiency reasons. ABAQUS
software, for example, uses lumped mass matrix formulations for linear truss and
solid elements, and consistent mass matrix for higher order solids, beams, and shell
elements.
In some structural analysis models, physical mass properties could be lumped to
selected translational DOF, while the rotational DOF of the system is assumed to be
massless. Vibration analysis problem of equation (10.35) can be solved by reducing
the full n × n system to a reduced set containing “m” master DOF that have nonzero
mass properties. Dividing the equations (10.35) into to subsets of master DOF (with
mass properties) and massless slave DOF:
kmm kms m 2 mmm 0 m 0
. i . (10.53)
ksm kss s 0 0 s 0
Vibration Frequency Analysis of Structures with FEA Model 259
where “m” refers to the master DOF (to be retained), “s” refers to the slave DOF (to
be reduced), and [ksm] = [kms]T. Using the bottom part of equations (10.53), the trans-
formation relationship between reduced and full system can be written as follows:
m I
Z m ; where Z 1
(10.54)
s
ss sm
k .k
The reduced stiffness and mass matrices are defined by using the transformation
matrix {Z}:
T T
k m x m Z k Z m m x m Z m Z (10.55)
Reduced stiffness and mass matrices, defined in equation (10.55), can be used in
equation (10.35) in lieu of full system matrices for vibration frequency analysis of
the structure. Condensation of the massless DOF, commonly known as Guyan reduc-
tion, is frequently used in earthquake response analysis of tall building frames.
As discussed in Section 1.4, a structure with un-constrained rigid body modes will
not produce a positive definite stiffness matrix [k]. A work-around for this issue is to
shift the eigenvalue problem of equation (10.35) by applying an arbitrary shift “α” as
shown in the following:
k . m . i i2 m . i 0
(10.56)
This “shifted” system equations can be solved by one of the numerical techniques
described in the following Section 10.8. The first vibration frequency, predicted by
solving equations (10.56), is the one closest to the applied shift value (α).
10.8 NUMERICAL CALCULATION OF VIBRATION MODE
SHAPES AND FREQUENCIES OF MDOF SYSTEMS
Assuming that system property matrices [k] and [m] in equation (10.35) are of n × n
size, there exist “n” number of frequencies and corresponding mode shapes for the
system. Several different numerical techniques exist, with varying degrees of effi-
ciency, for solving the eigenvalue problem of equation (10.35). Effectiveness of solu-
tion techniques to be used depends on the relative ratio between the number of
desired mode shapes “p” to the size of the overall problem “n” (p ≤ n). When all
mode shapes and frequencies are required for a small size problem, Jacobi transfor-
mation method provides effective solution to the Eigenvalue problem. In that method,
a transformation matrix [P] is developed through iterations to diagonalize the stiff-
ness and mass matrices of the system:
T
P k P diagonal kii
T (10.57)
P m P diagonal mii
260 Finite Element Analysis of Solids and Structures
The columns of matrix [P] represent the eigenvectors (i.e. mode shapes), and the
determination of eigenvalues becomes a straight-forward operation given by
i i2 kii (10.58)
mii
A systematic trial-and-error approach is followed to find the transformation matrix
[P] that progressively leads the system property matrices [k] and [m] to diagonal
forms. Householder-QR method is another matrix transformation method that ini-
tially transforms the matrices [k] and [m] to tri-diagonal forms, and eventually pro-
duces diagonal property matrices in the subsequent steps. Detail numerical steps for
the derivation of transformation matrices, with convergence checks of successive
iterations, can be found in Bathe (1996).
When a small number of mode shapes and frequencies are required for a large
system (p << n), vector iteration method provides a very efficiency solution strategy.
Starting with a trial vector {x1}k for mode shape {ϕ1}, an improved prediction for the
vector is obtained by solving the following re-arranged form of equation (10.35):
k . x1k 1 12 m . x1k (10.59)
where ω12 is the value predicted from equation (10.38) by using the trial vector {x1}k.
Successive iterations with equations (10.38) and (10.59) eventually lead to a con-
verged eigenpair solution of ω1 and {ϕ1}. Iterations for second eigenpair follow the
same iterative scheme, but with added interim step to enforce orthogonality condi-
tion between mode-1, {ϕ1}, and the next trial vector {x2}k:
x2 k x2 k 1 . x2 k . 1
T
(10.60)
Actual numerical implementation, however, does not need to follow the sequen-
tial extraction of mode shapes one by one. A set of trial vectors can be used simulta-
neously to progressively improve all predictions in each iteration step. Numerical
implementation of such a multi-vector trial scheme is commonly referred to as “sub-
space” iteration method (Bathe 1996). It is a good analysis practice to use a larger
number of trial vectors “q” for producing a good set of “p” mode shapes (p < q << n).
The method works efficiently when only a handful of eigenpairs are desired, e.g., 2
or 3 mode shapes are generally desired in dynamic analysis of tall building frame
structures. However, a much larger number of modal frequency values need to be
checked in vibration-sensitive structural designs, such as in automotive body struc-
tures. Computational efficiency of standard subspace iteration method goes down
significantly when “p” becomes high (p > 20). An accelerated form of subspace itera-
tion method is sometimes constructed by extracting a limited number of modes in
one step, and then by extracting additional mode steps in subsequent steps from the
solution of shifted eigenvalue problem (equation 10.56). Lanczos transformation
method, producing tri-diagonal forms of system property matrices ([k] and [m]), is
often used in conjunction with subspace iteration method to extract large number of
Vibration Frequency Analysis of Structures with FEA Model 261
eigenpairs efficiently. An important consideration in iterative solution of eigenvalue
problem is the verification with Sturm sequence check that all eigenpairs within a
frequency target range have been calculated (Bathe 1996). General-purpose finite
element software packages tend to offer many of the above referenced alternative
eigensolver options to be explicitly selected by the analyst when setting up a modal
analysis problem.
10.9 VIBRATION FREQUENCY ANALYSIS WITH ABAQUS
Finite element analysis models, described in earlier chapters for static load–deflec-
tion analysis, remain the same for stiffness matrix calculations in vibration analysis
models. Material property descriptions need additional data inputs for density values,
with the use of data block identifier *DENSITY as shown in the following:
*MATERIAL, NAME=mat-1
*ELASTIC
210,0.3
*DENSITY
7.8e-6
The specified material density value is used, with equation (10.45), for the calcula-
tions of element mass matrices [m]. Discrete mass values can be assigned to selected
nodes by defining virtual elements with TYPE=MASS:
*ELEMENT, TYPE=MASS, ELSET=aName
ID no, nodal ID
ID no, nodal ID
ID no, nodal ID
…..
Actual mass value to be assigned to the translational DOF of nodes selected in mass
element descriptions are defined with *MASS data block identifier:
*MASS, ELSET=aName
a numerical mass value
Discrete lumped mass values are combined with element mass matrices, calculated
from finite element analysis model, to get the overall mass matrix [m]. An ABAQUS
analysis step for vibration mode analysis can be constructed by using the following
commands:
*STEP
*FREQUENCY
p, fmin, fmax, λ
* END STEP
262 Finite Element Analysis of Solids and Structures
where “p” is the number of desired mode shapes from the analysis model. This field
can be left blank if the maximum frequency of interest (fmax) is specified and the
evaluation of all the eigenvalues in the given range is desired. The optional shift
parameter (λ) in ABAQUS analysis is specified in the unit of (f 2). Calculated modal
frequency values (fi) are saved by ABAQUS in standard output data file (*.dat). Mode
shape data are saved in the general binary database output file (*.ODB). When no
specific eigensolver routine is selected, ABAQUS uses the LANCZOS method by-
default for eigenpair extraction. However, a user may optionally select the subspace
iteration method, if desired, by using the following syntax:
*STEP
*FREQUENCY, EIGENSOLVER=SUBSPACE
p,, fmax, λ, q
* END STEP
The number of trial vectors (q), if omitted, is internally set by ABAQUS as the mini-
mum of (2p and p + 8). ABAQUS normalizes the mode shapes, by default, so that the
largest displacement or rotation value in each reported vector is unity. However, a
user may optionally specify the mass normalization method (equation 10.40) to be
used in eigenproblem solution:
*STEP
*FREQUENCY, EIGENSOLVER=SUBSPACE, NORMALIZATION=MASS
p
* END STEP
10.10 PRACTICE PROBLEMS: VIBRATION ANALYSIS
OF STRUCTURES
PROBLEM 1
Figure 10.13 presents geometry, material, and boundary condition data for a thin-
wall C-section beam. Using SDOF simplified modal analysis concepts, calculate by
Mid-plane dimensions of C-sec on beam: 58x96 mm
Uniform wall thickness: 4mm
Material : aluminum
Fully (E=70GPa, =0.33, =2.7x10-6kg/mm3)
constrained end
96
mm
58 mm
1200
Fully
z constrained end
y
x
FIGURE 10.13 Properties of a thin-wall C-section beam.
Vibration Frequency Analysis of Structures with FEA Model 263
hand the vibration frequencies of the member for: (1) lateral vibration mode in the
x-direction, (2) vertical mode in the z-direction, and (3) torsional mode about the
y-axis. Build a finite element model of the member using shell elements; and conduct
modal frequency analysis for the first five modes of vibration. Compare the hand
calculation results (frequency and mode shapes) with the finite element model results
PROBLEM 2
Roof panel of an automotive body structure is made of 0.65 mm thick steel panel –
spot welded @ approximately 60 mm spacing along the edges to the perimeter body
structure frame. Geometry data of the roof panel mid-surface is available in the file
Roof_Panel.step. Prepare a finite element analysis model of the roof panel assuming
fixed boundary conditions at the spot weld locations; calculate the lowest vibration
frequency of finite element shell model; and plot the corresponding mode shape.
Using equation (10.30), what will be the vibration frequency of an equivalent simply
supported flat panel having approximate panel dimensions shown in Figure 10.14?
What are the reasons for differences between FEA and hand calculation results?
PROBLEM 3
Re-analyze finite element analysis model of Problem-2 assuming 1.0 mm thick alu-
minum roof panel (E = 70 GPa, ν = 0.33, ρ = 2.7 × 10–6 kg/mm3). Compare the fun-
damental vibration frequency responses of 2 alternative material choices (steel versus
aluminum). What is the relative weight ratio of aluminum versus steel design?
PROBLEM 4
Re-analyze the roof panel finite element model, without support constraints at spot-weld
locations, by using the frequency shifting technique described by equation (10.56). Verify
the vibration frequency result of shifted model against that of base model.
PROBLEM 5
Beam structure in Figure 6.30 is made of steel (density, ρ = 7.8e−6 kg/mm3). Assume
an additional lumped mass of 10 kg supported at point C. Calculate the fundamental
vibration frequency of the system; and verify the result with hand calculations.
1050
1480
Mid-surface geometry of a vehicle roof panel: Roof_Panel.step
Uniform sheet metal thickness: 0.65 mm
Material : steel (E=210GPa, =0.3, =7.8x10-6 kg/mm3)
FIGURE 10.14 Properties of a vehicle roof panel (Roof panel extracted from vehicle FEA
model: courtesy of NHTSA 2020).
11 Linear Dynamic
Response Analysis
of Structures
SUMMARY
Structural response to cyclic dynamic loads and the management of dynamic response
through frequency separation have been discussed in Chapter 10. However, struc-
tures are also frequently subjected to non-cyclic short-duration dynamic events. This
chapter is dedicated to the analysis of structural response for such non-cyclic dynamic
load events. Duhamel integral formulation, to predict the elastic dynamic response of
SDOF systems to external impulse loads, is introduced in Section 11.1. This analysis
technique provides very useful conclusions on the range of dynamic amplifications
that a system can experience when subjected to impulse events of arbitrary duration.
The concept of design response spectra, based on the envelope of peak dynamic
responses of linear elastic systems, to pre-defined single loading function, is dis-
cussed in Section 11.2. Duhamel integration method, although very powerful for
predicting the linear dynamic response of SDOF systems, is not an efficient method
for predicting the time history response of MDOF systems. The alternative direct
integration techniques, for time-domain dynamic response simulations, are intro-
duced in Sections 11.3–11.6. Section 11.4 specifically focuses on the accuracy and
stability aspects of implicit time integration method. Relative efficiencies of direct
integration versus modal superposition methods, for linear elastic dynamic response
prediction, are discussed in Sections 11.5 and 11.6. Explicit time integration method,
which is more potent for nonlinear dynamic response analysis, is introduced in
Section 11.7. ABAQUS-specific commands for dynamic response analysis of finite
element models are reviewed in Section 11.8 followed by the presentation of practice
problems in Section 11.9.
11.1 LINEAR ELASTIC RESPONSE OF SDOF SYSTEMS
TO IMPULSIVE LOADING
Figure 11.1 shows a general dynamic loading F(τ) acting on an SDOF spring-mass
system. Re-arranging the terms of equation (10.1), the effect of dynamic force F(τ),
acting for a short time duration, dτ, can be expressed as follows:
1
du F .d (11.1)
m
265
266 Finite Element Analysis of Solids and Structures
Impulse:
F( ).d
support
F( )
Time, t S ffness, k
d Mass, m
u dů(τ)
du(t- ) = sin[ .(t )]
Time,
Time
e, t
FIGURE 11.1 Impact force acting on a flexible SDOF spring-mass system.
where du is the change in velocity of mass (m) caused by the impulse [F(τ). dτ].
Vibration response of the spring-mass system at a time (t), following the application
of infinitesimal impulse, is given by (Clough and Penzien 1975)
du
du t sin . t (11.2)
Substituting the expression from equation (11.1) into equation (11.2), displacement
of the spring-mass system due to short-duration impulse effect is given by
1
du t F .sin . t .d (11.3)
m
where ω is the natural vibration frequency of spring-mass system (defined by equa-
tion 10.7). Integrating both sides of equation (11.3) over the duration of impulse (τ:
0 to t), displacement response of the system, for a general dynamic load of duration
(t) and for the initial state of u(0) = 0, is given by
t
1
u t
m F .sin . t .d
0
(11.4)
Equation (11.4) is commonly known as Duhamel integral – a powerful expression for
calculating the undamped linear elastic dynamic response of an SDOF system when
subjected to a general dynamic loading F(τ). With known system properties ω and m,
elastic dynamic response u(t) can be calculated, for any given force function F(τ), by
conducting step-by-step numerical integration of the expression on the right side of
equation (11.4).
Linear Dynamic Response Analysis of Structures 267
Spring resistance,
(a) (c)
External force, F( ) k.u(t)
(b) Spring resistance, k.u(t) (d)
Spring resistance, k.u(t)
FIGURE 11.2 (a) External dynamic load F(τ); (b) spring force response to dynamic external
load; (c) response to a sudden impact force; (d) response to a slow external load.
An example case study is considered here assuming an SDOF with mass, m = 1,
and stiffness, k = 39.4784, such that the period of vibration of the system comes out
to be, T = 1. Figure 11.2(a) shows a dynamic force function reaching to unit peak
amplitude over a time duration of 0.5. Numerical integration of equation (11.4) pro-
duces the displacement response of the mass u(t), that, when multiplied with spring
stiffness k, gives the dynamic spring force response (shown by the dotted line in
Figure 11.2(b)). The dynamic amplification of the spring response, thus, happens to
be 1.64 when the force function ramps up to unit peak amplitude at half time of the
period of system vibration. Figure 11.2(c) shows the predicted spring force response
when the external load is applied suddenly over an infinitesimally short duration of
time. Peak spring force reaches a value of 2 times the amplitude of applied external
force, thus, defining a limit for the maximum dynamic amplification that can occur
when a sudden impact force is applied on a system. On the other hand, if the external
force is applied slowly, over a time span equal to or larger than the period of vibration
of the system (Figure 11.2(d)), spring force response oscillates about the applied
external force trend line before eventually approaching the value of applied external
force. This implies that, in the limit case of very slowly applied force (time to peak
force > period of structural vibration), system response can be predicted by solving
the static equilibrium equation for applied peak force (equation 1.1). Figure 11.3
summarizes the peak dynamic responses, obtained from equation (11.3), for the
external force function with variable time-to-peak values. Dynamic amplification of
system response diminishes as the time-to-peak of externally applied force becomes
268 Finite Element Analysis of Solids and Structures
FIGURE 11.3 Dynamic amplification of structural response as a function of the ratio
between time-to-peak force and vibration period of system.
longer than the period of vibration of the structure. For shorter duration impact events
(t < T), the Duhamel integral approach provides useful information about the dynamic
amplification of a system response for external loading function of any general shape.
11.2 RESPONSE SPECTRUM OF LINEAR DYNAMIC SYSTEMS
Dynamic amplification of a system response, as discussed in Section 11.1, is deter-
mined by the relative relationship between rate of loading and the period vibration of
structure. In engineering design practices, generic external loading functions are
often pre-defined, such as earthquake loading functions for building design in spe-
cific geographic locations, and road input loads for automotive system designs.
Linear dynamic response of a system, to the pre-defined loading function, can be
predicted by using the Duhamel integral technique described in Section 11.1. Figure
11.4(a) shows an example force function of trapezoidal shape (solid line spanning a
time duration of t = 1.5). The predicted force responses of two SDOF spring-mass
Spring force Spring response
Response spectrum
(for T=0.8) (for T=2)
External
force, F( )
(a) (b)
FIGURE 11.4 (a) External force and example response histories; (b) response spectrum
(envelop of maximum responses) as a function of vibration period of structure.
Linear Dynamic Response Analysis of Structures 269
systems, having period of vibration values T = 2 and T = 0.8, are shown by the dotted
lines in the same plot. Similar response plots can be generated, for the same force
function, but with many possible values of the structural vibration period (T). Taking
the peak values from such response histories, Figure 11.4(b) shows an envelope of
the peak force values for different values of the structural vibration period – gener-
ally known as response spectrum in structural design discipline. Like this example of
spring force response spectrum, similar envelopes can also be generated for accelera-
tion, velocity, or displacement responses of the system. From a pre-defined response
spectrum, the expected peak dynamic response can be estimated easily during pre-
liminary design iterations of a structure.
In practical engineering applications, engineers are often interested to know the
maximum response amplitude of a system for standard loading functions. The
response spectra method, described in the above for SDOF systems, becomes very
useful to predict the peak response of MDOF systems as well. Vibration periods of
an MDOF structure can be calculated, by using the techniques presented in Chapter
10, when system property matrices, [k] and [m], are known from a finite element
model of the system. Peak response for each of the known modal period of vibration
can be calculated from the response spectrum – assuming each mode responds inde-
pendently as an SDOF system. As it is evident from Figure 11.4(a), peak responses
of different modes (with different vibration period values) will occur at different
times. Superposition of the modal response values to generate the combined system
response requires special considerations (Tedesco et al. 1999, Chopra 2017). One
commonly used combination technique is to take the square root of the sum of the
squares of all relevant modal peak response values:
x
2
Max response i
(11.5)
i 1
where xi (i = 1…n) are the peak response values for individual mode shapes of an
MDOF system. A good estimate of the peak structural response can be produced by
considering a small number of structural vibration modes when the system response
is dominated by a handful of low-frequency modes (such as earthquake and wind
load effects on bridges and tall buildings). However, a pre-defined design response
spectrum may not always represent case-specific general dynamic load scenarios;
thus, requiring time-domain response analysis of structures in many engineering
development projects. Numerical integration of equation (11.4), as demonstrated
with elementary case studies, can predict time history of dynamic response for SDOF
systems. The method is, however, not convenient when numerous mode shapes may
need to be considered to predict solutions for large complex systems. More special-
ized numerical technique, based on direct step-by-step integration of the system
equilibrium equations (10.17), is preferred because of the generality of the method
that can be easily adapted to nonlinear problems as well when needed (to be dis-
cussed in Chapter 12). Sections 11.4–11.6 in the following review different aspects
of the numerical solution techniques for time-domain response history analysis of
structures.
270 Finite Element Analysis of Solids and Structures
11.3 TIME-DOMAIN ANALYSIS OF DYNAMIC
STRUCTURAL RESPONSE
Following the developments from Chapter 10, dynamic equilibrium of a damped
structure at a time (t) can be expressed as follows:
k.u t c.u t m.u t F t (11.6)
Solution of dynamic equilibrium equation (11.6), as an analytical function of time
(t), is difficult to achieve for general structural systems. Instead, solutions of dis-
cretized finite element models can be derived by considering system equilibrium at
discrete time steps. Re-writing equation (11.6), equilibrium after a small time incre-
ment, Δt, can be expressed as follows:
k.u t t c.u t t m.u t t F t t (11.7)
The derivation of actual solution at time (t + ∆t) requires certain assumptions about
the variations of displacement, velocity, and acceleration responses over the time
step Δt. As one of the simplest forms of numerical approximation, taking the average
of acceleration responses u t and u t t over the time step Δt, kinematic rela-
tions among displacement, velocity, and acceleration can be defined as follows:
t
u t t u t u t u t t (11.8)
2
1
u t t u t u t .t [u t u t t ] t
2
(11.9)
4
Assuming that the stiffness remains un-changed during the time interval Δt, resis-
tance to deformation can be expressed as follows:
k.u t t R t k.u (11.10)
where R(t) is the internal resistance of the structure to deformations, calculated from
the stress responses of all finite elements by using equation (2.42). Combining equa-
tions (11.7)-(11.10), and after re-arranging the terms:
4 2 4
t 2 m t c k u F t t R t m. t u t u t c. u t (11.11)
Equation (11.11), with known quantities on the right-hand side, can be solved to
determine the incremental displacement response Δu occurring over the time step Δt.
The dynamic response of an undamped spring-mass system (with T = 0.8, m = 1,
k = 61.685), for the trapezoidal force function of Figure 11.4(a), is analyzed with
Linear Dynamic Response Analysis of Structures 271
Spring response:
k.u(t)
External Inertia resistance: External force, F( ( k.u(t)
force, F( ( m. (t)
Inertia resistance: m.
(a) (b)
FIGURE 11.5 (a) Dynamic responses of a flexible SDOF spring-mass system; (b) force
responses of a very stiff system.
step-by-step application of equation (11.11); and the time histories of external force
and internal resistances (for both inertia and spring deformation) are shown in
Figure 11.5(a). Evidently, relative contributions of spring resistance and inertia com-
ponents vary with time. The system resistance components oscillate with time while
maintaining overall equilibrium with the externally applied force. When the spring
stiffness is increased by a factor of 100 (k = 6168.5), the mechanical resistance domi-
nates the overall system response; and the inertia component of resistance mecha-
nism becomes relatively very small (Figure 11.5(b)). In the limit case of relatively
negligible contribution from the inertia resistance, solution of the dynamic equilib-
rium of forces (equation 11.6) converges to the solution obtained by considering pure
static equilibrium state (equation 1.1), thus proving the validity of the step-by-step
numerical integration formulation presented in equation (11.11). Equations (11.8)
and (11.9) have used average value of accelerations at time “t” and “t + Δt”, to esti-
mate the velocity and displacement response changes over time step Δt. Evidently,
many other assumptions can be made to forecast the response variations over that
small discrete time step Δt. Accuracy and stability of various commonly used numer-
ical integration techniques are discussed in Section 11.4.
11.4 NUMERICAL INTEGRATION PARAMETERS FOR
TIME-DOMAIN ANALYSIS OF STRUCTURES
Assumptions made to approximate the variation of dynamic motions (i.e. of dis-
placement, velocity, and acceleration), over the times step Δt, have led to numerical
integration techniques of many different names in the field of structural dynamics.
Section 11.3 presented time-domain solution method using the assumption of aver-
age acceleration response over the time step Δt. A more general form of response
variation can be considered by re-writing the kinematic relations of equations (11.8)
and (11.9) as follows:
u t t u t 1 u t .u t t .t (11.12)
272 Finite Element Analysis of Solids and Structures
1
u t t u t u t .t
u t
.u t t t
2
(11.13)
2
where parameters β and γ provide a general framework for defining response varia-
tions over time step Δt; and the method is commonly known as Newmark integration
method (Newmark 1959). The average acceleration method, presented in Section 11.3,
is a special form of the Newmark method when integration parameters are defined as
β = 1/4, γ = 1/2. Linear acceleration method, another widely used assumption in
structural dynamics calculations, is derived from the Newmark method with para-
metric definitions of β = 1/6, γ = 1/2. A special form of linear acceleration method is
the Wilson θ method that assumes linear acceleration trend applies over a time span
from “t” to (t + θ.Δt) where θ ≥ 1 (Bathe and Wilson 1976). An enhanced form of
Newmark’s average acceleration method has been proposed by Hilber et al. (1977) to
include a numerical damping parameter α in the step-by-step calculation of structural
dynamic response by defining the numerical integration parameters as follows:
1
2
1 (11.14)
;
4 2
The value of α is usually defined in the range of (–1/3 ≤ α ≤ 0). This method degener-
ates to Newmark’s average acceleration method for α = 0. The constant or linear
acceleration assumptions, as used in the above integration techniques, are in fact
simplifications of infinite Taylor series expressions for continuous time-domain
response functions (Modak and Sotelino 2002). Modak’s T-method uses higher order
Taylor series expressions for displacement, velocity, and acceleration variations in
the time domain. The dynamic equilibrium equation is derived by using a weighted-
residual approach, and a recursive integration technique is derived with 9 numerical
parameters after truncating the Taylor series expansions. Since this method allows a
wide range of values for the nine parameters, it provides the opportunity for optimi-
zation of the parameters to make the algorithmic error minimum while keeping it
unconditionally stable and second-order accurate. The optimal form of this general-
ized method provides higher numerical accuracy compared to other integration
methods currently available in finite element software packages.
The direct step-by-step integration methods, commonly known as implicit time
integration methods, enforce system equilibrium at discrete time steps (t, t + Δt, …
etc.) – thus providing stable solution over the time domain. Accuracy of the solution
is somewhat affected by the assumptions made about the nature of response variation
over time step. However, the most critical parameter affecting the accuracy of solu-
tion appears to be the size of time step (Δt) selected for discrete solution steps. For
demonstration purpose, an SDOF spring-mass system, stretched to an initial defor-
mation state of u = 1 at time t = 0, and then released to undergo free vibration, is
considered for step-by-step time-domain analysis (Figure 11.6). The period of vibra-
tion of the system is calculated as follows:
1 m 21.54
T 2 . 2 50 ms (11.15)
f k 0.34
Linear Dynamic Response Analysis of Structures 273
Reference k=0.34 kN/mm
posi on, u=0 m=21.54 kg
c=0
Displacement
u=1
at me=0
velocity, ů = 0
accelera on, ü = 0
(a) (b)
FIGURE 11.6 (a) Spring-mass system stretched to an initial displacement of u = 1 at time =
0; (b) calculated displacement response with time step Δt = T/10.
Upon release of the mass from displaced state of u = 1, the spring-mass system is
expected go through simple harmonic free-vibration oscillations with an amplitude
of “1”. The response of the system can be calculated, by using the step-by-step
implicit integration of equation (11.11), with any assumed value of time step Δt. The
time history of predicted response with Δt = T/10 = 5 ms, shown in Figure 11.6(b),
shows harmonic response with a peak amplitude of 0.99999≈1. Figure 11.7 shows
the error in predicted peak amplitude as a function of the time step used in numerical
calculations. Evidently, the accuracy of predicted response goes down for time steps
Δt > T/10. The general analysis practice is to use a time step smaller than one 10th of
the important vibration period of a structure.
11.5 TIME DOMAIN ANALYSIS OF MDOF SYSTEMS
The time-domain dynamic equilibrium equation of SDOF system (equation 11.6)
can be re-written for an MDOF system by replacing the system property terms m, k
and c with corresponding property matrices, and by replacing the dynamic response
parameters u, u and u with corresponding nodal response vectors:
Implicit
method
FIGURE 11.7 Effect of time step on solution error.
274 Finite Element Analysis of Solids and Structures
k . u t c . u t m . u t F t (11.16)
Stiffness matrix of the structure, [k], is calculated by using the standard finite element
formulation given in equation (2.44); and the mass matrix [m] is also calculated from
the finite element model of a structure as described in Section 10.7. In many general
structural analysis cases, without the presence of distinct damping elements, the
damping matrix is usually defined proportional to stiffness and mass property
matrices:
c a. m b. k (11.17)
Damping matrix, defined in equation (11.17), is often known as Rayleigh damping
matrix; and the multiplication parameters “a” and “b” are known as Rayleigh damp-
ing coefficients. Since the stiffness and mass matrices are reduced to diagonal forms
by modal transformation, the damping matrix defined by equation (11.17) is also
reduced to diagonal form with same modal transformation. Pre-multiplication of
equation (11.17) with the transpose of a mode shape vector, {ϕi}T, and post multipli-
cation with {ϕi}, will provide the following expression for damping:
i
T T T
c i a. i m i b. i k i a b. i2 (11.18)
where ωi is the vibration frequency and {ϕ}i is the mode shape vector that is normal-
ized to give i m i 1. Equating the modal damping definition of equation
T
(11.18) with that of an SDOF system, c .2 k.m , (introduced in equation 10.19):
c i .2 k.m i .2.i a b.i2 (11.19)
Simplification of the equation (11.19) leads to the following relationship between
modal damping value (ξi) and Rayleigh damping coefficients:
1 a (11.20)
i b. i
2 i
Equation (11.20) can be solved to determine the unknown Rayleigh damping coef-
ficients (a and b) with damping factors for any two selected mode shapes (i = j,k).
The Rayleigh damping matrix [c], thus calibrated for two modal damping targets, can
be substituted in dynamic equilibrium equation (11.16) of an MDOF system. Step-
by-step time-domain analysis technique for predicting the response of SDOF system,
presented earlier, is equally applicable to MDOF systems as well where the system
property terms m, k and c in equation (11.11) are substituted with corresponding
property matrices, and the dynamic response parameters u, u and u are substituted
with corresponding nodal response vectors:
Linear Dynamic Response Analysis of Structures 275
4 2
t 2 m t c k u F t t R t
(11.21)
4
m .
t
u t u t c . u t
Solution of MDOF system equilibrium equations, with Rayleigh damping coeffi-
cients calculated for two selected mode shapes (equation 11.20), does not induce
uniform damping for all modal frequency values (Figure 11.8). The stiffness propor-
tional damping term (“b. ωi” in equation 11.20) shows linearly increasing damping
for higher frequency values – leading eventually to unrealistically high damping of
high-frequency modes. However, where few low-frequency modes dominate the sys-
tem response, high damping value imposed on high-frequency modes, by Rayleigh
damping formulation, does not cause a significant error in overall dynamic response
of the MDOF systems. Additional damping values can be added to the diagonal
terms of the Rayleigh damping matrix, [c], if discrete damping elements exist in an
MDOF system.
Solution of incremental displacement response, {Δu}, from equation (11.21),
requires the assembly of all 3 property matrices for the entire system ([k], [m], and
[c]). Enforcement of entire system equilibrium at discrete time steps (t, t + Δt, …etc.)
inherently maintains the stability of solution process although the accuracy can vary
depending on the time step used for step-by-step calculation of system response in
time domain. Evidently, the step-by-step integration of dynamic response with equa-
tion 11.21 implicitly computes the responses of all mode shapes present in an MDOF
system. The use of a constant time step value, Δt, implies that the modal response
values are computed with different degree of accuracy – with higher modes (i.e.
modes with shorter time periods) being calculated with lower accuracy according to
the analysis results of SDOF systems presented in Figure 11.7. In practical engineer-
ing analyses, time step Δt is chosen small enough to accurately calculate the response
of the highest mode of interest (equation 11.22):
FIGURE 11.8 Relationship between damping ratio and frequency (for Rayleigh damping).
276 Finite Element Analysis of Solids and Structures
Tk (11.22)
t
10
where “k” refers to the highest critical mode. Obviously, the identification of critical
mode “k” varies depending on loading history and dynamic properties of the system
being analyzed. In general engineering practice, the highest significant modal fre-
quency of structure is considered four times the highest significant frequency of a
loading function. This conclusion is derived from the observation of dynamic ampli-
fication plot in Figure 10.6 that implies that dynamic response of a structure
approaches the static response when the ratio of applied loading frequency to the
vibration frequency of structure is below 0.25. Fourier series analysis technique can
be used to identify the significant frequency content of a loading function. After
identifying the highest frequency of interest in a given loading function, the time step
for direct integration of system equilibrium equations can be set at (Δt < Tk/10);
where Tk refers to the period of vibration corresponding to four times the critical
loading frequency. Although modal frequency analysis of a structure is not a pre
requisite of the direct integration method, the finite element analysis model of struc-
ture, however, should possess detailed contents to represent up to “kth” mode shapes
correctly. In linear elastic system analyses, the large system property matrices ([k],
[m] and [c]) can be assembled once; however, the calculation steps (defined by equa-
tion 11.21) need to be repeated at discrete time steps Δt for the duration of dynamic
event. Direct time-domain integration of system equilibrium equations (11.21), with-
out the need for solving large eigenvalue problem, provides one way of calculating
the time-domain response of a large system. Modal superposition method, described
in the following Section 11.6, can be an efficient alternative technique for linear
elastic systems when the response is dominated by few low-frequency modes only.
11.6 MODE SUPERPOSITION METHOD FOR ANALYSIS
OF MDOF SYSTEMS
In Section 11.5, dynamic equilibrium equations (11.16) for MDOF system have been
solved by direct integration of responses involving all degrees of freedom in the sys-
tem. When the overall elastic response of a large structure is dominated by few low-
frequency modes (such as tall building and long-span bridges), the modal
superposition technique is used for the convenience and efficiency of combining few
modal response values. Theoretically, the dynamic response at nodal DOF can be
calculated by superposing the time-domain responses of multiple mode shapes:
u(t} xi t (11.23)
where [ϕ] is an n × p matrix of “p” mode shape vectors, each mode with “n” terms
corresponding to “n” DOF in the system (p ≤ n), and xi{t} are time function of
response amplitudes corresponding to mode shapes i = 1,2 … p (defined by equation
10.32). Substituting equation (11.23) into equation (11.16), and pre-multiplying both
sides with [ϕ]T:
Linear Dynamic Response Analysis of Structures 277
k . xi t c . xi t m .
T T T
xi t
T
F t (11.24)
Assuming that the damping matrix is defined by the Rayleigh damping coeffi-
cients (equation 11.17), orthogonality property of the modal shape vectors, ϕi, trans-
forms equations (11.24) into the following set of uncoupled equations in time
domain:
i2 . xi t 2. i . i . xi t
xi t ri t (11.25)
where ωi is the angular velocity of mode shapes (i = 1, 2 … p), ξi is the modal damp-
ing ratio, and ri(t) is the time function of transformed load function corresponding to
mode shape “i”. Equation (11.25) is analogous to the dynamic equilibrium of an
SDOF system that can be solved in time domain by using the Duhamel integral for
undamped systems (equation 11.4), or by using the direct integration technique
described in Sections 11.3–11.4. Time-domain response history of overall system is
calculated by superposing the modal response histories (equation 11.23). The mode
superposition method of dynamic response analysis, thus, involves three distinct
steps: (1) numerical solutions to identify mode shapes and frequencies (discussed in
Chapter 10); (2) time history solution of decoupled modal responses (equation
11.25); and (3) final superposition of modal responses to derive the overall system
response (equation 11.23). The choice between dynamic response history calculation
methods, either by using the modal transformation method, or by the direction inte-
gration of equilibrium equations of entire system, depends on the relative computa-
tional efficiency of two methods. As discussed earlier, direct integration of equations
(11.21) requires processing of an entire system of equations – posing significant
computational demand. If the response of a system is expected to be dominated by a
handful of low-frequency modes (such as earthquake response of tall buildings), few
modes of desired frequency range can be extracted with limited computational effort
(p << n), and the dynamic response of few individual modes can be calculated from
the uncoupled equations (11.25). The modal superposition method of dynamic
response history analysis, however, does not provide any computational benefit for
nonlinear systems that will be the topic of discussion in Chapter 12.
11.7 EXPLICIT TIME-DOMAIN ANALYSIS OF MDOF SYSTEMS
Implicit time-domain analysis of structural response requires the assembly and
numerical operations with the property matrices [k] and [m] of the entire system.
Modal superposition method, which produces a set of decoupled system of equations
(11.25) for dynamic response analysis, also requires the assembly of large system
property matrices at the beginning to find the modal frequencies and mode shapes.
Explicit time-domain analysis technique, on the other hand, provides an alternative
numerical method that considers each nodal DOF an independent entity. The time-
domain responses at future time step are forecasted explicitly from the past responses
of local DOF, without enforcing the equilibrium state over the entire system during
the forecasting process. The unbalanced forces produced during the response
278 Finite Element Analysis of Solids and Structures
forecasting process are equilibrated with iterative updates to the local response quan-
tities. The equilibrium state at time (t) of an MDOF system can be expressed by rear-
ranging the terms of equation (11.16) as follows:
m . u t F t R t X t
(11.26)
where {R(t)} refers to system resistance to deformation which is equal to [k].{u(t)}
for a linear elastic system; and X(t) is the force vector generated by local damping
resistance mechanism. Assuming that initial values {u(0)} and { u 0 } are known for
any given system, equations (11.26) can be readily solved to determine the initial
acceleration response for any unbalanced force on the right-hand side of equation
(11.26). This solution process becomes highly efficient for an undamped system hav-
ing a diagonal mass matrix – requiring no assemblies of large system property matri-
ces. Once the initial state of equilibrium is established, dynamic responses at a next
time step are forecasted based on the current response values of displacement, veloc-
ity, and acceleration. Among various explicit forecasting methods, the forward inte-
gration method is the simplest one that forecasts the velocity and displacement
response at each individual DOF based on the current known responses as follows:
u t t u t u t .t (11.27)
u t t u t u t t .t (11.28)
The displacement values at all DOF, forecasted locally by equations (11.27) and
(11.28), are used to calculate strains and stresses in finite elements. And the corre-
sponding system resistance to deformation, R(t+Δt), is calculated from integral
expression in equation (2.42). The damping resistance {X(t)}, if included in analysis,
is also assembled from local element properties. Equilibrium state at time (t+Δt) is
enforced by solving the equation (11.26) for updated acceleration response at time
(t+Δt). The process of explicit forecast of local responses, and subsequent correction
of acceleration response to take account of unbalanced force field, can proceed step-
by-step to predict the time-domain response of the entire system.
Forward forecasting method, described by equations (11.27) and (11.28), is prone
to numerical instability. A relatively more effective explicit integration method is
defined based on the central difference theorem where the relationships among dis-
placement, velocity, and accelerations are defined over two times steps as follows:
1 (11.29)
u t u t t u t t
2.t
1
u t u t t 2.u t u t t (11.30)
t
2
Substituting the expressions (11.29) and (11.30) into dynamic equilibrium equations
(11.16), and after rearranging the terms, the following explicit relationship is obtained
Linear Dynamic Response Analysis of Structures 279
to calculate the displacement response at time (t + Δt) based on the equilibrium states
at (t) and (t–Δt):
1 1
t 2 m 2. t c u t t
2 1 1
t
F t k 2 m u t 2 m
t 2 . t
c u t t (11.31)
Equation (11.31) can be solved at local DOFs provided that mass and damping matri-
ces are of the diagonal form. Explicit calculation of dynamic motion, thus, does not
require the assembly and decomposition of a positive definite stiffness matrix which
is a fundamental requirement in static load–deflection analysis with finite element
models. The very nature of explicit time integration method, which uses local
response variables to forecast the future response state, requires that the time step
size must not be large. Re-analysis of the free-vibration response of the spring-mass
SDOF system (Figure 11.6), using the explicit time integration method, produces
solution error much higher than that of implicit method (Figure 11.9). In fact, the
solution error increases exponentially as the time step size becomes larger than a
quarter of the period of vibration of the system. For stable and accurate response
prediction, enough number of calculation points must be used within the period of
vibration of the spring-mass system (Figure 11.10), thus defining the following crite-
rion for time step selection – commonly known as the Courant–Friedrichs–Lewy
(CFL) law (Courant et al. 1928):
T
t (11.32)
2
Considering the axial stress wave propagation in a one-dimensional member
(Figure 11.11), the CFL condition of equation (11.32) can be re-written in the follow-
ing from (equation 11.33):
Explicit
method Implicit
method
FIGURE 11.9 Solution errors for implicit and explicit integration methods.
280 Finite Element Analysis of Solids and Structures
FIGURE 11.10 Stable prediction of the vibration response of spring-mass system using
explicit time integration method with Δt = T/2π.
Stiffness, = l
c
Mass, m= .A. lc
lc
FIGURE 11.11 Axial stress wave propagation in a one-dimensional field.
T m Alc lc l (11.33)
t c
2 k AE E/ c
lc
where “c” is the velocity of stress wave propagation in a one-dimensional field.
Equation (11.33) establishes a very important relationship between the time step Δt,
for finite-difference calculation of response time history, and the finite element size
lc. The stability of explicit method requires that the time step limit, imposed by the
size of finite elements (lc), be satisfied – thus making the method conditionally stable.
Physical interpretation of the Courant condition is demonstrated with a one-dimen-
sional wave propagation problem in Figure 11.12. In explicit integration method,
local calculations at node “i” assume that the instantaneous effect of impact force
F(t) is felt by the mass DOF at node “i” only; and the impact wave does not reach the
next node “j” within the same step Δt:
c. t lc (11.34)
Linear Dynamic Response Analysis of Structures 281
F(t) wave travel:
c. t
F(t)
Time, t
length: l c
FIGURE 11.12 Impact wave propagation in a one-dimensional field.
where lc is the distance between adjacent nodes “i' and “j” – commonly known as
characteristic length of finite elements. For 1-D finite elements (truss, cable, etc.), lc
is simply equal to the element length (Figures 11.11 and 11.12). For surface-based
solid elements, such as the shell in Figure 11.13(a), element characteristic length is
often defined with one of the two alternative forms shown in equation (11.35) (LSTC.
COM 2021):
surface area surface area (11.35)
lc or, lc
Max D13,D24 Max L1,L2 ,L3 ,L4
where D13, D24, L1, L2, s…, etc., are geometric dimensions of shell element shown
in Figure 11.13(a). The value of lc is used in equation (11.33) to calculate the
critical time step where the stress wave velocity for planar solid is defined by
equation (11.36) – in terms of elastic modulus (E), material density (ρ), and
Poisson’s ratio (ν):
1 E (11.36)
c .
1 v
2
For solid elements (Figure 11.13(b)), characteristic length is defined by
volume
lc (11.37)
Max A1,A2,A3,A4,A5,A6
4
L3
3
L4 D13 A
L2 A1 2
D24
1 2
L1
(a) (b)
FIGURE 11.13 (a) Geometric dimensions of a shell element; (b) surface areas of a 3D solid
element.
282 Finite Element Analysis of Solids and Structures
where A1, A2 … etc., are surface areas of the solid element. The stress wave velocity
in 3D solid is given by
c
1 v . E (11.38)
1 v 1 2v
In complex structural analysis models, comprising of different element types,
critical time step limits are checked for all elements, and the smallest of all values
defines the time step limit for step-by-step time-domain solution of the MDOF sys-
tem. For an example shell element size of 5 mm, and for material properties of steel
(E = 210 GPa, ν = 0.3, and ρ = 7.8e–6 kg/mm3), equation (11.33) limits the explicit
integration time step to 0.0009 ms. Despite the very tight limit on allowable time step
size, explicit time-domain analysis tends to be very efficient since computations are
carried out at local element and nodal DOF levels without the need for assembly and
operations of large system property matrices. Local uncoupled analysis process also
makes the explicit technique particularly suitable for massive parallel computing
processing of highly nonlinear systems. Finite element model preparation for explicit
time-domain analysis must control the minimum element size so that the minimum
time step is not unduly burdensome for rest of the model analysis. Analysis software
packages, such as LS-DYNA (LSTC.COM 2021), allow artificial scaling of mass at
local DOF to maintain a minimum time step for efficient execution of the step-by-
step analysis of large system models. Mass scaling, however, is not recommended for
critical structural elements. Instead, FEA mesh should be revised to meet the time
step limit without mass scaling. Alternatively, large complex models can be divided
into subdomains; and explicit integration process can proceed independently with
different time steps in different subdomains – followed by periodic synchronization
of interface responses at larger time steps (Borrvall et al. 2014).
11.8 LINEAR DYNAMIC RESPONSE ANALYSIS WITH ABAQUS
Finite element analysis models, described in earlier chapters for static load and vibra-
tion frequency analyses, remain same for time-domain dynamic load analysis of
structures. Additional data inputs are required to specify: (1) time integration param-
eters, and (2) time history of load function. An example analysis step for implicit
dynamic analysis in ABAQUS can be described as follows:
*STEP
*DYNAMIC, TIME INTEGRATOR=HHT-TF
Δt0, tf, Δtmin , Δtmax
*END STEP
In the above example, selected time integration scheme “HHT-TF”, which is the
default ABAQUS method, refers to the α-integration method proposed by Hilber,
Hughes, and Taylor (1977), with slight numerical damping (α = −0.05). Numerical
Linear Dynamic Response Analysis of Structures 283
integration parameters β and γ are calculated by using equation (11.14). If desired,
time-domain integration parameters can be optionally specified for α, β, and γ as
follows:
*STEP
*DYNAMIC, ALPHA=α, BETA=β, GAMMA=γ
Δt0, tf, Δtmin, Δtmax
*END STEP
The value Δt0 in the above specifies the initial time step to be used by ABAQUS in
step-by-step time-domain calculations. ABAQUS has built-in capability to adjust the
time step during analysis, based on convergence characteristics of nonlinear solution
algorithms. The parameters Δtmin and Δtmax exert analyst’s control on the minimum
and maximum values to be considered by ABAQUS during automatic time step
adjustments. An analyst can choose fixed time step calculations by optionally defin-
ing a parameter, “DIRECT=value”, in the keyword line *DYNAMIC. The duration
of time-domain solution is specified by the input value tf.
External loads on finite element nodes or surfaces, described by *CLOAD or
*DLOAD keywords, can vary in time as defined by a user-defined time function. The
following example describes time-dependent concentrated load applications at
selected nodes of a finite element model:
*STEP
*DYNAMIC, ALPHA=α, BETA=β, GAMMA=γ
Δt0, tf, Δtmin, Δtmax
*CLOAD, AMPLITUDE=Loading-TH
node1, 3, c1
node2, 3, c2
node3, 3, c3
*END STEP
where, “Loading-TH” refers to a time function of load amplitude that is defined in
ABAQUS input data file by using *AMPLITUDE key word:
*AMPLITUDE, name=Loading-TH
0.0, 0.0, 5.0, 3.2, 10.0,0.0, 100.0, 0.0
3.2 kN
Time (ms)
0 5 10 100
FIGURE 11.14 Time history of an arbitrary load function.
284 Finite Element Analysis of Solids and Structures
In the above description, time history of load amplitude is described by time and
amplitude values, in pairs, with as many pairs as needed to describe the complete
time history of applied load. Figure 11.14 graphically describes the load function
described in the above example of input data block. The generic load function, thus
defined, is multiplied by a factor “c1”, and applied in coordinate direction “3” at node
“node1” (per the description of input data presented in the above example). Similarly
scaled load functions are also applied at nodes node2, node3, etc. Time history of
calculated finite element model responses can be saved in “odb” database file for
graphical post-processing of results, at specified time intervals by assigning a numer-
ical value to parameter “FREQUECNY” in *OUTPUT control command:
*STEP
*DYNAMIC, ALPHA=α, BETA=β, GAMMA=γ
Δt0, tf, Δtmin , Δtmax
*CLOAD, AMPLITUDE=Loading_TH
node1, 3, c1
node2, 3, c2
node3, 3, c3
*OUTPUT, FIELD, FREQUENCY=1
*NODE OUTPUT
U
*END STEP
In the discussions so far, time-domain dynamic analysis option has been initiated by
using the *DYNAMIC command. This analysis option can be repeatedly used to
generate target structural responses for many different load functions when needed.
In some practical applications, engineers often seek to determine the sensitivity of a
structural response parameter to the frequency of applied load (where the geometric
distribution and amplitude of loading do not change). Figure 11.15 presents a con-
ceptual description of the frequency response function for floor vibration of a hypo-
thetical automotive body structure subjected to external loads of variable input
Peak amplitude of response: R
Frequency response function
Peak response : R
F1 F2
Input cyclic load of frequency:
F(t)=Fi. sin (2 t)
Input load frequency:
FIGURE 11.15 Example of a frequency response function.
Linear Dynamic Response Analysis of Structures 285
frequency. Such frequency response functions can be efficiently generated in
ABAQUS by using the dedicated analysis procedure, *STEADY STATE
DYNAMICS. The analysis step for response function generation must be preceded
by a frequency analysis step to generate mode shape vectors. The following example
lists key ABAQUS command steps for steady-state dynamic analysis procedure:
*STEP
*FREQUENCY
, , ff
*END STEP
*STEP
*STEADY STATE DYNAMICS, Interval=Range, Frequency scale=linear
f0, ff, nsteps
*CLOAD
set1, 3, c1
set2, 3, c2
etc.
*OUTPUT, HISTORY
*NODE OUTPUT, NSET=set1
UT, VT, AT
*END STEP
Modal frequency and shape data, generated in the first analysis step up to the fre-
quency limit of ff, are used in the next steady-state dynamic analysis step for generat-
ing frequency response function by using the mode superposition technique.
Frequency-dependent displacement (UT), velocity (VT), and acceleration (AT)
responses, for the node set “set1” in the above example, are saved in the database file
for subsequent post-processing.
For explicit dynamic analysis, ABAQUS uses the central difference time integra-
tion method (Equations 11.29–11.31). Analysis step for explicit dynamic analysis is
specified in ABAQUS input data file by inserting the required parameter “EXPLICIT”
with *DYNAMIC keyword:
*STEP
*DYNAMIC, EXPLICIT, DIRECT USER CONTROL
Δt0, tf
*END STEP
The optional parameter “DIRECT USER CONTROL” specifies that the analysis is
conducted at a fixed user-defined time step of Δt0; and the time span of dynamic
response analysis is defined by tf. Automatic time step definition, based on element
wave characteristics (Equations 11.32–11.38) can be activated by choosing the optional
parameters “ELEMENT BY ELEMENT” in lieu of “DIRECT USER CONTROL”:
*STEP
*DYNAMIC, EXPLICIT, ELEMENT BY ELEMENT
tf, Δtmax
*END STEP
286 Finite Element Analysis of Solids and Structures
The input parameter Δtmax limits the maximum time step that ABAQUS can use from
element-by-element time step size calculations. Explicit dynamic analysis can be
conducted without direct initial load applications to nodes or elements. Instead, time-
domain system response can be calculated starting with an initial velocity state, or
with prescribed motion applied to a set of nodes. Following example shows how to
define initial velocity to a set of nodes:
*INITIAL CONDITIONS, TYPE=VELOCITY
seti, DOFi,vi
…
where initial velocity vi is applied to node set seti at degree of freedom DOFi. A user
can include additional lines of similar data to define initial velocity values for mul-
tiple node sets.
11.9 PRACTICE PROBLEMS: DYNAMIC RESPONSE ANALYSIS
OF STRUCTURES
PROBLEM 1
Static load–deflection analysis problem of Figure 6.28 is re-introduced in Figure
11.16. Analyze the problem, using implicit time-domain analysis technique, assum-
ing that the externally applied load amplitude follows the time function described in
Figure 11.14. Calculate the dynamic amplification factor for downward deflection at
monitoring point A (shown in Figure 11.16).
PROBLEM 2
8mm thick rigidly attached bracket –upper surface
(40x8 area) is subject to a time-dependent load
described by function in Figure 11.14
60
Monitoring point, A 100 uniform wall
thickness: 4 mm
600
z
y x
600
Assume fixed/fixed
at both ends
FIGURE 11.16 Load-deflection analysis for a time-dependent load.
Linear Dynamic Response Analysis of Structures 287
0.4 mm
Time (ms)
0 5 100
FIGURE 11.17 Prescribed downward displacement at the monitoring point A in Figure 11.16
ssuming that the loading in Problem-1 follows sine harmonic function, generate
A
frequency response function for peak downward deflection at monitoring point A
(shown in Figure 11.16) in the frequency range of 0.1 to 1 cycle/ms.
PROBLEM 3
Re-analyze the problem, described in Problem-1, for prescribed downward displace-
ment function defined in Figure 11.17 (in lieu of the external loading function applied
in Problem-1). Compare the relative computational efficiency of implicit vs explicit
time integration techniques for this problem.
12 Nonlinear Analysis
of Structures
SUMMARY
Finite element analysis methods, presented in Chapters 1–11 of this book, have been
built on four key assumptions: (a) material stress–strain relationship is linear elastic;
(b) strain and displacement responses are very small, thus, keeping the first-order
derivative relationship between strain and displacement constant throughout the
response history; (c) the boundary conditions and inter-body contact conditions do
not change over the course of load–displacement history; and (d) applied load vector
is independent of the structural displacement response. Assumptions (a), (b) and (c)
make the structural property matrices, [k] and [m], constant. The stress-based assess-
ment of structural strength and durability, thus, generally falls in the domain of linear
elastic finite element analysis. The vibration frequency analysis problem, discussed
in Chapter 10, is a linear eigen solution problem based on the constant structural
property matrices [k] and [m]. The time-domain load–displacement analysis tech-
nique, discussed in Chapter 11, is also a linear elastic finite element analysis method,
based on constancy assumptions for system property matrices, boundary conditions,
and geometric distribution of time-dependent loading. The buckling load analysis of
structures, frequently discussed in finite element literature, is another class of eigen-
value problem where the load capacity is estimated for a small perturbation to the
elastic stability of structure. This analysis technique is available in general-purpose
finite element software packages. However, this topic is omitted from this reference
book since the author has not seen an opportunity to use that technique in the past 25
years of professional practice in civil and automotive structural design. While many
of the structural engineering problems get analyzed and solved by using linear elastic
finite element simulation methods, practicing engineers do encounter frequent prob-
lems where one or more of the linearity assumptions get violated. And in certain
engineering problems, for example, in automotive structural design for crashworthi-
ness, nonlinear finite element simulation method is used as the daily analysis tool
soon after the development of preliminary design based on empirical design rules
and prior engineering experiences.
The volume of current knowledge on nonlinear finite element methods is way
beyond the scope of one entire book. This chapter here should be considered as par-
tial introductions to some of the key ideas and techniques that are used in the analysis
of practical engineering problems. Section 12.1 starts with brief references to the key
sources of nonlinearity that appear in structural analysis problems. The general idea
of deriving iterative solution for nonlinear load–displacement problem is introduced
in this Section. Section 12.2 summarizes the basic approach of how to include mate-
rial nonlinear behavior in finite element software implementation. Details of material
289
290 Finite Element Analysis of Solids and Structures
plasticity formulations are presented for demonstration of the key intricacies that get
implemented in constitutive models of finite element software packages. Section
12.3 is devoted to the formulation of geometric nonlinear problems – involving
higher order derivative relationships between strain and displacement responses of
solids. The basic ideas of contact formulations have been discussed in Chapter 8.
Section 12.4 in this chapter provides a brief description of how contact status changes
can be included in step-by-step simulation of nonlinear structural response analysis.
Section 12.5 describes the integration of nonlinear response mechanisms in step-by-
step implicit and explicit dynamic response analysis of structures. Nonlinear response
analysis, with deformation-dependent external loading description, is not explicitly
discussed in this book. However, it is understood that such changes can be integrated
into general nonlinear analysis steps when required. Section 12.6 is devoted to the
discussion of material fracture propagation in finite element simulation model – that
has experienced huge amount of research contributions over the past few decades.
Structural form simulation, a very specialized nonlinear structural engineering prob-
lem, is discussed in Section 12.7 followed by Section 12.8 presenting a set of practice
problems for nonlinear analysis.
12.1 SIMULATION OF NONLINEAR FORCE-DEFORMATION
RESPONSE OF STRUCTURES
Linear elastic finite element analysis technique, described thus far in Chapters 1–11,
rests on two critical assumptions: (1) material undergoes infinitesimal small defor-
mation resulting into a linear elastic stress–strain relationship matrix [C]; and (2)
deformation response does not effectively change the geometry of structure, thus
keeping the strain–displacement relationship matrix, [B], constant during the analy-
sis step. The constancy assumptions for element property matrices, [C] and [B], lead
to the calculation of a system stiffness matrix [k] (equation 2.44) that is independent
of the deformation response. Linear elastic solution of the static equilibrium state,
described by equation (2.43), uses two additional assumptions: (3) boundary condi-
tions embedded into the displacement vector {u} do not change, and (4) load vector
{P} is independent of the structural deformation response. These four assumptions,
together, imply that a scalar multiple of the load vector {P} will generate same scalar
multiple of the displacement response {u}. Equilibrium equations (2.43), for linear
elastic systems, are solved with one single step of load application – producing the
displacement response {u}, the elastic strain response {ε} = [B]{u}, and the stress
response {σ} = [C]{ε}. Assumptions of material and geometric linearity, as well as
the constancy of load distribution and boundary conditions, serve the purpose of
many engineering analysis problems. However, one or more deviations from the lin-
earity assumptions may appear in many structural design situations when the analyst
is required to consider nonlinear analysis techniques for safety assessment of
structures.
Nonlinear response of elementary structural members, for standard load and
boundary conditions, can be occasionally predicted with analytical techniques.
Figure 12.1 shows the ultimate load capacity calculation of a simply supported beam
Nonlinear Analysis of Structures 291
L
Mp= Fc*h h Mp
Stress
(GPa)
E
Fc=Sy*Ac 4
=
Strain (mm/mm)
Ft=Sy*At
(a) (b) (c)
Ideal elastic-plastic stress-strain Plastic moment capacity Ultimate load capacity of
response of ductile material of beam section a simply-supported beam
FIGURE 12.1 Ultimate strength analysis of a beam based on the concept of plastic moment
capacity of beam section.
assuming elastic-plastic material behavior. However, the progressive evolution of
nonlinear response mechanisms in complex structures is often very difficult to visu-
alize for analytical formulation. Finite element analysis, with step-by-step load
applications, provides an effective method to predict the nonlinear response mecha-
nism of structures. Analysis can start with usual linear elastic model, with a small
initial load application (F0 in Figure 12.2), providing preliminary estimate for dis-
placement response ui. This displacement response can be used to predict local strain
and stress responses using the preliminary assumption of linear elastic material
response. However, when locally calculated stresses exceed the elasticity limit of
material response, a nonlinear stress–strain relationship model can be invoked to
define the local resistance values. The integrated overall structural resistance, corre-
sponding to the first step of calculated displacement response (point A in Figure
12.2), may show an imbalance with the externally applied force F0. Newton-Raphson
iteration method provides a solution to this problem with the prediction of an incre-
mental displacement response for the unbalanced force value (equation 12.1):
Nonlinear force-deforma on response
F0
Internal resistance R(u)
B ≈ external force F
A
ui u0
FIGURE 12.2 Nonlinear force–deformation response of a system.
292 Finite Element Analysis of Solids and Structures
K i . u
i 1
F0 R i (12.1)
where Ki is the tangential stiffness matrix after iteration i, Ri is the internal resistance
of structure corresponding to displacement response ui, and (Δu)i+1 is the correction
to displacement response for the unbalanced force:
ui 1 ui u
i 1
(12.2)
The internal resistance of structure can be updated for the new displacement
response ui+1, and equation (12.1) can be solved repeatedly until the unbalance force
on the right-hand side becomes negligible. In stable structural configurations, the use
of tangential stiffness matrix Ki in iterative calculations, demonstrated graphically in
Figure 12.2, generally leads to a quick convergence to the target solution point
(F0,u0). The above procedure of iterative solution to a discrete load step can be
repeated for a second load step, and so on. For computational efficiency reasons, in
large structural system analyses, the stiffness matrix can be updated only once in the
first iteration of a load step, and the subsequent iterations in the same load step can
be conducted by re-using the same stiffness matrix – a method generally referred to
as modified Newton–Raphson method. In the above discussion, nonlinear material
behavior has been cited as the reason for stiffness matrix change. Large geometric
configuration change may also lead to the change of stiffness properties of a struc-
ture. Same can happen if the boundary conditions change due to part-to-part con-
tacts. And the fourth source on nonlinearity, as noted earlier, can appear when the
external loading condition changes due to changes in structural deformation response.
The step-by-step linearization of nonlinear structural response, presented by
equation (12.1), is a mere extension of the standard solution technique used in linear
elastic finite element analysis models (equation 2.43). The success of this analysis
technique depends on the positive definite property of tangent stiffness matrix Ki in
equation (12.1) which in indicated in Figure 12.2 by monotonically increasing struc-
tural resistance against deformation response. However, failure zone localization in
some structures, at or after the peak response point, may lead to post-peak softening
response (Figure 12.3). The tangent stiffness property of structure is no-longer posi-
tive definite at the post-peak state – meaning that standard technique of tangential
stiffness matrix decomposition and inversion cannot be used to predict the nonlinear
load–deformation response history beyond the peak resistance point (B). A solution
to this problem can be obtained by controlling the displacement response at one or
more control points – solution technique commonly known as displacement control
method. Structural DOF with pre-defined non-zero displacement responses are sepa-
rated from other DOF in the system equilibrium setup:
i i
Ka K as ua fa (12.3)
.
K sa Ks us rs
where suffix “s” refers to the DOF under specified displacement control, and suffix
“a” to all other non-zero DOF. Pre-defined displacement increment at control DOF is
Nonlinear Analysis of Structures 293
Post-peak so ening response
B
F0
A Unloading
path
u
B’
FIGURE 12.3 Post-peak softening response of a structure.
identified by Δus and the corresponding unknown reaction forces by Δrs. The term Δfa
refers to zero or nonzero externally applied forces at un-constrained DOF in the sys-
tem. Partitioning of the system equilibrium equations allows the calculation of
unknown displacement responses by solving the following reduced system of equa-
tions (12.4):
K a . ua fa K as . us
i i i
(12.4)
Solution of equation (12.4) provides the first approximation to unknown deformation
responses Δua that can be used to find out the unbalance between external and inter-
nal forces, and the iterative improvement to predicted solution can proceed by using
the Newton-Raphson method (or by using its modified form). The displacement con-
trol method, presented in the above, helps to predict nonlinear structural responses,
beyond the peak resistance point, provided that the controlled displacement response
increases monotonically. However, special situations exist where the load–displace-
ment response may not progress monotonically. Figure 12.4 shows a schematic shear
test setup for notched concrete beam that happens to show snap-back in load–deflec-
tion response (Arrea and Ingraffea 1981). Neither the step-by-step load control nor
the displacement control analysis techniques can numerically simulate such struc-
tural responses. Successful numerical techniques have been implemented in aca-
demic research to predict the snap-back behavior by using specially derived
monotonic response function, such as the crack tip displacement, to adjust the exter-
nally applied loads or displacements during incremental analyses (Bhattacharjee
1993).
Not all four nonlinear response mechanisms, listed earlier in this section, occur
simultaneously in a structural analysis problem. Nonlinear material and geometric
responses tend to be the commonly encountered problems followed by contact-
related nonlinearity. Section 12.2 in the following summarizes the basic approaches
of how material nonlinear response is handled in finite element formulations.
294 Finite Element Analysis of Solids and Structures
F, u
snap-back
u
(a) (b)
FIGURE 12.4 (a) Single notch shear test of a beam made of quasi-brittle concrete; (b) snap-
back in force-displacement response.
12.2 NONLINEAR MATERIAL MODELS FOR FINITE ELEMENT
SIMULATION OF STRUCTURES
Material stress–strain relationship, alternatively known as constitutive model, defines
the material elasticity matrix [C], that is used to calculate the stiffness property of
structures, and to calculate the stress responses from strains. In linear elastic finite
element simulations, both displacement and strain values are assumed to be very
small, thus retaining the initial reference coordinate system to define finite element
strain and stress responses (equations 2.5 and 2.25). Hooke’s law, using the principle
of superposition of stress–strain responses, relates the total strain responses to total
stress responses through equation (2.18). This linear relationship, between stresses
and strains, often changes as the deformation increases (Figure 12.5). Similar to the
piece-wise linearization of load–deformation response of structures, material stress–
strain law in nonlinear finite element analysis can also be expressed in the following
incremental form:
C .
i i i
(12.5)
where [C]i is the instantaneous elasticity matrix at a given state of deformation in the
material. In step-by-step nonlinear load–deformation analysis of structures, instanta-
neous elasticity matrix [C]i can be used, in lieu of the initial elasticity matrix [C], to
define the tangential stiffness matrix [K]i in equation (12.1). The total stress response,
at the newly deformed state of material, is calculated by adding the incremental stress
values to those of the previous equilibrium state:
i i 1 i
(12.6)
These updated stress values are used to calculate the internal resistance of structure
{R}i in equation (12.1), and the standard iterative correction for unbalanced forces
Nonlinear Analysis of Structures 295
Nonlinear elas c Elas c-bri le Elas c-so ening
(a) (b) (c)
Linear hardening plas city
Elastic-plastic (ideal) S ffening response
(d) (e) (f)
FIGURE 12.5 Nonlinear stress–strain responses of solids.
can continue as usual. Finite element formulations for multi-axial nonlinear stress
response, however, require special techniques to define the tangential elasticity
matrix [C]i.
One of the commonly used material models is the plasticity-based formulation for
simulation of ductile material behavior. Figure 12.6 schematically shows a nonlinear
stress–strain response curve, where material resistance, after initial yield point (Sy),
increases with increasing permanent deformation (plastic strain) in the material –
commonly known as hardening plastic behavior of ductile metals. The incremental
strain response Δεi in equation (12.5) comprises of two parts – an elastic (recover-
able) part, Δεe, and a permanent deformation (plastic) part, Δεp (equation 12.7):
Johnson-Cook plas city model:
Stress: = ( + )
i+1 =F/A
i
Sy e p) A
= L/L
p
Total strain:
FIGURE 12.6 Hardening plasticity response of a ductile metal specimen.
296 Finite Element Analysis of Solids and Structures
e p
i i i
(12.7)
For pure uniaxial stress–strain case, relationship between incremental stress and
strain is defined as follows:
E i . E. p
i i i
(12.8)
where “Ei “ refers to tangent elastic modulus of material, and “E” the Young’s modu-
lus that relates incremental stress to the elastic component of incremental strain. The
relationship between incremental stress and plastic strain is defined by using a mate-
rial parameter “H”, commonly known as strain hardening parameter:
p /H i
i i
(12.9)
Combining equations (12.8) and (12.9), the following definition is obtained for the
tangent modulus Ei:
E. H i E
Ei E 1 i (12.10)
E H i
E H
With known hardening parameter value Hi from nonlinear uniaxial stress–strain test
data, equation (12.10) gives the tangent modulus, Ei, which is used to define incre-
mental stress–strain relationship and tangential stiffness properties of one-dimen-
sional finite elements (truss, cable, etc.). The value of Hi, and thereby, the incremental
stress–strain relationship and stiffness property matrices can be updated during the
iterative calculations of nonlinear structural responses. Evidently, an ideal elastic-
plastic material law (Figure 12.5(d)) can be generated by setting Hi≈0; and a linear
hardening model (Figure 12.5(e)) can be generated by using a constant value of Hi.
Plastic hardening may occur only in the direction of straining without affecting other
directions – known as kinematic hardening of material. Alternatively, hardening can
be assumed to occur iso-tropically. Depending on the prior understanding of specific
material behavior, a user of general-purpose finite element software packages can
choose either of the hardening options, kinematic or isotropic, with material plastic-
ity models.
The accumulated plastic strain over a loading duration indicates the degree of
permanent deformation in the material – which is also used to assess the material
integrity at finite element calculation points:
p p p
i i 1 i
(12.11)
In one-dimensional finite element models (truss, cable, etc.), strains and stresses are
calculated in local element direction, thus explicitly defining the direction of plastic
strain in the same axis direction. The calculation of tangent elasticity matrix, [C]i in
Nonlinear Analysis of Structures 297
equation (12.5), is achieved with straight forward substitution of initial Young’s
modulus, E, with the tangent modulus Ei calculated from equation (12.10). In multi-
directional stress fields, with no apparent member axis direction, special analytical
assumptions are needed to define a comprehensive plasticity model: (a) an initial
yield function (σ, ε); (b) a plastic flow rule defining the direction and magnitude of
plastic strain; (c) material hardening rule; and (d) the tangent (incremental) elasticity
matrix [C]i. The von Mises stress function, defined in equation (9.19), is one possible
option to define a yield function in multi-axial stress field:
1
Y 1 2 2 2 3 2 3 1 2 S y 0 : elastic (12.12)
2
where σ1, σ2 and σ3 are principal stresses calculated at a finite element integration
point, and Sy is the yield strength of material determined from uniaxial material test
specimen. For hardening plasticity, a general form of yield function can be written as
Y , 0 (12.13)
where κ is a material hardening parameter. For plastic deformation occurring on the
yield surface, derivative of yield function gives:
Y Y Y (12.14)
. 1 . 2 . . 0
1 2
Using vector notation, equation (12.14) can be rewritten as:
T
Y Y
. . 0 (12.15)
When a stress state falls outside the yield function, plastic deformation is expected in
the material. In general constitutive models, a dedicated plastic potential function, Q
is defined as a function of stresses (σ) and hardening parameter (κ):
Q , 0 (12.16)
The incremental plastic strain vector {Δεp} is defined as (after dropping the itera-
tion index “i” for simplicity):
p . Q (12.17)
where Δλ is proportionality constant (to be determined). Equation (12.17) defines the
plastic strain increments in a multi-directional stress field. A constitutive model using
298 Finite Element Analysis of Solids and Structures
Stress: Plastic potential function: Q( )
Strain:
Plastic strain
Elastic components
behavior
Stress:
Strain:
FIGURE 12.7 Plastic strain potential function in a 2D strain field.
same analytical functions for Y and Q is known as associated plasticity model – com-
monly used for ductile metals. When Y and Q use different analytical functions, the
constitutive model is known as non-associated plasticity model – generally used for
soil and granular materials. For graphical simplicity, an example plastic potential
function is shown in Figure 12.7 for a two-dimensional stress field. In hardening
plasticity models, the plastic strain increment, defined by equation (12.17), grows in
normal direction to the original surface with components occurring in the original
directions of σ1 and σ2. A three-dimensional plastic potential function will include a
third dimension to this two-dimensional model.
The total strain increment in equation (12.7) can now be rewritten as
Q
C . .
1
(12.18)
where [C] is the standard elasticity matrix of material. Defining the work hardening
parameter κ (in Equations 12.13 and 12.16) as the amount of plastic work done dur-
ing plastic deformation, incremental plastic work is given by
. p
T
(12.19)
Substituting equation (12.17) into equation (12.19) gives:
T Q
. . (12.20)
Substituting equation (12.20) into equation (12.15) gives:
T
Y Y T Q
. . . . 0 (12.21)
Nonlinear Analysis of Structures 299
Equations (12.18) and (12.21) can now be combined to write the following
expression:
1 Q
C
. (12.22)
T Q
T
0 Y Y
. .
Inverting the relationship (12.22), the following relationship is obtained between
incremental stresses and strains (Zienkiewicz and Taylor 1991):
Q Y
T
.
. C
. (12.23)
C C . T
Y Q Y T Q
. C . . .
The incremental stress–strain relationship matrix, defined by the terms inside the
parenthesis on the right-hand side of equation (12.23), can be used to calculate tan-
gent stiffness matrix and incremental stress responses in finite element calculations.
Stress–strain curve generated from one-dimensional material test specimen is used to
define the basic material parameters embedded in the incremental formulation. The
post-yield stress–strain data can be described with piece-wise linearized steps, or can
also be described with an analytical function. Johnson–Cook model (Johnson and
Cook 1983), a commonly used isotropic plasticity formulation, describes the post-
yield hardening response with an analytical expression (Figure 12.6):
S y b pn (12.24)
In ABAQUS model input files, the Johnson–Cook plasticity model can be chosen
by a user by using the following data inputs:
*MATERIAL, NAME=mat1
*ELASTIC
E, ν
*PLASTIC, HARDENING=JOHNSON COOK
Sy, b, n
where E and ν are elastic material properties (Young’s modulus and Poisson’s ratio,
respectively); and Sy, b and n are Johnson–Cook material parameters that are gener-
ally calculated by fitting equation (12.24) to a uniaxial test data set. A step-by-step
nonlinear static analysis of the two-member truss assembly problem of Figure 2.18
can be constructed for a load amplitude of 960 kN by using the following input com-
mands in an ABAQUS input file:
300 Finite Element Analysis of Solids and Structures
*STEP
*STATIC
1,10,0.1,2.0
*CLOAD
2, 2,-960.0
*END STEP
The option for linear elastic small displacement – small strain analysis, keyword
“PERTURBATION”, has been excluded in the above analysis step definition. In this
example, a concentrated load of 960 kN is applied at node ID # 2 in negative direc-
tion of coordinate ID #2 (y-direction). The step parameters, described after the key-
word *STATIC, specifies that the first analysis step will apply 1/10th of the load; the
total load will be applied in 10 steps; and during automatic load adjustments,
ABAQUS is allowed to use a minimum step of 0.1 times the initial load step, but no
more than 2 times of that initial step. Figure 12.8 shows predicted load–deflection
responses of the system for both linear elastic material assumption and for Johnson–
Cook material plasticity model (Sy = 0.43 GPa, b = 0.824, and n = 0.51). The consid-
eration of material hardening plasticity has led to a hardening structural resistance
against increased deformation of the structure. In this simple system model of two
truss members, material nonlinearity has caused a very large downward displace-
ment response at the load application point – making the geometric configuration of
structure very different from the initial state (Figure 12.9). Large geometry change
requires consideration of geometric nonlinearity in finite element simulations – the
subject of discussion in next Section 12.3. It should be noted, however, that large
geometry change can also occur without causing nonlinear material response.
As noted earlier, isotropic plasticity model is relevant for ductile metal behavior
only. An analyst will need to choose an appropriate constitutive model depending on
the prior knowledge about nonlinear response mechanism of the specific material
being analyzed. Numerous material constitutive models have been proposed in the
literature to simulate nonlinear response mechanisms of different material types. The
Step-by-step load application
Linear elastic
analysis response
Downward load, P
960 kN
837
time
FIGURE 12.8 Elastic versus nonlinear load–deformation responses of two-member truss
assembly.
Nonlinear Analysis of Structures 301
12000 mm
60000 mm 60000 mm
P, u
12000 mm
> 6000 mm
Substantial changes to geometry make the initial strain-
displacement relationship [B] invalid
FIGURE 12.9 Nonlinear geometric response of a two-member truss assembly.
list of available options has become so large over the decades that a comprehensive
review of potential choices is beyond the scope of a limited size book. Descriptions
of few selected constitutive models for softening, creep and visco-elastic type mate-
rial behavior are available in Bathe (1996), Lemaitre and Chaboche (1994), and
Zienkiewicz and Taylor (1991). General-purpose finite element simulation software
packages include many commonly used material constitutive models. Based on prior
knowledge, an analyst can choose a constitutive model appropriate for the material
case study with known values of relevant model parameters.
12.3 SIMULATION OF LARGE DEFORMATION RESPONSE –
NONLINEAR GEOMETRIC PROBLEMS
Large geometric changes, with or without material nonlinear response, means that
the strain–displacement relationship matrix [B] (equation 2.6) does not remain con-
stant. Using the initial reference coordinate system, the Green-Lagrange strain com-
ponents for a nonlinear large deformation problem can be defined by extending the
definitions from equations (2.1) and (2.2) into the following forms (Fung 1965):
u 1 u v w
2 2 2
x
x 2 x x x
(12.25)
u v u u v v w w
xy . . .
y x x y x y x y
302 Finite Element Analysis of Solids and Structures
Other components of strain (εy, εz, γyz, γzx) can also be defined in similar formats of
equations (12.25). Evidently, the strain components can be separated into two terms:
0 L (12.26)
where ε0 refers to the linear first-order terms in equations (12.25), and εL to the sec-
ond-order nonlinear terms. Using the standard nomenclature for strain–deformation
relationships, described in earlier chapters, equation (12.26) can be re-written as:
B 0 B L u B u (12.27)
where, nonlinear strain–deformation relationship matrix comprises of two parts:
B B 0 B L (12.28)
[B]0 refers to the first-order linear components of strains, defined by Equations (2.5)
and (2.6), and [B]L is a new term referring to the higher order terms in equations
(12.25). The quadratic nonlinear strain terms can be conveniently expressed as
follows:
xT 0 0
0 yT 0
x
10 0 zT 1
L y A (12.29)
20 zT yT 2
T z
z 0 xT
yT xT 0
where, matrix [A] is of 6 × 9 size; and the terms in {θ} are defined as follows:
u v w u v w u v w (12.30)
xT ; y y
T
; zT
x x x y y z z z
The vector {θ} in equation (12.29), thus, can be defined in terms of nodal dis-
placements and finite element shape functions (defined in earlier Chapters 2–7):
G u (12.31)
The differential form of equation (12.31) leads to the following relationship:
d G du (12.32)
Nonlinear Analysis of Structures 303
Re-writing equation (12.29) in differential form leads to the following relation-
ship (Zienkiewicz and Taylor 1991):
1 1
d L 2 dA 2 A d A d (12.33)
Combining equations (12.32) and (12.33):
d L B L du (12.34)
where [B]L is a deformation-dependent relationship matrix for higher order strain
components:
B L A G (12.35)
The complete nonlinear strain–deformation relationship (equation 12.28), is thus,
known from the first-order relationship matrix [B]0 (equation 2.5) and the higher
order relationship, [B]L defined by equation (12.35).
The step-by-step solution of a nonlinear response, described by equation (12.1),
can be re-written as a minimization of difference between external force vector {F}i
and the internal resistance vector {R}i as follows:
F B . .dV 0
T
(12.36)
where [B] refers to the strain–displacement relationship matrix at the new deformed
state of a structure; and {F} to the external load vector at that instance. The varia-
tional form of equation (12.36) can be written as:
d dB . .dV B . d .dV K . du
T T i i
(12.37)
where [K]i is the tangential stiffness matrix of the system and {du}i is the incremental
displacement response for the unbalance between external force and internal resis-
tance. The tangent stiffness matrix, [K]i, is defined by two terms in equation (12.37):
(i) the first term arising from the resistance offered by existing stresses to the geom-
etry change [dB]; and (ii) the second term arising from the resistance offered by the
current geometry to stress change {dσ}. The change in stress response of a finite
element can be defined as follows:
d C . d C
i i i
B du C B 0 B L du (12.38)
where [C]i is the incremental stress–strain relationship matrix discussed in Section
(12.2). In the case of linear elastic material behavior, [C]i will simply be equal to the
initial elasticity matrix. Using the expression for stress increment from equation
304 Finite Element Analysis of Solids and Structures
(12.38), the associated stiffness contribution in equation (12.37) can be defined as
follows:
B . d .dV
T
B
T
0
T
B L . C . d .dV
i
(12.39)
Re-writing equation (12.27) in differential form, and using the incremental non-
linear strain definition from equation (12.34):
d B du B 0 B L du (12.40)
Substituting expression (12.40) into equation (12.39), finite element resistance
against stress change is defined as follows:
B . C . B .dV B . C . B .dV
T
0
i
0
T
0
i
L
B . C . B .dV B . C . B .dV .du
T i T i (12.41)
L L L 0
K 0 K L . du
where,
K 0
B
T
0
i
. C . B 0 .dV
K
L B
T
0
i
. C . B L .dV
B
T
L
i
. C . B L .dV B
T
L
i
. C . B 0 .dV
The first term in equation (12.41), [K]0, refers to the stiffness matrix for small
deformation–small strain system response, and the second term , [K]L, to the nonlin-
ear large deformation response of the system. The complete definition of tangential
stiffness matrix [K]i in equation (12.37) requires an evaluation of the first resistance
term associated with stress resistance to the geometry change:
dB . .dV K . du
T
(12.42)
Re-writing equations (12.28) and (12.35) in derivative forms, and combining the
expressions:
dB dB L dA G (12.43)
Substituting expression (12.43) into equation (12.42):
K . du G . dA . .dV
T T
(12.44)
Nonlinear Analysis of Structures 305
The product term [dA]T.{σ} inside equation (12.44) can be expressed as
(Zienkiewicz and Taylor 1991):
x .I 3 xy .I 3 xz .I 3
dA . xy .I 3 yz .I 3 . d M . d
T
y .I 3 (12.45)
xz .I 3 yz .I 3 z .I 3
where I3 is a 3 × 3 identity matrix. Combining the equations (12.32) and (12.45):
dA . M . d M . G . du
T
(12.46)
Finally, combining equations (12.44) and (12.46), following definition is obtained
for the stiffness matrix associated with current stress resistance to geometric changes:
T
K G . M . G .dV (12.47)
The tangential stiffness matrix, [K]i, in equation (12.37) is thus defined by com-
bining all resistance mechanisms:
i
K K K 0 K L (12.48)
where stiffness component [K]σ is defined by equation (12.47), and the other two
terms, [K]0 and [K]L, by equations (12.41).
Green-Lagrange strain components, defined by equations (12.25), as well as all
subsequent derivations presented in the above, refer to the initial reference coordi-
nate system defined in a standard finite element model description. The incremental
nonlinear structural response calculation process, referring to the original coordinate
reference system, is generally known as the total Lagrangian method. The stress
values, calculated by equation (12.38), and referring to the same initial coordinate
reference system, are known as second Piola–Kirchoff stresses. By virtue of defor-
mation gradient calculations based on iso-parametric formulations of finite elements,
discussed in earlier Chapters 2–7, both Green–Lagrange strain and second Piola–
Kirchoff stress values are invariant to rigid body rotations of a structural component.
This important invariant feature makes these strain and stress definitions particularly
suitable in finite element calculations. However, these response measurements, with
respect to the original reference system, are not easily useful in engineering interpre-
tations. This ambiguity related to interpretation of engineering stress–strain calcula-
tions is illustrated with a truss element example in Figure 12.10. During large
deformation response of a structural assembly, element A-B moves to a new deformed
position defined by position A′–B′. Stress values, calculated from Green-Lagrange
strains, referring to the original reference coordinate system, are shown by values σx
and σy. Understandably, these values are not directly meaningful to assess the strength
and integrity of the member. A more meaningful interpretation of the results can be
306 Finite Element Analysis of Solids and Structures
2
1
y
B ’
x
A’
A B
FIGURE 12.10 Coordinate reference systems (x–y vs. 1–2) for strain and stress definitions
during large deformation response.
made if the stress values refer to the post-deformation local axis system (1–2) – gen-
erally referred to as true or Cauchy stresses. This analysis objective can be achieved
by updating the element geometric reference system after each small incremental
deformation response of the system; and then using the latest configuration geometry
in next iteration of calculations. This continuously updated calculation process,
referred to as updated Lagrangian method, leads to true strain and stress responses
when calculations are conducted with very small load steps.
The consideration of geometric nonlinearity can be achieved in an ABAQUS sim-
ulation model by including the nonlinear geometric option (NLGEOM) in analysis
step description:
*STEP, NLGEOM
*STATIC
1,15,0.1,2.0
*CLOAD
2, 2,-1440.0
*END STEP
Figure 12.11 shows load–deformation response of problem described in Figure 12.9
including both material and geometric nonlinear response mechanisms. As expected,
Step-by-step load applicaon
Hardening resistance
1440 kN caused by geometry change
Downward Load, P
time
FIGURE 12.11 Load–deformation response of two-member truss assembly – including
geometric nonlinearity.
Nonlinear Analysis of Structures 307
Step-by-step displacement control
12000 mm “Hypothe cal” linear-elas c response
Upward displacement, u
Geometry change leading to
so ening of structural resistance
in this example case
time
FIGURE 12.12 Post-peak nonlinear load–deformation response prediction with displace-
ment control analysis.
downward deflection of load application point contributes to steeper geometric ori-
entation of the truss members, thus, increasing the structural resistance to downward
deflection of load application point. Nonlinear geometric formulation has reproduced
the hardening behavior of structure in this special example case study. However,
when the loading direction is reversed in this particularly simple structural configura-
tion, the geometric stiffness of structure diminishes with increasing upward deflec-
tion until the geometric assembly becomes nearly flat. As discussed in Section 12.1,
the post-peak softening response prediction requires special incremental step-by-step
analysis, such as by prescribing displacement controls at critical DOF:
*STEP, NLGEOM
*STATIC
1,30,0.1,2.0
*BOUNDARY, TYPE=DISPLACEMENT
2, 2,12000.0
*END STEP
Figure 12.12 shows step-by-step displacement control analysis results of the two-
member truss assembly with a prescribed upward displacement motion at node # 2.
12.4 NONLINEARITY ARISING FROM CHANGES
TO INTER-BODY CONTACTS
Body-to-body interface condition may change, from no-contact to contact or vice-
versa, during the progression of load–deformation analysis of structural assemblies.
Any change in the interface contact state leads to changes in stiffness matrix of the
overall structural assembly, and to the interface stress distributions. Changes of the
interface contact condition in finite element simulation models are simulated by
308 Finite Element Analysis of Solids and Structures
tracking the relative motion between nodes and element surfaces. Section 8.5 has
presented a detail discussion on two alternative numerical techniques, namely
Lagrangian multiplier and penalty methods, for simulation of interface contacts.
Either of these two methods can be incorporated in step-by-step nonlinear finite ele-
ment simulation models. Penalty method tends to be the preferred choice in step-by-
step simulation of nonlinear response of large structural systems because of improved
stability and overall computational efficiency. The success of iterative nonlinear
simulation depends on the proper definition of contact stiffness values. Special for-
mulation techniques are often required for the simulation of contact response between
soft and hard bodies. A low penalty stiffness typically results in better convergence
of the solution, while the higher stiffness keeps the interface penetration at an accept-
able level as the contact pressure builds up. Nonlinear contact stiffness formulations,
based on gap closure and interface sliding state, generally offer improved stability to
simulation models. Finite element simulation software packages often use pre-
defined formulations to set contact stiffness values based on material properties of
contacting bodies, provided contact check algorithm is activated with appropriate
user input instructions. Contact-related nonlinearity may or may not occur in con-
junction with material and geometric nonlinear response mechanisms, However, if
appropriate, all three nonlinear mechanisms can be included in a simulation model.
A finite element analysis model input data must include specifications for which
body parts be checked for potential contact condition changes during the incremental
step-by-step simulation process. Specific input data syntax, for activation of contact
consideration in ABAQUS simulation models, has been described in Section 8.5. It
is important to take care of good quality finite element model preparation with no
initial intersections and minimum interface penetrations (Figure 8.8), particularly for
simulation of large deformation problems. Initial positions of part surfaces must be
placed with adequate offset, taking account of the part thickness values, to minimize
initial interface penetrations. During the simulation of multipart shell models, inter-
face contact calculations using hypothetically scaled down (70–90%) values of part
thickness often provide stable solution results.
12.5 NONLINEAR DYNAMIC RESPONSE ANALYSIS
OF STRUCTURES
Step-by-step simulation technique of nonlinear structural response, as described ear-
lier in this chapter, is naturally adaptable to time-domain dynamic response analysis
techniques – that inherently rely on step-by-step simulation process to take account
of complex load-time histories. The implicit time-domain simulation technique,
described by equation (11.21), requires that the system property matrices [k], [m],
and [c] be updated frequently as necessitated by the nonlinear mechanisms arising in
a system. Similarly, in explicit time-domain solutions, described by equation (11.31),
nonlinear response mechanisms are taken care of during the calculations of local ele-
ment and nodal response variables. In both cases, implicit and explicit, time step size
can be affected by the nonlinear response mechanisms occurring in the system. Time
step selection criteria, such as those of Equations (11.22) and (11.33), are used to
Nonlinear Analysis of Structures 309
initiate the analysis process. However, there is no universally usable analytical crite-
rion to adapt the time step size as the nonlinear mechanisms emerge in a system.
Finite element software packages utilize various internal ad-hoc criteria to adjust the
time step size as nonlinearity emerges in a finite element simulation model.
Calculation of the energy error in overall system response provides an indication of
the quality of results. When the energy error fails to meet a certain threshold limit
within a preset number of iterations, software may scale down the time step size to
meet the energy error criterion. In explicit time-domain calculations, the penalty
stiffness at contact interfaces (defined by equation 8.8), can set the time step size
during step-by-step nonlinear dynamic response calculations. Finite element soft-
ware package may resort to the use of artificial mass scaling technique, discussed in
Section 11.7, to maintain a user-specified minimum time step during nonlinear
response simulation of structures. ABAQUS data input syntax for a nonlinear implicit
dynamic analysis problem is provided in the following:
*STEP, NLGEOM
*DYNAMIC
Δt0, tf, Δtmin, Δtmax
*END STEP
The input values for Δt0, tf, Δtmin, and Δtmax define time step control parameters as
discussed earlier in Section 11.8. The optional parameter “NLGEOM” turns on the
nonlinear geometric analysis method in this analysis step. Simulation of nonlinear
material behavior can be activated by using the appropriate constitutive model in
material input data under the keyword “*MATERIAL”. Constitutive models in simu-
lation software packages generally include options to describe the strain rate sensi-
tivity of material stress–strain response. For example, the rate sensitivity of the
John–Cook hardening plasticity model is generally expressed with a modified form
of equation (12.24) (Figure 12.13) (equation 12.49):
Strain rate-dependent resistance
= + 1+ .
Resistance @ quasi-sta c strain rate ( ):
=( + )
FIGURE 12.13 Strain-rate sensitivity of material stress–strain response.
310 Finite Element Analysis of Solids and Structures
S y b pn 1 r. ln p
0
(12.49)
where “r” is a material parameter defining the strain rate sensitivity of yield strength;
εp the instantaneous strain rate; and ε0 the reference strain rate defining the quasi-static
stress–strain response. This description in equation (12.49) is one of many different
possible formulations for rate-dependent constitutive models. A user of finite element
simulation software packages should consult the software-specific material theory
manual to identify strain rate dependent constitutive models relevant for the material
being analyzed. Users also need to consult the software manuals to identify critical
output data that can be used for assessing the reliability of nonlinear simulation results.
12.6 MATERIAL FAILURE SIMULATION IN
NONLINEAR FINITE ELEMENT ANALYSIS
Structural safety assessment often involves the simulation of material fracture initia-
tion and propagation in a continuum finite element model. This objective is often
achieved by allowing the stress resistance at finite element integration points to
degrade or disappear during step-by-step simulation of nonlinear structural response
(Figure 12.14). Ductile material failures are often simulated when the accumulated
plastic strain at finite element integration points (equation 12.11) exceeds a pre-
defined material failure strain. Sustained resistance during large plastic deformation
of ductile materials helps with the evolution of yielding zone beyond artificial
Simulated fracture
propaga on
FIGURE 12.14 Fracture response simulation with plane stress finite element model of a
concrete gravity dam (Bhattacharjee 1993).
Nonlinear Analysis of Structures 311
inter-element boundaries, thereby, preventing the emergence of mesh-dependent
sudden localization of failure at discrete finite element integration points. Strength-
based failure simulation for brittle and quasi-brittle materials generally leads to
mesh-dependent response predictions with uncertain practical application value.
Extensive research work has been pursued since 1990s to produce mesh-objective
fracture simulation results by making the softening part of quasi-brittle material
stress–strain response (Figure 12.5(c)) a function of the fracture energy dissipation
and finite element size parameter (Bhattacharjee and Leger 1994). Successful numer-
ical simulation case studies have been documented for both static and dynamic load
cases (ABAQUS Example Manual 2020). Alternative numerical techniques to simu-
late discrete crack propagation in solids, using linear and nonlinear fracture mechan-
ics criteria, have also been developed over the past decades where crack propagation
involves continuous updates to the finite element mesh to represent the geometric
profile of discrete crack and crack-tip. While research use of discrete crack simula-
tion has been very active over many decades, the practical use of it in everyday engi-
neering decision process remains very limited.
12.7 COMPUTATIONAL METHODS FOR STRUCTURAL FORM
SIMULATION
General-purpose finite element simulation method starts with the assumption that
geometric design data for parts and structures exist to build the initial configuration
model. However, engineering development process often requires that the forming
process of initial geometry itself be also assessed. For example, geometric shapes of
thin sheet metal parts are often formed by punching flat metal blanks to the desired
shape – a process involving large geometric deformation with significant plastic
straining of the original metal blanks (Altan and Tekkaya 2012). Sheet metal forming
simulation helps with the formability assessment of a desired geometric shape –
starting from flat blanks of cold or hot raw materials (Figure 12.15). Explicit
“BLANK”
“PART”
Stamping
FIGURE 12.15 Stamping of metal blank to a desired part geometry.
312 Finite Element Analysis of Solids and Structures
F, u F
FIGURE 12.16 Progressive folding of a thin-wall metal tube under axial compressive load.
time-domain analysis technique, described in Chapter 11, with geometric and mate-
rial nonlinear finite element formulations, provides a very useful tool for upfront
formability assessment, thus, helping to predict if excessive thinning and fracture
may appear during actual forming of geometric shapes. Computational simulation of
metal stamping process is an essential part of modern automotive product develop-
ment process for reducing the time spent on traditional trial-and-error approach used
in building and adjusting the physical punch and die (Andersson 2004).
Crashworthiness simulation of automotive body structures also falls in the domain of
explicit finite element simulation where thin-wall structures are engineered to achieve
a preferred post-crush geometry (Figure 12.16).
There exists, however, a special class of structural engineering problem where
initial geometric form of structure is unknown. Design of tensioned fabric structures,
generally, starts with a few known geometric points, such as the cable attachment
points in Figures 3.2 and 7.11. Engineering analysis task involves the finding of sta-
ble geometric form that will possess uniform acceptable tension in the deployed
configuration of cable-fabric assembly. As illustrated earlier with Figure 12.9, resis-
tance to out-of-plane forces is produced by the geometric configuration of prestressed
truss members. Similar resistance mechanism is produced by membranes in three-
dimensional space with bi-directional curvature at each point on the surface. The
greatest stiffness of a prestressed fabric surface is achieved in an anticlastic surface,
i.e. a surface with opposing curvatures at any point (producing a negative Gaussian
curvature) (Lewis 2008, 2013). Analysis techniques for structural form finding rely
on iterative discovery of a geometric shape without using the stiffness-based finite
element formulations. The computation starts with an initial fictitious surface con-
figuration, represented by a mesh of simple geometric elements (cable and mem-
branes). The unbalanced force at each node is calculated based on preset design
stress values in the connecting elements. Depending on the magnitude and direction
of unbalanced force, the node positions are incrementally adjusted in successive
iterations. Several iterations are generally required to find a geometry with overall
system equilibrium. Boundary configurations, defined with initial key points and
cable configuration, have a significant impact on the success of iterative form-finding
analysis. Computational form-finding, driving the discovery of detailed geometric
configuration of a surface, for given boundary constraints and target prestress values
Nonlinear Analysis of Structures 313
in the component elements, does not strictly fall within the domain of standard defor-
mation-based finite element method discussed in this book. The form-finding algo-
rithm basically iterates on geometric configuration to reach an equilibrium state
between assumed stress field and external forces and boundary constraints. Although
not explicitly stated in current research publications, the structural form-finding tech-
nique can be considered an iterative solution of the geometric nonlinear problem
with predefined stress functions for discrete finite element domains (representing
membranes and cables). Obviously, the merger of two simulation techniques, dis-
placement versus stress function-based approach, has not occurred in the currently
available commercial software packages.
12.8 PRACTICE PROBLEMS: NONLINEAR RESPONSE ANALYSIS
OF STRUCTURES
PROBLEM 1
Re-analyze the steel bumper beam problem #4 in Chapter 7 (Figure 7.16), assuming
nonlinear material behavior with Johnson-Cook hardening plasticity model defined by
S y 0.43 0.824 p0.51 GPa
(12.50)
onduct a step-by-step analysis by applying the pressure given in Figure 7.16. Plot
C
the contour of maximum plastic strain in the beam structure.
PROBLEM 2
Figure 12.17 shows a nonlinear spring-mass system impacting on a rigid wall.
Conduct hand calculations to predict the acceleration response of the mass by using
both implicit and explicit time integration methods; and compare the results.
200
Spring resistance, kN mm
m=
21.54 kg
k0=0.34kN/mm
v0=11.11
k0/2 mm/ms
0 40 80 Rigid wall
Spring deforma on, u (mm)
FIGURE 12.17 Impact of a nonlinear spring-mass system on the rigid wall.
314 Finite Element Analysis of Solids and Structures
Square section: 180x180 mm
Length: 800 mm
Wall thickness:2.1 mm
Material: elastic-plastic
(E=205GPa, =0.3, =7.8e-6 kg/mm3, SyLorem ipsumGPa) Rigid wall
Mass: 500 kg
rigidly attached to the end
of box section member
Initial impact velocity:
11.11 mm/ms
FIGURE 12.18 Simulation of the axial impact of a thin-wall box-section member on the
rigid wall.
PROBLEM 3
Figure 12.18 shows a thin-wall box-section member carrying a rigidly attached mass
of 500 kg at one end, and it is impacting on a rigid wall at the other end with an initial
velocity of 11.11 mm/ms. Average crush strength of a box section beam is approxi-
mately defined by the following empirical equation (Mahmood and Paluzsny 1981):
Pavg 19.796 t . b . Sy
1.86 0.14 0.57
(12.51)
where “b” is the side dimension of square section in mm, “t” the wall thickness in
mm, and Sy the material yield strength in GPa. Conduct a nonlinear finite element
simulation of the impact problem; assess the average dynamic crash strength from
simulation results, and compare that with the quasi-static strength predicted by the
empirical equation (12.51). Assume elastic-plastic material behavior.
PROBLEM 4
Re-analyze Problem-3 by using the Johnson–Cook hardening plasticity model
defined by equation (12.50); and compare the simulation results with that of
Problem-3.
References
ABAQUS Example Manual, Seismic Analysis of a Concrete Gravity Dam, 2020, https://
help.3ds.com/2020/English/DSSIMULIA_Established/SIMACAEEXARefMap/
simaexa-c-concretedam.htm?ContextScope=all&id=3b047001b85d4d5ba54899be58c0
cea7#Pg0.
Ahmad, S., Irons, B.M., and Zienkiewicz, O.C., “Analysis of thick and thin shell structures by
curved elements”, International Journal of Numerical Methods in Engineering, 2: 419–
451, 1970.
Altan, T. and Tekkaya, E., Sheet Metal Forming: Fundamentals, ASM International, 2012.
Altair University, HyperWorks 12.0 Student Edition, 2020. https://altairuniversity.com/
free-altair-student-edition/.
Altair University, 2D Meshing, 2014, https://www.altairuniversity.com/wp-content/
uploads/2014/02/2Dmeshing.pdf.
Andersson, A., “Comparison of sheet-metal-forming simulation and try-out tools in the design
of a forming tool”, Journal of Engineering Design, 15(6): 551–561, 2004.
ANSI, American National Standards Institute, The Initial Graphics Exchange Specification
(IGES), 1996, https://webstore.ansi.org/.
Arrea, M., and Ingraffea, A.R., “Mixed-mode crack propagation in mortar and concrete”,
Department of Structural Engineering Report 81-13, Cornell University, 1981.
ASM International, ASM Handbook Volume 20: Materials Selection and Design, 1997, https://
www. a sm i n t ern atio n a l. o rg /s e a rc h /-/jo u rn a l_c onte nt/56/10192/06481G /
PUBLICATION, 2021.
ASTM E8/E8M, Standard Test Methods for Tension Testing of Metallic Materials, 2021,
https://www.astm.org/Standards/E8.
ASTM E132-17, Standard Test Method for Poisson’s Ratio at Room Temperature, 2021,
https://www.astm.org/Standards/E132.htm.
ASTM E466-15, Standard Practice for Conducting Force Controlled Constant Amplitude
Axial Fatigue Tests of Metallic Materials, 2021, https://www.astm.org/Standards/E466.
htm.
ASTM E1049-85, Standard Practices for Cycle Counting in Fatigue Analysis, 2021, https://
www.astm.org/Standards/E1049.
ASTM E1820-20ae1, Standard Test Method for Measurement of Fracture Toughness, 2021,
https://www.astm.org/Standards/E1820.htm.
Bathe, K.J., Finite Element Procedures, Prentice Hall Inc., 1996.
Bathe, K.J. and Wilson, E.L., Numerical Methods in Finite Element Analysis, Prentice-Hall,
1976.
Bažant, Z.P., Scaling of Structural Strength, 2nd edition, Elsevier Butterwortg-Heinemann,
2005.
Belytschko, T., Lin, J., and Tsay, C.S., “Explicit algorithms for nonlinear dynamics of shells”,
Computer Methods in Applied Mechanics and Engineering, 42: 225–251, 1984.
Beta-CAE.com, META Post-Processor, https://www.beta-cae.com/meta.htm.
Bentley.com, STAAD Pro – 3D Structural Analysis and Design Software, https://www.bent-
ley.com/en/products/brands/staad.
Bhattacharjee, S.S., Smeared Fracture Analysis of Concrete Gravity Dams for Static and
Seismic Loads, Ph.D. Thesis, McGill University, Canada, 1993.
Bhattacharjee, S.S. and Chebl, C., Instrumentation de Stade Olympique de Montreal, Report
to SNC-Lavalin Inc., Montreal, Canada, 1997.
315
316References
Bhattacharjee, S.S. and Leger, P. “Seismic cracking and energy dissipation in concrete gravity
dams”, Journal of Earthquake Engineering and Structural Dynamics, 22(11): 991–
1007, 1993.
Bhattacharjee, S.S. and Leger, P., “Application of NLFM models to predict cracking on con-
crete gravity dams”, ASCE Journal of Structural Engineering, 120(4): 1994.
Borrvall, T., Bhalsod, D., Hallquist, J.O., and Wainscott, B., “Current status of sub-cycling and
multi-scale simulations in LS-DYNA”, Proceedings of the 13th International LS-DYNA
Users Conference, 1–14, 2014.
Broek, D., The Practical Use of Fracture Mechanics, Springer Netherlands, 2012.
BSSC (Building Seismic Safety Council), National Earthquake Hazards Reduction Program
(NEHRP) Recommended Seismic Provisions for New Buildings and Other Structures,
2020, FEMA P-1050-1/2015 Edition, https://www.nibs.org/page/bssc_pubs.
Budynas, R.G., Advanced Strength and Applied Stress Analysis, 2nd edition, McGraw-Hill, 1999.
Chopra, A.K., Dynamics of Structures, 5th edition, Pearson Education Inc., 2017.
Clough, R.W. and Penzien, J., Dynamics of Structures, McGraw-Hill, 1975.
Cook, R.D., Malkus, D.S., and Plesha, M.E., Concepts and Applications of Finite Element
Analysis, 3rd edition, John Wile & Sons, 1989.
Courant, R., Friedrichs, K., and Lewy, H., “Über die partiellen Differenzengleichungen der
mathematischen Physik”, Mathematische Annalen (in German), 100(1): 32–74, 1928.
Cowper, R.G., “The shear coefficient in Timoshenko’s beam theory”, Journal of Applied
Mechanics, 33: 335–340, 1966.
Crandall, S.H., Engineering Analysis, A Survey of Numerical Procedures, McGraw-Hill Book
Co, 1956.
CSIAmerica.com, Computers and Structures Inc., SAP2000 v22, https://www.csiamerica.
com/products/sap2000.
Dassault Systems, ABAQUS Student edition, 2020a, https://edu.3ds.com/en/get-software.
Dassault Systems, SIMULIA User’s Guides, 2020b. https://help.3ds.com/HelpProductsDS.
aspx.
Desai, C.S., Zaman, M.M., Lightner, J.G., and Siriwardane, H.J., “Thin-layer element for
interfaces and joints,” International Journal for Numerical and Analytical Methods in
Geomechanics, 8(1): 19–43, 1984.
Dow Automotive Systems, “Adhesives & Sealants”, “Structural-adhesives”, 2021, https://
www.adhesivesmag.com/articles/96128.
Fung, Y.C., Fundamentals of Solid Mechanics, Prentice Hall, 1965.
GRABCAD.COM, 2020, https://grabcad.com/dashboard.
Hilber, H.M., Hughes, T.J.R., and Taylor, R.L., “Improved numerical dissipation for time inte-
gration algorithms in structural dynamics”, Earthquake Engineering and Structural
Dynamics, 5: 283–292, 1977.
Hughes, T.J.R., The Finite Element Method, Prentice-Hall Inc., 1987.
Irwin, G.R., “Analysis of stresses and strains near the end of a crack traversing a plate”,
Journal of Applied Mechanics, 24: 361–364, 1957.
ISO, International Standards Organization, Standard 10303, STEP (Standard for the Exchange
of Product model data), 2020, https://www.iso.org/standards.html.
Irons, B., and Ahmad, S., Techniques of Finite Elements, Ellis Horwood Limited, Halsted
Press, John Wiley and Sons, Chichester, England, 1980.
Johnson, G.R. and Cook, W.H., “A constitutive model and data for metals subjected to large
strains and high strain rates”, Proceedings of the 7th International Symposium on
Ballistics, 541–547, 1983.
Kennedy, J.B. and Madugula, M.K.S., Elastic Analysis of Structures: Classical and Matrix
Methods, Harper & Row Publishers, 1990.
Lemaitre, J. and Chaboche, J.L., Mechanics of Solid Materials, Cambridge University Press,
1994.
References 317
Lewis, W.J., “Computational form-finding methods for fabric structures”, Proceeding of the
ICE. Engineering Computational Mechanics, 161: 139–149, 2008.
Lewis, W.J., “Modelling of fabric structures and associated design issues”, ASCE Journal of
Architectural Engineering, 19(2): 2013.
Logan, D.L., A First Course in the Finite Element Method, Cengage Learning, 2012.
LSTC.COM, LS-DYNA Theory Manual, ANSYS/LST, 2021. https://www.lstc.com/download/
manuals.
Mahmood, H.F. and Paluzsny, A., “Design of thin walled columns for crash energy
management-their strength and mode of collapse”, SAE Transactions, 90(4): 4039–
4050, 1981. https://www.jstor.org/stable/44725016.
Mindlin, R.D., “Influence of rotatory inertia and shear on flexural motions of isotropic elastic
plates”, ASME Journal of Applied Mechanics, 18(31): 38, 1951.
Miner, M.A., “Cumulative damage in fatigue”, Journal of Applied Mechanics, 12: 149–164,
1945.
Modak, S. and Sotelino, E.D., “The generalized method of structural dynamics applications”,
Advances in Engineering Software, 33: 7–10, 2002.
nCode, Software and solutions for fatigue and durability analysis, HBM Prenscia Inc, https://
www.ncode.com/.
Newmark, N.M., “A method of computation for structural dynamics”, ASCE Journal of
Engineering Mechanics, 85: 67–94, 1959.
NHTSA, National Highway Traffic Safety Administration, US Department of Transportation,
Crash Simulation Vehicle Models, 2010 Toyota Yaris, 2020, https://www.nhtsa.gov/
crash-simulation-vehicle-models.
Paris, P.C. and Erdogan, F., “A critical analysis of crack propagation laws”, Journal of Basic
Engineering, 85(4): 528–533, 1963.
Popov, E.P., Mechanics of Materials, 2nd edition, Prentice-Hall Inc., 1978.
Press, W.H., Teukolsky, S.A., Vetterling, W.T., and Flannery, B.P., Numerical Recipes – The
Art of Scientific Computing, 3rd edition, Cambridge University Press, 2007.
Reissner, E., “The effect of transverse shear deformation on the bending of elastic plates”,
ASME Journal of Applied Mechanics, 12: 68–77, 1945.
Rice, J.R., “A path independent integral and the approximate analysis of strain concentrations
by notches and cracks”, Journal of Applied Mechanics, ASME, 35: 379–386, 1968.
Siemens.COM, Products and Services, FEMAP Pre-and Post-processor, 2021, https://www.
plm.automation.siemens.com/global/en/products/simcenter/femap.html
Tada, H., Paris, P.C., and Irwin, G.R., The Stress Analysis of Cracks Handbook, 3rd edition,
American Society of Mechanical Engineers, 2000.
Tedesco, J.W., McDougal, W.G., and Ross, C.A., Structural Dynamics, Theory and
Applications, Addison Wesley Longman Inc., 1999.
Timoshenko, S.P. and Goodier, J.N., Theory of Elasticity, 3rd edition, McGraw-Hill Book Co.,
1982.
Timoshenko, S.P. and Gere, J.M., Theory of Elastic Stability, McGraw-Hill Book Co, 1963.
Timoshenko, S.P. and Woinowsky-Keiger, S., Theory of Plates and Shells, 2nd edition,
McGraw-Hill International Editions, 1970.
Tinawi, R., Leger, P., Ghrib, F., Bhattacharjee, S.S., and Leclerc, M., “Structural safety of
existing concrete dams: influence of construction joins. Final Report”, Canadian
Electricity Association; Montreal, PQ (Canada); CEA-9032 G 905, 1994.
Uflyand, Y.S., “Wave propagation by transverse vibrations of beams and plates”, Journal of
Applied Mathematics and Mechanics (in Russian), 12: 287–300, 1948.
Ugural, A.C. and Fenster, S.K., Advanced Mechanics of Materials and Applied Elasticity, 5th
edition, Prentice Hall, 2012.
World Auto Steel (WorldAutoSteel.org), Future Steel Vehicle Results and Reports, 2015.
https://www.worldautosteel.org/downloads/futuresteelvehicle-results-and-reports/.
318References
Wikipedia.org, Timoshenko-Ehrenfest Beam Theory, 2020. https://en.wikipedia.org/wiki/
Timoshenko-Ehrenfest_beam_theory.
Zehnder, A.T., Fracture Mechanics, Springer Netherlands, 2012.
Zienkiewicz, O.C. and Taylor, R.L., The Finite Element Method, Volume 1, Basic Formulation
and Linear Problems, 4th edition, McGraw-Hill Book Company, 1989.
Zienkiewicz, O.C. and Taylor, R.L., The Finite Element Method, Volume 2, Solid and Fluid
Mechanics, Dynamics and Non-linearity, 4th edition, McGraw-Hill Book Company,
1991.
Index
A convergence of finite element model solutions,
76, 191
analysis of forces crack-tip, 235
statically determinate, 5 crash strength, 314
statically indeterminate, 5 crashworthiness simulation, 258, 312
aspect ratio, 96 curvature of plate bending, 175
average acceleration method, of dynamic curved beams, 146
response analysis, 272 cumulative fatigue damage, 231
axisymmetric problems, 115 cut-plane, 212
cyclic load, 231
B
bending deflection of beams, 144 D
bending stiffness damping matrix, 274
of composite shell, 179 damping ratio, 247
of plate, 174 damping resistance, 247
bending stresses deep beam, 146
in beams, 128 deformation
in dissimilar materials, 129 axial, 10
in plate, 166 bending, 88
bending theory, see Euler–Bernoulli distortional, 20
body force, 13, 33 shear, 20
boundary constraints, 9, 34, 99 torsional, 134
buckling, thin-wall members, 4 degree of freedom
bulk modulus of elasticity, 28 master, 195
multi-degree of freedom (MDOF) system, 9
C single degree of freedom (SDOF) spring, 7
slave, 195
CAD (computer-aided-design) tools, 16 direction cosines, 137, 148, 181
CAD data exchange formats, 91 displacement
CAE (computer aided engineering) tools, 16 control method of structural analysis, 293
central difference theorem of explicit field, 32, 34
time-domain solution, 278 method of structural analysis, 7
CFL law (Courant–Friedrichs–Lewy) of time variation functions, 30
step limit in explicit dynamic virtual, 32
analysis, 279 distortional stress, 227
cohesive joint element, 207 distributed loads, 45
combined load effects on beam response, 142 ductile material, 3, 219
compatibility condition Duhamel integral, 266
among strains, 23 durability, 230
plane strain, 50 dynamic amplification factor
plane stress, 51 damped system, 247
in the polar coordinate, 111 to impact force, 267
completeness requirement, 191 un-damped system, 246
concentrated forces/loads, 33, 45, 117, 119 dynamic equilibrium
connector element, 195, 198, 203 MDOF, 273
constant stress/strain, 73, 122 SDOF, 247, 270
constitutive models, see stress–strain dynamic response
relationship nonlinear, 308
contact force, 204 of flexible spring-mass, 243, 271
contact simulation, 204 of free rigid body mass, 242
contact stiffness, 204 to impact force, 266
319
320Index
dynamic response analysis flexural rigidity, 144, 169
average acceleration method, 272 force method pf structural analysis, 6
modal superposition method, 276 form-finding, 312
time domain analysis, 270 fracture modes, 236
fracture response simulation, 310
E fracture toughness, 236
free-vibration response, 243
earthquake load, 249 frequency response function, 284
eigenpair, 255 frequency separation, 249
eigenvalue, 254 fundamental frequency of vibration, 252
eigenvalue shift, 259
eigenvector, 254 G
elasticity property
matrix, 27, 294 gage length, 20
tangential, 296 Gauss quadrature, 69
elastic modulus, 24, 219 Green-Lagrange strains, 301
element characteristic length, for explicit Guyan reduction, 259
dynamic analysis, 281
element compatibility, 191 H
element types
axisymmetric, 117 hardening plasticity, 295
beam element, 154, 159 Hooke’s law, 23, 110, 168, 294
mass, 261 hourglass deformation, 71, 90, 184
mixed, 192 householder-QR method, 260
shell elements, 186 high cycle fatigue, 230
3D elements, 122 hinge connection, 200
truss elements, 34, 39 hydrostatic pressure, 28
2D solids, 78, 88, 94 hysteresis, 248
energy error, 102
engineering strain, 24, 218 I
engineering stress, 24, 218
equations of implicit time integration, 272
dynamic equilibrium, 274 impulse, 266
equilibrium in polar coordinates, 109 in-compatible element deformation, 73, 80, 193
static equilibrium, 7, 11 initial penetration, 206
strain compatibility, 23 interface contact, 203
stress compatibility, 28 interface element, 207
stress equilibrium, 14 interpolation functions
stress–strain relationship, 23 beam deformation, 151
Euler–Bernoulli beam theory, 128, 146, 148, coordinate, 34
151, 154, 160, 251 displacement, 36
general 3D solids, 122
F 3-node truss elements, 40
2D deformation fields, 60
fatigue crack growth, 237 2-nodetruss element, 35
fatigue life, 230 intersection, 206
FEA solvers, 16
finite elements J
axisymmetric, 117
beam elements, 154, 160 Jacobian
plane-strain, 80 definition of, 66
plane-stress, 78 determinant, 96
shell, 176, 181, 186 distorted 2D solid element, 66, 70
3D solid, 122 inverse of, 68
truss, 34, 42 physical interpretation, 65
first moment of the beam section, 130 3D system, 122
flexibility method of structural analysis, 6 triangular elements, 74
Index 321
J-integral, 236 Miner’s rule, 231
Johnson–Cook plasticity model, 295, 299, 309 Modak’s T-method, for dynamic response
joint elements, deformable, 198 analysis, 272
modal damping, 274, 277
K modal superposition method of dynamic response
analysis, 269, 276
kinematic constraint, 194 model preparation
kinematic coupling, 196, 201 a-to-g key process steps, 102
kinematic joints, 194 axisymmetric problem, 116
kinematic relations, 270 for contact simulation, 206
Kronecker delta, 254 multi-component, 209
plate with hole, 92
L three-dimensional FEA, 123
modular organization of data, 209
Lagrangian element, 72, 80 Mohr circle of stress, 226
Lagrangian method of nonlinear moment-curvature relationship of plates, 176
structural analysis moment of inertia, 153
total, 305 moment release, 194
updated, 306 momentum, 242
Lagrangian multiplier, 195, 204 MPC (multi-point constraint), 195
Lame’s constants, 28 multi-component model, 209
Lame’s equations, 13
Lanczos transformation method, 260 N
large deformation response, 301
linear acceleration method, of dynamic neutral axis, 10, 56, 128, 130, 147
response analysis, 272 Newmark integration, 272
load-deformation response Newton–Raphson method, 291
monotonic, 293 Newton’s second law, 242
snap-back, 293 nodes, 33
load types, 82 nonlinear analysis
low cycle fatigue, 230 contact problems, 307
dynamic response, 308
M force-deformation response, 290
geometric problems, 301, 306
mass matrix material models, 294
calculation, 255 numerical damping, in dynamic response
consistent, 256 analysis, 272
lumped, 257 numerical integration
master–slave concept, 204 examples, 68
master surface, 204 Gauss quadrature, 69
material failure simulation, 310 of higher order elements, 72
material failure theories, 222 of mass matrix, 256
matrix equilibrium equations, 9 parameters for time-domain dynamic
measurement of deformation intensity, 20 response analysis, 271
mechanics of materials, 2 reduced, 90
membrane analogy, 138 of stiffness function, 70, 122
membrane element, 183
membrane stress resistance of shell, 172 P
mesh-dependent localization of failure, 311
mesh-independent fasteners, 201 parasitic shear, 89
mesh quality checks, 95, 123, 184 Paris law, 237
mesh refinement part property, 98
effect on predicted results, 76 part-to-part connections, 198, 201
h-adaptivity, 86 part-to-part contact, 203
p-adaptivity, 86 patch test, 76
selective, 90 penalty method, 204
Mindlin’s plate deformation theory, 179 penalty stiffness, 204
322Index
period of vibration, 244 simulation model limitations, 4
perturbation of stress flow, 86 singular element, 236
Piola–Kirchoff stresses, 305 size effect, 220
plane strain, 48 skeletal
plasticity models beam, 10, 148
associated, 298 structural members, 10
non-associated, 298 skew, 96
plastic potential function, 298 slave node, 204
plastic strain, 295 slender beam, 154, 158
plate bending theory S–N curve, 231
Kirchoff’s, 166 softening response of structures, 293
Mindlin’s, 179 software
Poisson’s ratio, 25 CAD, 16, 91–94, 102, 123, 186, 214
polar coordinates, 107 CAE tools, 16
polar moment of inertia, 141 FEA solvers, 16, 211, 212
post-processors, 16 finite element simulation, 78, 83
pre-processors, 16, 83 post-processor, 16
principal strain, 225 pre-processor, 16, 48, 83, 84, 90, 92, 93,
principal stress, 223 95, 98
principle of superposition, of stresses, 25 spot-welds, 197, 233
prismatic member, 148 steady state dynamics, 285
product development, automotive, 4 stiffness matrix
of beam element, 153
R of finite element, 33
of flat shell element, 180
radius of curvature, 147 fully integrated, 70
Rayleigh damping coefficient, 274 of general 3D solids, 122
Rayleigh’s quotient, 255 membrane resistance, 177
reference node, 196 positive definite, 9, 40
resonance, 245 reduced integration, 71
response spectrum, 268 spring elements, 7
result quality checks, 77, 103, 211, 234 system assembly, 8, 34, 39
rigid body, 196 tangent, 292, 305
rigid body motion, 9, 191 of 2-node truss element, 37
stiffness method of structural analysis, 7, 32, 60
S stiffness property
beam shear deformation, 156
SDOF (single degree of freedom) system, 7 calculation of finite elements, 32
self-contact, 206 Euler–Bernoulli beam element, 148
semi-infinite solid, 119 flat shell element, 177
shape functions joint element, 200
beam deflection, 151 spring, 7
general 3D solids, 121 truss element, 34
higher order elements, 71 strain
iso-parametric definition, 64 bi-directional, 21
2D deformation fields, 60 compatibility, 23
shear center, 143 components, 22
shear correction factor, 146, 156, 158, 183 energy, 9, 34, 219
shear deformation energy release rate, 237
of cantilever beam, 59 failure, 3
in plates, 179 Green-Lagrange, 301
shear flow through beam, 143 hardening parameter, 296
shear locking of beam, 158 measurement of, 20
shear modulus, 27 normal, 20
shear stress distribution plastic, 295
in beam, 131 rosette, 21
in plate, 168 shear, 20
Index 323
thermal, 83 structural form simulation, 311
virtual, 32 structural instability, 40
strain–displacement relationship strum sequence check, 261
axisymmetric problem, 115 stamping of metal blank, 311
beam element, 152 stress wave propagation, 280
in differential element, 22 subspace iteration method, 260
general 3D solid, 121 surface force, 33
Green-Lagrange, 301
iso-parametric 2D solid, 67 T
nonlinear, 302
plane-strain, 48, 61 tangent modulus, 297
plane-stress, 50 Taylor’s series
plate bending, 167 for strain definition, 13
in polar coordinates, 110 for time domain response, 272
shell element, 181 thin-layer joint, 207
two-node truss element, 36 thin shell, 179
triangular element, 76 time domain analysis
stress error in predicted response, 273, 279
at the crack tip, 235 explicit method, 277
beam section, 10 implicit method, 272
boundary conditions, 15, 50 of MDOF system, 273
components in 3D solids, 11 of SDOF system, 271
concentration factor, 86, 234 time step limit, 273, 276, 279
equilibrium equations, 14, 49 Timoshenko beam theory, 146, 156
initial, 33 torsional response of beam, 133
intensity factor, 235 torsional rigidity, 135
normal, 12, 129 torsional stiffness of member, 141
Piola–Kirchoff, 305 transformation matrix
shear, 12 of beam element stiffness matrix, 148
singularity, 78, 88 of global and local DOF of truss
tangential, 114 element, 37
thermal, 83 transverse normal stress in beam, 132
torsional, 134 true stress, 24, 219
transverse shear, 129 true strain, 24, 219
vector, 13
stress function U
in polar coordinates, 109
prismatic bar, 29 ultimate strength analysis of beam, 291
thin cantilever beam, 54
torsional problem, 136 V
for 2D stress field, 52, 84
stress flow variable amplitude stress cycle, 232
in a plate with hole, 86 vector of
torsion, 139 applied loads, 11, 33
stress–strain relationship nodal displacements, 11, 33
axisymmetric, 110, 116 vector iteration method, 260
for bending deformation of beam, 152 velocity of stress wave, 280, 282
for curved shell elements, 182 vibration frequency
incremental, 299 MDOF, 253
nonlinear, 295 numerical calculation, 259
plane-strain, 49, 63 SDOF, 244
plane-stress, 51, 63 vibration mode shape
for plate bending, 168 MDOF, 259
strain-rate effect, 309 numerical calculation, 259
truss element, 36 SDOF, 250
structural adhesive, 233 simply supported beam, 250
structural analysis, step-by-step load control, 293 simply supported plate, 252
324Index
virtual Wilson θ method of dynamic response
cut-plane, 212 analysis, 272
displacement, 32 Winker formula, 147
work, 32
von Mises stress, 228, 297
Y
W
yield function, 297
warpage, 96 yield strength, 3, 219
warpage-induced deformation, 135 Young’s modulus, 24
You might also like
- Advanced Mechanics of Solids Analytical and Numerical Solutions With MATLABNo ratings yetAdvanced Mechanics of Solids Analytical and Numerical Solutions With MATLAB582 pages
- Finite Element Modeling For Stress Analysis - Robert D. Cook100% (3)Finite Element Modeling For Stress Analysis - Robert D. Cook331 pages
- Advanced Mechanics of Materials and Applied Elasticity90% (10)Advanced Mechanics of Materials and Applied Elasticity108 pages
- Structural Analysis With The Finite Element Method Linear Statics Volume 1, Basis and Solids - E.Oñate96% (28)Structural Analysis With The Finite Element Method Linear Statics Volume 1, Basis and Solids - E.Oñate495 pages
- Numerical Structural Analysis - O'Hara, Steven (SRG) PDF100% (10)Numerical Structural Analysis - O'Hara, Steven (SRG) PDF302 pages
- Buckling and Postbuckling of Beams, Plates, and Shells (PDFDrive)No ratings yetBuckling and Postbuckling of Beams, Plates, and Shells (PDFDrive)603 pages
- Sanjay Govindjee-Engineering Mechanics of Deformable Solids - A Presentation With Exercises-Oxford University Press (2013)100% (1)Sanjay Govindjee-Engineering Mechanics of Deformable Solids - A Presentation With Exercises-Oxford University Press (2013)351 pages
- Computational Methods For Platicity-Souza100% (11)Computational Methods For Platicity-Souza816 pages
- Applied Elasticity and Plasticity-CRC Press - Taylor & Francis (2018)100% (4)Applied Elasticity and Plasticity-CRC Press - Taylor & Francis (2018)565 pages
- Foundations of Mechanics of Materials Part 1No ratings yetFoundations of Mechanics of Materials Part 1276 pages
- A Short Note On The Drag Correlation For SpheresNo ratings yetA Short Note On The Drag Correlation For Spheres4 pages
- CFD Analysis of Static Pressure and Dynamic Pressure For Naca 4412100% (1)CFD Analysis of Static Pressure and Dynamic Pressure For Naca 44128 pages
- Finite Element Analysis Method, Verification and Validation (Barna A. Szabó Ivo Babuška)100% (1)Finite Element Analysis Method, Verification and Validation (Barna A. Szabó Ivo Babuška)387 pages
- Mechanics of Anisotropic Materials (PDFDrive)No ratings yetMechanics of Anisotropic Materials (PDFDrive)328 pages
- 2015 Book EngineeringComputationOfStruct PDF100% (1)2015 Book EngineeringComputationOfStruct PDF325 pages
- (G. Lakshmi Narasaiah) Finite Element Analysis PDFNo ratings yet(G. Lakshmi Narasaiah) Finite Element Analysis PDF349 pages
- Structural Dynamics For The Practising Engineer - 1986100% (5)Structural Dynamics For The Practising Engineer - 1986221 pages
- Rao, H. S. Govinda - Finite Element Method vs. Classical Methods-New Age International (P) LTD., Publishers (2007)100% (3)Rao, H. S. Govinda - Finite Element Method vs. Classical Methods-New Age International (P) LTD., Publishers (2007)434 pages
- Zdeněk Bittnar - Jiří Šejnoha-Numerical Methods in Structural Mechanics-Amer Society of Civil Engineers (1996)100% (2)Zdeněk Bittnar - Jiří Šejnoha-Numerical Methods in Structural Mechanics-Amer Society of Civil Engineers (1996)429 pages
- Direct Methods For Limit States in Structures, (Franck Pastor, Joseph Pastor, Djimedo Kondo (Auth.), Konstantinos Spiliopoulos, Dieter Weichert (Eds.) )100% (2)Direct Methods For Limit States in Structures, (Franck Pastor, Joseph Pastor, Djimedo Kondo (Auth.), Konstantinos Spiliopoulos, Dieter Weichert (Eds.) )281 pages
- Wiggers S.L., Pedersen P. Structural Stability and Vibration. An Integrated Introduction by Analytical and Numerical Methods.2018. 160p100% (3)Wiggers S.L., Pedersen P. Structural Stability and Vibration. An Integrated Introduction by Analytical and Numerical Methods.2018. 160p169 pages
- B. W. Young (Auth.) - Energy Methods of Structural Analysis - Theory, Worked Examples and Problems-Macmillan Education UK (1981) PDF100% (8)B. W. Young (Auth.) - Energy Methods of Structural Analysis - Theory, Worked Examples and Problems-Macmillan Education UK (1981) PDF173 pages
- Finite Element Method With Applications in Engineering95% (19)Finite Element Method With Applications in Engineering532 pages
- Gerhard A. Holzapfel (Sin Apuntes) - Nonlinear Solid Mechanics, A Continuum Approach For Engineering (2001)No ratings yetGerhard A. Holzapfel (Sin Apuntes) - Nonlinear Solid Mechanics, A Continuum Approach For Engineering (2001)469 pages
- Crisfield M A Vol 2 Non-Linear Finite Element Analysis of Solids and Structures Advanced Topics0% (1)Crisfield M A Vol 2 Non-Linear Finite Element Analysis of Solids and Structures Advanced Topics509 pages
- STASA - Applied Finite Element Analysis For Engineers90% (10)STASA - Applied Finite Element Analysis For Engineers674 pages
- (FEM) Crisfield M.a., Non-Linear Finite Element Analysis of Solids and Structures, Vol.1,2 (Wiley, 1996)100% (1)(FEM) Crisfield M.a., Non-Linear Finite Element Analysis of Solids and Structures, Vol.1,2 (Wiley, 1996)869 pages
- Finite Element Analysis Theory and Programming by C. S. Krishnamoorthy82% (17)Finite Element Analysis Theory and Programming by C. S. Krishnamoorthy137 pages
- FiniteElementAnalysisByMirceaRades 1 PDFNo ratings yetFiniteElementAnalysisByMirceaRades 1 PDF274 pages
- Plate Analysis Theory and Application Volume 2 Numerical Methods CT4180No ratings yetPlate Analysis Theory and Application Volume 2 Numerical Methods CT4180136 pages
- Felippa C. Advanced Finite Element Methods (Draft, 2000) (O) (659s) - MNF - PDFNo ratings yetFelippa C. Advanced Finite Element Methods (Draft, 2000) (O) (659s) - MNF - PDF659 pages
- Elasticity and Plasticity: The Mathematical Theory of Elasticity and The Mathematical Theory of PlasticityFrom EverandElasticity and Plasticity: The Mathematical Theory of Elasticity and The Mathematical Theory of PlasticityNo ratings yet
- Introduction to the Explicit Finite Element Method for Nonlinear Transient DynamicsFrom EverandIntroduction to the Explicit Finite Element Method for Nonlinear Transient DynamicsNo ratings yet
- (Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 - Get the ebook instantly with just one click100% (2)(Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 - Get the ebook instantly with just one click49 pages
- Instant Access to (Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 ebook Full Chapters100% (4)Instant Access to (Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 ebook Full Chapters76 pages
- (Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 pdf download100% (1)(Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 pdf download59 pages
- Finite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee - The latest ebook edition with all chapters is now available100% (1)Finite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee - The latest ebook edition with all chapters is now available38 pages
- Download Complete Finite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee PDF for All Chapters100% (23)Download Complete Finite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee PDF for All Chapters60 pages
- Prediction of Burnout: An Artificial Neural Network ApproachFrom EverandPrediction of Burnout: An Artificial Neural Network ApproachNo ratings yet
- Automatic Speech and Speaker Recognition: Large Margin and Kernel MethodsFrom EverandAutomatic Speech and Speaker Recognition: Large Margin and Kernel MethodsJoseph KeshetNo ratings yet
- Immediate download Finite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee ebooks 2024100% (4)Immediate download Finite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee ebooks 202451 pages
- Basic Research and Technologies for Two-Stage-to-Orbit Vehicles: Final Report of the Collaborative Research Centres 253, 255 and 259From EverandBasic Research and Technologies for Two-Stage-to-Orbit Vehicles: Final Report of the Collaborative Research Centres 253, 255 and 259No ratings yet
- Chapter 12 Simple Harmonic Motion and Waves: Section 12.1 Hooke's LawNo ratings yetChapter 12 Simple Harmonic Motion and Waves: Section 12.1 Hooke's Law12 pages
- INTERNAL LOADINGS With Example ProblemsNo ratings yetINTERNAL LOADINGS With Example Problems11 pages
- Vehicle Noise, Vibration, and Sound Quality - (Pages 101 To 125)No ratings yetVehicle Noise, Vibration, and Sound Quality - (Pages 101 To 125)25 pages
- Reginald Hamman Metoda Smanjenja Smicne Cvrstoce Za HB Kriterij PDFNo ratings yetReginald Hamman Metoda Smanjenja Smicne Cvrstoce Za HB Kriterij PDF7 pages
- Co-Requisite: Prerequisite: Data Book / Codes/Standards Course Category Course Designed by ApprovalNo ratings yetCo-Requisite: Prerequisite: Data Book / Codes/Standards Course Category Course Designed by Approval3 pages
- Pump Calculation: Ozmara Angélica Ortiz OrozcoNo ratings yetPump Calculation: Ozmara Angélica Ortiz Orozco16 pages
- Creep Modeling For Concrete-Filled Steel TubesNo ratings yetCreep Modeling For Concrete-Filled Steel Tubes18 pages
- IMPACT ENGINEERING: Fundamentals, Experiments and Nonlinear Finite Elements 1st Edition Marcilio Alves 2024 Scribd Download100% (3)IMPACT ENGINEERING: Fundamentals, Experiments and Nonlinear Finite Elements 1st Edition Marcilio Alves 2024 Scribd Download79 pages
- Crystal Optics With Allowance For Spatial Dispersion Exciton Theory. IINo ratings yetCrystal Optics With Allowance For Spatial Dispersion Exciton Theory. II37 pages
- Concept of Uniform Flow: Lecture Note For Open Channel HydraulicsNo ratings yetConcept of Uniform Flow: Lecture Note For Open Channel Hydraulics26 pages
- PPTA354-2 Karakteristik Batuan Utuh & Massa Batuan FinalNo ratings yetPPTA354-2 Karakteristik Batuan Utuh & Massa Batuan Final125 pages
- Design of Eccentric Connection With Load Lying in Plane of Joint (Rivet & Bolt)No ratings yetDesign of Eccentric Connection With Load Lying in Plane of Joint (Rivet & Bolt)27 pages
- Advanced Mechanics of Solids Analytical and Numerical Solutions With MATLABAdvanced Mechanics of Solids Analytical and Numerical Solutions With MATLAB
- Finite Element Modeling For Stress Analysis - Robert D. CookFinite Element Modeling For Stress Analysis - Robert D. Cook
- Advanced Mechanics of Materials and Applied ElasticityAdvanced Mechanics of Materials and Applied Elasticity
- Structural Analysis With The Finite Element Method Linear Statics Volume 1, Basis and Solids - E.OñateStructural Analysis With The Finite Element Method Linear Statics Volume 1, Basis and Solids - E.Oñate
- Numerical Structural Analysis - O'Hara, Steven (SRG) PDFNumerical Structural Analysis - O'Hara, Steven (SRG) PDF
- Buckling and Postbuckling of Beams, Plates, and Shells (PDFDrive)Buckling and Postbuckling of Beams, Plates, and Shells (PDFDrive)
- Sanjay Govindjee-Engineering Mechanics of Deformable Solids - A Presentation With Exercises-Oxford University Press (2013)Sanjay Govindjee-Engineering Mechanics of Deformable Solids - A Presentation With Exercises-Oxford University Press (2013)
- Applied Elasticity and Plasticity-CRC Press - Taylor & Francis (2018)Applied Elasticity and Plasticity-CRC Press - Taylor & Francis (2018)
- CFD Analysis of Static Pressure and Dynamic Pressure For Naca 4412CFD Analysis of Static Pressure and Dynamic Pressure For Naca 4412
- Finite Element Analysis Method, Verification and Validation (Barna A. Szabó Ivo Babuška)Finite Element Analysis Method, Verification and Validation (Barna A. Szabó Ivo Babuška)
- (G. Lakshmi Narasaiah) Finite Element Analysis PDF(G. Lakshmi Narasaiah) Finite Element Analysis PDF
- Structural Dynamics For The Practising Engineer - 1986Structural Dynamics For The Practising Engineer - 1986
- Rao, H. S. Govinda - Finite Element Method vs. Classical Methods-New Age International (P) LTD., Publishers (2007)Rao, H. S. Govinda - Finite Element Method vs. Classical Methods-New Age International (P) LTD., Publishers (2007)
- Zdeněk Bittnar - Jiří Šejnoha-Numerical Methods in Structural Mechanics-Amer Society of Civil Engineers (1996)Zdeněk Bittnar - Jiří Šejnoha-Numerical Methods in Structural Mechanics-Amer Society of Civil Engineers (1996)
- Direct Methods For Limit States in Structures, (Franck Pastor, Joseph Pastor, Djimedo Kondo (Auth.), Konstantinos Spiliopoulos, Dieter Weichert (Eds.) )Direct Methods For Limit States in Structures, (Franck Pastor, Joseph Pastor, Djimedo Kondo (Auth.), Konstantinos Spiliopoulos, Dieter Weichert (Eds.) )
- Wiggers S.L., Pedersen P. Structural Stability and Vibration. An Integrated Introduction by Analytical and Numerical Methods.2018. 160pWiggers S.L., Pedersen P. Structural Stability and Vibration. An Integrated Introduction by Analytical and Numerical Methods.2018. 160p
- B. W. Young (Auth.) - Energy Methods of Structural Analysis - Theory, Worked Examples and Problems-Macmillan Education UK (1981) PDFB. W. Young (Auth.) - Energy Methods of Structural Analysis - Theory, Worked Examples and Problems-Macmillan Education UK (1981) PDF
- Finite Element Method With Applications in EngineeringFinite Element Method With Applications in Engineering
- Gerhard A. Holzapfel (Sin Apuntes) - Nonlinear Solid Mechanics, A Continuum Approach For Engineering (2001)Gerhard A. Holzapfel (Sin Apuntes) - Nonlinear Solid Mechanics, A Continuum Approach For Engineering (2001)
- Crisfield M A Vol 2 Non-Linear Finite Element Analysis of Solids and Structures Advanced TopicsCrisfield M A Vol 2 Non-Linear Finite Element Analysis of Solids and Structures Advanced Topics
- STASA - Applied Finite Element Analysis For EngineersSTASA - Applied Finite Element Analysis For Engineers
- (FEM) Crisfield M.a., Non-Linear Finite Element Analysis of Solids and Structures, Vol.1,2 (Wiley, 1996)(FEM) Crisfield M.a., Non-Linear Finite Element Analysis of Solids and Structures, Vol.1,2 (Wiley, 1996)
- Finite Element Analysis Theory and Programming by C. S. KrishnamoorthyFinite Element Analysis Theory and Programming by C. S. Krishnamoorthy
- Plate Analysis Theory and Application Volume 2 Numerical Methods CT4180Plate Analysis Theory and Application Volume 2 Numerical Methods CT4180
- Felippa C. Advanced Finite Element Methods (Draft, 2000) (O) (659s) - MNF - PDFFelippa C. Advanced Finite Element Methods (Draft, 2000) (O) (659s) - MNF - PDF
- Structural Reliability Analysis and PredictionFrom EverandStructural Reliability Analysis and Prediction
- Elasticity and Plasticity: The Mathematical Theory of Elasticity and The Mathematical Theory of PlasticityFrom EverandElasticity and Plasticity: The Mathematical Theory of Elasticity and The Mathematical Theory of Plasticity
- Introduction to the Explicit Finite Element Method for Nonlinear Transient DynamicsFrom EverandIntroduction to the Explicit Finite Element Method for Nonlinear Transient Dynamics
- Energy Principles and Variational Methods in Applied MechanicsFrom EverandEnergy Principles and Variational Methods in Applied Mechanics
- Nonlinear Finite Element Analysis of Solids and StructuresFrom EverandNonlinear Finite Element Analysis of Solids and Structures
- (Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 - Get the ebook instantly with just one click(Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 - Get the ebook instantly with just one click
- Instant Access to (Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 ebook Full ChaptersInstant Access to (Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 ebook Full Chapters
- Antenna Arraying Techniques in the Deep Space NetworkFrom EverandAntenna Arraying Techniques in the Deep Space Network
- Collapse analysis of externally prestressed structuresFrom EverandCollapse analysis of externally prestressed structures
- (Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 pdf download(Ebook) Finite Element Analysis of Solids and Structures by Sudip S. Bhattacharjee ISBN 9780367437053, 0367437058 pdf download
- Finite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee - The latest ebook edition with all chapters is now availableFinite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee - The latest ebook edition with all chapters is now available
- Download Complete Finite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee PDF for All ChaptersDownload Complete Finite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee PDF for All Chapters
- Prediction of Burnout: An Artificial Neural Network ApproachFrom EverandPrediction of Burnout: An Artificial Neural Network Approach
- Automatic Speech and Speaker Recognition: Large Margin and Kernel MethodsFrom EverandAutomatic Speech and Speaker Recognition: Large Margin and Kernel Methods
- Immediate download Finite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee ebooks 2024Immediate download Finite Element Analysis of Solids and Structures 1st Edition Sudip S. Bhattacharjee ebooks 2024
- Error-controlled Adaptive Finite Elements in Solid MechanicsFrom EverandError-controlled Adaptive Finite Elements in Solid Mechanics
- Basic Research and Technologies for Two-Stage-to-Orbit Vehicles: Final Report of the Collaborative Research Centres 253, 255 and 259From EverandBasic Research and Technologies for Two-Stage-to-Orbit Vehicles: Final Report of the Collaborative Research Centres 253, 255 and 259
- Chapter 12 Simple Harmonic Motion and Waves: Section 12.1 Hooke's LawChapter 12 Simple Harmonic Motion and Waves: Section 12.1 Hooke's Law
- Vehicle Noise, Vibration, and Sound Quality - (Pages 101 To 125)Vehicle Noise, Vibration, and Sound Quality - (Pages 101 To 125)
- Reginald Hamman Metoda Smanjenja Smicne Cvrstoce Za HB Kriterij PDFReginald Hamman Metoda Smanjenja Smicne Cvrstoce Za HB Kriterij PDF
- Co-Requisite: Prerequisite: Data Book / Codes/Standards Course Category Course Designed by ApprovalCo-Requisite: Prerequisite: Data Book / Codes/Standards Course Category Course Designed by Approval
- IMPACT ENGINEERING: Fundamentals, Experiments and Nonlinear Finite Elements 1st Edition Marcilio Alves 2024 Scribd DownloadIMPACT ENGINEERING: Fundamentals, Experiments and Nonlinear Finite Elements 1st Edition Marcilio Alves 2024 Scribd Download
- Crystal Optics With Allowance For Spatial Dispersion Exciton Theory. IICrystal Optics With Allowance For Spatial Dispersion Exciton Theory. II
- Concept of Uniform Flow: Lecture Note For Open Channel HydraulicsConcept of Uniform Flow: Lecture Note For Open Channel Hydraulics
- PPTA354-2 Karakteristik Batuan Utuh & Massa Batuan FinalPPTA354-2 Karakteristik Batuan Utuh & Massa Batuan Final
- Design of Eccentric Connection With Load Lying in Plane of Joint (Rivet & Bolt)Design of Eccentric Connection With Load Lying in Plane of Joint (Rivet & Bolt)