Mechanics of

Materials
Formulations and Solutions
with Python
Series in Computational Methods

Series Editor: Gui-Rong Liu (University of Cincinnati, USA)

Published

Vol. 3 Mechanics of Materials: Formulations and Solutions with Python

by G R Liu

Vol. 1 Numbers and Functions: Theory, Formulation, and Python Codes

by G R Liu
 Series in Computational Methods
Volume 3

Mechanics of
Materials
Formulations and Solutions
with Python

G. R. Liu
University of Cincinnati, USA

World Scientific
NEW JERSEY • LONDON • SINGAPORE • GENEVA • BEIJING • SHANGHAI • TAIPEI • CHENNAI
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

Library of Congress Cataloging-in-Publication Data

Names: Liu, G. R. (Gui-Rong), author.
Title: Mechanics of materials : formulations and solutions with Python /
G. R. Liu, University of Cincinnati, USA.
Description: Singapore ; Hackensack, NJ ; London : World Scientific, 2025. |
Series: Series in computational methods ; vol. 3 | Includes bibliographical references and index.
Identifiers: LCCN 2024028348 | ISBN 9789811294525 (hardcover) |
ISBN 9789811294532 (ebook for institutions) | ISBN 9789811294549 (ebook for individuals)
Subjects: LCSH: Strength of materials--Computer simulation. | Python (Computer program language)
Classification: LCC TA405 .L554 2025 | DDC 620.1/120113--dc23/eng/20241101
LC record available at [Link]

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

Copyright © 2025 by World Scientific Publishing Co. Pte. Ltd.

All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or
mechanical, including photocopying, recording or any information storage and retrieval system now known or to
be invented, without written permission from the publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center,
Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from
the publisher.

For any available supplementary material, please visit

[Link]

Desk Editors: Nambirajan Karuppiah/Steven Patt

Typeset by Stallion Press

Email: enquiries@[Link]

Printed in Singapore
 About the Author

Gui-Rong Liu received his Ph.D. from Tohoku Uni-

versity, Japan, in 1991. He was a Postdoctoral Fellow at
Northwestern University, USA, from 1991 to 1993. He
was a Professor at the National University of Singapore
until 2010. He is currently a Professor at the Department
of Aerospace Engineering and Engineering Mechanics,
University of Cincinnati, USA. He was the Founder of
the Association for Computational Mechanics (Singa-
pore) (SACM) and served as the President of SACM
until 2010. He served as the President of the Asia-Pacific
Association for Computational Mechanics (APACM) (2010–2013) and an
Executive Council Member of the International Association for Compu-
tational Mechanics (IACM) (2005–2010; 2020–2026). He authored a large
number of journal papers and books, including two bestsellers, Mesh Free
Method: Moving Beyond the Finite Element Method and Smoothed Par-
ticle Hydrodynamics: A Meshfree Particle Methods. He is the Editor-in-
Chief of the International Journal of Computational Methods and served
as an Associate Editor for IPSE and MANO. He is the recipient of numer-
ous awards, including the Singapore Defence Technology Prize, NUS Out-
standing University Researcher Award, NUS Best Teacher Award, APACM
Computational Mechanics Award, JSME Computational Mechanics Award,
ASME Ted Belytschko Applied Mechanics Award, Zienkiewicz Medal from
APACM, AJCM Computational Mechanics Award, Humboldt Research
Award, and SACM Medal from the Association of Computational Mechan-
ics (Singapore). He has been listed as one among the world’s top 1% most
influential scientists (Highly Cited Researchers) by Thomson Reuters for a
number of years.

v
This page intentionally left blank
 Contents

About the Author v

1. Introduction 1
1.1 Computational methods . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Why start and contribute to this book series . . . . . . . . . . . 2
1.3 Mechanics of materials: Essential for safe and effective use of
materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Who may read this book? . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Codes used in this book . . . . . . . . . . . . . . . . . . . . . . . 4
1.6 Use of external modules or dependences . . . . . . . . . . . . . . 5
1.7 Use of help() . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2. Preliminaries: Number, Vector, Tensor, and Functions 9

2.1 Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Vectors: Collection of numbers . . . . . . . . . . . . . . . . . . . 12
2.2.1 Vectors defined in a computer . . . . . . . . . . . . . . . 12
2.2.2 Basic properties of vectors . . . . . . . . . . . . . . . . . 13
2.2.3 Attributes of a vector in a computer . . . . . . . . . . . . 14
2.2.4 The standard basis vectors . . . . . . . . . . . . . . . . . 15
2.2.5 Basic arithmetic operations for vectors . . . . . . . . . . 16
2.2.6 Other arithmetic operations for vectors . . . . . . . . . . 17
2.3 Dot product of vectors, producing a scalar . . . . . . . . . . . . 18
2.3.1 Definition: A scalar produced using two vectors . . . . . 18
2.3.2 Direction cosine of two vectors . . . . . . . . . . . . . . . 19
2.3.3 Inner product associated norm . . . . . . . . . . . . . . . 20
2.3.4 Orthogonality and orthogonal projection . . . . . . . . . 20
2.3.5 Derivation of the law of cosines . . . . . . . . . . . . . . 21

vii
viii Mechanics of Materials: Formulations and Solutions with Python

2.3.6Power, energy, and work-done . . . . . . . . . . . . . . . 22

2.3.7Python examples . . . . . . . . . . . . . . . . . . . . . . 22
[Link] Dot product of an arbitrary pair of vectors . . 22
[Link] Dot product of a pair of parallel vectors . . . . 23
[Link] Dot product of two orthogonal vectors . . . . . 23
[Link] Direction cosine of two vectors . . . . . . . . . 23
2.3.8 Types of norms of a vector . . . . . . . . . . . . . . . . . 24
2.3.9 Direction cosine of an arbitrary vector . . . . . . . . . . 24
2.4 Outer product of two vectors, producing a matrix . . . . . . . . 25
2.4.1 Definition of outer product . . . . . . . . . . . . . . . . . 25
2.4.2 Rank-one matrix . . . . . . . . . . . . . . . . . . . . . . . 26
2.4.3 Python example . . . . . . . . . . . . . . . . . . . . . . . 27
2.5 Cross product of vectors, producing a vector . . . . . . . . . . . 27
2.5.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.5.2 Formulation . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.5.3 Volume of a parallelepiped . . . . . . . . . . . . . . . . . 29
2.5.4 Differences between the dot product and cross product . 30
2.5.5 Derivation of formulas . . . . . . . . . . . . . . . . . . . 30
2.5.6 Examples of computing cross products . . . . . . . . . . 33
2.6 Examples of vector operations and visualization . . . . . . . . . 34
2.6.1 Creation of vectors using the standard basis vectors, and
scaling vectors . . . . . . . . . . . . . . . . . . . . . . . . 35
2.6.2 Addition of vectors . . . . . . . . . . . . . . . . . . . . . 36
2.6.3 Element-wise multiplication and dot product of
two vectors . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.6.4 Orthogonal vectors . . . . . . . . . . . . . . . . . . . . . 39
2.7 Coordinate transformation . . . . . . . . . . . . . . . . . . . . . 41
2.8 Vectors in mechanics . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.8.1 Notation for coordinate systems . . . . . . . . . . . . . . 42
2.8.2 Description of vectors . . . . . . . . . . . . . . . . . . . . 42
2.8.3 Equivalence in representation, first-order tensor . . . . . 43
2.8.4 Tensor concept in general . . . . . . . . . . . . . . . . . . 44
2.8.5 Definition of tensor in mechanics of materials . . . . . . 44
2.9 Transformation rules . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.9.1 Definition of coordinate transformation . . . . . . . . . . 45
2.9.2 Representation of a tensor in a coordinate . . . . . . . . 46
2.9.3 Essential rule to obtain values on
a coordinate axis . . . . . . . . . . . . . . . . . . . . . . 46
2.10 Transformation matrix . . . . . . . . . . . . . . . . . . . . . . . 47
2.10.1 Two-dimensional (2D) cases . . . . . . . . . . . . . . . . 47
2.10.2 General definition of transformation matrix for
3D cases . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
 Contents ix

2.10.3Transformation via yaw, pitch, and roll in 3D . . . . . . 50

2.10.4Property of transformation matrix, a unitary matrix . . 52
2.10.5Numpy code for transformation matrices . . . . . . . . . 54
2.10.6Examples: Unitary property of transformation matrix . . 55
2.10.7Examples: Basic property preservation by transformation
matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.10.8 Example: Transformation matrix on arbitrary
matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.11 Rotation matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.11.1 Derivation for 2D cases . . . . . . . . . . . . . . . . . . . 58
2.11.2 Relationship between transformation and rotation
matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.12 Function, an object varying over space . . . . . . . . . . . . . . . 60
2.13 Gradient of a scalar function, a vector . . . . . . . . . . . . . . . 61
2.13.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.13.2 Python demonstration . . . . . . . . . . . . . . . . . . . 62
2.14 Gradient of a vector of function, a matrix . . . . . . . . . . . . . 64
2.14.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.14.2 Python examples . . . . . . . . . . . . . . . . . . . . . . 65
2.14.3 Code function for gradient of vector functions . . . . . . 66
2.15 Divergence of a vector function, a scalar . . . . . . . . . . . . . . 67
2.15.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.15.2 Python code and examples . . . . . . . . . . . . . . . . . 68
2.16 Curl of a vector function, a vector . . . . . . . . . . . . . . . . . 69
2.16.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.16.2 Python code and examples . . . . . . . . . . . . . . . . . 70
2.17 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

3. Forces and Stresses 73

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.2 Example: A bar subjected to a point force . . . . . . . . . . . . 74
3.3 Example: A bar subjected to a distributed force . . . . . . . . . 76
3.3.1 Derivation of equilibrium equation . . . . . . . . . . . . . 76
3.3.2 Stress solution to 1D equilibrium equation . . . . . . . . 77
3.4 General measure of force vectors in 3D space . . . . . . . . . . . 78
3.4.1 Graphic representation of force vectors . . . . . . . . . . 78
3.4.2 Components of a force, orthogonal projection . . . . . . 78
3.4.3 Equivalence of force vector and its components . . . . . . 78
3.4.4 Force vector in transformed coordinates . . . . . . . . . . 79
3.5 General measure of stresses in 3D solids . . . . . . . . . . . . . . 80
3.5.1 General definition of stress . . . . . . . . . . . . . . . . . 80
x Mechanics of Materials: Formulations and Solutions with Python

3.5.2 Shear stresses on a surface . . . . . . . . . . . . . . . . . 81

3.5.3 Stress representation in 3D solids . . . . . . . . . . . . . 83
3.5.4 Shear stress equivalence, symmetric tensor . . . . . . . . 84
3.5.5 Dimensionality and stresses . . . . . . . . . . . . . . . . . 85
3.6 Tractions on an arbitrary surface . . . . . . . . . . . . . . . . . . 86
3.6.1 Formulation . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.6.2 Stress vector projection . . . . . . . . . . . . . . . . . . . 87
3.6.3 Area projection . . . . . . . . . . . . . . . . . . . . . . . 88
3.6.4 Traction on a boundary surface . . . . . . . . . . . . . . 89
3.6.5 Stress vector projection, again . . . . . . . . . . . . . . . 90
[Link] Normal stress on a surface . . . . . . . . . . . 90
[Link] Shear stress on a surface . . . . . . . . . . . . . 91
[Link] The magnitude of the shear stress on a surface 91
3.6.6 Example: Stresses on a given surface . . . . . . . . . . . 91
3.7 Coordinate transformation of second order tensor (stress) . . . . 95
3.7.1 Formulation by simple inspection . . . . . . . . . . . . . 96
3.7.2 Inverse transformation . . . . . . . . . . . . . . . . . . . 97
3.7.3 Python example: Stress tensor under coordinate
transformation . . . . . . . . . . . . . . . . . . . . . . . . 97
[Link] Use of matrix formulation . . . . . . . . . . . . 98
[Link] Use of tensor operation . . . . . . . . . . . . . 99
3.7.4 Transformation of stress tensor in 2D . . . . . . . . . . . 99
[Link] Example: Uniaxial stress after a coordinate
transformation . . . . . . . . . . . . . . . . . . 100
[Link] Example: A general stress state under coordi-
nate transformation . . . . . . . . . . . . . . . 103
3.8 Principal stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.8.1 Eigenvalue problem . . . . . . . . . . . . . . . . . . . . . 106
3.8.2 Characteristic equation, polynomial . . . . . . . . . . . . 107
3.8.3 Python code for finding principal stresses and their
directions . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.8.4 Example: Computing the principal stresses and principal
directions . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.9 Stress invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.10 Special stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.10.1 Octahedral stress . . . . . . . . . . . . . . . . . . . . . . 110
3.10.2 The von Mises stress . . . . . . . . . . . . . . . . . . . . 111
3.10.3 The strain energy stress . . . . . . . . . . . . . . . . . . . 112
3.10.4 Python code for computing various stresses . . . . . . . . 112
3.10.5 Example: Computation of various stresses . . . . . . . . 114
3.11 Mohr circles, geometric representation . . . . . . . . . . . . . . . 115
3.11.1 Formulation for 2D stress states, double angle formula . 116
 Contents xi

3.11.2 Python code for coordinate transformation for 2D

stresses, double angle . . . . . . . . . . . . . . . . . . . . 117
3.11.3 Principal stresses for 2D stress states . . . . . . . . . . . 117
3.11.4 Example: Principal stresses of 2D stress states . . . . . . 118
3.11.5 Mohr circle for 2D and stress states . . . . . . . . . . . . 119
3.11.6 Sign convention for 2D Mohr circles . . . . . . . . . . . . 120
3.11.7 Python code for generating 2D Mohr circles . . . . . . . 121
3.11.8 Example: Mohr circle of a uniaxial stress state . . . . . . 122
3.11.9 Example: Mohr circle of a general 2D stress state . . . . 125
3.11.10 Mohr circle for 3D stress states . . . . . . . . . . . . . . 126
3.11.11 Python code for generating 3D Mohr circles . . . . . . . 128
3.11.12 Example: Mohr circle of a three-dimensional (3D)
stress state . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.12 Determination of the principal axis angle . . . . . . . . . . . . . 131
3.12.1 Transformation matrices for Yaw, Pitch, and Roll . . . . 131
3.12.2 Stress transformation via Yaw, Pitch, and Roll . . . . . . 132
3.12.3 Diagonalization of a stress tensor in 3D . . . . . . . . . . 133
3.12.4 Python code to find Yaw, Pitch, and Roll angles to diag-
onalize stress tensors . . . . . . . . . . . . . . . . . . . . 134
3.12.5 Example: Diagonalize a stress tensor via Yaw, Pitch,
and Roll . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3.13 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

4. Displacements and Strains 139

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.2 Displacement vector, representations . . . . . . . . . . . . . . . . 140
4.3 Displacement vector in transformed coordinates . . . . . . . . . 141
4.3.1 Forward transformation . . . . . . . . . . . . . . . . . . . 141
4.3.2 Inverse transformation . . . . . . . . . . . . . . . . . . . 142
4.4 Vector of displacement functions . . . . . . . . . . . . . . . . . . 142
4.4.1 Displacement functions, a vector function . . . . . . . . . 142
4.4.2 Code using tensor operation for transformation . . . . . 144
4.5 Strains derived via kinematic relations . . . . . . . . . . . . . . . 145
4.5.1 An ideal uniform bar under tension . . . . . . . . . . . . 145
4.5.2 Normal strains . . . . . . . . . . . . . . . . . . . . . . . . 147
4.5.3 Shear strains . . . . . . . . . . . . . . . . . . . . . . . . . 147
4.6 Strains, derivatives of displacement functions . . . . . . . . . . . 148
4.6.1 Displacement gradient . . . . . . . . . . . . . . . . . . . 148
4.6.2 Small strains, derived via displacement gradient . . . . . 150
4.6.3 Example: Computation of small strains . . . . . . . . . . 151
4.6.4 Rotation-displacement relations . . . . . . . . . . . . . . 152
xii Mechanics of Materials: Formulations and Solutions with Python

4.6.5 Large strains, derived via displacement gradient . . . . . 153

4.6.6 Example: Computation of large strains . . . . . . . . . . 155
4.6.7 Large strains derived via conventional deformation
gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
[Link] Deformation gradient, definition . . . . . . . . 157
[Link] Deformation gradient, in displacements . . . . 157
[Link] Cauchy–Green deformation tensor . . . . . . . 158
[Link] Green–Lagrange strain tensor . . . . . . . . . . 158
4.7 Strain–displacement relation, a general derivation . . . . . . . . 159
4.7.1 Normal strain along a fiber, magnification factor . . . . . 160
[Link] Evaluation of the length of a fiber in a
deformed solid . . . . . . . . . . . . . . . . . . 160
[Link] Engineering strain of a given fiber . . . . . . . 161
[Link] Magnification factor of a fiber, large normal
strain . . . . . . . . . . . . . . . . . . . . . . . 161
[Link] Magnification factor and the strain
components . . . . . . . . . . . . . . . . . . . . 162
[Link] Direction cosine of a fiber after
deformation . . . . . . . . . . . . . . . . . . . . 163
4.7.2 Shear strains caused by large deformation . . . . . . . . 164
4.7.3 Example: Symbolic computation of strains . . . . . . . . 165
4.7.4 Example: Computation of displacements, normal and
shear strains . . . . . . . . . . . . . . . . . . . . . . . . . 167
4.8 Coordinate transformation of strain tensor . . . . . . . . . . . . 171
4.9 Principal strains . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
4.10 Strain invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
4.11 Example: Strain components after coordinate rotation . . . . . . 174
4.12 Example: Computation of various strains . . . . . . . . . . . . . 175
4.13 Compatibility equations . . . . . . . . . . . . . . . . . . . . . . . 177
4.13.1 2D problems . . . . . . . . . . . . . . . . . . . . . . . . . 177
4.13.2 3D problems . . . . . . . . . . . . . . . . . . . . . . . . . 178
4.14 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

5. Elasticity Tensor and Constitutive Equations 181

5.1 Simplest case: 1D tests . . . . . . . . . . . . . . . . . . . . . . . 181
5.2 Comparison between Young’s modulus and spring constants . . 183
5.3 General formulation . . . . . . . . . . . . . . . . . . . . . . . . . 184
5.4 Coordinate transformation of fourth-order tensor . . . . . . . . . 184
5.5 Codes for fourth order tensor transformation . . . . . . . . . . . 185
5.6 Reduction of elastic constants . . . . . . . . . . . . . . . . . . . 186
5.6.1 Reduction due to stress and strain symmetry, 81 → 36 . 186
 Contents xiii

5.6.2 Reduction due to the smoothness of the strain energy

density . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
[Link] Strain energy density in solids . . . . . . . . . 187
[Link] Symmetry from the smoothness of the strain
energy density . . . . . . . . . . . . . . . . . . 188
[Link] Reduction due to strain energy density symme-
try, 36 → 21 . . . . . . . . . . . . . . . . . . . 189
5.6.3 Tensor vs. matrix, a discussion . . . . . . . . . . . . . . . 190
5.7 The Voigt notation . . . . . . . . . . . . . . . . . . . . . . . . . 190
5.8 Codes for conversion between a 2D matrix to fourth-order
tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
5.9 Monoclinic materials, 21 → 13 . . . . . . . . . . . . . . . . . . . 193
5.10 Orthotropic materials, 13 → 9 . . . . . . . . . . . . . . . . . . . 193
5.11 Transversely isotropic materials, 9 → 5 . . . . . . . . . . . . . . 193
5.12 Cubic materials, 5 → 3 . . . . . . . . . . . . . . . . . . . . . . . 194
5.13 Isotropic materials, 3 → 2 . . . . . . . . . . . . . . . . . . . . . . 194
5.13.1 For 3D problems . . . . . . . . . . . . . . . . . . . . . . . 194
5.13.2 For stress and strain invariants . . . . . . . . . . . . . . . 197
5.13.3 For 2D plane stress problems . . . . . . . . . . . . . . . . 197
5.13.4 For 2D plane strain problems . . . . . . . . . . . . . . . . 198
5.14 Mean and deviatoric stress and strain tensors . . . . . . . . . . . 199
5.15 Compliance matrix . . . . . . . . . . . . . . . . . . . . . . . . . . 200
5.15.1 Compliance matrix for orthotropic materials . . . . . . . 200
5.15.2 Python code for elasticity matrix for orthotropic materi-
als (3D problems) . . . . . . . . . . . . . . . . . . . . . . 201
5.15.3 Plane stress problems . . . . . . . . . . . . . . . . . . . . 203
5.15.4 Python code for elasticity matrix for orthotropic materi-
als (2D plane stress) . . . . . . . . . . . . . . . . . . . . . 203
5.15.5 Plane strain problems . . . . . . . . . . . . . . . . . . . . 204
5.15.6 Python code for elasticity matrix for orthotropic materi-
als (2D plane strain) . . . . . . . . . . . . . . . . . . . . 206
5.15.7 Compliance matrix for isotropic materials . . . . . . . . . 207
5.15.8 Derivation of the C matrix for isotropic materials . . . . 207
5.15.9 Compliance equations for 2D problems . . . . . . . . . . 208
5.15.10 Example: Proof of relation between G, E and ν . . . . . 209
5.16 A comprehensive example: Displacement, strain and
stress analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
5.16.1 Setting of the problem . . . . . . . . . . . . . . . . . . . 212
5.16.2 Solution to the problem . . . . . . . . . . . . . . . . . . . 213
[Link] Analysis of the problem . . . . . . . . . . . . . 213
[Link] Find the displacement functions . . . . . . . . 214
[Link] Compute the gradient of the displacement . . . 215
xiv Mechanics of Materials: Formulations and Solutions with Python

[Link] Find the strain tensor functions . . . . . . . . 215

[Link] Compute the normal strains . . . . . . . . . . . 216
[Link] Compute the shear strains . . . . . . . . . . . 216
[Link] Compute the principal strains and strain
invariants . . . . . . . . . . . . . . . . . . . . . 217
[Link] Displacements in the new coordinates
system . . . . . . . . . . . . . . . . . . . . . . . 218
[Link] Strains in the new coordinates system . . . . . 218
[Link] Compute the stresses . . . . . . . . . . . . . . 218
[Link] Compute the principal stresses, and stress
invariants . . . . . . . . . . . . . . . . . . . . . 219
[Link] Rotate the coordinates by 30◦ about z-axis,
and find the stresses at point B in the new
coordinates system . . . . . . . . . . . . . . . . 220
5.17 Thermal expansion effects . . . . . . . . . . . . . . . . . . . . . . 220
5.17.1 Thermal strains . . . . . . . . . . . . . . . . . . . . . . . 220
5.17.2 Strain-stress-temperature relation . . . . . . . . . . . . . 221
5.17.3 Stress-strain-temperature relation . . . . . . . . . . . . . 221
5.18 Examples for elasticity tensor transformation . . . . . . . . . . . 221
5.18.1 Examples on arbitrary symmetric 6 by 6 matrix . . . . . 221
5.18.2 Examples on elasticity matrices for steel . . . . . . . . . 224
5.18.3 Examples on elasticity matrices for steel for 2D
plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . 226
5.18.4 Examples on elasticity matrices for steel for 2D
plane strain . . . . . . . . . . . . . . . . . . . . . . . . . 227
5.18.5 Examples on elasticity matrices for carbon/epoxy
composite . . . . . . . . . . . . . . . . . . . . . . . . . . 228
5.19 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

6. Equilibrium Equations 233

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
6.2 Equilibrium equation in terms of stresses . . . . . . . . . . . . . 234
6.2.1 3D free-body diagram . . . . . . . . . . . . . . . . . . . . 234
6.2.2 The standard sign convention . . . . . . . . . . . . . . . 235
6.2.3 Equilibrium equation in x-direction . . . . . . . . . . . . 236
6.2.4 Equilibrium in y- and z-directions . . . . . . . . . . . . . 237
6.2.5 Equilibrium equation in indicial notation . . . . . . . . . 237
6.2.6 Equilibrium equation in 2D and 1D . . . . . . . . . . . . 238
6.2.7 Equilibrium equation via divergence of stress
tensor function . . . . . . . . . . . . . . . . . . . . . . . . 238
6.3 Equilibrium equations in displacements . . . . . . . . . . . . . . 240
6.3.1 General formulation in 3D . . . . . . . . . . . . . . . . . 240
 Contents xv

6.3.2 Formulations of 2D problems . . . . . . . . . . . . . . . . 242

[Link] Plane strain problems . . . . . . . . . . . . . . 242
[Link] Plane stress problems . . . . . . . . . . . . . . 243
6.3.3 Formulations of 1D bar problems . . . . . . . . . . . . . 244
[Link] 1D yz-constrained problem . . . . . . . . . . . 244
[Link] 1D y-constrained problem . . . . . . . . . . . . 245
[Link] 1D bar problem . . . . . . . . . . . . . . . . . 245
6.4 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . 246
6.4.1 Displacement boundary conditions . . . . . . . . . . . . . 246
6.4.2 Stress boundary conditions . . . . . . . . . . . . . . . . . 246
6.5 On techniques for solving solid mechanics problems . . . . . . . 247
6.6 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248

7. Solution to Axially Loaded Bars 249

7.1 Governing equation: Second-order
differential equation . . . . . . . . . . . . . . . . . . . . . . . . . 249
7.2 Solution procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 250
7.2.1 Example: A bar subjected to a force
at its right end . . . . . . . . . . . . . . . . . . . . . . . . 250
7.2.2 Python code for deriving solutions for bars . . . . . . . . 252
7.3 Example: Constant body force and concentrated force at
the right end . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
7.3.1 Method 1: Step-by-step approach integration . . . . . . . 253
7.3.2 Method 2: Use the Sympy differential
equation solver . . . . . . . . . . . . . . . . . . . . . . . . 256
7.3.3 Python code for plotting the solutions . . . . . . . . . . . 256
7.4 Example: Linear body force and concentrated
force at the right end . . . . . . . . . . . . . . . . . . . . . . . . 260
7.5 Example: A displacement imposed at the right end . . . . . . . . 261
7.6 Bars subjected to discontinuous body forces . . . . . . . . . . . . 264
7.6.1 Distributed forces . . . . . . . . . . . . . . . . . . . . . . 264
7.6.2 Point force . . . . . . . . . . . . . . . . . . . . . . . . . . 265
7.7 Statically indeterminate bars: Reaction force . . . . . . . . . . . 266
7.8 Thermal stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
7.8.1 Thermal displacement . . . . . . . . . . . . . . . . . . . . 267
7.8.2 Thermal forces . . . . . . . . . . . . . . . . . . . . . . . . 268
7.8.3 Computing thermal stresses . . . . . . . . . . . . . . . . 269
[Link] Approach 1: Use reaction and external
forces . . . . . . . . . . . . . . . . . . . . . . . 269
[Link] Approach 2: Use the combined
displacement . . . . . . . . . . . . . . . . . . . 270
xvi Mechanics of Materials: Formulations and Solutions with Python

7.9 Example: Stress in an overconstrained bar loaded mechanically

and thermally . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
7.10 Example: Thermal stress in an overconstrained bar . . . . . . . . 273
7.11 Example: A composite bar subjected to both mechanical and
thermal loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
7.11.1 Solution to (1) . . . . . . . . . . . . . . . . . . . . . . . . 276
7.11.2 Solution to (2) . . . . . . . . . . . . . . . . . . . . . . . . 278
7.11.3 Solution to (3) . . . . . . . . . . . . . . . . . . . . . . . . 280
7.11.4 Solution to (4) . . . . . . . . . . . . . . . . . . . . . . . . 281
7.12 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282

8. Solution to Bended Beams 283

8.1 Governing equation, fourth-order differential equation . . . . . . 284
8.1.1 Thin beam theory . . . . . . . . . . . . . . . . . . . . . . 284
8.1.2 Normal stress and the resulted moment . . . . . . . . . . 285
8.1.3 On second moment of area of beam cross-section . . . . . 288
8.1.4 Equilibrium equation for beams . . . . . . . . . . . . . . 289
8.2 Boundary conditions for beams . . . . . . . . . . . . . . . . . . . 291
8.2.1 Possible BCs for beams . . . . . . . . . . . . . . . . . . . 291
8.2.2 Admissible BCs for beams . . . . . . . . . . . . . . . . . 292
8.3 Example: A cantilever beam subjected to a force at its
right-end . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
8.4 A Python beam solver . . . . . . . . . . . . . . . . . . . . . . . . 294
8.4.1 Python code of beam solver . . . . . . . . . . . . . . . . 295
8.4.2 Example: Clamped-clamped beams . . . . . . . . . . . . 297
8.4.3 Use of the Sympy differential equation solver . . . . . . . 299
8.5 Python code for finding the maximum and minimum values . . . 300
8.6 Python code for plotting the solution marked with
extreme values . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
8.7 Solutions at specific locations . . . . . . . . . . . . . . . . . . . . 303
8.8 Example: Simple-simple supported beams . . . . . . . . . . . . . 304
8.9 Example: Simple-clamped beams . . . . . . . . . . . . . . . . . . 310
8.10 Example: Free-clamp supported beams . . . . . . . . . . . . . . 312
8.11 Example: Clamped-clamped beam subjected to a point force at
the middle span . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
8.12 Plot piecewise solutions . . . . . . . . . . . . . . . . . . . . . . . 319
8.13 Example: Simply supported beam with two spans . . . . . . . . 321
8.13.1 For span-1 . . . . . . . . . . . . . . . . . . . . . . . . . . 322
8.13.2 For span-2 . . . . . . . . . . . . . . . . . . . . . . . . . . 323
8.13.3 Continuity and equilibrium conditions . . . . . . . . . . . 324
8.13.4 Final solutions . . . . . . . . . . . . . . . . . . . . . . . . 325
 Contents xvii

8.14 Plot piecewise solutions . . . . . . . . . . . . . . . . . . . . . . . 326

8.15 Example: A symmetric frame subjected to a uniformly
distributed load . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
8.15.1 For member-1 . . . . . . . . . . . . . . . . . . . . . . . . 330
8.15.2 For member-2 . . . . . . . . . . . . . . . . . . . . . . . . 331
8.15.3 Continuity and equilibrium conditions at point B . . . . 332
8.15.4 Final solutions . . . . . . . . . . . . . . . . . . . . . . . . 333
8.16 Plot piecewise solutions . . . . . . . . . . . . . . . . . . . . . . . 335
8.17 Thermal stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
8.17.1 Setting of the problem . . . . . . . . . . . . . . . . . . . 337
8.17.2 Governing equation for deflection caused by the
temperature gradient . . . . . . . . . . . . . . . . . . . . 338
8.17.3 Solution of deflection resulting from temperature
gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
8.17.4 Solutions of indeterminate beams in temperature
gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
8.17.5 Example: Simple-clamped beams subjected to
temperature gradient . . . . . . . . . . . . . . . . . . . . 340
[Link] Find the general displacement solution
for an s-c beam . . . . . . . . . . . . . . . . . . 341
[Link] Add in the displacement caused by
temperature gradient . . . . . . . . . . . . . . 341
[Link] Find the combined displacement using BCs . . 341
[Link] Plot the distribution of solutions . . . . . . . . 342
8.17.6 Example: Beams simple-simple supported subjected to
temperature gradient . . . . . . . . . . . . . . . . . . . . 344
[Link] Find the general displacement solution for
an s-s beam . . . . . . . . . . . . . . . . . . . . 344
[Link] Add in the displacement caused by
temperature gradient . . . . . . . . . . . . . . 344
[Link] Find the combined displacement . . . . . . . . 344
[Link] Plot the distribution of solutions . . . . . . . . 345
8.18 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347

9. Torsion of Bars 349

9.1 Bars with circular cross-section . . . . . . . . . . . . . . . . . . . 351
9.1.1 Problem description . . . . . . . . . . . . . . . . . . . . . 351
9.1.2 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . 351
9.1.3 Displacements on the cross-section . . . . . . . . . . . . . 352
9.1.4 Shear strains and stresses on the cross-section . . . . . . 353
xviii Mechanics of Materials: Formulations and Solutions with Python

[Link] Divergence of the stress states in a

circular bar . . . . . . . . . . . . . . . . . . . . 355
[Link] Curl of the stress vector field in a
circular bar . . . . . . . . . . . . . . . . . . . . 355
9.1.5 Distribution of shear stress on the cross-section . . . . . 356
9.1.6 Boundary condition on the lateral surface of the bar . . . 357
9.1.7 Equilibrium on the cross-section, twist-torque
relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
9.1.8 On polar second moment of area . . . . . . . . . . . . . . 359
9.1.9 Twist-angle and torque relations . . . . . . . . . . . . . . 360
9.1.10 Continuous formulation . . . . . . . . . . . . . . . . . . . 361
9.1.11 Bars with hollow cylindrical cross-section . . . . . . . . . 361
9.1.12 Bars with thin-wall cross-sections . . . . . . . . . . . . . 362
9.1.13 Example: A shaft with hollow circular cross-section
subject to a torque . . . . . . . . . . . . . . . . . . . . . 363
9.1.14 Example: A shaft with two segments subject to
two torques . . . . . . . . . . . . . . . . . . . . . . . . . 365
9.2 Bars with enclosed thin-wall cross-sections . . . . . . . . . . . . 368
9.2.1 Concept of shear flow . . . . . . . . . . . . . . . . . . . . 369
9.2.2 Torque and shear-flow relation . . . . . . . . . . . . . . . 370
9.2.3 Twist-angle and torque relation . . . . . . . . . . . . . . 371
9.2.4 Example: Shear stress in a thin wall bar with box
cross-section . . . . . . . . . . . . . . . . . . . . . . . . . 372
9.2.5 A comparison of exact and thin-wall formulas . . . . . . 375
9.3 Bars with non-circular cross-sections . . . . . . . . . . . . . . . . 375
9.3.1 Setting of the problem . . . . . . . . . . . . . . . . . . . 375
9.3.2 Assumed displacements . . . . . . . . . . . . . . . . . . . 376
9.3.3 Shear strains in non-circular bars . . . . . . . . . . . . . 377
9.3.4 Equilibrium equations at any point in torsion bar and
stress function . . . . . . . . . . . . . . . . . . . . . . . . 378
9.3.5 Boundary conditions for the stress function . . . . . . . . 379
9.3.6 Shear stress on the cross-section by the boundary . . . . 381
9.3.7 Equilibrium conditions for the stress function . . . . . . 382
9.3.8 Linear elastic solution . . . . . . . . . . . . . . . . . . . . 383
9.3.9 Techniques for constructing the stress function . . . . . . 384
9.3.10 Example: Bars with elliptical cross-section . . . . . . . . 384
9.3.11 Comparison of torsional stiffness between elliptic and
circular bars . . . . . . . . . . . . . . . . . . . . . . . . . 386
9.3.12 Example: Bars with equilateral triangle cross-section . . 386
9.4 Prandtl’s elastic membrane analogy for torsion . . . . . . . . . . 387
9.4.1 General formulation . . . . . . . . . . . . . . . . . . . . . 387
9.4.2 A bar with narrow rectangular cross-section . . . . . . . 389
 Contents xix

9.4.3 Bars with cross-sections consisting of narrow

rectangles . . . . . . . . . . . . . . . . . . . . . . . . . . 392
9.5 Bars with rectangular cross-section . . . . . . . . . . . . . . . . . 393
9.5.1 Strategy for solution . . . . . . . . . . . . . . . . . . . . 393
9.5.2 Method of separation of variables . . . . . . . . . . . . . 394
9.5.3 Final solution of the stress function . . . . . . . . . . . . 396
9.5.4 Shear stress solution . . . . . . . . . . . . . . . . . . . . . 396
9.5.5 Torque twist-angle solution and equivalent polar
2nd moment of area . . . . . . . . . . . . . . . . . . . . . 396
9.5.6 The maximum shear stress . . . . . . . . . . . . . . . . . 397
9.5.7 Python code for computing k1 . . . . . . . . . . . . . . . 397
[Link] An example . . . . . . . . . . . . . . . . . . . . 398
[Link] Cut-off terms in the solution series . . . . . . . 398
[Link] Values as function of b/h . . . . . . . . . . . . 399
9.5.8 Comparison of equivalent J and polar moment of
area Jz . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
9.5.9 Comparison of torsional stiffness between rectangular
and square bars . . . . . . . . . . . . . . . . . . . . . . . 401
9.5.10 Comparison of torsional stiffness between circular and
square bars . . . . . . . . . . . . . . . . . . . . . . . . . . 403
9.5.11 Code to compute the stress function on rectangular
cross-section . . . . . . . . . . . . . . . . . . . . . . . . . 403
9.5.12 Plot of the stress function on rectangular cross-section . 404
9.5.13 Code to compute the shear stresses on rectangular cross-
section . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
9.5.14 Plot of the shear stresses on rectangular cross-section . . 405
9.5.15 Example: An I-beam . . . . . . . . . . . . . . . . . . . . 407
9.6 Hollow thin-wall torsion members, use of membrane analogy . . 408
9.6.1 Solution strategy . . . . . . . . . . . . . . . . . . . . . . 408
9.6.2 Formula for twist angle . . . . . . . . . . . . . . . . . . . 410
9.6.3 Torque stress relation . . . . . . . . . . . . . . . . . . . . 411
9.6.4 Equivalent polar 2nd moment of area . . . . . . . . . . . 411
9.6.5 Example: Bars with thin-wall circular cross-sections . . . 411
9.7 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414

10. Property of Areas 415

10.1 Area of areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
10.1.1 Integral formulas . . . . . . . . . . . . . . . . . . . . . . 415
10.1.2 Area of a composite area . . . . . . . . . . . . . . . . . . 417
10.2 First moment of area . . . . . . . . . . . . . . . . . . . . . . . . 418
10.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 418
xx Mechanics of Materials: Formulations and Solutions with Python

10.2.2 First moment of a composite area . . . . . . . . . . . . . 419

10.2.3 Example: The first moment of area of a rectangle . . . . 419
10.2.4 Centroid of an area . . . . . . . . . . . . . . . . . . . . . 421
10.2.5 Centroid for composite areas . . . . . . . . . . . . . . . . 422
10.2.6 Example: Centroid of an I-beam . . . . . . . . . . . . . . 422
10.2.7 Example: Compute the first moment of an area,
using centroid . . . . . . . . . . . . . . . . . . . . . . . . 424
10.2.8 First moment of area, coordinate rotation . . . . . . . . . 425
10.3 Second moment of area . . . . . . . . . . . . . . . . . . . . . . . 426
10.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 426
10.3.2 Second moment of area for regular areas . . . . . . . . . 426
10.3.3 Second moment of area, coordinate translation . . . . . . 428
10.3.4 Second moments of composite areas . . . . . . . . . . . . 429
10.3.5 Example: Second moments of an I-beam . . . . . . . . . 429
10.3.6 Example: Second moments of an S-beam . . . . . . . . . 431
10.3.7 Second moment of area, coordinate rotation . . . . . . . 433
10.3.8 Principal second moments of area . . . . . . . . . . . . . 434
10.3.9 Example: Second moments of area after coordinate
rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434
10.3.10 Example: The principal second moments of area . . . . . 435
10.3.11 Use of eigenvalue solver . . . . . . . . . . . . . . . . . . . 435
10.4 Polar moment of area . . . . . . . . . . . . . . . . . . . . . . . . 438
10.4.1 Perpendicular axis theorem . . . . . . . . . . . . . . . . . 438
10.4.2 Proof of the perpendicular axis theorem . . . . . . . . . . 438
10.4.3 Other useful cases . . . . . . . . . . . . . . . . . . . . . . 438
10.4.4 Equivalent polar second moment of area . . . . . . . . . 439
10.5 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440

11. Normal Stress by Moments 441

11.1 Loading plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
11.2 Setting of the problem . . . . . . . . . . . . . . . . . . . . . . . . 443
11.3 Symmetrical bending . . . . . . . . . . . . . . . . . . . . . . . . 443
11.3.1 Superposition of stresses . . . . . . . . . . . . . . . . . . 443
11.3.2 Neutral axis of the combined normal stress . . . . . . . . 445
11.3.3 Example: A cantilever beam subjected to a force
and a moment . . . . . . . . . . . . . . . . . . . . . . . . 445
11.3.4 Use gr.solver1D4() to find the moment . . . . . . . . . . 447
11.3.5 Example: A cantilever T-beam subjected to a force
and a moment . . . . . . . . . . . . . . . . . . . . . . . . 449
11.4 Beams under non-symmetrical bending . . . . . . . . . . . . . . 452
 Contents xxi

11.4.1 General formula for computing normal stress

distribution . . . . . . . . . . . . . . . . . . . . . . . . . 452
11.4.2 Neutral axis: general formula . . . . . . . . . . . . . . . . 454
11.4.3 Example: A cantilever L-beam subjected to a force
and a moment . . . . . . . . . . . . . . . . . . . . . . . . 455
[Link] Solution to 1 and 2 . . . . . . . . . . . . . . . . 456
[Link] Solution to 3: Determine the maximum stress
of the beam, using the formulas for non-
symmetric bending . . . . . . . . . . . . . . . . 456
[Link] Solution to 4: Find the principal axis of
the cross-section . . . . . . . . . . . . . . . . . 457
11.5 Effects of error in loading plane angle . . . . . . . . . . . . . . . 459
11.5.1 Example: A tall beam with rectangular cross-section . . 459
[Link] Python code for the solution . . . . . . . . . . 460
[Link] An important finding . . . . . . . . . . . . . . 461
11.5.2 Underlying reasons for error magnification . . . . . . . . 461
11.6 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463

12. Shear Stress by Shear Force 465

12.1 Setting of the problem . . . . . . . . . . . . . . . . . . . . . . . . 466
12.2 General formulation . . . . . . . . . . . . . . . . . . . . . . . . . 467
12.2.1 Free-body diagram . . . . . . . . . . . . . . . . . . . . . 467
12.2.2 Equilibrium conditions . . . . . . . . . . . . . . . . . . . 468
12.2.3 Formula for shear stress . . . . . . . . . . . . . . . . . . . 468
12.2.4 Shear-flow resulted from shear force . . . . . . . . . . . . 469
12.3 Shear stress on rectangular cross-section . . . . . . . . . . . . . . 469
12.3.1 Parabolic distribution . . . . . . . . . . . . . . . . . . . . 469
12.3.2 Alternative approach . . . . . . . . . . . . . . . . . . . . 471
12.4 Shear stress on thin-wall cross-sections . . . . . . . . . . . . . . 472
12.4.1 Setting of the problem . . . . . . . . . . . . . . . . . . . 472
12.4.2 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . 473
12.4.3 Free-body diagram and formulation . . . . . . . . . . . . 473
12.4.4 Shear-flow on thin wall cross-sections . . . . . . . . . . . 474
12.5 Example: Shear stress on the cross-section of a C-beam . . . . . 474
12.5.1 Shear-flow and stress at any point in the flange . . . . . 475
12.5.2 Shear stress at any point s in the web . . . . . . . . . . . 476
12.6 Example: Shear stress on the cross-section of an L-beam . . . . . 477
12.7 Shear stress on an I-beam cross-section . . . . . . . . . . . . . . 479
12.7.1 Problem analysis . . . . . . . . . . . . . . . . . . . . . . 479
12.7.2 Calculation procedure . . . . . . . . . . . . . . . . . . . . 480
xxii Mechanics of Materials: Formulations and Solutions with Python

[Link] Shear-flow and stress at any point in

the flange . . . . . . . . . . . . . . . . . . . . . 480
[Link] Shear-flow and stress at point D in
the flange . . . . . . . . . . . . . . . . . . . . . 480
[Link] Shear stress at point D in the web . . . . . . . 481
[Link] Shear stress at any point s in the web . . . . . 481
[Link] Shear stress at D on the top of the bottom
flange . . . . . . . . . . . . . . . . . . . . . . . 482
12.7.3 Example: Python code for computing the shear-flow and
stress in an I-beam . . . . . . . . . . . . . . . . . . . . . 483
[Link] Data preparation . . . . . . . . . . . . . . . . . 483
[Link] Compute shear-flow and stress at D in the top
flange . . . . . . . . . . . . . . . . . . . . . . . 484
[Link] Compute the shear-flow and stress at D on the
top of the web . . . . . . . . . . . . . . . . . . 484
[Link] Compute shear-flow and stress at any s
in web . . . . . . . . . . . . . . . . . . . . . . . 484
[Link] Compute shear stress at D at the top of the
bottom flange . . . . . . . . . . . . . . . . . . . 485
12.7.4 Plot of shear-flow in the web . . . . . . . . . . . . . . . . 485
12.7.5 Compute forces in channels . . . . . . . . . . . . . . . . . 486
12.8 Shear-center of a cross-section . . . . . . . . . . . . . . . . . . . 489
12.8.1 Setting of the problem . . . . . . . . . . . . . . . . . . . 489
12.8.2 Strategy to find the shear-center . . . . . . . . . . . . . . 490
12.8.3 Procedure to find the shear-center . . . . . . . . . . . . . 491
12.8.4 Example: Find the formula to compute e for a
C-beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491
12.8.5 Example: Python code for computing the shear-center
location for a box-beam . . . . . . . . . . . . . . . . . . . 492
[Link] Dividing the box to two open cross-sections . . 493
[Link] Splitting the shear force . . . . . . . . . . . . . 494
[Link] General purpose code for rectangular box
sections . . . . . . . . . . . . . . . . . . . . . . 495
[Link] Data preparation . . . . . . . . . . . . . . . . . 498
[Link] Distribution profile of the shear-flow . . . . . . 498
12.8.6 Example: The shear-center location for a C-beam . . . . 499
12.9 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500

13. Energy Methods 501

13.1 Energy principles and fundamental concepts . . . . . . . . . . . 501
13.2 Principle of virtual work by virtual displacement . . . . . . . . . 502
 Contents xxiii

13.2.1 Example: A loaded linear elastic spring . . . . . . . . . . 503

13.2.2 Evaluation of potential energy . . . . . . . . . . . . . . . 503
13.2.3 Imposition of infinitely small virtual displacement . . . . 505
13.2.4 Compatibility condition . . . . . . . . . . . . . . . . . . . 505
13.2.5 Potential energy principle for 1-DOF systems . . . . . . 506
13.2.6 Results of potential energy principle: Equilibrium
equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 506
13.2.7 On compatibility equation . . . . . . . . . . . . . . . . . 507
13.2.8 Clapeyron’s theorem . . . . . . . . . . . . . . . . . . . . 507
13.2.9 Formulas of potential energy for 1-DOF systems . . . . . 507
13.2.10 Formulas of potential energy for various members . . . . 509
13.2.11 Strain energies UN for bars with varying cross-section . . 509
13.2.12 Strain energies for beams with varying cross-section . . . 510
[Link] From moment, UM . . . . . . . . . . . . . . . . 510
[Link] From shear force, UV . . . . . . . . . . . . . . 510
13.2.13 Strain energies UT for bars under torsion . . . . . . . . . 511
13.2.14 Summary of formulas for U for various structural
members . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
13.3 Principle of virtual work by virtual force . . . . . . . . . . . . . 512
13.3.1 Application of an infinitely small virtual force . . . . . . 513
13.3.2 Evaluation of complementary energy . . . . . . . . . . . 513
13.3.3 Equilibrium condition . . . . . . . . . . . . . . . . . . . . 515
13.3.4 Complementary energy principle for 1-DOF systems . . . 515
13.3.5 Compatibility equation . . . . . . . . . . . . . . . . . . . 516
13.3.6 On equilibrium equation . . . . . . . . . . . . . . . . . . 516
13.3.7 General formulas for displacements of 1-DOF
members . . . . . . . . . . . . . . . . . . . . . . . . . . . 516
13.3.8 Example: Cantilever beam subjected to a concentrated
force at its tip . . . . . . . . . . . . . . . . . . . . . . . . 517
13.3.9 Comparison of PEP and CEP, linear elastic systems . . . 521
13.3.10 Comparison of PEP and CEP, nonlinear systems . . . . . 522
13.4 Strain energy formulation for 3D solids . . . . . . . . . . . . . . 522
13.4.1 Mathematical derivation . . . . . . . . . . . . . . . . . . 522
[Link] Work done by external forces . . . . . . . . . . 523
[Link] Strain energy expression . . . . . . . . . . . . . 524
[Link] Compatibility conditions . . . . . . . . . . . . 525
13.4.2 Pictorial derivation . . . . . . . . . . . . . . . . . . . . . 525
13.4.3 Strain energy density for 1D problem . . . . . . . . . . . 525
[Link] General formulas . . . . . . . . . . . . . . . . . 526
[Link] Formulas for linear elastic materials . . . . . . 528
13.4.4 Various formulas for strain energy density for
3D solids . . . . . . . . . . . . . . . . . . . . . . . . . . . 529
xxiv Mechanics of Materials: Formulations and Solutions with Python

13.5 Potential energy principle for structures . . . . . . . . . . . . . . 530

13.5.1 Virtual work via virtual displacement for structures . . . 530
13.5.2 Formulas of potential energy principle for structures . . . 530
13.5.3 Example: A two-bar truss subjected to two loads . . . . 531
[Link] Setting of the problem . . . . . . . . . . . . . . 531
[Link] Solutions to the problem . . . . . . . . . . . . 532
13.6 Complementary energy formulation for 3D solids . . . . . . . . . 537
13.6.1 General formulations . . . . . . . . . . . . . . . . . . . . 537
13.6.2 Equilibrium conditions . . . . . . . . . . . . . . . . . . . 537
13.6.3 Complementary energy 1D problems . . . . . . . . . . . 538
13.7 Complementary energy principle for structures . . . . . . . . . . 539
13.7.1 Virtual work via virtual forces for structures . . . . . . . 539
13.7.2 Formulas of complementary energy principle for
structures . . . . . . . . . . . . . . . . . . . . . . . . . . 539
13.7.3 Example: Linear spring–mass system . . . . . . . . . . . 540
13.7.4 Example: Nonlinear spring–mass system . . . . . . . . . 543
13.7.5 Example: A two-bar truss subjected to two loads . . . . 544
[Link] Setting of the problem . . . . . . . . . . . . . . 544
[Link] Solution to the problem . . . . . . . . . . . . . 544
13.8 Displacements of statically determinate structures . . . . . . . . 546
13.8.1 Principles for linear elastic structure members . . . . . . 546
13.8.2 Principles for structures with multiple multiple-type
members . . . . . . . . . . . . . . . . . . . . . . . . . . . 547
13.8.3 Examples: Cantilever beam subjected to two concen-
trated forces . . . . . . . . . . . . . . . . . . . . . . . . . 548
13.8.4 Examples: Cantilever beam subjected to
distributed force . . . . . . . . . . . . . . . . . . . . . . . 551
13.8.5 Fictitious unit load method . . . . . . . . . . . . . . . . . 553
13.8.6 Example: Cantilever beam subjected to
distributed force . . . . . . . . . . . . . . . . . . . . . . . 554
13.8.7 Example: A planar truss structure with three bars . . . . 555
13.9 Statically indeterminate structures . . . . . . . . . . . . . . . . . 557
13.9.1 Issues in dealing with indeterminate structures . . . . . . 557
13.9.2 Example: Cantilever beam with a simple support at the
other end . . . . . . . . . . . . . . . . . . . . . . . . . . . 558
13.10 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 559
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 559

14. Failure Criteria and Failure Analysis 561

14.1 Basic concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562
14.2 A case for this study: Yield analysis of a steel shaft . . . . . . . 562
14.2.1 Setting of the problem . . . . . . . . . . . . . . . . . . . 562
 Contents xxv

14.2.2 Formulas for computing the stresses . . . . . . . . . . . . 562

14.2.3 Python code to compute the stresses . . . . . . . . . . . 564
14.3 Maximum principal stress criterion . . . . . . . . . . . . . . . . . 565
14.3.1 Yield surface . . . . . . . . . . . . . . . . . . . . . . . . . 566
14.3.2 Application to the case study problem . . . . . . . . . . 567
14.4 Maximum principal strain criterion . . . . . . . . . . . . . . . . 567
14.4.1 Yield Surface . . . . . . . . . . . . . . . . . . . . . . . . . 568
14.4.2 Application to the case study problem . . . . . . . . . . 568
14.5 Strain-energy density criterion . . . . . . . . . . . . . . . . . . . 569
14.5.1 Strain-energy density . . . . . . . . . . . . . . . . . . . . 570
14.5.2 Application to the case study problem . . . . . . . . . . 570
14.6 Yielding criteria for ductile materials . . . . . . . . . . . . . . . 571
14.6.1 Maximum shear-stress (Tresca) criterion . . . . . . . . . 571
14.6.2 Tresca yield surface . . . . . . . . . . . . . . . . . . . . . 571
14.6.3 Application to the case study problem . . . . . . . . . . 572
14.6.4 Distortional energy density criteria . . . . . . . . . . . . 573
[Link] von Mises criterion . . . . . . . . . . . . . . . . 573
[Link] Application to the case study problem . . . . . 574
[Link] Octahedral shear-stress criterion . . . . . . . . 574
[Link] Application to the case study problem . . . . . 575
[Link] Difference between the von Mises and Tresca
criteria . . . . . . . . . . . . . . . . . . . . . . 575
14.6.5 π-plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . 576
14.6.6 Deviatoric planes . . . . . . . . . . . . . . . . . . . . . . 576
14.7 Yield criterion for cohesive materials . . . . . . . . . . . . . . . . 577
14.7.1 Mohr–Coulomb yield function . . . . . . . . . . . . . . . 577
14.7.2 Determination of material parameters . . . . . . . . . . . 577
14.7.3 Mohr–Coulomb yield surface . . . . . . . . . . . . . . . . 578
14.7.4 Drucker–Prager yield criterion . . . . . . . . . . . . . . . 578
14.8 Comparison of different criteria . . . . . . . . . . . . . . . . . . . 579
14.9 Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 582

Index 585
This page intentionally left blank
 Chapter 1

Introduction

1.1 Computational methods

This book is part of a series that delves into the exciting world of compu-
tational methods. Our goal is to equip you with a solid foundation in this
field, combining theoretical principles with practical applications.
The series encompasses general methods and techniques used throughout
science, technology, engineering, and math (STEM) education, as well as
research in various scientific and engineering disciplines. Think of it as your
one-stop encyclopedia for computational methods, covering the theory, for-
mulation, and even the code you’ll need to get started. We’ve designed the
series to be accessible. Readers who have completed elementary and middle
school (roughly nine years of education) can begin with the foundational vol-
umes on mathematics. Once you progress through the mechanics volumes,
you’ll be equipped to tackle research and design projects that leverage com-
putational methods.
Taking advantage of the rapid advancements in computer technology,
we’ll heavily integrate code examples throughout the series. This allows you
to see concepts, theories, and formulations come to life through practical
examples with clear visualizations. Initially, we’ll primarily use Python for
the fundamental volumes, potentially introducing other languages for more
advanced topics. By effectively utilizing code, you can dedicate more time
to understanding the core principles. Let the computer handle the complex
calculations and tedious derivations, allowing you to focus on the bigger
picture.
This approach empowers you to explore the fascinating world of compu-
tational methods with a strong theoretical grounding and practical skills.

1
2 Mechanics of Materials: Formulations and Solutions with Python

1.2 Why start and contribute to this book series

The editor-in-chief of this book series, Dr. G.R. Liu, has been working in
areas related to computational methods for over 40 years. He developed
his first FEM code for nonlinear problems in 1980 and has since published
more than 600 journal papers and 12 monographs in this area. After all
these years of studying, using, and developing computational methods, he
started to think about a means to help other interested individuals learn
computational methods in a more effective, systematic, and smoother way.
He has concluded that developing this book series is the best way to achieve
this objective.

1.3 Mechanics of materials: Essential for safe and

effective use of materials

Materials are used to build devices and structures for various purposes in
our everyday life. This book introduces the subject of mechanics of mate-
rials, which is the most basic and essential for safe and effective use of
materials in the fields of sciences and engineering. When a device (or struc-
ture) is loaded, the members in it experience force, stress, and strain, and
they undergo deformation. If the device is not designed properly, it may
either fail to carry the desired load or become too heavy with material not
fully utilized. The subject that studies the relationships among the forces,
stresses, strains, and deformation of the basic structural members
of materials is called mechanics of materials.
This book covers the fundamental principles of mechanics of materials,
focusing on the mechanical behavior of structural members under various
types of loads, including axial loading, bending, shearing, and torsion.
The members can have various shapes and can be constrained in different
ways. Concepts of energy and failure criteria will also be studied.
This book introduces the tensor concept at the beginning because it is
the most fundamental concept for all subjects in mechanics. Traditionally,
the tensor concept is introduced at a more advanced course, typically in
elasticity because of its difficulty level. Based on the author’s experience,
however, it is far better to teach this concept in the introductory course on
mechanics. Students may not be able to fully comprehend all the details,
but with the help of Python codes, the basic concepts can be reasonably
well set. Given time, these concepts will be fully established, leading to a
rigorous and solid foundation for future studies/courses. The principles of
mechanics of materials are explained in detail, formulated, and demonstrated
with numerous examples and illustrations. Python is intensively used
 Introduction 3

for deriving formulations, finding solutions symbolically or numerically, and

plotting the results graphically.
Proper connection between theory and real-world engineering prob-
lems will be made through these examples.
The materials of this book can be digested through different forms of the
learning process, including classroom teaching, online courses, and even
self-study. Since Python codes are provided, readers can easily see how the
theory is formulated and how the solutions are obtained in terms of formulas,
numerical values, and graphs. Readers may also deepen their understanding
via playing with codes and even further develop their own codes for solving
other related problems.
The book is written in Jupyter Notebook format, so that the descrip-
tion of theory, formulation, and coding can all be done in a unified document.
This provides an environment for easy reading, exercise, practice, and further
exploration.
This book is written in reference to Refs. [1–3], which were the textbooks
when the author was a university student, Refs. [4,5], which were frequently
used as reference books during his PhD and research work, and Refs. [6–9],
which were used as textbooks and reference books when teaching at uni-
versities. Some example problems given in these textbooks are used in this
book.

1.4 Who may read this book?

This book is written for beginners interested in learning computational meth-

ods for solving problems in nature, engineering, and science. Readers include
high school students, university students, graduate students, researchers, and
professionals in any discipline. Engineers and practitioners may also find the
book useful in establishing systematic basic concepts in computational meth-
ods. For beginners, it may be a good idea to first read or skim over Volume
1 of this book series: Numbers and Functions — Theory, Application, and
Python Codes and Volume 2: Calculus: Formulations and Solutions with
Python.
This volume contains a substantial amount of advanced concepts in the
mechanics of materials, which are marked with * or **. This provides better
completeness of the topic. When this book is used as a textbook for an
entry-level course on mechanics of materials, these materials may be treated
as optional. In the past 40 years of teaching, the author often found that
there would always be quite a number of students having more time and
capable of learning more materials than the curriculum defines. Therefore,
providing additional starred materials in the book can be beneficial to many
4 Mechanics of Materials: Formulations and Solutions with Python

students. When this book is used as a textbook for an advanced course on

mechanics of materials, these starred materials may be directly used. More
advanced materials will be covered in the next planned volume of this book
series: Solid Mechanics — Cartesian Coordinates.

1.5 Codes used in this book

Readers who purchased the book may contact the author directly at
liugr100@[Link] ([Link] to request a soft copy of
the book in Jupyter Notebook format with codes (which may be updated)
for free for academic use after registration. If the book is used as a text-
book, instructors may contact the author directly for sharing some of the
homework assignments and examination papers with codes.
The conditions for use of the book and codes developed by the author,
in both hard and soft copies, are as follows:

1. Users are entirely at their own risk when using any part of the codes
and techniques. The codes are written primarily for proof of concept and
not necessarily for efficiency and robustness. Many of the codes are not
thoroughly tested by third parties.
2. The book and codes are only for your own use. You are not allowed to
further distribute them without permission from the author.
3. There will not be any user support.
4. Proper reference and acknowledgment must be given for the use of the
book, codes, ideas, and techniques.

These codes are often run with various external packages/modules.

Therefore, care is needed when using these codes because the behavior of
the codes often depends on the versions of Python and all these pack-
ages/modules. When the code does not run as expected, version mismatch
could be one of the problems. When this book was written, the versions of
Python and some of the packages/modules were as follows:

• Python 3.9.16 :: Anaconda, Inc.,

• Jupyter Notebook 6.1.5.

When issues are encountered while running a code, readers may need
to check the versions of the packages/modules used. If Anaconda Navigator
is used, the versions of all these packages/modules installed in the Python
environment are listed when the Python environment is highlighted. You can
also check the version of a package in a code cell of the Jupyter Notebook.
For example, to check the version of the current environment of Python, one
may use the following:
 Introduction 5

1 !python -V #! is used to execute an external command

Python 3.9.16

1 !jupyter notebook --version

6.1.5

If the version is indeed an issue, one would need to either modify the
code to fit the version or install the correct version on your system. It
is very useful to query the web using the error message, and solutions or
leads can often be found. Online AI tools, such as ChatGPT, Gemini, and
Copilot can also be quite helpful. This is the approach that the author uses
most of the time facing issues in running a code.
This book will not discuss how to use Python. There is plenty of litera-
ture openly available online. Interested readers may also refer to Chapters 2
and 3 in Ref. [10] for a concise description of using Python for scientific
computations. Since Python is quite easy to learn, one may simply use the
codes and examples given in this book and jump-start learning Python in
the process of studying the technical subjects of this book.

1.6 Use of external modules or dependences

To use Python and codes provided in this volume, we import the necessary
modules and functions. The following are the most essential ones:

1 import sys # import "sys" module

2
3 [Link]('../grbin') # Relative directory.
4 # Or absolute folder like 'F:\\xxx\\...\\code'
5
6 #Author's own grcodes module is placed in folder grbin
7 import grcodes as gr

To view the codes in the imported module, one may just use the following
code by uncommenting it (which may produce a long output):

1 import inspect
2 #source_code = [Link](gr) # to view everything in gr module
3 source_code = [Link](gr.cheby_T) # to view any function in gr
4 print(source_code)
6 Mechanics of Materials: Formulations and Solutions with Python

def cheby_T(n, x):

'''Generate the first kind Chebyshev polynomials of degree (n-1) '''
if n == 0: return [Link]
elif n == 1: return x
else: return (2*x*cheby_T(n-1,x)-cheby_T(n-2,x)).expand()

Alternatively, one may use any text editor to view and change the
codes. To use any code function, say printx() in the imported module gr, for
example, use [Link]. Following is an example:

1 x = 8
2 [Link]('x') # when gr. is used, the code function is from the grcodes

x = 8

To avoid frequent import of lengthy external modules, we put all the

frequently used modules in the “[Link]” file, as follows:

1 from __future__ import print_function

2 import numpy as np # for numerical computation
3 import sympy as sp # sympy module for computation
4 import [Link] as lg # numpy linear algebra module
5 import [Link] as sg # scipy linear algebra module
6 import [Link] as si
7 from [Link] import ortho_group
8 import importlib
9 import itertools
10 import inspect
11 import csv
12 import pandas as pd
13
14 from grcodes import drawArrow, plotfig, printM, printx # frequently used
15
16 import math as ma
17 from sympy import sin, cos, symbols, lambdify, init_printing
18 from sympy import pi, Matrix, sqrt, oo, integrate, diff, Derivative
19 from sympy import MatrixSymbol, simplify, nsimplify, Function
20 from sympy import factor, expand, nsimplify, Matrix, ordered, hessian
21 from [Link] import plot as splt
22 init_printing(use_unicode=True) # for latex-like quality print format
23
24 from [Link] import MultipleLocator
25 import [Link] as plt # for plotting figures
26 import matplotlib as mpl

In the beginning of each Jupyter Notebook (Chapter), we simply import

everything in commonImports. This reduces the need to frequently import
modules in the coding process:
 Introduction 7

1 #Often used external modules are in commonImports placed in folder grbin

2 #Place cursor in this cell, and press Ctrl+Enter to import dependences.
3
4 import sys # for accessing the computer system
5 [Link]('../grbin/') # Change to the directory in your system
6
7 from commonImports import * # Import dependences from '../grbin/'
8 import grcodes as gr # Import the module of the author
9 [Link](gr) # When grcodes is modified, reload it
10
11 from continuum_mechanics import vector
12 from continuum_mechanics.solids import sym_grad, strain_stress
13 init_printing(use_unicode=True) # For latex-like quality printing
14
15 # Digits in print-outs
16 np.set_printoptions(precision=4,suppress=True,
17 formatter={'float_kind': '{:.4e}'.format})

In general, importing the same module multiple times does no harm. In

fact, Python ignores all the later importations if the module is already in
the cache. Due to this, if one makes changes to the imported module, the
module needs to be reloaded for the modification to take effect. This is done
using [Link](). For example:

1 # grcodes was imported as gr, reload it when grcodes is modified

2 [Link](gr);

1.7 Use of help()

To find more details on what a module or an object in the module does, use
help() after importing. For example:

1 help(gr.solver1D4)

Help on function solver1D4 in module grcodes:

solver1D4(E, I, by, l, v0, θ0, vl, θl, V0, M0, Vl, Ml, key='c-c')
Solves the Beam Equation for integrable distributed body force:
u,x4=-by(x)/EI, with various boundary conditions (BCs):
s-s, c-c, s-c, c-s, f-c, c-f.
Input: EI: bending stiffness factor; by, body force; l, the length
of the beam; v0, 0, V0, M0, deflection, rotation,
(internal) shear force, (internal) moment at x=0;
vl, l, Vl, Ml, are those at x=l.
Return: u, v_x, v_x2, v_x3, v_x4 up to 4th derivatives of v
8 Mechanics of Materials: Formulations and Solutions with Python

Note that in the code cells given in the book, necessary comments (start-
ing with “#”) are used to provide additional explanations. These comments
are placed on the right-hand side of the cells to avoid disrupting the reading
flow too much while still being available for any help when needed.

References
[1] J.M. Gere and S.P. Timoshenko, Mechanics of Materials, Van Nostrand Reinhold
Company, New York, 1972.
[2] S. Timoshenko and J.N. Goodier, Theory of Elasticity. McGraw Hill, New
York, 1970. [Link]
elasticity&ei=ICiKSsr3G4jwkQS bxMyPCg.
[3] Z.L. Xu, Elasticity, Vols. 1 & 2. People’s Publisher, China, 1979.
[4] J.D. Achenbach, Wave Propagation in Elastic Solids, 1984.
[5] Y.C. Fung, Foundations of Solid Mechanics, 1968.
[6] A.P. Boresi, K. Chong and J.D. Lee, Elasticity in Engineering Mechanics. John
Wiley & Sons, New York, 2010.
[7] F.P. Beer, Statics and Mechanics of Materials, 2011.
[8] A.P. Boresi, R.J. Schmidt and O.M. Sidebottom, Advanced Mechanics of Materials.
John Wiley & Sons, New York, 1985.
[9] T.H.G. Megson, Structural and Stress Analysis, 1996 ([Link]
books?id=qZykmEz4nzkC).
[10] G.R. Liu, Machine Learning with Python: Theory and Applications. World Scientific,
New York, 2023.
 Chapter 2

Preliminaries: Number, Vector,

Tensor, and Functions

1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.

2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 init_printing(use_unicode=True) # For latex-like quality printing
11 np.set_printoptions(precision=4,suppress=True) # Digits in print-outs

Mechanics of materials deals with and quantifies a family of objects, includ-

ing forces, displacements, stresses, and strains. These objects are constructed
as tensors that can be studied systematically with predictable behaviors.
This chapter provides some preliminary mathematical topics related to
objects in mechanics of materials, as well as their mathematical treatment.
This includes numbers, vectors, the tensor concept, coordinate representa-
tion, coordinate transformation, and various differential operations on func-
tions. Readers familiar with these topics may skip this chapter or quickly
skim over it. Readers having difficulty following the discussions in this chap-
ter may simply skim over it for some ideas, considerations, and terminolo-
gies. He/she may read in more detail the relevant parts of this chapter when
studying the later chapters.
We start with the most elementary object: numbers, which are 0th-order
tensors and used to form higher-order tensors.

9
10 Mechanics of Materials: Formulations and Solutions with Python

2.1 Numbers
In dealing with mechanics problems in a scientific manner, we must ensure
that variables and parameters are quantified precisely. For this, we use
numbers and operate on them in a number of ways. Therefore, understand-
ing the behavior of numbers subject to various operations is fundamentally
important to ensure that what we obtain is consistent, predictable, and
reliable.
Generally speaking, a number is a single element in a field ([Link]
[Link]/wiki/Field (mathematics)) defined with the standard arith-
metic operations (addition “+”, subtraction “−”, multiplication “×”
(often omitted in expression), and defined-division “÷” (alternatively “/”
or, sometimes, “:”), all satisfy necessary conditions, including commutativ-
ity, associativity, and distributivity for addition). In all our operations in
mechanics, we require that all these conditions hold for all numbers, which
may stand alone or as a part of other objects like vectors, matrices, or in
general tensors.
A number can be a real number or a complex number. Real numbers
form the set R that is a collection of numbers of different types including
zero, all integers,√ all rational numbers, and all irrational numbers, such as
√
2, π, e, and 1+2 5 ≈ 1.618 (the golden ratio). All these numbers can be
positive or negative, filling the entire one-dimensional (1D) coordinate space
R. With addition and scalar multiplication, it is a real vector space. Figure
2.1 shows a schematic 1D line space where real numbers live. Space R is
closed under standard arithmetic operations, collectively noted as  (addi-
tion, subtraction, multiplication, or defined-division). This means that any
of these operations on any numbers in R produces a new number d:
d=ab ∀a and b∈R (2.1)
which will still be in R. This enables the most fundamental property:
closure, implying that no numbers disappear under any of these operations.
Thus, a number becomes a building block for more complicated mathemat-
ical objects, such as tensors.
Clearly, R is an ordered field because all the real numbers can be ranked
based on their values. The square of a real number is always positive.

Figure 2.1. A 1D coordinate space where real numbers live. Image adapted from Wiki-
media Commons, the free media repository, public domain, by Phrood commonswiki.
 Preliminaries: Number, Vector, Tensor, and Functions 11

In theory, R is boundless (bounded by ±∞) and seamless. In a computer,

however, it is bounded and with gaps due to bit limitations in computers.
For example, in Numpy, this information for np.float64 can be found using:

1 print("Largest float in NumPy = ", [Link](np.float64).max)

2 print("Smallest float in NumPy =", [Link](np.float64).min)
3 print("Number of significant decimal digits = ",
4 [Link](np.float64).precision)

Largest float in NumPy = 1.7976931348623157e+308

Smallest float in NumPy = -1.7976931348623157e+308
Number of significant decimal digits = 15

Real numbers include integers by default. Due to its exactness in digital

representation in computers (within its bits limit [1]), an integer is also
defined as the integer type, which is often used for indexing, ordering, and
ranking.
All complex numbers form the set C. It is, in a way, an extension
√ of
the real set R, formed by introducing the imaginary number: i = −1. Any
member in C is made up of real numbers a and b in the form of c = a+ib. Set
C is closed under the above-defined arithmetic operations, transcendental
operations, such as exponents with generally a base in C and a power
in C and their inverse operation: various logarithms. Set C is known as
algebraically closed, implying that any polynomial of degree n > 0 defined
in C has always n roots in C. Simply put: no root will disappear, ensuring
solution existence, when algebraic operations are used.
More detailed discussion on the closure properties of numbers and their
representation in a computer can be found in Ref. [1]. The takeaway is as
follows:

1. When these defined arithmetic operations are used, we should be able

to find a solution if the problem is well posed. In other words, a numer-
ical solution exists under the defined operations if the solution exists
physically.
2. When algebraic operations are involved, we may need to switch to com-
plex numbers, and the solution could be multiple. In such a case, we may
need to choose a physically meaningful one(s).
3. The most commonly encountered failure in numerical computations is
undefined division: division by zero.
4. The second most commonly encountered is under- or overflow because
of the physical limits of finite digits in a computer.

Thus, when failing to obtain a solution, one may check these related
issues. All problems posed in mechanics of materials prohibit zero division.
When zero division or under- and overflow occurs, one may need to check the
12 Mechanics of Materials: Formulations and Solutions with Python

formulations and perform dimensional analysis of equations, data scaling, or

normalization [1].

2.2 Vectors: Collection of numbers

2.2.1 Vectors defined in a computer

Generally speaking, a vector is a collection of numbers (real or complex) in
properly arranged structures so that its behavior can be managed, analyzed,
and computed in a consistent manner for desired purposes. Each of these
numbers is often called an element or component. Depending on the field
of applications, precise definition may vary. In the framework of mechanics
in general, it is built from numbers vertically or horizontally. It becomes a
building block for more complicated mathematical objects, such as matrices
and tensors.
As a convention, we use boldface to denote a vector or a matrix, which
is a collection of vectors.
From the theory of standard linear algebra, a vector can have two
forms: line-up numbers called a row-vector or a stack of numbers called
a column-vector. Row- and column-vectors behave differently when used
in mathematical derivations and operations, in which dimensionality match
is required. A vector in Python is simply a 1D array by default. The num-
ber of components in a vector is called length and can be queried using
the Python function len(). Its dimension is 1, which can be confirmed using
ndim(). Consider an array v with n numbers, or len(v) = n. It has a shape
of (n,), as shown in the following example.

1 v = [Link]([5, 2, 8]) # create an 1D array with length 3

2
3 print(f'Array:{v}, its dimension:{[Link]}, length:{len(v)} and',
4 f'shape:{[Link]}')

Array:[5 2 8], its dimension:1, length:3 and shape:(3,)

It is an ndarray class in numpy. Readers may use help() to find out the
details about it, including the attributes and methods it has.

1 # help(v) # reader may use this to find out the details

Note that a numpy array behaves differently from the row- or column-
vector in standard linear algebra, where the transpose of a row-vector
becomes a column-vector and vice versa. However, transposing a numpy
1D array has no effect, as shown in the following example.
 Preliminaries: Number, Vector, Tensor, and Functions 13

1 print(f'Array v.T:{v.T}, its dimension:{[Link]}, length:{len(v.T)}',

2 f'and shape:{[Link]}')

Array v.T:[5 2 8], its dimension:1, length:3 and shape:(3,)

A row-vector needs, when it must be used, to be specifically defined as

a matrix with only one row with a shape of (1, n). It is a two-dimensional
(2D) matrix but has only one row.
1 v_row = v[None,:] # adds in a new dimension without elements
2
3 print(f'Array v_row:\n{v_row}, its dimension:{v_row.ndim},',
4 f' length:{len(v_row)}, and shape:{v_row.shape}')

Array v_row:
[[5 2 8]], its dimension:2, length:1, and shape:(1, 3)

Similarly, if a column-vector is needed, one must specifically define a

matrix with only one column with a shape of (n, 1).
1 v_row = v[:,None] # adds in a new dimension without elements
2
3 print(f'Array v_row:\n{v_row}, its dimension:{v_row.ndim}',
4 f', length:{len(v_row)}, and shape:{v_row.shape}')

Array v_row:
[[5]
[2]
[8]], its dimension:2, length:3, and shape:(3, 1)

With this kind of specific definition, the rules of operations follow largely
those defined in standard linear algebra. Care is needed when using the
Numpy 1D arrays because of some different behavior. A more detailed dis-
cussion can be found in Section 2.9.13 in Ref. [2].

2.2.2 Basic properties of vectors

Vectors are widely used in science and engineering, including mechanics of
materials. The two most basic features of a vector are its magnitude and
direction. The former is represented graphically by the length and the latter
by an arrow. When we treat a vector as an object in isolation, these two
features are sufficient to define the vector.
A vector in an n-D coordinate space Rn can have a general form of
⎡ ⎤
v1
⎢ v2 ⎥
⎢ ⎥
v=⎢ . ⎥ (2.2)
⎣ .. ⎦
vn
14 Mechanics of Materials: Formulations and Solutions with Python

where vi , i = 1, 2, . . . , n, is the ith element of the vector on the ith axis of

the generalized Cartesian coordinate system. The zero vector has zero in
all its components.
Direction of a vector: The direction of a vector is defined by all its ele-
ments in reference to the coordinate system. Each vi is the value on the
ith axis of the Cartesian coordinate system where the vector is defined. It
lives essentially in a 1D coordinate space. This means that a vector in Rn is
defined by n numbers each in R1 with ndim(v) = 1 and len(v) = n.
Magnitude of a vector: The magnitude or physical length of a vector
must be a non-negative real number. It is denoted as v2 and is computed
using
 n
v2 ≡ v12 + v22 + ··· + vn2 = vi2 (2.3)
i=1

where v2 is also known as the L2 norm and is a physical measure of the
length of the vector. It is zero if and only if all the components are zero. In
general, we can use different types of norms, but L2 is the default. Thus, we
often drop the subscript 2 and denote the magnitude of a vector using v.
Unit vectors: The unit vector of a vector is denoted as v̄. It has a magnitude
of 1, but with the same direction as v. It is also called a normalized vector.
It is obtained using
v
v̄ = (2.4)
v
Let us visualize vectors in 2D graphs using Python.

2.2.3 Attributes of a vector in a computer

1 # Define a vector in Python, in 3D x, y, z coordinates.
2 [Link](8) # use a seed to make the random number repeatable
3 v = [Link]([4,2,5]) # w.r.t origin (0, 0, 0)
4
5 print('The data type of vector v in Python:', type(v))
6 print('The size (number of elements) of v =', [Link](v))
7 print('The shape (structure of data) of v:', [Link](v))
8 print('The vector v is:', v, ', and its 3rd element=',v[2])
9 print('The direction of v is defined by all its elements.')
10
11 length_v = [Link]([Link](v*v))
12 v_bar = v/length_v
13
14 print(f"The length or magnitude of v = {length_v:.4f}")
15 print(f"The unit vector of v is: {v_bar}")
16 print(f"Length of unit vector = {[Link]([Link](v_bar*v_bar))}")
 Preliminaries: Number, Vector, Tensor, and Functions 15

The data type of vector v in Python: <class '[Link]'>

The size (number of elements) of v = 3
The shape (structure of data) of v: (3,)
The vector v is: [4 2 5] , and its 3rd element= 5
The direction of v is defined by all its elements.
The length or magnitude of v = 6.7082
The unit vector of v is: [0.5963 0.2981 0.7454]
Length of unit vector = 1.0

Note that the choice of the orientation of the coordinate system will
produce differences in the presentation of the vectors. This relates to
the coordinate transformation of vectors, which will be discussed in detail
later in this chapter.
Vectors may change with the location in mechanics problems. There is an
area of study called function and functional analysis dealing with vectors
that vary over space. This study is widely applied in the modeling and simu-
lation of solid mechanics and fluid dynamics problems [3–5]. The locational
variation is represented in terms of functions [1]. In this book, we will simply
use the concept of functions.

2.2.4 The standard basis vectors

Consider a space Rn . We can define a set of n unit vectors that can used
as basis vectors. This special set of unit vectors is known as the standard
basis vectors (SBVs), and it is of importance in vector analysis. It can be
expressed as follows:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
1 0 0
⎢0⎥ ⎢1⎥ ⎢0⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
i1 = ⎢ .. ⎥; i2 = ⎢ .. ⎥; . . . ; in = ⎢ .. ⎥ (2.5)
⎣.⎦ ⎣.⎦ ⎣.⎦
0 0 1
Clearly, each one of these SBVs has length 1 by the definition of Eq. (2.3).
These SBVs form a set of vectors for the coordinate system, and they are
always linearly independent. This means that any one of these basis vec-
tors in the set cannot be written as a linear combination of the remaining
vectors in the same set. This can be easily proven by simple observation. We
see clearly that in any of the basis vectors, there is always only one entry
of unit 1, and all the other entries are zero. Take any basis vector, say the
ith. Its ith entry is 1. All the remaining basis vectors have only zero at their
ith entry. Therefore, it is not possible to find a linear combination of the
remaining basis vectors to produce a vector that has a nonzero value at the
ith entry. Thus, the ith SBV is linearly independent of the remaining SBVs.
This means that any vector in the space can be written as a linear
combination of the SBVs.
16 Mechanics of Materials: Formulations and Solutions with Python

2.2.5 Basic arithmetic operations for vectors

Consider a space Rn that contains all possible vectors with length n and
each of these n elements are in R. We now define the basic properties and
operations for these vectors.
1. The zero vector: The zero vector 0 is a vector with all its elements as
0. A nonzero vector shall have at least one nonzero element.
2. Scaling: For a given arbitrary vector v in Rn , a new vector w can be
created through scaling the vector v by multiplying an arbitrary scaling
factor α ∈ R:
⎡ ⎤
αv1
⎢ αv2 ⎥
⎢ ⎥
w = αv = ⎢ .. ⎥ ∀α ∈ R (2.6)
⎣ . ⎦
αvn
where vi , i = 1, 2, . . . , n, is the ith element of the vector. The scaling is
applied to each of the elements of v. Thus, the ith element of the resultant
new vector w becomes αvi (i = 1, 2, . . . , n), each of which is a number in R.
Even if α = 0, the resulting w is still in Rn because Rn has a zero vector by
definition. Therefore, the new vector w is always a member of Rn . Since
αvi = vi α based on the commutative property of standard arithmetic
operation of multiplication for numbers, we thus have αv = vα. This
implies that the scalar and the vector are also commutative. This allows
us to move the scalar numbers back and forth when needed.
3. Addition: Given two arbitrary vectors v and u both in Rn , a new vector
w is created by adding them up:
⎡ ⎤
v1 + u1
⎢ v2 + u2 ⎥
⎢ ⎥
w = v+u = u+v = ⎢ .. ⎥ ∀v and u ∈ Rn (2.7)
⎣ . ⎦
vn + un
The addition is done to each of the corresponding elements of these two
vectors v and u. Thus, the resultant new vector w is also in Rn because
each of its elements, (vi + ui ), is a number in R. Due to the commutative
property of standard arithmetic operation of addition for scalar numbers,
vectors in addition operation are also commutative.
4. Linear combination: With the definition of the zero vector and the
above two basic operations defined in items 2 and 3, it is clear that the
space Rn formed by the vectors is closed under addition and scaling.
Any linear combination of any two vectors u ∈ Rn and v ∈ Rn with an
arbitrary scaling factor α ∈ R and an arbitrary scaling factor β ∈ R,
w = αu + βv, is in Rn . Therefore, the Rn of vectors is called a linear
 Preliminaries: Number, Vector, Tensor, and Functions 17

vector space. Its members are all vectors with n elements, which are all
in R. By default, all vectors dealt with in this book have all the above
basic operations defined and are vector spaces.

Note that a vector space does not mean a space formed just by vectors.
The space of real numbers R is also a vector space because it is closed under
addition and scaling based on the defined standard arithmetic operations.
The essential point is that any linear combination of members in the space
shall remain in the same space. Matrices can also form vector spaces.
Now, any arbitrary vector v in Rn can be expressed as multiples of the
SBVs:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
v1 1 0 0
⎢v ⎥ ⎢0⎥ ⎢1⎥ ⎢0⎥
⎢ 2⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
v = ⎢ . ⎥ = v1 ⎢ .. ⎥ + v2 ⎢ .. ⎥ + · · · + vn ⎢ .. ⎥
⎣ . ⎦
. ⎣.⎦ ⎣.⎦ ⎣.⎦ (2.8)
v 0 0 1

= v1 i1 + v2 i2 + · · · + vn in

where vi (i = 1, 2, . . . , n) are scalar numbers that are the n independent

elements of v, and v is a linear combination of all the SBVs.
Note that we omitted the standard × symbol for multiplication of real
numbers to avoid confusion because in vector operations, × has an entirely
different meaning, known as the cross product, which is discussed later in
the section on cross product of vectors.

2.2.6 Other arithmetic operations for vectors

For vectors, we often perform also the following element-wise operations.

1. Subtraction: Similarly, a new vector w can be created by subtracting a

vector u from another vector v all in Rn :

w = v − u = −u + v
= (v1 − u1 )i1 + (v2 − u2 )i2 + · · · + (vn − un )in ∀v and u ∈ Rn (2.9)

The subtraction is only applied to each of the corresponding elements of

v and u. Thus, the resultant new vector w is also in Rn because each of its
elements, (vi − ui ), is in R. Subtraction operations are also commutative for
the same reason. Note that subtraction is simply negated addition and can
be induced by a scaling operation with α = −1 and the defined addition.
18 Mechanics of Materials: Formulations and Solutions with Python

2. Element-wise multiplication (also known as the Hadamard prod-

uct or Schur product): Given two vectors v and u both in Rn , we perform
w =v∗u=u∗v
= (v1 u1 )i1 + (v2 u2 )i2 + · · · + (vn un )in ∀v and u ∈ Rn (2.10)
As the name suggests, the multiplication is performed, respectively, on
each of the corresponding elements of these two vectors v and u. Thus, the
resultant new vector w is also in Rn because each of its elements, (vi ui ), is
in R. We use the symbol ∗ to denote element-wise multiplication, following
the Python convention. Symbols and ◦ are also used in the literature for
this operation.
3. Element-wise division: Given two vectors v and u both in Rn , we
perform
w = v/u = (v1 /u1 )i1 + (v2 /u2 )i2 + · · · + (vn /un )in
∀v and u ∈ Rn with ui = 0 (2.11)
As the name suggests, the division is performed, respectively, on each of
the corresponding elements of these two vectors v and u. Thus, the resultant
new vector w is also in Rn because each of its elements, (vi /ui ) when ui = 0,
is in R. We use the symbol / to denote element-wise division, following
Python’s convention. Note that element-wise division can often fail because
all the elements in u have to be nonzero for the defined operation.

2.3 Dot product of vectors, producing a scalar

Dot product (or inner product) is essential for all linear algebraic opera-
tions and will be used frequently throughout the entire book. It is the most
important operation to produce tensors for mechanics of materials. Let us
introduce it in detail.

2.3.1 Definition: A scalar produced using two vectors

Consider two nonzero vectors u and v both in Rn , as shown in Fig. 2.2.
To find the angle θ between them, we define an operation using these two
vectors, called the dot product or inner product:
u · v = u v cos θ (2.12)
where u · v is computed using multiplications of their corresponding
elements:
u · v = (u1 v1 ) + (u2 v2 ) + · · · + (un vn ) ∀u and v ∈ Rn (2.13)
 Preliminaries: Number, Vector, Tensor, and Functions 19

Figure 2.2. Two vectors u and v and the angle θ between them.

It is the sum of elements of the vector of element-wise multiplication

defined above. The result is a scalar number in R. For this reason, dot prod-
uct is sometimes called the scalar product.
It is obvious that the dot product is commutative: u · v = v · u. This is
also called symmetry. If the vectors are complex, it are conjugated symmetry:
u · v = v · u. Widely used notations for dot product are given as follows:
u·v =v·u dot product notation
= u, v = v, u inner product notation
(2.14)
= v u = u v matrix product notation
= (v1 u1 ) + (v2 u2 ) + · · · + (vn un ) for computation
All these notations are widely used in the literature and will be used in
this book interchangeably.

2.3.2 Direction cosine of two vectors

Since we already know how to compute the norm of a vector using Eq. (2.3)
and it will not be zero for any nonzero vector, the angle θ can now be
computed using Eqs. (2.12) and (2.13), which gives
u·v
cos θ = cos(u, v) = = ū · v̄ (2.15)
  u v
direction cosine
20 Mechanics of Materials: Formulations and Solutions with Python

This gives an important concept of direction cosine of two vectors,

which is used very often in mechanics problems. The angle θ is found using
 
u·v
θ = arccos = arccos(ū · v̄) (2.16)
u v
Care is needed when computing θ using the foregoing equation because
θ can be positive or negative due to the multivalued nature of the cosine
function.
To make a better sense of the dot product defined, let us look at a few
special cases.

2.3.3 Inner product associated norm

First, assume u and v are identical. In this case, cos θ = 1 because the
in-between angle is zero, and Eq. (2.12) becomes
v · v = v v = v2
√ √ (2.17)
=⇒ v = v · v = v1 v1 + v2 v2 + · · · + vn vn
which is exactly Eq. (2.3). The inner product has an associated norm, which
is the L2 norm. Thus, the inner product is often used to compute the L2
norm of a vector.
Suppose u and v are in the same direction but have different magnitudes.
In this case, we still have cos θ = cos 0 = 1 because the in-between angle is
zero, and Eq. (2.12) gives
u · v = u v when u  v (2.18)
These two vectors are parallel. In this case, the inner product of these
two vectors is the product of their norms.
When u and v are in the opposite direction, cos θ = cos 180◦ = −1, they
are also parallel, but
u · v = −u v (2.19)

2.3.4 Orthogonality and orthogonal projection

Assume u is perpendicular (orthogonal) to v. In this case, cos θ =
cos ±90◦ = 0 because their in-between angle is either π/2 or −π/2, and
Eq. (2.12) leads to
u · v = u v × 0 = 0 (2.20)
 Preliminaries: Number, Vector, Tensor, and Functions 21

Vectors that satisfy this inner product property are mutual orthogonal
vectors. The inner product is often used as tool to check the orthogonality
among vectors. Equation (2.20) gives also a clear definition for orthogonal-
ity of two vectors.
It is obvious that the inner product of the set of SBVs satisfy

1 i=j
ii , ij  = ii · ij = i
i ij = i
j ii = δij = (2.21)
0 i=j

It is the so-called Kronecker delta function defined in the box.

Let u be a unit vector. In this case, u = 1, we have

u · v = u v cos θ = v cos θ (2.22)


1

This is the orthogonal projection of v on ū! It is the red arrow shown in

Fig. 2.2. The magnitude (length) of the projected vector is

|OA| = v · ū = v cos θ (2.23)

This can only happen when ∠OAC is a right angle, and hence it is an
orthogonal projection.
Similarly, one can project u on v, and the projected vector is the green
vector shown in Fig. 2.2. This time, the magnitude of the projected vector is

|OD| = u · v̄ = u cos θ (2.24)

This implies that ∠ODB is a right angle, and u is orthogonally projected

on v.
Dot products will be frequently used for orthogonal projections in this
book.

2.3.5 Derivation of the law of cosines

You may have studied the law of cosines for arbitrary triangles in high school.
Using the dot product, we can easily derive this law.
Consider two given vectors u and v with an angle θ in between. Let w
be the difference vector: w = u − v. Thus, u, v, and w form a triangle. If
we want to calculate the length of w, we can use its dot product and the
22 Mechanics of Materials: Formulations and Solutions with Python

standard arithmetic operations:

w · w = (u − v) · (u − v) = u · u − u · v − v · u + v · v
= u2 − u · v − u · v + v2 = u2 − 2u · v + v2 (2.25)

which gives the law of cosines:

w2 = u2 + v2 − 2uv cos θ (2.26)

2.3.6 Power, energy, and work-done

In science and engineering, an inner product is often the outcome of a pair
of complementary vectors. For example:

• The power of an object is the inner product of the force vector and the
velocity vector that the object possesses.
• The angular momentum is the inner product of moment of inertia and
the angular velocity.
• The traction on the boundary of a solid is the inner product of the stress
vector and the outward normal vector of the boundary.
• The strain energy (density) in a stressed solid is the inner product of
the stress and strain tensors.
• The work-done on an object is the inner product of the force vector
applied on it and the displacement vector of the object [3].

2.3.7 Python examples

[Link] Dot product of an arbitrary pair of vectors

1 # dot product of two arbitrary vectors

2 u = [Link](1,6) # numpy 1D array
3 print(f"Vector u: {u}; u⋅u = {[Link](u,u)}, norm(u) = {[Link](u)**2}")
4
5 v = [Link](len(u)) * 2 # another numpy 1D array of same length
6 print(f"Vector v: {v}")
7
8 udv, vdu = [Link](u, v), [Link](v, u)
9 print(f"Symmetry: u ⋅ v = {udv}; v ⋅ u = {vdu}")
10 print(f"Alternative calculation: u ⋅ v = {[Link](v)}; v ⋅ u = {[Link](v)}")

Vector u: [1 2 3 4 5]; u ⋅ u = 55, norm(u) = 55.0

Vector v: [2. 2. 2. 2. 2.]
Symmetry: u v = 30.0; v u = 30.0
Alternative calculation: u v = 30.0; v u = 30.0
 Preliminaries: Number, Vector, Tensor, and Functions 23

[Link] Dot product of a pair of parallel vectors

Vector u: [1 2 3 4 5]; Vector v parallel u: [ 2 4 6 8 10]

u v = 110; norm(u)*norm(v) = 110.0

[Link] Dot product of two orthogonal vectors

1 # dot product of two orthogonal vectors

2 u = [Link](1,3) # numpy 1D array
3 v = [Link]([-u[1], u[0]]) # make v that is orthogonal to u
4
5 # create an orthogonal vector in higher dimension:
6 #v = [Link](len(u) # create a random vector of same length
7 #v-= [Link](u)*u/[Link](u)**2 # make it orthogonal
8
9 print(f"Vector u: {u}; \nIts orthogonal vector: {v}")
10 print(f"u⋅v = {[Link](u, v)}") # check orthogonality

Vector u: [1 2];
Its orthogonal vector: [-2 1]
uv = 0

[Link] Direction cosine of two vectors

1 # Direction cosine of two orthogonal vectors

2 [Link](8)
3 n = 6
4 u = [Link](6) # create a random vector in numpy 1D array
5 v = [Link](6) # create another
6
7 u_norm = [Link](u)
8 v_norm = [Link](v)
9
10 cos = [Link](u)/(u_norm*v_norm) # The direction cosine
11 = [Link](cos ) # the angle in between
12
13 print(f"The direction cosine of u and v = {cos }")
14 print(f"The angle between u and v = { }rad, {np.rad2deg( )} ")

The direction cosine of u and v = 0.43319936686066546

The angle between u and v = 1.1227568333736386rad, 64.3292279717825∘

[Link]
 24 Mechanics of Materials: Formulations and Solutions with Python

2.3.8 Types of norms of a vector

Without detailed definition and discussion, readers may simply take a look
at the means to compute other types of norms of a vector. Readers may
encounter these norms in the literature.
1 # compute other norms of a vector
2 u = [Link](1,6) # numpy 1D array
3 p = 3 # for Lp-norms, one may change it
4
5 print(f"Vector u: {u}; p={p}")
6 print(f"Infinity norms: ‖u‖∞={[Link](u,[Link])};"
7 f" ‖u‖-∞={[Link](u,-[Link])}")
8
9 print(f"L2 norm: ‖u‖2={[Link](u)}; ‖u‖2={[Link](u,ord=2)}")
10 print(f"L1 norm: ‖u‖1={[Link](u,ord=1)}")
11 print(f"Lp norm: ‖u‖p={[Link](u, ord=p)}")

Vector u: [1 2 3 4 5]; p=3

Infinity norms: ‖u‖∞=5.0; ‖u‖-∞=1.0
L2 norm: ‖u‖2=7.416198487095663; ‖u‖2=7.416198487095663
L1 norm: ‖u‖1=15.0
Lp norm: ‖u‖p=6.082201995573399

2.3.9 Direction cosine of an arbitrary vector

Consider an arbitrary vector v defined in a Cartesian coordinate. It is
from point A with coordinates (xA , yA , zA ) to point B with coordinates
(xB , yB , zB ) in three-dimensional (3D) coordinate space, as shown in Fig. 2.3.

Figure 2.3. An arbitrary vector v in 3D Cartesian coordinate space.

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 25

Using Eq. (2.15), the direction cosine of vector v and the x-axis can be
computed using

v 1 xB − xA
n1 = cos(v, x) = · i1 = v · i1 = (2.27)
v v   v
xB −xA

Here, we used the fact that v · i1 is the orthogonal projection of v on i1 ,

which gives xB − xA .
The same applies to the direction cosines of v with the y-axis and v with
the z-axis. We thus obtain in summary:

xB − xA
n1 = cos(v, x) =
v
yB − yA
n2 = cos(v, y) = (2.28)
v
zB − zA
n3 = cos(v, z) =
v

The norm of v is computed using


v = (xB − xA )2 + (yB − yA )2 + (zB − zA )2 (2.29)

For 2D problems, say in the x–y plane, we need to use the first two
equations given above. These formulas will be frequently used in this book.

2.4 Outer product of two vectors, producing a matrix

2.4.1 Definition of outer product

An inner product of two vectors produces a scalar, as we have seen in the
previous section. In contrast, the outer product of two vectors produces a
matrix. It has applications in linear algebra, particularity in matrix oper-
ations. In this book, we will use it to generate the so-called displacement
gradient matrix that leads to strain–displacement relations.
Let two vectors u ∈ Rm and v ∈ Rn be
 
u = u1 u2 · · · um
  (2.30)
v = v1 v2 · · · vn

[Link]
 26 Mechanics of Materials: Formulations and Solutions with Python

The outer product u ⊗ v becomes a matrix in Rm×n :

A = u ⊗ v
m×n m×1 n×1
⎡ ⎤ ⎡ ⎤
u1 u1 v1 u1 v2 ... u1 vn
⎢u ⎥ ⎢u v u v ... u2 vn ⎥ (2.31)
⎢ 2⎥  ⎢ 2 1 2 2 ⎥
=⎢
⎢ . ⎥
.
⎥ v1 v2 · · · vn = ⎢
⎢ . . .. .. .. ⎥
⎥
⎣ . ⎦ ⎣ . . . . ⎦
um um v1 um v2 . . . um vn

2.4.2 Rank-one matrix

The ijth element in A is ui vj . The outer product is same as the matrix
multiplication of uv . Thus, the shapes of u and v are always compatible
for this operation. This type of matrix is known as a rank-one matrix
because it has essentially one independent row or column.
Now, if we use the SBVs to write vectors u and v, as in Eq. (2.8), we
shall have the following:

u = u1 i1 + u2 i2 + · · · + um im
(2.32)
v = v1 i1 + v2 i2 + · · · + vn in

The outer product u ⊗ v can be re-written as follows:

A = u ⊗ v = u1 v1 i1 ⊗ i1 +u1 v2 i1 ⊗ i2
m×n m×1 n×1    
I11 I12
(2.33)
+ u2 v1 i2 ⊗ i1 + · · · + um vn im ⊗ in
   
I21 Imn

where Iij is called standard basis matrix (SBM), given by the outer products
of the SBVs:

Iij = ii ⊗ ij ; i = 1, 2, . . . , m; j = 1, 2, . . . , n (2.34)
m×n m×n

Any matrix in Rm×n can be write as a linear combination of Iij .

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 27

2.4.3 Python example

Let us first look at an example.

1 u = [Link](1,4) # numpy 1D array

2 v = [Link](5) * 2 # numpy 1D array
3 print(f"Vector u: {u}; Vector v: {v}")
4
5 A, B = [Link](u, v), [Link](v, u)
6 print(f"v ⊗ u: \n{B}; its rank = {lg.matrix_rank(B)}")
7 print(f"u ⊗ v: \n{A}; its rank = {lg.matrix_rank(A)}")
8
9 uc = [Link]([Link][0],1) # change u to a column-vector
10 vc = [Link]([Link][0],1) # change v to a column-vector
11 print(f"uc@vc: \n{[email protected]}") # numpy matrix multiplication

Vector u: [1 2 3]; Vector v: [2. 2. 2. 2. 2.]

v ⊗ u:
[[2. 4. 6.]
[2. 4. 6.]
[2. 4. 6.]
[2. 4. 6.]
[2. 4. 6.]]; its rank = 1
u ⊗ v:
[[2. 2. 2. 2. 2.]
[4. 4. 4. 4. 4.]
[6. 6. 6. 6. 6.]]; its rank = 1
uc@vc:
[[2. 2. 2. 2. 2.]
[4. 4. 4. 4. 4.]
[6. 6. 6. 6. 6.]]

2.5 Cross product of vectors, producing a vector

A dot product produces a scalar, and an outer product produces a matrix. Is

there a vector operation that produces a vector? Yes. A cross product does
this. It is also called the vector product. In this book, it is used to produce
the curl of stress-vector functions leading to the torque on a torsional bar.
An excellent wiki page ([Link] product) is
available on this topic. This section briefly outlines the major points with
demonstrations in Python codes. Our notation follows largely that in the
wiki page so that readers can make easy connections.

[Link]
 28 Mechanics of Materials: Formulations and Solutions with Python

2.5.1 Definition
Given two linearly independent vectors a and b both in R3 , the cross product
is denoted as a × b and often reads “a crosses b”. The outcome is a vector
c also in R3 perpendicular (or orthogonal) to both a and b. Therefore, it is
normal to the plane of R2 spanned by a and b in R3 . Together with a and
b, it spans a space of R3 , as shown in Fig. 2.4.
The direction of the cross product has only two choices: either positive or
negative. This is why the cross product is called a pseudo vector. The positive
direction of the cross product is defined by the so-called right-hand rule, as
shown on the right in Fig. 2.4. When b crosses a, the resultant vector is in the
negative (downwards) direction. The right-hand rule is followed throughout
this book.

2.5.2 Formulation
When these two vectors are not linearly independent, meaning they are in-
line or collinear, their cross product is a zero vector. This is also true when
either of these two vectors is a zero vector.
The magnitude of the cross product equals the area of the parallelogram
formed by a and b, as shown in Fig. 2.5.

(a) (b)

Figure 2.4. (a) Cross product of two vectors. (b) The right-hand rule determines the
positive direction of the cross product of two vectors.
Source: [Link] product#/media/File:Right hand rule cross
[Link], made by Acdx, under the CC BY-SA 3.0 license.

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 29

Figure 2.5. Cross product of two vectors forms a parallelogram. The amplitude of the
cross product is the area of the parallelogram.
Source: ([Link] product [Link]).

The formula for the cross product definition is given by

c = a × b = a b sin(θ) n (2.35)
 
c

where 0◦ ≤ θ ≤ 180◦ is the angle measured from a to b, and n is the

unit vector perpendicular to the plane spanned by a and b following the
right-hand rule. As seen in Eq. (2.35), the positivity of the direction of c
is controlled by the sine function, which is an odd function. Therefore, if b
crosses a, θ will be negative, and the direction will reverse. In other words,
cross product is anti-commutative.

2.5.3 Volume of a parallelepiped

The volume V of a parallelepiped formed by three vectors a, b, and c as its
edges can be computed using a combination of a cross product and a dot
product, called scalar triplet product, as shown in Fig. 2.6.
The triplet product can be done in any of the three orders, and they all
gives a scalar:
a · (b × c) = b · (c × a) = c · (a × b) (2.36)
Note that the scalar triplet product can be negative, so the volume of the
parallelepiped should be the absolute value of the scalar:
V = |a · (b × c)| (2.37)

[Link]
 30 Mechanics of Materials: Formulations and Solutions with Python

Figure 2.6. A parallelepiped defined using three vectors. Image from Wikimedia
Commons in public domain, by Jitse Niesen.

2.5.4 Differences between the dot product and cross

product
1. The magnitude of the cross product is computed using the sine of the
angle between two vectors. Thus, the cross product can be a measure of
perpendicularity of two vectors. In contrast, the dot product is a measure
of parallelism, as mentioned earlier.
2. Given two vectors with unit length (unit vectors), if they are perpendic-
ular, the cross product has a magnitude of 1, and if the two are parallel,
the magnitude is zero. Conversely, the dot product of two perpendicular
unit vectors is zero and is 1 if they are parallel.
3. The magnitude of the cross product of two unit vectors gives the sine of
the angle between two vectors θ, which is always positive by definition
(θ is 0◦ ∼180◦ and measured counterclockwise). In contrast, the dot prod-
uct of two unit vectors gives the cosine of θ, and hence θ can be positive
or negative.

2.5.5 Derivation of formulas

Since the cross product involves direction change, the computation becomes
more complicated. For easy understanding, we first examine the cross prod-
ucts of the SBVs. Let us consider 3D spaces, and denote these three SBVs

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 31

as i, j, and k. Based on the right-hand rule, we have

i × j = k, j × i = −k
j × k = i, k × j = −i (2.38)
k × i = j, i × k = −j
Note the negative sign resulting from anti-commutativity. In addition,
because of collinearity, we have
i×i= j×j= k×k = 0 (2.39)
Using these SBVs, the cross product of any two vectors a and b can be,
respectively, written as
a = a1 i + a2 j + a3 k
(2.40)
b = b1 i + b2 j + b3 k
where ai , aj , and ak are the elements of a and bi , bj , and bk are the elements
of b, respectively, on its corresponding SBVs.
Now, following the standard arithmetic operations and the definition of
cross products, we have
c = a × b = (a1 i + a2 j + a3 k) × (b1 i + b2 j + b3 k)
= a1 b1 (i × i) + a1 b2 (i × j) + a1 b3 (i × k)
+ a2 b1 (j × i) + a2 b2 (j × j) + a2 b3 (j × k)
+ a3 b1 (k × i) + a3 b2 (k × j) + a3 b3 (k × k) (2.41)
Using Eqs. (2.38) and (2.39), the foregoing equation becomes
c = a × b = a1 b1 0 + a1 b2 k − a1 b3 j
− a2 b1 k + a2 b2 0 + a2 b3 i
+ a3 b1 j − a3 b2 i + a3 b3 0
= (a2 b3 − a3 b2 ) i + (a3 b1 − a1 b3 ) j + (a1 b2 − a2 b1 ) k
     
c1 c2 c3

= c1 i + c2 j + c3 k (2.42)
We observe a nice duality formula for computing the elements of vector
c if we re-denote c1 = c23 , c2 = c31 , c3 = c12 :
cij = ai bj − aj bi (2.43)

[Link]
 32 Mechanics of Materials: Formulations and Solutions with Python

When i, j = 1, 2, 3, the above duality formula gives exactly three distinct

values, which we denote as c1 , c2 , c3 to express vector c.
Alternatively, c can be expressed in the determinant of a matrix as
follows:
 
i j k 

 
× b = a1 a2
a  a3 
 
c  b1 b2 b3 
      (2.44)
a2 a3  a1 a3  a1 a2 
=   i −   j+
b b  k

b2 b3  b1 b3  1 2
     
c1 c2 c3

Moreover, we can use the following matrix-vector formulation in compu-

tation.
⎡ ⎤⎡ ⎤
0 −a3 a2 b1
⎢ ⎥⎢ ⎥
× b = ⎣ a3
a  0 −a1 ⎦ ⎣b2 ⎦ (2.45)
c −a2 a1 0 b3

where the matrix is anti-symmetric because the cross product is anti-

commutative.
The cross product has applications in various areas, including computa-
tional geometry, evaluation of angular momentum and torque, and calcula-
tion of Lorentz forces, calculation of torques by forces, just to name a few. It
is often used to determine the direction normal to any pair of vectors, which
is needed in many operations, such as coordinate transformation in the finite
element methods [3] for stress analysis in structures. Interested readers may
refer to the wiki page ([Link] product) and
the references therein for more detailed formulations, properties, history,
and applications.
Note that we did not discuss about cross products in higher dimensions.
This is because it would get into the concept of exterior algebra or tensor
algebra. From Eq. (2.43), we see that our formula works just right for 3D
when the dimension gets higher, 4D, for example, i and j shall vary from
1 to 4. In such cases, the duality formula would produce a lot more terms
than 4, implying it becomes a tensor for higher dimensions. Discussion on
exterior or tensor algebra is beyond the scope of this book. We thus stop
here.
[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 33

2.5.6 Examples of computing cross products

We now write some simple codes to examine cross products.

1 a = [2., 1, 3]
2 b = [5., 3, 2]
3
4 c1 = a[1]*b[2]-a[2]*b[1]
5 c2 = -(a[0]*b[2]-a[2]*b[0])
6 c3 = a[0]*b[1]-a[1]*b[0]
7 print(f"elements of vector c = a × b: c1={c1}, c2={c2}, c3={c3}")
8
9 c_ab = [Link](a, b)
10 c_ba = [Link](b, a)
11 print(f"c = a × b: {c_ab}\nc = b × a: {c_ba}")
12 print(f"c dot a: {c_ab.dot(a):.4f}, c dot b: {c_ab.dot(b):.4f}")

elements of vector c = a × b: c1=-7, c2=11.0, c3=1.0

c = a × b: [-7. 11. 1.]
c = b × a: [ 7. -11. -1.]
c dot a: 0.0000, c dot b: 0.0000

We observe all the major properties: 1. a × b is negative b × a; 2. c is

normal to both a and b because these dot products vanish.

1 # matrix-vector formulation:
2 A = [Link]([[ 0, -a[2], a[1]],
3 [ a[2], 0, -a[0]],
4 [-a[1], a[0], 0]])
5
6 print(f"elements of vector c = a x b = {A@b}")

elements of vector c = a x b = [-7. 11. 1.]

1 = [Link]([Link](a/[Link](a), b/[Link](b))) # can be negative

2 Area = abs([Link](a)*[Link](b)*[Link]( )) # thus, take abs()
3
4 print(f"Area of the parallelogram via definition: {Area:.4f}")
5 print(f"Area via [Link]: {[Link](c_ab):.4f}")

Area of the parallelogram via definition: 13.0767

Area via [Link]: 13.0767

We confirm another property that the norm of c is the area of the paral-
lelogram. We can do the same using randomly generated vectors.
[Link]
 34 Mechanics of Materials: Formulations and Solutions with Python

1 n = 3
2 [Link](8)
3 a = [Link](n) # randomly generated vector
4 b = [Link](n)
5
6 c_ab, c_ba = [Link](a, b), [Link](b, a)
7 print(f"c = a × b: {c_ab}\nc = b × a: {c_ba}")
8 print(f"c dot a: {c_ab.dot(a):.4f}, c dot b: {c_ab.dot(b):.4f}")
9
10 theta = [Link]([Link](a/[Link](a), b/[Link](b)))
11 Area = abs([Link](a)*[Link](b)*[Link](theta))
12 print(f"Area of the parallelogram via definition: {Area:.4f}")
13 print(f"Area via [Link]: {[Link](c_ab):.4f}")

c = a × b: [-1.8414 2.4794 1.3034]

c = b × a: [ 1.8414 -2.4794 -1.3034]
c dot a: 0.0000, c dot b: -0.0000
Area of the parallelogram via definition: 3.3522
Area via [Link]: 3.3522

For these randomly generated vectors, we observe again the same

properties.

2.6 Examples of vector operations and visualization

Now, we present examples to demonstrate vector operations using Python

code. In order to have the results easily visualizable, we study vectors in 2D
space.
For vectors in 2D space R2 , we shall have two basis vectors which can
be represented in graphs. In this case, the standard basis vectors, or SBVs,
become
   
1 0
i1 = ; i2 = (2.46)
0 1

Any vector in R2 can now be expressed using these basis vectors, as shown
in the following example:
     
v1 1 0
= v1 + v2 = v1 i1 + v2 i2 (2.47)
v2 0 1

Since v1 and v2 are scalar numbers in R, vector [ vv12 ] lives in R2 . This can
be demonstrated using the following codes.

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 35

2.6.1 Creation of vectors using the standard basis

vectors, and scaling vectors

1 o1 = [Link]([0,0])
2 i_1 = [Link]([1.,0]) # basis vector i1
3 i_2 = [Link]([0.,1]) # basis vector i2
4
5 v = [Link]([3., 4]) # any given vector
6 print('elements:', v[0], ' and ', v[1]) # vector elements
7 v1 = v[0]*i_1 + v[1]*i_2
8 print('v=', v, '; v_formed by two SBVs = v1', v1)
9 print('Norm of i_1=', [Link](i_1), ' Norm of i_2=',
10 [Link](i_2), ' Norm of v=', [Link](v))
11
12 print('L1-norm of v1=', [Link](v1,1), ' L2-norm of v1=',
13 [Link](v1), ' Inf-norm of v1=', [Link](v1,[Link]))
14
15 print('Normalized i_1=', i_1/[Link](i_1), ' Normalized i_2=',
16 i_2/[Link](i_2),' Normalized v=', v/[Link](v))
17
18 v2 = [Link]([-2, 3])
19 v4 = -v2
20 v3 = v1/[Link](v1)
21 v5, v6= 0.5 *v1, -0.5 *v2
22 print('v2=',v2,' -v2=',v4,' .5v1=',v5,' -.5v2=',v6)
23
24 fig, axis = [Link](1,2,figsize=(6.5,3))#3.8,1.8))#
25 Lb1 = drawArrow(axis,o1,i_1,color='r',iax=0,label='Basis')
26 Lb2 = drawArrow(axis,o1,i_2,color='r',iax=0,label='None')
27 Lb3 = drawArrow(axis,o1,v1,color='b',iax=0,label='v1=$2i_1+5i_2$')
28 Lb4 = drawArrow(axis,o1,v3,color='k',iax=0,label='Normalized v1')
29 Lb5 = drawArrow(axis,o1,v2,color='c',iax=0,label='v2')
30 Lb6 = drawArrow(axis,o1,v4,color='m',iax=0,label='$-$v2')
31 axis[0].legend(handles=[Lb1,Lb3,Lb4,Lb5,Lb6], loc='lower left')
32 plotfig(axis,iax=0, title='SBVs and scaled vectors',xylim=5)
33
34 Lb1 = drawArrow(axis,o1,i_1,color='r',iax=1,label='Basis')
35 Lb2 = drawArrow(axis,o1,i_2,color='r',iax=1,label='None')
36 Lb3 = drawArrow(axis,o1,v1,color='b',iax=1,label='v1=$2i_1+5i_2$')
37 Lb4 = drawArrow(axis,o1,v2,color='c',iax=1,label='v2')
38 Lb5 = drawArrow(axis,o1,v5,color='k',iax=1,label='0.5*v1')
39 Lb6 = drawArrow(axis,o1,v6,color='m',iax=1,label='$-$0.5*v2')
40 axis[1].legend(handles=[Lb1,Lb3,Lb4,Lb5,Lb6], loc='lower left')
41 plotfig(axis,iax=1, title='SBVs and scaled vectors',xylim=5)
42 [Link]('images/[Link]', dpi=500)
43 [Link]()

elements: 3.0 and 4.0

v= [3. 4.] ; v_formed by two SBVs = v1 [3. 4.]
Norm of i_1= 1.0 Norm of i_2= 1.0 Norm of v= 5.0
L1-norm of v1= 7.0 L2-norm of v1= 5.0 Inf-norm of v1= 4.0
Normalized i_1= [1. 0.] Normalized i_2= [0. 1.] Normalized v= [0.6 0.8]
v2= [-2 3] -v2= [ 2 -3] .5v1= [1.5 2. ] -.5v2= [ 1. -1.5]

[Link]
 36 Mechanics of Materials: Formulations and Solutions with Python

Figure 2.7. Vectors generated in various means.

2.6.2 Addition of vectors

1 v1 = [Link]([3,-1])
2 v2 = [Link]([1, 4])
3 v3 = v1 + v2
4 v4 = v1 - v2
5 print('v1=',v1,' v2=',v2,' v1+v2=',v3,' v1-v2=',v4)
6
7 fig, axis = [Link](1,2,figsize=(6.5,3))#(3.8,1.8))#(10.5,5))
8
9 Lb1 = drawArrow(axis,o1,v1,color='r',iax=0,label='v1')
10 Lb2 = drawArrow(axis,o1,v2,color='b',iax=0,label='v2')
11 Lb3 = drawArrow(axis,o1,v3,color='c',iax=0,label='v3=v1+v2')
12 axis[0].legend(handles=[Lb1,Lb2,Lb3], loc='upper left')
13 plotfig(axis,iax=0, title='Addition of two vectors',xylim = 5)
14
15 Lb1 = drawArrow(axis,o1,v1,color='r',iax=1,label='v1')
16 Lb2 = drawArrow(axis,o1,v2,color='b',iax=1,label='v2')
17 Lb3 = drawArrow(axis,o1,v4,color='c',iax=1,label='v4=v1-v2')
18 axis[1].legend(handles=[Lb1,Lb2,Lb3], loc='upper left')
19 plotfig(axis,iax=1, title='Subtraction of two vectors',xylim=5)
20 [Link]('images/[Link]', dpi=500)
21 [Link]()

v1= [ 3 -1] v2= [1 4] v1+v2= [4 3] v1-v2= [ 2 -5]

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 37

Figure 2.8. Vector additions.

2.6.3 Element-wise multiplication and dot product

of two vectors

1 # Define two pairs of vectors of the same shape

2 v1 = [Link]([3.,-1])
3 v2 = [Link]([1., 4])
4 v3 = v1*v2
5
6 print('v1=',v1,' v2=',v2,' v1+v2=',v3)
7 print('numpy dot product of two vectors =',[Link](v1,v2))
8 print('sum of products of their corresponding elements=',
9 v1[0]*v2[0]+v1[1]*v2[1])
10 print('sum of elements of vector element-wise multiplication =',
11 [Link](v1*v2))
12 print('')
13
14 v1 = [Link]([3.,-1, 4, 8, 2])
15 v2 = [Link]([1., 4,-3, 5, 3])
16 v3 = v1*v2
17 v4 = v1/v2
18
19 print('v1=',v1,' v2=',v2,'\n v1*v2=',v3,'\n v1/v2=',v4)
20 print('dot product of two vectors =',[Link](v1,v2))
21 print('sum of elements of vector element-wise multiplication =',
22 [Link](v1*v2))
23 print(f"sum of elements of vector element-wise division =",
24 f"{[Link](v1/v2):.4f}")

[Link]
 38 Mechanics of Materials: Formulations and Solutions with Python

v1= [ 3. -1.] v2= [1. 4.] v1+v2= [ 3. -4.]

numpy dot product of two vectors = -1.0
sum of products of their corresponding elements= -1.0
sum of elements of vector element-wise multiplication = -1.0

v1= [ 3. -1. 4. 8. 2.] v2= [ 1. 4. -3. 5. 3.]

v1*v2= [ 3. -4. -12. 40. 6.]
v1/v2= [ 3. -0.25 -1.3333 1.6 0.6667]
dot product of two vectors = 33.0
sum of elements of vector element-wise multiplication = 33.0
sum of elements of vector element-wise division = 3.6833

It is confirmed that the dot product is the same as the sum of the elements
of the vector created via element-wise multiplication of these two vectors,
which is defined in Eq. (2.14).
We write the following codes to plot the vectors produced by the element-
wise multiplications of two vectors and that of element-wise divisions of two
vectors.

1 # Define a pair of vectors in 2D space

2 v1 = [Link]([3,-1])
3 v2 = [Link]([1, 4])
4 v3 = v1*v2
5 v4 = v1/v2
6 print('v1=',v1,' v2=',v2,' v1*v2=',v3,' v1/v2=',v4)
7
8 fig, axis = [Link](1,2,figsize=(6.5,3))#(3.8,1.8))#(10.5,5))
9
10 Lb1 = drawArrow(axis,o1,v1,color='r',iax=0,label='v1')
11 Lb2 = drawArrow(axis,o1,v2,color='b',iax=0,label='v2')
12 Lb3 = drawArrow(axis,o1,v3,color='c',iax=0,label='v3=v1*v2')
13 axis[0].legend(handles=[Lb1,Lb2,Lb3], loc='upper left')
14 plotfig(axis,iax=0, title='element-wise multiplication of\
15 two vectors',xylim = 5)
16
17 Lb1 = drawArrow(axis,o1,v1,color='r',iax=1,label='v1')
18 Lb2 = drawArrow(axis,o1,v2,color='b',iax=1,label='v2')
19 Lb3 = drawArrow(axis,o1,v4,color='c',iax=1,label='v4=v1/v2')
20 axis[1].legend(handles=[Lb1,Lb2,Lb3], loc='upper left')
21 plotfig(axis,iax=1, title='Element-wise division of\
22 two vectors',xylim = 5)
23 [Link]('images/Mul_DivOfVectors.png', dpi=500)
24 [Link]()

v1= [ 3 -1] v2= [1 4] v1*v2= [ 3 -4] v1/v2= [ 3. -0.25]

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 39

Figure 2.9. Vector multiplication and division.

2.6.4 Orthogonal vectors

As mentioned earlier, if two vectors v, u in Rn are said to be orthogonal, it
means that the dot product of them is zero:

v · u = (v1 u1 ) + (v2 u2 ) + · · · + (vn un ) = 0 (2.48)

The obvious example of orthogonal vectors are the set of the SBVs. Let
us write a simple code to demonstrate this.

1 n = 4
2 I = [Link](n) # Create an identity matrix.
3
4 i_1 = [Link](I,n)[0].reshape(n,) #[1 0 0 0]
5 i_2 = [Link](I,n)[1].reshape(n,) #[0 1 0 0]
6 i_3 = [Link](I,n)[2].reshape(n,) #[0 0 1 0]
7 i_4 = [Link](I,n)[3].reshape(n,) #[0 0 0 1]
8
9 print(i_1, i_2, i_3, i_4)
10 print('Dot products of pairs of SBVs:\n',[Link](i_1,i_1),
11 [Link](i_1,i_2), [Link](i_1,i_3), [Link](i_2,i_4))

[1. 0. 0. 0.] [0. 1. 0. 0.] [0. 0. 1. 0.] [0. 0. 0. 1.]

Dot products of pairs of SBVs:
1.0 0.0 0.0 0.0

It is confirmed that the SBVs are orthogonal basis vectors.

Orthogonal vectors are not necessarily as “nice looking” as the SBVs. Let
us generate more orthogonal vectors randomly using the following snippet.
[Link]
 40 Mechanics of Materials: Formulations and Solutions with Python

1 # The orth function has already imported before from [Link]

2 A = [Link](5, 5) # generate a random matrix
3 OA = [Link](A) # compute orthogonal vectors using A
4
5 print(f"Any column vector is orthogonal to the rest:\n{OA}")
6 print('Dot products of column vectors:\n',[Link](OA[:,0],OA[:,1]),
7 [Link](OA[:,0],OA[:,2]),[Link](OA[:,1],OA[:,2]))

Any column vector is orthogonal to the rest:

[[-0.5411 -0.0601 -0.1721 -0.5599 -0.6004]
[-0.3929 0.1741 -0.6865 -0.0543 0.584 ]
[-0.2489 -0.9268 -0.0437 0.2656 0.082 ]
[-0.5318 0.3236 0.1047 0.7295 -0.2633]
[-0.4561 0.0498 0.6973 -0.2844 0.4715]]
Dot products of column vectors:
1.1102230246251565e-16 -2.220446049250313e-16 -1.8735013540549517e-16

As shown, these five vectors look not so nice, but they are orthogonal
to each other. They can be used as orthogonal basis vectors and span a 5D
vector space.
For vectors in 2D vector space, we can view clearly the orthogonality of
vectors.

1 # Define two pairs of orthogonal vectors in 2D space

2 v1, v2 = [Link]([3,-1]), [Link]([1,3])
3 v3, v4 = [Link]([2, 3]), [Link]([-3,2])
4 print('v1=',v1,' v2=',v2,' v1*v2=',v3,' v1/v2=',v4)
5 print('v1 dot v2=',[Link](v1,v2),' v3 dot v4=',[Link](v1,v2))
6
7 fig, axis = [Link](1,2,figsize=(6.5,3))
8
9 Lb1 = drawArrow(axis,o1,v1,color='r',iax=0,label='v1')
10 Lb2 = drawArrow(axis,o1,v2,color='b',iax=0,label='v2')
11 axis[0].legend(handles=[Lb1,Lb2], loc='upper left')
12 plotfig(axis,iax=0, title='Two orthogonal vectors',xylim = 5)
13
14 Lb1 = drawArrow(axis,o1,v3,color='r',iax=1,label='v3')
15 Lb2 = drawArrow(axis,o1,v4,color='b',iax=1,label='v4')
16 axis[1].legend(handles=[Lb1,Lb2], loc='upper left')
17 plotfig(axis,iax=1, title='Two orthogonal vectors', xylim = 5)
18 [Link]('images/[Link]', dpi=500)
19 [Link]()

v1= [ 3 -1] v2= [1 3] v1*v2= [2 3] v1/v2= [-3 2]

v1 dot v2= 0 v3 dot v4= 0

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 41

Figure 2.10. Orthogonal vectors in 2D.

2.7 Coordinate transformation

In solving a mechanics problem, we deal with objects including numbers, vec-

tors, and stress tensors. Such an object is usually represented in reference
to an orthogonal coordinate system, such as Cartesian, polar, cylindrical, or
spherical. We often perform coordinate transformation, which can result in
changes in the representation of the object. Coordinate transformation often
refers to rotation or a change from one type of (often orthogonal) coor-
dinate system to another type of (often orthogonal) coordinate system, for
example changes from the Cartesian to polar coordinate system. Coordinate
transformation of any object occurs usually at a point.
Therefore, when we say coordinate transformation, it does NOT include
translation. The change of an object due to the location change is dealt with
using the concept of function [1].
A number is a tensor of 0th order. The rule of coordinate transformation
for a number is the simplest:

A number does not change when the coordinate transforms. It does not
require an index with respect to the coordinates.

This is simply because of the transformation definition: When a number

is represented in a coordinate, we have only a single 1D axis, say x, in R,
as shown in Fig. 2.1. The number has a single value right on the axis of a
given 1D coordinate. In a 1D space, the axis can only be translated along its

[Link]
 42 Mechanics of Materials: Formulations and Solutions with Python

axis direction and cannot have rotation (otherwise, the space becomes 2D
or 3D).
The single value can still be viewed as obtained by an orthogonal projec-
tion, but the angle of the projection is zero. Any rotation (with zero angle)
of the coordinate has no effect at all to any number. In other words, the
transformation matrix T (discussed in detail later) is simply 1.
Due to the transformation invariant property of a number, it is called a
scalar.

2.8 Vectors in mechanics

In mechanics, vectors are often encountered important objects made of scalar

numbers, each of which represents a component of the vector. Typical ones
are the force vector and the displacement vector.

2.8.1 Notation for coordinate systems

To quantify a vector in mechanics, we use an orthogonal coordinate system.
A vector can then be equivalently described in a number of scalars in an
array.
Consider a 3D space R3 , in which we define a Cartesian coordinate system
x that is orthogonal: x-, y-, and z-axis are the three mutually orthogonal
axes of x. Each of these axes has a unit vector i, j, and k with unit length,
respectively.
In indicial rotation, these axes are denoted as x1 , x2 , and x3 , or xi
(i = 1, 2, 3), and its unit vectors are denoted as i1 , i2 , and i3 or ii (i = 1, 2, 3).
The unit vectors are also denoted as e1 , e2 , and e3 or ei (i = 1, 2, 3). Indicial
notation is more concise and often more convenient in derivations. Hence,
we often switch to it in this book when necessary. Readers will be getting
used to it after doing this for some time.

2.8.2 Description of vectors

Consider generally a vector in mechanics denoted as v in R3 , as shown in
Fig. 2.11. The vector can be a force vector, displacement vector, velocity
vector, or acceleration vector.
We present v uniquely under the Cartesian coordinate system. We first
decompose it into two vectors: a vertical component v3 i3 and an in-plane
vector vs following the orthogonal projection rule. vs is then further decom-
posed into two vectors: v1 i1 and v2 i2 following also the orthogonal projection
[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 43

Figure 2.11. Any vector v defined in an orthogonal coordinate system R3 can be decom-
posed into three components: v1 i1 , v2 i2 , and v3 i3 following the orthogonal projection rule.

rule. This process results in three components of v: v1 , v2 , and v3 , respec-

tively, on the x1 -, x2 -, and x3 -axis. The vector can now be written uniquely
as
⎡ ⎤
v1  
v = ⎣v2 ⎦ = v1 v2 v3 (2.49)
v3

where  stands for transpose. Any change to any of these three components
results in another different vector.

2.8.3 Equivalence in representation, first-order tensor

From Fig. 2.11, we found the following:

1. It is clear that this representation of a vector is unique under a fixed

coordinate, implying that any change, in length or in direction, in v will
result in unique changes in v1 , v2 , and v3 , and vice versa, as long as the
orthogonal projection rule is followed. This mutual unique representa-
 
tion ensures the equivalent representation, implying v and v1 v2 v3
are identical. Such an equivalence is ensured by the orthogonal projection
rule in an orthogonal coordinate system.

[Link]
 44 Mechanics of Materials: Formulations and Solutions with Python

2. When the coordinate x is rotated (transformed) to x , vector v is now

represented as
⎡ ⎤
v1
⎢ ⎥
v = ⎣v2 ⎦ (2.50)
v3

The values of v1 , v2 , and v3 are different from v1 , v2 , and v3 . How-
ever, these two basic attributes: length and direction of the vector v remain
unchanged. The vector under our transformation rule (orthogonal projec-
tion on orthogonal coordinates) is a tensor. Since we need only one index to
uniquely locate its components, it is called a first-order tensor. Since it is
in 3D, it is a 3D tensor.

2.8.4 Tensor concept in general

Tensor is a term we often hear about, but it confuses many, including the
author in the past for a long time. If you do not have a clear idea about
it, you are not alone. To put this concept in a short sentence, it states as
follows.
A tensor is an object defined in a system, where the change in the values
describing the object obeys a certain rule, when a transformation occurs in
the system.
This is a very general definition. It implies that there are different types
of tensors in different fields or systems with different transformation rules.
A more straightforward question may be: Is a square matrix a tensor?
The answer can be either no or yes. If the matrix contains merely stacks
of numbers, it is not. If there is a certain rule attached to it, so that the
numbers in it change following a system-wide rule when a transformation
occurs, it becomes a tensor. A second-order tensor can be in a square matrix
form, but a square matrix is not automatically a tensor.

2.8.5 Definition of tensor in mechanics of materials

In mechanics of materials (or solid mechanics, continuum mechanics, or elas-
ticity theory), all the objects we deal with are tensors of different orders.
Scalars are of zeroth order, forces and displacements are vectors and are
first order, stress and strain are second, and elasticity constants are fourth.
These objects are defined in an orthogonal coordinate that can transform
(rotate or type change). The rule for operating on these objects and to obtain
the values to describe them is the orthogonal projection in mathematic
operation and the principle of equilibrium in mechanics law.
[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 45

The order of the tensor is identified by the number of the indexes the
object carries as a subscript in indicial notation. Zeroth-order tensors (scalar
numbers) carry no index, first-order (vector) carries 1, second-order (stress)
carries 2, and nth n. These indexes refer to the coordinates. Thus, when the
coordinate transforms, the components of the object change accordingly.
The order of the tensor is also the times of operations needed to get the
coordinate-transformed values, and hence the mathematical formula of each
of them has a different form, but these mathematical rules are all obtained
following the orthogonal projection rule.
The first high-level question is may need to be answered is why we would
need such a concept of tensor. This is because with it, the mechanics family
can have responsible objects, and all the objects follow the same rule to
behave, and thus all these objects can be evaluated consistently and correctly
when a transformation takes place. This enables us to do whatever necessary
transformations needed to get what we want.
The second high-level question is why we transform the coordinate. This
is because we often want to find the values of the tensor objects at one
particular direction, or find the critical (maximum, minimum, etc.) values
and their corresponding directions.
In summary, in mechanics of materials, we deal with tensors: scalars, vec-
tors (force vector, displacement vector, etc.), stresses, strains, elasticity, just
to name most important ones. All these variable objects need to satisfy the
equilibrium conditions. They are often described in reference to a coordinate
that is orthogonal (at least locally), so that we can quantify and operate on
them in a meaningful manner in numerical values or symbolic expressions.
When the coordinate transforms, all the values of these variables may change.
Therefore, there must be a rule that governs such changes, so that the entire
system behaves in a manner for consistent and correct quantification. The
objects may have different forms of mathematical representations, includ-
ing scalar (single number), vector (a stack of numbers), matrix (2D stack of
numbers), an elasticity (four-dimensional stack of numbers), but they all are
tensors because they all follow the same rule governing the transformation.
It is the orthogonal projection rule in mathematical operations and the
principle of equilibrium in mechanics law.

2.9 Transformation rules

2.9.1 Definition of coordinate transformation

As mentioned earlier, when we say coordinate transformation, we always
refer to coordinate rotation or type change of the coordinate system.
This is because when we study object changes, we are interested in the
[Link]
 46 Mechanics of Materials: Formulations and Solutions with Python

change at the same location in solids. Therefore, the transformation can only
be a rotation or type change because the translational position is fixed. We
also require that when the coordinate rotates, the entire coordinate system
rotates rigidly together.
For changes of objects with the location change, we shall use the usual
function concept in mathematics, which is a separate issue from coordinate
rotation. In this case, the object becomes a variable over space, and we
require all the variables being sufficiently smooth functions of the coordinates
for differentiations. Functions of variables are another important aspect to be
studied, independently of coordinate transformation. This is because many
governing law of mechanics are the result of the function behavior of the
variables.
To put it more clearly: the tensor concept deals with the coordinate trans-
formation, and the function concept deals with the variation of the object
along with the coordinate.

2.9.2 Representation of a tensor in a coordinate

In mechanics of materials, we use the following assumptions for the repre-
sentation of a tensor in a coordinate system.

1. Orthogonal coordinate: We use orthogonal coordinate systems, imply-

ing all the axes in the coordinate system are, at least locally, mutually
perpendicular.
2. Right-hand rule: We require also these axes to follow the conventional
right-hand rule: x-cross-y gives z. This is for consistence and convenience
purposes.

In this book, we will use the Cartesian coordinate as the default, which
is also followed in many Python modules. Other coordinate systems may be
used in a similar manner. Conversion between coordinate systems can be
done based on the so-called curvilinear coordinates that are at least locally
orthogonal.

2.9.3 Essential rule to obtain values on

a coordinate axis
Orthogonal projection rule: When a tensor is represented in coordinates,
we follow the orthogonal projection rule to compute the component values of
the tensor, in addition to the equilibrium law of mechanics. This means that
the component value of the tensor on an axis is the value obtained always
by orthogonally projecting the tensor onto the axis. When the coordinate

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 47

transforms, the new component values of the tensor is obtained also follow
the same orthogonal projection rule in the entire operation process.

2.10 Transformation matrix

2.10.1 Two-dimensional (2D) cases

When the coordinate rotates to a new one, the representation of the tensor
in the new coordinate changes. To quantify such changes, we must figure out
the relationship between these two, the original and the rotated, coordinate
systems and in precise mathematical formulas.
For easy visualization, consider first a 2D case. Let the original x with
two axes be denoted in conventional manner as x and y. The rotated new
coordinate x with axes x and y  , as drawn in Fig. 2.12. The prime symbol
stands for objects in the new coordinate.
From Fig. 2.12, we find the following relationship via simple geometric
examination following the orthogonal projection rule:

x = x cos θ + y sin θ = x cos(x , x) + y cos(x , y)

   
cos θ cos(90◦ −θ) (2.51)
y  = −x sin θ + y cos θ = x cos(y  , x) + y cos(y  , y)
   
cos(90◦ +θ) cos θ

Figure 2.12. A point with coordinate values of x and y in coordinate x, and with x ,
y  in coordinate x that is obtained by rotating coordinate x by θ. All the operations here
are based on the orthogonal projection rule.

[Link]
 48 Mechanics of Materials: Formulations and Solutions with Python

Equations (2.51) can be written in a matrix form as

    
x cos(x , x) cos(x , y) x
=
y cos(y  , x) cos(y  , y) y (2.52)
   
x T x

These cosines are called direction cosines. Equation (2.52) can be writ-
ten simply as follows:
x = Tx (2.53)
where x is a vector of the original and x is a vector of the rotated coordi-
nate. Matrix T is called transformation matrix. It is clear that matrix T
contains only the angular relationship of these axes of the two original and
new coordinates. For 2D cases, it can be written as follows, which is seen
more often in the literature:
 
cos θ sin θ
T= (2.54)
−sin θ cos θ

It is important to note that although the values of the components in

x and x are different, they both represent the same point in 2D space
(a position vector), as shown in Fig. 2.12.
We emphasize that all these expressions obtained following the orthog-
onal projection rule and using orthogonal coordinates, as shown in
Fig. 2.12. Due to this, T has the following property:
  
cos θ sin θ cos θ −sin θ
TT =
−sin θ cos θ sin θ cos θ
 
cos2 θ + sin2 θ −cos θ sin θ + sin θ cos θ
=
−sin θ cos θ + cos θ sin θ sin2 θ + cos2 θ
 
1 0
= =I (2.55)
0 1
This property of T is called unitary or orthogonality. It is invertible,
and its inverse transformation can be simply done using the transpose of T:
T−1 = T (2.56)
where superscript stands for transpose. It is also easy to confirm that
T T = I.
[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 49

Due to this property of T, the inverse transformation of Eq. (2.53)

becomes

x = T x (2.57)

2.10.2 General definition of transformation matrix

for 3D cases
With the explicit examination of 2D cases of rotation matrix for coordinate
transformation, we are now ready to have a general definition. For conve-
nience of description and formulation, we use indicial notation.
Consider a Cartesian coordinate x with axes x1 , x2 , and x3 . It rotates
to x with axes x1 , x2 , and x3 , and the values of the components of an
object (obtained also via the orthogonal projection rule) will change. We
shall discuss the general mathematic rule of transformation to compute the
components of the object in the new coordinate x . To this end, we define
the rotation matrix for coordinate transformation that bridges the original
and new coordinates.
Naturally, the rotation matrix for coordinate transformation, T, shall
contain the angular relationship of these axes of the two coordinates. This
must be true regardless of dimensionality. Based on the orthogonal projection
rule, T for 3D cases should be formed by the direction cosines of the angle
between the axes of the original coordinate x1 , x2 , and x3 and the rotated
coordinates x1 , x2 , and x3 .
⎡ ⎤
cos(x1 , x1 ) cos(x1 , x2 ) cos(x1 , x3 )
⎢       ⎥
⎢ a11 a12 a13 ⎥
⎢ ⎥
⎢cos(x2 , x1 ) cos(x2 , x2 ) cos(x2 , x3 )⎥
⎢
T=⎢       ⎥ (2.58)
⎥
⎢ a21 a22 a23 ⎥
⎢    ⎥
⎣cos(x3 , x1 ) cos(x3 , x2 ) cos(x3 , x3 )⎦
     
a31 a32 a33

It contains all the direction cosines of these axes of these two coordinate
systems. This definition is a natural extension from the 2D cases discussed in
the previous section. Here, we introduced indicial notation aij , i, j = 1, 2, 3,
to denote the entries of T. The right-top 2 × 2 sub-matrix is the same T
given in Eq. (2.52), which can be re-written as
 
a11 = cos θ a12 = sin θ
T= (2.59)
a21 = −sin θ a22 = cos θ

[Link]
 50 Mechanics of Materials: Formulations and Solutions with Python

2.10.3 Transformation via yaw, pitch, and roll in 3D*

Consider coordinate transformation in 3D by yaw, pitch, and roll defined as
follows to create the new coordinates x , y  and z  based on x, y and z:

• Yaw: rotate the Cartesian coordinate by θ about x-axis, which gives Tx .

• Pitch: rotate the Cartesian coordinate by β about y-axis, which gives Ty .
• Roll: rotate the Cartesian coordinate by γ about z-axis, which gives Tz .

By extending Eq. (2.59), these transformation matrices are written as

follows:
⎡ ⎤
cos θ sin θ 0
⎢ ⎥
Tz (θ) = ⎣− sin θ cos θ 0⎦ (2.60)
0 0 1
⎡ ⎤
cos β 0 − sin β
⎢ 0 ⎥
Ty (β) = ⎣ 0 1 ⎦ (2.61)
sin β 0 cos β
⎡ ⎤
1 0 0
⎢ ⎥
Tx (γ) = ⎣0 cos γ sin γ ⎦ (2.62)
0 − sin γ cos γ

Note that when rotating about the y-axis by the right-hand rule, angle
β should be negative.
When yaw, pitch, and roll are performed together, the total transforma-
tion matrix T can be obtained from these three, using matrix multiplication:

T = Tz (θ) Ty (β) Tx (γ)

⎡ Tz (θ) for yaw ⎤ ⎡ Ty (β) for pitch ⎤ ⎡ Tx (γ) for roll ⎤

cos θ sin θ 0 cos β 0 −sin β 1 0 0
⎢ ⎥⎢ ⎥⎢ ⎥
= ⎣−sin θ cos θ 0⎦ ⎣ 0 1 0 ⎦ ⎣0 cos γ sin γ ⎦
0 0 1 sin β 0 cos β 0 −sin γ cos γ
⎡ ⎤
cos β cos θ sin β sin γ cos θ + sin θ cos γ −sin β cos γ cos θ + sin γ sin θ
⎢ ⎥
= ⎣−sin θ cos β −sin β sin γ sin θ + cos γ cos θ sin β sin θ cos γ + sin γ cos θ ⎦
sin β −sin γ cos β cos β cos γ
(2.63)

We write the following code for future use to create this type of transfor-
mation matrices in Sympy for symbolic computations.
[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 51

1 def transf_YPRs( , about = 'z'):

2 '''Create a transformation matrix for coordinate transformation\
3 Input: , rotation angle, in Sympy \
4 about, the axis of the rotation is about \
5 Return: Sympy matrix of transformation matrix of shape (3,3)
6 '''
7 c, s = [Link]( ), [Link]( )
8
9 if about == 'z':
10 # Yaw: rotates about z by
11 T = [Link]([[ c, s, 0],
12 [-s, c, 0],
13 [ 0, 0, 1]])
14 elif about == 'y':
15 # Pitch: rotates about y by
16 T = [Link]([[ c, 0,-s],
17 [ 0, 1, 0],
18 [ s, 0, c]])
19 elif about == 'x':
20 # roll: rotates about x by
21 T = [Link]([[ 1, 0, 0],
22 [ 0, c, s],
23 [ 0,-s, c]])
24 else: # randomly generated unitary matrix as transformation matrix:
25 print("Rotation axis must be given: z, y, or x")
26
27 return T, about

Following are examples on how to use this code.

1 , , = symbols(" , , ") # define variables for angle of rotation

2 Tz = transf_YPRs( , about = 'z')[0]
3 Ty = transf_YPRs( , about = 'y')[0]
4 Tx = transf_YPRs( , about = 'x')[0]
5 print(f'Ty=, {Ty}, \nTx=, {Tx}')
6 Tz

Ty=, Matrix([[cos( ), 0, -sin( )], [0, 1, 0], [sin( ), 0, cos( )]]),

Tx=, Matrix([[1, 0, 0], [0, cos( ), sin( )], [0, -sin( ), cos( )]])

⎡ ⎤
cos (θ) sin (θ) 0
⎣− sin (θ) cos (θ) 0⎦
0 0 1

1 T = Tz@Ty@Tx
2 T

[Link]
 52 Mechanics of Materials: Formulations and Solutions with Python

⎡
cos (β) cos (θ) sin (β) sin (γ) cos (θ) + sin (θ) cos (γ)
⎢
⎣− sin (θ) cos (β) − sin (β) sin (γ) sin (θ) + cos (γ) cos (θ)
sin (β) − sin (γ) cos (β)
⎤
− sin (β) cos (γ) cos (γ) + sin ( ) sin ( )
⎥
sin (β) sin (θ) cos (γ) + sin ( ) cos ( ) ⎦
cos (β) cos (γ)

1 TTT = [Link](T.T@T) # Test on orthogonality (see next section)

2 print(TTT)

Matrix([[1, 0, 0], [0, 1, 0], [0, 0, 1]])

2.10.4 Property of transformation matrix, a unitary

matrix
Since aij are obtained by the orthogonal projection rule, T is a tensor. It has
a matrix form and is in general asymmetric. Since the axes of the coordinate
systems are kept mutually orthogonal, we shall have the following orthogonal
condition:
aij akj = δik (2.64)
where δij is the Kronecker delta defined as

1, if i = j
δij = (2.65)
0, if i = j
In Eq. (2.64), we see that index j is repeated which implies a summation:
aij akj = ai1 ak1 + ai2 ak2 + ai3 ak3 (2.66)

It is known as the Einstein notation of summation: any repeated index in

a term in an expression implying a summation over the range of the repeated
index. As the result of such a summation, the index will disappear and is
said to be contracted. Therefore, any letter can be used for a repeated index,
as long as it differs from other indexes in the term and the range is the same.
For example, aij akj is the same as aim akm , and they all result in the same
second-order tensor with indexes ik. For these reasons, a repeated index is
called a dummy index.
Indexes i and k in Eq. (2.64) are all single in a term in the equation and
cannot be contracted. They are called free-indexes. Clearly, one can use
any other symbol for these free-indexes, as long as it does not interfere with
other indexes and the range is the same. For example, aij akj = δik is the
same as arj asj = δrs .
[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 53

Since the coordinate rotates rigidly, and coordinate x is also orthogonal,

we have

aij aik = δjk (2.67)

Due to Eq. (2.64), the transformation matrix T shall satisfy the following
orthogonal condition in matrix form, as seen earlier:

TT = I or T T = I (2.68)

where I is a 3 × 3 identify matrix in R3 . Here, we follow the standard matrix

multiplication rule. Eq. (2.68) means that T is an unitary matrix. We have
seen this property explicitly for 2D case in the previous section. The unitary
property T implies that:

• The determinant of T (with the right-hand rule enforced) is 1.

• When it operates on a vector, it will not change the vector length.
• When it operates on a square matrix, it will not change the determinant
of the matrix. Therefore, it explains also that why the determinant of a
matrix is invariant under coordinate transformation. This can be proven
easily: because det(TM) = det(T) det(M) and T has a determinant of 1.
The same holds if T operates from the right.

When T acts on a tensor in x, the new components of the same tensor

in the rotated coordinate x can be obtained. The length of x and x is the
same. This can easily be proven using Eq. (2.17) as follows.
Consider an arbitrary vector v defined in x. A transformation matrix T
is used to transfer the coordinate x to x . The length of the vector v in x
is evaluated as
v 2 = v v = (Tv) Tv = v T

T v = v v = v2
(2.69)
I

The length of the vector is unchanged, which implies that the vector
is a first order tensor and its length will not change after the coordinates
transform. Note that the component values in v will change, but such a
change is to ensure the vector direction remained unchanged, when measured
in reference to the new coordinates. This vector property preservation
is ensured by the orthogonal projection rule, which results in a unitary T.
It may be worth to note that this general definition of tensors for mechan-
ics of materials applies, in theory, to arbitrary dimension. In n-dimensions,
the ranges of i and j become n, and the dimension for T will be n × n. The
property of T will still hold, as long as the transformation rules defined in
Section 2.9 are followed.
[Link]
 54 Mechanics of Materials: Formulations and Solutions with Python

In indicial notation, Eq. (2.53) can be written as

xi = aij xj (2.70)

And Eq. (2.57) can be written as

xi = aji xj (2.71)

The transpose of a matrix is equivalent to swapping the indexes of a

second-order tensor.

2.10.5 Numpy code for transformation matrices

We write a Python function that can be used to generate some typical trans-
formation matrices or a random transformation matrix. With such a matrix,
one can rotate the coordinate, resulting in description change of objects,
such as vector, defined in the coordinate. This function will be used later for
multiple times to perform coordinate transformation.

1 def transferM(theta, about = 'z'):

2 '''Create a transformation matrix for coordinate transformation\
3 Input theta: rotation angle in degree \
4 about: the axis of the rotation is about \
5 Return: numpy array of transformation matrix of shape (3,3)'''
6 from [Link] import ortho_group
7
8 n = 3 # 3-dimensional problem
9 c, s = [Link](np.deg2rad(theta)), [Link](np.deg2rad(theta))
10
11 if about == 'z':
12 # rotates about z by theta
13 T = [Link]([[ c, s, 0.],
14 [-s, c, 0.],
15 [0.,0., 1.]])
16 elif about == 'y':
17 # rotates about y by theta
18 T = [Link]([[ c, 0.,-s],
19 [ 0, 1.,0.],
20 [ s, 0., c]])
21 elif about == 'x':
22 # rotates about x by theta
23 T = [Link]([[ 1.,0., 0.],
24 [ 0., c, s],
25 [ 0.,-s, c]])
26 else: # randomly generated unitary matrix as transformation matrix:
27 T = ortho_group.rvs(dim=n) # Generate a random matrix
28 T[2,:] = [Link](T[0,:], T[1,:]) # Enforce the right-hand rule
29
30 return T, about

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 55

2.10.6 Examples: Unitary property of transformation

matrix

Let us examine in detail T matrices created using transferM(), and check

whether the orthogonality and the right-hand rule are observed.

1 from [Link] import ortho_group

2
3 [Link](8) # set a seed, random number can be reproduced
4 about, theta = 'random', 30. # rotation angle and axis
5 Tr, _ = transferM(theta, about = about) # Or [Link]()
6
7 print(f'Transformation tensor {theta:3.2f}°, w.r.t. {about}:\n{Tr}')
8 print(f"Determinant of Tr= {[Link](Tr):.6f}\n")
9 print(f"Is Tr orthogonal? \n{[Link](Tr.T)}")
10 print(f"Is Tr.T orthogonal? \n{(Tr.T).dot(Tr)}\n")
11
12 print(f"Check right-hand rule:")
13 print(Tr[0,:],[Link](Tr[1,:], Tr[2,:])) #Check right-hand rule: 1=2X3
14 print(Tr[1,:],[Link](Tr[2,:], Tr[0,:])) #Check right-hand rule: 2=3X1
15 print(Tr[2,:],[Link](Tr[0,:], Tr[1,:])) #Check right-hand rule: 3=1X2

Transformation tensor 30.00°, w.r.t. random:

[[ 0.0411 0.9184 0.3935]
[-0.6253 -0.2835 0.7271]
[ 0.7793 -0.276 0.5626]]
Determinant of Tr= 1.000000

Is Tr orthogonal?
[[ 1. 0. -0.]
[ 0. 1. -0.]
[-0. -0. 1.]]
Is Tr.T orthogonal?
[[ 1. -0. -0.]
[-0. 1. 0.]
[-0. 0. 1.]]

Check right-hand rule:

[0.0411 0.9184 0.3935] [0.0411 0.9184 0.3935]
[-0.6253 -0.2835 0.7271] [-0.6253 -0.2835 0.7271]
[ 0.7793 -0.276 0.5626] [ 0.7793 -0.276 0.5626]

A coordinate transformation matrix is unitary, implying that its inverse

is its transpose. Let us confirm.

1 print(f"Is Tr unitary? {[Link]([Link](Tr), Tr.T, atol=1e-04)}")

Is Tr unitary? True

[Link]
 56 Mechanics of Materials: Formulations and Solutions with Python

2.10.7 Examples: Basic property preservation by

transformation matrix

Let us examine basic property preservation feature of a transformation

matrix T. Consider an arbitrary vector v, the basic property is its length
and direction. Computing the length of a vector is easy, and the value is
unique. However, when computing the angle of the vector care is needed
because cosine function is multi-valued, depending on which quadrant the
vector is in. In the following code, we set a vector in the first quadrant with
an randomly generated angle θx . When rotating the coordinate, we use a
rotation angle that satisfies θrotation ≤ θx . With this setting the computa-
tions of angles become unique. If such a setting is changed, care is needed
in computing the values of angles. With this said, the following code is self-
explanatory and easy to follow.

1 _x = [Link]()*90. # a random angle in [0, 90.]

2
3 # create a vector in x-y coordinate using _x:
4 v = [Link]([[Link](np.deg2rad( _x)), [Link](np.deg2rad( _x)), 0.])
5 v = [Link]()*v # make it with arbitrary length
6 print(f'Original vector v = {v} \n its length = {[Link](v)}')
7 print(f'Direction cosine of v = {v/[Link](v)}')
8
9 # angle w.r.t the base coordinate x where v is defined originally:
10 _x = [Link](v[0]/[Link](v))*180/[Link] # unique under the setting
11 print(f' of v w.r.t its base coordinate x = { _x:.4f}∘')
12
13 # rotate the base coordinate x-y about the z-axis to form x'-y' coords:
14 # rotation would be in [0, _x]:
15 about = 'z'; _rotation = [Link]()* _x
16
17 Tz, _ = transferM( _rotation, about = about)
18 vp = Tz@v
19 print(f'Transformed v = {vp}, \n its length = {[Link](vp)}')
20 print(f'Direction cosine of Tv = {vp/[Link](vp)}')
21
22 # angle w.r.t the rotated coordinate x':
23 _xp=[Link](vp[0]/[Link](vp))*180/[Link] # unique under the setting
24 print(f" of v' w.r.t rotated coordinate x' = { _xp:.4f}∘")
25
26 # angle w.r.t the base coordinate x where v is defined originally:
27 _x_= _xp+ _rotation
28 print(f" of v' w.r.t its base coordinate x = { _x_:.4f}∘")
29 print(f"Is the angle unchanged? {abs( _x_- _x)<1.e-6}")
30 print(f"Is the length unchanged? {abs([Link](v)-[Link](Tz@v))<1.e-6}")

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 57

Original vector v = [0.2397 0.3523 0. ]

its length = 0.4260917704761922
Direction cosine of v = [0.5625 0.8268 0. ]
of v w.r.t its base coordinate x = 55.7714∘
Transformed v = [0.3281 0.2719 0. ],
its length = 0.4260917704761921
Direction cosine of Tv = [0.77 0.6381 0. ]
of v' w.r.t rotated coordinate x' = 39.6493∘
of v' w.r.t its base coordinate x = 55.7714∘
Is the angle unchanged? True
Is the length unchanged? True

It is clear that both the basic properties of the vector, original length and
direction, remain unchanged under T. Since the random angle is set random,
readers may execute the code for multiple times, and think about the results
obtained.

2.10.8 Example: Transformation matrix on arbitrary

matrices
Transformation matrix is unitary, and a unitary matrix does not change
the determinant of a matrix it acts on. We shall test this using a randomly
generated matrix, which is a lined-up vectors.

1 [Link](9) # set a seed for reproducibility

2 n = 3
3 M = [Link](n,n) # Generate a random matrix
4 print(f'Randomly generated matrix M: \n{M}')
5 print(f'Determinant of M = {[Link](M)}')
6 print(f'Determinant of Tr@M = {[Link](Tr@M)}')

Randomly generated matrix M:

[[ 0.0011 -0.2895 -1.1161]
[-0.0129 -0.3784 -0.4811]
[-1.5173 -0.4909 -0.2407]]
Determinant of M = 0.4230331712701571
Determinant of Tr@M = 0.4230331712701568

It is seen that the determinant of the matrix is unchanged, as proven

earlier.
Let us check the effect of T on the eigenvalues of a matrix. We first look
at a symmetric matrix.

[Link]
 58 Mechanics of Materials: Formulations and Solutions with Python

1 Ms = (M + M.T)/2 # random symmetric matrix

2 print(f'Symmetric random square matrix Ms: \n{Ms}')
3 print(f'Eigenvalues of Ms = {[Link](M)[0]}')
4 print(f'Eigenvalues of Tr@Ms = {[Link](Tr@M)[0]}')

Symmetric random square matrix Ms:

[[ 0.0011 -0.1512 -1.3167]
[-0.1512 -0.3784 -0.486 ]
[-1.3167 -0.486 -0.2407]]
Eigenvalues of Ms = [ 1.2042 -1.603 -0.2191]
Eigenvalues of Tr@Ms = [ 0.5069 -1.5005 -0.5562]

It is shown that the eigenvalues of Tr@M are different from those of M.

This is because M is a second-order tensor, we need to let T act on both
sides in a “symmetric” manner. With such a two-side application of T, the
eigenvalues of the resultant new matrix will be unchanged because T is
unitary. This is shown in the following code, using an asymmetric matrix.
1 M = [Link](n,n) # random asymmetric square matrix
2 print(f'Random asymmetric square matrix M: \n{M}')
3 print(f'Eigenvalues of M = {[Link](M)[0]}')
4 print(f'After forward & inverse transformation: \n{Tr@[email protected]}')
5 print(f'Eigenvalues of Tr@[email protected] = {[Link](Tr@[email protected])[0]}')

Random asymmetric square matrix M:

[[-0.6479 0.6359 1.7401]
[ 0.2967 0.7075 1.8228]
[ 0.4308 1.5427 -0.9007]]
Eigenvalues of M = [ 2.0292 -0.7299 -2.1405]
After forward & inverse transformation:
[[ 1.7428 0.3881 0.7523]
[-0.3824 -2.188 -0.9215]
[ 1.195 0.136 -0.396 ]]
Eigenvalues of Tr@[email protected] = [ 2.0292 -0.7299 -2.1405]

With a two-side application of T on a matrix, the eigenvalues of the resul-

tant new matrix are unchanged, even if the matrix is a random asymmetric
matrix.
This two-side action of T will be performed to stress and strain tensors
in later Chapters.

2.11 Rotation matrix

2.11.1 Derivation for 2D cases

Readers may be familiar with the rotation matrix, often denoted as R. It is
used to rotate a vector in a coordinate system counterclockwise, which results
in new components of the vector in the same coordinate system. Assume the
[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 59

Figure 2.13. A vector with coordinate values of x and y in a coordinate system. It is

rotated counterclockwise by θ, which results in new vector with coordinate values x , y  in
the same coordinate system. All the operations here are based on the orthogonal projection
rule.

vector, before rotation, is denoted as v = (x, y) in the coordinate system.

After it is rotated, it becomes v = (x , y  ), as shown in Fig. 2.13. The prime
symbol stands for the new coordinate values.
Upon simple analysis based on Fig. 2.13, we have

x = r  cos(α + θ) = r  cos 
 α cos θ − r sin
 α sin θ
x y

= x cos θ − y sin θ
(2.72)
y  = r  sin(α + θ) = r  sin 
 α cos θ + r cos
 α sin θ
y x

= x sin θ + y cos θ

where α is the angle of the vector with respect to the x-axis, and θ is the
rotation angle to the vector. The foregoing equation can be written in matrix
form:
    
 x cos θ −sin θ x
v =  = = Rv (2.73)
y sin θ cos θ y
   
v R v

where R is the rotation matrix given as

 
cos θ −sin θ
R= (2.74)
sin θ cos θ

[Link]
 60 Mechanics of Materials: Formulations and Solutions with Python

2.11.2 Relationship between transformation and

rotation matrix
Comparing Eqs. 2.54 and 2.74, it is clear that
R = T (2.75)
This is simply because T rotates the coordinate with the object (vector)
remaining still, but R rotates the object (vector) within the same coordinate.
Thus, the effects of their actions are mutual reverse. Thus, Eq. (2.75) is valid
for any higher dimensions.
Due to Eq. (2.75), the rotation matrix has all the properties of the trans-
formation matrix.
One may note that deriving Eq. (2.74) is in fact easier than deriv-
ing Eq. (2.54) for 2D cases. However, for higher dimensions, the trans-
formation matrix T can be much easier to obtain because it can be
systematically obtained using the direction cosine concept, as shown in
Section 2.3.2. A quite comprehensive article on rotation matrices can be
found at the wikepage on Rotation matrix ([Link]
Rotation matrix).

2.12 Function, an object varying over space

Note that when we talk about coordinate transformations we stand at a

point in solid and examine the change of the components of an object in
mechanics of materials.
We now discuss about functions that deal with the variation of an object
with respect to changes of coordinate values. In a problem of mechanics
materials, a variable, such as a component of displacement or stress or strain
tensor, can change from location to location. This means that it can vary
with the coordinates in a coordinate system. The variation of a variable with
respect to coordinates is called function of coordinates. It can be defined as
an operator f () that takes inputs of coordinates in an n-dimensional space
Rn and produces a new number in R corresponding to the coordinates. It is a
mapping from Rn → R. In the 3D Cartesian coordinate system, for example,
the function is often expressed as f (x) or f (x, y, z), where x, y, and z are
the coordinates as the independent variables of the function f [1].
In higher dimensions, there are multiple components. Each of which can
be a function of coordinates. For a displacement vector in 3D solids, for exam-
ple, it has 3 scalar component functions. It thus forms a vector function of 3
scalar functions. For a stress tensor in 3D solids, we have 9 (6 independent)
scalar functions.
[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 61

Note that the function for an object is a concept for accounting for the
variation of the object along the coordinates. It is a separate concept of the
coordinate transformation. Therefore, a tensor at a location is still a tensor
at any other location in the coordinate system, as long as the same rule is
followed for coordinate transformation at locations. However, operations to
a tensor object with respect to the coordinates can produce a new tensor
object of different order. We will discuss about this in great detail because it
happens all the time in mechanics formulation. In our discussion, we assume
all the scalar functions involved are sufficiently continuous and smooth and
hence differentiable to the required order, allowing us to take limits when
necessary.

2.13 Gradient of a scalar function, a vector

A function is a device that takes in an input (or argument) and produces
an output, as defined in Ref. [1]. When the argument is a coordinate, the
function determines how an variable function changes with the coordinate.
For example, a function of a displacement component gives the variation of
the component against the coordinate. In general, the argument can be a
scalar, a vector, or even a matrix. The function itself can also be a scalar, a
vector or a matrix.
Gradient of a scalar function with scalar number argument converts
the scalar number (0th first order tensor) to a vector (first order tensor).
Consider a scalar function defined in a domain in the coordinate system.
A typical such a function is one of the components of a displacement vector
or stress tensor.

2.13.1 Definition
The gradient of an arbitrary scalar differentiable function f (x) of the inde-
pendent variables (that is a vector) becomes a vector of three functions
defined as
⎡ ⎤ ⎡ ∂f (x) ⎤
∂
∂x
⎢ ∂x ⎥ ⎢ ∂f (x) ⎥
∇f (x) := ⎢ ∂ ⎥
⎣ ∂y ⎦ f (x) = ⎢
⎣ ∂y ⎦
⎥ (2.76)
∂ ∂f (x)
∂z
  ∂z
∇

where ∇ (pronounced “nabla”) stands for gradient operator, and has a

vector form. The components in the resultant vector ∇f (x) are, respectively,
the partial derivatives of the scalar function with respect to these 3 coordi-
nates. The shape of ∇f (x) is the same as that of x. We see now, f is a scalar
[Link]
 62 Mechanics of Materials: Formulations and Solutions with Python

number, a 0th order tensor, but its gradient becomes a vector of first order
tensor! It can be subjected to the same rule of coordinate transformation for
a vector.

2.13.2 Python demonstration

Let us use the following Python codes to demonstrate this. We write first a
function to compute the gradient of a given scalar function.

1 def grad_f(f, X):

2 '''Compute the gradient of a given scalar function f.
3 input: f, [Link].
4 X, coordinates, array-like of symbols.
5 return: grad_f, [Link], a matrix of 2×1 or 3×1 or len(X)×1.
6 ## Example:
7 x, y, z = [Link]("x, y, z")
8 X = [x, y, z]
9 f = 8*x**2 + 5*x*y -2*y + 3*y**2 + 2*z**2 - 5*z
10 f_g = grad_f(f, X)
11 f_g
12 '''
13 gradf = [Link]([[Link](f,xi) for xi in X])
14 return gradf

1 # Define an arbitrary scalar function of coordinates

2 x, y, z = symbols("x, y, z")
3 X = Matrix([x, y, z])
4
5 f = Function('f')(x, y, z)
6 #f = 8*x*[Link](y)+5*x**2-2*y*[Link](x)+3*y**2-5*z # alternative
7 f = 8*x**2 + 5*x*y -2*y + 3*y**2 + 2*z**2 - 5*z
8 f # print out

8x2 + 5xy + 3y 2 − 2y + 2z 2 − 5z

This function f takes any value of x, y, and z in R3 , and produces a single

value in R. For example,

4.375
[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 63

The gradient of the scalar function becomes a vector of 3 functions:

⎡ ⎤
16x + 5y
⎣5x + 6y − 2⎦
4z − 5

⎡ ⎤
16x + 5y
⎣5x + 6y − 2⎦
4z − 5
It is a vector because of the definition given in Eq. (2.76). Each of which of
3 functions in the vector is the derivative of f (x) with respect to, respectively,
x, y, and z. The same can be obtained using [Link]ff() method.

1 # Or simply use:
2 [Link](X) # This produces a column vector of 3 functions

⎡ ⎤
16x + 5y
⎣5x + 6y − 2⎦
4z − 5

1 print(f_g.shape,[Link]) # The shapes of f_g & the coordinate vector

(3, 1) (3, 1)

It is seen that the shape of the gradient of the scalar function is the same
as that of the coordinate vector.
The vector of functions can be simplified by substituting some of the
coordinates with fixed values.
1 f_gs = f_g.subs({x:1/2})
2 f_gs

⎡ ⎤
5y + 8.0
⎣6y + 0.5⎦
4z − 5
[Link]
 64 Mechanics of Materials: Formulations and Solutions with Python

It is seen now f gs is still a vector of the same shape, but the components
are no longer functions of x. This is because of the use of subs() function,
which simplifies the functions at a given x coordinate. We can simplify it
further.

1 f_gs = f_gs.subs({y:2}) # fix y

2 f_gs

⎡ ⎤
18.0
⎣ 12.5 ⎦
4z − 5

We can also get the numerical values of this gradient vector by fixing z:

1 f_gv = f_gs.subs({z:1/4}).evalf() # get the vector values

2 printM(f_gv, 'Values of function gradient',n_dgt=5) # printed pretty

Values of function gradient

⎡ ⎤
18.0
⎣ 12.5 ⎦
−4.0

Remark: The gradient of a scalar function defined in the coordinate

becomes a vector. It has the same shape as the coordinate vector.

2.14 Gradient of a vector of function, a matrix

2.14.1 Definition
Gradient of a vector of functions converts a vector (first order tensor) to a
matrix (second-order tensor).
Consider a differentiable vector field in 3D Cartesian coordinates defined
by a vector of three scalar functions:
⎡ ⎤
fx (x)
⎢ ⎥
f (x) = fx (x)i + fy (x)j + fz (x)k = ⎣fy (x)⎦ (2.77)
fz (x)

where i, j and k are, respectively, the unit vector on x, y and z axes, fi (x)
(i = x, y, z) are the components of the vector function f (x). The vector func-
tion creates a vector field in the domain where these functions are defined.
[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 65

Each of the component function is a function of the coordinates. The gradient

of a vector of functions is defined using the outer product of ∇ and f (x):
⎡ ∂ ⎤ ⎡ ⎤
∂x fx (x)
⎢∂⎥ ⎢ ⎥
∇f (x) := ⎢ ⎥ ⎢
⎣ ∂y ⎦ ⊗ ⎣fy (x)⎦
⎥
∂ fz (x)
∂z  
 
∇ f (x)
⎡ ∂ ⎤
∂x
⎢ ∂ ⎥  (2.78)
=⎢ ⎥
⎣ ∂y ⎦ fx (x) fy (x) fz (x)
∂
∂z
⎡ ∂ ∂ ∂ ⎤
∂x fx (x) ∂x fy (x) ∂x fz (x)
⎢∂ ⎥
=⎢
⎣ ∂y fx (x)
∂
∂y fy (x)
∂ ⎥
∂y fz (x)⎦
∂ ∂ ∂
∂z fx (x) ∂z fy (x) ∂z fz (x)

2.14.2 Python examples

Let us evaluate the gradient of a vector function, assume each of the com-
ponent function is differentiable, such as the vector functions given in the
following:

1 vf = Matrix([y*[Link](x), z*[Link](y), x*[Link](z)])

2 vf

⎡ ⎤
y sin (x)
⎣ z sin (y) ⎦
x cos (z)

Using [Link] vec() from the continuum mechanics module, the gra-
dient of the vector function can be found as follows:

1 vf_g_ = vector.grad_vec(vf) # produces 3 column vectors of 3 functions

2 vf_g_

⎡ ⎤
y cos (x) 0 cos (z)
⎣ sin (x) z cos (y) 0 ⎦
0 sin (y) −x sin (z)
[Link]
 66 Mechanics of Materials: Formulations and Solutions with Python

It is a 3 by 3 matrix. This is because we have three component functions,

and the gradient of each of the function is a vector by the definition of
Eq. (2.76). These three vectors can form a matrix.

2.14.3 Code function for gradient of vector functions

Note that [Link] vec() works only for 3D vector functions. One may
use [Link] vf() which is given in the following code function. It works for
both 2D and 3D vector functions.

1 def grad_vf(vf, X):

2 '''Compute the gradient of a given vector function, vf
3 input: vf, [Link], a column vector of 2 or 3.
4 X, coordinates, array-like of symbols
5 return: grad_f, [Link], a matrix of 2×1 or 3×1 or len(X)×1.
6 ## Example:
7 x, y, z = [Link]("x, y, z")
8 vf = [Link]([y*[Link](x), z*[Link](y), x*[Link](z)])
9 vf_g = grad_vf(vf, X)
10 '''
11 grad_f = [Link]([[Link](vf,xi).T for xi in X])
12 return grad_f

1 # 3D vector function example:

2 vf_g = grad_vf(vf, X) # or gr.grad_vf(vf)
3 vf_g

⎡ ⎤
y cos (x) 0 cos (z)
⎣ sin (x) z cos (y) 0 ⎦
0 sin (y) −x sin (z)

1 # 2D vector function example:

2 vf2d = Matrix([y*[Link](x), z*[Link](y)])
3 vf2d_g = gr.grad_vf(vf2d, X[:2]) # or gr.grad_vf(vf2d, X[:2]
4 vf2d_g

 
y cos (x) 0
sin (x) z cos (y)

When the independent variables are set with fixed values, we shall obtain
a matrix with fixed values.

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 67

1 vf_gs = vf_g.subs({x:1/2,y:2,z:1/4}).evalf()
2 vf_gs
3 printM(vf_gs, 'Gradient of vector at a fixed location',n_dgt=5)

Gradient of vector at a fixed location

⎡ ⎤
1.7552 0 0.96891
⎣0.47943 −0.10404 0 ⎦
0 0.9093 −0.1237
It is still a matrix. The entries become fixed numbers, as expected.
Gradient of vector of functions will be used in deriving the strains in solid
using displacement vectors.

2.15 Divergence of a vector function, a scalar

2.15.1 Definition
The divergence of a vector of functions produces a scalar function. It converts
a first order tensor object to a 0th order one.
Consider a differentiable vector field in 3D Cartesian coordinates defined
by a vector of three scalar functions, given in Eq. (2.77). The divergence of
a vector of functions is defined as the dot product of ∇ with f :

div f := ∇ · f
⎡ ⎤
  fx
∂ ∂ ∂ ⎢ ⎥
= ⎣ fy ⎦
∂x ∂y ∂z
fz
(2.79)
∂fx ∂fy ∂fz
= + +
∂x ∂y ∂z
∂fi
= = fi,i
∂xi

This is a scalar function because of the contraction of index i. The dot in

Eq. (2.79) stands for the dot product of vectors: a del operator vector and
a given vector of functions. This is a measure of the vector diverges away
from a point. The divergence carries different physical meaning depending
on application. For example,

• If the vector of functions is concentration of a species, it is the spatial rate

of loss of mass at that point. The compensation to that loss should be the
birth rate.
[Link]
 68 Mechanics of Materials: Formulations and Solutions with Python

• If the vector is temperature, it is the spatial loss of heat at that point.

The compensation to that loss should be the heat source.
• If the vector has stress components (all in the same direction) at a point,
it is the loss of stresses in that direction. Such a loss must be compensated
externally by an applied body force in that direction. This leads to equi-
librium at the point together with the externally applied total body force
(force density) there.

2.15.2 Python code and examples

Readers may use [Link] vf() to compute the divergence of a vector function,
which is given in the following.

1 def div_vf(vf, X):

2 '''Compute the divergence of a vector function vf.
3 input: vf, vector function in, [Link]
4 X, vector of in dependent variables, [Link]
5 return: div_f
6 '''
7 div_f = [Link](*[[Link](vf[i], X[i]) for i in range(len(X))])
8 return div_f # divergence of a vector function

Let us demonstrate this using the vector of three simplest functions. Con-
sider first a vector with three stress components in a direction that are all
constants (independent of the coordinates x, y, and z):

1 x, y, z = symbols('x, y, z') # define coordinates

2 X = Matrix([x, y, z])
3
4 a, b, c = symbols('a, b, c') # define constants
5 vc = Matrix([a, b, c]) # constant vector function
6 vc

⎡ ⎤
a
⎣b⎦
c

1 # compute the divergence of a vector function

2 div_f = div_vf(vc, X) # Or use: [Link](vc)
3 div_f

0
The divergence of the constant stress vector is zero. This is because a
field of constant stays still and no loss will occur. This means also that the
[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 69

body force in that direction must be zero. We will see an example in the
next chapter.
Consider now a vector with three stress components in a direction that
are linear functions of x, y, and z:

1 a1, b1, c1 = symbols('a1, b1, c1') # define constants

2 a2, b2, c2 = symbols('a2, b2, c2') # define constants
3 a3, b3, c3 = symbols('a3, b3, c3') # define constants
4
5 vl=Matrix([a1*x+b1*y+c1*z, a2*x+b2*y+c2*z, a3*x+b3*y+c3*z]) # linear
6 vl

⎡ ⎤
a1 x + b1 y + c1 z
⎣a2 x + b2 y + c2 z ⎦
a3 x + b3 y + c3 z

1 # compute the divergence of a vector function

2 div_f = div_vf(vl, X) # Or use: [Link](vl)
3 div_f

a1 + b2 + c3

Thus, if a vector of the stresses in a direction that are all linear functions,
its divergence is a constant. This means that the body force in that direc-
tion must be a constant, which is the sum of the slopes in the respective
coordinates. Therefore, if a solid is subjected to a constant gravity force in
a direction, the stresses in that direction is linear.
Generally speaking, the function of stresses is one order higher than the
body force function.
This relationship of stress vector divergence and body force may be used
to check stress solutions. We shall do so in some of the examples in the
following chapters. Most importantly, it offers a convenient way to derive
the equilibrium equations for solids.

2.16 Curl of a vector function, a vector

2.16.1 Definition
Consider a vector function f defined in Eq. (2.77). Using the gradient oper-
ator vector ∇, the curl of a vector function can be defined in a similar way
[Link]
 70 Mechanics of Materials: Formulations and Solutions with Python

as the cross product of two vectors. Using Eq. (2.44), we can write:
 
 i j k
 
∂ ∂ ∂
× f :=  ∂x
∇ ∂y ∂z 

 
curl f  fx fy fz 
⎡ ∂ ∂ ⎤⎡ ⎤
0 − ∂z ∂y f
⎢ ∂ ⎥ ⎢ x⎥
=⎢⎣ ∂z 0 − ∂x ⎥
∂
⎦ ⎣fy ⎦ (2.80)
∂
− ∂y ∂
0 fz
∂x
⎡ ∂f ∂fy
⎤
z
−
⎢ ∂y ∂z
⎥
⎢ ∂fx ∂fz ⎥
= ⎢ ∂z − ∂x ⎥
⎣ ⎦
∂fy ∂fx
∂x − ∂y

As shown, curl f is a vector. It is a measure of the components of rate

of rotation or rate of circulation in the coordinate directions at a point
in a vector function field. In mechanics of materials, a typical example is
a bar subject to pure torsion in the axial direction. The rate of the twist
angle in the axial direction of the bar is proportional to the curl of the shear
stress vector function on the cross-section of the bar. We will discuss this in
Chapter 9. Here, let us take a look at the curl of two simple vector functions
created earlier: the constant vector function and the linear vector function,
which can offer some intuitive understanding of the curl of a vector function
field.

2.16.2 Python code and examples

1 def curl_vf(vf, X):
2 '''Compute the curl of a given vector function vf w.r.t X.
3 For 3D only.
4 '''
5 curl_f = [Link]([[Link](vf[2], X[1]) - [Link](vf[1], X[2]),
6 [Link](vf[0], X[2]) - [Link](vf[2], X[0]),
7 [Link](vf[1], X[0]) - [Link](vf[0], X[1])])
8 return curl_f

1 curl_F = curl_vf(vc, X) # Or use: [Link](vc)

2 curl_F # Curl of the vector function

[Link]
 Preliminaries: Number, Vector, Tensor, and Functions 71

⎡ ⎤
0
⎣0⎦
0
As shown, 1. the curl of a vector function is a vector; 2. when the vector
function field is a constant, the curl will be zero because there is no rotation
or circulation at all at any point in a constant field.
1 curl_F = curl_vf(vl, X) # Or use: [Link](vl)
2 curl_F # Curl of the vector function

⎡ ⎤
b3 − c2
⎣−a3 + c1 ⎦
a2 − b1
For linear vector functions, the curl will be constant in all three directions
because these three vector functions vary linearly in its own way, creating
rotations at any point in the field. The condition for linear vector function
field has no rotation with respect to x-axis would be b3 = c2 , no rotation
with respect to y-axis is a3 = c1 , and no rotation with respect to z-axis is
a2 = b1 . The proof is given in the following codes.
1 vl_0curl = Matrix([a1*x+a2*y+a3*z, a2*x+b2*y+c2*z, a3*x+c2*y+c3*z])
2 vl_0curl # linear vector function without rotation

⎡ ⎤
a1 x + a2 y + a3 z
⎣ a2 x + b2 y + c2 z ⎦
a3 x + c2 y + c3 z

1 curl_vf(vl_0curl, X) # Or use: [Link](vl_0curl)

⎡ ⎤
0
⎣0⎦
0

This is a curl-free field. The field described in this set of specially designed
linear vector functions does not have rotation or circulation.

2.17 Remarks
In concluding this chapter, we mention a few high level remarks.
1. Objects in solid mechanics are constructed as tensors. The scalar variables
are of 0th order, vectors are first order, and matrices are second order,
[Link]
 72 Mechanics of Materials: Formulations and Solutions with Python

etc. The rules for the construction and transformations are the equilib-
rium in mechanics law and orthogonal projection in mathematical
operations. Thus, all these objects can behave at any point in solid in a
consistent and predictable manner.
2. We proved and demonstrated via code example that the basic proper-
ties of a vector (first order tensor), the length and direction, remains
unchanged under coordinate transformation. The component values in v
will change to ensure the vector direction remaining unchanged, when
measured in reference to the new coordinates. This basic property
preservation is ensured by the unitary property of T. This applies to
both force and displacement vectors to be discussed in the next chapter.
3. We will see also the similar basic property preservation feature of a
stress or strain tensor (second order) under coordinate transformation.
Their principal stresses and strains (and hence the Mohr circles) remain
unchanged under coordinate transformation using T.
4. Functions are in general devices that take objects and produce new
objects. Objects in solid mechanics vary over the coordinates defined for
the solid, and these can be expressed in forms of functions with coor-
dinates as the independent variables. Mathematical operations, such as
different forms of differentiations, are closely related to mechanics laws.
Note also that integrations are frequently used in solid mechanics.
Examples include derivation of displacements from strain/stresses, and in
evaluation of strain energy, etc. Readers may review the concepts and tech-
niques for integration detailed in Ref. [6]. We will use these in later relevant
chapters.
All these will be elaborated in great detail in the following chapters.

References
[1] G.R. Liu, Numbers and Functions: Theory, Formulation, and Python Codes. World
Scientific, New Jersey, 2024.
[2] G.R. Liu, Machine Learning with Python: Theory and Applications. World Scientific,
New Jersey, 2023.
[3] G.R. Liu and S.S. Quek, The Finite Element Method: A Practical Course.
Butterworth-Heinemann, 2013.
[4] G.R. Liu and T.T. Nguyen, Smoothed Finite Element Methods. Taylor and Francis
Group, New York, 2010.
[5] G.R. Liu, Mesh Free Methods: Moving Beyond the Finite Element Method. Taylor
and Francis Group, New York, 2010.
[6] G.R. Liu, Calculus: A Practical Course with Python, World Scientific, New Jersey,
2025.

[Link]
 Chapter 3

Forces and Stresses

Let us import the necessary external modules or dependencies for later use
in this chapter.
1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.
2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 init_printing(use_unicode=True) # For latex-like quality printing
11 np.set_printoptions(precision=4,suppress=True) # Digits in print-outs

3.1 Introduction
A force is a typical vector familiar to readers in almost all disciplines, not
just in mechanics of materials. When a solid material is subject to the action
of a force, stresses may result in the material. This chapter examines the
force-stress relations.
Types of forces: There are mainly two types of force in terms of its spatial
distribution features: concentrated force (or point force), and distributed
force.
A concentrated force acts at a point in a solid material or at a point
on the surface of a solid material. It carries a force unit, typically Newton
denoted as [N ].

[Link]
73
74 Mechanics of Materials: Formulations and Solutions with Python

A distributed force acts over area or volume in three-dimensional (3D)

solids, or volume of two-dimensional (2D) solids, or a line in 2D solids, or
volume of one-dimensional (1D) solid. It is also often called body force. If
it is over a volume, the unit is typically [N/m3 ] with [m] (meter) being the
typical length unit. If it is over an area, the unit can be [N/m2 ] and over a
line/curve [N/m].
An often encountered distributed force is the pressure on the surface of
a solid, which can be exerted by air, water, or granular materials. Forces
distributed over an volume, include the gravity force, centrifugal force, and
electromagnetic force over solids.
Note that truly concentrated or point force does not exist in practical
applications in engineering. It is an idealized case of a distributed force over
a very small area or volume. For conveniences in analysis, we lump it to a
point, and treat it as a concentrated or point force, when our interest is at
a distance away from it.
To study how a force is related to stress, let us first look at a simple
1D case.

3.2 Example: A bar subjected to a point force

Consider a solid with simple geometry of a straight bar with uniform cross-
section made of isotropic linear elastic material. Its dimension of the cross-
section is far smaller than its length L. The bar is fixed at the left-end, and
subjected to a force at the right-end in the longitudinal direction denoted
by the 1D coordinate x, as shown in Fig. 3.1.

Figure 3.1. A uniform bar with a cross-sectional area of Ax fixed at its left-end and
subjected to a force fx at its right-end. One-dimensional coordinate x is taken along the
axial direction of the bar. When a cut is made to expose the cross-section surface, we see
the stress σxx .
 Forces and Stresses 75

Under these conditions, we can assume that stress within the bar is uni-
form over its cross-section. The stress can be calculated as

fx
σxx = (3.1)
Ax

where fx is the point force in the x-direction. It is applied at a fixed loca-

tion: x = L and Ax is the area of the cross-section that is perpendicular
to its axial axis x. Thus, the stress on the cross-section has a unit of typ-
ically [N/m2 ]. The stress distribution along the x coordinate is assumed
uniform.
Note here that the stress σxx carries two subscripts. The first x stands
for the surface direction of the cross-section on which the stress is on, and
the second x stands for the direction of the stress. We note:

Remark: Stress in solid material depends on both direction of the force

that causes the stress, and the direction of the surface on which the stress
is on. It carries two subscripts to fully denote its feature.

With this conventional use of subscripts, the stress is precisely described

in a given coordinate system. We call σxx a normal stress because its
direction is normal to the cross-section surface.
If our concern is confined to 1D problems, we can drop the two subscripts
and denote the stress simply as σ. One may see this notation often in the
literature. For us, however, the concern in this chapter is 3D in general, and
1D is just a special case to start our discussion. We shall stick on the two
subscripts notation. Also, in general stresses are a second order tensor, and
shall always carry two subscripts. We drop the subscripts and use simple
notation of σ only when there is no possibility of misleading.
For this 1D case, we have the simple force-stress relationship Eq. (3.1).
If the bar is uniform in cross-section, Ax is constant. Since fx is at a fixed
location, it is a constant too (does not change with x). Therefore, σxx is also
constant. Its derivative with respect to x will be zero.

dσxx
=0 (3.2)
dx

Note that since our problem is 1D, the shape of the cross-section of the
bar is immaterial. For example, if the cross-section is square instead of cir-
cular, the foregoing equations shall still hold. All we need is the bar is long
compared to its largest cross-sectional dimension, so that the stress can be
assumed uniformly distributed over the cross-section.
76 Mechanics of Materials: Formulations and Solutions with Python

3.3 Example: A bar subjected to a distributed force

3.3.1 Derivation of equilibrium equation

Assume the external force applied to the bar is a body force bx distributed
along the bar axial axis, as shown in Fig. 3.2.
Assume that stress is uniform over its cross-section, but it will not be
uniform along its axial direction x because of the distributed body force.
The stress will be a function of the coordinate x.
To analyze the stress at a location x in the bar, we assume that the
function of stress is sufficiently smooth and hence differentiable at least for
once. We can then cut off a free-body with an infinitely small dx, and look
at the equilibrium condition for the free-body. The diagram is shown at the
bottom of Fig. 3.2. We then put all the forces involved on the free-body
diagram.

Sign convention: As a general rule, the direction of a force put on a

free-body diagram should follow the standard convention. For example,
the internal force Nx + dNx is on the positive surface, it should point to
the positive x-direction. The internal force Nx is on the negative surface, it
should point to the negative x-direction. The externally applied force bx Ax dx
should always in the positive x-direction. This general rule applies to all the
free-body diagrams we use in this book, including the ones used in the future
chapters.

Figure 3.2. A bar fixed at its left-end and subjected to a body force bx distributed along
the bar axial axis. A free-body diagram with length dx is shown with all the forces on it.
The normal forces Nx and Nx + dNx are called internal force because it can only be seen
after the free-body is cut out. Force (bx Ax dx) is the externally applied force. It is there
regardless of whether the free-body is cut out or not.
 Forces and Stresses 77

Using the equilibrium condition that all forces in the x-direction must be
zero, we have

Fx = 0 : −Nx + (Nx + dNx ) + bx Ax dx = 0 (3.3)

In the summation above, forces in the positive x-direction are positive.

A minus sign is given to any force pointing to the negative direction of x.
Equation (3.3) can be rewritten as
dNx
+ bx Ax = 0 (3.4)
dx
The stress σxx on the cross-section should be
Nx
σxx = (3.5)
Ax
Substituting Eq. (3.5) into Eq. (3.4) gives

dσxx
+ bx = 0 (3.6)
dx
This is the equilibrium equation in terms of stress for 1D bars.
Equation (3.6) can be re-written as
dσxx
= −bx (3.7)
dx
This means that dσdxxx is the negated externally applied body force. Thus,
for 1D cases, the derivative of the stress is called divergence of the stress.
Equilibrium equation (3.6) states that the stress divergence is compen-
sated by the externally body force. If there is no body force, the stress will
not diverge, its derivative will be zero, as seen in Eq. (3.2) for the previous
example where no body force is involved.

3.3.2 Stress solution to 1D equilibrium equation

Equation (3.7) is a differential equation (DE) in stress. If the body force is
an integrable function of x, we can find the stress solution analytically. As
a simple example, if bx is constant, we have the general stress solution:
σxx = −bx x + c (3.8)
where c is an integral constant[2]. It can be determined using boundary
condition (BC). For our problem defined in Fig. 3.2, the stress BC is
At x = L : σxx = 0 (3.9)
78 Mechanics of Materials: Formulations and Solutions with Python

Using BC Eq. (3.9) in Eq. (3.8), we found

c = bx L (3.10)
The final solution becomes
σxx = −bx x + bx L = bx (L − x) (3.11)
It is a linear function of x. It is zero at x = L and has a maximum of
σxx = bx L at x = 0.
Readers may set body force bx to another integrable function, such as a
polynomial or trigonometric function and then find the stress solution along
the bar. Readers may consider cases with concentrated force fx at x = L, in
addition to the body force.

3.4 General measure of force vectors in 3D space

Forces are vectors. Thus, in higher dimensions we need a proper description
for it.

3.4.1 Graphic representation of force vectors

Consider an arbitrary force acting at the center of a rigid block denoted as
f in R3 , as shown in Fig. 3.3. It can be decomposed into three components,
as discussed Section 2.8.2.

3.4.2 Components of a force, orthogonal projection

We first decompose the force vector to two vectors: a vertical component f3 i3
and fs following the orthogonal projection rule. The fs is then further
decomposed into two vectors: f1 i1 and f2 i2 following also the orthogonal
projection rule. This process results in three components of f : f1 , f2 , and
f3 , respectively, on x1 -, x2 -, and x3 -axis. The force vector can be written
uniquely as
⎡ ⎤
f1
f = ⎣f2 ⎦ (3.12)
f3

3.4.3 Equivalence of force vector and its components

It is clear that this representation of a vector is unique under a fixed
coordinate, implying that any change, in length or in direction, in f will
result in unique changes in values of f1 , f2 and f3 , and vice versa, as long as
the orthogonal projection rule is followed, and an orthogonal coordinate
system is used. This mutual unique representation ensures that force vector
f are equivalent to its components f1 , f2 , and f3 .
 Forces and Stresses 79

Figure 3.3. A force f acting at the center of a rigid block. It can be decomposed into
three components f1 , f2 and f3 on three axes of an orthogonal coordinate system, respec-
tively, following the orthogonal projection rule.

3.4.4 Force vector in transformed coordinates

Note that when the coordinate system transforms (rotates), the values of
f1 , f2 , and f3 change according to the orthogonal projection rule, but the
basic property, amplitude and direction, of vector f do not change. Therefore,
the force vector is a tensor. It is of order one because its components have
only one index.
Consider a force vector f in coordinate x. Using the transformation tensor
T defined in the previous chapter, the values of force vector f  in the rotated
coordinate x become

fi = aij fj (3.13)

where fi is obtained following the orthogonal projection rule. On the right-
hand side of Eq. (3.13), we see that the 2nd index j of T is contracted
with the index of f . This leaves one index i, implying that the outcome is a
new vector.
In matrix notation, we have

f  = Tf (3.14)
80 Mechanics of Materials: Formulations and Solutions with Python

which achieves the correct contraction through the standard matrix multi-
plication rule.
As proven in the previous chapter, T is unitary. Thus, the inverse trans-
formation can be simply done using the transpose of T:

f = T f  (3.15)

Or in the indicial notation:

fi = aji fj (3.16)

All we need is to swap the order of the two indexes of aij , which is
equivalent to using T in matrix multiplication.

3.5 General measure of stresses in 3D solids

3.5.1 General definition of stress

Consider a 3D solid with arbitrary geometry. It is constrained so that it
cannot have any rigid-body displacements, and is loaded by point forces
and/or distributed forces like pressure, resulting in stresses distributed over
the solid. A schematic drawing is given in Fig. 3.4.
To view the stresses inside the solids, we use an imaginary plane-P to cut
off a part of the solid and expose a surface with unit normal direction of n.
Since any surface can be identified by its normal, let us call it surface-n for
convenience in our analysis. The effect of the cut-off part is replaced by a
force vector f that can be in arbitrary direction changing from location to
location. This force results in stresses distributed over the surface-n, implying
that the value and direction of the stresses vary also from location to location.
Now, look at a small area ΔAn on the surface-n. The force on ΔAn
becomes Δf , and a stress vector σn on it caused by Δf can be obtained
using

Δf
σn = lim (3.17)
ΔAn →0 ΔAn

The direction of stress vector σn is the same as f at the limit. Here,

n stands for the surface of the stress vector is on (not the direction of the
stress!). If n is taken as the direction of the x-axis of the Cartesian coordinate
system, we shall have σx defined, which is the stress vector is on surface-x.
The same argument applies to σj on surface-y, and σk on surface-z.
 Forces and Stresses 81

Figure 3.4. A (3D) solid with arbitrary geometry. It is constrained and loaded by forces
and/or distributed forces like the pressure. An imaginary plane-P cuts off a part of the
solid and exposes the surface-n.

3.5.2 Shear stresses on a surface

The shear force vector Δfs results in a stress vector expressed as
Δfs
σs = lim (3.18)
ΔAn →0 ΔAn

The direction of stress vector σs is the same as fs at the limit.

Note that σn and σs are at the same point in the same surface, following
the orthogonal projection rule.
Next, to quantify the shear stress vector σs , we introduce a 2D orthogonal
coordinate system (r, q) on the surface, as shown in Fig. 3.5.
The force component Δfr results in the following shear stress component
on surface-n:
Δfr
σnr = lim (3.19)
ΔAn →0 ΔAn

where σnr is the stress component on surface-n and in the r-direction.

82 Mechanics of Materials: Formulations and Solutions with Python

Figure 3.5. The shear force on surface-n (plane P) and its components on the r − q
coordinates that are orthogonal. We require that r crosses q gives n, leading to a local
orthogonal coordinates obeying the right-hand rule.

The force component Δfq results in:

Δfq
σnq = lim (3.20)
ΔAn →0 ΔAn
where σnq is the stress component on surface-n and in the q-direction.
The choice of coordinates r and q can be arbitrary, as long as they both
are orthogonal, and follow the right-hand-rule with n: n = ir × iq , where ir
and iq are unit vectors, respectively, on the r- and q-axes.
In summary, we obtain three stress components on surface-n at the center
of ΔAn : normal stress σnn , shear stresses σnr , and σnq . The directions of
these three stress components are orthogonal. These three stress components
quantify completely and uniquely the stress vector σn on surface-n at the
center of ΔAn for a given local Cartesian coordinate system.
Now, suppose n is taken as the direction of the z-axis, r is taken as the
x-axis, and q as the y-axis, we shall have normal stress σzz , shear stress
σzx and shear stress σzy . These stresses all on the same surface-z pointing,
respectively, to x, y, and z directions.
By the same argument, when n is taken as x-axis, we shall have σxx , σxz
and σxy . When n is taken as y-axis, it gives σyy , σyz , and σyx .
Let us state some important remarks:

• All stress measure depends on the area of the surface, in addition to its
direction. Thus, there must be two indexes for a stress component.
This is the root reason for stress to be a second-order tensor.
• At any point on a surface in 3D solid, there are in general three stress
components: one normal and two shears. The directions of these
three stress components are orthogonal, as long as the coordinates
used are orthogonal.
 Forces and Stresses 83

3.5.3 Stress representation in 3D solids

Consider a point in a stressed solid. To view all the stress components at

the point, we must expose the surfaces surrounding the point. From the
discussion above, we know that each surface can have a total of three stress
components: one normal and two shears. To present these stress components
on surfaces that are orthogonal with each other, we cut a small cube from
the solid surrounding that point, as shown in Fig. 3.6. These surfaces on the
cube are along with the Cartesian coordinates and hence orthogonal.
There are six surfaces on the cube, forming three pairs of surfaces with
opposite direction. All these pairs of surfaces are infinitely close to represent
the point. A Cartesian coordinate is defined with three axes: x, y, and z, all
being perpendicular to these pair of surfaces. On each of these six surfaces,
we put three stress components on it. All these stress components are now
shown in Fig. 3.6, all following our standard sign convention.
On any pair of opposite surfaces, the stress components should be the
same because these surfaces are infinitely close. But they point to opposite
directions because of the opposite surface direction. Therefore, there are a
total of nine stress components on the cube representing the stress states at
the point. The notation convention for a stress component σij is as follows:

Figure 3.6. Stress components on surfaces of a cube cut along a Cartesian coordinate
system. There are three pairs of orthogonal surfaces: surface-x, -y, and -z. Each pair has
two opposite normals. The directions of these three stress components on a surface are
orthogonal.
84 Mechanics of Materials: Formulations and Solutions with Python

σ (3.21)
i j

surface that stress is on direction of the stress

in which i, j = x, y, z (or 1, 2, 3). Thus, these indexes covers all the nine
stress components uniquely.

3.5.4 Shear stress equivalence, symmetric tensor

Let us examine the relationship between a pair of shear stresses: σyz and σzy ,
by considering the mechanics law of equilibrium conditions, using Fig. 3.7.
We look at the moment equilibrium about the axis of A-A in the x-
direction. The stress components that contribute to such a moment are in
red: a pair of σyz on two surfaces-y, and pair of σzy on two surfaces-z. Due to
the orthogonality, all other stresses contribute nothing to the moment about
A-A . The equilibrium law requires the total moment caused by the forces
resulting from these stresses being zero, we have

MA−A = 0 : (σyz dxdz) dy − (σzy dxdy)  dz = 0 (3.22)
    
force arm force arm

which leads to the shear stress equivalence for σyz and σzy :
σyz = σzy (3.23)
Note that the foregoing equation holds regardless of the actual length of
dx or dy, or dz, as long as they are small and the coordinates are orthogonal.

Figure 3.7. Moment equilibrium about the axis of A-A . All the stress components on
the cube that contribute to the moment are in red.
 Forces and Stresses 85

Similarly, examining the moment equilibrium about other two axes in

other two directions, we obtain
σxz = σzx
(3.24)
σxy = σyx
Readers may try to get one of these equations in Eq. (3.24). Equations
(3.23) and (3.24) are known as the shear stress equivalence. It can be
written in indicial notation as
σij = σji ∀i, j = 1, 2, 3 (3.25)
Due to these three equivalence conditions, the number of independent
stress components become six.
Equations (3.25) is known as the symmetry of stress components, and
hence the stress tensor is thus said to be a symmetric tensor: their indexes
are interchangeable. It is all rooted at the shear stress equivalence, as
the consequence of the mechanics equilibrium condition. As stated in
Noether’s theorem “every differentiable symmetry of the action of a physical
system with conservative forces has a corresponding conservation law”. In
mechanics of materials, the conservative law is equilibrium.

3.5.5 Dimensionality and stresses

Stress in mechanics of materials in general is a symmetric tensor of second
order, as discussed above. It can be written in indicial notation simply as
σij . For 1D problems, i and j have a range of 1, for 2D problems, i and j a
range of 2, and for 3D problems, a range of 3.
It has also the following three form of matrix representations:
⎡ ⎤ ⎡ ⎤
σ11 σ12 σ13 σxx σxy σxz
⎢ ⎥ ⎢ ⎥
σ = [σij ] = ⎣σ21 σ22 σ23 ⎦ = ⎣σyx σyy σyz ⎦ (3.26)
σ31 σ32 σ33 σzx σzy σzz
Since the stress tensor is symmetric, in space R3 , the matrix is a 3 × 3
symmetric matrix [1,3,4]. It is clear again that there are only 6 independent
components. This book uses all these three notation interchangeably.
In space R2 , it is a 2 × 2 symmetric matrix (considering symmetry), it
has only 3 components. Eq. (3.26) becomes
 
σ11 σ12 σxx σxy
σ = [σij ] = = (3.27)
σ21 σ22 σyx σyy
In space R1 , it is a scalar, and hence there is only 1 component. One may
write it in any of these forms: σ11 , σxx , or simply σ. This is the case we
examined at the beginning of this chapter.
86 Mechanics of Materials: Formulations and Solutions with Python

The analysis given above shall establish the basic concept about the
stresses and stress tensor. It is the right time to reinforce the concept, by
writing the following code to create in computer the stress tensor. First,
define stress tensor for later symbolic computations.

1 # define stress components as symbolic variable

2 s11, s12, s13 = "\u03C3_11", "\u03C3_12", "\u03C3_13"
3 s21, s22, s23 = "\u03C3_21", "\u03C3_22", "\u03C3_23"
4 s31, s32, s33 = "\u03C3_31", "\u03C3_32", "\u03C3_33"
5
6 S = MatrixSymbol('S', 3, 3)
7
8 S = Matrix([[s11, s12, s13],
9 [s21, s22, s23],
10 [s31, s32, s33]])
11 S # a stress tensor

⎡ ⎤
σ11 σ12 σ13
⎣σ21 σ22 σ23 ⎦
σ31 σ32 σ33
Stresses in 2D can be obtained by slicing in Python.

1 S2D = S[:2,:2]
2 S2D


σ11 σ12
σ21 σ22
Further, the stress in 1D can be obtained by indexing in Python.

1 S[0,0]

σ11

3.6 Tractions on an arbitrary surface

For a given stress tensor σ at a point in solid, we often need to know the
traction tn on a pre-specified surface with unit normal direction n. The
traction tn , as shown in Fig. 3.8, on the surface of the solid is the externally
applied force (per unit area) there. It carries the same unit of the stress.
When the point of interest is inside the solid, it is the stress vector denoted
as σn when the surface at that point is exposed. Positive traction component
is in the positive direction of the coordinates, as shown in Fig. 3.8.
We need to derive the formulas for this task.
 Forces and Stresses 87

Figure 3.8. As schematic draw of tractions on the surface at a point with normal n on
the boundary of the solid, or on a surface at a point inside a 2D solid.

3.6.1 Formulation
Consider in general a 3D solid, and suppose we know the stress tensor σ (in
nine components) at a point in solid. We would like to know the traction on a
pre-specified surface with unit normal direction of n = (n1 , n2 , n3 ), in which
n1 , n2 , n3 are the direction cosines of the unit normal vector. We shall have
n1 = cos(x, n)
n2 = cos(y, n) (3.28)
n3 = cos(z, n)
The setting is shown in Fig. 3.9. Note in the literature, some authors may
use n = (l, m, n) to denote these direction cosines.

3.6.2 Stress vector projection

First, the stress vector σn on surface-n can be projected orthogonally on the
x, y and z-axis, or x1 , x2 and x3 -axis in indicial notation. It can be expressed
as follows:
σn = σnx i1 + σny i2 + σnz i3 (3.29)
where i1 , i2 , and i3 are, respectively, the unit vectors on the x1 , x2 and
x3 -axes. These unit vectors are also denoted in literature as i, j, and k.
88 Mechanics of Materials: Formulations and Solutions with Python

Figure 3.9. Stresses on an arbitrary surface-n for a point in solid with stress tensor
given.

Equation (3.29) can be written as

σn = σni ii (3.30)
in which σn1 = σnx , σn2 = σny , and σn3 = σnz . Equation (3.30) or (3.29) is
a result from stress vector projection.

3.6.3 Area projection

The forces resulting from σni , i = 1, 2, 3 should be in equilibrium with the
forces resulting from the given nine stress components. Consider first the
components in the x-direction, we have
σnx An = Ax σxx + Ay σyx + Az σzx (3.31)
where An is the area of surface-n, Ax is the area of surface-x, Ay the area
of surface-y, and Az the area of surface-z, as shown in Fig. 3.9. These areas
have the following relationship:
Ax = n1 An
Ay = n2 An (3.32)
Az = n3 An
This is because Ax is the orthogonal projection of An along the x
direction, Ay is that along y direction, and Az is that along z. Equation
(3.31) becomes
σnx = n1 σxx + n2 σyx + n3 σzx on surface-n in x-direction (3.33)
 Forces and Stresses 89

This is the stress on surface-n in the x-direction. Similarly, we have

σny = n1 σxy + n2 σyy + n3 σzy on surface-n in y-direction
(3.34)
σnz = n1 σxz + n2 σyz + n3 σzz on surface-n in z-direction
It is clear that Eqs. (3.33) and (3.34) is the consequence of the orthogonal
projections of the area, and of the equilibrium condition. Three expressions
in Eqs. (3.33) and (3.34) can be written simply in a single equation in indicial
notation:
σnj = ni σij (3.35)
Note that the contraction occurs on the first axis of the stress tensor,
leaving a single index j, producing a vector on the given surface-n. In matrix
notation, we have the following formula with dot product to achieve the
needed contraction on the first axis of the stress tensor.
σn = n · σ (3.36)
Or in matrix form (following matrix-vector multiplication rule):
⎡ ⎤ ⎡ ⎤⎡ ⎤
σnx σxx σyx σzx n1
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣σny ⎦ = ⎣σxy σyy σzy ⎦ ⎣n2 ⎦ (3.37)
σnz σxz σyz σzz n3
    
σn σ n

This means the stress vector on surface-n is the orthogonal projection

of the transposed stress matrix on the unit normal vector n. Note that the
stress vector are written in components in the x-, y-, and z-axes.

3.6.4 Traction on a boundary surface

When the surface-n is on a boundary surface of a solid, Eq. (3.35), σni shall
give the traction applied on the boundary surface and is often denoted as
tni . It is written as
tnj = ni σij (3.38)
or explicitly
tnx = n1 σxx + n2 σyx + n3 σzx
tny = n1 σxy + n2 σyy + n3 σzy (3.39)
tnz = n1 σxz + n2 σyz + n3 σzz
This traction-stress relation bridges the internal stresses σij with the
externally applied tractions on the boundary of the solid tnj . It is thus
frequently used in expressing the stress boundary conditions for solid
mechanics problems.
90 Mechanics of Materials: Formulations and Solutions with Python

Equation (3.39) satisfies the equilibrium conditions, implying that the

tractions on the boundary at a point in solid is uniquely determined by
stress states at the point. If, for example, surface-n is the surface-x (the
surface normal is the x-axis), we have n = (1, 0, 0), and
tnx = σxx tny = σxy tnz = σxz (3.40)
If surface-n is the negative surface-x, we have n = (−1, 0, 0), and
tnx = −σxx tny = −σxy tnz = −σxz (3.41)
Equations (3.40) and (3.41) is a typical relationship between the external
stress tni and internal stress σij . Readers may take a moment examine these
two equations for better understanding traction. Readers may try to figure
out similar relations, if the boundary is on surface-y or on surface-z, both
positive and negative.

3.6.5 Stress vector projection, again

[Link] Normal stress on a surface
Once the stress vector on a surface-n is obtained, one can then compute
the normal stress on surface-n, by projecting the stress vector onto the unit
normal n of the surface.
σnn = ni σij nj (3.42)
This time the contraction occurs on the second axis of the stress tensor.
It can be achieved with the dot product of the traction vector on the left
with the unit normal n of the plane on the right. This is the orthogonal
projection of σn on direction n. At the end, both indexes of the stress
tensor are contracted, producing just a scalar, which is the normal stress on
the normal direction of the surface-n. Its explicit expression can be found
using the following code.
1 11, 22, 33, 23, 13, 12 = symbols(" 11, 22, 33, 23, 13, 12")
2 n1, n2, n3, nn = symbols("n1, n2, n3, _nn")
3
4 σ = Matrix([[ 11, 12, 13], # ε: stress t nsor
5 [ 12, 22, 23],
6 [ 13, 23, 33]])
7
8 n = Matrix([n1, n2, n3])
9 nn = (n.T@σ@n)[0].simplify()
10 [Link]( nn, "The mormal stress on surface-n:")

The mormal stress on surface-n:

n1 (n1 σ11 + n2 σ12 + n3 σ13 ) + n2 (n1 σ12 + n2 σ22 + n3 σ23 )

+ n3 (n1 σ13 + n2 σ23 + n3 σ33 )
 Forces and Stresses 91

Equation (3.42) can also be written in vector operation as

σnn = σn · n (3.43)
Equations (3.42) and (3.43) are identical.

[Link] Shear stress on a surface

To find the shear stress in a direction on surface-n, all we need to know is the
direction cosine of the direction vector. Assume it is given as q = (q1 , q2 , q3 ).
Since q is on the surface-n, q and n is orthonormal (orthogonal and all with
unit length), q · n = 0. The shear stress is thus the orthogonal projection
the stress vector σn onto q:
σnq = ni σij qj (3.44)
Note that there are infinite number of q on the surface satisfying the
orthonormal condition. Among which, there must be a unit vector that is
orthonormal to q, satisfying also the right-hand rule: r = q × n. Let us
denote it as r = (r1 , r2 , r3 ). The shear stress on the direction of r becomes
σnr = ni σij rj (3.45)
In this setting, r, q, and n form a Cartesian coordinate system obeying
the right-hand-rule.

[Link] The magnitude of the shear stress on a surface

The magnitude of the shear stress vector on a surface-n can be computed
using the Pythagorean theorem (also a type of orthogonal projection),

|σns | = |σn |2 − σnn
2

2 + σ2 + σ2 − σ2
= σn1 n2 n3 nn (3.46)

2 + σ2
= σnq nr

Let us write the following Numpy codes to compute stresses on a given

surface.

3.6.6 Example: Stresses on a given surface

This example computes stresses on a given surface for a given stress tensor
with nine stress components. For the convenience in future possible uses, we
write a simple Python function for this kind of task.
92 Mechanics of Materials: Formulations and Solutions with Python

1 def stressesOnN(σ, n, q=False):

2 '''Compute the traction (a stress vector), normal stress, and \
3 magnitude of the shear stress, and the shear stress on a surface
4 with normal vector n.
5 Input σ: stress tensor in matrix form with 9 components\
6 n: normal vector of the surface
7 q: a unit vector on the surface; an vector r orthogonal to q
8 will be generated based on the right-hand rule, and
9 the shear stresses in both q&r directions will be computed.
10 If not provided, the shear stresses are not computed.
11 Return: σ_n (traction), σ_nn (normal stress), σ_ns (shear stress)
12 σ_nq (shear stress in given q-direction)
13 σ_nr(shear stress in r-direction)
14 '''
15 n = n/[Link](n) # normalize vector n -> unit normal vector
16
17 # Compute the traction vector on the surface with normal n
18 σ_n = [Link](n,σ)
19
20 # Compute the normal stress on surface in the normal n direction
21 σ_nn = [Link](σ_n,n)
22
23 # compute the magnitude of the shear stress on the surface
24 σ_ns = [Link](σ_n.dot(σ_n)-σ_nn**2)
25
26 # Compute the shear stresses on the surface, if q is given
27 if isinstance(q, (list, tuple, [Link])):
28 q = q/[Link](q) # normalize vector q
29 ndq = [Link](n,q)
30
31 if ndq > 1e-12:
32 print(f'n and q are not orthogonal: {ndq}; σnq & σnr=None.')
33 σ_nq = None; σ_nr = None
34 else:
35 r = [Link](n, q) # generate r via right-hand rule
36 print(f'Unit vector r on surface-n = {r}\n')
37
38 # Compute the shear stress on the surface direction q
39 σ_nq = [Link](σ_n,q)
40
41 # Compute the shear stress on the surface direction q
42 σ_nr = [Link](σ_n,r)
43 elif q == False:
44 print(f'Shear stress will not be computed.')
45 σ_nq = None; σ_nr = None
46 else:
47 print(f'q must be array-like')
48 σ_nq = None; σ_nr = None
49
50 return σ_n, σ_nn, σ_ns, σ_nq, σ_nr
 Forces and Stresses 93

1 # Uniaxial loading case studied at the beginning of the chapter.

2
3 # Given stress tensor (matrix) in MPa
4 St = [Link]([[1.,0., 0.],[0., 0., 0.],[0.,0.,0.]]) # only xx.
5 print(f'Stress Tensor (MPa) :\n{St}')
6
7 n = [Link]([-1.,0.,0.]) # the surface normal is positive x-direction
8 #n = [Link]([1/[Link](3),1/[Link](3),1/[Link](3)]) # Octahedral
9 print(f'A surface with normal direction n = {n}\n')
10
11 q = [Link]([0. ,1.,0.]) # must be orthogonal to n
12 q = q/[Link](q) # normalize vector q
13 print(f'Normal vector q on surface-n = {q}\n')
14
15 σ_n, σ_nn, σ_ns, σ_nq, σ_nr = stressOnN(St, n, q) # or [Link]()
16 print(f'Traction (stress vector) on the surface = {σ_n} MPa')
17 print(f'normal stress on the surface = {σ_nn:2.4f} MPa')
18 print(f'Magnitude of the shear stress on the surface = {σ_ns:2.4f} MPa')
19 print(f'shear stress on the surface in q-direction = {σ_nq} MPa')
20 print(f'shear stress on the surface in r-direction = {σ_nr} MPa')

Stress Tensor (MPa) :

[[1. 0. 0.]
[0. 0. 0.]
[0. 0. 0.]]
A surface with normal direction n = [-1. 0. 0.]

Normal vector q on surface-n = [0. 1. 0.]

Unit vector r on surface-n = [ 0. 0. -1.]

Traction (stress vector) on the surface = [-1. 0. 0.] MPa

normal stress on the surface = 1.0000 MPa
Magnitude of the shear stress on the surface = 0.0000 MPa
shear stress on the surface in q-direction = 0.0 MPa
shear stress on the surface in r-direction = 0.0 MPa

In this example, we purposely set the normal vector of the surface in neg-
ative x-direction (with n = [Link]([−1., 0., 0.])), and found that the com-
ponent of the traction in the x-axis on surface-n is negative. This is because
the sign of the traction follows the sign of the coordinates. The normal stress
on this negative surface is however positive because it follows the direction of
surface-n. This example is ideal in revealing the difference between the trac-
tion on a surface and stresses on the same surface. This subtlety confuses
many, but these kinds of behavior are correct, and are resulting from the
rules of operations for all tensors in solid mechanics: orthogonal coordinates,
orthogonal projections, equilibrium laws, and the standard sign convention.
94 Mechanics of Materials: Formulations and Solutions with Python

Readers may consider the surface is in the positive x-direction (with

n = [Link]([1., 0., 0.])) (or any other direction), and run the code. Then,
think carefully about the results obtained for this simple problem, until
it makes sense. Readers may also create artificial bi-axial stress state say,
σxx = 1 and σyy = 2 with all other components zero, and then perform the
tests again, using the code given. Such an exercise may significantly improve
the understanding on the behavior of stresses in solids.
The following is a more general stress state case.

1 # stress state used in Example 2.1 in

2 # Ref: Advanced Mechanics of Materials, by Boresi and Schmidt
3 np.set_printoptions(precision=4,suppress=True)
4
5 # Given stress matrix in MPa
6 Sv = [Link]([[-10.,15., 0.],[15.,30., 0.],[0.,0.,0.]]) #Example 2.1
7 print(f'Stress Tensor (MPa) :\n{Sv}')
8
9 # normal vector of the surface:
10 n = [Link]([1.,1.,0.]) #[1.,1.,2.] # one may try this also.
11 n = n/[Link](n)
12 print(f'A surface with normal direction n = {n}\n')
13
14 q = [Link]([-1., 1., 0]) # must be orthogonal to n
15 q = q/[Link](q) # normalize vector q
16 print(f'Normal vector q on surface n = {q}\n')
17
18 σ_n, σ_nn, σ_ns, σ_nq, σ_nr = stressOnN(Sv, n, q)
19 print(f'Traction (stress vector) on the surface = {σ_n} MPa')
20 print(f'Normal stress on the surface = {σ_nn:2.4f} MPa')
21 print(f'Magnitude of the shear stress on the surface = {σ_ns:2.4f} MPa')
22 print(f'Shear stress on the surface in q-direction = {σ_nq} MPa')
23 print(f'Shear stress on the surface in r-direction = {σ_nr} MPa')

Stress Tensor (MPa) :

[[-10. 15. 0.]
[ 15. 30. 0.]
[ 0. 0. 0.]]
A surface with normal direction n = [0.7071 0.7071 0. ]

Normal vector q on surface n = [-0.7071 0.7071 0. ]

Unit vector r on surface-n = [ 0. -0. 1.]

Traction (stress vector) on the surface = [ 3.5355 31.8198 0. ] MPa

Normal stress on the surface = 25.0000 MPa
Magnitude of the shear stress on the surface = 20.0000 MPa
Shear stress on the surface in q-direction = 20.0000 MPa
Shear stress on the surface in r-direction = 0.0 MPa
 Forces and Stresses 95

Note that the code stressOnN() works for problems in any dimension.
Following is an example for 2D stress state, which is the same as the previous
3D example.

1 # Given stress matrix in MPa

2 Sv2d = Sv[:2,:2]
3 #Sv2d = [Link]([[88., 38.], [38.,-35.]]) # one may try this also.
4 print(f'Stress Tensor (MPa) :\n{Sv2d}')
5
6 n = [Link]([1., 1.]) # normal vector of the surface
7 n = n/[Link](n)
8 print(f'A surface with normal direction n = {n}\n')
9
10 q = [Link]([-1.,1.]) # must be orthogonal to n
11 q = q/[Link](q)
12 print(f'Unit vector q on surface n = {q}\n')
13
14 σ_n, σ_nn, σ_ns, σ_nq, σ_nr = stressOnN(Sv2d, n, q)
15 print(f'Traction (stress vector) on the surface = {σ_n} MPa')
16 print(f'Normal stress on the surface = {σ_nn:2.4f} MPa')
17 print(f'Magnitude of the shear stress on the surface = {σ_ns:2.4f} MPa')
18 print(f'Shear stress on the surface in q-direction = {σ_nq} MPa')
19 print(f'Shear stress in r-direction is ignored for 2D cases.')

Stress Tensor (MPa) :

[[-10. 15.]
[ 15. 30.]]
A surface with normal direction n = [0.7071 0.7071]

Unit vector q on surface n = [-0.7071 0.7071]

Unit vector r on surface-n = 1.0000000000000002

Traction (stress vector) on the surface = [ 3.5355 31.8198] MPa

Normal stress on the surface = 25.0000 MPa
Magnitude of the shear stress on the surface = 20.0000 MPa
Shear stress on the surface in q-direction = 20.0000 MPa
Shear stress in r-direction is ignored for 2D cases.

Note that in 2D cases, the unit vector r is only the third component of
the cross product q × n. The shear in r-direction should be ignored.
With stressOnN(), readers can try more cases, 2D and 3D ones.

3.7 Coordinate transformation of second order

tensor (stress)

Consider a coordinate transformation from coordinate system of x1 , x2 , and

x3 to a new coordinate x1 , x2 , and x3 . We can now readily derive the rule for
96 Mechanics of Materials: Formulations and Solutions with Python

coordinate transformation of stress tensor. It is nothing but a generalization

of Eq. (3.44).

3.7.1 Formulation by simple inspection

Let us now re-write Eq. (3.44) to include very carefully the precise meaning
of each of the indexes.
σnq = ni σij qj
σ n
 q = ni σ i
 j qj
   
arbitrary-surface arbitrary-direction a surface comp. contract contract with a direction
denoted by r denoted by s with surface direction comp.

⎡ r ⎤ ⎡ s ⎤ ⎡i⎤ ⎡j⎤
x1 x1 x1 x1
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣x2 ⎦ ⎣x2 ⎦ ari ⎣x2 ⎦ ⎣x2 ⎦ asj
x3 x3 x3 x3
(3.47)

⎡ ⎤
cos(x1 , x1 ) cos(x , x ) cos(x , x )
⎢ a   1 2  1 3 ⎥
⎢ 11 a12 a13 ⎥
⎢ ⎥
⎢cos(x2 , x1 ) cos(x2 , x2 ) cos(x2 , x3 )⎥
 
T=⎢
⎢ a     ⎥
⎥ (3.48)
⎢ 21 a22 a23 ⎥
⎢ ⎥
⎣cos(x3 , x1 )  
cos(x3 , x2 ) cos(x3 , x3 )⎦
     
a31 a32 a33

After the coordinate transformation from the base coordinates x1 , x2 , and

x3 to the new coordinates x1 , x2 , and x3 , the ni should be replaced by ari
because the surface could be one of the three axes x1 , x2 , and x3 . The qj
should be replaced by asj because the direction could also be one of the
three directions x1 , x2 , and x3 . The surface-index can use free index r, and
the direction-index can use free index s. All the contraction operations stay
unchanged, leaving with 2 indexes, resulting in a new description of the same
stress tensor in the new coordinate system. After this indexation matching,
the new components of the stress tensor in the new coordinate are computed
using

σrs = ari σij asj (3.49)

This is the formulation for stress tensor transformation rule using

indicial notation. This is because the transformation is in reference to the
original base coordinates x1 , x2 and x3 as defined in Eq. (3.48), which must
be contracted so as to get into the new coordinate system x1 , x2 and x3 .
 Forces and Stresses 97

Alternatively, we may use the matrix notation and matrix multiplication

rule that achieves the same contraction effects, and the formula should be:

σ = TσT (3.50)

The T on the left ensures the contraction is on the first index of σ (with
the 2nd index of T), and T on the right ensures the contraction is on the
2nd index of σ (also with the 2nd index of T), and thus the transpose is
needed.

3.7.2 Inverse transformation

Since T is unitary, the inverse transformation can be simply done using the
transpose of T:

σ = T σ  T (3.51)

Or in the indicial notation:


σrs = air σij ajs (3.52)

All we need is to swap the order of the two indexes of aij for reverse
transformation. This is because the transformation is in reference to x1 , x2
and x3 as defined in Eq. (3.48), which must be contracted so as to get back
to the coordinate system x1 , x2 and x3 .
It is seen that transformation of a second-order stress tensor requires two
times of operation by the transformation matrix T.

3.7.3 Python example: Stress tensor under coordinate

transformation
Consider a stress tensor in 3D solids represented originally in a coordinate
system x. When the coordinates is transformed (rotated) to a new coordinate
x , we compute the new components of the same stress tensor in the new
coordinate system.
Let us write Numpy codes to compute stresses after a coordinate trans-
formation, using both matrix and indicial formulations.
First, an arbitrary stress tensor in the coordinate system x can be written
using the following code. In coordinate transformation operations, it is more
convenient to use the indicial notation: x:1, y:2, and z:3. We shall also make
use of the symmetry properties of the stress tensor.
98 Mechanics of Materials: Formulations and Solutions with Python

[Link] Use of matrix formulation

1 T = [Link]()
2 T

⎡ ⎤
a11 a12 a13
⎢ ⎥
⎣a21 a22 a23 ⎦
a31 a32 a33

1 Sp=T@[email protected] # S: stress tensor; @: the shorthand of [Link]

2 Sp[:,0] # print out the first column of S-prime

⎡ ⎤
a11 (a11 σ11 + a12 σ21 + a13 σ31 ) + a12 (a11 σ12 + a12 σ22 + a13 σ32 )
⎢ + a13 (a11 σ13 + a12 σ23 + a13 σ33 ) ⎥
⎢ ⎥
⎢ a (a σ + a σ + a σ ) + a (a σ + a σ + a σ ) ⎥
⎢ 11 21 11 22 21 23 31 12 21 12 22 22 23 32 ⎥
⎢ ⎥
⎢ + a13 (a21 σ13 + a22 σ23 + a23 σ33 ) ⎥
⎢ ⎥
⎣ a11 (a31 σ11 + a32 σ21 + a33 σ31 ) + a12 (a31 σ12 + a32 σ22 + a33 σ32 ) ⎦
+ a13 (a31 σ13 + a32 σ23 + a33 σ33 )
This is the general formula for the first column stress components after
the coordinate transformation with T. Others can be easily printed out as
follows. Note again that T is applied from both sides because S is a second-
order tensor.

1 Sp[:,1] # print out the 2nd column

⎡ ⎤
a21 (a11 σ11 + a12 σ21 + a13 σ31 ) + a22 (a11 σ12 + a12 σ22 + a13 σ32 )
⎢ + a23 (a11 σ13 + a12 σ23 + a13 σ33 ) ⎥
⎢ ⎥
⎢ a (a σ + a σ + a σ ) + a (a σ + a σ + a σ ) ⎥
⎢ 21 21 11 22 21 23 31 22 21 12 22 22 23 32 ⎥
⎢ ⎥
⎢ + a23 (a21 σ13 + a22 σ23 + a23 σ33 ) ⎥
⎢ ⎥
⎣ a21 (a31 σ11 + a32 σ21 + a33 σ31 ) + a22 (a31 σ12 + a32 σ22 + a33 σ32 ) ⎦
+ a23 (a31 σ13 + a32 σ23 + a33 σ33 )

1 Sp[:,2] # print out the 3rd column

⎡ ⎤
a31 (a11 σ11 + a12 σ21 + a13 σ31 ) + a32 (a11 σ12 + a12 σ22 + a13 σ32 )
⎢ + a33 (a11 σ13 + a12 σ23 + a13 σ33 ) ⎥
⎢ ⎥
⎢ a (a σ + a σ + a σ ) + a (a σ + a σ + a σ ) ⎥
⎢ 31 21 11 22 21 23 31 32 21 12 22 22 23 32 ⎥
⎢ ⎥
⎢ + a33 (a21 σ13 + a22 σ23 + a23 σ33 ) ⎥
⎢ ⎥
⎣ a31 (a31 σ11 + a32 σ21 + a33 σ31 ) + a32 (a31 σ12 + a32 σ22 + a33 σ32 ) ⎦
+ a33 (a31 σ13 + a32 σ23 + a33 σ33 )
 Forces and Stresses 99

[Link] Use of tensor operation

Alternatively, we can use indicial notation, and [Link]() to have the

same transformation done.

1 def Tensor2_transfer(T,S):
2 '''Coordinate transformation for 2nd order tensors
3 '''
4 S_ = [Link](T , S, axes=([1],[0])) # specify contracting exes
5 S_ = [Link](S_, T, axes=([1],[1]))
6 return S_

1 TS = Tensor2_transfer(T, S)
2 TS = Matrix(TS)
3 print(f"Transformed stresses (first column only):")
4 TS[:,0]

Transformed stresses (first column only):

⎡ ⎤
a11 (a11 σ11 + a12 σ21 + a13 σ31 ) + a12 (a11 σ12 + a12 σ22 + a13 σ32 )
⎢ + a13 (a11 σ13 + a12 σ23 + a13 σ33 ) ⎥
⎢ ⎥
⎢ a (a σ + a σ + a σ ) + a (a σ + a σ + a σ ) ⎥
⎢ 11 21 11 22 21 23 31 12 21 12 22 22 23 32 ⎥
⎢ ⎥
⎢ + a13 (a21 σ13 + a22 σ23 + a23 σ33 ) ⎥
⎢ ⎥
⎣ a11 (a31 σ11 + a32 σ21 + a33 σ31 ) + a12 (a31 σ12 + a32 σ22 + a33 σ32 ) ⎦
+ a13 (a31 σ13 + a32 σ23 + a33 σ33 )

3.7.4 Transformation of stress tensor in 2D

For stress states at any point in 2D, the expression is more manageable, and
can be derived explicitly. The transformation rule is the same as Eq. (3.50).
Assuming the new coordinates are X-Y rotated from the original x-y coor-
dinates by θ, the Sympy derivation for the expressions for stresses in X-Y is
as follows:

1 # define stress components as symbolic variable for 2D stresses

2 sxx, syy, sxy, syx = "\u03C3_xx", "\u03C3_yy", "\u03C3_xy", "\u03C3_yx"
3
4 S2D = Matrix([[sxx, sxy],
5 [syx, syy]])
6 S2D # a 2D stress tensor


σxx σxy
σyx σyy
100 Mechanics of Materials: Formulations and Solutions with Python

1 = symbols(" ") # rotation angle from x-y

2 T2D = [Link]( )[0:2,0:2]
3 T2D


cos (θ) sin (θ)
− sin (θ) cos (θ)

1 S2Dp = T2D @ S2D @ T2D.T # S2D: stress tensor in 2D, obtained earlier
2 sXX = S2Dp[0,0].expand()
3 [Link](sXX, 'σ_XX in the rotated coordinate X-Y:')

σ_XX in the rotated coordinate X-Y:

σxx cos2 (θ) + σxy sin (θ) cos (θ) + σyx sin (θ) cos (θ) + σyy sin2 (θ)

1 sYY = S2Dp[1,1].expand()
2 [Link](sYY, 'σ_YY in the rotated coordinate X-Y:')

σ_YY in the rotated coordinate X-Y:

σxx sin2 (σ) − σxy sin (θ) cos (θ) − σyx sin (θ) cos (θ) + σyy cos2 (θ)

1 sXY = S2Dp[0,1].expand()
2 [Link](sXY, 'σ_XY in the rotated coordinate X-Y:')

σ_XY in the rotated coordinate X-Y:

−σxx sin (θ) cos (θ) + σxy cos2 (θ) − σyx sin2 (θ) + σyy sin (θ) cos (θ)
The formulas are summarized, with regrouping, as follows:
σXX = σxx cos2 θ + σyy sin2 θ + 2σxy sin θ cos θ
σY Y = σxx sin2 θ + σyy cos2 θ − 2σxy sin θ cos θ (3.53)
σXY = σxy cos2 θ − σxy sin2 θ + (σyy − σxx ) sin θ cos θ
Readers may find these formulas frequently in the literature. It will be
used to derive the Mohr circle later. Note that the 2D stresses can also
computed easily by treating 2D stresses as a special case of 3D ones.
Let us look at some specific examples.

[Link] Example: Uniaxial stress after a coordinate transformation

Consider a simple uniaxial loading case studied at the beginning of the
chapter.
 Forces and Stresses 101

1 # Given stress matrix in MPa

2 S_xx = Matrix([[1.,0., 0.],[0., 0., 0.],[0.,0.,0.]]) # It has only xx.
3 printM(S_xx, 'Given stress matrix', n_dgt=4)

Given stress matrix

⎡ ⎤
1.0 0 0
⎣ 0 0 0⎦
0 0 0
Let us rotate the coordinate by 45 degrees about the z-axis. The trans-
formation matrix becomes
1 # Create transformation matrix.
2 about, = 'z', 45. # rotation axis and angle
3 Tr, _ = [Link]( , about = about)
4 print(f'Transformation tensor { :3.2f}°, w.r.t. {about}:\n{Tr}')

Transformation tensor 45.00°, w.r.t. z:

[[ 0.7071 0.7071 0. ]
[-0.7071 0.7071 0. ]
[ 0. 0. 1. ]]

1 TTS = Matrix(Tensor2_transfer(Tr, S_xx))

2 printM(TTS, 'Stress after forward transfer', n_dgt=4)

Stress after forward transfer

⎡ ⎤
0.5 −0.5 0
⎣−0.5 0.5 0⎦
0 0 0
This means that uniaxial stress tensor with only one nonzero component
of σxx = 1 in (x, y, z) has a shear stress components of σXY = −0.5 in
45-degree rotated coordinate (X, Y, Z) with respect to z, shown in Fig. 3.10.

Figure 3.10. Stresses in (X, Y, Z) in a bar subjected to a uniaxial force in the x-direction.
Rotation angle: 45◦ . The arrows for the stress is the actual direction of the stresses (not
following the standard sign convention for free-body diagrams).
102 Mechanics of Materials: Formulations and Solutions with Python

The failure of many ductile materials is controlled by shear stress. Thus,

such materials fail at an angle close to 45 degrees when they are subjected to
uniaxial tension or compression. Readers may pay attention to the negative
sign of σXY , which explains why the shear stress needs to carry a negative
sign in the Mohr circle (see later).
We can now transform the coordinate back:

1 TTSb = Matrix(Tensor2_transfer(Matrix(Tr).T, TTS))

2 printM(TTSb, 'Stress after forward & inverse transfer', n_dgt=4)

Stress after forward & inverse transfer

⎡ ⎤
1.0 0 0
⎣ 0 0 0⎦
0 0 0

Let us generate a random transformation matrix, and use it do a coordi-

nate transformation.

1 # Create random transformation matrix.

2 about, = 'random', 0. # rotation angle is not used
3 Tr, _ = [Link]( , about = about)
4 print(f'Transformation tensor { :3.2f}°, w.r.t. {about}:\n{Tr}')

Transformation tensor 0.00°, w.r.t. random:

[[-0.2402 0.9696 -0.0463]
[-0.0036 0.0468 0.9989]
[ 0.9707 0.2401 -0.0078]]

1 TTS = Matrix(Tensor2_transfer(Tr, S_xx))

2 printM(TTS, 'Stress after forward transfer', n_dgt=4)

Stress after forward transfer

⎡ ⎤
0.05768 0.0008532 −0.2331
⎣0.0008532 1.262 · 10−5 −0.003449⎦
−0.2331 −0.003449 0.9423

We see that the simple uniaxial stress state becomes a full matrix, and
all the components become nonzero, in some rotated coordinates. Since the
transformation is invertible, one can also expect that a complicated stress
state can have a simpler representation in a rotated coordinates.
 Forces and Stresses 103

Let us rotate the coordinates back.

1 TTSb = Matrix(Tensor2_transfer(Matrix(Tr).T, TTS))

2 printM(TTSb, 'Stress after forward & inverse transfer', n_dgt=4)

Stress after forward & inverse transfer

⎡ ⎤
1.0 1.388 · 10−16 −8.327 · 10−17
⎣ 1.397 · 10−16 1.311 · 10−17 −5.517 · 10−19 ⎦
−8.383 · 10−17 −6.37 · 10−19 2.679 · 10−20

We recovered the original uniaxial stress state (to machine accuracy).

[Link] Example: A general stress state under

coordinate transformation

Let us look at a more general stress state.

1 # define the stress state:

2
3 S_value=Matrix([[8, 2, 3],
4 [2, 7, 1],
5 [3, 1, 5]]) # stress values are given arbitrarily
6 print(f'Stress state in values:')
7 [Link](list(zip(S,S_value)))

Stress state in values:

⎡ ⎤
8 2 3
⎣2 7 1⎦
3 1 5

Let us rotate the coordinates by 60 degrees about the z axis. The trans-
formation matrix becomes

1 # Create transformation matrix.

2 about, = 'z', 60 # rotation axis and angle
3 Tr, _ = [Link]( , about = about)
4 print(f'Transformation tensor { :3.2f}°, w.r.t. {about}:\n{Tr}')

Transformation tensor 60.00°, w.r.t. z:

[[ 0.5 0.866 0. ]
[-0.866 0.5 0. ]
[ 0. 0. 1. ]]
104 Mechanics of Materials: Formulations and Solutions with Python

1 # This time use the formula derived earlier for transformed stresses.
2 # The results can be obtained by simple substitutions.
3 TS = Matrix(Tensor2_transfer(T, S)) # re-generate the formula
4 T = [Link]() # re-generate the transformation matrix
5
6 # Substitute matrix T with values in Tr. We use zip to create a
7 # list of tuples, which tuple pairs T_ij with Tr_ij for substitution:
8 TTS = [Link](list(zip(T, Matrix(Tr))))
9 [Link](TTS[:,0], 'Stress (1st column) formulas in new coordinates')

Stress (1st column) formulas in new coordinates

⎡ ⎤
0.25σ11 + 0.433σ12 + 0.433σ21 + 0.75σ22
⎣−0.433σ11 − 0.75σ12 + 0.25σ21 + 0.433σ22 ⎦
0.5σ31 + 0.866σ32

1 TTSv=Matrix([Link](list(zip(S, Matrix(S_value))))) # Subs. stresses

2 [Link](TTSv, 'Stresses in the new coordinates')

Stresses in the new coordinates

⎡ ⎤
8.982 −1.433 2.366
⎣−1.433 6.018 −2.098⎦
2.366 −2.098 5.0

These are the stress components after transformation. We can now trans-
form the coordinate back:

1 TTSvb = Matrix(Tensor2_transfer(Tr.T, TTSv))

2 printM(TTSvb, 'Stress after forward & inverse transfer', n_dgt=4)

Stress after forward & inverse transfer

⎡ ⎤
8.0 2.0 3.0
⎣2.0 7.0 1.0⎦
3.0 1.0 5.0

The stresses are again recovered.

Let us now look a randomly generated matrix that may or may not be a
stress state, and subject it to coordinate transformation, using a randomly
generated transformation matrix.
 Forces and Stresses 105

1 [Link](8) # set a seed for reproducibility

2 M = [Link](3,3) # Generate a random matrix
3
4 about, theta = 'random', 0. # angle theta is not used
5 Tr, _ = [Link](theta, about = about) # random T.
6
7 print(f'Randomly generated matrix M (asymmetric): \n{M}')

Randomly generated matrix M (asymmetric):

[[ 0.0912 1.0913 -1.947 ]
[-1.3863 -2.2965 2.4098]
[ 1.7278 2.2046 0.7948]]

1 TM = Matrix(gr.Tensor2_transfer(Tr,M)) # use random transformation

2 printM(TM, 'After forward coordinate transformation', n_dgt=4)

After forward coordinate transformation

⎡ ⎤
−1.886 −3.557 1.229
⎣−0.2874 −1.248 0.507⎦
−1.742 0.9192 1.724
1 TMb = Matrix(Tensor2_transfer(Tr.T,TM))
2 printM(TMb, 'After forward & inverse transformation', n_dgt=4)

After forward & inverse transformation

⎡ ⎤
0.0912 1.091 −1.947
⎣−1.386 −2.296 2.41 ⎦
1.728 2.205 0.7948
We obtained the original stress state.
Remark: To perform an inverse transformation, all we need is to use the
transposed transformation matrix, or reverse the indexes of the trans-
formation tensor. This applies to arbitrary matrix. This is because of the
unitary property of the transformation matrix.

This example shows that any matrix can be a second-order tensor, pro-
vided the rule of transformation defined in Eq. (3.49) is attached to it. The
content of the matrix is not the point. The behavior of the content under a
coordinate transformation is.

3.8 Principal stresses

For a given stress state, one can find a set of principal stresses, by rotating
the coordinates to a particular angle, where all the shear stress components
vanish and only normal stresses exist. Such a normal stress is called the prin-
cipal stress. Assuming such a plane is found with a unit normal denoted as
106 Mechanics of Materials: Formulations and Solutions with Python

n = [n1 n2 n3 ] . Such a plane is called principal plane, and n is called

a principal axis. On a principal plane, the stress vector σn must be in the
same direction of n because there is only a normal stress on that plane. It
can be written as
σn = σn (3.54)
where σ is the amplitude of the normal stress. It is yet unknown. The com-
ponents of the normal stress vector σn should be
σnx = σn1 ; σny = σn2 ; σnz = σn3 (3.55)
Substituting these into Eqs. (3.33) and (3.34), we obtain:
σn1 = n1 σxx + n2 σyx + n3 σzx
σn2 = n1 σxy + n2 σyy + n3 σzy (3.56)
σn3 = n1 σxz + n2 σyz + n3 σzz
which can be re-arranged to
n1 (σxx − σ) + n2 σyx + n3 σzx = 0
n1 σxy + n2 (σyy − σ) + n3 σzy = 0 (3.57)
n1 σxz + n2 σyz + n3 (σzz − σ) = 0
In matrix form:
⎛ ⎞
⎜⎡σ ⎤ ⎡ ⎤⎟ ⎡ ⎤ ⎡ ⎤
⎜ xx σyx σzx 1 0 0 ⎟ n1 0
⎜⎣ ⎟
⎜ σxy σyy σzy ⎦ −σ ⎣0 1 0⎦⎟ ⎣n2 ⎦ = ⎣0⎦ (3.58)
⎜ ⎟
⎝ σxz σyz σzz 0 0 1 ⎠ n3 0
     
σ = σ I n 0

Here, we used the symmetry property of the stress tensor.

3.8.1 Eigenvalue problem

Equation (3.58) is a typical eigenvalue equation. It can be written in a concise
form of
[σ − σI]n = 0 (3.59)

where σ is an unknown stress called eigenvalue in mathematical term, I

is the identity matrix that has the same dimension as stress tensor σ, and
n is called eigenvector related to the vector of the direction cosines of n.
Equation (3.59) is in fact a linear algebraic system equation with n being
the unknown vector, but the system matrix contains a unknown σ. We need
to solve the eigenvalue equation to find σ and its corresponding vector n.
 Forces and Stresses 107

Since this linear algebraic system is homogeneous (the right-hand side is

zero). The solution exists only if
det(σ − σI) = 0 or |σ − σI| = 0 (3.60)
Equation (3.60) can then be used to find the eigenvalue σ.

3.8.2 Characteristic equation, polynomial

For stresses in 3D solids, the determinate in Eq. (3.60) results in a polyno-
mial equation of order of 3 in σ (implying the problem becomes a nonlinear
one). The polynomial equation is known as the characteristic equation,
for which there will always be 3 roots (complex in general, real if σ is sym-
metric) of σ. Therefore, σ is also called characteristic value. These three
characteristic values are the principal stresses denoted as σ1 , σ2 , and σ3 . This
essentially proves the existence of 3 principal stresses.
Substituting each of these eigenvalues back to Eq. (3.59), n can be solved
with the help of the normalization condition:

n21 + n22 + n23 = 1 or ni ni = 1 (3.61)

Note that this equation is nonlinear in ni .
The eigenvector corresponds to the principal axes. Since Eq. (3.59) is
homogeneous and σ is a square matrix, the solutions for eigenvectors always
exist. In fact, the eigenvector is not unique up to a scalar. This is simply
because an eigenvector that satisfies the homogeneous Eq. (3.59), any its
scalar multiples will also satisfy it.
Since the problem becomes nonlinear, we are expecting multiple solu-
tions for the eigenvectors (in ni ). We shall choose these eigenvectors relevant
to mechanics of materials problems following the right-hand rule. This can
be ensured by, for example, computing n3 using the cross product of n1
toward n2 :
n3 = n1 × n2 (3.62)
These eigenvectors correspond to the principal stresses, and give the direc-
tion cosines of the principal axes.
Note that the cosine function is multivalued. For example, cos(θ) =
cos(−θ). When a direction cosine value, say n1 , is used to find the angle
θ1 , the actual angle could be either −θ1 or θ1 . In addition because both n
and −n are all eigenvectors, n1 and −n1 can also be used to compute the
angle θ1 . Since cos(θ1 ) = −cos(180◦ − θ1 ), (180◦ − θ1 ) may also be the angle.
Therefore, one needs to find the correct angle from the possible ones. This
can be easily done by using the possible angles to perform coordinate trans-
formations to the stress tensor. The correct angle should produce zero shear
stress. We will do so multiple time in later examples.
108 Mechanics of Materials: Formulations and Solutions with Python

3.8.3 Python code for finding principal stresses

and their directions
Eigenvalue problems can be solved routinely. We write the following code
function for solving eigenvalue problems for stresses in 3D solid materials.

1 def principalS(S):
2 '''Compute the principal stresses/strains & their direction cosines.
3 inputs:
4 S: given stress/strain tensor, numpy array
5 return:
6 principal stresses/strain (eigenValues), their direction cosines
7 (eigenVectors) ranked by its values. Right-hand-rule is enforced
8 '''
9 eigenValues, eigenVectors = [Link](S)
10
11 #Sort in order
12 idx = [Link]()[::-1]
13 eigenValues = eigenValues[idx]
14 eigenVectors = eigenVectors[:,idx]
15 print('Principal stress/strain (Eigenvalues):\n',eigenValues,'\n')
16
17 # make the first element in the first vector positive (optional):
18 #eigenVectors[0,:] = eigenVectors[0,:]/[Link](eigenVectors[0,0])
19
20 # Determine the sign for given eigenVector-1 and eigenVector-3
21 eigenVectors[:,2] = [Link](eigenVectors[:,0], eigenVectors[:,1])
22
23 angle = [Link](eigenVectors[0,0])*180/[Link] # in degree
24 print(f'Principal stress/strain directions:\n{eigenVectors}\n')
25 print(f"Possible angles (n1,x)={angle}∘ or {180-angle}∘")
26
27 return eigenValues, eigenVectors

3.8.4 Example: Computing the principal stresses

and principal directions
This example is studied in Ref. [1]. We write the following code for the same
problem, using our code function principalS().

1 # Advanced Mechanics of Materials, by Boresi and Schmidt

2 # Example 2.1 Principal Stresses and Principal Directions
3
4 St = [Link]([[-10, 15, 0], [15, 30, 0], [0, 0, 0]]) # stress tensor
5
6 print(f'Stress state:\n{St}\n')
7
8 eigenValues, eigenVectors = principalS(St) # or [Link](St)
 Forces and Stresses 109

The solution is found with a single line of code. The computation takes
microseconds.
We got two angles because of the non-uniqueness of the eigenvectors and
the multivalued feature of cosine functions. To find out which one is correct,
we use these two angles, form the transformation matrices, and perform the
transformations. The correct one gives zero shear stresses.

1 # Form the transformation matrix using the first angle:

2 T = [Link](108.43494882292201, about = 'z')[0]
3
4 # Perform the transformation to the stress tensor
5 T @St@T .T

array([[ 17., -24., 0.],

[-24., 3., 0.],
[ 0., 0., 0.]])

This angle is not correct. Let us try the second one:

1 # Form the transformation matrix using the 2nd angle:

2 T = [Link](71.56505117707799, about = 'z')[0]
3
4 # Perform the transformation to the stress tensor
5 T @St@T .T

array([[ 35., -0., 0.],

[ 0., -15., 0.],
[ 0., 0., 0.]])

This time we got the correct principal stresses and the shear stress is
zero. This θ is the correct one.
Alternative and more systematic means and code to find the angles for
principal axes are given in the last section of this chapter.
110 Mechanics of Materials: Formulations and Solutions with Python

3.9 Stress invariants

Nine stress components at any point in solid can form three stress invariants
that do not change with the coordinates used. For example, these three prin-
cipal stresses do not change with coordinate transformations. The invariant
property is essentially the feature of any 3 × 3 matrix, including the stress
tensor matrix.
In mechanics of materials, these three stress invariants are defined as
follows, based on nine stress components or these three principal stresses.
First, the sum of the diagonal terms in σ, also known as the trace of σ,
gives the first stress invariant:
I1 = σ11 + σ22 + σ33 = σ1 + σ2 + σ3 (3.63)
Second, the sum of the three minor determinants of σ gives the second
stress invariant:
     
σ11 σ12  σ11 σ13  σ22 σ23 

I2 =   +   +  
σ12 σ22  σ13 σ33  σ23 σ33 
2 2 2
(3.64)
= σ11 σ22 + σ11 σ33 + σ22 σ33 − σ12 − σ13 − σ23
= σ1 σ2 + σ1 σ3 + σ2 σ3
Third, the determinant of σ gives the third stress invariant:
 
σ11 σ12 σ13 

 
I3 = σ12 σ22 σ23  = σ1 σ2 σ3 (3.65)
 
σ13 σ23 σ33 
For 2D cases, say in the x–y plane, we shall have two stress invariants:
I1 = σ11 + σ22 = σ1 + σ2 (3.66)
and
 
σ11 σ12 

I2 =   = σ11 σ22 − σ12
2
= σ1 σ2 (3.67)
σ12 σ22 
Due to the coordinate independent feature, the stress invariants are useful
in establishing failure criteria (see Chapter 14). A solid material failing or
not should not depend on the coordinates used to find the stress values. It
is a physical property of the material itself.

3.10 Special stresses

3.10.1 Octahedral stress

Once the principal stresses are found, we know the principal axes. The
so-called octahedral surface is defined in the principal axes with direction
 Forces and Stresses 111

Figure 3.11. The octahedral surface is defined on the

√ principal axes of stresses. Its three
normal components are the same: n1 = n2 = n3 = 1/ 3.

cosines of n = (n1 , n2 , n3 ) = ( √13 , √13 , √13 ), as shown in the figure on the left
of Fig. 3.11.
Note that there is a total of eight octahedral surfaces for a point in solid,
as shown in the figure on the right in Fig. 3.11.
On the octahedral surface, there is a special stress called octahedral stress.
It has two components, octahedral normal stress σoct and octahedral shear
stress τoct . They can be written as
1 1
σoct = I1 = (σ1 + σ2 + σ3 )
3 3
 (3.68)
2 2 2 1 2 2 2
τoct = I − I2 = (σ1 − σ2 ) + (σ1 − σ3 ) + (σ2 − σ3 )
9 1 3 3
Note that these octahedral stresses are written in stress invariants, and
hence are also invariants. Alternatively, octahedral stress can also be written
in nine components.
1
σoct = (σ11 + σ22 + σ33 )
3

1 2 − σ2 − σ2 )
τoct = (σ11 − σ22 )2 + (σ22 − σ33 )2 + (σ33 − σ11 )2 + 6(σ11 13 23
3
(3.69)
These stresses can be used in failure criteria (see Chapter 14).

3.10.2 The von Mises stress

The well-known von Mises stress is widely used in failure criteria for ductile
materials. It is related to the distortional energy density in stressed solids
112 Mechanics of Materials: Formulations and Solutions with Python

and the octahedral shear stress τoct . The von Mises stress is computed using

1 3
σvm = [(σ1 − σ2 )2 + (σ1 − σ3 )2 + (σ2 − σ3 )2 ] = √ τoct (3.70)
2 2
The von Mises stress relates to the octahedral shear stress by a constant
factor.
Alternatively, the von Mises stress can also be written in nine
components.

1 2 − σ 2 − σ 2 )]
σvm = [(σ11 − σ22 )2 + (σ22 − σ33 )2 + (σ33 − σ11 )2 + 6(σ12 13 23
2
(3.71)

3.10.3 The strain energy stress

The strain-energy stress is used in the failure criteria based on the strain-
energy density (see Chapter 13). It is the effective stress that gives the strain-
energy density:
1 1 2
U0 = [σ12 + σ22 + σ32 − 2ν(σ2 σ3 + σ1 σ3 + σ1 σ2 )] = σ (3.72)
2E   2E energy
2
σenergy

where the strain-energy stress is defined as


σenergy = σ12 + σ22 + σ32 − 2ν(σ2 σ3 + σ1 σ3 + σ1 σ2 ) (3.73)

We will study the strain energy in greater detail in Chapter 13. For now,
we use it as defined above.

3.10.4 Python code for computing various stresses

We write some useful functions for easy computation of various special
stresses, for a given stress state with nine stress components at a point,
including the principal stresses, stress invariants, the octahedral stress, mean
stress, deviatoric stress (see Section 5.14 for detailed definition), and the von
Mises stress.
Our idea is to first conduct eigenvalue analysis for the given stress matrix,
and then compute all the other stresses using the eigenvalues. Since the stress
matrix is a very small (3×3) symmetric matrix, the time taken for eigenvalue
computation is not noticeable. This makes the formulas for all the other types
of stress simple and the code is much easier to read.
 Forces and Stresses 113

1 def M_eigen(stressMatrix):
2 '''Compute the eigenvalues & eigenvectors for a given stress
3 (or strain) matrix.
4 Input stressMatrix: the stress/strain matrix with 9 components.
5 Return: the eigenvalues are sorted in descending
6 order, and the eigenvectors will be sorted accordingly. The sign
7 for eigenVectors are also adjusted based on the right-hand rule.
8 '''
9 eigenValues, eigenVectors = [Link](stressMatrix)
10
11 # Sort in descending order
12 idx = [Link]()[::-1]
13 eigenValues = eigenValues[idx]
14 eigenVectors = eigenVectors[:,idx]
15
16 # Determine the sign for given eigenVector-1 and eigenVector-3
17 eigenVectors[2,:] = [Link](eigenVectors[0,:], eigenVectors[1,:])
18 return eigenValues, eigenVectors
19
20 def von_Mises(e):
21 '''Compute the von Mises stress using principal stresses
22 Input e: eigenvalues of the stress matrix or principal stresses
23 '''
24 vonMises=[Link](((e[0]-e[1])**2+(e[0]-e[2])**2+(e[1]-e[2])**2)/2.)
25 return vonMises
26
27 def Oct_stress(e):
28 '''Compute the octahedral shear stress using principal stresses
29 Input e: eigenvalues of the stress matrix or principal stresses
30 '''
31 sigma_oct= [Link](e)/3.
32 tau_oct=[Link]((e[0]-e[1])**2+(e[0]-e[2])**2+(e[1]-e[2])**2)/3.
33 return sigma_oct, tau_oct
34
35 def energy_stress(e,nu):
36 '''Compute the strain-energy using principal stresses
37 Input e: eigenvalues of the stress matrix or principal stresses
38 nu: Poisson's ratio.
39 '''
40 U_stress=[Link](e[0]**2+e[1]**2+e[2]**2-2*nu*(e[1]*e[2]+
41 e[0]*e[2]+e[0]*e[1]))
42 return U_stress
43
44 def M_invariant(e):
45 '''Compute the 3 stress/strain invariants using principal stresses
46 Input e: eigenvalues of the stress matrix or principal stresses
47 '''
48 I1 = [Link](e)
49 I2 = e[0]*e[1]+e[1]*e[2]+e[0]*e[2]
50 I3 = e[0]*e[1]*e[2]
51 return I1, I2, I3
52
53 def meanDeviator(stressM,e):
114 Mechanics of Materials: Formulations and Solutions with Python

54 '''Compute the mean stress, mean stresses, deviatoric stresses.

55 input stressM: given stress (strain or any) matrix with 9 components
56 Input e: eigenvalues of stress/strain matrix or principal stresses
57 Input stressM: The stress/strain matrix
58 '''
59 meanStress = [Link](e)
60 meanMatrix = [Link]([meanStress,meanStress,meanStress])
61 deviatorM = stressM - meanMatrix
62 return meanStress, meanMatrix, deviatorM

3.10.5 Example: Computation of various stresses

Let us do an numerical example using the codes written above.

1 # Example 2.5 Stress Invariants:

2 # from Ref: Advanced Mechanics of Materials, by Boresi and Schmidt
3
4 np.set_printoptions(precision=4,suppress=True)
5
6 # Given stress matrix
7 #Sv = [Link]([[-10.,15., 0.],[15.,30., 0.],[0.,0.,0.]]) #Example 2.1
8 Sv = [Link]([[80.,20, 40],[20., 60, 10],[40., 10, 20]]) #Example 2.5
9 print(f'The given stress tensor =\n{Sv}\n')
10
11 eigenValues, eigenVectors = M_eigen(Sv) # or gr.M_eigen(Sv)
12 print('Principal stresses (Eigenvalues) = ',eigenValues)
13 print(f'Eigenvectors:\n{eigenVectors}')
14 print(f'Orthogonal:\n{[Link](eigenVectors.T)}\n')
15
16 I1, I2, I3 = M_invariant(eigenValues)
17 print(f'The first stress invariant I1={I1:9.5f}')
18 print(f'The second stress invariant I2={I2:9.5f}')
19 print(f'The third stress invariant I3={I3:9.5f}\n')
20
21 sigma_oct, tau_oct = Oct_stress(eigenValues)
22 print(f'The octahedral normal stress is {sigma_oct:9.5f}')
23 print(f'The octahedral shear stress is {tau_oct:9.5f}\n')
24
25 vonMises = von_Mises(eigenValues)
26 print(f'The von Mises stress is {vonMises:9.5f}\n')
27
28 U_stress = energy_stress(eigenValues, 0.3)
29 print(f'The strain-energy stress is {U_stress:9.5f}\n')
30
31 Tau_max =0.5*(eigenValues[0]-eigenValues[2])
32 print(f'The maximum shear stress is {Tau_max:9.5f}\n')
33
34 meanStress, meanMatrix, deviatorM = meanDeviator(Sv,eigenValues)
35 print(f'The mean stress = {meanStress:9.5f}')
36 print(f'The mean stress tensor:\n{meanMatrix}')
37 print(f'The deviatoric stress tensor:\n{deviatorM}')
 Forces and Stresses 115

The given stress tensor =

[[80. 20. 40.]
[20. 60. 10.]
[40. 10. 20.]]

Principal stresses (Eigenvalues) = [110. 50. 0.]

Eigenvectors:
[[-0.8165 0.3651 -0.4472]
[-0.4082 -0.9129 -0. ]
[-0.4082 0.1826 0.8944]]
Orthogonal:
[[ 1. 0. -0.]
[ 0. 1. -0.]
[-0. -0. 1.]]

The first stress invariant I1=160.00000

The second stress invariant I2=5500.00000
The third stress invariant I3= 0.00000

The octahedral normal stress is 53.33333

The octahedral shear stress is 44.96913

The von Mises stress is 95.39392

The strain-energy stress is 106.30146

The maximum shear stress is 55.00000

The mean stress = 53.33333

The mean stress tensor:
[[53.3333 0. 0. ]
[ 0. 53.3333 0. ]
[ 0. 0. 53.3333]]
The deviatoric stress tensor:
[[ 26.6667 20. 40. ]
[ 20. 6.6667 10. ]
[ 40. 10. -33.3333]]

3.11 Mohr circles, geometric representation

Mohr circle is a geometric representation of tensors of second order. This

section discusses construction of Mohr circles for both 2D and 3D stress
states. It is written in reference to the wikipage on Mohr circle ([Link]
[Link]/wiki/Mohr%27s circle) and Ref. [1].
116 Mechanics of Materials: Formulations and Solutions with Python

3.11.1 Formulation for 2D stress states, double angle

formula
To derive the equations for the Mohr circle, we consider first 2D cases, includ-
ing the plane stress and plane strain, as long as the stress components are
given in the x–y coordinate system.
Consider a point in 2D solid. Its state of stress is represented by an
infinitely small square shown in Fig. 3.12.
Using Eq. (3.53) and the double angle formulas for sine and cosine, we
can derive formulas for stresses in X–Y system that is rotated by θ with
respect to the x–y system. The code to derive these formulas is given as
follows:
1 # Define the double angle identities in a dictionary:
2
3 to2 = {[Link]( )**2:(1+[Link](2* ))/2, [Link]( )**2:([Link](2* ))/2,
4 [Link]( )*cos( ):[Link](2* )/2}
5
6 # Use sXX, sYY, sXY obtained in code given earlier.
7 sXX_ = [Link](to2 ).factor().expand()
8 print(sXX_)

σ_xx*cos( )**2 + σ_xy*sin( )*cos( ) + σ_yx*sin( )*cos( ) + σ_yy*sin( )**2

1 sYY_ = [Link](to2θ).factor().expand()
2 print(sYY_)

σ_xx*sin( )**2 - σ_xy*sin( )*cos( ) - σ_yx*sin( )*cos( ) + σ_yy*cos( )**2

Figure 3.12. 2D stress state on the surfaces of an infinitely small square, and a surface-X
with inclined angle θ.
 Forces and Stresses 117

1 sXY_ = [Link](to2θ).factor([Link](2* )).expand()

2 print(sXY_)

-σ_xx*sin( )*cos( ) + σ_xy*cos( )**2 - σ_yx*sin( )**2 + σ_yy*sin( )*cos( )

These results are summarized and re-arranged as

1 1
σXX = (σxx + σyy ) + (σxx − σyy ) cos 2θ + σxy sin 2θ
2 2
1 1
σY Y = (σxx + σyy ) − (σxx − σyy ) cos 2θ − σxy sin 2θ (3.74)
2 2
1
σXY = − (σxx − σyy ) sin 2θ + σxy cos 2θ
2
where θ is the rotation angle of the coordinate from x–y to X–Y . We write
the following Python function for future use.

3.11.2 Python code for coordinate transformation for

2D stresses, double angle
1 def tensorT2angle( xx, yy, xy, , py='sympy'):
2 '''
3 Perform coordinate transformation for 2nd order tensors (stress and
4 strain) in 2D, using double angle of (rad).
5 inputs: xx, yy, xy in x-y,
6 : rad, angle of rotation (counterclockwise) to form X-Y
7 py: if 'sympy', return sympy formulas
8 if 'numpy', return numpy object
9 return: XX, YY, XY in X-Y
10 '''
11 if py == 'sympy':
12 c2 = [Link](2* ); s2 = [Link](2* )
13 elif py == 'numpy':
14 c2 = [Link](2* ); s2 = [Link](2* )
15 else:
16 print("Please set py == 'sympy' or 'numpy'")
17
18 XX = σxx*c2 /2 + σxx/2 + σxy*s2 /2 + σxy*s2 /2 - σyy*c2 /2 + σyy/2
19 YY =-σxx*c2 /2 + σxx/2 - σxy*s2 /2 - σxy*s2 /2 + σyy*c2 /2 + σyy/2
20 XY =-σxx*s2 /2 + σxy*c2 /2 + σxy/2 + σxy*c2 /2 - σxy/2 + σyy*s2 /2
21
22 return XX, YY, XY

3.11.3 Principal stresses for 2D stress states

For given 2D stress states, one can find the principal stresses by extending
it to a 3D stress state, and use principalS() to find the principal stresses.
118 Mechanics of Materials: Formulations and Solutions with Python

Alternatively, we can make use of the double angle formula. In this way, we
can find the angle of the principal axis with ease.
Since at the principal axes, the shear stress is zero, using the third equa-
tion in Eq. (3.74), we obtain
2σxy
tan 2θp1 = (3.75)
σxx − σyy

where 2θp1 is the angle of the principal X-axis with respect to x-axis. This
angle found is unique because arctan() is single-valued for arguments in
[−∞, ∞].
After θp1 is found, the principal stresses σXX and σY Y can be computed
using Eq. (3.74). Following is the code.

1 def principal2angle( 2d):

2 '''Compute the principal stresses for a given 2D stress state 2d.
3 '''
4 p1 = 0.5*[Link](2* 2d[0,1]/( 2d[0,0]- 2d[1,1]))
5 #print(f'Angle (X,x) = { p1}(rad), {np.rad2deg( p1)}∘')
6
7 XX, YY, XY=tensorT2angle( 2d[0,0], 2d[1,1], 2d[0,1], p1,py='numpy'
8 #print(f'Principal stresses (MPa) :\n{ XX, YY, XY}')
9 return XX, YY, XY, p1

3.11.4 Example: Principal stresses of 2D stress states

In this example, we define a 2D stress state, and compute the principal
stresses using the procedure described above. We then compare the results
with that obtained using the eigenvalue solver principalS() for 3D stress state
extended from the same 2D one.
1 # Define a 2D stress state:
2 2d = [Link]([[88., 38.], # in MPa
3 [38.,-35.]]) # one may use other stresses
4 print(f'Stress Tensor (MPa) :\n{ 2d}')
5
6 XX, YY, XY, p1 = principal2angle( 2d) # Or: gr.principal2angle( 2d)
7 print(f'Angle (X,x) = { p1}(rad), {np.rad2deg( p1)}∘')
8 print(f'Principal stresses (MPa) :\n{ XX, YY, XY}')

Stress Tensor (MPa) :

[[ 88. 38.]
[ 38. -35.]]
Angle (X,x) = 0.2767336977935464(rad), 15.85567293261899∘
Principal stresses (MPa) :
(98.79280738773396, -45.792807387733944, 8.881784197001252e-15)
 Forces and Stresses 119

1 # Expand 2d to 3d, by appending zeros:

2 3d = [Link]((3,3))
3 3d[:2,:2] = 2d
4 print(f'Stress Tensor (MPa) :\n{ 3d} \n Principal stresses:')
5 eigenValues, eigenVectors = principalS( 3d)
6 print(f' Principal stresses: \n{eigenValues}')

Stress Tensor (MPa) :

[[ 88. 38. 0.]
[ 38. -35. 0.]
[ 0. 0. 0.]]
Principal stresses:
Principal stress/strain (Eigenvalues):
[ 98.7928 0. -45.7928]

Principal stress/strain directions:

[[ 0.962 0. 0.2732]
[ 0.2732 0. -0.962 ]
[ 0. 1. 0. ]]

Possible angles (n1,x)=15.85567293261899∘ or 164.144327067381∘

Principal stresses:
[ 98.7928 0. -45.7928]

As shown, we obtained the same principal stress results using these two
approaches. The double angle approach principal2angle() gives an unique
angle, but the eigenvalue solver principalS() gives two choices.

3.11.5 Mohr circle for 2D and stress states

Using the stress invariants Eqs. (3.66) and (3.67), we shall have
σXX + σY Y = σxx + σyy (I1 )
2 2
(3.76)
σXX σY Y − σXY = σxx σyy − σxy (I2 )
Using Eq. (3.74) and with the help of Eq. (3.76), we obtain,
 2 2
 2 2
σXX − 12 (σxx + σyy ) + σXY = 12 (σxx − σyy ) + σxy (3.77)
or
(σXX − σavg )2 + σXY
2
= R2 (3.78)
where σavg = 12 (σxx + σyy ) is the average or mean stress.
Equation (3.78) is clearly for a circle in the σXX ∼ σXY coordinate system
with the center at (σavg , 0) and the radius given by
 2
1 2
R= 2 (σxx − σyy ) + σxy (3.79)
120 Mechanics of Materials: Formulations and Solutions with Python

Figure 3.13. Mohr circle for 2D stress states at a point in solid. The shear stress on the
surface-x (or surface-X), which is at point B, carries a minus sign. Image modified based
on those from Wikimedia Commons, by Sanpaz under the CC BY-SA 3.0 license.

Figure 3.13 shows the detailed relationship between the a Mohr circle
and the 2D stress states. It is constructed using the equations given in the
previous section. All the states of 2D stresses are located on the circle. The
initial stress state before the coordinate transformation is at points B and
A. The stress state after the coordinate rotating counterclockwise an arbi-
trary θ corresponds to point D. Note that the angle θ in the x–y coordinate
corresponds to 2θ on the Mohr circle. The principal stress state corresponds
to the points C and E on the axis with zero shear stress. The angle of 2θp1
to the principal stress state is computed using Eq. (3.75).

3.11.6 Sign convention for 2D Mohr circles

Note that at any state, there are 2 normal stresses: σXX and σY Y , while σXX
corresponds to point B (surface-X), and σY Y corresponding to the opposite
point A (surface-Y ) on the circle, as shown in Fig. 3.13. However, at any
state, there is only one shear stress σXY that must show on the σY Y  σXY
coordinate system with different signs on the upper and lower quadrants.
Therefore, we need a sign convention to allow the shear stress change
signs. There are a few ways to do so, but no standard sign convention for
this. Here, we let the shear stress at point A be positive on the Mohr circle:
A(σY Y , σXY ), and at point B be negative: B(σXX , −σXY ). This is true also
 Forces and Stresses 121

for the initial state x−y: A(σyy , σxy ) and B  (σxx , −σxy ). This allows the
shear stress on the surface-X to carry a reversed sign. On the surface-Y , the
sign for the shear stress is unchanged. With this sign reversal, the shear
stress on the Mohr circle gives the same results, as that obtained using the
coordinate transformation rule.
One can easily confirm this sign convention by plotting the Mohr circle
for the problem shown in Fig. 3.10. When x−y rotates 45◦ to X−Y , we
obtain a negative shear stress. This will be seen in the example later when
plotting the Mohr circle for this problem.
More detailed discussions on the sign convention can be found at the
wikipage on Mohr circle.

3.11.7 Python code for generating 2D Mohr circles

Python codes are made by the author for conveniently producing the Mohr
circles for both 2D and 3D stress (or strain) states. The 2D codes are given
as follows, and are self-explanatory.

1 def Mohr_circle2D( , lw=1.0):

2 '''
3 Plot a 2D Mohr circle, for given 2×2 stress/strain matrix (tensor).
4 input: , numpy array.
5 lw, the line width of the circle.
6 Example:
7 [Link](figsize=(6, 6))
8 Mohr_circle2D([Link]([[88.,-35], [-35, 38.]]), lw=2.0)
9 '''
10 _avg = ( [0,0]+ [1,1])/2 # mean stress
11 center = ( _avg, 0) # center of the Mohr circle
12 R = [Link](( [0,0]- [1,1])**2/4+ [0,1]**2) # radius Mohr circle
13
14 [Link]( _avg, 0, c='r', marker="o", zorder=2) # circle center
15 [Link](( [1,1], [0,0]),( [0,1],- [0,1])) #current state A & B
16
17 [Link](( [1,1], [0,0]),( [0,1],- [0,1]), color='blue', lw=0.5)
18 [Link]( [1,1]+R/25., [0,1]+R/25., "A") # current state A
19 [Link]( [0,0]+R/25.,- [0,1]+R/25., "B") # current state B
20
21 circle = [Link](center, R, color='blue', fill=False, lw=lw)
22 [Link]().add_patch(circle) #[Link]().invert_yaxis()
23
24 [Link]('scaled')
25 [Link]( _avg-R-R/10, _avg+R+R/10)
26 [Link]('$ _{XX}$'); [Link]('$ _{XY}$')
27 #[Link]()
122 Mechanics of Materials: Formulations and Solutions with Python

3.11.8 Example: Mohr circle of a uniaxial stress state

Consider the uniaxial stress state defined it as a 2D stress state in x–y plane.
 
σxx σxy 1 0
= (3.80)
σyx σyy 0 0
This is the stress state shown in Fig. 3.10. This stress tensor can be
represented in a single Mohr circle, by simply using
1 # Consider a bar subject to tension xx = 1.0:
2 xx = 1.0; yy = 0.; xy= yx=0. # stress tensor at initial state
3 _tensile = [Link]([[ xx, xy], # [Link]([[1, 0], [0, 0]])
4 [ yx, yy]])
5
6 [Link](figsize=(3, 3)) # for change the figsize, when needed
7 Mohr_circle2D( _tensile, lw=2.0) # or gr.Mohr_circle2D(SM, lw=2.0)

The initial stress state before a coordinate transformation is at points B

and A. The center is the mean stress (0.5), and the radius is the absolute
value of the maximum shear stress (also 0.5). The principal stress state
corresponds to the two points on the axis with zero shear stress.
When the coordinates rotates by θ = 45 degrees, the shear stress found
using coordinate transformation formula is −0.5, as shown in Fig. 3.10 (or
by the following code). It is the minimum shear stress state.
 Forces and Stresses 123

Based on our sign convention for the Mohr circle given earlier, point
B is at (σXX , −σXY ). Therefore, it is at the top of the Mohr circle, due
to the sign reversal. The following code computes the transformed stress
components and draws the point on the same Mohr circle.

1 # Rotate _deg:
2 _deg = 45. # this gets to the top point on the Mohr circle
3 = np.deg2rad( _deg)
4 XX, YY, XY = tensorT2angle( xx, yy, xy, , py='numpy')
5
6 _45 = [Link]([[ XX, XY],
7 [ XY, YY]])
8 print(f" At = { _deg}∘, stress tensor via coordinate transfer: "
9 f" \n{ _45} the shear stress should be = { XY} (the minimum)")
10
11 print(f" Point B( XX,- XY) is at the circle top (sign reversal)")
12
13 [Link](figsize=(3, 3)) # for change the figsize, when needed
14 Mohr_circle2D( _45, lw=2.0) # or gr.Mohr_circle2D(SM, lw=2.0)

At = 45.0∘, stress tensor via coordinate transfer:

[[ 0.5 -0.5]
[-0.5 0.5]] the shear stress should be = -0.5 (the minimum)
Point B( XX,- XY) is at the circle top (sign reversal)
124 Mechanics of Materials: Formulations and Solutions with Python

Figure 3.14. Stresses in (X, Y, Z) in a bar subjected to a uniaxial force in the x-direction.
Rotation angle: −45◦ . The arrows are the actual direction of the stresses.

If we rotate θ = −45 degrees, the shear stress value found using the coor-
dinate transformation formula should be σXY = 0.5, which is the maximum
shear stress. On the Mohr circle, point B gets on the bottom at (σXX , −σXY ),
based on our sign convention. The actual directions of all these stress com-
ponents are drawn in Fig. 3.14.
The following code computes the transformed stress components and
draws the Mohr circle.

1 # Rotate _deg:
2 _deg =-45. # this gets to the bottom point on the Mohr circle
3 = np.deg2rad( _deg)
4 XX, YY, XY = tensorT2angle( xx, yy, xy, , py='numpy')
5
6 _135 = [Link]([[ XX, XY],
7 [ XY, YY]])
8 print(f" At = { _deg}∘, stress tensor via coordinate transfer: "
9 f" \n{ _135} the shear stress should be = { XY} (the maximum)")
10
11 print(f" Point B( XX,- XY) is at the circle bottom (sign reversal)")
12
13 [Link](figsize=(3, 3)) # for change the figsize, when needed
14 Mohr_circle2D( _135, lw=2.0) # or gr.Mohr_circle2D(SM, lw=2.0)

At = -45.0∘, stress tensor via coordinate transfer:

[[0.5 0.5]
[0.5 0.5]] the shear stress should be = 0.5 (the maximum)
Point B( XX,- XY) is at the circle bottom (sign reversal)
 Forces and Stresses 125

This example is simple, but ideal for revealing the differences of the values
of stresses obtained by the coordinate transformation rule and their presen-
tations on the point on the same Mohr circle.

3.11.9 Example: Mohr circle of a general 2D stress

state
Consider 2D solid in x–y plane with general stress state given as
 
σxx σxy 3 −1
= (3.81)
σyx σyy −1 2

The Mohr circle can be produced by simply using

1 SM = [Link]([[3, -1],
2 [-1, 2]])
3 #[Link](figsize=(6, 6)) # for change the figsize, when needed
4 Mohr_circle2D(SM, lw=2.0)
126 Mechanics of Materials: Formulations and Solutions with Python

We observe again that the initial stress state before the coordinate trans-
formation is at points B and A. The center is the mean stress, and the radius
is the absolute value of the maximum shear stress. The principal stress state
corresponds to the two points on the axis of zero shear stress.

3.11.10 Mohr circle for 3D stress states*

To construct the Mohr circle for stress states at a point in a 3D solid, we
assume that the values of the principal stresses σ1 , σ2 , σ3 and their principal
directions n1 , n2 , n3 are known. Those can be found via an eigenvalue anal-
ysis, which is easy to do, and we have done multiple times. These principal
stresses are ranked so that σ1 > σ2 > σ3 .
We can then consider the principal stress state with principal axes as the
initial state. Using Eq. (3.42), the normal stress on the plane with normal n
with respect to the principal axes is given by
σnn = ni σij nj = n21 σ1 + n22 σ2 + n23 σ3 (3.82)
This is because all σij = 0 for i = j and σij = σi for i = j on the principal
axes.
Using Eq. (3.46) the share stress on the plane with n is given by
2
σns = σ12 + σ22 + σ32 − σnn
2
(3.83)
 Forces and Stresses 127

In addition, we have (from the orthogonality of the coordinates):

n21 + n22 + n23 = 1 (3.84)
We can solve the system of three Eqs. (3.82), (3.83), (3.84) for three
unknowns n21 , n22 , and n23 . The results are as follows:
2 + (σ
σns nn − σ2 )(σnn − σ3 )
n21 = ≥0
(σ1 − σ2 )(σ1 − σ3 )
2 + (σ
σns nn − σ3 )(σnn − σ1 )
n22 = ≥0 (3.85)
(σ2 − σ3 )(σ2 − σ1 )
2 + (σ
σns nn − σ1 )(σnn − σ2 )
n23 = ≥0
(σ3 − σ1 )(σ3 − σ2 )
The “≥0” on the right-hand side of the equation is because of the square
of ni .
Next because σ1 > σ2 > σ3 , the numerators in Eq. (3.85) satisfy
2
σns + (σnn − σ2 )(σnn − σ3 ) ≥ 0, ∵ σ1 − σ2 > 0, σ1 − σ3 > 0
2
σns + (σnn − σ3 )(σnn − σ1 ) ≤ 0, ∵ σ2 − σ3 > 0, σ2 − σ1 < 0 (3.86)
2
σns + (σnn − σ1 )(σnn − σ2 ) ≥ 0, ∵ σ3 − σ1 < 0, σ3 − σ2 < 0
The foregoing equations can be rewritten as
2
 2  2
σns + σnn − 12 (σ2 + σ3 ) ≥ 12 (σ2 − σ3 )
2
 2  2
σns + σnn − 12 (σ1 + σ3 ) ≤ 12 (σ1 − σ3 ) (3.87)
2
 2  2
σns + σnn − 12 (σ1 + σ2 ) ≥ 12 (σ1 − σ2 )
These are the equations of three circles in the coordinate σnn −σns denoted
as C1 , C2 , and C3 . Their radii are given respectively as
R1 = 12 (σ2 − σ3 )
R2 = 12 (σ1 − σ3 ) (3.88)
R3 = 12 (σ1 − σ2 )
The corresponding centers are at
1 
2 (σ2 + σ3 ), 0
1 
2 (σ1 + σ3 ), 0 (3.89)
1 
2 (σ1 + σ2 ), 0
128 Mechanics of Materials: Formulations and Solutions with Python

Figure 3.15. Mohr circle for 3D stress states at a point in solid. Image modified base
on that from the Wikimedia Commons in public domain, by Sanpaz.

Equations (3.87) are the Mohr circles for 3D stress states. The inequality
means that the stress states at any point in solid (σnn , σns ) shall be on these
circles or within the shaded green area enclosed by these circles shown in
Fig. 3.15.

3.11.11 Python code for generating 3D Mohr circles

1 def Mohr_circle3D( , lw=1.0):

2 '''
3 Plot a 3D Mohr circle, for given 3×3 stress/strain matrix (tensor)
4 in numpy array. lw: the line width of the circle.
5 Example:
6 [Link](figsize=(6, 6))
7 Mohr_circle3D([100., 20., -50.], lw=2.0)
8 '''
9 p , _ = [Link]( ) # ranked principal stresses
10
11 _avg12 = (p [0]+p [1])/2 # 3 average stresses
12 _avg13 = (p [0]+p [2])/2
13 _avg23 = (p [1]+p [2])/2
14
15 center1 = ( _avg12, 0) # 3 centers
16 center2 = ( _avg13, 0)
17 center3 = ( _avg23, 0)
18
 Forces and Stresses 129

19 R1 = (p [0]-p [1])/2 # 3 radii

20 R2 = (p [0]-p [2])/2
21 R3 = (p [1]-p [2])/2
22
23 # centers of 3 circles:
24 circle2 = [Link](center2, R2, edgecolor='blue', \
25 facecolor='lightgreen', lw=lw)
26 circle1 = [Link](center1, R1, edgecolor='c', \
27 facecolor='white', lw=lw)
28 circle3 = [Link](center3, R3, edgecolor='orange', \
29 facecolor='white', lw=lw)
30
31 [Link]().add_patch(circle2) # plot on the same plot
32 [Link]().add_patch(circle1)
33 [Link]().add_patch(circle3)
34
35 [Link]('scaled')
36 [Link]('$ _{nn}$'); [Link]('$ _{ns}$')
37
38 # three principal stresses:
39 [Link](p , (0,0,0), c='r', marker="o", s=6, zorder=2)
40
41 # three centers:
42 centers = ( _avg12, _avg13, _avg23)
43 [Link](centers, (0,0,0), c='k', \
44 marker="o", facecolor='none', s=4, zorder=2)
45
46 [Link](centers,( R1, R2, R3), c='r', marker="o",\
47 s=6, zorder=2) # maximum shear stress (positive)
48 [Link](centers,(-R1,-R2,-R3), c='r', marker="o",\
49 s=6, zorder=2) # maximum shear stress (negative)
50 [Link]()

3.11.12 Example: Mohr circle of a three-dimensional

(3D) stress state
Consider general problems in 3D. The stress tensor will be 3D, and can be
represented in Mohr circles, and there are in general three of them connected
together. The 3D codes are given as follows, and are self-explanatory.
Following is an example.
1 matrix = [Link]([[2, 1, 3], # stresses at a point in 3D solid
2 [1, 9, 2],
3 [3, 2, 7]])
4
5 #[Link](figsize=(6, 6)) # for adjusting figsize
6 Mohr_circle3D( matrix, lw=2.0) # or gr.Mohr_circle3D( matrix, lw=2.0)
130 Mechanics of Materials: Formulations and Solutions with Python

Principal stress/strain (Eigenvalues):

[10.9578 6.4478 0.5945]

Principal stress/strain directions:

[[-0.2832 0.3133 -0.9065]
[-0.7523 -0.6588 0.0074]
[-0.5948 0.684 0.4222]]

Possible angles (n1,x)=106.4510639474349 or 73.5489360525651

The three principal stresses (the initial stress state) are at the “kissing”
points of these three circles, corresponding to zero shear stress. The centers
of these three circles are the average values of a pair of the principal stresses.
The radius of each circle is the absolute value of a maximum shear stress.
The maximum shear stress is the radius of the largest circle.
Alternatively, readers may use the following external modules contin-
uum mechanics to plot Mohr circles. The sample codes are given as follows:
1 #from continuum_mechanics.visualization import mohr2d, mohr3d
2 #mohr3d(Matrix([[2, 1, 3], [1, 9, 2], [3, 2, 7]]));
3 #mohr3d(Matrix([[2, 1, 3], [1, 9, 2], [3, 2, 7]]))
4 #SM = Matrix([[1, 0], [0, 0]]); mohr2d(SM);
 Forces and Stresses 131

3.12 Determination of the principal axis angle*

Solving the eigenvalue problem given in Eq. (3.59) gives only the possible
direction cosines of the principal axis. Also, the eigenvector is not unique. For
example, an n satisfies Eq. (3.59), −n also satisfies it. Therefore, a direction
cosine, say n1 , can be set as both negative and positive, as long as n2 and
n3 change their signs accordingly. This creates problems if one would want
to find precise angle of the principal axis.
For 2D cases, we can use Eq. (3.75), the angle found is unique because
arctan(x) is single-valued for x in [−∞, ∞]. Alternatively, we can perform
tests as done in Section 3.8.4 to find the angle.
For stresses in 3D, there will be three angles, and hence finding those can
be a challenge. This section discusses techniques for 3D cases.

3.12.1 Transformation matrices for Yaw, Pitch,

and Roll
Let us now consider the general cases of 3D stress states. In such a case,
all these 6 stress components may not be zero. We need to find the prin-
cipal stress and the corresponding precise angles. This problem becomes
quite complicated, and the solution cannot be found in existing textbooks
and open literature, to the best knowledge of the author. The author has
found an effective method to get this done, and the strategy is given as
follows:

1. Create the transformation matrix using the Yaw, Pitch, and Roll formu-
lation given in the previous chapter in, for example, Sympy formulas.
2. Perform the coordinate transformation to any given general stress state,
which leads to three expressions for shear stresses on all the off-diagonal
terms. These expressions will be in the Yaw angle θ, Pitch angle β, and
Roll angle γ, and are in general quite complicated full-page long formulas.
3. Force these three expressions to zero to find the angles, θ, β, γ. This can
be done using an equation solver.

Due to the tensor properties, we know that the solution for θ, β, γ exists.
This procedure is essentially a diagonalization procedure. It produces a
diagonal stress matrix. It not only finds the angles, θ, β, γ, but also these
principal stresses that are on the diagonal.
The following are the codes to get it done.
132 Mechanics of Materials: Formulations and Solutions with Python

1 # define variables for angle of rotation

2 , , = symbols(" , , ")
3
4 # Create the matrices for Yaw, Pitch, and Roll for transformations:
5 Tz = gr.transf_YPRs( , about = 'z')[0]
6 Ty = gr.transf_YPRs( , about = 'y')[0]
7 Tx = gr.transf_YPRs( , about = 'x')[0]
8
9 print(f'Ty = {Ty}, \nTx = {Tx}')
10 Tz # take a look at one of these

Ty = Matrix([[cos( ), 0, -sin( )], [0, 1, 0], [sin( ), 0, cos( )]]),

Tx = Matrix([[1, 0, 0], [0, cos( ), sin( )], [0, -sin( ), cos( )]])
⎡ ⎤
cos (θ) sin (θ) 0
⎣− sin (θ) cos (θ) 0⎦
0 0 1

1 # Construct the transformation matrix for Yaw, Pitch and Roll together.
2 T_ypr = Tz@Ty@Tx
3 T_ypr

⎡
cos (β) cos (θ) sin (β) sin (γ) cos (θ) + sin (θ) cos (γ)
⎢
⎣− sin (θ) cos (β) − sin (β) sin (γ) sin (θ) + cos (γ) cos (θ)
sin (β) − sin (γ) cos (β)
⎤
− sin (β) cos (γ) cos (θ) + sin (γ) sin (θ)
⎥
sin (β) sin (θ) cos (γ) + sin (γ) cos (θ) ⎦
cos (β) cos (γ)

3.12.2 Stress transformation via Yaw, Pitch, and Roll

We now perform stress transformation using all three Yaw, Pitch, and Roll
angles.

1 # Define a stress tensor:

2 St = [Link]([[10, 15, 8], [15, -30, 5], [8, 5, 9]]) # stress matrix
3
4 # Perform coordinate transformation to the given stress tensor.
5 S_ypr = T_ypr@St@T_ypr.T
6
7 #S_ypr[0,1] # reader may take a look at one shear stress (1/3 page)

These three shear stress components will be in Sypr [0, 1], Sypr [0, 2],
Sypr [1, 2]. These are all quite lengthy.
 Forces and Stresses 133

3.12.3 Diagonalization of a stress tensor in 3D

To obtain the principal stresses, we can find a set of Yaw, Pitch, and Roll
angles, under which the stress tensor becomes diagonal. This means that all
these off-diagonal shear stresses must be zero. This provides three equations.
We shall solve these equations for these three angles: θ, β, γ.
This set of equations are nonlinear. There are many Python methods to
solve this type of equations. Here, we use [Link]() for this task.

1 # Import the external solver: fsolve:

2 from [Link] import fsolve
3
4 # Create these three equations:
5 eqns = lambda x: [S_ypr[0,1].subs({ :x[0], :x[1], :x[2]}),
6 S_ypr[0,2].subs({ :x[0], :x[1], :x[2]}),
7 S_ypr[1,2].subs({ :x[0], :x[1], :x[2]})]
8
9 eqns([0.1, 0.1, 0.1]) # test the lambda function defined

[10.7685926296899, 7.56582183504705, 9.59290586151537]

For solving an nonlinear equations, we often need to give an initial guess,
which can be essentially arbitrary. We set it with random values, but make
these values small so that the solutions found can be near zero. This is
because our functions are nonlinear, sinusoidal and periodical. There are
infinite number of solutions in (−∞, ∞). We would like to have the primary
solutions within (−π, π).

1 init_guess = [Link](3)/999. # make it small

2
3 # Solve the set of three equations numerically:
4 sln = fsolve(eqns, init_guess)
5
6 print(f'Solution, , , = {sln} (rad)')
7 print(f'Solution, , , = {sln*180/[Link]} (degree)')

Solution, , , = [ 0.3839 -0.6111 0.173 ] (rad)

Solution, , , = [ 21.9951 -35.0143 9.9117] (degree)

We find the solutions in a few seconds.

Finally, we can check the results, by using these θ, β, γ angles found to
perform coordinate transformation:
1 T_ypr_ = T_ypr.subs({ :sln[0], :sln[1], :sln[2]})
2 S_ypr_ = T_ypr_@St@T_ypr_.T
3 [Link](S_ypr_, 'Principal stresses found',n_dgt=5)

Principal stresses found

134 Mechanics of Materials: Formulations and Solutions with Python

⎡ ⎤
21.679 −3.4318 · 10−10 −1.8484 · 10−10
⎣−3.4318 · 10−10 −35.115 2.3496 · 10−10 ⎦
−1.8483 · 10−10 2.3496 · 10−10 2.4368
We obtained all the expected results. This process is summarized into a
code function for future use.

3.12.4 Python code to find Yaw, Pitch, and Roll angles

to diagonalize stress tensors
1 def principalS_ypr(St):
2 '''Diagonalization of a stress tensor in 3D, through coordinate
3 transformation via three Yaw, Pitch and Roll angles.
4 Input: St: Stress tensor, array like, 3 by 3
5 return: , , : Yaw, Pitch and Roll angles, in rad.
6 TS_ypr: Diagonalized stress matrix array like, 3 by 3
7 '''
8 from [Link] import fsolve
9
10 , , = [Link](" , , ") # Angles for Yaw, Pitch and Roll
11
12 # Create the matrices for Yaw, Pitch, and Roll for transformations:
13 Tz = gr.transf_YPRs( , about = 'z')[0]
14 Ty = gr.transf_YPRs( , about = 'y')[0]
15 Tx = gr.transf_YPRs( , about = 'x')[0]
16
17 # Construct the transformation matrix for Yaw, Pitch and Roll.
18 T_ypr = Tz@Ty@Tx
19
20 # Perform coordinate transformation to the given stress tensor.
21 S_ypr = T_ypr@St@T_ypr.T
22
23 # Create these three equations:
24 eqns = lambda x: [S_ypr[0,1].subs({ :x[0], :x[1], :x[2]}),
25 S_ypr[0,2].subs({ :x[0], :x[1], :x[2]}),
26 S_ypr[1,2].subs({ :x[0], :x[1], :x[2]})]
27
28 # Give an initial guess randomly:
29 init_guess = [Link](3)/999. # make it small
30
31 # Solve the set of three equations numerically:
32 sln = fsolve(eqns, init_guess)
33 print(f'Solution, , , = {sln} (rad)')
34 print(f'Solution, , , = {sln*180/[Link]} (degree)')
35
36 # Finally, we can check the results, use , , found to
37 # perform coordinate trans formation:
38 T_ypr_ = T_ypr.subs({ :sln[0], :sln[1], :sln[2]})
39 S_ypr_ = T_ypr_@St@T_ypr_.T
40
41 return , , , S_ypr_
 Forces and Stresses 135

3.12.5 Example: Diagonalize a stress tensor via Yaw,

Pitch, and Roll

1 St = [Link]([[10, 15, 8], [15, -30, 5], [8, 5, 9]]) # stress matrix
2 , , , S_ypr_ = principalS_ypr(St)
3
4 [Link](S_ypr_, 'Principal stresses found',n_dgt=5)

Solution, , , = [ 0.3839 -0.6111 0.173 ] (rad)

Solution, , , = [ 21.9951 -35.0143 9.9117] (degree)
Principal stresses found

⎡ ⎤
21.679 −5.7027 · 10−11 −4.9118 · 10−11
⎢ ⎥
⎣ −5.703 · 10−11 −35.115 2.4794 · 10−11 ⎦
−4.9117 · 10−11 2.4795 · 10−11 2.4368

The code function can, of course, be used for finding the angles for these
2D examples studied earlier:

1 St = [Link]([[-10, 15, 0], [15, 30, 0], [0, 0, 0]]) # stress tensor
2 , , , S_ypr_ = principalS_ypr(St)
3
4 [Link](S_ypr_, 'Principal stresses found',n_dgt=5)

Solution, , , = [-0.3218 -0. -0. ] (rad)

Solution, , , = [-18.4349 -0. -0. ] (degree)
Principal stresses found

⎡ ⎤
−15.0 9.77 · 10−15 2.1503 · 10−12
⎢ ⎥
⎣8.8818 · 10−15 35.0 6.3698 · 10−13 ⎦
2.1503 · 10−12 6.3698 · 10−13 −2.9666 · 10−25

Since this example is a 2D stress state, the pitch and roll angles are zero.
We can also test it for stress tensor generated randomly:

1 n = 3
2 St = [Link](n,n) # Generate a random matrix
3 St = (St + St.T)/2 # random symmetric matrix
4 print(f'Stress state:\n{St}\n')
5
6 , , , S_ypr_ = gr.principalS_ypr(St)
7 [Link](S_ypr_, 'Principal stresses found',n_dgt=5)
136 Mechanics of Materials: Formulations and Solutions with Python

Stress state:
[[ 1.1661 0.8021 -0.2455]
[ 0.8021 -0.2215 -0.0158]
[-0.2455 -0.0158 -0.5416]]

Solution, , , = [ 0.3917 -0.1998 -0.7082] (rad)

Solution, , , = [ 22.4422 -11.4482 -40.5775] (degree)
Principal stresses found
⎡ ⎤
1.5581 9.3682 · 10−13 −4.8572 · 10−15
⎢ ⎥
⎣ 9.369 · 10−13 −0.66529 1.2548 · 10−12 ⎦
−4.8295 · 10−15 1.2549 · 10−12 −0.48968

We obtained all these yaw, pitch, and roll angles, and the principal
stresses in the diagonal terms in the output matrix.

3.13 Remarks
This busy chapter discusses first how the stress tensor (order 2) result from
the external force tensor (order 1) all defined in an orthogonal coordinate
system, and then the states of stress and its change when the coordinate
system rotates.
1. The value of the stress tensor depends on both the force tensor and the
surface of the stress is on. This is the root reason why a stress tensor has
two indexes and hence being a second-order tensor.
2. In all the formulas derived for the stress tensor, orthogonal projections
for both forces and areas are strictly followed.
3. When coordinate rotates, the presentation of the stress tensor changes. It
uses twice the transformation matrix because it is a second-order tensor.
4. The basic properties of a stress tensor (three stress invariants, principal
stresses, and the Mohr circle) do not change with the coordinate trans-
formation. Special stresses are defined using these invariant property of
stresses. These special stress will be used in establishing failure criteria.
5. At the principal axes, the shear stress components vanish, and only the
normal stresses present, which are called principal stresses. These are
invariant.
6. Principal stresses and the axes can be found using the eigenvalue solver
provided in this chapter.
7. Mohr circle is a geometric representation of stresses in 2D and 3D. For a
given stress tensor, the circle is determined. Stress states of all possible
rotated coordinates stay on the circle (for 2D) or in the region on and
bounded by these circles (for 3D). It is useful to intuitively locate the
principal stresses and maximum shear stresses.
 Forces and Stresses 137

8. The Mohr circle applies exactly the same way to the strain tensor to be
discussed in the following chapter.
9. A code principalS ypr() has been developed to find the three angles: Yaw,
Pitch and Roll, for the principal axes for 3D stress states. It produces also
a diagonalized stress tensor.
With the detailed formulations, discussions, examples, and code provided
to play with in the chapter, we shall have a good understanding on forces,
stresses, stress states in the solids, on how to perform coordinate transfor-
mation with them, and how to find the special stresses for given a stress
state.
The following chapter will discuss another important pair of tensors: dis-
placement vector (first order) and strain tensor (second order).

References
[1] A.P. Boresi, R.J. Schmidt and O.M. Sidebottom, Advanced Mechanics of Materials.
John Wiley & Sons, New York, 1985.
[2] G.R. Liu, Calculus: A Practical Course with Python, World Scientific, New Jersey,
2025.
[3] G.R. Liu and S.S. Quek, The Finite Element Method: A Practical Course.
Butterworth-Heinemann, Amsterdam, 2013.
[4] G.R. Liu and T.T. Nguyen, Smoothed Finite Element Methods. Taylor and Francis
Group, New York, 2010.
This page intentionally left blank
 Chapter 4

Displacements and Strains

Let us import the necessary external modules or dependencies for later use
in this chapter.
1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.
2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 init_printing(use_unicode=True) # For latex-like quality printing
11 np.set_printoptions(precision=4,suppress=True) # Digits in print-outs

4.1 Introduction
When a solid material is loaded, points in the solid are displaced (in a con-
tinuous form). This results in a field of displacements over the solid body.
The displacement at a point has three components, each representing the
displacement in a direction. Hence, displacement is a vector, a tensor of first
order. When the coordinates transform (rotate), the displacement transforms
in the same way as the force vector discussed in the previous chapter.
If the field of displacements is not uniform (otherwise, it is trivial and
the solid undergoes a rigid body movement), it results in a field of strains
over the solid body. In general, displacement components are functions of
coordinates, forming a vector of functions. In this book, we assume that
all the displacement functions are continuous. This means that there is not
breakage in the material under loading. The gradient of the displacement
vector function leads to the strain field in the solid.

139
140 Mechanics of Materials: Formulations and Solutions with Python

Strains are tensors of second order. The components of the strains change
when the coordinate transforms. The transformation rule for strains is the
same as the stress tensor discussed in the previous chapter. We will find out
why this is true.
This chapter discusses the details about these important displacement-
strain relations, including the representation of displacement vector in
3D space, coordinate transformation rule, how the strains are resulted
from the derivatives of the displacement components, derivation of the
strain transformation rule, and strain invariants. This chapter is written
in reference to [1–6].
As usual, our discussion is accompanied with Python codes written in ref-
erence to the documentations of Python, Sympy, and other Python modules
with assistance from the AI tools, especially ChatGPT.

4.2 Displacement vector, representations

Forces on a solid can result in displacements. Displacement is a vector

because it has two basic properties: magnitude and direction. Consider an
infinitely small block representing a point in a solid defined in a 3D Carte-
sian coordinate system. The block is somehow moved (displaced) to a new
location, as shown in Fig. 4.1.

Figure 4.1. An infinitely small block is displaced in space. A vector u is used to repre-
sent the displacements. The displacement vector can be decomposed to three components
following the orthogonal projection rule.
 Displacements and Strains 141

We now present the displacements of the block using a vector u under

a Cartesian coordinate system, in the same way for a force vector. The dis-
placement vector is first decomposed to two vectors: a vertical component
u3 i3 and us following the orthogonal projection rule. The us is then further
decomposed into two vectors: u1 i1 and u2 i2 following also the orthogonal
projection rule. This process results in three components of u: u1 , u2 , and
u3 , respectively, on x1 -, x2 -, and x3 -axis. The displacement vector can be
written as
⎡ ⎤
u1
u = u2 ⎦
⎣ (4.1)
u3
It is clear that this representation of a vector is complete and unique
under the Cartesian coordinate system, implying that any change, in length
or direction, in u will result in unique changes in u1 , u2 , and u3 , and vice
versa, as long as the orthogonal projection rule is followed.
When we denote the coordinates as (x, y, z), the corresponding displace-
ment components are often denoted as ux , uy , uz or in shorthand u, v, w.
In solid mechanics, we deal with deformable solids. Therefore, any point
in a solid body can have displacements, and they often change with the
coordinates. Thus, we have in general displacement vector functions.
Note also that there are many other vectors, such as velocity and accel-
eration, rooted at the displacement vector. The same rules of representation
apply.

4.3 Displacement vector in transformed coordinates

4.3.1 Forward transformation

Consider a displacement vector u in coordinate x, and the transformed
(rotated) coordinate x . Using the transformation tensor T defined in Section
2.10, the values of vector u becomes
ui = aij uj (4.2)

where ui is obtained following the orthogonal projection rule. On the right-
hand side of Eq. (4.2), we see that the 2nd index j of T is contracted with
the index of u. This leaves one index i, implying that the outcome is a vector.
In matrix notation, we have
u = Tu (4.3)
which achieves the correct contraction through the standard matrix multi-
plication rule.
142 Mechanics of Materials: Formulations and Solutions with Python

4.3.2 Inverse transformation

Since T is unitary (see Section 2.10), the inverse transformation can be
simply done using the transpose of T:

u = T u (4.4)

Or in the indicial notation:

ui = aji uj (4.5)

All we need is to swap the order of the two indexes of aij . This can also
be achieved by contracting the first index of aij .

4.4 Vector of displacement functions

4.4.1 Displacement functions, a vector function

Since each displacement component is a function of coordinates, and the
displacement components form a vector, we have displacement vector func-
tions. Let us define symbolic functions for each displacement components,
and then form the displacement vector function in a Sympy code.

1 # Define each component of the displacement vector as a function of

2 # coordinates
3 x1, x2, x3 = [Link]('x1, x2, x3') # symbolic coordinates
4 u1 = Function('u1')(x1, x2, x3) # symbolic function of coordinates
5 u2 = Function('u2')(x1, x2, x3)
6 u3 = Function('u3')(x1, x2, x3)
7
8 u = Matrix([u1, u2, u3]) # form a vector of symbolic functions
9 u # print out
⎡ ⎤
u1 (x1 , x2 , x3 )
⎣u2 (x1 , x2 , x3 )⎦
u3 (x1 , x2 , x3 )

In addition, at any point in solid, the components of the displacement

vector may change when the coordinate at that point transforms. Thus, we
need to evaluate such a change due to the coordinate transformation.
Let us create a coordinate transformation matrix and compute the new
displacement vector function symbolically.

1 T = [Link]() # form a symbolic transformation matrix

2 T # print out
 Displacements and Strains 143

⎡ ⎤
a11 a12 a13
⎣a21 a22 a23 ⎦
a31 a32 a33

1 up=T@u # transformed displacements; @ is shorthand for multiplication

2 printM(up, 'Formulas for vector components of u',n_dgt=4)

Formulas for vector components of u

⎡ ⎤
a11 u1 (x1 , x2 , x3 ) + a12 u2 (x1 , x2 , x3 ) + a13 u3 (x1 , x2 , x3 )
⎣a21 u1 (x1 , x2 , x3 ) + a22 u2 (x1 , x2 , x3 ) + a23 u3 (x1 , x2 , x3 )⎦
a31 u1 (x1 , x2 , x3 ) + a32 u2 (x1 , x2 , x3 ) + a33 u3 (x1 , x2 , x3 )

This gives closed-form formulas for computing the new displacement com-
ponent functions. If we substitute the displacement vector with a vector of
fixed values for the transformation matrix, we have simplified formulas. Let
us show this by first generating randomly such a vector with fixed numerical
values.

1 # A test on a random vector with fixed numerical values.

2 [Link](8) # use a seed for re-producibility
3 n = 3
4 Vr = [Link](n,1) # Generate a random vector at a point
5 print(Vr)

[[ 0.0912]
[ 1.0913]
[-1.947 ]]

Now, compute the transformed displacement component functions at the

point.

1 upT = [Link](list(zip(u, Matrix(Vr))))

2 printM(upT, 'Reduced formulas for vector components of u', n_dgt=4)

Reduced formulas for vector components of u

⎡ ⎤
0.0912a11 + 1.091a12 − 1.947a13
⎣0.0912a21 + 1.091a22 − 1.947a23 ⎦
0.0912a31 + 1.091a32 − 1.947a33

These are reduced closed-form formulas for computing new displacement

components for any given transformation matrix T. If we further substi-
tute T with fixed numerical values, we obtain the displacement vector in
numerical values. Let us generate a T with random numbers.
144 Mechanics of Materials: Formulations and Solutions with Python

1 theta = 0.
2 Tr, about = [Link](theta, about = "random")
3 print(Tr)

[[-0.5727 -0.3829 0.7249]

[ 0.7137 0.2022 0.6706]
[-0.4033 0.9014 0.1575]]

1 upv = [Link](list(zip(T, Matrix(Tr))))

2 printM(upv, 'Transformed random vector by a random T:', n_dgt=4)

Transformed random vector by a random T:

⎡ ⎤
−1.881
⎣ −1.02 ⎦
0.6404

The above example shows that one can perform coordinate transforma-
tion for displacement vector function at any point in the solid, and at the
same time evaluate its variation with the coordinates.

4.4.2 Code using tensor operation for transformation

Alternatively, the same transformation can be done using tensor notation.
The following is a Python code for transformation for a first-order tensor,
such as a displacement vector.

1 def Tensor1_transfer(T,u):
2 '''Transformation for 1st order tensor u using transformation
3 matrix T.
4 u and T are array alike.
5 '''
6 S = [Link](T, u, axes=([1],[0])) # use numpy tensordot()
7 # contracting axis 1 of T with axis 0 of u
8 return S

1 Tu = Tensor1_transfer(T, u) # or: gr.Tensor1_transfer(T, u)

2 TuS = Matrix(Tu) #Covert to sympy matrix
3 print(f"Transformed vector:")
4 TuS # print out
 Displacements and Strains 145

Transformed vector:
⎡ ⎤
a11 u1 (x1 , x2 , x3 ) + a12 u2 (x1 , x2 , x3 ) + a13 u3 (x1 , x2 , x3 )
⎢ ⎥
⎣a21 u1 (x1 , x2 , x3 ) + a22 u2 (x1 , x2 , x3 ) + a23 u3 (x1 , x2 , x3 )⎦
a31 u1 (x1 , x2 , x3 ) + a32 u2 (x1 , x2 , x3 ) + a33 u3 (x1 , x2 , x3 )

These formulas are the same as those obtained earlier using matrix mul-
tiplication.
We can also perform numerically the transformation using the same func-
tion. Following is an example for randomly generated transformation matrix
and vector:

1 TrV = Tensor1_transfer(Tr, Vr)

2 p_text = 'Random vector in a transformed coordinates with random T:'
3 printM(Matrix(TrV), p_text, n_dgt=4)

Random vector in a transformed coordinates with random T:

⎡ ⎤
−1.881
⎣ −1.02 ⎦
0.6404

We obtained the same results as we did when using matrix transforma-

tion. We can transform it back:

1 print(Tensor1_transfer(Tr.T, TrV)) # one may use Tr with swapping axes

[[ 0.0912]
[ 1.0913]
[-1.947 ]]

This is the vector before coordinate transformation.

4.5 Strains derived via kinematic relations

Strains are often defined based on the following mechanics principles [1,5,6].
We do so here starting with a simple case.

4.5.1 An ideal uniform bar under tension

Consider a uniform bar with length L and height h. It is constrained at the
left-end, and subjected to a uniformly distributed stress at the right-end, as
shown in Fig. 4.2.
146 Mechanics of Materials: Formulations and Solutions with Python

Figure 4.2. A uniform bar constrained at the left-end: the middle point is fixed, but
other points on the cross-section is allowed to move freely vertically. The bar with dashed
line is the undeformed bar. It is subjected to a uniformly distributed stress at the right-
end. Physical observation finds that the bar is elongated in the x-direction and shrunk in
the y-direction. Such deformation results in strains in both x- and y-directions.

The particles at any point in the bar are displaced by the applied stress.
Assume the displacement changes linearly in both x- and y-directions. The
length of the bar becomes L and the height of the bar becomes h . Thus,
the right-end of the bar has a displacement of ux in the x-direction, and the
top (or bottom) surface has a displaced uy in the y-direction. The strains in
the bar are defined as
L − L ux
εxx = = (4.6)
L L
which is the rate of length change of the bar in x-direction when its right-end
has a displacement ux in the x-direction. Note that the strain measure here
is in reference to the original length of the bar. Similarly,

h /2 − h/2 uy
εyy = = (4.7)
h/2 h/2
where uy is the displacement in the y-direction, resulting from ux at the right-
end. Strain εyy is the rate of length change of the bar in the y-direction. The
strain measure is in reference to the original half-height of the bar.
Note that strain is dimensionless because the displacement carries a
length dimension.
In a general problem of mechanics of materials, the displacements are
usually not linearly distributed, and have a gradient field. Thus, the general
definition for a strain is defined for a local point in the deformed material as

the rate of change in length of an infinitely small fiber at a point in the

material.
 Displacements and Strains 147

Therefore, the strains will be written in the derivatives of the displace-

ments. Since the tip of the fiber can move in x-, y-, and z-directions, we
shall have normal and shear strains. Also, the fiber can have x-, y-, and
z-orientations; we thus shall have a tensor of second order (two indexes).
These expressions are derived as follows.

4.5.2 Normal strains

Consider in general a small line segment (fiber) in the s direction has length
of Δs. It is elongated in a deformed solid. Assume the two tips of Δs has
a relative displacement Δus also in the s direction. The strain along that
direction is defined as
Δus ∂us
εss = lim = (4.8)
Δs→0 Δs ∂s
In particular, if the fiber is Δx in the x-direction, we shall have the normal
strain as
Δux ∂ux
εxx = lim = (4.9)
Δx→0 Δx ∂x
Similarly, if the fiber is in the y- and z-direction, respectively, we shall
have the normal strains εyy and εzz :

Δuy ∂uy
εyy = lim =
Δy→0 Δy ∂y
(4.10)
Δuz ∂uz
εzz = lim =
Δz→0 Δz ∂z

4.5.3 Shear strains

Consider a small square solid abdc experiencing a small deformation to
ab d c , as shown in Fig. 4.3. The angles αxy and αyx can be calculated
using

∂uy ∂ux
αxy = ; αyx = (4.11)
∂x ∂y

The total angular change becomes

∂uy ∂ux
γxy = αxy + αyx = + (4.12)
∂x ∂y
148 Mechanics of Materials: Formulations and Solutions with Python

Figure 4.3. A small square solid abdc is deformed to ab d c .

where γxy is the so-called engineering shear strain. The shear strain εxy is
defined as

1 1 ∂uy ∂ux
εxy := γxy = + (4.13)
2 2 ∂x ∂y
It is obvious that the strains are symmetric. Similarly, we can obtain
shear strains εyz and εxz .
In summary, all the strains (normal or shear) can be written in a unified
form of

1 ∂ui ∂uj 1
εij = + = (ui,j + uj,i ) (4.14)
2 ∂xj ∂xi 2
Here, we used “,” in the subscript to represent a partial differentiation. As
shown, the small strain is linear in displacement derivatives. The foregoing
equation is also known as the strain–displacement relation for small
strains. It will be used frequently in this book.

4.6 Strains, derivatives of displacement functions

4.6.1 Displacement gradient

The displacement vector in solid mechanics has three components, each of
which is a function of coordinates. It can be defined symbolically in Python
code as follows:
 Displacements and Strains 149

1 x, y, z = symbols("x, y, z") # define symbolic coordinates

2 X = Matrix([x, y, z])
3
4 ux = Function('u_x')(x, y, z) # define symbolic displacement functions
5 uy = Function('u_y')(x, y, z)
6 uz = Function('u_z')(x, y, z)
7
8 U = Matrix([ux, uy, uz]) # displacement vector of symbolic functions
9 U
⎡ ⎤
ux (x, y, z)
⎣uy (x, y, z)⎦
uz (x, y, z)

The displacement U is a vector, and each of its components is a function

of coordinates. When we get the gradient of the displacement vector, we
are getting the gradient of each of the 3 component functions. Since the
gradient of one component function gives a vector, we shall now obtain a
matrix as follows (see Section 2.14), using the outer product of ∇ (vector
of differential operators) and U(x):

⎡ ⎡ ⎤ ⎤
∂
∂x ux (x)
⎢∂⎥ ⎢
∇u(x) := ⎢ ⎥ ⊗ ⎣uy (x)⎥ ⎦
⎣ ∂y ⎦
∂ uz (x)
∂z
∇ U(x)
⎡ ⎤
∂ ∂ ∂
ux (x, y, z) ∂x uy (x, y, z) ∂x uz (x, y, z)
⎢ ∂x ⎥
⎢ ux (x, y, z)
∂ ∂ ∂ ⎥
= ⎣ ∂y ∂y uy (x, y, z) ∂y uz (x, y, z) ⎦ (4.15)
∂ ∂ ∂
∂z ux (x, y, z) ∂z uy (x, y, z) ∂z uz (x, y, z)

It is a matrix of differentials of the displacements, known as the dis-

placement gradient. We can use the following code to obtain it.

1 def grad_vf(vf, X):

2 '''Compute the gradient of a given vector function, vf
3 input: vf, [Link], a column vector of 2 or 3.
4 X, coordinates, array-like of symbols
5 return: grad_f, [Link], a matrix of 2 by 2 or 3 by 3.
6 '''
7 grad_f = [Link]([[Link](vf,xi).T for xi in X])
8 return grad_f
150 Mechanics of Materials: Formulations and Solutions with Python

1 U_g = grad_vf(U, X) # displacement gradient. Or use vector.grad_vec(U)

2 U_g # Or gr.grad_vf(U, X)

⎡ ∂ ∂ ∂ ⎤
∂x ux (x, y, z) ∂x uy (x, y, z) ∂x uz (x, y, z)
⎢ ⎥
⎢∂ ∂ ∂ ⎥
⎢ ∂y ux (x, y, z) ∂y uy (x, y, z) ∂y uz (x, y, z) ⎥
⎣ ⎦
∂ ∂ ∂
∂z ux (x, y, z) ∂z uy (x, y, z) ∂z uz (x, y, z)

4.6.2 Small strains, derived via displacement gradient

The displacement gradient matrix given in Eq. (4.15) is in general asymmet-
ric. It can be used to produce symmetric tensors. There are two approaches
to do this. One is to decompose ∇u(x) into two matrices: one symmetric
and another antisymmetric. The symmetric part is as follows:

1 
∇u(x) + [∇u(x)] (4.16)
2

This leads to a symmetric tensor that will be the small strains in solids.
In indicial notation, it is given in Eq. (4.14).
We write the following simple code to derive this relation.

1 E_l = (U_g + U_g.T)/2 # Or use vector.sym_grad(U)

2 E_l # strain tensor, symmetric

⎡ ∂
u (x,y,z) ∂ ∂ ∂ ⎤
∂ ∂y x u (x,y,z)
∂x y
u (x,y,z)
∂z x
u (x,y,z)
∂x z
⎢ ∂x ux (x, y, z) 2 + 2 2 + 2 ⎥
⎢ ∂ ⎥
⎢ ∂y ux (x,y,z) ∂
u (x,y,z) ∂
u (x,y,z)
∂
u (x,y,z) ⎥
⎢ ∂x y ∂ ∂z y ∂y z ⎥
⎢ 2 + 2 ∂y uy (x, y, z) 2 + 2 ⎥
⎣ ⎦
∂ ∂ ∂ ∂
u (x,y,z) u (x,y,z) u (x,y,z) u (x,y,z)
∂y z ∂
∂z x ∂x z ∂z y
2 + 2 2 + 2 ∂z uz (x, y, z)

This gives the analytic formulas for the linear strain tensor in terms of
the derivatives of the displacement components. It is valid only when strain
is small.
The strain tensor given in the above matrix is of second order and sym-
metric. We now write the strain tensor as symbolic variables.
 Displacements and Strains 151

1 Exx, Exy, Exz = "\u03B5_xx", "\u03B5_xy", "\u03B5_xz"

2 Eyx, Eyy, Eyz = "\u03B5_yx", "\u03B5_yy", "\u03B5_yz"
3 Ezx, Ezy, Ezz = "\u03B5_zx", "\u03B5_zy", "\u03B5_zz"
4
5 E = Matrix([[Exx, Exy, Exz], # E: stress tensor
6 [Eyx, Eyy, Eyz],
7 [Ezx, Ezy, Ezz]])
8 E
⎡ ⎤
εxx εxy εxz
⎣εyx εyy εyz ⎦
εzx εzy εzz

We can write the strain–displacement relations (small strains) in matrix

form as follows [1, 5, 6].

⎡ ⎤ ⎡ ∂uy ⎤
εxx εxy εxz 2 ∂u
∂x
x ∂ux
∂y + ∂x
∂ux
∂z + ∂uz
∂x
⎢ ⎥ 1 ⎢ ∂ux ∂uy ⎥
⎣εyx εyy εyz ⎦ = ⎢ ∂uz ⎥
∂uy ∂uy
+ ∂x 2 + (4.17)
2 ⎣ ∂y ∂y ∂z ∂y ⎦
εzx εzy εzz ∂ux ∂uy
+ ∂uz
∂z ∂x ∂z + ∂y
∂uz
2 ∂u
∂z
z

In indicial notation, it is the same as that given in Eq. (4.14). We obtained

the displacement-strain relations via purely mathematic operations.

4.6.3 Example: Computation of small strains

Following is an example of computing small strains for given functions for
the three displacement components.

1 # Define displacement vector function:

2 #uv=Matrix([.058*x-.025*y+.065*x*y,.016*x+.025*y-.03*x*y,.005*z])#large
3 uv=Matrix([.0058*x-.0025*y+.0065*x*y, .0016*x+.0025*y-.003*x*y,.005*z])
4 uv # the given functions for three displacement components

⎡ ⎤
0.0065xy + 0.0058x − 0.0025y
⎣−0.003xy + 0.0016x + 0.0025y ⎦
0.005z

Compute the strains, using the formula E l obtained earlier and the subs()
function.
152 Mechanics of Materials: Formulations and Solutions with Python

1 e_small = E_l.subs(list(zip(U,uv))).doit().expand()
2 [Link](e_small*1000., 'Strains (e^-3)', n_dgt=4)

Strains (e^-3)
⎡ ⎤
6.5y + 5.8 3.25x − 1.5y − 0.45 0
⎣3.25x − 1.5y − 0.45 2.5 − 3.0x 0⎦
0 0 5.0

Further, we find the strains at a point with coordinates, say at (1., 1., 1.).

1 e_v = e_small.subs({x:1.,y:1,z:1.})
2 [Link](e_v*1000., 'Strains (e^-3)', n_dgt=4)

Strains (e^-3)
⎡ ⎤
12.3 1.3 0
⎣ 1.3 −0.5 0 ⎦
0 0 5.0

Remark: The gradient of the displacement vector at a point in a solid

gives the strain tensor/matrix at that point.

4.6.4 Rotation-displacement relations*

An antisymmetric matrix can be obtained using the displacement gradient
matrix:

1 
∇u(x) − [∇u(x)] (4.18)
2

This will be the rotation in solids. The following Python code creates the
matrix of rotations.

1 eps = (U_g-U_g.T)/2 # compute rotations, anti-symmetric

2 eps

⎡ ∂ ∂ ∂ ∂
⎤
ux (x,y,z) u (x,y,z) ux (x,y,z) u (x,y,z)
∂x y ∂x z
⎢ 0 − ∂y 2 + 2 − ∂z 2 + 2 ⎥
⎢ ∂ ux (x,y,z) ∂ ⎥
⎢ ∂y ∂
u (x,y,z)
∂x y
∂
u (x,y,z)
∂z y
u (x,y,z)
∂y z ⎥
⎢ − 0 − + ⎥
⎣ 2 2
∂
2 2
⎦
∂ ∂ ∂ u (x,y,z)
u (x,y,z)
∂z x
u (x,y,z)
∂x z
u (x,y,z)
∂z y ∂y z
2 − 2 2 − 2 0
 Displacements and Strains 153

We see that the matrix of rotations is anti-symmetric. It is obvious that

the rotation components do not contribute anything to the strains defined
earlier. Therefore, anti-symmetric part of the gradient of the displacement
vector must be excluded in the measure of strains. This will be done in the
formulations for both small and large strains. The matrix of rotations can
also be obtained using the [Link] grad().

4.6.5 Large strains, derived via displacement gradient*

Using the displacement gradient, one can also construct a symmetric matrix
as follows:

1
∇u(x)[∇u] (4.19)
2

This gives a symmetric tensor, the nonlinear part of the large strains.
The formula for the large strain, known as the so-called Green–Lagrange
Strain tensor denoted as εGL , can then be obtained by adding up the linear
and nonlinear strains:

1 
εGL = ∇u(x) + [∇u(x)] + ∇u(x)[∇u] (4.20)
2

In indicial notation, it can be written as


1 ∂ui ∂uj ∂uk ∂uk 1
εij = + + = (ui,j + uj,i + uk,i uk,j ) (4.21)
2 ∂xj ∂xi ∂xi ∂xj 2

Note the contraction over k. This means that the nonlinear part is a sum
of three nonlinear terms.

1 E_GL = (U_g+U_g.T + U_g@U_g.T)/2

2 [Link](E_GL[:,0], ' Formulas for large strains (1st column):')

Formulas for large strains (1st column):

⎡ 
∂uy
2 ⎤
2 2
( ∂u x
∂x ) ∂ux ∂x ( ∂u z
∂x )
⎢ + ∂x + + ⎥
⎢ 2 2 2 ⎥
⎢ ∂ux ∂ux ∂ux ∂uy ∂uy ∂uy ∂uz ∂uz ⎥
⎢ ∂x ∂y ∂y ∂x ∂y ∂x ∂x ∂y ⎥
⎢ 2 + 2 + 2 + 2 + 2 ⎥
⎣ ∂uy ∂uy
⎦
∂ux ∂ux ∂ux ∂uz ∂uz ∂uz
∂x ∂z ∂z ∂x ∂z ∂x ∂z ∂x
2 + 2 + 2 + 2 + 2
154 Mechanics of Materials: Formulations and Solutions with Python

Strains given in Eq. (4.21) depend purely on the derivatives of the dis-
placement components. Thus, it should be some kind of strain measure. Since
it has terms of products of the displacement derivatives, it is clearly non-
linear. Equation (4.21) is a measure of large strains containing both linear
and nonlinear parts. In addition, if the derivatives of the displacements are
small, the higher-order 3rd term in Eq. (4.21) can be omitted. In that case,
Eq. (4.21) is the small strain given in Eq. (4.14).
The nonlinear part can be obtained easily using the following code.

1 E_nl = (U_g@U_g.T)/2
2 [Link](E_nl[:,0],'Formulas for nonlinear strains only (1st column):')

Formulas for nonlinear strains only (1st column):

⎡ ∂ 2 ∂ 2 ∂ 2 ⎤
( ∂x ux (x,y,z)) ( ∂x uy (x,y,z)) ( ∂x uz (x,y,z))
⎢ 2 + 2 + 2 ⎥
⎢ ∂ u (x,y,z) ∂ u (x,y,z) ⎥
⎢ ∂x x ∂y x
∂ ∂
uy (x,y,z) ∂y uy (x,y,z) ∂
∂x z
∂
u (x,y,z) ∂y uz (x,y,z) ⎥
⎢ + ∂x + ⎥
⎣ 2 2 2 ⎦
∂ ∂ ∂ ∂ ∂ ∂
u (x,y,z) ∂z
∂x x
ux (x,y,z) u (x,y,z) ∂z
∂x y
uy (x,y,z) u (x,y,z) ∂z
∂x z
uz (x,y,z)
2 + 2 + 2

The nonlinear part is still too large to print out in whole, and hence we
printed out only the first column. To write it out in whole, we simplify the
expressions, which leads to the same formulas given in Refs. [1, 3]:

⎡ 2  ∂v 2  ∂w 2
∂u ∂u ∂u ∂v ∂v ∂w ∂w
∂x + ∂x + ∂x ∂x ∂y + ∂x ∂y + ∂x ∂y
1⎢
⎢  2  2  2
Enl = ⎢ ∂u ∂u
+ ∂v ∂v
+ ∂w ∂w ∂u ∂v
+ ∂y + ∂w
2⎣ ∂x ∂y ∂x ∂y ∂x ∂y ∂y ∂y
∂u ∂u ∂v ∂v ∂w ∂w ∂u ∂u ∂v ∂v ∂w ∂w
∂x ∂z + ∂x ∂z + ∂x ∂z ∂y ∂z + ∂y ∂z + ∂y ∂z
(4.22)
∂u ∂u ∂v ∂v ∂w ∂w ⎤
∂x ∂z + ∂x ∂z + ∂x ∂z
∂u ∂u ∂v ∂v
⎥
∂w ∂w ⎥
∂y ∂z + ∂y ∂z + ∂y ∂z ⎦
 ∂u 2  2  ∂w 2
∂z + ∂v
∂z + ∂z

Note that the nonlinear strains given in Eq. (4.21) is an exact expression,
implying that it is derived without any assumption or approximation. It is
not obtained from a Taylor series expansion by cutting off the third-order
terms and beyond.

Remark: The strains at a point in a solid are determined by the displace-

ments derivatives at that point.
 Displacements and Strains 155

4.6.6 Example: Computation of large strains*

Following is an example for computing large strains for displacement func-
tions given below, simply using subs().

1 # Define displacement vector function:

2 #df=Matrix([.058*x-.025*y+.065*x*y,.016*x+.025*y-.03*x*y,.005*z])#large
3 df=Matrix([.0058*x-.0025*y+.0065*x*y, .0016*x+.0025*y-.003*x*y,.005*z])
4 df
⎡ ⎤
0.0065xy + 0.0058x − 0.0025y
⎣−0.003xy + 0.0016x + 0.0025y ⎦
0.005z

1 egl = E_GL.subs(list(zip(U,df))).doit().expand()
2 [Link](egl*1000., 'Green-Lagrange Strains (e^-3)', n_dgt=4)

Green-Lagrange Strains (e^-3)

⎡ ⎤
0.02562xy + 3.266x
⎢ 0.02562y 2 + 6.533y + 5.818 0 ⎥
⎢ − 1.512y − 0.4552 ⎥
⎢ 0.02562xy + 3.266x ⎥
⎢ 0.02562x − 3.024x + 2.506
2 0 ⎥
⎣ − 1.512y − 0.4552 ⎦
0 0 5.013
Further, we find the solution at a given coordinate, say at (1., 1., 1.).

1 egln = [Link]({x:1.,y:1,z:1.})
2 [Link](egln*1000., 'Values of the GL Strains (e^-3)', n_dgt=4)

Values of the GL Strains (e^-3)

⎡ ⎤
12.38 1.325 0
⎣1.325 −0.4919 0 ⎦
0 0 5.013

It is found that the Green–Lagrange strain (large strain) is very close

to the small strain obtained in the previous section. Readers may make the
displacements larger (by 10 times for example), and compute both small and
large strains again to see the differences.

4.6.7 Large strains derived via conventional

deformation gradient**
For problems with large displacements and hence large strains, the strain–
displacement relation will be nonlinear [1, 6]. Traditionally, one often derive
156 Mechanics of Materials: Formulations and Solutions with Python

large strains using the deformation gradient. Let us do so using symbolic

computation.
Assume a solid is initially at state 0 and is described using coordinates
x, y, and z. We define the initial coordinate vector x0 as follows:

1 x, y, z = symbols("x, y, z") # define symbolic coordinates

2 X = Matrix([x, y, z])
3 x0 = Matrix([x, y, z]) # define initial coordinates (symbolic)
4 x0

⎡ ⎤
x
⎣y ⎦
z

Vector x0 can be treated as a vector of (simple) functions. The gradient

of x0 with respect to x, y, and z becomes an identity matrix as expected:

1 x0_g = gr.grad_vf(x0, X) # gradient of the initial coordinates

2 x0_g

⎡ ⎤
1 0 0
⎣0 1 0⎦
0 0 1

Assume now the solid is stressed and deformed heavily to a new state t.
At any point in the solid, the displacements are denoted as vector U
with components not necessarily small. The coordinates of the point now
become xt .

1 x_t = Function('x_t')(x,y,z) # define symbolic functions

2 y_t = Function('y_t')(x,y,z)
3 z_t = Function('z_t')(x,y,z)

1 xt = Matrix([x_t, y_t, z_t]) # form a vector of symbolic functions

2 xt

⎡ ⎤
xt (x, y, z)
⎣ yt (x, y, z) ⎦
zt (x, y, z)
 Displacements and Strains 157

[Link] Deformation gradient, definition

Note that xt is the vector of functions of x, y, and z. The deformation
gradient, often denoted as F, can be defined as the gradient of xt , that is a
vector function, with respect to x, y, and z. It becomes a matrix as follows:
⎡ t t t
⎤
∂x1 ∂x1 ∂x1
0 ∂x02 ∂x03 ⎥
  ⎢ ∂x1
t
∂xi ⎢ ⎥
⎢ ∂xt ∂xt2 ∂xt2 ⎥
F= = ⎢ ∂x20 ∂x02 ∂x03 ⎥
(4.23)
∂xj
0 ⎢ 1 ⎥
⎣ t ⎦
∂x3 ∂xt3 ∂xt3
∂x01 ∂x02 ∂x03

The code to compute it is given as follows:

1 F = vector.grad_vec(xt).T
2 F

⎡ ∂ ∂ ∂ ⎤
∂x xt (x, y, z) ∂y xt (x, y, z) ∂z xt (x, y, z)
⎢∂ ⎥
⎢ yt (x, y, z) ∂ ∂ ⎥
⎣ ∂x ∂y yt (x, y, z) ∂z yt (x, y, z) ⎦
∂ ∂ ∂
∂x zt (x, y, z) ∂y zt (x, y, z) ∂z zt (x, y, z)

Note the deformation gradient matrix is in general asymmetric. Since it

contains rigid body movements and that may result in false strains. Thus,
we need to go through the following process.

[Link] Deformation gradient, in displacements

First, we write the deformation gradient in terms of the displacement gradi-
ent. Because the new coordinates xt is the result of the displacement U, we
shall have (assume the material is continuous):
xt = x0 + U (4.24)
Let us use a code to do this.
1 xt = x0 + U
2 F = gr.grad_vf(xt, X).T # Or use vector.grad_vec(xt).T
3 [Link](F, 'F in displacement gradient:', n_dgt=1)

F in displacement gradient:
⎡ ∂ ∂ ∂ ⎤
∂x ux (x, y, z) + 1 ∂y ux (x, y, z) ∂z ux (x, y, z)
⎢ ∂ ⎥
⎢ ∂ ∂ ⎥
⎣ ∂x uy (x, y, z) ∂y uy (x, y, z) + 1 ∂z uy (x, y, z) ⎦
∂ ∂ ∂
∂x uz (x, y, z) ∂y uz (x, y, z) ∂z uz (x, y, z) + 1
158 Mechanics of Materials: Formulations and Solutions with Python

As seen, the deformation gradient is the transpose of the displacement

gradient (see Eq. (4.15)) plus an identity matrix. These 1s on the diagonal
are the rigid body movements, and hence have nothing to do with the dis-
placements derivatives, and hence needs to be removed in a proper manner.

[Link] Cauchy–Green deformation tensor

Using the deformation gradient, we can form the so-called Cauchy–Green

deformation tensor.

FCG = F F (4.25)

It is clear that FCG is a symmetric tensor of second order. We can use

the following code to form it.

1 F_CG = F.T@F # the Cauchy-Green deformation tensor

2 F_CG[:,0] # the 1st column (too large to print out in whole)

⎡
∂ux
2  x 2  ∂ux 2 ⎤
∂x + 1 + ∂u ∂y + ∂z
⎢   ⎥
⎢  ∂ux  x ∂uy ⎥
⎢ + 1 ∂u + + 1 ∂ux ∂ux ∂uz ⎥
⎣ ∂x ∂y ∂y ∂y + ∂z ∂y ⎦
 ∂ux  x  ∂uz  ∂ux ∂ux ∂uy
∂x + 1 ∂u
∂z + ∂z + 1 ∂z + ∂y ∂z

1 #F_CG[:,1] # the 2nd column

2 #F_CG[:,2] # the 3rd column

Note that the 1s are still there.

[Link] Green–Lagrange strain tensor

Using the Cauchy–Green deformation tensor, we can form the so-called

Green–Lagrange strain tensor, denoted as εGL .

1
εGL = (FCG − I) (4.26)
2

The subtraction of identity matrix I removes these 1s. We can use the
following code to form εGL .
 Displacements and Strains 159

1 E_GL=[Link]*(F_CG-[Link](3)).expand() # E_GL: Green—Lagrange Strain

2 E_GL[:,0] #E_GL[:,1] #E_GL[:,2]

⎡ 2

∂uy
2
2
⎤
( ∂u x
∂x ) ∂ux ∂x ( ∂u z
∂x )
⎢ + ∂x + + ⎥
⎢ 2 2 2
⎥
⎢ ∂ux ∂ux ∂ux ∂uy ∂uy ∂uy ∂uz ∂uz ⎥
⎢ ∂x ∂y ∂y ∂x ∂y ∂x ∂x ∂y ⎥
⎢ + + + + ⎥
⎣ 2 2 2
∂uy ∂uy
2 2
⎦
∂ux ∂ux ∂ux ∂uz ∂uz ∂uz
∂x ∂z ∂z ∂x ∂z ∂x ∂z ∂x
2 + 2 + 2 + 2 + 2

In the code above, we used [Link] that is a singleton. It is the Rational

(1, 2) in Sympy. Green–Lagrange strain tensor is for large strains. It contains
quadratic terms and hence it is nonlinear in derivatives of the displace-
ments.
Note that εGL given in Eq. (4.26) is the same as that in Eq. (4.20)
obtained earlier via a much simpler procedure, via an argument of sym-
metry.
Recall that we have obtained the formula for linear strain tensor earlier,
which is given in Eq. (4.17). We can subtract the linear strain from the
Green–Lagrange strain and obtain the nonlinear part of the strain, which will
be exactly the same as the one obtained earlier, via also a simple symmetry
argument.

4.7 Strain–displacement relation, a general derivation

Similar to the stress tensor, the strain tensor has both indicial and matrix
representations. Since the strain tensor is symmetric, the matrix is a sym-
metric matrix. It is also similar to the stress tensor, in terms of formulation
and operations.

• In space R3 , it is a 3 × 3 matrix, and hence has, in general, a total of 9

components. Due to the symmetry, we have only 6 components.
• In space R2 , it is a 2 × 2 matrix, and hence has, in general, a total of 4
components. Due to the symmetry, we have only 3 components.
• In space R1 , it is a scalar, and hence there is only 1 component.

We present now a general approach to derive the strain–displacement

relation, considering large deformation in 3D solids. This derivation shall
lead to the strain tensor transformation rule.
160 Mechanics of Materials: Formulations and Solutions with Python

4.7.1 Normal strain along a fiber, magnification

factor*
Consider a solid occupying domain Ω in the Cartesian coordinate system
with mutual orthogonal axes (x1 , x2 , x3 ). It undergoes a large deformation
and its domain becomes a deformed one, Ω∗ , as shown in Fig. 4.4.
An infinitely small fiber in the undeformed solid is denoted as P-Q. The
coordinates of the two ends of the fiber are P (x1 , x2 , x3 ) and Q(x1 +dx1 , x2 +
dx2 , x3 + dx3 ). After the deformation the fiber becomes P*-Q*, and its coor-
dinates become P ∗ (x∗1 , x∗2 , x∗3 ) and Q∗ (x∗1 + dx∗1 , x∗2 + dx∗2 , x∗3 + dx∗3 ). The
displacements at point P is denoted as (u1 , u2 , u3 ).

[Link] Evaluation of the length of a fiber in a deformed solid

Assume the solid is a continuum media and also the deformation is continu-
ous, implying that each of the coordinates (x∗1 , x∗2 , x∗3 ) is differentiable with
respect to x1 , x2 , and x3 . Hence, we have
∂x∗i ∂x∗ ∂x∗ ∂x∗i
dx∗i = dx1 + i dx2 + i dx3 = dxj = x∗i,j dxj (4.27)
∂x1 ∂x2 ∂x3 ∂xj
where we used again the summation convention for repeated indexes in a
term, and the subscript “,” stands for partial differentiation.
Since the deformation is assumed continuous, the coordinates at a point
in the deformed solid should be the sum of the coordinates of the undeformed

Figure 4.4. A solid continuum media undergoes large deformation.

 Displacements and Strains 161

solid plus the displacements at the same point:

x∗i = xi + ui , i = 1, 2, 3 (4.28)
In addition, each of the displacement components will also be differen-
tiable, which gives
x∗i,j = xi,j + ui,j (4.29)
Since the solid is a continuum media, the length of the infinitely small
fiber can be measured using
(ds)2 = (dx1 )2 + (dx2 )2 + (dx3 )2 = dxi dxi (4.30)
The length of the infinitely small fiber in the deformed solid can also be
obtained using
(ds∗ )2 = (dx∗1 )2 + (dx∗2 )2 + (dx∗3 )2 = dx∗i dx∗i (4.31)

[Link] Engineering strain of a given fiber

Assume now the infinitely small fiber has an orientation in the Cartesian
coordinate system defined as n = [n1 , n2 , n3 ] , where ni is the direction
cosine of the fiber: ni = cos(n, xi ). The engineering strain of the fiber is a
measure with respect to the original length of the fiber, and is expressed as,
ds∗ − ds ds∗
εE = = −1 (4.32)
ds ds
which gives
ds∗
= εE + 1 (4.33)
ds

[Link] Magnification factor of a fiber, large normal strain

We now define the so-called magnification factor as
   
1 (ds∗ )2 − (ds)2 1 ds∗ 2
M := = −1 (4.34)
2 (ds)2 2 ds
Readers may take a note of the difference between the expressions of the
engineering strain and the magnification factor, which can be expressed as
1
M = εE + ε2E (4.35)
2
The magnification factor M is known as the Green–Lagrange normal
strain for a fiber in a solid undergoing large deformation. We see the non-
linear (squared) term, due to the large strain. When εE is small, we have
M ≈ εE .
162 Mechanics of Materials: Formulations and Solutions with Python

[Link] Magnification factor and the strain components

Now, we are ready to derive the relationship between the magnification factor
of the given fiber P-Q and the strain components at point P. From Eqs. (4.27)
and (4.29), we have

dx∗i = x∗i,j dxj = (xi,j + ui,j ) dxj (4.36)

x∗i,j

Note that
∂xi
xi,j = = δij (4.37)
∂xj

Equation (4.36) becomes

dx∗i = x∗i,j dxj = (δij + ui,j )dxj (4.38)

Let us evaluate the difference of the squared length of the fiber before
and after deformation. Using Eqs. (4.30) and (4.31) we shall have
 ∗ 2 
(ds ) − (ds)2 = dx∗i dx∗i − dxi dxi
⎛ ⎞
⎜ ⎟
= ⎝(δij + ui,j ) (δil + ui,l )⎠ dxj dxl − dxi dxi
dx∗i dx∗i
⎛ ⎞
⎜ ⎟
= ⎝δij δil + δij ui,l + δil ui,j +ui,j ui,l ⎠ dxj dxl − dxi dxi
δjl uj,l ul,j

= (δjl + uj,l + ul,j + ui,j ui,l ) dxj dxl − dxi dxi

= δjl dxj dxl + (uj ,l + ul ,j + ui,j ui,l )dxj dxl − dxi dxi
dxj dxj 2εjl

= 2εjl dxj dxl = 2εij dxi dxj (4.39)

In the last line in the above derivation, we used Eq. (4.21).

Finally, using Eq. (4.34), we obtain:
 
1 (ds∗ )2 − (ds)2 1 ∗ 2  1 1
M= = (ds ) − (ds)2
2 (ds) 2 2 ds ds
dxi dxj
= εij = ni εij nj (4.40)
ds ds
 Displacements and Strains 163

In the foregoing equation, we used

dxi
ni = (4.41)
ds
Magnification factor M and the strain components εij are related as

M = ni εij nj (4.42)

where ni is the direction cosine of the fiber: ni = cos(n, xi ), before deforma-

tion. The explicit formula for M can be obtained using the following code.

1 11, 22, 33, 23, 13, 12 = symbols(" 11, 22, 33, 23, 13, 12")
2 n1, n2, n3, nn = symbols("n1, n2, n3, _nn")
3
4 ε = Matrix([[ 11, 12, 13], # ε: stress t nsor
5 [ 12, 22, 23],
6 [ 13, 23, 33]])
7
8 n = Matrix([n1, n2, n3])
9 M = (n.T@ε@n)[0].simplify()
10 M # nn

n1 (n1 ε11 + n2 ε12 + n3 ε13 ) + n2 (n1 ε12 + n2 ε22 + n3 ε23 )

+ n3 (n1 ε13 + n2 ε23 + n3 ε33 )
Comparing to Eq. (3.42), it is clear that M is the normal strain of the
fiber with direction cosine n, εnn .

[Link] Direction cosine of a fiber after deformation*

Let us now examine the direction cosine of the fiber after deformation. We
start with the definition of the direction cosine for the deformed fiber, which
gives

dx∗i
n∗i = (4.43)
ds∗
which can be written as
dx∗i ds
n∗i = (4.44)
ds ds∗
Using Eq. (4.38), we have

dx∗i dxj
= x∗i,j = (δij + ui,j )nj (4.45)
ds ds
164 Mechanics of Materials: Formulations and Solutions with Python

From Eq. (4.33), we have

ds∗ = (1 + εE ) ds (4.46)
length change rate

where (εE + 1) accounts for the length change rate in a fiber. Using
Eqs. (4.44), (4.46), and (4.45), we have

(1 + εE )n∗i = (δij + ui,j )nj (4.47)

4.7.2 Shear strains caused by large deformation*

Consider two orthogonal infinitely small fibers in a solid undergoes large
deformation, as shown in Fig. 4.5. Fiber P-A has direction cosines of
n = (n1 , n2 , n3 ), and Fiber P-B has m = (m1 , m2 , m3 ). These two fibers are
deformed, respectively, to P*-A* and P*-B*. The direction cosines become,
respectively, n∗ = (n∗1 , n∗2 , n∗3 ) and m∗ = (m∗1 , m∗2 , m∗3 ). Our task now is to
evaluate the shear strain between these two fibers.
Since P-A and P-B are orthogonal, we shall have initially

n i mi = 0 (4.48)

Figure 4.5. Two orthogonal fibers in a solid continuum media undergoes large defor-
mation. The angular change of these two fibers gives the shear strains.
 Displacements and Strains 165

After deformation, P*-A* and P*-B* are not necessarily orthogonal, and
their angle changed from π/2 can be evaluated using

cos θ ∗ = n∗i m∗i (4.49)

The shear strain between fiber n and m is then defined as

1
εnm = (1 + εEn ) (1 + εEm ) cos θ ∗
2
length change rate of n length change rate of m

1
= (1 + εEn )(1 + εEm )n∗i m∗i
2
1
= (1 + εEn )n∗i (1 + εEm )m∗i (4.50)
2
Using Eqs. (4.50), and (4.47), we have
1
εnm = (δik + ui,k )nk (δil + ui,l )ml
2
1
= (δik δil + δik ui,l + ui,k δil + ui,k δil ) nk ml
2
1
= (δkl + uk,l + ul,k + ui,k ui,l ) nk ml
2
1 1
= δkl nk ml + (uk,l + ul,k + ui,k ui,l ) nk ml
2 2
nk mk =0
εkl

= nk εkl ml (4.51)

where εnm is the shear strain between two fibers with direction cosines n and
m, which are initially orthogonal. The resulting strain expression εkl in the
foregoing equation is the same as that obtained in Eq. (4.21), using gradient
of displacement vector functions.
It may be worth to note that when m = n, εnn becomes the magnifi-
cation factor obtained in Eq. (4.42). This is because the length change of
the infinitely small fibers are taken into consideration in Eq. (4.50), while
considering the angular change. Therefore, Eq. (4.51) is more general than
Eq. (4.42). It gives both normal and shear strains.

4.7.3 Example: Symbolic computation of strains

Suppose we know the strain tensor ε (in 9 components) at a point in
solid. We want to know the strain along an arbitrary fiber with direction
of n = (l, m, n), which are the direction cosines of the fiber with respect to
166 Mechanics of Materials: Formulations and Solutions with Python

the coordinates x1 , x2 , x3 . In indicial notation, we denote n = (n1 , n2 , n3 ).

The normal strain of this fiber is the magnification factor given in Eq. (4.42),
and can be computed using

εnn = ni εij nj (4.52)

In matrix notation, we have

εnn = nεn (4.53)

Consider now another arbitrary fiber orthogonal to n with direction

cosines denoted as m = (m1 , m2 , m3 ). The shear strain between fibers n
and m is then computed using Eq. (4.51):

εnm = ni εij mj (4.54)

In matrix notation, we have

εnm = nεm (4.55)

Let us write the following Numpy codes to show this. We first give the
symbolic computation code.

1 #symbolic computation code

2 n1, n2, n3 = "n1", "n2", "n3"
3 n = Matrix([[n1, n2, n3]]) #Direction cosines for fiber n
4 n
 
n1 n2 n3

1 nn = n@ε@n.T # _nn: normal strain of fiber n

2 nn # the magnification factor M

[n1 (n1 ε11 + n2 ε12 + n3 ε13 ) + n2 (n1 ε12 + n2 ε22 + n3 ε23 )

+ n3 (n1 ε13 + n2 ε23 + n3 ε33 )]

1 [Link]({n3:0}) # Formula of normal strain for n3 = 0 (2D)

 
n1 (n1 ε11 + n2 ε12 ) + n2 (n1 ε12 + n2 ε22 )

1 m1, m2, m3 = "m1", "m2", "m3"

2 m = Matrix([[m1, m2, m3]]) # Direction cosines for fiber m
3 m
 
m1 m2 m3
 Displacements and Strains 167

1 nm = n@ε@m.T # Shear strain of between fiber n and m

2 nm

[m1 (n1 ε11 + n2 ε12 + n3 ε13 ) + m2 (n1 ε12 + n2 ε22 + n3 ε23 )

+ m3 (n1 ε13 + n2 ε23 + n3 ε33 )]

1 [Link]({n3:0, m3:0}) # Formula of shear strain for n3 = m3 = 0 (2D)

 
m1 (n1 ε11 + n2 ε12 ) + m2 (n1 ε12 + n2 ε22 )

4.7.4 Example: Computation of displacements, normal

and shear strains
Consider a 3D solid brick shown in Fig. 4.6.
The displacements are given in the following formulas [1].
u = c1 xyz, v = c2 xyz, w = c3 xyz (4.56)
where c1 , c2 and c3 are unknown constants. The displacements at point E
are found by a measurement as (0.004, 0.002, −0.004) (m).

1. Determine the functions for all the displacement components.

2. Compute the gradient of the displacement vector functions.
3. Compute the strain functions in the solid, and the values of the strains
at point E.

Figure 4.6. A 3D solid brick.

168 Mechanics of Materials: Formulations and Solutions with Python

−→
4. Compute the normal strains at point E along the EA direction, and along
−→
EA direction.
−→ −−→
5. Compute the shear strains at point E between EA and EF .

Solution: The formulas for the displacement components contain unknown

constants c1 , c2 , c3 . Therefore, we need to determine these constants. This
can be done using the conditions given for the displacements at point E,
which gives the following three algebraic equations:

c1 xE yE zE = uE
c2 xE yE zE = vE (4.57)
c3 xE yE zE = wE

where xE , yE , zE are the coordinates at point E, and uE , vE , wE are the dis-

placements at point E. These three equations can be solved easily one-by-one
for c1 , c2 , c3 because these are not coupled. In general, this type equation may
be coupled. Thus, we use Sympy solver to obtain the solution, assuming these
equations are a coupled system.
After the displacements are determined, we can answer all the remaining
questions with ease. We write a code to complete the task.

1 # Ref: Advanced Mechanics of Materials, by Boresi and Schmidt

2 # Example 2.8 Three-Dimensional Strain State
3
4 # 1) Determine the functions for all the displacement components.
5 x, y, z = symbols("x, y, z") # define symbolic coordinates
6 X = Matrix([x, y, z])
7 c1, c2, c3 = symbols("c_1, c_2, c_3")
8
9 u = c1*x*y*z
10 v = c2*x*y*z
11 w = c3*x*y*z
12
13 xE = {x:3/2, y:1, z:2} # Coordinates at point E
14 dE = [4, 2, -4] # × 1e-3 # displacement at point E
15 sln_cs=[Link]([[Link](xE)-dE[0],[Link](xE)-dE[1],[Link](xE)-dE[2]],
16 [c1, c2, c3])
17 sln_cs # should be: { 1:0.004/3, 2:0.002/3, 3:−0.004/3}

{c1 : 1.33333333333333, c2 : 0.666666666666667, c3 : −1.33333333333333}

 Displacements and Strains 169

1 # substuting the constants back to the formulas for the displacements:

2 U = Matrix([[Link](sln_cs), [Link](sln_cs), [Link](sln_cs)])
3 [Link](U.T, 'The displacement vector functions are found as:')

The displacement vector functions are found as

 
1.333xyz 0.6667xyz −1.333xyz

The displacement functions are found as

⎡ ⎤ ⎡ 0.004 ⎤
u 3 xyz
⎢ ⎥ ⎢ 0.002 ⎥
⎢v⎥ = ⎢ ⎥
⎣ ⎦ ⎣ 3 xyz ⎦ (4.58)
w 3 xyz
− 0.004

1 # 2) Compute the gradient of the displacement vector functions.

2 np.set_printoptions(precision=4,suppress=True)
3
4 # the gradient of the displacement vector functions
5 U_g = gr.grad_vf(U, X) # Or use: vector.grad_vec(U)
6 printM(U_g, 'Displacement gradient (matrix):', n_dgt=4)

Displacement gradient (matrix):

⎡ ⎤
1.333yz 0.6667yz −1.333yz
⎣1.333xz 0.6667xz −1.333xz ⎦
1.333xy 0.6667xy −1.333xy

1 # 3) Compute the strain functions in the solid

2 strains = 0.5*(U_g + U_g.T) # symmetric part of displacement gradient
3 printM(strains,'Strain tensor (small) functions', n_dgt=4)

Strain tensor (small) functions

⎡ ⎤
1.333yz 0.6667xz + 0.3333yz 0.6667xy − 0.6667yz
⎢ ⎥
⎣0.6667xz + 0.3333yz 0.6667xz 0.3333xy − 0.6667xz ⎦
0.6667xy − 0.6667yz 0.3333xy − 0.6667xz −1.333xy
170 Mechanics of Materials: Formulations and Solutions with Python

1 # 3) the values of the strains at point E.

2
3 _val = [Link](xE) # values of the strains at point E
4 printM( _val, 'Strain tensor values (e-3):', n_dgt=4)

Strain tensor values (e-3):

⎡ ⎤
2.667 2.667 −0.3333
⎢ ⎥
⎣ 2.667 2.0 −1.5 ⎦
−0.3333 −1.5 −2.0

1 # 4) Compute the normal strains at point E along the fiber EA direction:

2
3 _val = [Link]( _val)
4
5 # compute the direction cosine for EA using Eq.(2.28):
6 # E(1.5, 1.0, 2.0); A(1.5, 0.0, 0.0);
7 length_EA = [Link]((1.5-1.5)**2+(0.0-1.0)**2+(0.0-2.0)**2)
8 n_EA = [Link]([(1.5-1.5),(0.0-1.0),(0.0-2.0)])/length_EA # ni, E->A
9 print(f'The fiber direction N = {n_EA}')
10
11 nn_EA = n_EA@ _val@n_EA # Normal strain of fiber E->A
12 print(f'Normal strain on fiber N = { nn_EA:.4f}e-3')
13
14 # normal strain at point E along EF direction:
15 n_EF = [Link]([-1., 0, 0]) # fiber along E->F
16 print(f'The fiber direction M = {n_EF}')
17
18 nn_EF = n_EF@ _val@n_EF #Normal strain of fiber E->F
19 print(f'Normal strain on fiber M = { nn_EF:.4f}e-3')

The fiber direction N = [0 -0.447213595499958 -0.894427190999916]

Normal strain on fiber N = -2.4000e-3
The fiber direction M = [-1. 0. 0.]
Normal strain on fiber M = 2.6667e-3

1 #5) Compute the shear strains at point E between fibers EA and EF.
2
3 _NM = n_EA@ _val@n_EF #Shear strain between fiber E->A and E->F
4 print(f'Shear strain between two fibers = { _NM:.4f}e-3')
5 print(f'Engineering shear strain between two fibers = {2.* _NM:.4f}e-3')

Shear strain between two fibers = 0.8944e-3

Engineering shear strain between two fibers = 1.7889e-3
 Displacements and Strains 171

4.8 Coordinate transformation of strain tensor

Consider a coordinate transformation from base coordinate of x1 , x2 , and x3

to a new coordinate x1 , x2 , and x3 . We can now readily derive the rule for
coordinate transformation of strain tensor. It is a generalization of Eq. (4.54).
After the coordinate transformation from the coordinate system of x1 , x2 ,
and x3 to a new coordinate x1 , x2 , and x3 , the ni should be replaced by ari
(with r range from 1 to 3). This is because the fiber direction could be one of
the three axes x1 , x2 , and x3 . The mj should be replaced by asj (with s range
from 1 to 3) because the fiber-tip-moving direction could also be one of the
three directions x1 , x2 , and x3 . The fiber directions can be indexed by r, and
the tip-moving directions can be indexed by s. All the contraction operations
stay unchanged, leaving with 2 indexes, resulting in a new description of the
same strain tensor. After this indexation matching, we re-write Eq. (4.54)
to include the precise meaning of each of the indexes.

εnm = ni εij mj
ε n m = ni ε i j mj
fiber-dir tip-move-dir controls fiber-dir contract contract with controls
denoted by r denoted by s with fiber-dir tip-move-dir tip-move-dir

⎡ r ⎤ ⎡ s ⎤ ⎡i⎤ ⎡j⎤
x1 x1 x1 x1
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ x2 ⎦ ⎣ x2 ⎦ ari ⎣ x2 ⎦ ⎣ x2 ⎦ asj
x3 x3 x3 x3
(4.59)

⎡ ⎤
cos(x1 , x1 ) cos(x1 , x2 ) cos(x1 , x3 )
⎢ ⎥
⎢ a11 a12 a13 ⎥
⎢    ⎥
⎢cos(x2 , x1 ) cos(x2 , x2 ) cos(x2 , x3 )⎥
⎢
T=⎢ ⎥ (4.60)
⎥
⎢ a21 a22 a23 ⎥
⎢ ⎥
⎣cos(x3 , x1 ) cos(x3 , x2 ) cos(x3 , x3 )⎦
a31 a32 a33

We obtain:

εrs = ari εij asj (4.61)

172 Mechanics of Materials: Formulations and Solutions with Python

This is the formula for strain tensor transformation using indicial nota-
tion. Note the indexes i and j are all contracted, leaving with the free indexes
r and s, resulting in a second-order tensor, the transformed strain tensor.
Alternatively, we may use the matrix notation and matrix multiplication
rule, and the formula should be:

ε = TεT (4.62)

The T on the left ensures the contraction is on the first index of ε (with
the 2nd index of T), and T on the right ensures the contraction is on the
2nd index of ε and the transpose is needed so that the contraction is always
with the 2nd index of T, related to the base coordinate axes.
Since T is unitary, the inverse transformation can be simply done using
the transpose of T:

ε = T  ε T (4.63)

Or in the indicial notation:

εrs = air εij ajs (4.64)

All we need is to swap the order of the two indexes of aij .

It is seen again that transformation of a second order strain tensor
requires two times of operation by the rotation matrix T.
It is clear now that the coordinate transformation rule of strain tensor
is exactly the same as that for the stress tensor discussed in the previous
chapter. After all, these are all second-order tensors. Thus, the same codes
given in the previous chapter can be used. In addition, the Mohr circle
presentation applies also to strain tensors in exactly the same manner.
Let us write the following Python codes to perform coordinate transfor-
mation for strains, using both matrix notation and indicial notation.

1 T = [Link]() # form a symbolic transformation matrix

2 T # print out
⎡ ⎤
a11 a12 a13
⎢ ⎥
⎣a21 a22 a23 ⎦
a31 a32 a33

1 εp = T@ε@T.T # ε: Strain tensor defined earlier

2 εp[:,0] # print out the first column
 Displacements and Strains 173

⎡ ⎤
a11 (a11 ε11 + a12 ε12 + a13 ε13 ) + a12 (a11 ε12 + a12 ε22 + a13 ε23 )
⎢ + a (a ε + a ε + a ε ) ⎥
⎢ 13 11 13 12 23 13 33 ⎥
⎢ ⎥
⎢ a11 (a21 ε11 + a22 ε12 + a23 ε13 ) + a12 (a21 ε12 + a22 ε22 + a23 ε23 ) ⎥
⎢ ⎥
⎢ + a (a ε + a ε + a ε ) ⎥
⎢ 13 21 13 22 23 23 33 ⎥
⎢ ⎥
⎣ a11 (a31 ε11 + a32 ε12 + a33 ε13 ) + a12 (a31 ε12 + a32 ε22 + a33 ε23 ) ⎦
+ a13 (a31 ε13 + a32 ε23 + a33 ε33 )

The computation can be done using the same function Tensor2 transfer()
defined earlier.

1 εp_ = gr.Tensor2_transfer(T,ε) # use the same code function for stress

2 εp_ = Matrix(εp_)
3
4 print(f"Transformed strain via tensor operation (first column only):")
5 εp_[:,0] # print first column only; εp_[:,1] # εp_[:,12]

Transformed strain via tensor operation (first column only):

⎡ ⎤
a11 (a11 ε11 + a12 ε12 + a13 ε13 ) + a12 (a11 ε12 + a12 ε22 + a13 ε23 )
⎢ + a (a ε + a ε + a ε ) ⎥
⎢ 13 11 13 12 23 13 33 ⎥
⎢ ⎥
⎢ a11 (a21 ε11 + a22 ε12 + a23 ε13 ) + a12 (a21 ε12 + a22 ε22 + a23 ε23 ) ⎥
⎢ ⎥
⎢ + a (a ε + a ε + a ε ) ⎥
⎢ 13 21 13 22 23 23 33 ⎥
⎢ ⎥
⎣ a11 (a31 ε11 + a32 ε12 + a33 ε13 ) + a12 (a31 ε12 + a32 ε22 + a33 ε23 ) ⎦
+ a13 (a31 ε13 + a32 ε23 + a33 ε33 )

4.9 Principal strains

Similar to stress states, for a given strain state, one can find a set of prin-
cipal strains, by rotating the coordinates to a particular angle, where all
the shear strain components vanish. Obtaining such a coordinate direction
and the principal strains can also be casted to an eigenvalue problem, which
can be written in the exact form by replacing the stress tensor with strain
tensor.

[ε − εI]n = 0 (4.65)

All the formulations and theory for the strain tensor is the same as that
for the stress tensor. Thus, the same codes given in the previous chapter can
be used.
174 Mechanics of Materials: Formulations and Solutions with Python

4.10 Strain invariants

Similar to stress tensor, the strain tensor has also nine strain components.
Thus, it can also form three strain invariants that do not change with the
coordinate change. The invariant property is essentially the feature of any
3 × 3 matrix, including the stress and strain tensor matrix.
These three strain invariants are defined as follows, based on nine strain
components or these three principal strains.
First, the trace of the strain tensor gives the first strain invariant:
I¯1 = ε11 + ε22 + ε33 = ε1 + ε2 + ε3 (4.66)
Second, the sum of the three minor determinants of the strain tensor
gives the second strain invariant:
ε11 ε12 ε11 ε13 ε22 ε23
I¯2 = + +
ε12 ε22 ε13 ε33 ε23 ε33
(4.67)
= ε11 ε22 + ε11 ε33 + ε22 ε33 − ε212 − ε213 − ε223
= ε1 ε2 + ε1 ε3 + ε2 ε3
Third, the determinant of the strain tensor gives the third strain invariant:
ε11 ε12 ε13
I3 = ε12 ε22 ε23 = ε1 ε2 ε3
¯ (4.68)
ε12 ε23 ε33
For 2D cases, say in the x–y plane, we shall have two strain invariants:
I¯1 = ε11 + ε22 = ε1 + ε2 (4.69)
and
ε11 ε12
I¯2 = = ε11 ε22 − ε212 = ε1 ε2 (4.70)
ε12 ε22
Let us look at a specific example.

4.11 Example: Strain components after coordinate

rotation

Consider the same matrix used earlier.

1 # Define a strain tensor:
2 v=Matrix([[2.667,2.667,-0.3333],[2.667,2.0,-1.5],[-0.3333,-1.5,-2.0]])
3 printM( v, 'Strain tensor values (e-3):', n_dgt=4)
 Displacements and Strains 175

Strain tensor values (e-3):

⎡ ⎤
2.667 2.667 −0.3333
⎢ ⎥
⎣ 2.667 2.0 −1.5 ⎦
−0.3333 −1.5 −2.0

Let us rotate the coordinate by 60 degrees about the z-axis. The trans-
formation matrix becomes

1 about, theta = 'z', 60. # rotation angle and axis

2 Tz, about = [Link](theta, about = about)
3 print(f'Transformation tensor {theta:3.2f}°, w.r.t. {about}:\n{Tz}')

Transformation tensor 60.00°, w.r.t. z:

[[ 0.5 0.866 0. ]
[-0.866 0.5 0. ]
[ 0. 0. 1. ]]

1 60 = [Link](list(zip(T, Matrix(Tz))))
2 [Link]( 60, 'Stress formulas in the new coordinates',n_dgt=4)

Stress formulas in the new coordinates

⎡ ⎤
2.667 2.667 −0.3333
⎢ ⎥
⎣ 2.667 2.0 −1.5 ⎦
−0.3333 −1.5 −2.0

These are the strain components after 60 degree of transformation. Note

that to compute the engineering shear strains, we simply use:

γij = 2εij i = j (4.71)

4.12 Example: Computation of various strains

This example computes

• principal strains, strain invariants, mean strain, and deviatoric strains (see
Chapter 5 for detailed definition)

for a given strain state with 9 strain components. The definitions and for-
mulas are essentially the same as those for stresses discussed in the previous
chapter. Thus, all these are done using the functions made earlier for com-
puting various stresses.
176 Mechanics of Materials: Formulations and Solutions with Python

1 # Ref: Advanced Mechanics of Materials, by Boresi and Schmidt

2 # Example 2.9 State of Strain in a Torsion-Tension Member
3
4 import numpy as np
5 np.set_printoptions(precision=4,suppress=True)
6
7 #Given strain matrix in e-3 unit
8 #Ev=[Link]([[2.67, 1.335,.335],[1.335, 2.,-1.5],[.335,-1.5,-2.]]) #e-3
9 Ev = [Link]([[-.832, 0., 0.],
10 [0., -.832,.8725],
11 [0., .8725,1.667]]) #e-3
12 print(f'strain tensor (e-3) :\n{Ev}')
13
14 eigenValues, eigenVectors = gr.M_eigen(Ev)
15 print('Principal strains (Eigenvalues) (e-3) =',eigenValues)
16 print(f'Eigenvectors:\n{eigenVectors}')
17 print(f'Orthogonal:\n{[Link](eigenVectors.T)}')
18
19 I1, I2, I3 = gr.M_invariant(eigenValues)
20 print(f'The first strain invariant I1={I1:9.5f}e-3')
21 print(f'The second strain invariant I2={I2:9.5f}e-6')
22 print(f'The third strain invariant I3={I3:9.5f}e-9')
23
24 meanstrain, meanMatrix, deviatorM = [Link](Ev,eigenValues)
25 print(f'The mean strain = {meanstrain:9.5f}')
26 print(f'The mean strain tensor:\n{meanMatrix}')
27 print(f'The deviatoric strain tensor:\n{deviatorM}')

strain tensor (e-3) :

[[-0.832 0. 0. ]
[ 0. -0.832 0.8725]
[ 0. 0.8725 1.667 ]]
Principal strains (Eigenvalues) (e-3) = [ 1.9415 -0.832 -1.1065]
Eigenvectors:
[[ 0. 1. 0. ]
[ 0.3001 0. 0.9539]
[ 0.9539 0. -0.3001]]
Orthogonal:
[[1. 0. 0.]
[0. 1. 0.]
[0. 0. 1.]]
The first strain invariant I1= 0.00300e-3
The second strain invariant I2= -2.84292e-6
The third strain invariant I3= 1.78730e-9
The mean strain = 0.00100
The mean strain tensor:
[[0.001 0. 0. ]
[0. 0.001 0. ]
[0. 0. 0.001]]
The deviatoric strain tensor:
[[-0.833 0. 0. ]
[ 0. -0.833 0.8725]
[ 0. 0.8725 1.666 ]]
 Displacements and Strains 177

4.13 Compatibility equations

4.13.1 2D problems
For 2D problems, there are two displacement components u and v, but we
have three equations for strain–displacement relations. Using Eq. (4.14),
these can be written as
∂u
εxx =
∂x
∂v
εyy = (4.72)
∂y

1 ∂u ∂v
εxy = +
2 ∂y ∂x

Therefore, these two displacement components cannot in general be deter-

mined, if the three strain components are chosen arbitrarily. These strain
components must be somehow related. Such relation is called compatibility
equation.
Based on the Schwarz’s theorem which states that if a function is sec-
ond order continuously differentiable, the mixed partial derivatives must be
equal, regardless of the order of differentiation.
To obtain such a relation, we perform twice partial differentiation to the
εxx with respect to y, twice partial differentiation to the εyy with respect
to x, and twice partial differentiation to εxy with respect to x and y, which
gives

∂ 2 εxx ∂3u
=
∂y 2 ∂x∂y 2
∂ 2 εyy ∂3v
= (4.73)
∂x2 ∂y∂x2
 3
∂ 2 εxy 1 ∂ u ∂3v
= +
∂x∂y 2 ∂x∂y 2 ∂y∂x2

Substituting the first two expressions in the 3rd expression gives

∂ 2 εxx ∂ 2 εyy ∂ 2 εxy

+ = 2 (4.74)
∂y 2 ∂x2 ∂x∂y

This is the compatibility equation for 2D problems. All the strains must
satisfy this equation for correct displacements.
178 Mechanics of Materials: Formulations and Solutions with Python

4.13.2 3D problems*
For 3D problems, there are three displacement components (u, v, w), but we
have six equations for strain–displacement relations as given in Eq. (4.14).
Therefore, these three displacement components must also be compatible.
These compatibility equations can be derived through a similar procedure
used for 2D cases, and have been found as [1]
∂ 2 εxx ∂ 2 εyy ∂ 2 εxy
+ = 2
∂y 2 ∂x2 ∂x∂y
∂ 2 εxx ∂ 2 εzz ∂ 2 εxz
+ =2
∂z 2 ∂x 2 ∂x∂z
∂ εyy
2
∂ εzz
2 ∂ 2 εyz
+ = 2
∂z 2 ∂y 2 ∂y∂z
(4.75)
∂ 2 εxy ∂ 2 εzz ∂ 2 εyz ∂ 2 εxz
+ = +
∂z 2 ∂x∂y ∂x∂z ∂y∂z
∂ 2 εxz ∂ 2 εyy ∂ 2 εxy ∂ 2 εyz
+ = +
∂y 2 ∂x∂z ∂y∂z ∂x∂y
∂ 2 εyz ∂ 2 εxx ∂ 2 εxz ∂ 2 εxy
+ = +
∂x2 ∂y∂z ∂x∂y ∂x∂z
These compatibility equations can also be written in terms of stresses,
using the constitutive equations (to be discussed in the next chapter). These
compatibility equations need to be satisfied when solving a solid mechanics
problem.

4.14 Remarks

This chapter discusses first how the strain tensor (order 2) results from the
gradient of displacement tensor (order 1) all defined in a Cartesian coordinate
system. The derivation is done in two ways: a mathematical one that used
the gradient of the function of the displacement tensor, and mechanics one
that uses the kinematics of motion. The states of strain and its change are
also discussed, when the coordinate system transforms (rotates).

1. The value of the strain tensor depends on both the gradient of the dis-
placement tensor and the orientation of the fiber for which the strain is
measured. This is the root reason why a strain tensor has two indexes
and hence being a second-order tensor.
2. In all the formulas derived for the strain tensor, orthogonal projections
are strictly followed.
 Displacements and Strains 179

3. When coordinate rotates, the presentation of the strain tensor changes. It

uses twice the transformation matrix because it is a second-order tensor.
The process is the same as for the stress tensors.
4. The basic properties of a strain tensor (three strain invariants, principal
strains, and the Mohr circle) do not change with the coordinate transfor-
mation. These invariants can be used in establishing some failure criteria.
5. At the principal axes, the shear strain components vanish, and only the
normal strains are present, which are called principal strains. These are
invariant.
6. Principal strains and the axes can also be found using the eigenvalue
solver provided in this and the previous chapters.
7. Mohr circle can also be used as a geometric representation of strains in
2D and 3D, but less often compared to stresses. The codes for producing
the Mohr circle for strains are the same as those for stresses given in the
previous chapter.
8. Since the number of strain components is more than the number of
displacement components, compatibility equations are derived, which
need to be satisfied by the strains to ensure the deformation in solid
is compatible (no gaps and overlaps).

References
[1] A.P. Boresi, R.J. Schmidt and O.M. Sidebottom, Advanced Mechanics of Materials.
John Wiley & Sons, New York, 1985.
[2] J.M. Gere and S.P. Timoshenko, Mechanics of Materials, 1972.
[3] S. Timoshenko and J.N. Goodier, Theory of Elasticity. McGraw Hill, New York,
1970. [Link]
asticity&#38;ei=ICiKSsr3G4jwkQSbxMyPCg
[4] Z.L. Xu, Elasticity, Vols. 1&2. People’s Publisher, China, 1979.
[5] G.R. Liu and S.S. Quek, The Finite Element Method: A Practical Course.
Butterworth-Heinemann, New York, 2013.
[6] G.R. Liu and T.T. Nguyen, Smoothed Finite Element Methods. Taylor and Francis
Group, New York, 2010.
This page intentionally left blank
 Chapter 5

Elasticity Tensor and

Constitutive Equations

We import the necessary external modulus or dependencies for later use in

this chapter.

1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.

2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 init_printing(use_unicode=True) # For latex-like quality printing
11 #np.set_printoptions(precision=4,suppress=True) # Digits in print-outs
12 np.set_printoptions(formatter={'float': '{: 0.3e}'.format})

Chapter 3 defines the stresses and their relations with forces, and
Chapter 4 defines strains and their relationship with the displacements. We
need to know the relationship between the stress and strain tensors. The
elasticity tensor plays this role. This chapter is written in reference to [1–5].
As usual, our discussion is accompanied with Python codes written in ref-
erence to the documentations of Python, Sympy, and other Python modules.

5.1 Simplest case: 1D tests

To begin with, we look at the simplest uniaxially loaded 1D specimen that
often has a rectangular or circular cross-section. The cross-sectional dimen-
sion is much smaller than its length.

181
182 Mechanics of Materials: Formulations and Solutions with Python

Its normal strain εxx can be obtained as

ΔL
εxx = (5.1)
L
where x stands for the x-axis, which is vertical if the specimen is pulled ver-
tically using a tensile machine, as shown in the left figure in Fig. 5.1. The ΔL
is the length change in L under stress σxx applied by the tensile test machine.
The length change can be measured during experiment corresponding to the
level of stress.
Equation (5.1) shows that the strain is in fact an averaged one over the
base length L, which depends on the measurement method. If the displace-
ments between the two points are measured, L will be the distance of these
two points. If a strain gauge is used, it is the effective length of the gauge.
Assume the specimen or bar is made of linear elastic isotropic material,
implying that the stress and strain are proportional, and related thorough a
material constant E called Young’s modulus.
σxx = Eεxx (5.2)

Figure 5.1. Left: A tensile spacemen; Middle: A universal testing machine (Hegewald &
Peschke) for measuring the elastic constants and tensile strength for materials. Right:
A specimen is mounted in the machine.
Source: (Middle) Image permission by Vom Hofe Kaltstauchdraht GmbH, Altena from
[Link] Wikimedia Commons under the CC BY-SA 2.0 de license. (Right) Instron
under the CC0 license.
 Elasticity Tensor and Constitutive Equations 183

When the specimen is stretched in the x-direction, it may shrink in the

y-direction (horizontal). The shrink strain εyy is calculated using
ΔW
εyy = (5.3)
W
where ΔW is the length change in W under the same stress σxx . It can also
be measured during an experiment. This behavior of material is quantified
by Poisson’s ratio, denoted often by ν. It gives the relationship between two
mutually orthogonal normal strains:
εyy = −νεxx (5.4)
where εyy is a normal strain in the y-direction. Clearly Poisson’s ratio is
dimensionless. The minus sign ensures that Poisson’s ratio is positive for
usual materials that shrink under tension.
Note that for this 1D bar problem, σyy will be zero because the width
of the specimen is much smaller compared to its length, and no force or
pressure is applied in the y-direction.
Using a test machine shown in Fig. 5.1, Young’s modulus E, Poisson’s
ratio ν, yield stress, failure stress, etc. can be measured.
Equation (5.2) is well known as Hooke’s Law. Since strain is dimension-
less, E carries the same unit as stress: force/area and [P a = N/m2 ] is often
used. Young’s modulus is the most important elastic constant of a mate-
rial. For materials that are made through a standardized processes, Young’s
modulus can be found in material databases. A list of approximate or nom-
inal Young’s modulus for commonly used materials can be found at Young
modulus ([Link] modulus).
Interested readers may take a look at the online video Tensile Test Exper-
iment ([Link]

5.2 Comparison between Young’s modulus and spring

constants

To make a good sense of Young’s modulus, we can compare it with the

spring constant that we learned in high school, known also as Hooke’s Law.
Consider a spring fixed at one end, and it is pulled at another by a force
f . The spring is then elongated and the elongation is denoted as e. The
relationship between the force and displacement is then expressed as
f = ke (5.5)
where k is the spring constant. It carries a unit of force/length or [ N
m ]. It is
the stiffness of the spring.
184 Mechanics of Materials: Formulations and Solutions with Python

Consider now an uniform bar with cross-section area A and length L.

It is fixed also at one end, and pulled at another by a force f . The bar is
elongated and the elongation is denoted as e. Using Eq. (5.2), σxx = Eεxx ,
we have
AE
Aσxx = Lεxx (5.6)
   L   
f
 e
k
AE
It is clear that L is equivalent to the spring constant k. The unit of
AE m2 N2
L is also m m
= Nm . It is the stiffness of the bar. Clearly, E and A are
proportional, and L is inversely proportional to the bar stiffness.

5.3 General formulation

For a general problem in mechanics of materials, there are usually 9 stress

and strain components. Therefore, a general material property law is needed
to relate these stresses and strains. Such a law is called the constitutive
equation. It is assumed that each stress component is related to all the 9
strain components. Since there are 9 stress components, we shall have a total
of 9 × 9 = 81 material constants. For linear elastic materials, it can be given
in the most general form in tensor notation as [6–8]
σij = Cijkl εkl (5.7)

This is also known as the generalized Hooke’s Law [5], where Cijkl is a
fourth-order tensor, carrying 4 indexes. It carries the same unit as Young’s
modulus E: [P a]. It bridges two second-order tensors, σij and εij . This
ensures that each of the stress components depends on all the strain com-
ponents with independent elastic constants. Since each of the four indexes
of Cijkl has the same range of 3. There are a total of (3 × 3 × 3 × 3 =) 81
elasticity constants for materials in 3D. Thus, Eq. (5.7) is the most general
constitutive equation for solid materials.
Due to the symmetry of the stress and strain tensors, and the special
features the materials may have, some of these constants may equal, and
some of them may be zero. The actual number of the independent elastic
constants needed will be drastically reduced, which is discussed in detail
later.

5.4 Coordinate transformation of fourth-order tensor*

For many materials, their elastic constants change with directions, such as
anisotropic composite materials widely used in various industries and most
 Elasticity Tensor and Constitutive Equations 185

materials in nature [8, 9]. For analyzing these materials, we must find out
the rule of coordinate transformation for fourth-order tensors, so that we
can compute these constants when the coordinate rotates. This process is
discussed in this section.
Assume Eq. (5.7) is defined in coordinate system x, and now it is rotated
to x . Using the rule of the second-order stress transformation, we obtain:
 
σrs = ari asj σij =⇒ σrs = ari asj Cijkl εkl (5.8)

Cijkl εkl

On the other hand, the corresponding strain transformation rule gives

εkl = akp alq εpq (5.9)

Its reverse strain transformation becomes

εkl = apk aql εpq (5.10)

The swaps of the subscripts in aij are the result of the inverse transfor-
mation that is from coordinate system x to x, as discussed in the previous
chapter.
Substituting Eq. (5.10) into the 2nd equation in Eq. (5.8), we obtain:

σrs = ari asj Cijkl apk aql εpq (5.11)
  

Crspq

It is clear that i, j, k and l are contracted, resulting in free indexes r, s, p, q.

We finally obtain

Crspq = ari asj Cijkl apk aql (5.12)

This is the important rule of transformation of fourth-order tensor

of elastic stiffness, from coordinate system x to x .
When inverse transformation is needed, from x to x, all we need is to
swap the two indexes of aij . We will demonstrate this in the later examples.
It is seen that transformation of a fourth-order tensor requires 4 times
of operation by aij .

5.5 Codes for fourth order tensor transformation*

We write, for the first time, a Python code to perform transformations of

fourth-order tensors. To the best knowledge of the author, no such code have
186 Mechanics of Materials: Formulations and Solutions with Python

been written. This is made possible, by using the [Link]() function

that allows contraction operations by specifying individual axis in a multi-
dimensional numpy array.

1 def Tensor4_transfer(T,C4):
2 '''Transformation of 4th order tensors
3 input: T, transformation matrix (aij); C4: tensor to be transformed
4 return:C4, transformed tensor of 4th order'''
5
6 C4 = [Link]( T, C4, axes=([1],[0])) # contract i
7 C4 = [Link]( T, C4, axes=([1],[1])) # contract j
8 C4 = [Link](C4, T, axes=([3],[1])) # contract l
9 C4 = [Link](C4, T, axes=([2],[1])) # contract k
10
11 return C4

5.6 Reduction of elastic constants

5.6.1 Reduction due to stress and strain symmetry,

81 → 36
As discussed earlier, in general, have a total of (3 × 3 × 3 × 3 =) 81 elas-
ticity constants for materials in 3D. This means that we must find these 81
constants via experiments for a material. This is not practical in engineer-
ing applications, even if it is possible. This number needs to be and can be
drastically reduced. First, we can use the symmetry of the stress and strain
tensors to prove the symmetry of the fourth-order tensor.
Due to the symmetry of the stress tensors σij = σji , implying that ij can
be swapped. Thus, these constants satisfy the following symmetry equations.

Cijkl = Cjikl from stress tensor symmetry (5.13)

This reduces the number by (3 × 9) = 27 constants. Next, because of the

symmetry of the strain tensors εkl = εlk , implying that kl can be swapped.
These constants satisfy the following additional symmetry equations.

Cijkl = Cijlk from strain tensor symmetry (5.14)

This reduces the number by (3 × 6) = 18 constants. The total number of

the elasticity constants becomes 36, and Eq. (5.7) can now be written in the
following 6 × 6 matrix form, which indexes are still properly matched.
 Elasticity Tensor and Constitutive Equations 187

⎡ ⎤ ⎡ ⎤ ⎡ ⎤
σ11 C1111 C1122 C1133 C1123 C1113 C1112 ε11
⎢σ ⎥ ⎢C C2222 C2233 C2223 C2213 C2212 ⎥ ⎢ ⎥
⎢ 22 ⎥ ⎢ 2211 ⎥ ⎢ ε22 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢σ33 ⎥ ⎢C3311 C3322 C3333 C3323 C3313 C3312 ⎥ ⎢ ε33 ⎥
⎢ ⎥=⎢ ⎥=⎢ ⎥ (5.15)
⎢σ ⎥ ⎢C C2322 C2333 C2323 C2313 C2312 ⎥ ⎢ ⎥
⎢ 23 ⎥ ⎢ 2311 ⎥ ⎢2ε23 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣σ13 ⎦ ⎣C1311 C1322 C1333 C1323 C1313 C1312 ⎦ ⎣2ε13 ⎦
σ12 C1211 C1222 C1233 C1223 C1213 C1212 2ε12

Note that the stress and strain tensors are now written as vectors because
of its symmetry and thus the 9 components is reduced to 6.
This number can be further reduced, by making use of the smoothness
of the strain energy density in a stressed elastic solid.

5.6.2 Reduction due to the smoothness of the strain

energy density
[Link] Strain energy density in solids
When a solid is stressed, it stores some energy known as the strain energy.
The strain energy per unit volume is called strain energy density, carrying a
unit of mN2 m J
m = m3 . In general, the strain energy density will be different at
different locations in solid, and hence is a function of coordinates. Here, let
us evaluate the strain energy density at a point in solid.
Consider first a simple case, where at a point in solid there is only a
single component of stress σ and its corresponding strain ε all in the same
direction. The stress-strain relationship can be schematically shown Fig. 5.2.
The stress-strain relation is the function represented by the thick black
curve, which gives the stress as a function of strain: σ(ε). Assume the strain
compatible for variation, implying the operation of dε is possible, and thus
the integration of stress-strain function can be carried out. The strain energy
density U0 is then defined by the area under the curve:
ε
U0 = σ(ε )dε (5.16)
0

Here we use prime to indicate the integration variable.

Note that the complementary energy density C0 is the area above the
thick black curve shown Fig. 5.2. We will make use of it in developing an
energy method in Chapter 13. For now, our focus is on the strain energy
density U0 .
Assume U0 is second order continuously differentiable with respect to the
strains, we can then perform this following analysis.
188 Mechanics of Materials: Formulations and Solutions with Python

Figure 5.2. Strain energy density U0 and complementary energy density C0 at a point
in a stressed solid.

First, for 1D solids, we shall have

∂U0
=σ (5.17)
∂ε
In a general 3D solid, we shall have a total of 9 mutually orthogonal
stress components, and corresponding 9 mutually orthogonal strain com-
ponents. Each pair of these 9 stress-strain component pairs is in the same
direction. Therefore, the strain energy density should be the sum of all the
strain energy densities of the corresponding stress-strain component pairs,
which can be expressed as
ε11 ε22 ε33
U0 = σ(ε11 )dε11 + σ(ε22 )dε22 + σ(ε33 )dε33
0 0 0
ε23 ε13 ε12 (5.18)
+2 σ(ε23 )dε23 + 2 σ(ε13 )dε13 + 2 σ(ε12 )dε12
0 0 0

This is the most general formula for evaluating the strain energy den-
sity in a 3D solid. Note that this is possible because of the use of orthogonal
coordinates, over which the stress and strain tensors are defined. This for-
mula is used not only in this chapter, but also in Chapter 13 when discussing
energy methods. It is a widely used fundamental formula in solid mechanics.

[Link] Symmetry from the smoothness of the strain energy density*

In Eq. (5.18), we used the shear stress equivalence, and hence 9 terms is
reduced to 6. Partial differentiations of the strain energy density with respect
 Elasticity Tensor and Constitutive Equations 189

to each of these 6 strains give

∂U0 ∂U0 ∂U0

= σ11 ; = σ22 ; = σ33
∂ε11 ∂ε22 ∂ε33
(5.19)
∂U0 ∂U0 ∂U0
= 2σ23 ; = 2σ13 ; = 2σ12
∂ε23 ∂ε13 ∂ε12

Using the first two equations in Eq. (5.15), we obtain

∂U0
= σ11 = C1111 ε11 + C1122 ε22 + C1133 ε33 + 2C1123 ε23
∂ε11
+ 2C1113 ε13 + 2C1112 ε12
∂U0
= σ22 = C2211 ε11 + C2222 ε22 + C2233 ε33 + 2C2223 ε23
∂ε22
+ 2C2213 ε13 + 2C2212 ε12 (5.20)

Next, we perform the following 2nd derivatives to U0 , which leads to

∂ 2 U0 ∂ 2 U0
= C1122 ; = C2211 (5.21)
∂ε11 ∂ε22 ∂ε22 ∂ε11
Based on Schwarz’s theorem, if a function is second order continuously
differentiable, the mixed partial derivatives must be equal, regardless of the
sequential order of differentiation. This gives

C1122 = C2211 (5.22)

Similarly, using the first and fourth equations in Eq. (5.15), we obtain

C1123 = C2311 (5.23)

Repeating the same, for other pairs of equations in Eq. (5.15), we shall
find the following additional symmetry condition for the fourth-order elas-
ticity tensor:

Cijkl = Cklij (5.24)

This means that ij and kl can be swapped.

[Link] Reduction due to strain energy density symmetry, 36 → 21*

Condition Eq. (5.24) makes the C in Eq. (5.15) symmetric, and hence
reduces number of elasticity constants by 15. The independent constants
become 21. The stress-strain relation can be written in the following
190 Mechanics of Materials: Formulations and Solutions with Python

symmetric matrix of 6 × 6.
⎡ ⎤ ⎡ ⎤⎡ ⎤
σ11 C1111 C1122 C1133 C1123 C1113 C1112 ε11
⎢σ ⎥ ⎢ C2222 C2233 C2223 C2213 C2212 ⎥ ⎢ ⎥
⎢ 22 ⎥ ⎢ ⎥ ⎢ ε22 ⎥
⎢σ ⎥ ⎢ C3333 C3323 C3313 C3312 ⎥ ⎢ ε33 ⎥
⎥ ⎢
⎢ 33 ⎥ ⎢ ⎥
⎢ ⎥=⎢ ⎥⎢ ⎥ (5.25)
⎢σ23 ⎥ ⎢ C2323 C2313 C2312 ⎥ ⎢2ε23 ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣σ13 ⎦ ⎣ sym. C1313 C1312 ⎦ ⎣2ε13 ⎦
σ12 C1212 2ε12
        

σ C 
ε

5.6.3 Tensor vs. matrix, a discussion*

Note in Eq. (5.25) that the stress and strain tensors are now written as vec-
tors. Due to symmetry the 9 components in the tensor is reduced to 6. In this
form of vector stress or strain, it is no longer a tensor because our transfor-
mation rule does not apply any more. Therefore, when we want to perform
coordinate rotation and examine the stress and strain components, we would
still need to use their tensor form of stress, Eq. (5.8), or strain Eq. (5.9).
The same situation is for the elasticity constant matrix C. It is no longer
a tensor. The transformation of its components needs to go back to use
the tensor transformation rule given in Eq. (5.12), or use our function Ten-
sor4 Transfer() given earlier. Therefore, conversion between 2D matrix C
and fourth-order tensor Cijkl is needed.

5.7 The Voigt notation*

By substituting pairs of indexes 11 → 1, 22 → 2, 33 → 3, 23 → 4, 13 → 5,
12 → 6, Eq. (5.25) can be further expressed using the so-called Voigt notation
[9, 10]:
⎡ ⎤ ⎡ ⎤⎡ ⎤
σ11 C11 C12 C13 C14 C15 C16 ε11
⎢σ ⎥ ⎢ C22 C23 C24 C25 C26 ⎥ ⎢ ⎥
⎢ 22 ⎥ ⎢ ⎥ ⎢ ε22 ⎥
⎢σ ⎥ ⎢ C33 C34 C35 C36 ⎥ ⎢ ⎥
⎢ 33 ⎥ ⎢ ⎥ ⎢ ε33 ⎥
⎢ ⎥=⎢ ⎥⎢ ⎥ (5.26)
⎢σ23 ⎥ ⎢ C44 C45 C46 ⎥ ⎢2ε23 ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣σ13 ⎦ ⎣ sym. C55 C56 ⎦ ⎣2ε13 ⎦
σ12 C66 2ε12
        

σ C 
ε
which can be written concisely as
 = C
σ ε (5.27)
This is the general matrix form of stress–strain or constitutive equation
for anisotropic materials. The C matrix is symmetric. It can be computed
 Elasticity Tensor and Constitutive Equations 191

for anisotropic composite materials [8] and for isotropic materials using its
Young’s modulus and Poisson’s ratio (see Section 2.2.2 in textbook [5]). The
C matrix can also be expressed as [Cαβ ], where α, β = 1, 2, . . . , 6. The use of
Voigt notation is important for engineering applications because it enables
practical means to obtain these material constants through engineering mea-
surements. We can now produce a 6 × 6 stiffness matrix first, and then
convert it to a 3 × 3 × 3 × 3 tensor.
We mention, but without further elaboration, that there is an alternative
handcrafted way to convert between a 6 × 6 compliance matrix and a 3 ×
3 × 3 × 3 4th strain tensor. Interested readers may refer to Ref. [9].

5.8 Codes for conversion between a 2D matrix to

fourth-order tensor*
We write the following codes for conversion between a C(6,6) matrix and a
4th tensor C(3,3,3,3). The code function C2toC4() converts 2D C(6,6) matrix
(in the Voigt notation) to its 4th tensor C(3,3,3,3) form. Code function
C4toC2() converts a 4th tensor C(3,3,3,3) to its 2D C(6,6) matrix.

1 def C2toC4(C2):
2 '''To convert C(6,6) matrix (the Voigt notation) to
3 4th tensor C(3,3,3,3).
4 '''
5 C4 = [Link]((3,3,3,3)) # Initialization
6
7 # Pass over all C(6,6) to parts of C(3,3,3,3)
8 C4[0,0,0,0],C4[0,0,1,1],C4[0,0,2,2] = C2[0,0],C2[0,1],C2[0,2]
9 C4[0,0,1,2],C4[0,0,0,2],C4[0,0,0,1] = C2[0,3],C2[0,4],C2[0,5]
10
11 C4[1,1,0,0],C4[1,1,1,1],C4[1,1,2,2] = C2[1,0],C2[1,1],C2[1,2]
12 C4[1,1,1,2],C4[1,1,0,2],C4[1,1,0,1] = C2[1,3],C2[1,4],C2[1,5]
13
14 C4[2,2,0,0],C4[2,2,1,1],C4[2,2,2,2] = C2[2,0],C2[2,1],C2[2,2]
15 C4[2,2,1,2],C4[2,2,0,2],C4[2,2,0,1] = C2[2,3],C2[2,4],C2[2,5]
16
17 C4[1,2,0,0],C4[1,2,1,1],C4[1,2,2,2] = C2[3,0],C2[3,1],C2[3,2]
18 C4[1,2,1,2],C4[1,2,0,2],C4[1,2,0,1] = C2[3,3],C2[3,4],C2[3,5]
19
20 C4[0,2,0,0],C4[0,2,1,1],C4[0,2,2,2] = C2[4,0],C2[4,1],C2[4,2]
21 C4[0,2,1,2],C4[0,2,0,2],C4[0,2,0,1] = C2[4,3],C2[4,4],C2[4,5]
22
23 C4[0,1,0,0],C4[0,1,1,1],C4[0,1,2,2] = C2[5,0],C2[5,1],C2[5,2]
24 C4[0,1,1,2],C4[0,1,0,2],C4[0,1,0,1] = C2[5,3],C2[5,4],C2[5,5]
25
26 # Impose (minor) symmetric conditions
27 apply_symmetry(C4, key = "all", tol=1.e-4)
28
29 return C4
192 Mechanics of Materials: Formulations and Solutions with Python

1 def apply_symmetry(C4, key = "all", tol=1.e-2):

2 '''Impose (minor) symmetric conditions
3 '''
4 if key == "all" or key == "ij":
5 for k in range(3):
6 for l in range(3):
7 for i in range(3):
8 for j in range(i+1,3):
9 if abs(C4[j,i,k,l]) <= tol:
10 C4[j,i,k,l]=C4[i,j,k,l]
11
12 if key == "all" or key == "kl":
13 for k in range(3):
14 for l in range(k+1,3):
15 for i in range(3):
16 for j in range(3):
17 if abs(C4[i,j,l,k]) <= tol:
18 C4[i,j,l,k]=C4[i,j,k,l]
19
20 if key == "all" or key == "ijkl":
21 for k in range(3):
22 for l in range(3):
23 for i in range(k+1,3):
24 for j in range(l+1,3):
25 if abs(C4[i,j,k,l]) <= tol:
26 C4[i,j,k,l]=C4[k,l,i,j]
27 return C4

1 def C4toC2(C4):
2 '''To convert 4th tensor C(3,3,3,3) to C(6,6) matrix
3 (the Voigt notation).
4 '''
5 C2 = [Link]((6,6))
6 C2[0,0],C2[0,1],C2[0,2]=C4[0,0,0,0],C4[0,0,1,1],C4[0,0,2,2]
7 C2[0,3],C2[0,4],C2[0,5]=C4[0,0,1,2],C4[0,0,0,2],C4[0,0,0,1]
8
9 C2[1,0],C2[1,1],C2[1,2]=C4[1,1,0,0],C4[1,1,1,1],C4[1,1,2,2]
10 C2[1,3],C2[1,4],C2[1,5]=C4[1,1,1,2],C4[1,1,0,2],C4[1,1,0,1]
11
12 C2[2,0],C2[2,1],C2[2,2]=C4[2,2,0,0],C4[2,2,1,1],C4[2,2,2,2]
13 C2[2,3],C2[2,4],C2[2,5]=C4[2,2,1,2],C4[2,2,0,2],C4[2,2,0,1]
14
15 C2[3,0],C2[3,1],C2[3,2]=C4[1,2,0,0],C4[1,2,1,1],C4[1,2,2,2]
16 C2[3,3],C2[3,4],C2[3,5]=C4[1,2,1,2],C4[1,2,0,2],C4[1,2,0,1]
17
18 C2[4,0],C2[4,1],C2[4,2]=C4[0,2,0,0],C4[0,2,1,1],C4[0,2,2,2]
19 C2[4,3],C2[4,4],C2[4,5]=C4[0,2,1,2],C4[0,2,0,2],C4[0,2,0,1]
20
21 C2[5,0],C2[5,1],C2[5,2]=C4[0,1,0,0],C4[0,1,1,1],C4[0,1,2,2]
22 C2[5,3],C2[5,4],C2[5,5]=C4[0,1,1,2],C4[0,1,0,2],C4[0,1,0,1]
23
24 return C2
 Elasticity Tensor and Constitutive Equations 193

5.9 Monoclinic materials, 21 → 13

For special materials, the 2D Voigt matrix can be further simplified if there
is some plane of symmetry in terms of material properties. Linearly elas-
tic monoclinic solid has one plane of material symmetry, say z = 0. This
means that the shear strains ε13 and ε23 should not result in any normal
stresses. Thus, C14 = C24 = C34 = 0 and C15 = C25 = C35 = 0. In addi-
tion, the shear strains ε12 should not result in any shear stresses σ23 and
σ23 , due to the mirror symmetry with respect to z = 0. This forces 8 con-
stants to zero, and C has now 13 independent elasticity constants as shown
below:
⎡ ⎤
C11 C12 C13 0 0 C16
⎢ C22 C23 0 0 C26 ⎥
⎢ ⎥
⎢ ⎥
⎢ C33 0 0 C36 ⎥
⎢ ⎥
C = [Cαβ ] = ⎢ ⎥ (5.28)
⎢ C44 C45 0 ⎥
⎢ ⎥
⎢ ⎥
⎣ sym. C55 0 ⎦
C66

5.10 Orthotropic materials, 13 → 9

If the material has three mutually orthogonal planes of mirror symmetry,

the number of independent elastic constants reduces to 9. The arguments on
this are similar to these we made in the previous sub-section. The C matrix
becomes
⎡ ⎤
C11 C12 C13 0 0 0
⎢ C22 C23 0 0 ⎥
⎢ 0 ⎥
⎢ ⎥
⎢ C33 0 0 0 ⎥
⎢ ⎥
C = [Cαβ ] = ⎢ ⎥ (5.29)
⎢ C44 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎣ sym. C55 0 ⎦
C66

5.11 Transversely isotropic materials, 9 → 5

When the mechanical properties of the material are axial-symmetric about

an axis that is normal to a plane of isotropy, say x–y plane, the independent
194 Mechanics of Materials: Formulations and Solutions with Python

number of elasticity constants becomes 5, and the C matrix can be given by

⎡ ⎤
C11 C12 C13 0 0 0
⎢ C11 C13 ⎥
⎢ 0 0 0 ⎥
⎢ ⎥
⎢ C33 0 0 0 ⎥
⎢ ⎥
C = [Cαβ ] = ⎢ ⎥ (5.30)
⎢ C44 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎣ sym. C44 0 ⎦
1
2 (C11 − C12 )

5.12 Cubic materials, 5 → 3

Cubic materials possess three mutually orthogonal planes of mirror symme-

try, and the elasticity constants are the same along these three orthogonal
directions. Its independent number of elasticity constants becomes 3, and
the C matrix can be written as
⎡ ⎤
C11 C12 C12 0 0 0
⎢ C11 C12 0 0 ⎥
⎢ 0 ⎥
⎢ ⎥
⎢ C11 0 0 0 ⎥
⎢ ⎥
C = [Cαβ ] = ⎢ ⎥ (5.31)
⎢ C44 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎣ sym. C44 0 ⎦
C44

5.13 Isotropic materials, 3 → 2

The material properties of an isotropic material are direction independent.

Therefore, all the elastic constants are invariant with respect to coordinate
transformation. The constitutive equation for isotropic materials is thus the
simplest and also most widely used. It has a number of different forms. This
section provides the major ones.

5.13.1 For 3D problems

Due to the direction-independence, any plane would be a plane of mirror
symmetry for isotropic materials. Its independent number of elasticity con-
stants becomes 2, and the C matrix can be given by
 Elasticity Tensor and Constitutive Equations 195

⎡ ⎤
C11 C12 C12 0 0 0
⎢ C11 C12 0 0 0 ⎥
⎢ ⎥
⎢ C11 ⎥
⎢ 0 0 0 ⎥
C = [Cαβ ] = ⎢ ⎥
⎢ (C11 − C12 )/2 0 0 ⎥
⎢ ⎥
⎣ sym. (C11 − C12 )/2 0 ⎦
(C11 − C12 )/2

(5.32)
where
E(1 − ν)
C11 = = 2G + λ
(1 − 2ν)(1 + ν)
Eν 2Gν G(E − 2G)
C12 = λ = = = (5.33)
(1 − 2ν)(1 + ν) 1 − 2ν 3G − E
E
G=μ=
2(1 + ν)

In which, E is Young’s Modulus, ν is Poisson’s ratio, and G the shear

modulus. They are also called the engineering constants, and can be obtained
from engineering experiments. Formel definition for the shear modulus G,
and the derivation of relations in Eq. (5.33) will be given later, when dis-
cussing about the compliance matrix where the engineering constants are
defined.
Due to the last equation in Eq. (5.33), we have only two independent
elastic constants, which is any pair of these three: E, ν, and G. Due to the
various formulations for solid mechanics problems, these elastic constants are
also written in other forms, such as the bulk modulus K for volume deforma-
tion, P-wave modulus M for wave propagation problems. A comprehensive
list of conversion formulas for elastic constants can be found at the Young
modulus ([Link] modulus) page.
The constitutive equation for isotropic materials can be written explic-
itly as
σ11 = λe + 2Gε11
σ22 = λe + 2Gε22
σ33 = λe + 2Gε33
(5.34)
σ23 = 2Gε23
σ12 = 2Gε12
σ13 = 2Gε13
196 Mechanics of Materials: Formulations and Solutions with Python

where
e = ε11 + ε22 + ε33 (5.35)
Equation (5.34) can written in a concise indicial form:
σij = 2Gεij + λδij εkk (5.36)
This equation contains only two constants: G and λ. Both are called
Lamé’s constants.
Using Eq. (5.36), the 4th elasticity tensor for isotropic materials can be
given as
Cijkl = G(δik δjl + δil δjk ) + λδij δkl (5.37)
Equation (5.37) can be confirmed easily by multiplying it with εkl from
the right, which shall lead to Eq. (5.36).
By setting j = i in Eq. (5.36), we obtain:
σkk = (2G + 3λ)εkk = 3Kεkk (5.38)
where K known as the bulk modulus, given by
2 E
K =λ+ G= (5.39)
3 3(1 − 2ν)
We have 3 in front of K because σkk is 3 times the hydrostatic stress. It
is seen that ν cannot be 0.5. When ν approaches to 0.5, the bulk modulus
approaches to infinity, implying that the material becomes incompressible.
This the root of the so-called volumetric locking issue in displacement meth-
ods for solid mechanics problems [11, 12].
Substituting Eq. (5.38) into Eq. (5.36), we obtain strains in terms of
stresses:
λδij 1
εij = σkk + σij (5.40)
2G(2G + 3λ) 2G
Equation (5.36) can also be written in matrix form:
σ = 2Gε + λtr(ε)I (5.41)
Alternatively, C matrix in Eq. (5.32) for isotropic materials can be written
explicitly in these engineering constants:
⎡ ⎤
λ + 2G λ λ 0 0 0
⎢ λ + 2G λ 0 0 0⎥
⎢ ⎥
⎢ ⎥
⎢ λ + 2G 0 0 0 ⎥
C = [Cαβ ] = ⎢
⎢
⎥ (5.42)
⎢ G 0 0⎥ ⎥
⎢ ⎥
⎣ sym. G 0⎦
G
 Elasticity Tensor and Constitutive Equations 197

Using Eqs. (5.33) and (5.42), the stress-strain relations for isotropic mate-
rials can be written in the following explicit form, in terms of Young’s mod-
ulus and Poisson’s ratio.
E
σ11 = [(1 − ν)ε11 + ν(ε22 + ε33 )]
(1 + ν)(1 − 2ν)
E
σ22 = [(1 − ν)ε22 + ν(ε11 + ε33 )]
(1 + ν)(1 − 2ν)
(5.43)
E
σ33 = [(1 − ν)ε33 + ν(ε11 + ε22 )]
(1 + ν)(1 − 2ν)
E E E
σ23 = ε23 ; σ13 = ε13 ; σ12 = ε12
(1 + ν) (1 + ν) (1 + ν)
We emphasize that for isotropic materials, the elastic constants and hence
their fourth-order elastic tensor are direction-independent, will not change
with the coordinate rotation, and no coordinate transformation is needed
for the elasticity tensor for isotropic materials. We will demonstrate this
important feature in later examples.

5.13.2 For stress and strain invariants

We can also write the constitution equations for stress and strain invariants.
Such relations are particularly useful because all their invariants are inde-
pendent of coordinates. They can be obtained using the definitions of stress
invariants (see Chapter 3) and strain invariants (see Chapter 4), and the
constitutive equations Eq. (5.41). These are found as
I1 = (2G + 3λ)I¯1
I2 = λ(4G + 3λ)I¯12 + 4G2 I¯2 (5.44)
I3 = λ2 (2G + λ)I¯13 + 4λG2 I¯1 I¯2 + 8G3 I¯3
Studying general 3D problems can often be difficult. Many problems can
be simplified as 2D, in which many stress or strain components are zero. The
next sections derive various constitutive equations for 2D problems.

5.13.3 For 2D plane stress problems

One type of 2D problem is called 2D plane stress problems. The nonzero
stresses stay in a plane, say in the x1 –x2 plane, we shall have σ33 = σ23 =
σ13 = 0. Thus, the 3rd, 4th and 5th equations in Eq. (5.43) give
−ν
ε33 = (ε11 + ε22 ); ε23 = 0; ε13 = 0 (5.45)
1−ν
198 Mechanics of Materials: Formulations and Solutions with Python

Note that the normal strain ε33 is nonzero. Substituting the first equation
in Eqs. (5.45) to (5.43), with a little lengthy derivation, the stress-strain
relations reduces to:

E
σ11 = (ε11 + νε22 )
(1 − ν 2 )
E
σ22 = (ε22 + νε11 ) (5.46)
(1 − ν 2 )
E
σ12 = ε12
(1 + ν)

This is often written in matrix form:

⎡ ⎤ ⎡ ⎤⎡ ⎤
σ11 1 ν 0 ε11
⎢ ⎥ E ⎢ν 1
⎣σ22 ⎦ = ⎣ 0 ⎥ ⎢
⎦ ⎣ ε22 ⎦
⎥
(1 − ν 2 ) (1−ν)
σ12 0 0 2
2ε12
  
C
⎡ ⎤⎡ ⎤ (5.47)
c11 c12 0 ε11
⎢ ⎥⎢ ⎥
= ⎣c12 c22 0 ⎦ ⎣ ε22 ⎦
0 0 c33 2ε12
  
C

5.13.4 For 2D plane strain problems

Another type of 2D problem is the 2D plane strain problem. In this case
the nonzero strains stay in a plane, say in the x1 –x2 plane, and we shall have
ε13 = ε23 = ε33 = 0. Equation (5.43) becomes

E
σ11 = [(1 − ν)ε11 + νε22 ]
(1 + ν)(1 − 2ν)
E
σ22 = [νε11 + (1 − ν)ε22 ]
(1 + ν)(1 − 2ν)
(5.48)
E
σ12 = ε12
(1 + ν)
νE
σ33 = [ε22 + ε11 ]; σ23 = 0; σ13 = 0
(1 + ν)(1 − 2ν)
 Elasticity Tensor and Constitutive Equations 199

Note in this case that σ33 is nonzero. The first 3 equations in Eq. (5.48)
is often written in matrix form:
⎡ ⎤ ⎡ ν ⎤⎡ ⎤
σ11 1 1−ν 0 ε11
⎢ ⎥ E(1 − ν) ⎢ ν ⎥⎢ ⎥
⎣σ22 ⎦ = ⎣ 1−ν 1 0 ⎦ ⎣ ε22 ⎦
(1 + ν)(1 − 2ν) (1−2ν)
σ12 0 0 2(1−ν)
2ε12
  
C
⎡ ⎤⎡ ⎤ (5.49)
c11 c12 0 ε11
⎢ ⎥⎢ ⎥
= ⎣c12 c22 0 ⎦ ⎣ ε22 ⎦
0 0 c33 2ε12
  
C

Note that we can obtain Eq. (5.49) simply using the following substitu-
tions in Eq. (5.47) (plane stress to plane strain).
ν
first: substitute ν by
(1 − ν)
E
then: substitute E by (5.50)
(1 − ν 2 )
no change in G

5.14 Mean and deviatoric stress and strain tensors

The mean stress is defined as the averaged normal stresses:

σkk σ11 + σ22 + σ33

σm = = (5.51)
3 3
It is essentially the same as the so-called hydrostatic stress. The mean
stress tensor is formed using the mean stress:

σij = σm δij (5.52)

Or in matrix form:
⎡ ⎤
σm 0 0
σm = ⎣ 0 σm 0 ⎦ (5.53)
0 0 σm
200 Mechanics of Materials: Formulations and Solutions with Python

The deviatoric stress tensor is defined as the stress tensor minus the
mean stress tensor:
⎡ ⎤
σ11 − σm σ12 σ13
σd = [sij ] = ⎣ σ12 σ22 − σm σ23 ⎦ (5.54)
σ13 σ23 σ33 − σm

Or
σkk
sij = σij − δij (5.55)
3
The deviatoric strain tensor is defined in the similar way as
εkk
eij = εij − δij (5.56)
3
The deviatoric stress and strain tensors are related as follows:

sij = 2Geij (5.57)

They are related only by the shear modulus.

The mean stress and mean strain are related as
2
σm = λ + G εkk (5.58)
3
  
K

where K is the bulk modulus given in Eq. (5.39).

5.15 Compliance matrix

The C matrix is also called stiffness matrix of materials because it represents

how stiff the materials are. Frequently, we use also compliance matrix of
materials. It can be obtained by inverting the stiffness matrix, as long as the
material is stable (it does not flow like a fluid, C is positive definite, hence
invertible).

S = [Sαβ ] = C−1 (5.59)

5.15.1 Compliance matrix for orthotropic materials

Many composite materials used in engineering are orthotropic. We often need
to obtain first the compliance matrix through measuring Young’s modulus
and Poisson’s ratios, and then compute the stiffness matrix via inversion.
 Elasticity Tensor and Constitutive Equations 201

For orthotropic materials, we have

⎡ ⎤
⎡ ⎤ 1 −ν21 −ν31
0 0 0 ⎡ ⎤
ε11 E1
⎢ −ν12
E2 E3
⎥ σ11
⎢ε ⎥ ⎢ E 1 −ν32
0 0 0 ⎥ ⎢σ ⎥
⎢ 22 ⎥ ⎢ 1 E2 E3 ⎥ ⎢ 22 ⎥
⎢ ⎥ ⎢ −ν13 −ν23 1 ⎥⎢ ⎥
⎢ ε33 ⎥ ⎢ E1 E2 E3 0 0 0 ⎥ ⎢σ33 ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢2ε ⎥ = ⎢ 0 1
0 ⎥ ⎢ ⎥ (5.60)
⎢ 23 ⎥ ⎢ 0 0 G23 0 ⎥ ⎢σ23 ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣2ε13 ⎦ ⎢ 0 1
0 ⎥ ⎣σ ⎦
⎣ 0 0 0 G13 ⎦ 13
2ε12 1 σ12
   0 0 0 0 0 G12   

ε
   σ
S

Or

ε = S
σ (5.61)

In Eq. (5.60), Ei is Young’s modulus in the xi direction, Gij is the shear

modulus between fibers in xi -direction and xj -direction defined as
σij σij
Gij := = (no summation) (5.62)
γij 2εij

Poisson’s ratio, νij , is defined as

εjj
νij := − (no summation) (5.63)
εii
Due to the symmetry, we have the following symmetric relations:
ν12 ν21 ν13 ν31 ν23 ν32
= ; = ; = (5.64)
E1 E2 E1 E3 E2 E3
Due to the anisotropy, experiments discussed in Section 5.1 need to be
conducted using samples in three directions. Since Young’s moduli, Poisson’s
ratios, and shear moduli can be measured from experiments, it is often easier
to obtain the compliance matrix, and then convert it to stiffness matrix.

5.15.2 Python code for elasticity matrix for orthotropic

materials (3D problems)
We write the following code function to form a 6 × 6 compliance matrix
using Young’s moduli and Poisson’s ratios, and then compute the 6 × 6
stiffness matrix via matrix inversion. We will use it multiple times in the
later examples.
202 Mechanics of Materials: Formulations and Solutions with Python

1 def E_SnC3D(E1, E2, E3, m12, m13, m23, G23, G13, G12):
2 '''Numpy code to compute the S and C matrix in Voigt notation for
3 given Young's moduli and Poisson's ratios of orthotropic materials
4 for 3D problems.
5 '''
6 S = [Link]((6,6)) #initialization
7
8 m21, m31, m32 = m12/E1*E2, m13/E1*E3, m23/E2*E3
9
10 # compute the compliance matrix S
11 S[0,0], S[1,1], S[2,2] = 1./E1, 1./E2, 1./E3
12 S[0,1], S[0,2], S[1,2] = -m21/E2, -m31/E3, -m32/E3
13 S[3,3], S[4,4], S[5,5] = 1./G23, 1./G13, 1./G12
14 S[1,0], S[2,0], S[2,1] = S[0,1], S[0,2], S[1,2]
15
16 # compute C matrix
17 C = [Link](S)
18
19 return C, S

1 def E_SnC3Dsp(E1, E2, E3, m12, m13, m23, G23, G13, G12):
2 '''Sympy code to compute the S and C matrix in Voigt notation for
3 given Young's moduli and Poisson's ratios of orthotropic materials
4 for 3D problems.
5 '''
6 S = [Link](6,6) #initialization
7
8 m21, m31, m32 = m12/E1*E2, m13/E1*E3, m23/E2*E3
9
10 # compute the compliance matrix S
11 S[0,0], S[1,1], S[2,2] = 1/E1, 1/E2, 1/E3
12 S[0,1], S[0,2], S[1,2] = -m21/E2, -m31/E3, -m32/E3
13 S[3,3], S[4,4], S[5,5] = 1/G23, 1/G13, 1/G12
14 S[1,0], S[2,0], S[2,1] = S[0,1], S[0,2], S[1,2]
15
16 # compute C matrix
17 C = [Link]()
18
19 return C, S
 Elasticity Tensor and Constitutive Equations 203

5.15.3 Plane stress problems

For plane stress problems, we shall have σ33 = σ23 = σ13 = 0. Equation
(5.60) can be simplified as

⎡ ⎤
⎡ ⎤ 1
E1 − νE212 0 ⎡ ⎤
ε11 ⎢ ⎥ σ11
⎢ ⎥ ⎢ ν12 1 ⎥⎢ ⎥
⎣ 22 ⎦ ⎢− E1
ε = E2 0 ⎥ ⎣σ22 ⎦ (5.65)
⎣ ⎦
2ε12 1 σ12
0 0 G12
  
S

Strain ε33 is given by


ν13 ν23
ε33 =− σ11 + σ22 (5.66)
E1 E2

In reverse, Eq. (5.65) becomes

⎡ ⎤ ⎡ ⎤⎡ ⎤
σ11 E1 ν21 E1 0 ε11
⎢ ⎥ 1 ⎢ ⎥⎢ ⎥
⎣σ22 ⎦ = ⎣ν12 E2 E2 0 ⎦ ⎣ ε22 ⎦ (5.67)
1 − ν12 ν21
σ12 0 0 G12 (1 − ν12 ν21 ) 2ε12
  
C

For isotropic materials, simply set Ei = E, νij = ν, and Gij = G.

5.15.4 Python code for elasticity matrix for orthotropic

materials (2D plane stress)
The following code function forms a 3 × 3 compliance matrix using Young’s
moduli, Poisson’s ratios, and shear moduli. And then computes the 3 ×
3 stiffness matrix via inversion. We will use it multiple times in the later
examples. This code is for 2D plane stress problems.
204 Mechanics of Materials: Formulations and Solutions with Python

1 def E_SnC2D_stress(E1, E2, 12, G12):

2 '''Numpy code to compute S and C matrix in Voigt notation for given
3 Young's moduli Ei, and Poisson's ratios 12, shear modulus G12
4 of orthotropic materials for 2D plan stress problems.
5 '''
6 S = [Link]((3,3)) # sp: S = [Link](3,3) #initialization
7 21 = 12/E1*E2
8
9 # compute the compliance matrix S
10 S[0,0], S[1,1] = 1/E1, 1/E2
11 S[0,1] = - 21/E2
12 S[1,0] = S[0,1]
13 S[2,2] = 1/G12
14
15 # compute C matrix
16 C = [Link](S) # sp: S = [Link](S)
17
18 return C, S

5.15.5 Plane strain problems

For plane strain problems (ε13 = ε23 = ε33 = 0), the derivation of compli-
ance matrix for orthotropic materials is quite tedious and prone to making
mistakes. We thus use Sympy to help in the derivation. We define first some
symbolic variables.

1 E1, E2, E3 = symbols("E1, E2, E3")

2 11, 22 = symbols(" 11, 22")
3 12, 21, 13, 31, 23, 32 = symbols(" 12, 21, 13, 31, 23, 32")

Since ε33 = 0 for plane strain problems, we can compute the stress com-
ponent σ33 , using the 3rd equation in Eq. (5.60):

1 33 = E3/E1* 13* 11 + E3/E2* 23* 22

2 33

E3 ν23 σ22 E3 ν13 σ11

+
E2 E1
 Elasticity Tensor and Constitutive Equations 205

Using now the first equation in (5.60), we compute ε11 :

1 11 = 11/E1 - 22/E2* 21 - 33/E3* 31

2 [Link]( 11).simplify()

ν21 σ22 ν23 ν31 σ22 ν13 ν31 σ11 σ11

− − − +
E2 E2 E1 E1
Next, using the second equation in (5.60), compute ε22 :

1 22 = - 11/E1* 12 + 22/E2 - 33/E3* 32

2 [Link]().factor( 22)

E2 ν12 σ11 + E2 ν13 ν32 σ11 + σ22 (E1 ν23 ν32 − E1 )

−
E1 E2
We form the strain-stress matrix as
⎡ ⎤ ⎡ 1−ν13 ν31 ⎤⎡ ⎤
ε11 E1 − ν12 +ν
E1
13 ν32
0 σ11
⎢ ⎥ ⎢ ν +ν ν 1−ν23 ν32 ⎥⎢ ⎥
⎣ ε22 ⎦ = ⎣− 12 E113 32 E2 0 ⎦ ⎣σ22 ⎦ (5.68)
2ε12 1 σ12
0 0 G12
  
S

Stress σ33 is given by

E3 ν23 σ22 E3 ν13 σ11
σ33 = + = ν32 σ22 + ν31 σ11 (5.69)
E2 E1
Equation (5.68) can be inverted to obtain the stress-strain relation. We
again use Sympy to help us get this done. First, we form the 2 by 2 sub-
matrix in in the upper-left corner in Eq. (5.68):

1 S_2by2 = Matrix([[(1- 13* 31)/E1, -( 12+ 13* 32)/E1],

2 [-( 12+ 13* 32)/E1, (1- 23* 32)/E2 ]])
3 A = S_2by2.det()*E1*E1*E2 # a coefficient to simplify expression
4 A

2 2 2
E1 ν13 ν23 ν31 ν32 − E1 ν13 ν31 − E1 ν23 ν32 + E1 − E2 ν12 − 2E2 ν12 ν13 ν32 − E2 ν13 ν32

1 C_2by2 = S_2by2.inv()*A
2 C_2by2


−E12 ν23 ν32 + E12 E1 E2 ν12 + E1 E2 ν13 ν32
E1 E2 ν12 + E1 E2 ν13 ν32 −E1 E2 ν13 ν31 + E1 E2
206 Mechanics of Materials: Formulations and Solutions with Python

In reverse, Eq. (5.68) becomes

⎡ ⎤ ⎡ ⎤⎡ ⎤
σ11 −E12 ν23 ν32 + E12 E1 E2 ν12 + E1 E2 ν13 ν32 0 ε11
⎣σ22 ⎦ = 1 ⎢
⎣E1 E2 ν12 + E1 E2 ν13 ν32 −E1 E2 ν13 ν31 + E1 E2
⎥
0 ⎦ ⎣ ε22 ⎦
A
σ12 0 0 G12 A 2ε12
  
C

(5.70)

where
2
A = E1 ν13 ν23 ν31 ν32 − E1 ν13 ν31 − E1 ν23 ν32 + E1 − E2 ν12
2 2
− 2E2 ν12 ν13 ν32 − E2 ν13 ν32 (5.71)

For isotropic materials, we simply set Ei = E, νij = ν, and Gij = G. It is

trivial but a little lengthy to confirm that such a setting leads to Eq. (5.49).

5.15.6 Python code for elasticity matrix for orthotropic

materials (2D plane strain)
The following function forms a 3 × 3 compliance matrix using Young’s mod-
uli, Poisson’s ratios, and shear modulus, and then computes the 3×3 stiffness
matrix via inversion. This code is for 2D plane strain problems.

1 def E_SnC2D_strain(E1, E2, E3, 12, 13, 23, G12):

2 '''Compute the S and C matrix in Voigt notation for given Young's
3 moduli Ei, Poisson's ratios ij, and shear modulus G12 of
4 orthotropic materials for 2D plan strain problems.
5 '''
6 S = [Link]((3,3)) #initialization
7 21 = 12/E1*E2
8 31 = 13/E1*E3
9 32 = 23/E2*E3
10
11 # compute the compliance matrix S
12 S[0,0] = (1. - 13* 31)/E1
13 S[0,1] =-( 12+ 13* 32)/E1
14 S[1,1] = (1. - 23* 32)/E2
15 S[1,0] = S[0,1]
16 S[2,2] = 1./G12
17
18 # compute C matrix
19 C = [Link](S)
20
21 return C, S
 Elasticity Tensor and Constitutive Equations 207

5.15.7 Compliance matrix for isotropic materials

For isotropic materials, we have a much simpler form.
⎡1 −ν −ν
⎤
E E E 0 0 0
⎢ ⎥
⎢ 1 −ν
0 0 0⎥
⎢ E E ⎥
⎢ ⎥
⎢ 1
0 0 0 ⎥
⎢ E ⎥
S = [Sαβ ] = ⎢ ⎥ (5.72)
⎢ 1
0 0 ⎥
⎢ G ⎥
⎢ ⎥
⎢ sym. 1
0 ⎥
⎣ G ⎦
1
G

Using Eqs. (5.61) and (5.72), the constitutive equation in terms of strain
components can be written in a familiar form of
1
ε11 = [σ11 − νσ22 − νσ33 ]
E
1
ε22 = [σ22 − νσ11 − νσ33 ]
E
1
ε33 = [σ33 − νσ11 − νσ22 ]
E
(5.73)
1 1+ν
ε23 = σ23 = σ23
2G E
1 1+ν
ε13 = σ13 = σ13
2G E
1 1+ν
ε12 = σ12 = σ12
2G E

5.15.8 Derivation of the C matrix for isotropic

materials
Using Eq. (5.72), we are now ready to derive the C matrix for isotropic mate-
rials defined in Eq. (5.32), assuming Young’s modulus E, shear modulus G,
and Poisson’s ratio ν are given. We shall use the code function E SnC3Dsp(),
and set Ei = E, Gij = G, and νij = ν because of the isotropy.

1 E, , G = symbols('E, , G', positive=True)

2 C, S = E_SnC3Dsp(E, E, E, , , , G, G, G)
3 C
208 Mechanics of Materials: Formulations and Solutions with Python

⎡ Eν−E ⎤
2ν 2 +ν−1
− 2ν 2Eν
+ν−1
− 2ν 2Eν
+ν−1
0 0 0
⎢ ⎥
⎢− 2ν 2Eν Eν−E
− 2ν 2Eν 0 0⎥
0
⎢ +ν−1 2ν 2 +ν−1 +ν−1 ⎥
⎢ Eν ⎥
⎢− 2ν 2 +ν−1 − 2ν 2Eν Eν−E
0 0 0⎥
⎢ +ν−1 2ν 2 +ν−1 ⎥
⎢ ⎥
⎢ 0 0 0 G 0 0⎥
⎢ ⎥
⎢ ⎥
⎣ 0 0 0 0 G 0⎦
0 0 0 0 0 G

This matrix gives the relations in Eq. (5.33).

5.15.9 Compliance equations for 2D problems

For 2D plane stress problems, the nonzero stresses stay, say in the x1 –x2
plane, we shall have σ33 = σ23 = σ13 = 0. In this case, Eq. (5.73) becomes

1
ε11 = (σ11 − νσ22 )
E
1
ε22 = (σ22 − νσ11 ) (5.74)
E
1 1+ν
ε12 = σ12 = σ12
2G E

The strain ε33 is not zero, and is given by

ν
ε33 = − [σ11 + σ22 ] (5.75)
E

after σ11 and σ22 are found.

For 2D plane strain problems, the nonzero strains stay, say in the x1 –x2
plane, we shall have ε33 = ε23 = ε13 = 0. In this case, the constitutive
equation for strain components can be written in a familiar form of

1
ε11 = [(1 − ν 2 )σ11 − ν(1 + ν)σ22 ]
E
1
ε22 = [(1 − ν 2 )σ22 − ν(1 + ν)σ11 ] (5.76)
E
1 1+ν
ε12 = σ12 = σ12
2G E
 Elasticity Tensor and Constitutive Equations 209

Stress σ33 is given by

σ33 = νσ11 + νσ22 (5.77)
which is the same as Eq. (5.69).
We can obtain Eq. (5.76) using the substitution defined in Eqs. (5.50)
to (5.74) (plane stress to plane strain). In reverse, we can obtain Eq. (5.74)
using the following substitutions in Eq. (5.76) (plane strain to plane stress).
E
first: substitute by E
(1 − ν 2 )
ν
then: substitute ν by (5.78)
(1 + ν)
No change in G

5.15.10 Example: Proof of relation between G, E and ν

Consider a small thin square cell of solid that undergoes pure shearing, as
shown in Fig. 5.3. Since it is thin, and no force is applied in the z-direction,
all the stress components σzz = σzy = σzx = 0. It is thus a plane stress
problem.
For the pure shear stress state given in Fig. 5.3, we have σxx = σyy =
0, σxy = τ . Perform a coordinate transformation from x−y to X−Y with
θ = 45◦ , using the double angle formula (derived in Chapter 3):
1 1
σXX = (σxx + σyy ) + (σxx − σyy ) cos 2θ + σxy sin 2θ = τ
2 2
1 1
σY Y = (σxx + σyy ) − (σxx − σyy ) cos 2θ − σxy sin 2θ = −τ (5.79)
2 2
1
σXY = − (σxx − σyy ) sin 2θ + σxy cos 2θ = 0
2

Figure 5.3. Schematic view of a small square solid subjected to pure shearing.
210 Mechanics of Materials: Formulations and Solutions with Python

Figure 5.4. Stress state after 45-degree coordinate rotation from the pure stress state.

The transformed stress state of σXX = τ, σY Y = −τ, σXY = 0 is shown

in Fig. 5.4.
The code used to find this is given below:

1 xx, yy, xy = symbols(" _xx, _yy, _xy")

2 E, , , = symbols("E, , , ")
3
4 dic = { xx:0, yy:0, xy: , :[Link]/4} # = 45∘
5 XX, YY, XY = gr.tensorT2angle( xx, yy, xy, , py='sympy')
6
7 XX_= [Link](dic)
8 YY_= [Link](dic)
9 XY_= [Link](dic)
10 print(f" XX={ XX_}, YY={ YY_}, XY={ XY_}") # XX= , YY=- , XY=0

XX= , YY=- , XY=0

Using Eq. (5.74), the strain component in the X−Y coordinates can be
found as
ν+1
εXX = τ (5.80)
E

1 XX = ( XX_ - * YY_)/E # use strain-stress relation

2 print(f" XX={ XX}")

XX=( * + )/E
 Elasticity Tensor and Constitutive Equations 211

the length EC (see Fig. 5.3) becomes |EC| =

Considering small strains, √
εXX |AE|, where |AE| = |AB| 2, which gives

√
|EC| = εXX |AB| 2 (5.81)

√
On the other hand, |F C| = |EC| 2, and the shear strain becomes

|F C|
γ= = 2εXX (5.82)
|AB|

With the help of Eq. (5.80), we then found

τ τ E
G := = = (5.83)
γ 2εXX 2(ν + 1)

or

E
G= (5.84)
2(ν + 1)

This is the last equation given in Eq. (5.33). It is a widely used formula
that relates E, G and ν.
The code to get this done is given as follows:

1 AB = symbols("AB")
2 AE = [Link](2)*AB
3 EC = XX*AE
4 FC = EC*[Link](2)
5 = FC/AB
6 G = ( / ).simplify()
7 [Link](G, 'Shear modulus G=')

Shear modulus G = E
2(ν+1)

Equations given in this chapter are not difficult to derive, as long as the
concepts are clear. Some of the equations are quite lengthy in derivation. In
this chapter, we tried to provide various forms of equations for convenience
in use.
212 Mechanics of Materials: Formulations and Solutions with Python

5.16 A comprehensive example: Displacement, strain

and stress analysis

With relations between displacements-strain and strain-stress established,

we are now ready to look at a comprehensive example created in reference
to an exercise problem given in Ref. [1].

5.16.1 Setting of the problem

Consider 2D square solid cell with dimension of 1 cm × 1 cm defined in x−y
coordinates. It is formed by four vertices: A(0, 1), B(1, 1), C(1, 0), O (0, 0).
The displacement in the z-direction is zero: w = 0. The displacements in the
x–y plane are given in the following formulas.

u = c1 x + c2 y + c3 xy
(5.85)
v = c4 x + c5 y + c6 xy

where c1 , c2 , . . . , c6 are constants. Each of the displacement components is

given in a so-called bi-linear function in x and y because if one fixes x,
the function is linear in y, and vice versa. It is linear in two directions.
Through measurements, the displacements (u, v) at points A, B, and C
are found as, respectively, (0.004, 0.002), (0.002, 0.003), and (0.001, 0.001), all
in millimeters. Point O stays at the origin. %(.0, .00125), (−.00125, .00125),
and (.0025, .00125),
The data for material property are given as E = 200 GPa, ν = 0.3.
Our tasks are as follows:

1. Determine the functions for all the displacement components.

2. Compute the gradient of the displacement vector functions.
3. Compute the strain functions in the solid, and the values of the strains
at point B.
−−→
4. Compute the normal strains at point B along the OB direction, and that
−→
along AC direction.
5. Compute the shear strains at point B between BA and BC.
6. Compute the principal strains and strain invariants.
7. Rotate the coordinates by 30◦ about z-axis, and find the displacements
at point B in the new coordinates system.
8. Rotate the coordinates by 30◦ about z-axis, and find the strains at point
B in the new coordinates system.
 Elasticity Tensor and Constitutive Equations 213

9. Assume the square cell is in plane strain condition, compute the stress
functions in the solid, and the values of the strains at point B.
10. Compute the principal stresses, and stress invariants.
11. Rotate the coordinates by 30◦ about z-axis, and find the stresses at point
B in the new coordinates system.

5.16.2 Solution to the problem

[Link] Analysis of the problem

For a clearer intuitive understanding of the problem, we first plot the domain
of the problem: the original and the deformed square cell.

1 x, y = symbols("x, y ") # 2D coordinates

2 O = {x:0., y:0.} # the origin
3 dO = [Link]([ .0, .0])
4
5 A = {x:0., y:1.} # Coordinates at vertex A
6 dA = [Link]([.004, .002]) # displacement at A
7
8 B = {x:1., y:1.} # Coordinates at vertex B
9 dB = [Link]([.002, .003]) # displacement at B
10
11 C = {x:1., y:0.} # Coordinates at vertex C
12 dC = [Link]([.001, .001]) # displacement at C
13
14 # Create a list of the x- and y-coordinates of the vertices
15 xi = [A[x], B[x], C[x], O[x], A[x]]
16 yi = [A[y], B[y], C[y], O[y], A[y]]
17
18 [Link](xi, yi, marker='o') # Plot the undeformed square cell
19
20 k = 20. # to magnify the displacements to be visible.
21 xp = [A[x]+dA[0]*k,B[x]+dB[0]*k,C[x]+dC[0]*k,O[x]+dO[0]*k,A[x]+dA[0]*k]
22 yp = [A[y]+dA[1]*k,B[y]+dB[1]*k,C[y]+dC[1]*k,O[y]+dO[1]*k,A[y]+dA[1]*k]
23
24 [Link](xp, yp, linestyle='--', marker='*') # deformed quadrilateral
25
26 [Link](A[x], A[y], 'A', fontsize=12, ha='right')# Label the vertices
27 [Link](B[x], B[y], 'B', fontsize=12, ha='left')
28 [Link](C[x], C[y], 'C', fontsize=12, ha='left')
29 [Link](O[x], O[y], 'O', fontsize=12, ha='right')
30
31 [Link]('x-axis'); [Link]('y-axis')
32 [Link]('equal')
33 [Link]([-0.1, 1.1, -0.1, 1.1]) # Set the axis limits
34 [Link]()
214 Mechanics of Materials: Formulations and Solutions with Python

[Link] Find the displacement functions

1 # 1 Determine the functions for all the displacement components.

2 x, y = symbols("x, y ")
3 X = Matrix([x, y])
4
5 c1, c2, c3, c4, c5, c6 = symbols("c1, c2, c3, c4, c5, c6")
6
7 u = c1*x + c2*y + c3 *x*y
8 v = c4*x + c5*y + c6 *x*y
9
10 solution=[Link]([[Link](A)-dA[0],[Link](B)-dB[0],[Link](C)-dC[0],\
11 [Link](A)-dA[1],[Link](B)-dB[1],[Link](C)-dC[1]],\
12 [c1, c2, c3, c4, c5, c6])
13 solution

{c1 : 0.001, c2 : 0.004, c3 : −0.003, c4 : 0.001, c5 : 0.002, c6 : 0.0}

1 # The displacement functions are found as:

2 u = [Link](solution)
3 v = [Link](solution)
4
5 U = Matrix([u, v])
6 [Link](U.T, '\nThe displacement functions are found as:')

The displacement functions are found as

 
−0.003xy + 0.001x + 0.004y 0.001x + 0.002y
 Elasticity Tensor and Constitutive Equations 215

[Link] Compute the gradient of the displacement

1 # 2: Compute the gradient of the displacement vector functions.

2
3 U3D = Matrix([u, v, 0]) # set 0 for w, to create a 3D displacements
4
5 # Find the gradient of the displacement vector functions
6 U_gradient = gr.grad_vf(U3D[:2,:2], X) # get 2D displacement gradient
7 # or use: vector.grad_vec(U3D)[:2,:2]
8 printM(U_gradient,'Gradient of displacement vector functions',n_dgt=4)

Gradient of displacement vector functions


0.001 − 0.003y 0.001
0.004 − 0.003x 0.002

Since the displacements are bi-linear functions in the entire cell, the gra-
dient contains linear functions. This means that the strain field will be linear.

[Link] Find the strain tensor functions

1 #3 Compute the strain functions in the solid, and strain values at B.

2
3 strains = 0.5*(U_gradient +U_gradient.T)
4 printM(strains,'Strain tensor', n_dgt=4)

Strain tensor


0.001 − 0.003y 0.0025 − 0.0015x
0.0025 − 0.0015x 0.002

The strains are indeed linear functions.

1 Bv = [Link](B) # B: coordinates at point B

2 printM( Bv, 'Strain tensor values at point B:', n_dgt=5)

Strain tensor values at point B:


−0.002 0.001
0.001 0.002
216 Mechanics of Materials: Formulations and Solutions with Python

[Link] Compute the normal strains

1 # 4 Compute the normal strains at point B along OB, and AC.

2 B_M = [Link]( Bv, dtype=float) # convert to numpy array
3
4 # For OB, find the unit vector along OB
5 vector_B = [Link]([B[x],B[y]],dtype=float) # vector of OB
6 unit_OB = vector_B/[Link](vector_B) # Normal vector OB
7
8 NS_OB = unit_OB@ B_M@unit_OB # Normal strain of fiber O->B
9 print(f'Unit vector along OB = {unit_OB}')
10 print(f'Normal strain on vector OB = {NS_OB}')
11
12 # Similar process for AC, use Eq.(2.28)
13 unit_AC = [Link]([C[x]-A[x],C[y]-A[y]],dtype=float)
14 unit_AC = unit_AC/[Link]((C[x]-A[x])**2+(C[y]-A[y])**2)
15 print(f'Unit vector along AC = {unit_AC}')
16
17 NS_AC = unit_AC@ B_M@unit_AC
18 print(f'Normal strain on vector AC = {NS_AC:.5f}')

Unit vector along OB = [ 7.071e-01 7.071e-01]

Normal strain on vector OB = 0.001
Unit vector along AC = [ 7.071e-01 -7.071e-01]
Normal strain on vector AC = -0.00100

[Link] Compute the shear strains

1 # 5 Compute the shear strains at point B between BA and BC.

2 # vector BA:
3 unit_BA = [Link]([A[x]-B[x], A[y]-B[y]]) # use Eq.(2.28)
4 unit_BA = unit_BA/[Link]((A[x]-B[x])**2+(A[y]-B[y])**2)
5 print(f'The unit vector along vector BA = {unit_BA}')
6
7 NS_BA = unit_BA@ B_M@unit_BA # Normal strain of fiber BA
8 print(f'Normal strain along vector BA = {NS_BA:.5f}')
9
10 # vector BC xB-xA, yB-yA, zB-zA,
11 unit_BC = [Link]([C[x]-B[x], C[y]-B[y]]) # fiber along BC
12 unit_BC = unit_BC/[Link]((C[x]-B[x])**2+(C[y]-B[y])**2)
13 print(f'The unit vector BC = {unit_BC}')
14
15 E_BA_BC = unit_BA@ B_M@unit_BC # Shear strain between fiber BA and BC
16 print(f'Shear strain between BA and BC = {E_BA_BC:.5f}')

The unit vector along vector BA = [-1.000e+00 0.000e+00]

Normal strain along vector BA = -0.00200
The unit vector BC = [ 0.000e+00 -1.000e+00]
Shear strain between BA and BC = 0.00100
 Elasticity Tensor and Constitutive Equations 217

[Link] Compute the principal strains and strain invariants

1 # 6 Compute the principal strains, and strain invariants.

2 B_3D = [Link]((3,3)) # create a 3D tensor with zeros
3 B_3D[:2,:2] = B_M # create a 3D tensor
4 print( B_3D)
5 eigenValues, eigenVectors = [Link]( B_3D)
6
7 I1, I2, I3 = gr.M_invariant(eigenValues)
8 print(f'Principal strains (Eigenvalues) ={eigenValues}')
9 print(f'The first strain invariant I1={I1:.6e}')
10 print(f'The second strain invariant I2={I2:.6e}')
11 print(f'The third strain invariant I3={I3:.6e}, discard.')

[[-2.000e-03 1.000e-03 0.000e+00]

[ 1.000e-03 2.000e-03 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00]]
Principal stress/strain (Eigenvalues):
[ 2.236e-03 0.000e+00 -2.236e-03]

Principal stress/strain directions:

[[-2.298e-01 0.000e+00 -9.732e-01]
[-9.732e-01 0.000e+00 2.298e-01]
[ 0.000e+00 1.000e+00 0.000e+00]]

Possible angles (n1,x)=103.28252558853899∘ or 76.71747441146101∘

Principal strains (Eigenvalues) =[ 2.236e-03 0.000e+00 -2.236e-03]
The first strain invariant I1=0.000000e+00
The second strain invariant I2=-5.000000e-06
The third strain invariant I3=-0.000000e+00, discard.

One may complete this task using gr.M eigen2D(εB M), and
gr.M invariant2D(eigenValues).
We obtained 2 possible angles. To determine which one, we simply use:

1 #Tr, about = [Link](103.2825, about = 'z')

2 Tr, about = [Link](76.71747441146101, about = 'z')
3 p_3D = Tr@ [email protected]
4
5 print(f'Transformed strains: \n{ p_3D}')

Transformed strains:
[[ 2.236e-03 -6.728e-19 0.000e+00]
[-6.246e-19 -2.236e-03 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00]]
218 Mechanics of Materials: Formulations and Solutions with Python

Thus, 76.71◦ is the correct one (to machine accuracy). Readers may test
the other one.

[Link] Displacements in the new coordinates system

1 #7 Rotate the coordinates by 30° about z-axis, and find the
2 # displacements at point B in the new coordinates system.
3
4 Tr, about = [Link](30, about = 'z')
5 print(Tr)
6 Tu = Tr[:2,:2]@dB
7 print(f'Transformed displacements, Tu: {Tu}')

[[ 8.660e-01 5.000e-01 0.000e+00]

[-5.000e-01 8.660e-01 0.000e+00]
[ 0.000e+00 0.000e+00 1.000e+00]]
Transformed displacements, Tu: [ 3.232e-03 1.598e-03]

[Link] Strains in the new coordinates system

1 # 8) Rotate the coordinates by 30° about z-axis, and find the strains
2 # at point B in the new coordinates system.
3 p_3D = Tr@ [email protected]
4
5 print(f'Transformed strains: \n{ p_3D}')

Transformed strains:
[[-1.340e-04 2.232e-03 0.000e+00]
[ 2.232e-03 1.340e-04 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00]]

[Link] Compute the stresses

Since displacement is in the z-direction, w, is set to zero, all the strain
components related to z are zero. Therefore, the problem is a plane strain
problem. Using Eq. (5.48) or (5.49), the stress tensor is found using the
following code.
 Elasticity Tensor and Constitutive Equations 219

1 # 2D plane strain problem:

2
3 E_ = 200e9; _=0.3
4 c11 = E_*(1- _)/(1+ _)/(1-2* _)
5 c12 = c11* _/(1- _)
6 c33 = E_/(1+ _)/2 # G
7
8 C = [Link]([[c11, c12, 0],
9 [c12, c11, 0],
10 [ 0, 0, c33]])
11
12 v = [Link]([ B_M[0,0], B_M[1,1],2* B_M[0,1]]) # strain vector
13 print(f"Strains at B in vector form = { v}")
14
15 v = C@ v # stress vector
16 print(f"Stresses at B in vector form = { v} Pa")
17
18 33 = c11* _*( B_M[0,0] + B_M[1,1])
19 print(f"Stress 33 at B = { 33:.3e} Pa")

Strains at B in vector form = [-2.000e-03 2.000e-03 2.000e-03]

Stresses at B in vector form = [-3.077e+08 3.077e+08 1.538e+08] Pa
Stress 33 at B = 0.000e+00 Pa

[Link] Compute the principal stresses, and stress invariants

1 _M = [Link]((3,3)) # create a 3D tensor with zeros

2 _M[0,0] = v[0] # create a 3D tensor
3 _M[1,1] = v[1]
4 _M[0,1] = v[2]
5 _M[1,0] = v[2]
6 _M[2,2] = 33
7 print( _M)
8 eigenValues, eigenVectors = gr.M_eigen( _M)
9
10 I1, I2, I3 = gr.M_invariant(eigenValues)
11 print(f'Principal stress (Eigenvalues) ={eigenValues}')
12 print(f'The first stress invariant I1={I1:.6e}')
13 print(f'The second stress invariant I2={I2:.6e}')
14 print(f'The third stress invariant I3={I3:.6e}')

[[-3.077e+08 1.538e+08 0.000e+00]

[ 1.538e+08 3.077e+08 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00]]
Principal stress (Eigenvalues) =[ 3.440e+08 0.000e+00 -3.440e+08]
The first stress invariant I1=5.960464e-08
The second stress invariant I2=-1.183432e+17
The third stress invariant I3=-0.000000e+00
220 Mechanics of Materials: Formulations and Solutions with Python

[Link] Rotate the coordinates by 30◦ about z-axis, and find the
stresses at point B in the new coordinates system

1 p_M = Tr@ [email protected] # transformed stress tensor

2
3 print(f'Transformed stress tensor: \n{ p_M}')

Transformed stress tensor:

[[-2.061e+07 3.434e+08 0.000e+00]
[ 3.434e+08 2.061e+07 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00]]

This example demonstrated that when the displacement functions are

found, it is straightforward to obtain strains, and then stresses. Various
states of stresses and strains, including the principal stresses and strains,
can also be obtained with ease. Therefore, in most applications we often
would like to find the displacements for given externally applied forces. This
would need other important equations: equilibrium equations, which will be
discussed in the next chapter.

5.17 Thermal expansion effects

5.17.1 Thermal strains

Structural components may work in environments where the temperature
changes. Since most of the materials expand (or contract) when temperature
changes, additional strains are introduced inside the materials.
If the structure component is not overconstrained, the thermal strain just
results in (often small) dimension change. It will not introduce additional
stress. In such cases, it is often less a problem in terms of stress induced
failure.
If, however, the structure component is overconstrained (indeterminate),
the material is not allowed to expand freely. The thermal strain can result
in additional stress, known as thermal stress. Even if the thermal strain is
small, the resulting stress can be significant, which can be part of the cause
to failure. This section addresses such a problem for bar members that are
indeterminate.
We assume the temperature change is small and hence the thermal expan-
sion is small and linear in temperature change. It is quantified by αΔT ,
where α is the thermal expansion coefficient of the material carrying a unit
typically of [(m/m)/◦ C], which is strain (unit 1) per degree. And ΔT is the
change in the temperature that carries a unit typically [◦ C]. Therefore, αΔT
gives a strain. This strain is simply added to the strain resulting from the
derivatives of the displacement.
 Elasticity Tensor and Constitutive Equations 221

For orthotropic materials, the thermal expansion coefficient may differ in

coordinate directions. The three normal strains resulting from the thermal
expansions are given as follows:
εT 11 = α1 ΔT ; εT 22 = α2 ΔT ; εT 33 = α3 ΔT (5.86)
where αi is the thermal expansion coefficient of the material in the ith coor-
dinate direction. No shear strain will be generated from the thermal expan-
sions.
For isotropic materials, αi is set as α.

5.17.2 Strain-stress-temperature relation

The constitutive equation Eq. (5.60) is rewritten as
⎡ 1 ⎤
⎡ ⎤ −ν21 −ν31
0 0 0 ⎡ ⎤ ⎡ ⎤
ε11 E 1
⎢ −ν12
E 2 E 3
⎥ σ11 α1 ΔT
⎢ε ⎥ ⎢ E 1 −ν32
0 0 0 ⎥ ⎢σ ⎥ ⎢α ΔT ⎥
⎢ 22 ⎥ ⎢ 1 E2 E3 ⎥ ⎢ 22 ⎥ ⎢ 2 ⎥
⎢ ⎥ ⎢ −ν13 −ν23 1 ⎥⎢ ⎥ ⎢ ⎥
⎢ ε33 ⎥ ⎢ E1 E2 E3 0 0 0 ⎥ ⎢σ33 ⎥ ⎢α3 ΔT ⎥
⎢ ⎥=⎢ ⎥⎢ ⎥+⎢ ⎥ (5.87)
⎢2ε ⎥ ⎢ 0 1 ⎥ ⎢σ ⎥ ⎢ 0 ⎥
⎢ 23 ⎥ ⎢ 0 0 G23 0 0 ⎥⎢ ⎥ ⎢23 ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥
⎣2ε13 ⎦ ⎢ 0 0 0 0 1
0 ⎥ ⎣ σ 13 ⎦ ⎣ 0 ⎦
⎣ G13 ⎦
2ε12 σ12 0
   0 0 0 0 0 G112      

ε
   σ  εT
S
Or
ε = S
σ + εT (5.88)

5.17.3 Stress-strain-temperature relation

Stress-strain-temperature relation can be rewritten as

 = C [
σ ε − εT ] (5.89)
where C = S−1 .
The same concept and similar formulas of constitutive relations apply for
all other cases discussed in previous section in this chapter. Hence, we will
not repeat here.

5.18 Examples for elasticity tensor transformation*

5.18.1 Examples on arbitrary symmetric 6 by 6 matrix

First, we conduct tests on functions C2toC4() and C4toC2(C40) created
earlier using an arbitrary symmetric 6 by 6 matrix. We create such a matrix:
222 Mechanics of Materials: Formulations and Solutions with Python

1 # Produce a simple symmetric matrix for code testing purposes.

2 C_test = [Link]((6,6))
3
4 for i in range(6):
5 for j in range(6):
6 C_test[i, j] = (i+1)*(j+1)
7
8 print(C_test, C_test.shape)

[[ 1.000e+00 2.000e+00 3.000e+00 4.000e+00 5.000e+00 6.000e+00]

[ 2.000e+00 4.000e+00 6.000e+00 8.000e+00 1.000e+01 1.200e+01]
[ 3.000e+00 6.000e+00 9.000e+00 1.200e+01 1.500e+01 1.800e+01]
[ 4.000e+00 8.000e+00 1.200e+01 1.600e+01 2.000e+01 2.400e+01]
[ 5.000e+00 1.000e+01 1.500e+01 2.000e+01 2.500e+01 3.000e+01]
[ 6.000e+00 1.200e+01 1.800e+01 2.400e+01 3.000e+01 3.600e+01]] (6, 6)

1 # Test on reproducibility of conversion C(6,6) <-> C(3,3,3,3)

2 # for arbitrary symmetric matrix.
3 print(C_test,"\n") # Original C(6,6) matrix
4
5 C40 = C2toC4(C_test) # Convert C(6,6) to C(3,3,3,3)
6 C20 = C4toC2(C40) # Convert C(3,3,3,3) to C(6,6)
7
8 print(C20) # Back-converted C(6,6) matrix
9 print(f"Is close? {[Link]([Link](C20),[Link](C_test),l =1e-9)}")

[[ 1.000e+00 2.000e+00 3.000e+00 4.000e+00 5.000e+00 6.000e+00]

[ 2.000e+00 4.000e+00 6.000e+00 8.000e+00 1.000e+01 1.200e+01]
[ 3.000e+00 6.000e+00 9.000e+00 1.200e+01 1.500e+01 1.800e+01]
[ 4.000e+00 8.000e+00 1.200e+01 1.600e+01 2.000e+01 2.400e+01]
[ 5.000e+00 1.000e+01 1.500e+01 2.000e+01 2.500e+01 3.000e+01]
[ 6.000e+00 1.200e+01 1.800e+01 2.400e+01 3.000e+01 3.600e+01]]

[[ 1.000e+00 2.000e+00 3.000e+00 4.000e+00 5.000e+00 6.000e+00]

[ 2.000e+00 4.000e+00 6.000e+00 8.000e+00 1.000e+01 1.200e+01]
[ 3.000e+00 6.000e+00 9.000e+00 1.200e+01 1.500e+01 1.800e+01]
[ 4.000e+00 8.000e+00 1.200e+01 1.600e+01 2.000e+01 2.400e+01]
[ 5.000e+00 1.000e+01 1.500e+01 2.000e+01 2.500e+01 3.000e+01]
[ 6.000e+00 1.200e+01 1.800e+01 2.400e+01 3.000e+01 3.600e+01]]
Is close? True

The conversion forth and back worked fine.

Next, test C2toC4() and C4toC2() using a random transformation
matrix. The testing steps are given in the comments to the code line.
 Elasticity Tensor and Constitutive Equations 223

1 # Test C2toC4() and C4toC2()

2 T, _ = [Link](60., about = 'random') # Transformation Matrix
3
4 C0 = C2toC4(C_test) # Convert C(6,6) to C(3,3,3,3)
5 Cp = Tensor4_transfer(T, C0) # Perform transformation
6
7 print(f"Is close? {[Link]([Link](C0),[Link](Cp),atol=1e-4)}")

Is close? True

The fourth-order tensor transformation did not change the norm. This is
expected because of the unitary property of the transformation matrix.

1 C20 = C4toC2(Cp) # Convert C(3,3,3,3) to C(6,6)

2 print(C20)
3 print(f"Close? {[Link]([Link](C20),[Link](C_test),atol=1e-4)}")

[[ 1.699e-03 2.492e-01 -3.604e-03 2.809e-01 6.418e-02 1.692e-01]

[ 2.492e-01 3.656e+01 -5.286e-01 4.120e+01 9.414e+00 2.482e+01]
[-3.604e-03 -5.286e-01 7.644e-03 -5.958e-01 -1.361e-01 -3.590e-01]
[ 2.809e-01 4.120e+01 -5.958e-01 4.644e+01 1.061e+01 2.798e+01]
[ 6.418e-02 9.414e+00 -1.361e-01 1.061e+01 2.424e+00 6.392e+00]
[ 1.692e-01 2.482e+01 -3.590e-01 2.798e+01 6.392e+00 1.686e+01]]
Close? False

The components in the matrix changed after the transformation, as they

should be. The norm does not change.

1 C40 = C2toC4(C20) # Convert C(6,6) to C(3,3,3,3)

2 Cp = Tensor4_transfer(T.T, C40) # Perform inverse transformation
3 C0 = C4toC2(Cp) #Convert C(3,3,3,3) to C(6,6) to print out
4
5 print(C0)
6 print(f"Close? {[Link]([Link](C0),[Link](C_test),atol=1e-4)}")

[[ 1.000e+00 2.000e+00 3.000e+00 4.000e+00 5.000e+00 6.000e+00]

[ 2.000e+00 4.000e+00 6.000e+00 8.000e+00 1.000e+01 1.200e+01]
[ 3.000e+00 6.000e+00 9.000e+00 1.200e+01 1.500e+01 1.800e+01]
[ 4.000e+00 8.000e+00 1.200e+01 1.600e+01 2.000e+01 2.400e+01]
[ 5.000e+00 1.000e+01 1.500e+01 2.000e+01 2.500e+01 3.000e+01]
[ 6.000e+00 1.200e+01 1.800e+01 2.400e+01 3.000e+01 3.600e+01]]
Close? True
Using the codes above, we have confirmed the following.

• The transformed stiffness matrix will change, for a given general 6 by

6 symmetric matrix.
• The fourth-order tensor transformation does not change the norm.
• The inverse-transformed stiffness matrix recovers the original one.
224 Mechanics of Materials: Formulations and Solutions with Python

5.18.2 Examples on elasticity matrices for steel

Steel can be treated as isotropic materials at macroscale. Let us compute its

C matrix using the function E SnC3D() defined earlier, and then compute
its compliance matrix.

1 E_steel, _steel=77.1e9,0.3333 # Young's modulus, GPa & Poisson's ratio

2 G_s = .5*E_steel/(1+ _steel) # shear modulus (GPa)
3 = 7600. # density (Kg/m^3) # not used here
4 print(f"Shear modulus = {G_s:2.4e}")
5
6 C_steel, S_steel = E_SnC3D(E_steel,E_steel,E_steel,\
7 _steel, _steel, _steel,G_s,G_s,G_s)
8 # (77.1e9,77.1e9,77.1e9,.3333,.3333,.3333,28.91e9,28.91e9,28.91e9)
9
10 S_steel_ = [Link](C_steel)
11
12 print(f"Matrix of elastic constants C {C_steel.shape}:\n{C_steel}")
13 print(f"Matrix of compliance constants S_{S_steel_.shape}:\n{S_steel_}")
14 print(f"Matrix of original S {S_steel.shape}:\n{S_steel}")

Shear modulus = 2.8913e+10

Matrix of elastic constants C (6, 6):
[[ 1.156e+11 5.781e+10 5.781e+10 0.000e+00 0.000e+00 0.000e+00]
[ 5.781e+10 1.156e+11 5.781e+10 0.000e+00 0.000e+00 0.000e+00]
[ 5.781e+10 5.781e+10 1.156e+11 0.000e+00 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 2.891e+10 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 2.891e+10 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00 2.891e+10]]
Matrix of compliance constants S_(6, 6):
[[ 1.297e-11 -4.323e-12 -4.323e-12 0.000e+00 0.000e+00 0.000e+00]
[-4.323e-12 1.297e-11 -4.323e-12 0.000e+00 0.000e+00 0.000e+00]
[-4.323e-12 -4.323e-12 1.297e-11 0.000e+00 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 3.459e-11 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 3.459e-11 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00 3.459e-11]]
Matrix of original S (6, 6):
[[ 1.297e-11 -4.323e-12 -4.323e-12 0.000e+00 0.000e+00 0.000e+00]
[-4.323e-12 1.297e-11 -4.323e-12 0.000e+00 0.000e+00 0.000e+00]
[-4.323e-12 -4.323e-12 1.297e-11 0.000e+00 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 3.459e-11 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 3.459e-11 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00 3.459e-11]]

We prepare some transformation matrices for testing our codes for 4th
tensor transformation.
 Elasticity Tensor and Constitutive Equations 225

1 # Test on reproducibility of conversion C(6,6) <-> C(3,3,3,3)

2 # for real elastic material, using C2toC4() and C4toC2()
3
4 C4_steel = C2toC4(C_steel) #Convert C(6,6) to C(3,3,3,3)
5 C2_steel = C4toC2(C4_steel) #Convert C(3,3,3,3) to C(6,6)
6
7 print(C2_steel) #Back-converted C(6,6) matrix
8 print([Link](C2_steel) == [Link](C_steel))

[[ 1.156e+11 5.781e+10 5.781e+10 0.000e+00 0.000e+00 0.000e+00]

[ 5.781e+10 1.156e+11 5.781e+10 0.000e+00 0.000e+00 0.000e+00]
[ 5.781e+10 5.781e+10 1.156e+11 0.000e+00 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 2.891e+10 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 2.891e+10 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00 2.891e+10]]
True

The conversion forth and back worked correctly.

1 # Test C2toC4() and C4toC2() with a random transformation matrix

2 T, _ = [Link](60., about = 'random') #transformation matrix
3
4 C4_steel = C2toC4(C_steel) #Convert C(6,6) to C(3,3,3,3)
5 C4p_steel = Tensor4_transfer(T, C4_steel) # Perform transformation
6 print(f"Is close? "
7 f"{[Link]([Link](C4_steel),[Link](C4p_steel),atol=1e-4)}")
8
9 C2_steel = C4toC2(C4p_steel) #Convert C(3,3,3,3) to C(6,6)
10 print(f"Close? "
11 f"{[Link]([Link](C2_steel),[Link](C_steel),atol=1e-4)}")

Is close? True
Close? True

The fourth-order tensor transformation did not change the norm. Let us
check on forward and backward transformations.
1 # Test our 4th order tensor transformation, forward and backward:
2 print(C_steel,'\n') #Original C(6,6) matrix
3
4 C4_steel = C2toC4(C_steel) # Convert C(6,6) to C(3,3,3,3)
5 C4p_steel= Tensor4_transfer(T,C4_steel) # forward transformation
6 C2_steel = C4toC2(C4p_steel) # Convert C(3,3,3,3) to C(6,6)
7
8 print(C2_steel) #Transformed stiffness matrix
9 print(f"Is C_steel symmetric?"
10 f" {[Link](C2_steel,C2_steel.T,atol=1e-04)}\n")
11
12 C4p_steel= C2toC4(C2_steel) #Convert C(6,6) to C(3,3,3,3)
13 C4_steel = Tensor4_transfer(T.T,C4p_steel) # back-transformation
14 C2_steel = C4toC2(C4_steel) # Convert C(3,3,3,3) to C(6,6)
15
16 print(C2_steel) #Back-transformed stiffness matrix
17 print(f"Is C2 back to original? "
18 f"{[Link](C_steel,C2_steel,atol=1e-04)}")
226 Mechanics of Materials: Formulations and Solutions with Python

[[ 1.156e+11 5.781e+10 5.781e+10 0.000e+00 0.000e+00 0.000e+00]

[ 5.781e+10 1.156e+11 5.781e+10 0.000e+00 0.000e+00 0.000e+00]
[ 5.781e+10 5.781e+10 1.156e+11 0.000e+00 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 2.891e+10 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 2.891e+10 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00 2.891e+10]]

[[ 1.156e+11 5.781e+10 5.781e+10 -3.759e-06 -5.847e-06 -8.269e-05]

[ 5.781e+10 1.156e+11 5.781e+10 1.931e-06 -9.491e-06 -8.362e-05]
[ 5.781e+10 5.781e+10 1.156e+11 -1.174e-05 -3.947e-06 -3.801e-05]
[-6.211e-06 -4.498e-06 -6.426e-06 2.891e+10 -2.286e-05 -4.059e-06]
[-1.807e-06 -2.649e-06 -6.417e-06 -2.118e-05 2.891e+10 -2.093e-06]
[-7.874e-05 -7.977e-05 -4.143e-05 -2.603e-06 2.734e-06 2.891e+10]]
Is C_steel symmetric? True

[[ 1.156e+11 5.781e+10 5.781e+10 -4.150e-05 -8.103e-05 -1.110e-04]

[ 5.781e+10 1.156e+11 5.781e+10 -7.889e-05 -4.730e-05 -1.099e-04]
[ 5.781e+10 5.781e+10 1.156e+11 -7.497e-05 -8.225e-05 -5.216e-05]
[-4.130e-05 -7.285e-05 -7.307e-05 2.891e+10 -3.012e-05 -1.997e-05]
[-7.656e-05 -3.787e-05 -8.583e-05 -2.465e-05 2.891e+10 -1.898e-05]
[-1.224e-04 -1.161e-04 -5.348e-05 -2.686e-05 -2.175e-05 2.891e+10]]
Is C2 back to original? False

Using the codes above, we have confirmed the following.

• The transformed stiffness matrix will not change if the material is
isotropic.

5.18.3 Examples on elasticity matrices for steel for

2D plane stress
1 E_steel, _steel=77.1e9,0.3333 # Young's modulus,GPa & Poisson's ratio
2 G_s = .5*E_steel/(1+ _steel) # shear modulus (GPa)
3 print(f"Shear modulus = {G_s:2.4e}")
4
5 C_steel, S_steel = E_SnC2D_stress(E_steel, E_steel, _steel, G_s)
6 S_steel_ = [Link](C_steel)
7
8 print(f"Matrix of elastic constants C {C_steel.shape}:\n{C_steel}")
9 print(f"Matrix of compliance constants S_{S_steel_.shape}:\n{S_steel_}")
10 print(f"Matrix of original S {S_steel.shape}:\n{S_steel}")
 Elasticity Tensor and Constitutive Equations 227

Shear modulus = 2.8913e+10

Matrix of elastic constants C (3, 3):
[[ 8.674e+10 2.891e+10 0.000e+00]
[ 2.891e+10 8.674e+10 0.000e+00]
[ 0.000e+00 0.000e+00 2.891e+10]]
Matrix of compliance constants S_(3, 3):
[[ 1.297e-11 -4.323e-12 0.000e+00]
[-4.323e-12 1.297e-11 0.000e+00]
[ 0.000e+00 0.000e+00 3.459e-11]]
Matrix of original S (3, 3):
[[ 1.297e-11 -4.323e-12 0.000e+00]
[-4.323e-12 1.297e-11 0.000e+00]
[ 0.000e+00 0.000e+00 3.459e-11]]

5.18.4 Examples on elasticity matrices for steel for

2D plane strain
1 E_steel, _steel=77.1e9,0.3333 # Young's modulus,GPa & Poisson's ratio
2 G_s = .5*E_steel/(1+ _steel) # shear modulus (GPa)
3 print(f"Shear modulus = {G_s:2.4e}")
4
5 C_steel, S_steel = E_SnC2D_strain(E_steel, E_steel, E_steel,\
6 _steel, _steel, _steel, G_s)
7 S_steel_ = [Link](C_steel)
8
9 print(f"Matrix of elastic constants C {C_steel.shape}:\n{C_steel}")
10 print(f"Matrix of compliance constants S_{S_steel_.shape}:\n{S_steel_}")
11 print(f"Matrix of original S {S_steel.shape}:\n{S_steel}")

Shear modulus = 2.8913e+10

Matrix of elastic constants C (3, 3):
[[ 1.156e+11 5.781e+10 0.000e+00]
[ 5.781e+10 1.156e+11 0.000e+00]
[ 0.000e+00 0.000e+00 2.891e+10]]
Matrix of compliance constants S_(3, 3):
[[ 1.153e-11 -5.764e-12 0.000e+00]
[-5.764e-12 1.153e-11 0.000e+00]
[ 0.000e+00 0.000e+00 3.459e-11]]
Matrix of original S (3, 3):
[[ 1.153e-11 -5.764e-12 0.000e+00]
[-5.764e-12 1.153e-11 0.000e+00]
[ 0.000e+00 0.000e+00 3.459e-11]]
228 Mechanics of Materials: Formulations and Solutions with Python

The compliance matrix obtained above can be used to double checked

against Eq. (5.76):

1 print(f's11={(1.- _steel**2)/E_steel}')
2 print(f's12={- _steel*(1.+ _steel)/E_steel}')
3 print(f's33={2*(1.+ _steel)/E_steel}')

s11=1.1529326977950713e-11
s12=-5.763798832684825e-12
s33=3.4586251621271074e-11

5.18.5 Examples on elasticity matrices for

carbon/epoxy composite**
Carbon/epoxy composite has a very strong anisotropy because of the
reinforcement with carbon fibers. Let us compute the C matrix using the
function E SnC3D() defined above.

1 #Carbon-Reinforced Plastic Composite:(E1,E2,E3,m12,m13,m23,G23,G13,G12)

2 C_CRP, S_CRP = E_SnC3D(142.17e9,9.255e9,9.255e9,.334,.25,.25,\
3 4.795e9,4.795e9,4.795e9)
4 rho = 1900. # density (Kg/m^3)
5 print(f"Material stiffness matrix:\n{C_CRP}")
6 print(f"Material compliance matrix:\n{S_CRP}")

Material stiffness matrix:

[[ 1.443e+11 3.974e+09 3.342e+09 0.000e+00 0.000e+00 0.000e+00]
[ 3.974e+09 9.981e+09 2.560e+09 0.000e+00 0.000e+00 0.000e+00]
[ 3.342e+09 2.560e+09 9.949e+09 0.000e+00 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 4.795e+09 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 4.795e+09 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00 4.795e+09]]
Material compliance matrix:
[[ 7.034e-12 -2.349e-12 -1.758e-12 0.000e+00 0.000e+00 0.000e+00]
[-2.349e-12 1.080e-10 -2.701e-11 0.000e+00 0.000e+00 0.000e+00]
[-1.758e-12 -2.701e-11 1.080e-10 0.000e+00 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 2.086e-10 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 2.086e-10 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00 2.086e-10]]

1 # Test on reproducibility of conversion C(6,6) <-> C(3,3,3,3)

2 # for real elastic material
3
4 C4_CRP = C2toC4(C_CRP) # Convert C(6,6) to C(3,3,3,3)
5 C2_CRP = C4toC2(C4_CRP) # Convert C(3,3,3,3) to C(6,6)
6
7 print(C2_CRP) # Back-converted C(6,6) matrix
8 print([Link](C2_CRP) == [Link](C_CRP))
 Elasticity Tensor and Constitutive Equations 229

[[ 1.443e+11 3.974e+09 3.342e+09 0.000e+00 0.000e+00 0.000e+00]

[ 3.974e+09 9.981e+09 2.560e+09 0.000e+00 0.000e+00 0.000e+00]
[ 3.342e+09 2.560e+09 9.949e+09 0.000e+00 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 4.795e+09 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 4.795e+09 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00 4.795e+09]]
True

Conversion forth and back worked fine.

1 # Test C2toC4() and C4toC2()
2 T, _ = [Link](60., about = 'random')
3
4 C4_CRP = C2toC4(C_CRP) #Convert C(6,6) to C(3,3,3,3)
5 C4p_CRP = Tensor4_transfer(T, C4_CRP) #4th order tensor transformation
6 C2_CRP = C4toC2(C4p_CRP)
7
8 print(C2_CRP)
9 print(f"Is C2 symmetric? {[Link](C2_CRP,C2_CRP.T,atol=1e-04)}\n")
10 print(f"Is close? "
11 f"{[Link]([Link](C4_CRP),[Link](C4p_CRP),atol=1e-4)}")

[[ 4.748e+10 8.336e+09 3.033e+10 -1.221e+10 -3.160e+10 1.387e+10]

[ 8.336e+09 1.128e+10 7.002e+09 -2.512e+09 -5.005e+09 2.503e+09]
[ 3.033e+10 7.002e+09 3.392e+10 -9.660e+09 -2.514e+10 1.101e+10]
[-1.221e+10 -2.512e+09 -9.660e+09 8.846e+09 1.077e+10 -4.670e+09]
[-3.160e+10 -5.005e+09 -2.514e+10 1.077e+10 3.127e+10 -1.191e+10]
[ 1.387e+10 2.503e+09 1.101e+10 -4.670e+09 -1.191e+10 1.006e+10]]
Is C2 symmetric? True

Is close? True

The stiffness matrix is changed, but still symmetric, and the norm is
unchanged.

1 # Test our 4th order tensor transformation

2 print(f"Original C_CRP: \n{C_CRP}\n") #Original C(6,6) matrix
3
4 C4_CRP = Tensor4_transfer(T.T,C4p_CRP) #Inverse transformation
5 C2_CRP = C4toC2(C4_CRP) #Convert C(3,3,3,3) to C(6,6)
6
7 print(f"After transformation:\n{C2_CRP}\n") #Transformed matrix
8 print(f"Is C2 symmetric? {[Link](C2_CRP,C2_CRP.T,atol=1e-04)}\n")
9 print(f"Is C2 back to original? {[Link](C2_CRP,C_CRP,atol=1e-03)}")
230 Mechanics of Materials: Formulations and Solutions with Python

Original C_CRP:
[[ 1.443e+11 3.974e+09 3.342e+09 0.000e+00 0.000e+00 0.000e+00]
[ 3.974e+09 9.981e+09 2.560e+09 0.000e+00 0.000e+00 0.000e+00]
[ 3.342e+09 2.560e+09 9.949e+09 0.000e+00 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 4.795e+09 0.000e+00 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 4.795e+09 0.000e+00]
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00 4.795e+09]]

After transformation:
[[ 1.443e+11 3.974e+09 3.342e+09 2.085e-06 -8.959e-06 2.150e-05]
[ 3.974e+09 9.981e+09 2.560e+09 -6.372e-06 -4.398e-06 6.379e-06]
[ 3.342e+09 2.560e+09 9.949e+09 -3.480e-06 -3.051e-06 6.145e-07]
[-6.205e-07 -5.577e-06 -3.720e-06 4.795e+09 8.087e-07 -1.662e-07]
[-1.384e-05 -4.524e-07 -1.816e-06 1.084e-07 4.795e+09 -4.319e-06]
[ 2.431e-05 5.402e-07 2.402e-07 1.142e-06 -3.561e-07 4.795e+09]]

Is C2 symmetric? True

Is C2 back to original? True

We have confirmed the following:

• The transformed stiffness matrix will change if the material is anisotropic,

such as the CRP composite.
• The transformation does not change the norm.
• The back-transformed stiffness matrix recovers the original one.

5.19 Remarks

1. Constitutive equation or stress-strain equations are dealt with in this

chapter. The theoretically most general fourth-order tensor with 81 elastic
constants is first introduced. The symmetry of the tensor is proven. The
symmetry leads to 21 elastic constants for anisotropic solid materials. The
fourth-order elasticity tensor can then be represented by a 6 × 6 matrix,
which is much easier to use in engineering applications.
2. Procedure and codes are provided to reduce the fourth-order tensor form
to symmetric matrix form that is more often used in engineering applica-
tions.
3. For isotropic elastic materials, the material constants become 2, which
is most often encountered in engineering structures and devices. Major
forms of constitutive equations and codes are thus provided.
4. The concept of stiffness matrix and compliance matrix are introduced. For
stable solid materials (that do not flow), both matrices are symmetric
positive definite (SPD), and hence one can be inverted to produce the
other.
 Elasticity Tensor and Constitutive Equations 231

5. To obtain these material constants through experimental measurements,

we use often the compliance matrix that contains explicitly Young’s mod-
uli, shear moduli, and Poisson’s ratios.
6. Thermal effects can be taken into consideration via thermal strains in the
constitutive equations.
The following chapter discusses equilibrium equations.

References
[1] A.P. Boresi, R.J. Schmidt and O.M. Sidebottom, Advanced Mechanics of Materials.
John Wiley & Sons, New York, 1985.
[2] J.M. Gere and S.P. Timoshenko, Mechanics of Materials, Van Nostrand Reinhold
Company, New York, 1972.
[3] S. Timoshenko and J.N. Goodier, Theory of Elasticity. McGraw Hill, New York,
1970. [Link]
lasticity&#38;ei=ICiKSsr3G4jwkQSbxMyPCg.
[4] Z.L. Xu, Elasticity, Vols. 1&2. People’s Publisher, China, 1979.
[5] G.R. Liu and S.S. Quek, The Finite Element Method: A Practical Course.
Butterworth-Heinemann, New York, 2013.
[6] W. Michael, David Rubin and Erhard Krempl. The Elastic Solid [Chapter 5], 2010.
[7] Lev Davidovich Landau, Evegnii Mikhailovich Lifshitz, A.M. Kosevich et al., Theory
of Elasticity: Volume 7, 1986.
[8] G.R. Liu and Z.C. Xi, Elastic Waves in Anisotropic Laminates. CRC Press, New
York, 2001.
[9] T.T.C. Ting, Anisotropic Elasticity: Theory and Applications, 1996.
[10] A.P. Boresi, K. Chong and J.D. Lee, Elasticity in Engineering Mechanics. John
Wiley & Sons, New York, 2010.
[11] G.R. Liu and T.T. Nguyen, Smoothed Finite Element Methods. Taylor and Francis
Group, New York, 2010.
[12] G.R. Liu, Mesh Free Methods: Moving Beyond the Finite Element Method. Taylor
and Francis Group, New York, 2010.
This page intentionally left blank
 Chapter 6

Equilibrium Equations

1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.

2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 init_printing(use_unicode=True) # For latex-like quality printing
11 np.set_printoptions(precision=4,suppress=True) # Digits in print-outs

6.1 Introduction
Consider a three-dimensional (3D) solid material, occupying a physical
domain Ω bounded by surface S and defined in a 3D Cartesian coordinate.
The solid is subjected to body forces b over Ω. The displacements of parts
of the boundary Su are constrained, so that the solid cannot have rigid body
movements. The rest of the boundary St is subjected to tractions t that are
given or zero, and S = Su ∪ St , as shown in Fig. 6.1.
In the previous chapters, we discussed where the stresses are coming from,
how strains relate to displacements, and then how the stress and strain are
related. We need now one more important piece: how the stress and forces are
working together to satisfy the equilibrium conditions. This chapter provides
this piece. The conclusion is that the divergence of the stresses at any point
in solid material is compensated by the body force distributed in the solid,
leading to the equilibrium equation in the form of partial differential
equations (PDEs). It governs the behavior of the loaded solid material inside
its domain and is hence known as the governing equation.

233
234 Mechanics of Materials: Formulations and Solutions with Python

Figure 6.1. A 3D solid constrained on the displacement boundary Su . It is loaded by

body forces b over the domain and tractions t over St .

At the end of this chapter, the different types of boundary conditions

(BCs) will also be discussed. The BCs must be satisfied when solving the
governing equation for loaded solids.

6.2 Equilibrium equation in terms of stresses

Stresses in loaded solid materials will in general vary from location to loca-
tion. In a coordinate system, they will be functions of coordinates. Therefore,
a stress tensor becomes a tensor function. Such variations shall relate to the
equilibrium state at any point within the loaded solid. We examine such
variations, leading to one of the most important equations governing the
behavior of the solid: the equilibrium equation.

6.2.1 3D free-body diagram

Consider a solid material subjected to external forces, resulting in stresses
distributed over the solids, in a way detailed in Chapter 3. Assume that the
solid media is continuous and the stresses are sufficiently smooth, and hence
their functions are differentiable at least up to first order.
We first derive the equilibrium based on the laws of mechanics: The sum
of all the forces at any point in the solid must vanish. To do this, we isolate
 Equilibrium Equations 235

Figure 6.2. A representative infinitely small brick from a solid with dimensions of dx,
dy, and dz. It serves as a typical 3D free-body diagram. All stresses are shown on the all
the six surfaces of the free-body diagram, following the standard sign convention. Stresses
in the x-direction are highlighted in red. Since the stresses are functions, a change in
coordinates results in a change in the stresses.

a small representative cell from the solid, which is a 3D free-body diagram,

as shown in Fig. 6.2. The directions of all these surfaces of the brick are in
x-, y-, or z-directions.

6.2.2 The standard sign convention

The standard sign convention defined in this book is as follows:

1. Positive displacement components are in the positive axis direction.

2. Positive rotation or twist with respect to an axis is counterclockwise about
the axis.
3. All the externally applied forces point to the positive axis direction.
4. All the externally applied moments or torques with respect to an axis are
counterclockwise with respect to the axis.
5. Positive internal forces or stresses on a surface follow the direction of the
surface.
6. Positive internal moments or toques on a surface are with respect to the
direction of the surface.

The stress components drawn in Fig. 6.2 follow exactly this convention,
which will be used in all the chapters.
236 Mechanics of Materials: Formulations and Solutions with Python

6.2.3 Equilibrium equation in x-direction

Let us now consider only, as an example, all the stress components in the
x-direction. The sum of the forces resulting from these stress components
and the externally applied body force must vanish. We thus have
(σxx + dσxx )dydz − σxx dydz
+ (σyx + dσyx )dxdz − σyx dxdz (6.1)
+ (σzx + dσzx )dxdy − σzx dxdy + bx dxdydz = ρüx dxdydz
where bx is the x component of the external body-force vector b, which may
be a function of the coordinates and time t, ρ is the density of the solid
material, and üx is the x component of the acceleration due to b changing
with time, which is also a function of the coordinates.
Here, we used an overhead “ ˙ ” to represent the first derivative with
respect to time, and thus “¨” becomes the second time derivative.
Equation (6.1) can be simplified to
dσxx dydz + dσyx dxdz + dσzx dxdy + bx dxdydz = ρüx dxdydz (6.2)
Since the stress components are assumed differentiable, we have
∂σxx ∂σyx ∂σzx
dσxx = dx; dσyx = dy; dσzx = dz (6.3)
∂x ∂y ∂z
Substituting Eq. (6.3) into Eq. (6.2) leads to
∂σxx ∂σyx ∂σzx
dxdydz + dydxdz + dzdxdy + bx dxdydz = ρüx dxdydz (6.4)
∂x ∂y ∂z
which can be simplified (the volume dxdydz is nonzero) to
∂σxx ∂σyx ∂σzx
+ + + bx = ρüx (6.5)
∂x ∂y ∂z
This is the equation obtained by enforcing the equilibrium of all these
forces in the x-direction.
Note that all three stress components in Eq. (6.5) are the ones on three
different surfaces but in the same x-direction. Thus, we can form the follow-
ing vector:
⎡ ⎤
σxx
⎢ ⎥
σx = ⎣σyx ⎦ (6.6)
σzx
 Equilibrium Equations 237

These stress components are on, respectively, x, y, and z surfaces and can
∂σ
vary with the coordinates. The sum of their derivatives ∂σ∂xxx + ∂yyx + ∂σ∂zzx
(with respect to the coordinate variations) forms the so-called divergence
of the stress vector. It is the spatial rate of losses of all stresses in the
x-direction. It must be compensated by the external force (density) in the
same x-direction in order to satisfy the equilibrium, which is Eq. (6.5). Note
that we treat the inertial force as an externally applied force but with a
negative sign.

6.2.4 Equilibrium in y- and z-directions

The same can be done for the forces in the y- and z-directions, which gives

∂σxy ∂σyy ∂σzy

+ + + by = ρüy
∂x ∂y ∂z
(6.7)
∂σxz ∂σyz ∂σzz
+ + + bz = ρüz
∂x ∂y ∂z

Equations (6.5) and (6.7) are the set of three equilibrium equations of 3D
solids subjected to external body forces bi , i = x, y, z, each of which can be
a function of the coordinates.

6.2.5 Equilibrium equation in indicial notation

In indicial notation, this set of equilibrium equations can be written
simply as

σji,j + bi = ρüi ∈ Ω (6.8)

This is the governing equation at any point in a loaded solid. Note that the
contraction is on the first index of the stress tensor (standing for surfaces).
The unknowns in the governing equation are these six stress components.
We assume that the application of all the forces is sufficiently slow and
hence has no dynamic effects. For such a static problem, we simply drop
the inertia term on the right-hand side. Equation (6.8) becomes

σji,j + bi = 0 ∈ Ω (6.9)

In this book, our focus is on static problems.

238 Mechanics of Materials: Formulations and Solutions with Python

6.2.6 Equilibrium equation in 2D and 1D

For 2D problems in the x–y plane, for example, the equilibrium equation
can be obtained by simply removing all terms related to z, which gives

∂σxx ∂σyx
+ + bx = ρüx
∂x ∂y
(6.10)
∂σxy ∂σyy
+ + by = ρüy
∂x ∂y

Equation (6.9) still holds for 2D problems, but the range of i and j is
from 1 to 2.
For 1D problems, the equilibrium equation is simply

∂σxx
+ bx = ρü (6.11)
∂x

which is the equation we derived at the beginning of Chapter 3 (when omit-

ting the inertia term). For 1D cases, we may drop the indexes when all
variables are treated as scalars.

6.2.7 Equilibrium equation via divergence of stress

tensor function
In the previous section, we found that the divergence of the vector function
becomes a single function of coordinates. Expanding this finding, we may
expect that a divergence of a tensor function produces a vector function.
We also found that a stress vector divergence is directly related to the body
force. We shall expect that the stress tensor divergence relates to the body
force in all directions, leading to the set of equilibrium equations, but in a
pure mathematical way.
Consider a stress tensor that can be represented by a 3 × 3 matrix in
space R3 . We first define the stress tensor symbolically as follows:

1 Sxx, Sxy, Sxz = symbols("\u03C3_xx \u03C3_xy \u03C3_xz", cls=Function)

2 Syx, Syy, Syz = symbols("\u03C3_yx \u03C3_yy \u03C3_yz", cls=Function)
3 Szx, Szy, Szz = symbols("\u03C3_zx \u03C3_zy \u03C3_zz", cls=Function)

Since each of the stress components could be a function of the coordinates,

we now define a tensor function as a matrix of functions:
 Equilibrium Equations 239

1 x, y, z = symbols("x, y, z")
2 X = Matrix([x, y, z])
3
4 Sf = Matrix([[Sxx(x, y, z), Sxy(x, y, z), Sxz(x, y, z)],
5 [Syx(x, y, z), Syy(x, y, z), Syz(x, y, z)],
6 [Szx(x, y, z), Szy(x, y, z), Szz(x, y, z)]])
7 Sf # a tensor of 9 functions: stress tensor function

⎡ ⎤
σxx (x, y, z) σxy (x, y, z) σxz (x, y, z)
⎢ ⎥
⎣σyx (x, y, z) σyy (x, y, z) σyz (x, y, z) ⎦
σzx (x, y, z) σzy (x, y, z) σzz (x, y, z)
Note that a column in the stress matrix is a vector of stresses in a direc-
tion. Let us now take the divergence of the tensor function:

1 def div_tensorF(σ, X):

2 '''Compute the divergence of a tensor function σ.
3 input: σ, tensor function in [Link]
4 return: the divergence of a tensor function
5 '''
6 n = σ.shape[0]
7 div_tf = [([Link](*[[Link](σ[i,j], X[i]) for i in range(n)]))\
8 for j in range(n)]
9
10 return [Link](div_tf) # divergence of a tensor function
11

1 div_Sf = div_tensorF(Sf, X) # Or use: vector.div_tensor(Sf.T)

2 div_Sf

⎡ ∂ ∂ ∂ ⎤
∂x σxx (x, y, z) + ∂y σyx (x, y, z) + ∂z σzx (x, y, z)
⎢ ⎥
⎢∂ ∂ ∂ ⎥
⎢ ∂x σxy (x, y, z) + ∂y σyy (x, y, z) + ∂z σzy (x, y, z) ⎥
⎣ ⎦
∂ ∂ ∂
∂x σxz (x, y, z) + ∂y σyz (x, y, z) + ∂z σzz (x, y, z)

1 div_Sf.shape # it should be a vector

(3, 1)
It now becomes a vector of length 3, as expected. Each component in this
vector is the divergence of the vector of stresses in one of the three directions.
This divergence is the spatial rate of loss of the stresses in the direction at any
point in the solid, which should be compensated by the body force applied
240 Mechanics of Materials: Formulations and Solutions with Python

at that point. Their sum shall vanish for the solid staying in equilibrium.
This is the equilibrium equation in terms of the stress components and given
body forces:
⎡∂ ∂ ∂ ⎤
∂x σxx (x, y, z) + ∂y σyx (x, y, z) + ∂z σzx (x, y, z)
⎡ ⎤ ⎡ ⎤
⎢ ⎥ bx üx
⎢∂ ∂ ∂ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ∂x σxy (x, y, z) + ∂y σyy (x, y, z) + ∂z σzy (x, y, z) ⎥ + ⎣by ⎦ = ρ ⎣üy ⎦ (6.12)
⎣ ⎦
∂ ∂ ∂ bz üz
∂x σxz (x, y, z) + ∂y σyz (x, y, z) + ∂z σzz (x, y, z)

which is the same as Eqs. (6.5) and (6.7). Note again that the inertia force
is treated as a body force (with a negative sign).

Remark. The divergence of the stress vector (formed by stress components

in one of the x-, y-, and z-directions) at any point in the solid, together with
the applied body force at that point, must vanish to ensure equilibrium. In
other words, the divergence of the stress tensor function together
with the body-force vector shall vanish for a solid staying in
equilibrium.

Finally, we mention that when working with other coordinate systems,

deriving the equilibrium equations using the technique given in Section 6.2
can be quite challenging and prone to error. The use of divergence of stress
tensor can be much easier in one code line, which can be done in future
volumes of the book series when working with other curvilinear coordinate
systems for more general high-dimensional solid mechanics problems.

6.3 Equilibrium equations in displacements

6.3.1 General formulation in 3D*

In Chapter 4, we derived a set of equations for small strains in terms of the
derivatives of the displacements. In indicial notation, it is given as
1
εij = (ui,j + uj,i ) (6.13)
2
In Chapter 5, we introduced the constitutive equation for elastic solids
that can be given in the most general form with a fourth-order elasticity
tensor:

σij = Cijkl εkl (6.14)

We shall now put all these together with Eq. (6.9) to derive the equi-
librium equations in terms of displacements that have only three unknown
 Equilibrium Equations 241

displacement functions instead of six unknowns in Eq. (6.9). First, substi-

tuting Eqs. (6.13) into (6.14) gives
1
σij = Cijkl (uk,l + ul,k ) (6.15)
2
Since Cijkl = Cijlk (due to the symmetry of the strain tensor discussed
in Chapter 5), the foregoing equation becomes
1 1
σij = Cijkl (uk,l + ul,k ) = [Cijkl uk,l + Cijkl ul,k ]
2 2
(6.16)
1 1
= [Cijkl uk,l + Cijlk uk,l ] = [Cijkl uk,l + Cijkl uk,l ]
2 2
or

σij = Cijkl uk,l (6.17)

We obtain the stresses in terms of the displacements. Next, substituting

Eq. (6.17) into Eq. (6.9), we arrive at

Cijkl uk,lj + bi = ρüi ∈ Ω (6.18)

For static problems, Eq. (6.18) becomes

Cijkl uk,lj + bi = 0 ∈ Ω (6.19)

This is the equilibrium equation purely in displacements. It is rather a

quite complicated set of PDEs because of the fourth-order tensor. It can be
reduced using the Viot notation for specific types of materials, as discussed
in detail in the previous chapter.
For isotropic materials, it was found that the fourth-order tensor has the
following form:

Cijkl = G(δik δjl + δil δjk ) + λδij δkl (6.20)

Substituting Eq. (6.20) into (6.17), we obtain stresses in terms of the

displacements:

σij = G(δik δjl uk,l + δil δjk uk,l ) + λδij δkl uk,l
= G(ui,j + uj,i ) + λδij uk,k (6.21)

Using Eq. (6.8), we obtain the equilibrium equation in terms of the dis-
placements:

Gui,jj + (G + λ)uj,ji + bi = ρüi (6.22)

242 Mechanics of Materials: Formulations and Solutions with Python

If we write Eq. (6.22) explicitly, it has the form of

 2  2
∂ u ∂2u ∂2u ∂ u ∂2v ∂2w
G + + + (G + λ) + + + bx = ρü
∂x2 ∂y 2 ∂z 2 ∂x2 ∂x∂y ∂x∂z
 2  2
∂ v ∂2v ∂2v ∂ u ∂2v ∂2w
G 2
+ 2
+ 2
+ (G + λ) + 2+ + by = ρv̈
∂x ∂y ∂z ∂x∂y ∂y ∂y∂z
 2 
∂ w ∂2w ∂2w ∂2u ∂2v ∂2w
G + + + (G + λ) + + + bz = ρẅ
∂x2 ∂y 2 ∂z 2 ∂x∂z ∂y∂z ∂z 2
(6.23)
This is truly a set of “monster” equations. Finding an analytic solution
is difficult, except for a very few drastically simplified problems.

6.3.2 Formulations of 2D problems

There is a class of problems in solid mechanics having special features which
allow us to simplify it from 3D to 2D. Such 2D problems are of two major
types: plane stress and plane strain.

[Link] Plane strain problems

When one of the displacement components, say w, is zero, we have a plane
strain problem. In this case, all the nonzero strains stay within the x−y
plane, which are called in-plane strains. In this case, all the variables are
independent of z. A typical case is a dam built to block water flow, as shown
in Fig. 6.3.

Figure 6.3. The cross-section of a water dam subjected to water pressure.

 Equilibrium Equations 243

Since the dam is very long along the z-direction, it is rigid in the
z-direction. In addition, the water pressure does not change in the
z-direction, and w can be regarded as zero. Therefore, the problem can
be simplified as 2D by considering only the 2D cross-section of the dam.
Equation (6.23) can be reduced to
 2  2
∂ u ∂2u ∂ u ∂2v
G + + (G + λ) + + bx = ρü
∂x2 ∂y 2 ∂x2 ∂x∂y
 2  2 (6.24)
∂ v ∂2v ∂ v ∂2u
G + + (G + λ) + + by = ρv̈
∂x2 ∂y 2 ∂y 2 ∂x∂y
For plane strain problems, the stress in the z-direction is not necessarily
zero and can be obtained from the in-plane stresses, as discussed in the
previous chapter.

[Link] Plane stress problems

When the stress components in one of the coordinate directions, say in the
z-direction, are zero, we have a plane stress problem. In such a case, all
the nonzero stresses stay in the x–y plane, and all the variables also do not
depend on z. A typical case is a thin plate. Its thickness in the z-direction is
small compared to that in its other dimensions. In addition, all the external
forces on it is only within the x–y plane. Since the two surfaces of the thin
plate are free of stresses, all the stress components in the z-direction can be
regarded as zero, as shown in Fig. 6.4.

Figure 6.4. The cross-section of a thin plate subjected to loadings only within the plane
of the plate.
244 Mechanics of Materials: Formulations and Solutions with Python

To obtain the governing equations for 2D plane stress problems, we sim-

2Gν
ply replace λ = 1−2ν in Eq. (6.23) with 2Gν
1−ν because ν needs to be replaced
ν
by 1+ν when converting a plane strain problem into a plane stress one, as
discussed in the previous chapter. After such replacements followed by sim-
plification, we obtain the equilibrium equations for 2D plane stress problems:
 2 
∂ u ∂2u 1 + ν ∂2u ∂2v
G + + G + + bx = ρü
∂x2 ∂y 2 1 − ν ∂x2 ∂x∂y
 2  (6.25)
∂ v ∂2v 1 + ν ∂2v ∂2u
G + +G + + by = ρv̈
∂x2 ∂y 2 1 − ν ∂y 2 ∂x∂y
The strain in the z-direction, εzz , is not zero because the thickness of the
plate can change when it is loaded in the x–y plane due to Poisson’s effect.
It can be calculated using these in-plane strains.
The difference between these two types of 2D problems is only in the
material constants at the front of the derivative terms. Therefore, they can
be dealt with in largely the same way. Plane strain problems behave stiffer
than the plane stress ones because of the displacement constraints on the
z-direction.
Note that even for 2D problems, seeking an analytical solution is still a
challenge. Solution can be found only feasible for some simple problems with
regular geometry.

6.3.3 Formulations of 1D bar problems

Consider one-dimensional (1D) problems defined in the x-coordinate. We
can have three types:
1. 1D yz-constrained (both y- and z-directions are constrained),
2. 1D y-constrained (or z-constrained),
3. 1D bar (with a free circumferential surface).

[Link] 1D yz-constrained problem

In this case, the displacements v and w are zero. The loading does not change
in the y- and z-directions. The nonzero strain is only εxx . All the variables
do not change with y and z. This is a 1D bar with all its circumferential
surfaces fully constrained. In this case, Eq. (6.23) is drastically reduced by
simply removing all terms with respect to y and z:
∂2u
(2G + λ) + bx = ρü (6.26)
∂x2
or
E(1 − ν) ∂2u
+ bx = ρü (6.27)
(1 − 2ν)(1 + ν) ∂x2
 Equilibrium Equations 245

Figure 6.5. A 1D bar constrained in the y-direction. It becomes a 1D strain problem:

the nonzero strains are in the x- and z-directions due to Poisson’s effect.

[Link] 1D y-constrained problem

If the displacement, say v, is zero, for example, if a 2D domain is infinite in
the y-direction, and also the loading does not change in the y-direction, as
shown in Fig. 6.5, we then have a 1D y-constrained problem. The nonzero
strain is εxx with εzz = −νεxx . All the variables do not change with y.
In this case, Eq. (6.25) is further reduced to

1 + ν ∂2u
G 1+ + bx = ρü (6.28)
1 − ν ∂x2
or
E ∂2u
+ bx = ρü (6.29)
1 − ν 2 ∂x2
The problem defined in Eq. (6.29) is a 1D y-constrained problem.
In this case, the stress σyy is not zero because the bar is not allowed to
shrink or expand freely in the y-direction, resulting in some stress σyy . Also,
εzz = −νεxx is nonzero because the z-direction is not constrained. This
problem is equivalent to the 1D z -constrained problem, where the stress
σzz is not zero because the bar is not allowed to shrink or expand freely
in the z-direction, and εyy = νεxx is nonzero because the y-direction is not
constrained.

[Link] 1D bar problem

If the 1D problem is a longitudinal bar with its circumferential surface free
from any stress, we have a 1D stress problem. This is the case we studied
at the beginning of Chapter 3, and we derived the governing equation in
terms of stress. To obtain its governing equation in displacement, we would
E
need to, once again, do a substitution: replacing 1−ν 2 with E in Eq. (6.29),
which gives
∂2u
E + bx = ρü (6.30)
∂x2
246 Mechanics of Materials: Formulations and Solutions with Python

Note in this case that the strains εyy and εzz are not zero because the
bar is allowed to shrink or expand freely in the y- and z-directions (Poisson’s
effect).
Equations (6.27), (6.29), and (6.30) are 1D deferential equations (DEs)
that can be solved with ease for many problems. Comparing Eqs. (6.29) and
1
(6.30), we find that the y-constraint makes the bar stiffer by a factor of 1−ν 2.

If ν = 0.3, the factor is about 1.1, or ∼10% stiffer. Comparing Eqs. (6.27)
and (6.30), we find that the yz-constraint makes the bar stiffer by a factor
(1−ν)
of (1−2ν)(1+ν) . If ν = 0.3, the factor is about 1.35, or ∼35% stiffer. The
constraint effects are only because of Poisson’s ratio. Such a stiffing effect
disappears only if Poisson’s ratio is zero.
Since Eqs. (6.27) and (6.29) differ from Eq. (6.30) by a material constant,
we only need to solve the latter because it is far more common in engineering
applications. We will do so in great detail in the following chapter.

6.4 Boundary conditions

The equilibrium equations (in either stresses or displacements) govern the

solid material inside the domain, as shown in Eq. (6.19). In actual engineering
problems of mechanics of solid materials, there are always some kinds of
conditions given at the boundary of the domain. These BCs must also be
satisfied. In general, there are two major types of BCs for solids: displacement
BCs and stress (or force) BCs.

6.4.1 Displacement boundary conditions

The displacement BCs can be expressed as follows:
ui = uSi on Su ∈ S (6.31)
where uSi denote the specified displacements ui on the boundary surface Su ,
which can be zero, but may not necessarily be.

6.4.2 Stress boundary conditions

The stress BCs can be expressed as follows:
σij ni = tnj on St ∈ S (6.32)
where nj is the jth component of the unit normal at a point on the boundary
surface St and tnj is the specified traction on the boundary St with the nor-
mal vector n, which can be zero, but may not necessarily be. Equation (6.32)
is essentially Eq. (3.38) used on the stress boundary.
 Equilibrium Equations 247

6.5 On techniques for solving solid mechanics problems

To find a solution to a solid mechanics problem, we need to solve a system
of equations consisting of the following:
• the equilibrium equation (also called the governing equation)
Eq. (6.22);
• the boundary conditions Eqs. (6.31) and/or (6.32);
• the compatibility equations derived in Section 4.13.
Since the governing equation can be written in terms of stress, as in
Eq. (6.9), or in terms of displacement, as in Eq. (6.19), we have largely two
possible methods to solve this system of equations: the stress method and the
displacement method. These methods are useful for higher-dimensional prob-
lems (2D and 3D). Each method has its own strategy, tricks, and strengths in
solving different types of problems. We will discuss these methods in future
volumes of this book series.
In this book, our objective is to solve problems for structural members,
which can often be reduced to 1D. In these cases, the stress and displacement
methods are essentially the same. We will use the displacement method in
the later chapters for various types of problems because it is straightforward
and easy to code to obtain systematic solutions for various types of problems,
including indeterminate ones.

6.6 Remarks
1. This chapter presents a set of system equations for solid mechanics prob-
lems. Using a free-body diagram of a stressed 3D solid, we establish a
set of equilibrium equations in all directions. It is a set of PDEs with the
stress components as the unknowns. Thus, there are six unknowns.
2. Since the stress tensor is a function of coordinates, we also thus use the
divergence of the stress tensor function together with the body-force vec-
tor to derive mathematically the same set of equilibrium equations.
3. The PDEs are then converted into ones with displacements as the
unknowns. This reduces the number of unknowns to three.
4. The PDEs are reduced to 2D and 1D.
5. Both displacement and traction BCs are introduced. The BCs need to be
satisfied in addition to the set of equilibrium PDEs.
Various solution methods will be introduced in future volumes on how to
solve these sets of system equations for 2D and 3D solid mechanics problems.
The following chapters of this book will focus on problems in mechanics of
materials, which can often be simplified into 1D problems or easily solvable
248 Mechanics of Materials: Formulations and Solutions with Python

2D problems. Hence, efficient solvers can be devised for various problems

involving elastic solid materials. Techniques on how to derive the simplified
equations will be discussed in great detail, followed by solution procedures
and Python codes, in later chapters.

References
[1] A.P. Boresi, R.J. Schmidt and O.M. Sidebottom, Advanced Mechanics of Materials,
New York, 1985.
[2] G.R. Liu and S.S. Quek, The Finite Element Method: A Practical Course.
Butterworth-Heinemann, New York, 2013.
[3] G.R. Liu and T.T. Nguyen, Smoothed Finite Element Methods. Taylor and Francis
Group, New York, 2010.
 Chapter 7

Solution to Axially Loaded Bars

1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.

2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 init_printing(use_unicode=True) # For latex-like quality printing
11 np.set_printoptions(precision=4,suppress=True,
12 formatter={'float': '{:0.4e}'.format}) # Digits in print-outs

Bars or rods are the most widely used structural members. A bar takes axial
loads that are applied along the axial direction of the bar. Solutions to it in
terms of both axial displacement and stress are useful in many applications.
A typical example is the widely used truss structures. Therefore, a bar is
also referred as a truss member.

7.1 Governing equation: Second-order

differential equation

Consider a bar subjected to a distributed body force bf , as shown in Fig. 7.1.

It is fixed at x = 0. At x = l, it may be subjected to a concentrated force
F (force driven) or with a fixed given displacement (displacement driven).
The problem is essentially one-dimensional (1D), implying that our concern
is mainly with the unknown displacement and stress/strain varying with the
coordinate x.

249
250 Mechanics of Materials: Formulations and Solutions with Python

Figure 7.1. A 1D bar fixed at the left end, subjected to a distributed body force bf
[N/m] and a concentrated force F [N] at the right end.

The governing equation for 1D bar problems has been derived in Chapter
6 and is rewritten as follows in terms of displacement:
∂2u −bf
2
= (7.1)
∂x EA
where bf (N/m) is the body force within a 1D bar or truss member in the
axial direction. For bars with uniform cross-sectional area A (m2 ), it relates
to the body force per volume bx (N/m3 ) as

bf = bx A (7.2)

Equation (7.1) is a second-order differential equation (DE). It can be

solved easily by simply integrating twice. As long as the bf function can
be integrated analytically, we can obtain the exact solutions for bars with
different types of boundary conditions (BCs). Since the DE is of second
order, the integrations result in two unknown constants. Since the 1D bar
always has two boundaries (left and right ends), which provide us the exactly
needed number of BCs to determine these two constants, we have a logical
path to find the exact solution effectively. Let us use the following example
to demonstrate this process.

7.2 Solution procedure

First, we use an example to derive, by hand, the formulas for displacement

and stress solutions, considering the simplest case where the body force is
zero, so that we can have a good idea about how Eq. (7.1) can be solved.
We will then write a Python code that can be used to derive formulas for
more complicated problems. The code follows essentially the same process,
but allows the computer to do the heavy lifting of derivation.

7.2.1 Example: A bar subjected to a force

at its right end
Consider a bar with length L and a uniform cross-sectional area A. It is made
of a material with Young’s modulus E. It is fixed at the left end, where x = 0,
 Solution to Axially Loaded Bars 251

and subjected to a force at its right end, denoted as F at x = L. Derive the

formulas for displacement and stress in the bar.
Solution: Since there is no body force, Eq. (7.1) is reduced to
∂2u
=0 (7.3)
∂x2
The general solution of u that satisfies Eq. (7.3) can be found by integra-
tion. Integrating once gives
∂u
= c0 (7.4)
∂x
where c0 is an integral constant. Integrating again gives
u = c0 x + c1 (7.5)
where c1 is another integral constant. Equation (7.5) is the general solution
for any bar without body force. It has a total of two constants. For a given
specified problem defined in this example, we shall use the BCs to determine
these constants.
At x = 0, the bar is fixed, and hence, we have the following BC:
u = 0 at x = 0 (7.6)
To satisfy Eq. (7.6), c1 in Eq. (7.5) has to be zero, which leads to
u = c0 x (7.7)
At x = L, the bar is loaded by F , and hence, we have the following BC:
∂u
σA = F, or AEε = F, or AE =F at x = L (7.8)
∂x
Substituting Eq. (7.7) into Eq. (7.8), we obtain
F
AEc0 = F, or c0 = (7.9)
AE
Finally, the displacement solution is found as
F
u= x (7.10)
AE
At x = L, we have displacement uR as
L
uR = F (7.11)
AE
where AEL is known as the axial stiffness of a bar and AE is the axial
L
stiffness per length. In reverse, AE is called the axial flexibility of a bar.
252 Mechanics of Materials: Formulations and Solutions with Python

Equation (7.11) can also be written as

AE
F = uR (7.12)
L

The stress solution is next found as

∂u F
σ = Eε = E = (7.13)
∂x A
Clearly, this is the correct solution, as expected. This is in fact the solution
we found at the beginning of Chapter 3. Here, we found it by using a standard
solution procedure for DEs.
The same procedure can find solutions for bars subjected to body forces
as long as the force function is integrable analytically. We shall use Sympy
to do the derivation for us.

7.2.2 Python code for deriving solutions for bars

We write a Python code function to get it done in a step-by-step approach
integrating Eq. (7.1), the same procedure that we did by hand in the previous
example. The code is self-explanatory with comment lines:

1 def solver1D2(E, A, bf, l, bc0, bcl, key='disp'):

2
3 '''Solves mechanics problems for 1D bars subjected to body
4 force bf: u_xx=-bf(x)/EA, with BCs: u(0)=bc0; u(l)=bcl or fR=bcl.
5 Input: bf, body force; l, the length of the bar;
6 bcl, condition at x=l;
7 key="disp", displacement condition at a boundary;
8 otherwise a force boundary condition
9 Return: u (displacement), E*u_x (stress), u_xx (body force term)
10 Example: Consider a bar fixed at x=0, loaded with fR at x=L and q:
11 bf = q
12 u = solver1D2(E, A, bf, l, uL, fR, key='force')
13 '''
14 x, c0, c1 = [Link]('x, c0, c1')
15
16 # Integrate twice to obtain the general solutions in u_x and u:
17 # u_x: du/dx
18 u_x = [Link](-bf/E/A,(x))+ c0 #c0: integration constant
19 u = [Link](u_x,(x))+ c1 #c1: integration constant
20
21 # Solve for the integration constants:
22 if key == "disp":
23 consts=[Link]([[Link](x,0)-bc0, [Link](x,l)-bcl], [c0, c1])
24 else:
25 consts=[Link]([[Link](x,0)-bc0,u_x.subs(x,l)-bcl/E/A],[c0,c1])
26
 Solution to Axially Loaded Bars 253

27 # Substitute the constants back to the general solutions, to obtain

28 # solutions that satisfy the BCs:
29 u = [Link]({c0:consts[c0],c1:consts[c1]})
30 u = [Link]().simplify()
31 u_x = u_x.subs({c0:consts[c0],c1:consts[c1]})
32 u_x = u_x.expand().simplify()
33 u_xx = [Link](u_x,x)
34
35 print(f'output from solver1D2(): u, , u_xx')
36 return [Link](), (E*u_x).expand(), u_xx.expand()

Let us use this code function to find solutions for two 1D problems: one
with force boundary condition on the right end and the other with a displace-
ment condition there. For both cases, the bar may be subjected to a given
body force. We fix the bar at x = 0 because it is needed to have our mechanics
problem well defined, indicating that it has a physically meaningful solu-
tion with nonzero strains or stresses. By “fix”, we mean that the displacement
is pre-described with a fixed value that can be zero or a finite value.

7.3 Example: Constant body force and concentrated

force at the right end

Consider a bar subjected to a constant body force and a concentrated force

at x = l, all in its axial direction, as shown in Fig. 7.1. We compute the full
solution: axial displacement, strain, and stress in the bar.

7.3.1 Method 1: Step-by-step approach integration

1 E, A, l = symbols('E, A, l', nonnegative=True) #l: length of the bar

2 x, q, bf, c0, c1 = symbols('x, q, b_f, c0, c1')
3
4 fL, fR, uL,uR = symbols('f_L, f_R, u_L, u_R')
5 # fL: internal force at the Left-end; fR: that at the Right-end
6 # uL: displacement at the Light-end; uR: displacement at the Right-end
7
8 bf = [Link]('b_f')(x) # 1D body-force (N/m)
9 title = ["Displacement", "Stress"]
10
11 # Consider the distributed force is a constant q
12 bf = q #0 #(2*[Link](x)+8)/E one may try other force functions
13 u = solver1D2(E, A, bf, l, uL, fR, key='force') # or: gr.solver1D2()
14
15 # check whether the DE is satisfied.
16 print(f'Is solution correct? {u[2] == -bf/E/A}')
17 u # solutions: u; E*u,x; u,xx

output from solver1D2(): u, σ, u xx

Is solution correct? True
254 Mechanics of Materials: Formulations and Solutions with Python

 
fR x lqx qx2 fR lq qx q
uL + + − , + − , −
AE AE 2AE A A A AE

We found a full set of formulas for solutions to the 1D bar fixed with uL
at the left end (x = 0), subjected to distributed force q and point force fR
at the right end (x = l). The formula for the displacement is summarized as
follows:

fR x qlx qx2
u(x) = uL + + − (7.14)
AE AE 2AE

It is quadratic in x and has a maximum value at x = l:

fR l ql2
u(x) = uL + + (7.15)
AE 2AE

When setting q = 0, uL = 0, this is Eq. (7.11).

The formula for the stress is

fR ql qx
σ(x) = + − (7.16)
A A A

When setting q = 0, this is Eq. (7.13).

Equation (7.16) is linear in x and has a maximum value at x = 0:

fR ql fR + ql
σ(x) = + = (7.17)
A A A

This is expected because the root of the bar receives the total force of
(fR + ql).
Equations (7.14) and (7.16) are useful formulas for computing the solution
to a bar subjected to uniform body force. The stress solution is exactly the
same as we obtained in the first example, by setting q = 0, uL = 0. It is also
exactly the same obtained in the second example in Chapter 3 (by setting
fR = 0).
To obtain solutions for a bar with specified parameters, we simply sub-
stitute these parameters with fixed values. For example:
 Solution to Axially Loaded Bars 255

1 ux_cq=[u[i].subs({E:1,A:1,l:1,q:1,uL:0,fR:0}) for i in range(len(u))]

2 ux_cq # solution for constant body force q

 
x2
− + x, 1 − x, −1
2

If the distributed body force is another type function of x, we simply

change the expression for bf in the foregoing code. For example, if the force is
a sine function, we use the following code to obtain the formulas of solutions:

1 # Consider the distributed force is a constant q

2 bf = q*[Link]([Link]*x)
3 u = solver1D2(E, A, bf, l, uL, fR, key='force')
4 print(f'Is solution correct? {u[2] == -bf/E/A}')
5 u

output from solver1D2(): u, σ, u xx

Is solution correct? True

 
fR x qx cos (πl) q sin (πx) fR q cos (πl) q cos (πx) q sin (πx)
uL + − + , − + , −
AE πAE π 2 AE A πA πA AE

1 # For a specified set of parameters (unit values):

2 ux=[u[i].subs({E:1,A:1,l:1,q:1,uL:0,fR:0}) for i in range(len(u))]
3 ux

 
x sin (πx) cos (πx) 1
+ , + , − sin (πx)
π π2 π π

Here, the displacement at the left end of the bar is set to zero. All the
deformation and stress/strain in this case are all driven by the forces (body
force and the force at the right end). This types of problem is known as a
force-driven problem.
256 Mechanics of Materials: Formulations and Solutions with Python

7.3.2 Method 2: Use the Sympy differential

equation solver
One can may use the built-in [Link]() to solve the differential equation
directly:
1 E, A, l = symbols('E, A, l', nonnegative=True)
2 x, q = symbols('x, q ')
3 fR, uL, uR = symbols('f_R, u_L, u_R')
4 u = [Link]('u')(x) # 1D displacement (m)
5
6 # Consider the distributed force is a constant q
7 bf = q*[Link]([Link]*x)
8
9 # Define the 2nd differential equation
10 diff_eq = [Link](x, 2) + bf/E/A
11
12 # Define the boundary conditions
13 BCs = {[Link](x, 0): uL, [Link](x).subs(x, l): fR/E/A}
14
15 # Solve the differential equation using [Link]()
16 solution = [Link](diff_eq, u, ics=BCs) # This is an [Link] object
17
18 # check whether the DE is satisfied.
19 print(f'Is solution correct? {[Link](x,2) == -bf/E/A}')
20 solution # solution equation

Is solution correct? True

q sin (πx) x (πfR − q cos (πl))
u(x) = uL + +
π 2 AE πAE
The result obtained is the same as that obtained earlier. Note that this
approach only gives the displacement.

7.3.3 Python code for plotting the solutions

Next, we write the following code to plot the results:
1 def plot2curveS(u, xL=0., xR=1., title="f_title"):
2
3 '''Plot the distribution of function u, the maximum values and
4 locations, as well as stationary points, and the values at the
5 stationary points, and boundaries.
6 input: u, sympy function defined in [xL, xR]
7 '''
8 x = symbols('x')
9 dx = 0.01; dxr = dx*10 # x-step
10 xi = [Link](xL, xR+dx, dx)
11 uf = [Link]((x), u[0], 'numpy') #convert Sympy f to numpy f
12 yi = uf(xi)
 Solution to Axially Loaded Bars 257

13 if type(yi) != [Link]: #in case, uf is a constant

14 #type(yi) == int or type(yi) == float: # or len(yi)==1:
15 xi = [Link](xL, xR+dxr, dxr)
16 yi = float(yi)*np.ones_like(xi)
17
18 fig, ax1 = [Link](figsize=(5.,1.), dpi=300)
19 fs = 8 # fontsize
20 color = 'black'
21 ax1.set_xlabel('location x', fontsize=fs)
22 ax1.set_ylabel(title[0], color=color, fontsize=fs)
23 [Link](xi, yi, color=color)
24 [Link](color='r',ls=':',lw=.3, which='both') # Use both tick
25 ax1.tick_params(axis='x', labelcolor=color, labelsize=fs)
26 ax1.tick_params(axis='y', labelcolor=color, labelsize=fs)
27
28 vmax = yi[[Link]()]
29 max_l = [Link](yi == vmax)
30 [Link](xi[max_l], yi[max_l], 'r*', markersize=4)
31 print(f'Maximum {title[0]} value={vmax:.3e},at x={xi[max_l][0][0]}')
32
33 uf = [Link]((x), u[1], 'numpy') #convert Sympy f to numpy f
34 xi = [Link](xL, xR+dx, dx)
35 yi2 = uf(xi)
36 if type(yi2) != [Link]: # or len(yi2) == 1:
37 xi = [Link](xL, xR+dxr, dxr)
38 yi2 = float(yi2)*np.ones_like(xi)
39
40 m1, m2, m3 = [Link](abs(yi2), 2)[0:3]
41 msl=[[Link](abs(yi2)==m1)[0][0],[Link](abs(yi2)==m2)[0][0],
42 [Link](abs(yi2)==m3)[0][0]]
43
44 vmax = yi2[[Link]()]
45 max_l = [Link](yi2 == vmax)
46 print(f'Maximum {title[1]} value={vmax:.3e},at x={xi[max_l][0][0]}')
47
48 if abs(xi[msl[2]]-xi[msl[1]])<2*dx:
49 if abs(yi2[msl[2]]-0.)<abs(yi2[msl[1]]-0.): [Link](1)
50 else: [Link](2)
51 if len(msl) > 2:
52 if abs(xi[msl[2]]-xi[msl[0]])<2*dx:
53 if abs(yi2[msl[2]]-0.)<abs(yi2[msl[0]]-0.): [Link](0)
54 else: [Link](2)
55 if len(msl) > 1:
56 if abs(xi[msl[1]]-xi[msl[0]])<2*dx:
57 if abs(yi2[msl[1]]-0.)<abs(yi2[msl[0]]-0.): [Link](0)
58 else: [Link](1)
59
60 ax2 = [Link]() # instantiate second axes sharing the same x-axis
61
62 color = 'blue'
63 ax2.set_ylabel(title[1], color=color, fontsize=fs)
64 [Link](xi, yi2, color=color)
258 Mechanics of Materials: Formulations and Solutions with Python

65 [Link](xi[max_l], yi2[max_l], 'r*', markersize=4)

66 [Link](xi[msl], yi2[msl], 'r*', markersize=4)
67 [Link](xi[msl], yi[msl], 'r*', markersize=4)
68 [Link](xi[0], yi2[0], 'ro', markersize=2)
69 [Link](xi[0], yi[0], 'ro', markersize=2)
70 [Link](xi[-1], yi2[-1], 'ro', markersize=2)
71 [Link](xi[-1], yi[-1], 'ro', markersize=2)
72 [Link](color='r',ls=':',lw=.5, which='both') # Use both tick
73 ax2.tick_params(axis='x', labelcolor=color, labelsize=fs)
74 ax2.tick_params(axis='y', labelcolor=color, labelsize=fs)
75 np.set_printoptions(formatter={'float': '{: 0.3e}'.format})
76 print(f'Extreme {title[0]} values={yi[msl]},\n at x={xi[msl]}')
77 print(f'Critical {title[1]} values={yi2[msl]},\n at x={xi[msl]}')
78 print(f'{title[0]} values at boundary ={yi[0], yi[-1]}')
79 print(f'{title[1]} values at boundary ={yi2[0], yi2[-1]}\n')

We plot the solutions for the bar subjected to a constant body force q = 1
using the code function given above:

1 title = ["Displacement", "Stress"]

2 # use ux_cq, solution from method 1
3 plot2curveS(ux_cq, title=title) # or gr.plot2curveS()
4 #[Link]('images/beam_cq.png', dpi=500) # save the plot to a file
5 [Link]()

Maximum Displacement value=5.000e-01,at x=1.0

Maximum Stress value=1.000e+00,at x=0.0
Extreme Displacement values=[ 5.000e-01],
at x=[ 1.000e+00]
Critical Stress values=[ 0.000e+00],
at x=[ 1.000e+00]
Displacement values at boundary =(0.0, 0.5)
Stress values at boundary =(1.0, 0.0)
 Solution to Axially Loaded Bars 259

It is seen that for a constant body force, the displacement is quadratic

and the stress is linear along x. Let us print out specific functions and values.

1 [Link](ux_cq[0].subs(x,l), 'Displacement at the right-end',n_dgt=4)

l2
Displacement at the right end: − +l
2

1 [Link](ux_cq[1], 'The stress in the bar',n_dgt=4)

The stress in the bar: 1 − x

1 [Link]((ux_cq[1]/E).expand(), 'The strain in the bar',n_dgt=4)

x 1
The strain in the bar: − +
E E

1 fL = (ux_cq[1]).subs(x,0).expand()
2 [Link](fL, 'Internal force at left-end:',n_dgt=4)

Internal force at left-end: 1

1 [Link]((ux_cq[1]).subs(x,l), 'Internal force at right-end:')

Internal force at right-end: 1−l

1 [Link](-ux_cq[2], 'The distributed force along the bar:')

The distributed force along the bar: 1

260 Mechanics of Materials: Formulations and Solutions with Python

7.4 Example: Linear body force and concentrated

force at the right end

Consider a bar subjected to a linearly distributed body force and a con-

centrated force fR at x = l, both in its axial direction. The setting of the
problem is the same as that shown in Fig. 7.1, but the function for the body
force is given by

x
bf = q0 + q (7.18)
l

where q0 and q are given constants.

The solution can then be solved easily by simply integrating Eq. (7.1)
twice by hand. Here, we use solver1D2(), which gives formulas for the dis-
placement, stress, and the second derivative of the displacement:

1 # Consider the distributed force is a linear function

2 l, E, A = symbols('l, E, A')
3 q0, q, x = symbols('q0, q, x ')
4 bf = q0 + q*x/l
5
6 u = solver1D2(E, A, bf, l, uL, fR, key='force')
7 print(f'Is solution correct? {u[2] == -(bf/E/A).expand()}')
8 u

output from solver1D2(): u, σ, u xx

Is solution correct? True
 
fR x lqx lq0 x q 0 x2 qx3 fR lq lq0 q0 x qx2 q0 qx
uL + + + − − , + + − − , − −
AE 2AE AE 2AE 6AEl A 2A A A 2Al AE AEl

This is the set of formulas for calculating the displacement in the bar in
the x direction and the stress in the bar. The last expression is the term
on the right-hand side of Eq. (7.1), indicating that the solution satisfies the
differential equation.

1 # For a set of specified parameters:

2 ux=[u[i].subs({E:1,A:1,l:1,q0:.1,q:1,uL:0,fR:0}) for i in range(len(u))]
3 ux=[[Link](ux[i]) for i in range(len(ux))]
4 ux

 
x3 x2 3x x2 x 3 1
− − + , − − + , −x −
6 20 5 2 10 5 10
 Solution to Axially Loaded Bars 261

1 # Plot the distribution of displacement, stress and the maximum values:

2 plot2curveS(ux, title=title)
3 #[Link]('images/bar_cq.png', dpi=500) # save the plot to a file
4 [Link]()

Maximum Displacement value=3.833e-01,at x=1.0

Maximum Stress value=6.000e-01,at x=0.0
Extreme Displacement values=[ 3.833e-01],
at x=[ 1.000e+00]
Critical Stress values=[ 0.000e+00],
at x=[ 1.000e+00]
Displacement values at boundary =(0.0, 0.3833333333333333)
Stress values at boundary =(0.6, 0.0)

For this linear body-force case, the displacement is cubic and the stress
is quadratic in x.

7.5 Example: A displacement imposed at the right end

Consider a bar with its displacement is imposed as a fixed value ul at

x = l. The deformation and stress/strains in this case are all driven by
the imposed displacement at the boundary. This types of problem is known
as displacement-driven problem. This is the typical case when we do a
tensile test to a material specimen:

1 # Consider the distributed force is constant

2 bf = q # One may try this; q0 + q*x/l
3 u = solver1D2(E, A, bf, l, uL, uR, key='disp')
4 print(f'Is solution correct? {u[2] == -bf/E/A}')
5 u

output from solver1D2(): u, σ, u xx

Is solution correct? True
 
uL x uR x lqx qx2 EuL EuR lq qx q
uL − + + − , − + + − , −
l l 2AE 2AE l l 2A A AE
262 Mechanics of Materials: Formulations and Solutions with Python

1 # For a set of specified parameters:

2 ux=[u[i].subs({q:0, uL:0}) for i in range(len(u))]
3 ux # General solution for u, in a bar when right-end displaced by uR

 
uR x EuR
, , 0
l l
This gives the general solution for u and σ in a bar when right end is
displaced by uR .
If we set specified parameters with numerical values, we can plot the
solution:

1 # For a set of specified parameters:

2 ux=[u[i].subs({E:1,A:1,l:1, q:0,uL:0,uR:1}) for i in range(len(u))]
3 ux

[x, 1, 0]

1 # Plot the distribution of the displacement and stress:

2 plot2curveS(ux, title=title)
3 #[Link]('images/bar_cq.png', dpi=500)
4 [Link]()

Maximum Displacement value=1.000e+00,at x=1.0

Maximum Stress value=1.000e+00,at x=0.0
Extreme Displacement values=[ 0.000e+00],
at x=[ 0.000e+00]
Critical Stress values=[ 1.000e+00],
at x=[ 0.000e+00]
Displacement values at boundary =(0.0, 1.0)
Stress values at boundary =(1.0, 1.0)
 Solution to Axially Loaded Bars 263

For this displacement-driven problem, the displacement is linear and the

stress is constant in x:

1 [Link](u[0].subs(x,l), 'Displacement at right-end:',n_dgt=4)

Displacement at right end: uR

1 [Link](u[1], 'The stress in the bar:',n_dgt=4)

EuL EuR lq qx
The stress in the bar: − + + −
l l 2A A

1 [Link]((u[1]/E).expand(), 'The strain in the bar:',n_dgt=4)

uL uR lq qx
The strain in the bar: − + + −
l l 2AE AE

1 N0 = (u[1]*A).subs(x,0).expand()
2 [Link](N0, 'Internal force at left-end: f0=',n_dgt=4)

AEuL AEuR lq
Internal force at left end: f0 = − + +
l l 2

1 Nl = (u[1]*A).expand().subs(x,l)
2 [Link]((Nl).subs(x,l), 'Internal force at the right-end:')

AEuL AEuR lq
Internal force at the right end: − + −
l l 2

1 [Link](-u[2]*E*A, 'The distributed force along the bar')

The distributed force along the bar: q

264 Mechanics of Materials: Formulations and Solutions with Python

7.6 Bars subjected to discontinuous body forces

7.6.1 Distributed forces

In deriving the equilibrium equations, we assumed that the stresses are suffi-
ciently smooth and their functions are differentiable at least at first order.
When the body force is discontinuous, the stresses may not be differentiable
at the force discontinuity points. Figure 7.2 shows a bar fixed at the left
end, subjected to two distributed body forces q1 and q2 (N/m), and a con-
centrated force fR at the right end.
In this case, we need to treat the bar being formed by two bars: bar-1 has
length l1 and bar-2 has length l2 . Since this problem is statically determinate,
meaning that the internal forces can be obtained only by equilibrium, the
forces applied at the right end of each of these two sub-bars can be obtained
by simple equilibrium conditions:
For bar-2: f2R = fR that is the externally applied force at the right end of
the bar.
For bar-1: f1R = fR + q2 l2 .
Using Eq. (7.14), the displacement solution for bar-1 is found as
f1R x l1 q1 x q 1 x2
u1 (x) =
+ − (7.19)
AE AE 2AE
where we used boundary condition uL = 0.
At x = l1 , we have
f1R l1 q1 l12
u1 (l1 ) = +
AE 2AE
 
l1 q1 l1
= f1R + (7.20)
AE 2
 
l1 q1 l1
= fR + q2 l2 +
AE 2

Figure 7.2. A 1D bar fixed at the left end, subjected to two distributed body-force, and
a concentrated force at the right end.
 Solution to Axially Loaded Bars 265

The displacement solution for bar-2 is found as

f2R x q2 l2 x q 2 x2
u2 (x) = u1 (l1 ) + + −
AE AE 2AE
(7.21)
fR x q2 l2 x q 2 x2
= u1 (l1 ) + + −
AE AE 2AE
Here, we used the boundary condition for bar-2 at its left end: uL = u1 (l1 ).
Using Eq. (7.16), the stress solution for bar-1 is found as
fR + q2 l2 q1 l1 q1 x
σ1 (x) = + − (7.22)
A A A
Here, we used f1R = fR + q2 l2 . The stress solution for bar-2 is found as
fR q2 l2 q2 x
σ2 (x) = + − (7.23)
A A A
Clearly, if q1 = q2 = 0, we have the solution for the whole bar subjected to
fR .

7.6.2 Point force

For point forces, the discontinuities are at the points where the forces are
applied. Figure 7.3 shows a bar fixed at the left end. It is subjected to two
point forces: fI at an interior point and fR at the right end.
In this case, we also need to treat the bar as two bars: bar-1 has length
l1 and bar-2 has length l2 . This problem is also statically determinate. The
forces applied at the right-end of each of these two sub-bars can be obtained
by simple equilibrium conditions:
For bar-2: f2R = fR that is the externally applied force at the right-end of
the bar.
For bar-1: f1R = fR + fI . Here, we need to look at the left-hand side of the
point where fI is applied.

Figure 7.3. A 1D bar fixed at the left end, subjected to two point forces: one at an
interior point, and one at the right end.
266 Mechanics of Materials: Formulations and Solutions with Python

Using Eq. (7.14), the displacement solution for bar-1 is found as

f1R x (fR + fI )x
u1 (x) = = (7.24)
AE AE
where we used boundary condition uL = 0. At x = l1 , we have
f1R l1 (fR + fI )l1
u1 (l1 ) = = (7.25)
AE AE
The displacement solution for bar-2 is found as
(fR + fI )l1 fR x
u2 (x) = + (7.26)
AE AE
Here, we used the boundary condition for bar-2 at its left end: uL = u1 (l1 ).
Using Eq. (7.16), the stress solution for bar-1 is found as
fR + fI
σ1 (x) = (7.27)
A
Here, we used, f1R = fR + fI . The stress solution for bar-2 is found as
fR
σ2 (x) = (7.28)
A

7.7 Statically indeterminate bars: Reaction force

If the problem is statically indeterminate, implying the forces applied at
the right end of each of the bars cannot be obtained only by equilibrium
conditions, extra work is needed. Figure 7.4 shows a bar fixed at both ends,
subjected to two distributed body forces q1 and q2 (N/m).

Figure 7.4. Top: a 1D bar fixed at both ends, subjected to two distributed body-force.
Bottom: the constraint at the right end is removed, and replaced by an unknown reaction
force R.
 Solution to Axially Loaded Bars 267

For this problem, we still need to treat the bar as two bars: bar-1 has
length l1 and bar-2 has length l2 because of the body force discontinuity.
This problem is statically indeterminate, and hence, the forces applied at the
right end of each of these two sub-bars cannot be obtained by equilibrium
conditions alone. Therefore, we remove the constraint at the right end, and
replace it with an unknown force called reaction force R. After this is done,
the displacements at the right end of the bar-2 can be found using the same
procedure in the previous two examples. Using Eq. (7.21), and replacing fR
by R, we obtain

l1 Rx q2 l2 x q 2 x2
u2 (x) = (2R + 2q2 l2 + q1 l1 ) + + − (7.29)
2AE AE AE 2AE

At x = l2 , u2 = 0, we then found

−1
R= (q1 l12 + q2 l22 + 2q2 l1 l2 ) (7.30)
2(l1 + l2 )

It is clear that the reaction force is negative for positive body forces. Consider
a special case of q1 = q2 = q, Eq. (7.30) becomes

−q −q −q(l1 + l2 )
R= (l12 + l22 + 2l1 l2 ) = (l1 + l2 )2 =
2(l1 + l2 ) 2(l1 + l2 ) 2
(7.31)

Finally, substituting Eq. (7.30), for (7.19), (7.21) we obtain the displace-
ment solution. Substituting Eq. (7.30), for (7.22), (7.23) we obtain the stress
solution.
The above technique works well for simple problems with a small number
of overconstraints. For large structural system, the overconstraints can be
very large in number. More effective approaches such as the finite element
method [1] should be used.

7.8 Thermal stress

7.8.1 Thermal displacement

Structural components may work in an environment where the temperature
changes. Due to the thermal expansion of materials and the overconstrained
268 Mechanics of Materials: Formulations and Solutions with Python

(indeterminate) nature of the bar, the material is not allowed to expand

freely. This can result in additional stress, known as thermal stress. The
general constitutive equation is given in Chapter 5. For 1D bars, we have

εT xx = αΔT (7.32)

where εT xx is the thermal strain. The displacement corresponding to the

temperature change can be found from the following differential equations:
∂uT
= εT xx = αΔT (7.33)
∂x
Integrating the foregoing equation gives

uT = αΔT x + c (7.34)

where c is the integral constant. Assuming uT = 0 at x = 0, we found c = 0,

and we have

uT = αΔT x (7.35)

This is the displacement at x, when the bar is fixed at x = 0 and subjected

to a temperature change ΔT .

7.8.2 Thermal forces

Such a thermal displacement can be simply added into the displacement
expression obtained earlier. For the example of fixed–fixed bar subjected to
distributed body forces discussed in the previous section, we shall use the
same trick of replacing the constraint on the left end by an unknown reaction
force R. Using Eq. (7.29), we have the combined displacement in bar-2 as

l1 Rx q2 l2 x q 2 x2
u2 (x) = (2R + 2q2 l2 + q1 l1 ) + αΔT l1 + + − + αΔT x
2AE   AE AE 2AE  
uT 1 uT 2

(7.36)

where the last term is the extra displacement caused by the temperature
change.
At x = l2 , u2 = 0, we then found
−1
R= (q1 l12 + q2 l22 + 2q2 l1 l2 ) − AE(αΔT ) (7.37)
2(l1 + l2 )
The last term is the thermal forces resulting from the temperature change.
Even if all the distributed forces are zero, there will still be a stress built up
 Solution to Axially Loaded Bars 269

in the bar. For positive temperature change, the reaction force is negative
as expected. The formula for R is found using the following code:

1 E, A, l1, l2, x = symbols('E, A, l_1, l_2, x')

2 R, ΔT, q1, q2, , x = symbols('R, ΔT, q1, q2, , x', nonnegative=True)
3
4 eqn = l1*(2*R+2*q2*l2+q1*l1)/(2*A*E) + *ΔT*l1 + \
5 l2*(2*R +q2*l2)/(2*A*E) + *ΔT*l2
6 sln = [Link](eqn, R)
7 sln

l12 q1 l22 q2
−AEl1 ΔT α − AEl2 ΔT α − 2 − l1 l2 q2 − 2
l1 + l2

After the reaction force is found, the displacements in these two bars
should be obtained using Eqs. (7.19)–(7.21). The combined displacement
formula for bar-1 becomes
(R + q2 l2 )x l1 q1 x q 1 x2
u1 (x) = + − + αΔT x (7.38)
AE AE 2AE
At x = l1 , we have
(R + q2 l2 )l1 q1 l12 l1
u1 (l1 ) = + = (2fR + q1 l1 ) + αΔT l1 (7.39)
AE 2AE 2AE
The combined displacement formula for bar-2 is
Rx q2 l2 x q 2 x2
u2 (x) = u1 (l1 ) + + − + αΔT x (7.40)
AE AE 2AE

7.8.3 Computing thermal stresses

[Link] Approach 1: Use reaction and external forces
The thermal stress can be obtained by treating the reaction force R as the
externally applied force in addition to other externally applied forces. In this
approach, the formulas for stress remain unchanged. For stress in bar-2, for
example, the formula becomes
R q2 l2 q2 x
σxx = + − (7.41)
A A A
The thermal effects to the stress solution are introduced thorough the reac-
tion forces resulting from the temperature change. Again, the thermal stress
should be be computed from all forces, including the mechanical forces and
the thermal forces.
270 Mechanics of Materials: Formulations and Solutions with Python

[Link] Approach 2: Use the combined displacement

Alternatively, we can use the combined displacement and the strain–
displacement relation to compute the strain. Then, compute the stress using
the constitutive equation. In this case, we need to use the constitutive equa-
tion given in Chapter 5:
   
ducombined duT ducombined
σxx = E − =E − αΔT (7.42)
dx dx dx
This is because the displacement gradient caused by temperature change
should not result in stresses and hence needs to be removed. Using the fore-
going equation for bar-2, for example, we shall get the same formula as given
in Eq. (7.41).

7.9 Example: Stress in an overconstrained bar loaded

mechanically and thermally

Consider a bar with length 1.0 subjected to q1 = 1.0 over 0 < x < 0.5 and
q2 = 0.0 over 0.5 < x < 1.0. Let A = 1.0 and E = 1.0. Compute the reaction
forces, the displacement and stress distributed over the entire bar. The bar
may be loaded mechanically and thermally.
Solution: We first write a code function to calculate the displacement and
stress at x in an arbitrary 1D bar with length of l1 + l2 , subjected to 2
distributed body-forces q1 and q2 , respectively, over l1 and l2 , for any given
force at the right end. This code also considers the thermal expansion, which
is set as zero as default.
1 def twobars_bf(x, A, E, l1, l2, q1, q2, fR, aT=0, barId=1):
2 '''Calculate the displacement and stress at x in a 1D bar subjected
3 to 2 distributed body-force q1 & q2 over l1 and l2.
4 fR: force applied at the right-end of the bar
5 A and E: area and Youngs modulus of the bar
6 aT: *ΔT (thermal expansion coefficient * change in temperature)
7 barId: if 1 (default), calculate the results over l1
8 if 2, calculate the results over l2
9 Return: displacement u, and stress
10 '''
11 if barId == 1:
12 u = ((fR+q2*l2)*x+l1*q1*x-q1*x**2/2)/(A*E) + aT*x
13 = (fR+l2*q2 + q1*l1-q1*x)/A
14 else:
15 u1R = l1*((fR+q2*l2) + l1*q1/2)/(A*E) + aT*l1
16 u = u1R + (fR*x+l2*q2*x-q2*x**2/2)/(A*E) + aT*x
17 = (fR + q2*l2-q2*x)/A
18 return u,
 Solution to Axially Loaded Bars 271

Set the values for the variables involved in this problem, assuming there
is no temperature change.

1 q1 = 1.0; q2 = 0.0
2 l1 = 0.5; l2 = 0.5
3 E = 1.0; A = 1.0
4 = 1.0e-4 # thermal expansion coefficient
5 dT = 0.0 # temperature change
6 T = *dT
7 RT = A*E* T # Reaction force by temperature change
8
9 # Compute the reaction force at the right-end of the bar
10 fR = -(q1*l1**2 + q2*l2**2+2*q2*l1*l2)/(2*l1 + 2*l2)-RT # fR=-0.125
11 print(f"Reaction force at the right-end = {fR}")
12
13 dx = 0.002
14 x1 = [Link](0.0, l1+dx, dx) # coordinates in l1
15 x2 = [Link](0.0, l2+dx, dx) # coordinates in l2
16
17 # Compute the displacement and stress:
18 u1, 1 = twobars_bf(x1, A, E, l1, l2, q1, q2, fR, aT= T, barId=1)
19 u2, 2 = twobars_bf(x2, A, E, l1, l2, q1, q2, fR, aT= T, barId=2)
20
21 x = [Link]((x1, l1+x2)) # put the results in a single array
22 ux = [Link]((u1, u2))
23 sx = [Link](( 1, 2))
24
25 ux_max = [Link](abs(ux)) # find the absolute maximum value
26 sx_max = [Link](abs(sx))
27 print(f"The absolute maximum displacement: {ux_max}")
28 print(f"The absolute maximum stress: {sx_max}")
29
30 step = max(int(len(x)/5), 1)
31 print(f"Location x:\n {x[::step]}") # print out some results
32 print(f"Displacement ux:\n {ux[::step]}")
33 print(f"Stress :\n {sx[::step]}")
34 print(f"Normal force N:\n {A*sx[::step]}")

Reaction force at the right end = −0.125

The absolute maximum displacement: 0.07031200000000001
The absolute maximum stress: 0.375
Location x:
[ 0.000e+00 2.000e-01 4.000e-01 5.980e-01 7.980e-01
9.980e-01]
Displacement ux:
[ 0.000e+00 5.500e-02 7.000e-02 5.025e-02 2.525e-02
2.500e-04]
272 Mechanics of Materials: Formulations and Solutions with Python

Stress σ:
[ 3.750e-01 1.750e-01 -2.500e-02 -1.250e-01 -1.250e-01
-1.250e-01]
Normal force N:
[ 3.750e-01 1.750e-01 -2.500e-02 -1.250e-01 -1.250e-01
-1.250e-01]
We now plot the distribution of the solutions. In order to plot both dis-
placement and stress together in the same figure, we divide the solutions by
its corresponding absolute maximum value. Therefore, the unit of the value
on these curves are their corresponding maximum values.

1 # Plot the distribution curves:

2
3 x_data = [x, x] # put x coordinate values to a list
4 y_data = [ux/max(ux_max,1.e-9), sx/max(sx_max,1.e-9)] # put to a list
5 labels=['Displacement', 'Stress'] # define the label
6 gr.plot_fs(x_data, y_data, labels, 'x', 'u(x), $ $(x)', 'temp')
7 # x-label, y-label, file-name

Note in the foregoing examples (and some examples in the future), we

used unit values for the parameters involved in the problem. Users may use
the actual values for a given practical problem. Any unit system can be used,
but all the units for the same problem must be consistent.
The code given above can be used for many other problems. For example,
if we set l1 = 1.0 and l2 = 0 and keep others unchanged, we obtain the
reaction forces, displacement and stress for a fixed–fixed bar subjected to a
uniform body force. Readers may play with it.
 Solution to Axially Loaded Bars 273

It is a good exercise for readers to write a code to compute a similar case,

considering a fixed–fixed bar with a point force in the interior of the bar, as
shown in Fig. 7.3.

7.10 Example: Thermal stress in an overconstrained bar

Consider a bar with length 1.0 subjected to a temperature change of ΔT =

1.0. Let A = 1.0 and E = 1.0. Compute the reaction forces, the displacement
and stress distributed over the entire bar.

Solution: We use the same code function twobars bf() written earlier, but
with the following setting, so that we see the stress resulting purely by a
temperature change:

1 q1 = 0.0; q2 = 0 # no body force, consider only the thermal effects

2 l1 = 1.0; l2 = 0.0
3 E = 1.0; A = 1.0
4 = 1.0; dT = 1.0 # with temperature change dT: ΔT
5 aT = *dT # thermal expansion per length
6 RT = A*E*aT # Reaction force by temperature change
7
8 # Compute the reaction force at the right-end of the bar
9 fR = -(q1*l1**2 + q2*l2**2+2*q2*l1*l2)/(2*l1 + 2*l2) - RT
10 print(f"Reaction force at the right-end = {fR}")
11
12 dx = 0.02
13 x1 = [Link](0.0, l1+dx, dx) # coordinates in l1
14 x2 = [Link](0.0, l2+dx, dx) # coordinates in l2
15
16 # Compute the displacement and stress:
17 u1, 1 = twobars_bf(x1, A, E, l1, l2, q1, q2, fR, aT, barId=1)
18 u2, 2 = twobars_bf(x2, A, E, l1, l2, q1, q2, fR, aT, barId=2)
19
20 x = [Link]((x1, l1+x2)) # put the results in a single array
21 ux = [Link]((u1, u2))
22 sx = [Link](( 1, 2))
23
24 ux_max = [Link](abs(ux)) # find the absolute maximum value
25 sx_max = [Link](abs(sx))
26 print(f"The absolute maximum displacement: {ux_max}")
27 print(f"The absolute maximum stress: {sx_max}")
28
29 step = max(int(len(x)/5), 1)
30 print(f"Location x:\n {x[::step]}") # print out some results
31 print(f"Displacement ux:\n {ux[::step]}")
32 print(f"Stress :\n {sx[::step]}")
274 Mechanics of Materials: Formulations and Solutions with Python

Reaction force at the right end = −1.0

The absolute maximum displacement: 0.0
The absolute maximum stress: 1.0
Location x:
[ 0.000e+00 2.000e-01 4.000e-01 6.000e-01 8.000e-01
1.000e+00]
Displacement ux:
[ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00
0.000e+00]
Stress σ:
[-1.000e+00 -1.000e+00 -1.000e+00 -1.000e+00 -1.000e+00
-1.000e+00]

1 # Plot the distribution curves:

2
3 x_data = [x, x] # put x coordinate values to a list
4 y_data = [ux/max(ux_max,1.e-9), sx/max(sx_max,1.e-9)] # put to a list
5 labels=['Displacement', 'Stress'] # define the label
6 gr.plot_fs(x_data, y_data, labels, 'x', 'u(x), $ $(x)', 'temp')
7 # x-label, y-label, file-name

It is seen that the displacement is zero in the entire bar because it is fixed
at both ends. The stress is −1 which is induced by the temperature change.
This stress is purely caused by the temperature change because there are no
external forces applied on the bar.
 Solution to Axially Loaded Bars 275

It is a good exercise for readers to write a code to compute a similar case,

considering a fixed–fixed bar with a distributed force or point force in the
interior of the bar, as shown in Figs. 7.3 and 7.4, but also with a temperature
change.

7.11 Example: A composite bar subjected to both

mechanical and thermal loads

Consider a bar of length L consisting of three thin bars of homogeneous

isotropic materials, as shown in Fig. 7.5. The bars are all fixed on a wall at the
left ends and are rigidly connected at the right ends. The bars on the top and
bottom (bar-1) are identical with Young’s modulus E1 , thermal expansion
coefficient of α1 , and each with area A1 . The bar in the middle (bar-2) is
with Young’s modulus E2 , area A2 , and thermal expansion coefficient of α2 .
The data are given as follows: E1 = 200 GPa, E2 = 70 GPa, A1 = 5 mm2 ,
A2 = 8 mm2 and L = 100 mm, α1 = 11.0E −6 /◦ C, α2 = 22.0E −6 /◦ C,
ΔT = 300◦ C, and F = 50 N.
Considering only the displacement in the x-direction, complete the fol-
lowing tasks:

1. Consider only external force F applied at the right end of the bar, derive
the formulas for computing the normal stresses and the internal normal
forces in each of the three bars as well as their elongations.
2. Consider only temperature change ΔT over the entire bars, derive the
formulas for computing the normal stresses and the internal normal forces
in each of the three bars as well as their elongations.
3. Using the data given, compute numerical values for normal stresses, inter-
nal normal forces in each of the three bars, and the elongation of the bars,

Figure 7.5. A bar consisting of three thin bars of homogeneous isotropic material con-
nected by a rigid block.
276 Mechanics of Materials: Formulations and Solutions with Python

when the composite bar is subjected to both external force and temper-
ature changes.
4. Compute the principal stresses and the maximum stress and their direc-
tions in bar-1.

7.11.1 Solution to (1)

Consider only external force F applied at the right end of the bar and derive
the formulas for computing the normal stresses and the internal normal forces
in each of the three bars as well as their elongations.
Since the problem is statically indeterminate, using equilibrium condi-
tions alone is not enough to find the solution. We thus release the rigid
connector, and replace its interaction with the three bars by three unknown
forces, F1 , F2 , and F1 , as shown in Fig. 7.6.
For the free body diagram of the rigid block, we shall have the following
equilibrium equation:

−2F1 − F2 + F = 0 (7.43)

From the diagram for the three bars, we have

F1 L
u1L =
A1 E1
F2 L
u2L = (7.44)
A2 E2
Since these three bars are connected together by the rigid block, we shall
have

u1L = u2L (7.45)

This is known as the so-called continuity equation, implying in this case

that the structure is still in one piece and is not broken. These three bars are

Figure 7.6. Free body diagram of the bar consisting of three thin bars and the reactions
with the rigid block.
 Solution to Axially Loaded Bars 277

always attached to the rigid connector without gaps and overlaps. Equations
(7.44) and (7.45) give
A1 E1
F1 = F2 (7.46)
A2 E2
Substituting Eq. (7.46) into (7.43), we obtain
A2 E2
F2 = F (7.47)
2A1 E1 + A2 E2
Substituting Eq. (7.47) into (7.46) gives
A1 E1
F1 = F (7.48)
2A1 E1 + A2 E2
After these two forces are found, the displacement at x = L is found as
FL
u1L = u2L = (7.49)
2A1 E1 + A2 E2
This is the elongation of these three bars.
The strains in these three bars are
F
εbar−1 = εbar−2 = (7.50)
2A1 E1 + A2 E2
The stresses in these three bars are found as
E1 F
σbar−1 =
2A1 E1 + A2 E2
E2 F
σbar−2 = (7.51)
2A1 E1 + A2 E2
This process is simple enough to work out by hand, as shown. Simple codes
can be written to obtain the same results for checking to ensure no error is
made:
1 E1, A1, E2, A2, L = symbols('E_1, A_1, E_2, A_2, L', positive=True)
2 bar1F, bar2F, F, F1F, F2F = symbols(' _bar1F, _bar2F, F, F_1F, F_2F')
3 u1L, u2L, u1LF, u2LF = symbols('u_1L, u_2L, u_1LF, u_2LF')
4
5 u1L = F1F*L/(A1*E1) # displacementof bar-1 at x=L by force F
6 u2L = F2F*L/(A2*E2) # displacementof bar-2 at x=L by force F
7
8 slnFF = [Link]([-2*F1F-F2F+F, u1L-u2L], [F1F,F2F])
9 slnFF # solution of internal forces in bars by external Force F

A1 E1 F A2 E2 F
F1F : , F2F :
2A1 E1 + A2 E2 2A1 E1 + A2 E2
278 Mechanics of Materials: Formulations and Solutions with Python

1 sln F = { bar1F: slnFF[F1F]/A1, bar2F:slnFF[F2F]/A2}

2 sln F # solution of stresses in bars by external Force F

E1 F E2 F
σbar1F : , σbar2F :
2A1 E1 + A2 E2 2A1 E1 + A2 E2

1 slnUF = {u1LF: slnFF[F1F]*L/(A1*E1), u2LF:slnFF[F2F]*L/(A2*E2)}

2 slnUF # solution of displacements of bars at x=L, i.e. elongation by F

FL FL
u1LF : , u2LF :
2A1 E1 + A2 E2 2A1 E1 + A2 E2

7.11.2 Solution to (2)

Consider only temperature change ΔT over the entire bars and derive the
formulas for computing the normal stresses and the internal normal forces
in each of the three bars, as well as their elongations.
When the bars experiences temperature change ΔT , the displacements
in each of these bars shall have an additional term. Equation (7.44) becomes

F1 L
u1L = + α1 ΔT L
A1 E1
F2 L
u2L = + α2 ΔT L (7.52)
A2 E2

Using Eqs. (7.52) and (7.45) gives

F2 A1 E1
F1 = + A1 E1 (α2 − α1 )ΔT (7.53)
A2 E2

Since F = 0, Eq. (7.43) gives

−2F1 − F2 = 0 (7.54)

Substituting Eq. (7.54) into (7.53) gives

F1 A2 E2 = −2F1 A1 E1 + A1 E1 A2 E2 (α2 − α1 )ΔT (7.55)

A1 E1 A2 E2 (α2 − α1 )ΔT
F1 = (7.56)
2F1 A1 + A2 E2
 Solution to Axially Loaded Bars 279

Substituting it into Eq. (7.54), we find

2A1 E1 A2 E2 (α2 − α1 )ΔT

F2 = − (7.57)
2F1 A1 + A2 E2
The displacements are then found using Eq. (7.52):

A2 E2 (α2 − α1 )ΔT L
u1L = + α1 ΔT L
2E1 A1 + A2 E2
(7.58)
2A1 E1 (α2 − α1 )ΔT L
u2L = − + α2 ΔT L
2E1 A1 + A2 E2
These are the elongations. The stresses are found as
F1 E1 A2 E2 (α2 − α1 )ΔT
σbar−1 = =
A1 2E1 A1 + A2 E2
(7.59)
F2 2A1 E1 E2 (α2 − α1 )ΔT
σbar−2 = =−
A2 2E1 A1 + A2 E2
The codes to get the same formulas are as follows:

1 1, 2, T, F1T, F2T = symbols(' _1, _2, T, F_1T, F_2T') # T denotes ∆

2 u1LT, u2LT, bar1T, bar2T = symbols('u_1LT, u_2LT, _bar1T, _bar2T')
3
4 # Due to the rigid connector, T shall also results in forces: F1T & F2T
5 u1L = F1T*L/(A1*E1)+ 1*T*L # displacement of bar-1 at x=L by T
6 u2L = F2T*L/(A2*E2)+ 2*T*L # displacement of bar-2 at x=L by T
7
8 slnFT = [Link]([-2*F1T-F2T, u1L-u2L], [F1T,F2T]) # set F = 0
9 slnFT[F1T] = (slnFT[F1T]).simplify()
10 slnFT[F2T] = (slnFT[F2T]).simplify()
11
12 slnFT # solution of internal forces in bars by T

A1 A2 E1 E2 T (−α1 + α2 ) 2A1 A2 E1 E2 T (α1 − α2 )

F1T : , F2T :
2A1 E1 + A2 E2 2A1 E1 + A2 E2

It is clear that the forces result from the difference of the thermal expan-
sion coefficients of these two bars. If α1 = α2 , there will not be forces (and
hence thermal stresses) resulting from the temperature change:

1 sln T = { bar1T:(slnFT[F1T]/A1).simplify(), \
2 bar2T:(slnFT[F2T]/A2).simplify()}
3 sln T # solution of stresses in bars by T
280 Mechanics of Materials: Formulations and Solutions with Python

A2 E1 E2 T (−α1 + α2 ) 2A1 E1 E2 T (α1 − α2 )

σbar1T : , σbar2T :
2A1 E1 + A2 E2 2A1 E1 + A2 E2

1 slnUT = { u1LT: (slnFT[F1T]*L/(A1*E1)).simplify()+ 1* *L, \

2 u2LT: (slnFT[F2T]*L/(A2*E2)).simplify()+ 2* *L}
3 print(slnUT[u1LT].simplify() == slnUT[u2LT].simplify())
4 slnUT # solution of displacements of bars at x=L, i.e. elongation by T

True

A2 E2 LT (−α1 + α2 ) 2A1 E1 LT (α1 − α2 )

u1LT : + LT α1 , u2LT : + LT α2
2A1 E1 + A2 E2 2A1 E1 + A2 E2
It is clear that even if α1 = α2 , there will still be displacements in these
bars because of the thermal expansion.
The elongation of the bars must be the same, we double check using the
following code:
1 print("Same u_L?:", slnUT[u1LT].simplify()==slnUT[u2LT].simplify())

Same u_L?: True

7.11.3 Solution to (3)

Using the data given, compute numerical values for normal stresses, internal
normal forces in each of the three bars, and the elongation of the bars when
the bar is subjected to both external force and temperature changes.
We can find the formulas for effects from both force F and temperature
change T by simply superpositioning. Let us do this via coding:
1 F1, F2 = symbols('F1, F2') # total forces by both F and T
2 Fs = {F1:(slnFF[F1F]+slnFT[F1T]).simplify(), \
3 F2:(slnFF[F2F]+slnFT[F2T]).simplify()}
4 Fs # solution of forces in bars by both F and T

A1 E1 (−A2 E2 T (α1 − α2 ) + F ) A2 E2 · (2A1 E1 T (α1 − α2 ) + F )

F1 : , F2 :
2A1 E1 + A2 E2 2A1 E1 + A2 E2

1 bar1, bar2 = symbols(' _bar1, _bar2') # total stresses by F and T

2 bars = { bar1:(sln F[ bar1F]+sln T[ bar1T]).simplify(), \
3 bar2:(sln F[ bar1F]+sln T[ bar2T]).simplify()}
4 bars # solution of stresses in bars by both F and T
 Solution to Axially Loaded Bars 281

E1 (−A2 E2 T (α1 − α2 ) + F ) E1 · (2A1 E2 T (α1 − α2 ) + F )

σbar1 : , σbar2 :
2A1 E1 + A2 E2 2A1 E1 + A2 E2

1 u1L, u2L = symbols('u_1L, u_2L ') # total displacements at x=L by F&T

2 UsL = {u1L:(slnUF[u1LF]+slnUT[u1LT]).simplify(), \
3 u2L:(slnUF[u2LF]+slnUT[u2LT]).simplify()}
4 UsL # solution of displacements of bars at x=L by both F and T

L(2A1 E1 T α1 + A2 E2 T α2 + F ) L(2A1 E1 T α1 + A2 E2 T α2 + F )
u1L : , u2L :
2A1 E1 + A2 E2 2A1 E1 + A2 E2

1 print("Same u_L?:", UsL[u1L].simplify()==UsL[u1L].simplify())

Same u_L?: True

For given data, we can compute the numerical results with ease:
1 # Assign values to all parameters. The unites are in Pa, m^2, N, C
2
3 data = {E1:2e11, E2:7e10, A1:5/1000/1000, A2:8/1000/1000, \
4 L:100/1000, F:50, T:300, 1:11e-6, 2:22e-6}
5
6 Fs_val = [Link]([Fs[F1].subs(data), Fs[F2].subs(data)], dtype=float)
7 print("Normal Force in bars =",Fs_val,"(N)")
8
9 bars_val = [Link]([ bars[ bar1].subs(data), \
10 bars[ bar2].subs(data)], dtype=float)
11 print(f"Normal stress in bars = { bars_val} (Pa)")
12
13 UsL_val = [Link]([UsL[u1L].subs(data), \
14 UsL[u1L].subs(data)], dtype=float)
15 print(f"Displacement of bars at x=L: {UsL_val} (m)")

Normal Force in bars = [7.414e+02 -1.433e+03] (N)

Normal stress in bars = [1.483e+08 -1.766e+08] (Pa)
Displacement of bars at x=L: [4.041e-04 4.041e-04] (m)
It is clear that the equilibrium of all the forces and the continuity of the
displacements are satisfied. Using the same code, readers may compute the
results by setting F = 0 or T = 0 to examine the numerical results to see if
it makes sense.

7.11.4 Solution to (4)

Compute the principal stresses and the maximum stress and their directions
in bar-1.
282 Mechanics of Materials: Formulations and Solutions with Python

Since the stress in the bars is only in the x-direction, the shear stress
is zero in the x–y coordinate. Therefore, the largest principal stress is the
normal stress we found earlier. The other principal stresses are all zero:
1 PrincipalStress_1 = [Link]([ bars_val[0], 0., 0.])
2 print('Principal Stress in bar-1 = ', PrincipalStress_1,'(Pa)')
3 PrincipalStress_2 = [Link]([ bars_val[1], 0., 0.])
4 print('Principal Stress in bar-2 = ', PrincipalStress_2,'(Pa)')
5 print('The direction is all in the axial direction of the bars')

Principal Stress in bar-1 = [ 1.483e+08 0.000e+00 0.000e+00] (Pa)

Principal Stress in bar-2 = [-1.766e+08 0.000e+00 0.000e+00] (Pa)
The direction is all in the axial direction of the bars

7.12 Remarks
This chapter solves a class of problems of mechanics of materials: 1D bars
subjected to types of mechanical and thermal loadings. The bar is assumed
long and the loading is axial. The full set of solutions are obtained, including
displacements, stresses and strains. A number of useful Python codes are
provided and examined using various examples.
The following chapter solves another class of problem of structure mem-
bers: beams subjected to bending.

Reference
[1] G.R. Liu and S.S. Quek, The Finite Element Method: A Practical Course.
Butterworth-Heinemann, New York, 2013.
 Chapter 8

Solution to Bended Beams

1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.

2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 from continuum_mechanics.solids import sym_grad, strain_stress
11 from matplotlib import colors
12 from collections import OrderedDict
13 init_printing(use_unicode=True) # For latex-like quality printing
14 #np.set_printoptions(precision=4,suppress=True) # Digits in print-outs
15 np.set_printoptions(formatter={'float': '{: 0.4e}'.format})

Beams may be the most widely used structure members. When one sees a
structure, most likely there are some beam members in it. A beam takes
transverse loads and moments, resulting in deflections, rotations, and inter-
nal shear force and moments. This chapter develops equations that govern
the behavior of beams, methods and formulas and codes leading to solutions
to loaded beams in terms of deflection, rotation, and internal shear force and
moments.

283
284 Mechanics of Materials: Formulations and Solutions with Python

8.1 Governing equation, fourth-order differential

equation

8.1.1 Thin beam theory

Consider a thin beam member with its axial axis long the x-direction, as
shown in Fig. 8.1. Its cross-section dimensions are much smaller than its
length. It is similar to a 1D bar studied in the previous chapter in terms of
slender geometry. The difference is that a beam is subjected to forces in the
transverse y-direction, or moments with respect to z axis, resulting in bend-
ing deformation called deflection and internal shear force and moments.
The primary variable for the beam is the deflection, the displacement in
the y-direction v of its neutral axis, where the normal stress σxx is zero.
The deflection is primary because all other variables such as cross-section
rotation, normal stress, shear forces, and moments can be obtained from
it. Since the deflection v changes only along the x axis, it is essentially a
one-dimensional (1D) problem, after a proper simplification process.
Consider transverse loads are only in the y-direction. The defection will
not occur in the z-direction. Therefore, it can be treated as at most a 2D
problem, for which there are two unknown displacement components: u and
v. To further simplify the problem to 1D, we must find a way to relate one of
the displacements, say u, to v. This needs to be done in a way that captures
the major mechanics behavior of the beam under bending. The so-called
Euler–Bernoulli beam theory captures well such behavior of thin beams,
and hence most widely used. This theory makes two major assumptions:

Figure 8.1. A 1D beam member. It is subjected to forces in the transverse y-direction,

resulting in bending deformation.
 Solution to Bended Beams 285

1. The cross-section of the beam remains straight during bending.

2. The cross-section of the beam (perpendicular to its axis) remains perpen-
dicular to its axis during bending.

Figure 8.2 is a schematic diagram of a deflected beam.

8.1.2 Normal stress and the resulted moment

These two assumptions mean that the shear strain εxy is negligible. Other-
wise the cross-section and the neutral axis will not be perpendicular, when
the beam is bent. This leads to
 
1 ∂u ∂v
εxy = + =0 (8.1)
2 ∂y ∂x
∂v
where ∂x is the slope of the deflection. Due to the second assumption, the
rotation of the cross-section θ is the same as the slope of the deflection (see
Fig. 8.2). We thus have
∂v
θ= (8.2)
∂x
Equation (8.1) becomes
∂u
= −θ (8.3)
∂y

Figure 8.2. A thin beam under bending. This neutral axis that representing the beam
has a deflection, while the cross-section of the beam has a rotation. The slope of the
deflection and the rotation angle is the same due to the Euler–Bernoulli assumptions.
286 Mechanics of Materials: Formulations and Solutions with Python

Integration of Eq. (8.3) with respect to y gives

u = −yθ + c (8.4)

where c is an integral constant. Since the neutral axis (y = 0) does not

deform in the x direction, u = 0. This condition leads to c = 0. Equation
(8.4) becomes

∂v
u = −yθ = −y (8.5)
∂x

The displacement u is written in the derivative of deflection v. Thus, two

unknown displacements become only one, which is deflection v. We need to
write other variables also in terms of only v.
Using Eq. (8.5), the normal strain can now be evaluated using

∂u ∂2v
εxx = = −y 2 (8.6)
∂x ∂x

The normal strain εxx is written in terms of the second derivative of

deflection v. It is linear in y, implying that the normal strain distribution
along y is a straight line.
Note that there is only one independent variable x involved for beams
under bending. Therefore, the partial derivative can be replaced by total
derivative.
For this 1D beam, the normal stress relates the normal strain by Hooke’s
law, and hence can be obtained using

d2 v
σxx = Eεxx = −yE (8.7)
dx2

This means the normal stress changes also linearly with y, as shown in
Fig. 8.3.
The linearly distributed normal stress results in moment Mz on the cross-
section, which can be evaluated using,
  2
d v d2 v
Mz = − yσxx dA = E y dA 2
= EI z (8.8)
A dx2 dx2
 A  
Iz
 Solution to Bended Beams 287

Figure 8.3. Normal stress (positive) distribution along the cross-section of beams, and
the resultant moment (positive), following the standard sign convention.

or
d2 v
Mz = EIz (8.9)
dx2
Note that on a positive plane, positive moment must be counter-
clockwise. The minus sign is because the positive normal stress results in
negative moment. Equation (8.9) shows that the moment is related to the
deflection in the second derivative, in which Iz is the second moment of
area of the cross-section of the beam about z axis. It is calculated using

Iz = y 2 dA (8.10)
A

This is the general equation for computing Iz of the cross-section of a

beam. Its value dependents on shape and dimensions of cross-section over
which the integration is performed. Note that because y is squared, the
height of the cross-section influences significantly the Iz value.
Using Eqs. (8.9) and (8.7), the normal stress relates the moment as
Mz
σxx = −y (8.11)
Iz
This is a useful equation for computing normal stress on the cross-section
of the beam for given moment. We see that:
1. The normal stress is proportional to the moment on the cross-section,
implying that larger moment results in large normal stress.
288 Mechanics of Materials: Formulations and Solutions with Python

2. The normal stress is linear in y, and the largest stress is registered at ymax
or −ymax on the cross-section.
3. Increasing Iz reduces the stress. Since Iz relates to the shape of the cross-
section, it is thus important to design a good shape with large Iz value.

Since the moment relates to the deflection v in Eq. (8.9). The normal
stress is also related to displacement v.

8.1.3 On second moment of area of beam cross-section

The normal stress is inversely proportional to Iz of the cross-section, imply-
ing that a larger Iz reduces the normal stress. In other words, we would like
to have a beam with large Iz .
For beams with rectangular cross-section with height h and base b, as
shown in Fig. 8.4 (see also Fig. 8.1), Iz can be computed easily using
Eq. (8.10):
  h/2  h/2
1 h/2 bh3
Iz = y dA =
2
y bdy = 2
2
y 2 bdy = 2b y 3 = (8.12)
A −h/2 0 3 0 12

To make Iz larger, we can increase b and/or h. Since h has a cubic power,

increasing h is far more effective than increasing b for given same amount of
material. This means that we would like to make the beam as tall as possible,
which is what the engineering designer is doing. However, there is a limit to

Figure 8.4. Rectangular cross-section of a beam.

 Solution to Bended Beams 289

it because beam with too large h/b ratio will have stability issues, which will
be discussed in Chapter 12.
A better alternative is to design the beam cross-section with a better
shape. The objective is to move the material as farther as possible away
from the neutral axis of the beam. A typical such design is the I-beam. This
kind of study relates to the property of an area, which will be discussed in
more detail in Chapter 10.

8.1.4 Equilibrium equation for beams

We now need to establish the equilibrium equation of beams subjected to
transverse loading. This is done using the usual trick of free body diagram,
which is created by isolating a small cell from the beam with loads on. Figure
8.5 shows such a cell.
Here, ρ is the density of the material. The external body-force is assumed
to be by that can change along x. It carries a unit of [N/m]. When the
force is dynamic or the beam is subjected to dynamic excitation, there can
also be inertia force, which depends on the mass of the free-body and its
acceleration, as shown in Fig. 8.5. On the cross-section, there are a possible
moment and a shear force caused by body-force in the transversal direction.
Both moments and shear forces are internal because we see them only when
the cross-section is exposed. They both may vary along x, and thus have
increments corresponding to the coordinate increment dx.
First, all the moments about z axis must be in equilibrium. This leads to
1
dMz + V dx − by (x) − ρAv̈ dx dx = 0 (8.13)
2

Figure 8.5. A small free-body diagram from a beam subjected to loading, including
possible inertia force. All the forces (moments) shown are all positive by our sign conven-
tion.
290 Mechanics of Materials: Formulations and Solutions with Python

Since the third term is second order in the small dx, it can be neglected.
We obtain
dMz
= −V (8.14)
dx
This means that the shear force on the cross-section is the negated deriva-
tive of the moment.
Next, all these forces in the y direction must be in equilibrium. This gives
dV + by (x) − ρAv̈ dx = 0 (8.15)
or
dV
= −by (x) + ρAv̈ (8.16)
dx
Using Eq. (8.9), we found the shear force in terms of deflection:

d3 v
V = −EIz (8.17)
dx3
The shear force relates to the deflection in its third derivative. Substi-
tuting Eqs. (8.17) to (8.16), we finally arrived at the dynamic equilibrium
equation for beams:
d4 v
EIz + ρAv̈ = by (8.18)
dx4
For static cases, we have

d4 v by
= (8.19)
dx 4 EIz
where EIz is often called bending stiffness (per length) of the beam. Equation
(8.19) is the well-known beam equation based on the Euler–Bernoulli the-
ory, and is the governing equation for thin beams with only one unknown
function v. It is quite similar to Eq. (7.1), but it is a fourth-order differential
equation (DE). It can be solved through integration as long as the body
force by is integrable. The integration will result in four integral constants,
which need to be determined using boundary conditions (BCs). The process
is essentially the same as we did in the previous chapter for second-order
DEs, but a little more complicated because of the higher-order differential.
In this section, we put subscript z for I and M to indicate the bending
is with respect to the z-axis. In the following sections, we may drop z to
lighten the subscripts.
 Solution to Bended Beams 291

8.2 Boundary conditions for beams

8.2.1 Possible BCs for beams

Based on the thin beam theory, at a cross-section (a point on its neutral
axis to be more precise) has two displacement degrees of freedom (DoFs):
deflection v and rotation θ. Their corresponding force variables are the shear
force V and moment M . Thus, there are a total of four possible variables: v, θ,
V , and M , at any cross-section. There are always two paired complementary
relations. The first pair is v with V , implying that if we know v at a cross-
section, we do not in general know V , and if we know V we do not generally
know v. The same applies to the pair of θ with M . Thus, we can only pre-
scribe two variables at a cross-section.
Now, any given beam will have two boundaries: left- and right-ends. At
each of these two boundary points, there are two possible BCs. The possible
BCs for a boundary point can be given in Fig. 8.6. As seen, we cannot
pre-scribe any complementary pair of variables at a boundary point.
Since there will always be two boundary points for a given beam, we
always have four BCs. This is just the right number we need to solve the
fourth-order DE given in Eq. (8.19).

(a) (b)

(c) (d)

Figure 8.6. Possible boundary conditions for a beam based on the thin beam theory.
In many cases, these two variables are given as zero, but in general these can also be
pre-specified values.
292 Mechanics of Materials: Formulations and Solutions with Python

8.2.2 Admissible BCs for beams

It is important to note that in a set of four BCs, there must be at least
two conditions on these two displacement DoFs. This is because we must
constrain the beam for rigid body movements, or we will not have a unique
solution. For example, we cannot have both ends of the beam have condition
(d) shown in Fig. 8.6. Since the beam can move freely without any force
being generated. If condition (d) is given for one end, the other end must
be condition (a), which is the often-encountered cantilever beam. Also, (d)
cannot go with (b) or (c) because such cases have only one displacement
variable constrained. The other rigid motion is still possible.
On the other hand, we can have overprescribed with displacement DoFs
because our method is displacement based. It works naturally well for inde-
terminate beams. For example, it is perfectly fine for beams with both ends
clamped that is known as c-c condition. Simple-clamp (s-c) and roller-clamp
(r-c) are also fine, implying that unique solutions can be found.

8.3 Example: A cantilever beam subjected to a force at

its right-end

Considering a simplest case of cantilever beam fixed at the left-end and

subjected to a concentrated force F [N] at the right-end, without body force,
as shown in Fig. 8.7. The beam has length L and a uniform cross-section
with its second moment of area given as I.
We derive, by hand, the formulas for solutions to a beam as functions of
x, so that we can have a good idea on how Eq. (8.19) can be solved. We
will then write a Python code that can be used to derive formulas for more

Figure 8.7. A cantilever beam fixed at the left-end, subjected to a concentrated force
at the right-end.
 Solution to Bended Beams 293

complicated problems. The code follows essentially the same process, but
allows the computer to do the heavy lifting of derivation.
Solution: Since there is no body force, Eq. (8.19) is reduced to
d4 v
=0 (8.20)
dx4
The general solution of the deflection v that satisfies Eq. (8.20) can be
found by integration four times, which gives
1 1
v = c3 x3 + c2 x2 + c1 x + c0 (8.21)
6 2
where ci , i = 0, 1, 2, 3 are integral constants. Equation (8.21) is the general
solution for any beam without body force. It has a total of four constants
because the DE is of fourth order. For a given specified problem defined in
this example, we shall use the BCs to determine these constants.
At x = 0, the beam is fixed and hence we have the following BCs:
v=0
at x = 0 (8.22)
θ=0
To satisfy Eq. (8.22), c0 = 0 and c1 = 0. Equation (8.21) is reduced to
1 1
v = c3 x3 + c2 x2 (8.23)
6 2
At x = L, we have the following BCs:
V =F
at x = L (8.24)
M =0
Using Eqs. (8.17) and (8.9), we have
−EIc3 = F
at x = L (8.25)
EI(c3 L + c2 ) = 0
which gives
⎧
⎪ F
⎪
⎨c3 = − EI
at x = L (8.26)
⎪
⎪ F L
⎩c2 =
EI
Substituting the foregoing equation to Eq. (8.23), we obtain the solution
for the deflection as follows:
F 3 FL 2
v=− x + x (8.27)
6EI 2EI
294 Mechanics of Materials: Formulations and Solutions with Python

At the right-end of the beam, x = L, the deflection is found as

L3
vL = F (8.28)
3EI
which is the largest deflection on the cantilever beam subject to point force
at its free end.
Once the deflection is found, we can easily find all other solutions (typical
for a displacement method). The solution for the rotation is obtained
using Eq. (8.2):
∂v F 2 FL
θ= =− x + x (8.29)
∂x 2EI EI
At the right-end of the beam, x = L, the rotation is found as

L2
θL = F (8.30)
2EI
The moment is found using Eq. (8.9):
d2 v
M = EI = F (L − x) (8.31)
dx2
It changes linearly along the beam. The shear force is found using
Eqs. (8.17) and (8.27):
d3 v
V = −EI =F (8.32)
dx3
It is constant along the beam, which can be easily confirm by equilibrium.
The correctness of these solutions can be confirmed using the general
solver: gr.solver1D4() to be developed in the next section.
The normal stress is found using Eq. (8.11):
F (L − x)
σxx = −y (8.33)
I
The same procedure can find solutions for beams subjected to body forces,
as long as the force function is integrable analytically. We shall use Sympy
to do the derivation for us.

8.4 A Python beam solver

Our Python beam solver, solver1D4(), is developed based on the following

considerations:
 Solution to Bended Beams 295

1. Since Eq. (8.19) is a fourth-order DE, and we need to integrate it four

times using Sympy. This results in 4 integral constants. For beams of
length l, we have two boundaries at x = 0 (left-end) and x = l (right-end).
At each boundary, we have two of the four possible variables: deflection
v, rotation θ, shear force V and moment M . Therefore, we shall have just
enough 4 equations to determine these integral constants. After these 4
dv d2 v d3 v
constants are determined, we can have analytic formulas for v, dx , dx2 , dx3 ,
d4 v
and dx4 . These formulas give the full solution set in a Python list.
2. The solver1D4() accepts the following combinations of BCs: c-c, s-c, c-s,
f-c, and c-f. In which c stands for clamp, s for simple, and f for free. These
set of BCs are all admissible.
3. As discussed before, because the mechanics problem is complimentary,
the displacement variables are naturally paired with the force variables.
For example, for the s-c BC, v and M are specified at x = 0, and v and
θ are specified at x = l. This gives 4 equations for determining these 4
integral constants. Their corresponding counterparts V and θ at x = 0,
and V and M at x = l can then be computed using the output formulas
from solver1D4().
4. The computation is done symbolically. Hence, we can set all the variables
specified on the BCs as symbolic variables, and the derived general formu-
las are in closed-form. Such a set of formulas can then be used to compute
many problems of beams with complicated loading, with multiple spans,
and even frame structures. If some or all of these variables are zero, we
can simply set so using [Link](). Therefore, it is flexible and useful.
5. The shear force and moment used in solver1D4() are internal. When
replacing those by the applied external force/moment at the boundary,
the standard sign convention applies: on the right-boundary (positive
boundary), the directions of the internal force/moment are the same as
these of the external point force and moment applied there; On the left-
boundary (negative boundary), the directions the internal force/moment
are opposite to these of the externally applied point force/moment. This
is because positive externally applied force always follows the direction
of the coordinates, and positive externally applied moment always fol-
lows the right-hand rule. Internal forces/moments, however, follow the
direction of the surface the forces/moments.

8.4.1 Python code of beam solver

The following code uses a step-by-step approach to perform the integration
four times using Sympy, and then uses BCs of symbolic variables to deter-
mine the integral constants. The code is self-explanatory with comments
given.
296 Mechanics of Materials: Formulations and Solutions with Python

1 def solver1D4(E, I, by, l, v0, 0, vl, l, V0, M0, Vl, Ml, key='c-c'):
2
3 '''Solves the Beam Equation for integrable distributed body force:
4 u,x4=-by(x)/EI, with various boundary conditions (BCs):
5 s-s, c-c, s-c, c-s, f-c, c-f.
6 Input: EI: bending stiffness factor; by, body force; l, the length
7 of the beam; v0, 0, V0, M0, deflection, rotation,
8 (internal) shear force, (internal) moment at x=0;
9 vl, l, Vl, Ml, are those at x=l.
10 Return: u, v_x, v_x2, v_x3, v_x4 up to 4th derivatives of v
11 '''
12 x = [Link]('x')
13 c0, c1, c2, c3 = [Link]('c0, c1, c2, c3') #integration constant
14 EI = E*I # I is the Iz in our chosen coordinates.
15
16 # Integrate 4 times:
17 v_x3= [Link](by/EI,(x, 0, x))+ c0 #ci: integration constant
18 v_x2= [Link]( v_x3,(x, 0, x))+ c1
19 v_x = [Link]( v_x2,(x, 0, x))+ c2
20 v = [Link]( v_x,(x, 0, x))+ c3
21
22 # Solve for the 4 integration constants:
23 if key == "s-s":
24 cs=[Link]([[Link](x,0)-v0, v_x2.subs(x,0)-M0/EI,
25 [Link](x,l)-vl, v_x2.subs(x,l)-Ml/EI],[c0,c1,c2,c3])
26 elif key == "c-c":
27 cs=[Link]([[Link](x,0)-v0, v_x.subs(x,0)- 0,
28 [Link](x,l)-vl, v_x.subs(x,l)- l], [c0,c1,c2,c3])
29 elif key == "s-c":
30 cs=[Link]([[Link](x,0)-v0, v_x2.subs(x,0)-M0/EI,
31 [Link](x,l)-vl, v_x.subs(x,l)- l], [c0,c1,c2,c3])
32 #print('solver1D4:',cs[c0],cs[c1],cs[c2],cs[c3])
33 elif key == "c-s":
34 cs=[Link]([[Link](x,0)-v0, v_x.subs(x,0)- 0,
35 [Link](x,l)-vl, v_x2.subs(x,l)-Ml/EI],[c0,c1,c2,c3 ])
36 elif key == "c-f":
37 cs=[Link]([[Link](x,0)-v0, v_x.subs(x,0)- 0,
38 v_x3.subs(x,l)+Vl/EI, v_x2.subs(x,l)-Ml/EI], [c0,c1,c2,c3])
39 elif key == "f-c":
40 cs=[Link]([v_x3.subs(x,0)+V0/EI, v_x2.subs(x,0)-M0/EI,
41 [Link](x,l)-vl, v_x.subs(x,l)- l], [c0,c1,c2,c3])
42 else:
43 print("Please specify boundary condition type.")
44 [Link]()
45
46 # Substitute the constants back to the integral solutions
47 v = [Link]({c0:cs[c0],c1:cs[c1],c2:cs[c2],c3:cs[c3]})
48 v = [Link]().simplify().expand()
49
50 v_x = v_x.subs({c0:cs[c0],c1:cs[c1],c2:cs[c2],c3:cs[c3]})
51 v_x = v_x.expand().simplify().expand()
52
53 v_x2=v_x2.subs({c0:cs[c0],c1:cs[c1],c2:cs[c2],c3:cs[c3]})
 Solution to Bended Beams 297

54 v_x2=v_x2.expand().simplify().expand()
55
56 v_x3=v_x3.subs({c0:cs[c0],c1:cs[c1],c2:cs[c2],c3:cs[c3]})
57 v_x3=v_x3.expand().simplify().expand()
58
59 v_x4 = [Link](v_x3,x).expand()
60
61 print("Outputs form solver1D4(): v, , M, V, qy")
62 return v,v_x,(EI*v_x2).expand(),(-EI*v_x3).expand(),(v_x4).expand()

Let us use this function to find solutions for some bending problems. We
define first necessary symbolic variables.
1 E, I, EI, A, l, = symbols('E, I, EI, A, l, ', nonnegative=True)
2 x, q = symbols('x, q ') # geometry & force
3 c0, c1, c2, c3 = symbols('c0, c1, c2, c3') #integration constant
4
5 # for displacement boundary conditions (DBCs):
6 v0, 0, vl, l = symbols('v0, 0, v_l, _l') # DBCs
7
8 # for force boundary conditions (FBCs):
9 V0, M0, Vl, Ml = symbols('V0, M0, V_l, M_l') # FBCs
10
11 by = [Link]('b_y')(x) # body force function

8.4.2 Example: Clamped-clamped beams

Consider beams of length l with clamped-clamped (c-c) displacement bound-
ary conditions (DBCs) given in symbolic variables, as shown in Fig. 8.8. The
cross-section has a second moment of area I. It is made of material with
Young’s modulus E. It is subjected to a distributed force q(x), with the
x-axis starting at the left-end of the beam and being leftwards. Derive for-
mulas for solutions, including deflection, rotation, moment, and shear force
in the beam.

Figure 8.8. A beam with clamped-clamped (c-c) displacement boundary conditions

(DBCs). It is subjected to a distributed force q(x).
298 Mechanics of Materials: Formulations and Solutions with Python

We consider first uniformly distributed force.

1 # The title strings are used later for plots.

2 title = [["Deflection", "Rotation"], ["Moment", "Shear force"]]
3
4 # Consider constant distributed force
5 by = q # [N/m], one may try other force functions
6
7 v = solver1D4(E, I, by, l, v0, 0, vl, l, V0, M0, Vl, Ml,key='c-c')
8 print(f'Is solution correct? {by/E/I==v[4]}; The solution v(x) is:')
9 v[0] # solution of general displacement function

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? True; The solution v(x) is

2x2 θ0 x2 θl 3v0 x2 3vl x2 x3 θ0 x3 θl 2v0 x3 2vl x3

v0 + xθ0 − − − 2 + 2 + 2 + 2 + 3 − 3
l l l l l l l l
l2 qx2 lqx3 qx4
+ − +
24EI 12EI 24EI

This expression is a little lengthy with all the symbolic variables. It is a

general solution and hence is useful to many beam problems. In the solution
expression, v0 , θ0 are the deflection and rotation at the left-end, and vl , θl are
the deflection and rotation at the right-end of the beam. This allows users
to replace these variables at the boundary with known values. For example,
if these variables are all zero, we can simply set so, as in the following code.

1 # Set all the pre-scribed displacements to zero:

2
3 vx=[v[i].subs({v0:0, 0:0,vl:0, l:0}) for i in range(len(v))]
4 vx # Outputs are: v, , Mx, Vx, qy


l2 qx2 lqx3 qx4 l2 qx lqx2 qx3
− + , − + ,
24EI 12EI 24EI 12EI 4EI 6EI

l2 q lqx qx2 lq q
− + , − qx,
12 2 2 2 EI
This is a set of solutions that may be found in textbooks or other open
literature. A better presentation of these solutions is in a so-called natural
coordinate system. We can obtain these using the following code.

1 # Convert to the natural coordinate: = x/l (dimensionless)

2 v = [vx[i].subs({x: *l}) for i in range(len(vx))]
3 v
 Solution to Bended Beams 299


l4 qξ 4 l4 qξ 3 l4 qξ 2 l3 qξ 3 l3 qξ 2 l3 qξ
− + , − + ,
24EI 12EI 24EI 6EI 4EI 12EI

l2 qξ 2 l2 qξ l2 q lq q
− + , −lqξ + ,
2 2 12 2 EI
The above results can be written in the following easy-to-manage forms.
 
l4 q ξ4 ξ3 ξ2 l4 q
v(ξ) = − + = v̄(ξ)
EI
 24 12 24 EI
  
displacement unit v̄(ξ)
 
l3 q ξ3 ξ2 ξ l3 q
θ(ξ) = − + = θ̄(ξ)
EI
 6 4 12 EI
  
rotation unit
θ̄(ξ)
  (8.34)
l2 q ξ2 ξ 1 l2 q
M (ξ) = − + = M̄ (ξ)
EI
 2 2 12 EI
  
moment unit
M̄ (ξ)
 
1
V (ξ) = ql −ξ + = qlV̄ (ξ)
 2
force unit   
V̄ (ξ)

where ξ = x/l is the so-called natural coordinate, which is normalized coor-

dinate by the length of the beam l. Thus, ξ is dimensionless and varies in
[0 1]. Each variable has its own unit. For example, the unit for the displace-
l4 q
ment is EI and it will be naturally with a length unit (m). The bar on top
of a variable stands for dimensionless function depending only on ξ ∈ [0 1],
and hence convenient to analyze, especially in symbolic computations. For
example, we need only to deal with the dimensionless function of variables to
find the locations of the maximum or minimum values. In such an analysis
we can set all the variable units to 1, without affecting the generality. We will
practice this in the following analysis. After the solutions are found using
the dimensionless functions, one can re-install physical solutions simply by
multiplying the corresponding units, using Eq. (8.34).

8.4.3 Use of the Sympy differential equation solver

Alternatively, one my use the built-in [Link]() to solve the differential
equation directly.
300 Mechanics of Materials: Formulations and Solutions with Python

1 E, I, EI, l = symbols('E, I, EI, l', nonnegative=True)

2 x, q, M0 = symbols('x, q, M0 ')
3 v0, vl, 0, l = symbols('v0, v_l, 0, _l')
4 vv = [Link]('vv')(x) # deflection of the beam, v (m)
5
6 # Consider the distributed force is a constant q
7 by = q #[Link]([Link]*x)
8
9 # Define the 2nd differential equation
10 diff_eq = [Link](x, 4) - by/EI
11
12 # Define the boundary conditions, consider clamped-clamped, "c-c":
13 BCs={[Link](x,0): v0, [Link](x).subs(x,0): 0,
14 [Link](x,l): vl, [Link](x).subs(x,l): l}
15
16 #simple-simple supports,"s-s"
17 #BCs={[Link](x,0): v0, [Link](x,2).subs(x,0): M0/EI,
18 # [Link](x,l): vl, [Link](x,2).subs(x,l): Ml/EI}
19
20 # Solve the differential equation using [Link]()
21 solution = [Link](diff_eq, vv, ics=BCs) # This is an [Link] object
22
23 # check whether the DE is satisfied.
24 print(f'Is solution correct? {[Link](x,4) == by/EI}')
25 [Link]() # solution

Is solution correct? True

2x2 θ0 x2 θl 3v0 x2 3vl x2 x3 θ0 x3 θl 2v0 x3

v0 + xθ0 − − − 2 + 2 + 2 + 2 + 3
l l l l l l l
2vl x3 l2 qx2 lqx3 qx4
− + − +
l3 24EI 12EI 24EI

The solution obtained is the same as that using method 1. Since our beam
solver given in method 1 has all the different BCs built in, it is much easier
to use. We will use solver1D4() in the following examples. For conveniences,
we write a few more utility codes.

8.5 Python code for finding the maximum

and minimum values

We write a code to find the maximum and minimum values of a given

function.
 Solution to Bended Beams 301

1 def roundS(expr, n_d): # to be used in maxminS()

2 '''To limit the number of digits to keep in a variable.
3 Usage: roundS(expr, n_d), where n_d: number of digits to keep.
4 '''
5 return [Link]({[Link]():round(n,n_d)
6 for n in [Link]([Link])})

1 def maxminS(f, title="value"):

2 '''Find maximum location and values of a symbolic function f.
3 '''
4 ndg = 8 # number of digits
5 df = [Link](f, x)
6 ddf = [Link](df, x)
7 df0_points = [Link](df, x) #find stationary points
8 df0s = [roundS([Link](),ndg) for point in df0_points]
9 print(f"Stationary points for {title}: {df0s}")
10
11 for point in df0s:
12 fv = roundS(([Link](x,point)).evalf(), ndg)
13 ddfv = [Link]({x:point})
14 if ddfv < 0:
15 print(f"At x={point}, local maximum {title}={fv}")
16 elif ddfv > 0:
17 print(f"At x={point}, local minimum {title}={fv}")
18 else:
19 print(f"At x={point}, max or min {title}={fv}")
20
21 [Link](0) # Add in points on the boundaries
22 [Link](1)
23 fs_df0 = [[Link](x, point).evalf() for point in df0s]
24 f_max = roundS(max(fs_df0), ndg)
25 x_max=[pnt for pnt in df0s if [Link]([Link](x,pnt),max(fs_df0))]
26
27 print(f"\nAt x={x_max}, Max {title}={f_max}\n")

Since the location of the extreme values, maximum and/or minimum val-
ues, depends only on the function part of the solution. We can set all the
parameters to unity and then find these extreme values and their correspond-
ing locations.

1 # Set all factors to unit to the solution obtained using beam solver.
2 dic = {E:1,I:1,l:1,q:1,v0:0, 0:0,vl:0, l:0}
3 vx=[v[i].subs(dic) for i in range(len(v))]
4 vx # Outputs are: v, , Mx, Vx, qy

 
x4 x3 x2 x3 x2 x x2 x 1 1
− + , − + , − + , − x, 1
24 12 24 6 4 12 2 2 12 2
302 Mechanics of Materials: Formulations and Solutions with Python

1 maxminS(vx[0], title="deflection")
2 maxminS(vx[1], title="rotation")

Stationary points for deflection: [0, 0.50000000, 1.00000000000000]

At x=0, local minimum deflection=0
At x=0.50000000, local maximum deflection=0.00260417
At x=1.00000000000000, local minimum deflection=0

At x=[0.50000000], Max deflection=0.00260417

Stationary points for rotation: [0.21132487, 0.78867513]

At x=0.21132487, local maximum rotation=0.00801875
At x=0.78867513, local minimum rotation=-0.00801875

At x=[0.21132487], Max rotation=0.00801875

1 maxminS(vx[2], title="moment")
2 maxminS(vx[3], title="shear force")

Stationary points for moment: [0.50000000]

At x=0.50000000, local minimum moment=-0.04166667

At x=[0, 1], Max moment=0.08333333

Stationary points for shear force: []

At x=[0], Max shear force=0.50000000

8.6 Python code for plotting the solution marked with

extreme values

The most effective way to analyze the solution obtained is to plot it in

graphs. The following code plots the distribution along the beam for all four
solutions: deflection, rotation, moment and shear force.

1 # plot the distribution of the solutions:

2
3 for i in range(len(title)):
4 gr.plot2curveS(vx[2*i:2*(i+1)], 0., 1., title=title[i])
5
6 #fig.tight_layout()
7 #[Link]('images/beam_cq.png', dpi=500) # save the plot to file
8 [Link]()
 Solution to Bended Beams 303

Maximum Deflection value=2.604e-03, at x=0.5

Maximum Rotation value=8.018e-03, at x=0.21
Extreme Deflection values=[ 0.000e+00 2.604e-03 0.000e+00],
at x=[ 0.000e+00 5.000e-01 1.000e+00]
Critical Rotation values=[ 0.000e+00 -6.939e-18 -1.388e-17],
at x=[ 0.000e+00 5.000e-01 1.000e+00]
Deflection values at boundary =(0.0, 0.0)
Rotation values at boundary =(0.0, -1.3877787807814457e-17)

Maximum Moment value=8.333e-02, at x=0.0

Maximum Shear force value=5.000e-01, at x=0.0
Extreme Moment values=[-4.167e-02],
at x=[ 5.000e-01]
Critical Shear force values=[ 0.000e+00],
at x=[ 5.000e-01]
Moment values at boundary =(0.08333333333333333, 0.08333333333333333)
Shear force values at boundary =(0.5, -0.5)

8.7 Solutions at specific locations

We can obtain solutions at different specific locations. This is useful not

only for obtaining the desired solutions, but also for building up equations
for solving more complicated problems. The codes are as follows:

1 printM(v[0].subs(x,l), 'Deflection at x=l',n_dgt=4)

Deflection at x=l : vl

1 printM(v[0].subs(x,l/2), 'Deflection at x=l/2:',n_dgt=4)

lθ0 lθl v0 vl l4 q
Deflection at x=l/2: − + + +
8 8 2 2 384EI
304 Mechanics of Materials: Formulations and Solutions with Python

1 printM(v[0].subs({x:l/2,v0:0, 0:0, vl:0, l:0}),

2 'Deflection at x=l/2 when all DBCs are set to zero:',n_dgt=4)

Deflection at x=l/2 when all DBCs are set to zero:

l4 q
384EI
This is the well-known formula for computing the largest deflection in a
c-c beam subject to uniformly distributed force.
1 printM((v[2].subs(x,0)).expand(), 'Moment at x=0:')

4EIθ0 2EIθl 6EIv0 6EIvl l2 q

Moment at x=0: − − − + +
l l l2 l2 12

1 printM((v[2].subs(x,l)).expand(), 'Moment at x=l:')

2EIθ0 4EIθl 6EIv0 6EIvl l2 q

Moment at x=l: + + − +
l l l2 l2 12

1 printM((v[2].subs(x,l/2)).expand(), 'Moment at x=l/2:')

EIθ0 EIθl l2 q
Moment at x=l/2: − + −
l l 24
1 printM((v[3].subs(x,l)).expand(), 'Shear force at x=l:')

6EIθ0 6EIθl 12EIv0 12EIvl lq

Shear force at x=l: − − − + −
l2 l2 l3 l3 2

1 printM((v[3].subs({x:l,v0:0, 0:0, vl:0, l:0})).expand(),

2 'Shear force at x=l when all fixed:')

lq
Shear force at x=l when all fixed: −
2

1 printM(v[4], 'Body force on the beam:')

Body force on the beam:

q
EI
This recovers the right-hand side of the fourth-order DE for beams.

8.8 Example: Simple-simple supported beams

Consider beams of length l with cross-section having a second moment of
area I. It is made of material with Young’s modulus E. Its boundaries are
simple-simple supports (s-s), as shown in Fig. 8.9.
 Solution to Bended Beams 305

Figure 8.9. A beam with simple-simple (s-s) displacement boundary conditions (DBCs).
It is subjected to a distributed force q(x) over the beam and moment ML at its left- and
MR at its right-end.

Solution: Use our solver: solver1D4() to obtain the general solutions for the
beam, considering uniformly distributed force.

1 v = solver1D4(E, I, by, l, v0, 0, vl, l, V0, M0, Vl, Ml, key='s-s')

2
3 print(f'Is solution correct? {by/E/I==v[4]}; The solution u(x) is:')
4 v[0] # solution of general displacement function

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? True; The solution u(x) is

v0 x vl x M0 lx M0 x2 M0 x3 Ml lx Ml x3 l3 qx lqx3 qx4
v0 − + − + − − + + − +
l l 3EI 2EI 6EIl 6EI 6EIl 24EI 12EI 24EI
In the solution expression, v0 , M0 are the deflection and (internal)
moment at the left-end, and vl , Ml are those at the right-end of the beam.
This allows users to replace these variables at the boundary with known val-
ues. For example, if v0 = 0, M0 = −ML , and vl = 0, Ml = MR , where ML
and MR are the externally applied moments at the beam ends, as shown in
Fig. 8.9, we can simply set so, as in the following code.

1 ML, MR = symbols('M_L, M_R')

2 dic = {v0:0, vl:0, Ml:MR, M0:-ML}
3 # at negative boundary ⇑, sign reversal
4 vM=[v[i].subs(dic) for i in range(len(v))]
5 vM[0] # Outputs v

ML lx ML x2 ML x3 MR lx MR x3 l3 qx lqx3 qx4
− + − + + − +
3EI 2EI 6EIl 6EI 6EIl 24EI 12EI 24EI
Note that setting M0 = −ML with a minus sign is due to our standard
sign convention. At the left-end, the boundary is in the negative x-direction,
and hence the internal moment is clockwise. However, the external moment
ML is counterclockwise (positive), as shown in Fig. 8.9. Therefore, we shall
306 Mechanics of Materials: Formulations and Solutions with Python

set M0 = −ML . Let us we check the solution of moment distributed along

the beam:

1 [Link](vM[2], "Moment solution in the beam as a function of x:")

Moment solution in the beam as a function of x:

ML x MR x lqx qx2
−ML + + − +
l l 2 2
It is clear that at x = 0, we obtain the correct result of M0 = −ML . We
can confirm also the moment solution at x = l, by setting q = 0, Ml = 0:

1 [Link](vM[2].subs({x:l, ML:0, q:0}), "Moment solution at x=l:")

Moment solution at x=l: MR

We obtain again the correct result. Let us now check the solution for the
shear force.
1 v[3].subs(dic
) # solution of general shear force function

ML MR lq
− − + − qx
l l 2
If q = 0, the applied moment MR (or ML ) results in constant shear force
of V = MR /l (or ML /l) along the beam. If MR = ML = 0, the shear force
is a linear function of x, at x = 0, V = lq/2, and at x = l, V = −lq/2. All
the results are correct.
Let us now check the solution for deflection.

1 # Check the solution at the boundary:

2 printM(v[0].subs(x,l), 'Deflection at x=l:',n_dgt=4)

Deflection at x=l: vl

We obtain the correct solution (which were set as the boundary condi-
tion). The distribution of the deflection over the beam by only q is found as
follows.

1 [Link](v[0].subs({v0:0, vl:0,M0:0, Ml:0}), "Deflection v(x):")

l3 qx lqx3 qx4
Deflection v(x): − +
24EI 12EI 24EI
 Solution to Bended Beams 307

This is the well-known formula for computing the deflection function

in an s-s beam subject to uniformly distributed force. It is a fourth-order
polynomial.

1 # get the solution at a location within the beam:

2 printM(v[0].subs(dic).subs(x,l/2), 'Deflection at x=l/2:',n_dgt=4)

ML l2 MR l2 5l4 q
Deflection at x=l/2: − +
16EI 16EI 384EI

1 # get the deflection at the middle span by only distributed force q:

2 printM(v[0].subs(dic).subs({ML:0, MR:0,x:l/2}), 'Deflection at x=l/2:')

Deflection at x=l/2:
5l4 q
384EI

This is the well-known formula for computing the largest deflection in an

s-s beam subject to uniformly distributed force. It is five times that of the
c-c beam.

1 # get the rotation at beam's left-end by only distributed force q:

2 printM(v[1].subs(dic).subs({ML:0, MR:0,x:0}), 'Rotation at x=s:')

Rotation at x=s:
l3 q
24EI

This is the well-known formula for computing the largest rotation in an

s-s beam subject to uniformly distributed force, which is registered at the
left-end (or at the right-end, but with a negative sign).

1 # Check whether the governing equation is satisfied:

2 [Link](v[4], 'Body force on the beam:')

Body force on the beam:

q
EI

1 # Get the distribution of the solutions, by setting variables values:

2 vx=[v[i].subs({E:1,I:1,l:1,q:1,v0:0,vl:0,Ml:0,M0:0})
3 for i in range(len(v))]
4 vx
 
x4 x3 x x3 x2 1 x2 x 1
− + , − + , − , − x, 1
24 12 24 6 4 24 2 2 2
308 Mechanics of Materials: Formulations and Solutions with Python

1 # Plot the distributions of solutions.

2 for i in range(len(title)):
3 gr.plot2curveS(vx[2*i:2*(i+1)], 0., 1., title=title[i])
4
5 [Link]()

Maximum Deflection value=1.302e-02, at x=0.5

Maximum Rotation value=4.167e-02, at x=0.0
Extreme Deflection values=[ 1.302e-02],
at x=[ 5.000e-01]
Critical Rotation values=[-6.939e-18],
at x=[ 5.000e-01]
Deflection values at boundary =(0.0, 0.0)
Rotation values at boundary =(0.041666666666666664, -0.04166666666666668)

Maximum Moment value=0.000e+00, at x=0.0

Maximum Shear force value=5.000e-01, at x=0.0
Extreme Moment values=[-1.250e-01],
at x=[ 5.000e-01]
Critical Shear force values=[ 0.000e+00],
at x=[ 5.000e-01]
Moment values at boundary =(0.0, 0.0)
Shear force values at boundary =(0.5, -0.5)

Let us get the formulas for rotations of the s-s beam.

1 v=solver1D4(E, I, by, l, v0, 0, vl, l, V0, M0, Vl, Ml, key = 's-s')
2 print(f'Is solution correct? {by/E/I==v[4]}; The solution u(x) is: ')
3 v[1] # solution of general function for rotation
 Solution to Bended Beams 309

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? True; The solution u(x) is

v0 vl M0 l M0 x M0 x2 Ml l Ml x2 l3 q lqx2 qx3
− + − + − − + + − +
l l 3EI EI 2EIl 6EI 2EIl 24EI 4EI 6EI

1 printM(v[1].subs(dic).subs(x,0), 'Rotation at x=0:',n_dgt=4)

ML l MR l l3 q
Rotation at x=0: − +
3EI 6EI 24EI

We found correct solutions of rotation at x = 0, which are resulting from

the external loads.

1 printM(v[1].subs(dic).subs(x,l), 'Rotation at x=l:',n_dgt=4)

ML l MR l l3 q
Rotation at x=l: − + −
6EI 3EI 24EI

1 # Deflection solutions when setting all the fixed DBCs to 0:

2 v_ss = v[0].subs({v0:0, 0:0, vl:0, l:0})
3
4 printM(v_ss, 'Deflection function when all DBCs=0:',n_dgt=4)

Deflection function when all DBCs=0:

M0 lx M0 x2 M0 x3 Ml lx Ml x3 l3 qx lqx3 qx4
− + − − + + − +
3EI 2EI 6EIl 6EI 6EIl 24EI 12EI 24EI

1 # Deflection when setting all the fixed DBCs to 0, and q=0, M0=0
2 v_ss_Ml = v[0].subs({v0:0, 0:0, vl:0, l:0, q:0, M0:0}).subs(dic)
3
4 printM(v_ss_Ml, 'Deflection function, only Ml is applied at x=l:')

MR lx MR x3
Deflection function, only Ml is applied at x=l: − +
6EI 6EIl

This is the formula for computing the deflection of an s-s beam when
a concentrated moment is applied at the right-end. As seen, solver1D4()
can produce all kinds of solutions for beams. Readers may play with it
further.
310 Mechanics of Materials: Formulations and Solutions with Python

8.9 Example: Simple-clamped beams

Consider beams of length l with cross-section having a second moment of

area I. It is made of material with Young’s modulus E. It is simple-clamped
(s-c) at its two ends, as shown in Fig. 8.10.

Solution: Use our solver: solver1D4() to obtain the general solutions for the
beam, considering uniformly distribute force q.

1 ML, q = symbols('M_L, q')

2 by = q
3 v=solver1D4(E, I, by, l, v0, 0, vl, l, V0, M0, Vl, Ml, key= 's-c')
4 print(f'Is solution correct? {by/E/I==v[4]}; The solution u(x) is:')
5 dic = {v0:0,vl:0, l:0,M0:-ML} # for setting boundary conditions
6 # at negative boundary ⇑, sign reversal
7
8 v[0].subs(dic) # solution for deflection as function of x

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? True; The solution u(x) is

ML lx ML x2 ML x3 l3 qx lqx3 qx4
− + + − +
4EI 2EI 4EIl 48EI 16EI 24EI
Similar to the previous example, one can find various solutions for the
beam by setting desired values to these variables. For example, solutions at
x = l caused by the body force q can be found as follows:

1 M_l = [v[i].subs(dic).subs({x:0, ML:0}) for i in range(len(v))]

2 M_l # solutions for v, , M, V, qy at x=0

 
l3 q 3lq q
0, , 0, ,
48EI 8 EI

Figure 8.10. A beam with simple-clamped (s-c) displacement boundary conditions

(DBCs). It is subjected to a distributed force q(x) and moment ML at the left-end.
 Solution to Bended Beams 311

This time we plot all the solutions in graphs.

1 dic1 = {E:1,I:1,l:1,q:1,ML:0} # set values for variables

2 vx=[v[i].subs(dic).subs(dic1) for i in range(len(v))]
3 vx

 
x4 x3 x x3 3x2 1 x2 3x 3
− + , − + , − , − x, 1
24 16 48 6 16 48 2 8 8

1 # Plot all these curves for v, , M, V:

2 for i in range(len(title)):
3 gr.plot2curveS(vx[2*i:2*(i+1)], 0., 1., title=title[i])
4
5 [Link]()

Maximum Deflection value=5.416e-03, at x=0.42

Maximum Rotation value=2.083e-02, at x=0.0
Extreme Deflection values=[-3.469e-18 5.416e-03],
at x=[ 1.000e+00 4.200e-01]
Critical Rotation values=[-1.041e-17 1.063e-04],
at x=[ 1.000e+00 4.200e-01]
Deflection values at boundary =(0.0, -3.469446951953614e-18)
Rotation values at boundary =(0.020833333333333332, -1.0408340855860843e-17)

Maximum Moment value=1.250e-01, at x=1.0

Maximum Shear force value=3.750e-01, at x=0.0
Extreme Moment values=[-7.030e-02],
at x=[ 3.700e-01]
Critical Shear force values=[ 5.000e-03],
at x=[ 3.700e-01]
Moment values at boundary =(0.0, 0.125)
Shear force values at boundary =(0.375, -0.625)
312 Mechanics of Materials: Formulations and Solutions with Python

1 dicML = {v0:0, 0:0, vl:0, l:0, q:0, M0:-ML}

2 printM(v[0].subs(dicML),
3 'Deflection function by moment ML, when all DBCs=0:',n_dgt=4)

ML lx ML x2 ML x3
Deflection function by moment ML, when all DBCs=0: − +
4EI 2EI 4EIl

1 printM(v[1].subs(dicML),
2 'Rotation distribution by moment ML, when all DBCs=0:',n_dgt=4)

Rotation distribution by moment ML, when all DBCs=0:

ML l ML x 3ML x2
− +
4EI EI 4EIl

1 printM((v[2].subs(dicML)).expand(),
2 'Moment distribution by moment ML, when all DBCs=0:')

3ML x
Moment distribution by moment ML, when all DBCs=0: −ML +
2l

1 printM((v[3].subs(dicML)).expand(),
2 'Shear force distribution by moment ML, when all DBCs=0:')

3ML
Shear force distribution by moment ML, when all DBCs=0: −
2l

The shear force is constant along the beam because the body force is set
to zero.

8.10 Example: Free-clamp supported beams

Consider beams with cross-section having a second moment of area I. It is

made of material with Young’s modulus E. It is free-clamp supported (f-c),
as shown in Fig. 8.11.
A full set of solutions is found using solver1D4(). First, let us find the
general solution.
 Solution to Bended Beams 313

Figure 8.11. A cantilever beam with free-clamp (s-c) displacement boundary condi-
tions (DBCs). It is subjected to a distributed force q(x), a concentrated force P and a
concentrated moment M at the left-end.

1 q, M0, v0, vl, 0, l = symbols('q, M0, v0, vl, 0, l')

2 # Consider the distributed force is a constant q
3 by = q
4 v = solver1D4(E, I, by, l, v0, 0, vl, l, V0, M0, Vl, Ml, key='f-c')
5 print(f'Is solution correct? {by/E/I==v[4]}; The solution u(x) is:')
6 v[0]

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? True; The solution u(x) is

M0 l2 M0 lx M0 x2 V0 l3 V0 l2 x V0 x3
−lθl + vl + xθl + − + − + −
2EI EI 2EI 3EI 2EI 6EI
l q
4 l qx
3 qx 4
+ − +
8EI 6EI 24EI

Let us consider q = 0, M = 0, and only P is on.

1 # Cantilever beam under tip load P (positive upward) at the left-tip:

2 P, M = symbols('P, M')
3 dicP = {q:0, vl:0, l:0, M0:0, V0:-P}
4 # at negative boundary ⇑, sign reversal
5 vx= [v[i].subs(dicP) for i in range(len(v))]
6 vx # solution in [v, , M, V, qy]
 
P l3 P l 2 x P x3 P l2 P x2
− + , − + , P x, −P, 0
3EI 2EI 6EI 2EI 2EI

It is seen that the shear force in the beam is a constant of −P along

the beam. The moment in the beam varies linearly, P x, along the beam.
314 Mechanics of Materials: Formulations and Solutions with Python

The deflection distribution is third-order polynomial, and at x = 0 it can be

found as follows:

1 vx[0].subs(x,0)

P l3
3EI
This is the well-known formula for computing the largest deflection in a
cantilever beam subject to a point force at its free end, which is registered
at the free-end.

1 # Cantilever beam under external M (positive counterclockwise) at x=0:

2 ML = symbols('M_L')
3 vx= [v[i].subs({q:0, vl:0, l:0, M0:-ML, V0:0}) for i in range(len(v))]
4 vx # at negative boundary ⇑, sign reversal
 
ML l2 ML lx ML x2 ML l ML x
− + − , − , −ML , 0, 0
2EI EI 2EI EI EI

It is seen in this case that the moment in the beam is a constant of −ML
along the beam. The rotation at x = 0 is found as follows:

1 vx[1].subs(x,0)

ML l
EI
We have the widely used rotation-moment relation for cantilever beam.
We write it down for later reference.

l EI
θL = ML or ML = θL (8.35)
EI l

where EIl is called bending stiffness for beams with length l. It is equivalent
to the spring constant k.
Let us plot the solutions in graphs for q = 0 and ML = 1.

1 vx1 = [vx[i].subs({E:1,I:1,l:1,q:1,ML:1}) for i in range(len(v))]

2 vx1
 
x2 1
− + x − , 1 − x, −1, 0, 0
2 2
 Solution to Bended Beams 315

1 title = [["Deflection", "Rotation"], ["Moment", "Shear force"]]

2 for i in range(len(title)):
3 gr.plot2curveS(vx1[2*i:2*(i+1)], 0., 1., title=title[i])
4
5 [Link]()

Maximum Deflection value=0.000e+00, at x=1.0

Maximum Rotation value=1.000e+00, at x=0.0
Extreme Deflection values=[ 0.000e+00],
at x=[ 1.000e+00]
Critical Rotation values=[ 0.000e+00],
at x=[ 1.000e+00]
Deflection values at boundary =(-0.5, 0.0)
Rotation values at boundary =(1.0, 0.0)

Maximum Moment value=-1.000e+00, at x=0.0

Maximum Shear force value=0.000e+00, at x=0.0
Extreme Moment values=[-1.000e+00],
at x=[ 0.000e+00]
Critical Shear force values=[ 0.000e+00],
at x=[ 0.000e+00]
Moment values at boundary =(-1.0, -1.0)
Shear force values at boundary =(0.0, 0.0)

It is seen that the moment solution is constant (−1 because the positive
external moment is on the negative boundary of the beam), and shear force
316 Mechanics of Materials: Formulations and Solutions with Python

is zero. This means that the bean is under pure bending without any shear
force been generated, as expected for an f-c beam loaded with a moment at
the free-end.
Next, let us plot the case of q = 0, M = 0, and only P is on the f-c beam.

1 # Cantilever beam under external P (positive upwards) at x=0:

2 P = symbols('P')
3 vx= [v[i].subs({q:0, vl:0, l:0, M0:0, V0:-P}) for i in range(len(v))]
4 vx # at negative boundary , sign reversal

 
P l3 P l 2 x P x3 P l2 P x2
− + , − + , P x, −P, 0
3EI 2EI 6EI 2EI 2EI

1 vx1 = [vx[i].subs({E:1,I:1,l:1,q:1,P:1}) for i in range(len(v))]

2 vx1

 
x3 x 1 x2 1
− + , − , x, −1, 0
6 2 3 2 2

1 title = [["Deflection", "Rotation"], ["Moment", "Shear force"]]

2 for i in range(len(title)):
3 gr.plot2curveS(vx1[2*i:2*(i+1)], 0., 1., title=title[i])
4
5 [Link]()

Maximum Deflection value=3.333e-01, at x=0.0

Maximum Rotation value=0.000e+00, at x=1.0
Extreme Deflection values=[-5.551e-17],
at x=[ 1.000e+00]
Critical Rotation values=[ 0.000e+00],
at x=[ 1.000e+00]
Deflection values at boundary =(0.3333333333333333, -5.551115123125783e-17)
Rotation values at boundary =(-0.5, 0.0)

Maximum Moment value=1.000e+00, at x=1.0

Maximum Shear force value=-1.000e+00, at x=0.0
Extreme Moment values=[ 0.000e+00],
at x=[ 0.000e+00]
Critical Shear force values=[-1.000e+00],
at x=[ 0.000e+00]
Moment values at boundary =(0.0, 1.0)
Shear force values at boundary =(-1.0, -1.0)
 Solution to Bended Beams 317

This time, the shear force solution is constant (−1 because the posi-
tive external force is on the negative boundary of the beam), and moment
increases from linearly from 0. This is the case we studied by-hand in Section
8.3. The difference is the orientation of the beam is reversed.

8.11 Example: Clamped-clamped beam subjected to a

point force at the middle span
Consider a beam with length l with c-c BC, as shown in Fig. 8.12(a). The
beam has the same Young’s modulus E, and its cross-section has a second
moment of area I. It is subjected to a point force at the middle span A (x =
l/2), with x-axis starting at the left-end of the beam and being leftwards.
Derive formulas for solutions, including deflection, rotation, moment, and
shear force in the beam.

(a)

(b)

Figure 8.12. (a) A beam with clamped-clamped (c-c) displacement boundary conditions
(DBCs). It is subjected to a concentrated force P at the middle span. (b) Consider only
half of the beam by making use of the symmetry.
318 Mechanics of Materials: Formulations and Solutions with Python

Solution: This problem looks simple, but it needs some considerations. First
because force P is a point force, the solution may not be continuous and
needs to be given piecewise. Since the problem is symmetric with respect to
x = l/2, the rotation at A must be zero. Thus, we can consider only a half of
the beam, say the left half, with also the force being halved. Now, we have a
beam with length of l/2 and clamp-roller (c-r) BCs, as shown in Fig. 8.12b.
The general solution for the c-r beam can be found using our beam solver,
solver1D4(), with c-c BC and then set the rotation at A to zero. The process
and codes are as follows:

1 title = [["Deflection", "Rotation"], ["Moment", "Shear force"]]

2
3 by = 0 # Not distributed force
4 # the length is halved
5 v = solver1D4(E, I, by, l/2, v0, 0, vl, l, V0, M0, Vl, Ml,key='c-c')
6 print(f'Is solution correct? {by/E/I==v[4]}; The solution u(x) is:')
7 v[0] # solution in a general displacement function

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? True; The solution u(x) is

4x2 θ0 2x2 θl 12v0 x2 12vlx2 4x3 θ0 4x3 θl 16v0 x3 16vlx3

v0 + xθ0 − − − + + 2 + 2 + −
l l l2 l2 l l l3 l3

1 # Set DBCs at the left-end, and rotation at A to zero:

2
3 vx=[v[i].subs({v0:0, 0:0, l:0}) for i in range(len(v))]
4 vx

 
12vlx2 16vlx3 24vlx 48vlx2 24EIvl 96EIvlx 96EIvl
− , − , − , , 0
l2 l3 l2 l3 l2 l3 l3

This set of solution has an unknown vl . At point A, we can find the shear
force using the solution obtained:

1 fy2=vx[3].subs(x,l/2).expand() # at point A
2 [Link](fy2, 'Shear force at point A, fy2:',n_dgt=4)

Shear force at point A, fy2:

96EIvl
l3
 Solution to Bended Beams 319

The shear force at point A must be P/2, we then have

1 P = symbols("P")
2 sln = [Link](fy2 - P/2, vl)
3 [Link](sln[0], 'deflection at A, v_l = ',n_dgt=4)

deflection at A, v_l =
P l3
192EI

This is the well-known formula for computing the largest deflection in a

c-c beam subject to a point force at the middle span.
Finally, we put vl back to the solution set:

1 # substitute v_l to the solution obtained earlier:

2
3 sln_final =[vx[i].subs({vl:sln[0]}) for i in range(len(v))]
4 sln_final # solution set in [v, , M, V, qy]

 
P lx2 P x3 P lx P x2 P l P x P
− , − , − , , 0
16EI 12EI 8EI 4EI 8 2 2

This is the final full solution set for a c-c beam subject to a point
force at the middle span, which can be found in textbooks and the open
literature.

8.12 Plot piecewise solutions

For the foregoing problem, the solution has two pieces: one for left-span and
one for the right-span of the beam. We write the following code to plot the
solutions.

1 # Set the parameters to unit values:

2
3 sln_x =[sln_final[i].subs({P:1,l:1,E:1,I:1}) for i in range(len(v))]
4 sln_x
320 Mechanics of Materials: Formulations and Solutions with Python

 
x3 x2 x2 x 1 x 1
− + , − + , − , , 0
12 16 4 8 8 2 2

1 # set parameters: #P = 1.0; E = 1.0; I = 1.0

2 l1 = 0.5; l2 = 0.5
3
4 dx = 0.02
5 x1 = [Link](0.0, l1, dx) # coordinates in l1, left span
6 x2 = [Link](0.0, l2, dx) # coordinates in l2, right span
7 X2 = -x2 + l1 # coord. transfer, right-span is opposite to left-span
8 signs=[1, -1, 1, -1] # symmetry of solutions: v is sym, is anti-sym
9
10 # convert Sympy solution func. to numpy arrays.
11 n = len(sln_x)-1
12 x = symbols("x")
13 sln = [[Link](x, sln_x[i], 'numpy') for i in range(n)]
14 sx1 = [ sln[i](x1) for i in range(n)]
15 sx2 = [signs[i]*sln[i](X2) for i in range(n)]
16
17 # The shear force function could be a constant. Need to generate array
18 if type(sx1[-1]) != [Link]: # in case, func. is a constant
19 sx1[-1] = float(sx1[-1])*np.ones_like(x1)
20 sx2[-1] = float(sx2[-1])*np.ones_like(X2)
21
22 X = [Link]((x1, l1+x2)) # put the results in a single array
23 sx = [[Link]((sx1[i], sx2[i])) for i in range(n)]
24
25 sx_max = [[Link](abs(sx[i])) for i in range(n)] # absolute max values
26 print(f"The absolute maximum solutions: {sx_max}")

The absolute maximum solutions: [0.005208333333333334, 0.0156, 0.125, 0.5]

1 # Plot the distribution curves:

2 x_data = [X for i in range(n)] # put x values to a list
3 y_data = [sx[i]/sx_max[i] for i in range(n)] # func. values to a list
4
5 labels=["Deflection", "Rotation", "Moment", "Shear force"]
6 gr.plot_fs(x_data, y_data, labels, 'x', 'Solutions', 'temp')
7 # x-label, y-label, file-name
 Solution to Bended Beams 321

The discontinuity of the solutions especially for the moment and shear
force is clearly shown.
Readers may try to find the solution to the same problem, but replacing
P with M at the same location.

8.13 Example: Simply supported beam with two spans

Consider a beam with two spans. Span-1 has a length l1 , and span-2 has a
length l2 . All the members has the same Young’s modulus E, and its cross-
section has a second moment of area I. The beam is simple-simple-clamp
supported (s-s-c), as shown in Fig. 8.13. It is subjected to a distributed force
q over span-2.

1. Derive formulas for solutions, including deflection, rotation, moment, and

shear force in the beam.
2. Plot the distribution of all these solutions.

Figure 8.13. A beam with two spans. It is simple-simple-clamp supported.

322 Mechanics of Materials: Formulations and Solutions with Python

Solution: Since the beam has two spans, the solution will be piecewise. Let
the x-axis start at point A for span-1 and at point B for span-2. Both are
positive rightwards.
For each of the spans, we can find the general solution using our beam
solver solver1D4(). The process and codes are as follows:

1 # Define variables:
2 title = [["Deflection", "Rotation"], ["Moment", "Shear force"]]
3 E, I, l1, l2, q = symbols("E, I, l1, l2, q")
4 x = symbols("x")
5
6 # for displacement boundary conditions (DBCs):
7 v10, 10, v1l, 1l = symbols('v10, 10, v_1l, _1l') # DBCs for span-1
8 v20, 20, v2l, 2l = symbols('v20, 20, v_2l, _2l') # DBCs for span-2
9
10 # for force boundary conditions (FBCs):
11 V10, M10, V1l, M1l = symbols('V10, M10, V_1l, M_1l') # FBCs for span-1
12 V20, M20, V2l, M2l = symbols('V20, M20, V_2l, M_2l') # FBCs for span-2

8.13.1 For span-1

Let us seek the general solution for span-1.

1 # Distributed force is zero for span-1

2 by = 0 # BC is s-s
3 v1 = solver1D4(E,I,by,l1, v10, 10,v1l, 1l, V10, M10, V1l, M1l,key='s-s')
4 print(f'Is solution correct? {by/E/I==v1[4]}; The solution v(x) is:')
5 v1[0] # solution of general displacement function for span-1

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? True; The solution v(x) is

v10 x v1l x M10 l1 x M10 x2 M10 x3 M1l l1 x M1l x3

v10 − + − + − − +
l1 l1 3EI 2EI 6EIl1 6EI 6EIl1

1 # Set known BCs for span-1:

2
3 v1x=[v1[i].subs({v10:0,M10:0,v1l:0}) for i in range(len(v1))]
4 v1x # solution of v, , Mx, Vx, qy
 Solution to Bended Beams 323

 
M1l l1 x M1l x3 M1l l1 M1l x2 M1l x M1l
− + , − + , , − , 0
6EI 6EIl1 6EI 2EIl1 l1 l1

This set of solutions has an unknown M1l . At point B, we can find the
rotation using the general solution:

1 1B = v1x[1].subs(x,l1).expand() # at point B of span-1

2 [Link]( 1B, 'rotation at point B of span-1, 1B:',n_dgt=4)

rotation at point B of span-1, 1B:

M1l l1
3EI

8.13.2 For span-2

Next, let us seek the general solution for span-2.

1 # General solution for span-2: (constant q)

2 # BC is simple-clamp
3 v2 = solver1D4(E,I,q,l2, v20, 20,v2l, 2l, V20,M20, V2l,M2l,key='s-c')
4 print(f'Is solution correct? {by/E/I==v2[4]}; The solution u(x) is: ')
5 v2[0] # solution of general displacement function for span-2

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? False; The solution u(x) is

xθ2l 3v20 x 3v2l x x3 θ2l v20 x3 v2l x3 M20 l2 x M20 x2

v20 − − + + + − − +
2 2l2 2l2 2l22 2l23 2l23 4EI 2EI
M20 x 3 l qx
3 l2 qx 3 qx 4
− + 2 − +
4EIl2 48EI 16EI 24EI

1 # Set known BCs for span-2:

2
3 v2x=[v2[i].subs({v20:0,v2l:0, 2l:0}) for i in range(len(v2))]
4 v2x[1] # solution of

M20 l2 M20 x 3M20 x2 l3 q 3l2 qx2 qx3

− + − + 2 − +
4EI EI 4EIl2 48EI 16EI 6EI
324 Mechanics of Materials: Formulations and Solutions with Python

This set of solutions has an unknown M20 . At point B of span-2, we can

find the rotation using the general solution:

1 2B = v2x[1].subs(x,0).expand() # at point B of span-2

2 [Link]( 2B, 'rotation at point B of span-2, 2B:',n_dgt=4)

M20 l2 l3 q
rotation at point B of span-2, 2B: − + 2
4EI 48EI

8.13.3 Continuity and equilibrium conditions

At point B of the beam, we must have the following conditions:

θ1B = θ2B continuity in rotation

(8.36)
−M1B + M2B = 0 equilibrium in moment

The free-body diagram is shown in Fig. 8.14. We now use these conditions
to solve for unknowns M1B and M2B , using the following code.

1 sln = [Link]([ 1B - 2B, -M1l + M20], [M1l, M20])

2 sln # solution for these unknowns

 
l23 q l23 q
M20 : , M1l :
16l1 + 12l2 16l1 + 12l2

Figure 8.14. Free-body diagram of an infinitely small beam segment on support B with
length dx.
 Solution to Bended Beams 325

8.13.4 Final solutions

Substituting these unknown moments back to the general solutions, gives

the final solutions.

1 # Final solution in span-1:

2 vs1x = [v1x[i].subs({M1l:sln[M1l]}) for i in range(len(v1x))]
3 vs1x[0] # solution of deflection v as function of x for span-1.

l1 l23 qx l23 qx3

− +
6EI (16l1 + 12l2 ) 6EIl1 · (16l1 + 12l2 )

1 vs1x[1] # solution of rotation

l1 l23 q l23 qx2

− +
6EI (16l1 + 12l2 ) 2EIl1 · (16l1 + 12l2 )

1 vs1x[2:] # solution of moment M and shear force V

 
l23 qx l23 q
, − , 0
l1 · (16l1 + 12l2 ) l1 · (16l1 + 12l2 )

1 # Final solution in span-2:

2 vs2x=[v2x[i].subs({M20:sln[M20]}) for i in range(len(v2x))]
3 vs2x[0] # solution v(x)

l24 qx l23 qx2 l3 qx l22 qx3

− + + 2 −
4EI (16l1 + 12l2 ) 2EI (16l1 + 12l2 ) 48EI 4EI (16l1 + 12l2 )
l2 qx 3 qx 4
− +
16EI 24EI
326 Mechanics of Materials: Formulations and Solutions with Python

1 vs2x[1] # solution of (x)

l24 q l23 qx l3 q 3l22 qx2

− + + 2 −
4EI (16l1 + 12l2 ) EI (16l1 + 12l2 ) 48EI 4EI (16l1 + 12l2 )
3l2 qx 2 qx 3
− +
16EI 6EI

1 vs2x[2:] # solution of moment M(x) and shear force V(x)


l23 q 3l22 qx 3l2 qx qx2
− − + ,
16l1 + 12l2 2 · (16l1 + 12l2 ) 8 2

3l22 q 3l2 q q
+ − qx,
2 · (16l1 + 12l2 ) 8 EI

This is the final solution to our problem, which may not be found in
textbooks and the open literature. When setting l1 = l2 , one can obtain
simplified solution that can be found in textbooks and the open literature.
Readers may do so to check the solutions obtained.

8.14 Plot piecewise solutions

For the foregoing problem, the solution has two pieces: one for left-span and
one for the right-span of the beam. We write the following code to plot the
solutions together.

1 # Set the parameters to unit values:

2
3 sln1x=[vs1x[i].subs({q:1,l1:1,l2:1,E:1,I:1}) for i in range(len(vs1x))]
4 sln1x

 
x3 x x2 1 x 1
− , − , , − , 0
168 168 56 168 28 28
 Solution to Bended Beams 327

1 # Set the parameters to unit values:

2
3 sln2x=[vs2x[i].subs({q:1,l1:1,l2:1,E:1,I:1}) for i in range(len(vs2x))]
4 sln2x

 
x4 x3 x2 x x3 3x2 x 1 x2 3x 1 3
− + + , − + + , − + , − x, 1
24 14 56 84 6 14 28 84 2 7 28 7

1 np.set_printoptions(formatter={'float': '{: 0.4e}'.format})

2 # set parameters: #q = 1.0; E = 1.0; I = 1.0
3 l1 = 1.0; l2 = 1.0
4
5 dx = 0.002
6 x1 = [Link](0.0, l1, dx) # coordinates in l1, left span
7 x2 = [Link](0.0, l2, dx) # coordinates in l2, right span
8
9 # convert Sympy solution func. to numpy arrays.
10 n1 = len(sln1x)-1
11 n2 = len(sln1x)-1
12 x = symbols("x")
13 sln1 = [[Link](x, sln1x[i], 'numpy') for i in range(n1)]
14 sln2 = [[Link](x, sln2x[i], 'numpy') for i in range(n2)]
15 sx1 = [sln1[i](x1) for i in range(n1)]
16 sx2 = [sln2[i](x2) for i in range(n2)]
17
18 # The shear force function could be a constant. Need to generate array
19 if type(sx1[-1]) != [Link]: # in case, func. is a constant
20 sx1[-1] = float(sx1[-1])*np.ones_like(x1)
21
22 if type(sx2[-1]) != [Link]: # in case, func. is a constant
23 sx2[-1] = float(sx2[-1])*np.ones_like(x2)
24
25 X = [Link]((x1, l1+x2)) # put the results in a single array
26 sx = [[Link]((sx1[i], sx2[i])) for i in range(n1)]
27
28 sx_max = [[Link](abs(sx[i])) for i in range(n1)] # absolute max values
29 print(f"Absolute maximum values: {[Link](sx_max)}")

Absolute maximum values: [ 4.1733e-03 1.3507e-02 1.0600e-01 5.6943e-01]

1 # Plot the distribution curves:

2 x_data = [X for i in range(n1)] # put x coordinates to a list
3 y_data = [sx[i]/sx_max[i] for i in range(n1)] # func. values to a list
4
5 labels=["Deflection", "Rotation", "Moment", "Shear force"]
6 gr.plot_fs(x_data, y_data, labels, 'x', 'Solutions', 'temp')
7 # x-label, y-label, file-name
328 Mechanics of Materials: Formulations and Solutions with Python

The discontinuity of the solutions especially for the moment and shear
force is clearly shown.

8.15 Example: A symmetric frame subjected to

a uniformly distributed load*

Consider a simple frame consisting of three members with uniform cross-

section. It is symmetric, as shown in Fig. 8.15(a). Its pillar has a length
l1 , and top-beam has a length 2l2 . All the members have the same Young’s
modulus E, and its cross-section area is A and second moment of area I. The
frame is clamped on the ground. It is subjected to a uniformly distributed
force q on the top-beam of the frame.

1. Derive formulas for solutions, including deflection, rotation, moment, and

shear force in the frame.
2. Plot the distribution of all these solutions, using unit for all its
parameters.

Solution: Since the frame is symmetric, we can consider only half of it, as
shown in in Fig. 8.15(b). The support for point C becomes a roller, implying
that it constrains the horizontal movement, but allows it to move vertically
at that point. The frame now has two members: member-1 is the pillar,
and member-2 is the (halved) top-beam. We can now use similar strategy
 Solution to Bended Beams 329

(a) (b)

Figure 8.15. (a) A symmetric frame fixed on the grounds. It is subjected to a uniformly
distributed force on the top. (b) Only half a frame needs to be analyzed, making use of
the symmetry.

as for the previous example, by obtaining the general solutions for these
two members separately, and then use the continuity conditions at point B
and equilibrium conditions to determine these unknowns.
Let the x-axis start at point A for member-1 (mem-1) with upwards being
positive, and x-axis starts at point B for member-2 (mem-2) with rightwards
positive.
For each of the members, we can find the general solution using our beam
solver solver1D4(). The process and codes are as follows:

1 # Define variables:
2 title = [["Deflection", "Rotation"], ["Moment", "Shear force"]]
3 E, I, A, l1, l2, q = symbols("E, I, A, l1, l2, q")
4 x = symbols("x")
5
6 # for displacement boundary conditions (DBCs):
7 v10, 10, v1l, 1l = symbols('v10, 10, v_1l, _1l') # DBCs for mem-1
8 v20, 20, v2l, 2l = symbols('v20, 20, v_2l, _2l') # DBCs for mem-2
9
10 # for force boundary conditions (FBCs):
11 V10, M10, V1l, M1l = symbols('V10, M10, V_1l, M_1l') # FBCs for mem-1
12 V20, M20, V2l, M2l = symbols('V20, M20, V_2l, M_2l') # FBCs for mem-2
330 Mechanics of Materials: Formulations and Solutions with Python

8.15.1 For member-1

Let us seek the general solution for member-1.

1 # Distributed force is zero for member-1

2 by = 0 # BC is clamp-clamp
3 v1 = solver1D4(E,I,by,l1, v10, 10,v1l, 1l, V10,M10, V1l,M1l,key='c-c')
4 print(f'Is solution correct? {by/E/I==v1[4]}; The solution v(x) is: ')
5 v1[0] # solution of general displacement function

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? True; The solution v(x) is

2x2 θ10 x2 θ1l 3v10 x2 3v1l x2 x3 θ10 x3 θ1l 2v10 x3 2v1l x3

v10 +xθ10 − − − 2 + 2 + 2 + 2 + 3 − 3
l1 l1 l1 l1 l1 l1 l1 l1

1 # Set known BCs for member-1:

2
3 v1x=[v1[i].subs({v10:0, 10:0}) for i in range(len(v1))]
4 v1x[:2] # solution of v,

 
x2 θ1l 3v1l x2 x3 θ1l 2v1l x3 2xθ1l 6v1l x 3x2 θ1l 6v1l x2
− + + − , − + 2 + −
l1 l12 l12 l13 l1 l1 l12 l13

1 v1x[2:] # solution of Mx, Vx, qy

 
2EIθ1l 6EIv1l 6EIxθ1l 12EIv1l x 6EIθ1l 12EIv1l
− + + − , − + , 0
l1 l12 l12 l13 l12 l13

This set of solutions has 2 unknowns θ1l and v1l . At point B, we can find
the rotation using the general solution:

1 M1B = v1x[2].subs(x,l1).expand() # at point B of mem-1

2 [Link](M1B, 'Moment at point B of member-1, M1B:',n_dgt=4)

4EIθ1l 6EIv1l
Moment at point B of member-1, M1B: −
l1 l12
 Solution to Bended Beams 331

We can also find the shear force at point B of member-1.

1 V1B = v1x[3].subs(x,l1).expand() # at point B of mem-1
2 [Link](V1B, 'Shear force at point B of member-1, V1B:',n_dgt=4)

6EIθ1l 12EIv1l
Shear force at point B of member-1, V1B: − +
l1
2 l13

8.15.2 For member-2

Next, let us seek the general solution for member-2.
1 # General solution for member-2: (constant q)
2 # BC is clamp-clamp
3 v2 = solver1D4(E,I,q,l2, v20, 20,v2l, 2l, V20, M20, V2l, M2l,key='c-c')
4 print(f'Is solution correct? {by/E/I==v2[4]}; The solution u(x) is:')
5 v2[0] # solution of general displacement function v(x)

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? False; The solution u(x) is

2x2 θ20 x2 θ2l 3v20 x2 3v2l x2 x3 θ20 x3 θ2l 2v20 x3

v20 + xθ20 − − − + + 2 + 2 +
l2 l2 l22 l22 l2 l2 l23
2v2l x3 l22 qx2 l2 qx3 qx4
− + − +
l23 24EI 12EI 24EI

1 # Set known BCs for member-2:

2
3 v2x=[v2[i].subs({ 2l:0}) for i in range(len(v2))]
4 v2x[1] # solution of (x)

4xθ20 6v20 x 6v2l x 3x2 θ20 6v20 x2 6v2l x2

θ20 − − 2 + 2 + + −
l2 l2 l2 l22 l23 l23
l2 qx l2 qx2 qx3
+ 2 − +
12EI 4EI 6EI
This set of solutions has 3 unknowns θ20 , v20 , v2l . At point B of member-2,
we can find the moment using the general solution:
332 Mechanics of Materials: Formulations and Solutions with Python

1 M2B = v2x[2].subs(x,0).expand() # at point B of mem-2

2 [Link](M2B, 'Moment at point B of top-beam, M2B:',n_dgt=4)

Moment at point B of top-beam, M2B

4EIθ20 6EIv20 6EIv2l l22 q

− − + +
l2 l22 l22 12

The shear force at point B of member-2 is found as follows.

1 V2B = v2x[3].subs(x,0).expand() # at point B of mem-2

2 [Link](V2B, 'Shear force at point B of member-2, V2B:',n_dgt=4)

Shear force at point B of member-2, V2B

6EIθ20 12EIv20 12EIv2l l2 q

− − + +
l2
2 l2
3 l2
3 2

Also, the shear force at point C of member-2 is found as follows.

1 V2l = v2x[3].subs(x,l2).expand() # at point B of mem-2

2 [Link](V2l, 'Shear force at point C of member-2, V2l:',n_dgt=4)

Shear force at point C of member-2, V2l

6EIθ20 12EIv20 12EIv2l l2 q

− − + −
l2
2 l2
3 l2
3 2

8.15.3 Continuity and equilibrium conditions at point B

In the solutions for the frame, we have a total of 5 unknowns: θ1l and v1l ,
θ20 , v20 , and v2l . We need 5 conditions to determine those.
At point B of the frame, we have the following conditions:

θ1B = θ2B continuity

−M1B + M2B = 0 equilibrium in moment
−V1B l2
= v1l elongation of member-2 = deflection of member-1
EA
−V2B l1
= −v20 elongation of member-1 = deflection of member-2
EA
V2B = ql2 shear force = sum of load on member-2 (8.37)
 Solution to Bended Beams 333

These are the 5 conditions we need to solve for these 5 unknowns

θ1l , v1l , θ20 , v20 and v2l . The following code sets these 5 equations and then
finds these unknowns.

1 eq1 = 1l- 20
2 eq2 = -M1B+M2B
3 eq3 = -V1B*l2/(E*A)-v1l
4 eq4 = -V2B*l1/(E*A)+v20
5 eq5 = V2B-q*l2
6 sln = [Link]([eq1, eq2, eq3, eq4, eq5], [ 1l, v1l, 20, v20, v2l])
7 sln[ 1l] # solution for unknown 1l

Al14 l23 q + 12Il1 l24 q

3AEIl14 + 12AEIl13 l2 + 36EI 2 l1 l2 + 36EI 2 l22

1 sln[v1l] # solution for unknown v1l

2l12 l24 q
AEl14 + 4AEl13 l2+ 12EIl1 l2 + 12EIl22

1 sln[ 20] # solution for unknown 20

Al14 l23 q + 12Il1 l24 q

3AEIl14 + 12AEIl13 l2 + 36EI 2 l1 l2 + 36EI 2 l22

1 sln[v20] # solution for unknown v20

l1 l2 q
AE

1 sln[v2l] # solution for unknown v21

5A2 l14 l24 q + 4A2 l13 l25 q + 24AIl15 l2 q + 96AIl14 l22 q + 60AIl1 l25 q
+ 12AIl26 q + 288I 2 l12 l22 q + 288I 2 l1 l23 q
24A2 EIl14 + 96A2 EIl13 l2 + 288AEI 2 l1 l2 + 288AEI 2 l22

8.15.4 Final solutions

Substituting these unknowns found back to the general solution for the frame
members, gives the final solutions.
334 Mechanics of Materials: Formulations and Solutions with Python

1 # Final solution in mem-1:

2 vs1x=[v1x[i].subs({ 1l:sln[ 1l],v1l:sln[v1l]}) for i in range(len(v1x))
3 vs1x[0].simplify() # solution of deflection v(x)

l23 qx2 −Al13 + Al12 x + 6Il2

3EI Al14 + 4Al13 l2 + 12Il1 l2 + 12Il22

1 vs1x[1].simplify() # solution of rotation (x)

l23 qx −2Al13 + 3Al12 x + 12Il2

3EI Al14 + 4Al13 l2 + 12Il1 l2 + 12Il22

1 vs1x[2].simplify() # solution of moment M(x)

2l23 q −Al13 + 3Al12 x + 6Il2

3 Al14 + 4Al13 l2 + 12Il1 l2 + 12Il22

1 vs1x[3].simplify() # solution of shear force V(x)

2Al12 l23 q
−
Al14 + 4Al13 l2 + 12Il1 l2 + 12Il22

1 vs1x[4] # solution of transverse load by

This is zero as expected. Let us find the solutions for member-2. Since
the formulas are very long. We set l2 = l1 , so that these can be shortened
for displaying.

1 # Final solution in mem-2:

2 dic = { 20:sln[ 20],v20:sln[v20],v2l:sln[v2l]}
3 vs2x=[v2x[i].subs(dic) for i in range(len(v2x))]
4 vs2x[0].subs({l2:l1}).simplify().collect(x**4).collect(x**3)\
5 .collect(x**2).collect(x) # solution of v(x)
 Solution to Bended Beams 335

q 120AIl14 + 576I 2 l12 + x4 · 5A2 l12 + 24AI + x3 −20A2 l13 − 96AIl1

+ x2 · 16A2 l14 + 48AIl12 + x 8A2 l15 + 96AIl13
24AEI 5Al12 + 24I

1 vs2x[1].subs({l2:l1}).simplify().collect(x**3)\
2 .collect(x**2).collect(x) # solution of (x)

q 2Al15 + 24Il13 + x3 · 5Al12 + 24I + x2 −15Al13 − 72Il1 + x 8Al14 + 24Il12

6EI 5Al12 + 24I

1 vs2x[2].subs({l2:l1}).simplify().collect(x**2).collect(x) # moment M(x)

q 8Al14 + 24Il12 + x2 · 15Al12 + 72I + x −30Al13 − 144Il1

6 · 5Al12 + 24I

1 vs2x[3].subs({l2:l1}).simplify() # shear force V(x)

q (l1 − x)

The normal forces in these members are found as

1 N_mem1 = vs2x[3].subs({x:0}) # normal forces in member-1

2 N_mem1.simplify()

l2 q

1 N_mem2 = vs1x[3].subs({x:l1}) # normal forces in member-2

2 N_mem2.simplify()

2Al12 l23 q
−
Al14 + 4Al13 l2 + 12Il1 l2 + 12Il22

This is the final solution to our frame structure problem, which may not
be found in textbooks and the open literature.

8.16 Plot piecewise solutions

For the foregoing problem, the solution has two pieces: one for pillar and one
for the top-beam. We write the following code to plot the solutions.
336 Mechanics of Materials: Formulations and Solutions with Python

1 # Set the parameters to unit values:

2 dic = {q:1,l1:1,l2:1,E:1,I:1,A:1} # set all parameters to unit
3 sln1x=[vs1x[i].subs(dic) for i in range(len(vs1x))]
4 sln1x

 
x3 5x2 x2 10x 2x 10 2
+ , + , + , − , 0
87 87 29 87 29 87 29

1 # Set the parameters to unit values:

2
3 sln2x=[vs2x[i].subs(dic) for i in range(len(vs2x))]
4 sln2x

 
x4 x3 8x2 13x x3 x2 16x 13 x2 16
− + + + 1, − + + , − x + , 1 − x, 1
24 6 87 87 6 2 87 87 2 87

1 np.set_printoptions(formatter={'float': '{: 0.4e}'.format})

2 # set parameters: #q = 1.0; E = 1.0; I = 1.0
3 l1 = 1.0; l2 = 1.0
4
5 dx = 0.002
6 x1 = [Link](0.0, l1, dx) # coordinates in l1, pillar
7 x2 = [Link](0.0, l2, dx) # coordinates in l2, top-beam
8
9 # convert Sympy solution func. to numpy arrays.
10 n1 = len(sln1x)-1
11 n2 = len(sln1x)-1
12 x = symbols("x")
13 sln1 = [[Link](x, sln1x[i], 'numpy') for i in range(n1)]
14 sln2 = [[Link](x, sln2x[i], 'numpy') for i in range(n2)]
15 sx1 = [sln1[i](x1) for i in range(n1)]
16 sx2 = [sln2[i](x2) for i in range(n2)]
17
18 # The shear force function could be a constant. Need to generate array
19 if type(sx1[-1]) != [Link]: # in case, func. is a constant
20 sx1[-1] = float(sx1[-1])*np.ones_like(x1)
21
22 if type(sx2[-1]) != [Link]: # in case, func. is a constant
23 sx2[-1] = float(sx2[-1])*np.ones_like(x2)
24
25 X = [Link]((x1, l1+x2)) # put the results in a single array
26 sx = [[Link]((sx1[i], sx2[i])) for i in range(n1)]
27
28 sx_max = [[Link](abs(sx[i])) for i in range(n1)] # absolute max values
29 print(f"Absolute maximum values: {[Link](sx_max)}")

Absolute maximum values: [ 1.1164e+00 1.6755e-01 3.1609e-01 1.0000e+00]

 Solution to Bended Beams 337

1 # Plot the distribution curves:

2 X_data = [X for i in range(n1)] # put x coordinates to a list
3 y_data = [sx[i]/sx_max[i] for i in range(n1)] # func. values to a list
4 labels=["Deflection", "Rotation", "Moment", "Shear force"]
5 gr.plot_fs(X_data, y_data, labels, 'x', 'Solutions', 'temp')
6 # x-label, y-label, file-name

As shown, we can have endless examples done using our powerful

solver1D4() for thin beams. The best approach for readers to learn may be
to spend some time to play with the code: first to repeat the examples given
above, and then try other examples. The following section uses solver1D4()
for analysis of thermal stress in beams.

8.17 Thermal stress*

8.17.1 Setting of the problem

For beams that are overconstrained (indeterminate), a temperature gradient
across the beam thickness can also result in additional thermal stresses,
when the material is not allowed to expand freely. The general constitutive
equation is given in Eq. (5.87). For thin beams, a temperature gradient
relates to the curvature (derivative of the rotation of the cross-section) of
the beam. The relationship can be derived as follows.
Consider a free-body with length dx and height h from a thin beam, as
shown in Fig. 8.16.
338 Mechanics of Materials: Formulations and Solutions with Python

Figure 8.16. Free-body diagram for a thin beam experiencing a temperature change
across its thickness.

8.17.2 Governing equation for deflection caused by the

temperature gradient
The temperature change on the top surface of the beam is assumed T and
that on the bottom surface is −T , and the temperature varies linearly across
the thickness of the beam. Consider small displacements resulting from the
temperature difference, the change in rotation θT of the cross-section is eval-
uated as follows:
du+ dx − du− dx αT − (−αT ) 2T
dθT = − =− dx = −α dx (8.38)
h h h

Tg

where du+ and du− are, respectively, the incremental displacements on the
top and bottom surfaces of the beam caused the temperature change there,
α is the thermal expansion coefficient of the material of the beam, and Tg is
the temperature gradient. The negative sign on the front of the right-hand
side is because positive gradient results in negative rotation. Equation (8.38)
can be rewritten as
dθT
= −αTg (8.39)
dx
dθT
where dx is the curvature of the beam.
 Solution to Bended Beams 339

Using Eqs. (8.2) and (8.39), we obtain:

∂ 2 vT
= −αTg (8.40)
∂x2
where vT is the deflection resulting from the temperature gradient.

8.17.3 Solution of deflection resulting from

temperature gradient
Assuming Tg is constant along x-axis, the foregoing equation can be inte-
grated twice, which gives
1
vT = − αTg x2 + c0 x + c1 (8.41)
2
Case-I: Assuming the beam is clamped at x = 0, we have vT = 0 and
dvT
dx = 0. This leads to both c0 = 0 and c1 = 0. Hence, the deflection
resulting from the temperature gradient is found as
αTg 2
vT = − x (8.42)
2
Case-II: Assuming the beam is clamped at x = l, we have vT = 0 and
dvT αTg l2
dx = 0. This leads to c0 = αTg l and c1 = − 2 . Hence, the deflection
resulting from the temperature gradient is found as
αTg
vT = − (l − x)2 (8.43)
2
The code to find this expression is given as follows:

1 x, l = symbols("x, l")
2 vT, Tg, , c0, c1 = symbols("v_T, T_g, , c0, c1 ")
3 vT = - *Tg/2*x**2 + c0*x+c1
4 sln = [Link]([[Link](x,l), [Link](x).subs(x,l)], [c0, c1])
5 sln
 
Tg l2 α
c0 : Tg lα, c1 : −
2

1 vTl = [Link]({c0:sln[c0], c1:sln[c1]})

2 [Link](Tg)

Tg α −l2 + 2lx − x2
2
340 Mechanics of Materials: Formulations and Solutions with Python

8.17.4 Solutions of indeterminate beams in temperature

gradient
Since we have already the beam solver, and the just derived formulas for
the deflection caused by a temperature gradient, finding the solutions of
deflection, rotation, moment and shear force in beams is straightforward.
The procedure is as follows:

1. Find solution to beams for mechanical effects using the solver1D4(), as

we did earlier, which gives the mechanical deflection v for the beam, in
analytic formulas.
2. Add the thermal deflection vT to the mechanical deflection v.
3. Enforce BCs to the combined deflection to determine the relevant
unknowns, and to obtain the solution for the final deflection.
4. After the final deflection is obtain, use Eq. (8.2) to compute the rotation.
5. The moment can be computed using Eqs. (8.9) and (8.40):
 2   2 
d vcombined ∂ 2 vT d vcombined
Mz = EIz − = EIz + αTg (8.44)
dx2 ∂x2 dx2
This is because of the constitutive equation change due to the temperature
gradient. In other words, we need to remove the thermal curvature to
compute the moment. This treatment is similar to the equation derived
in the previous chapter for 1D bars.
6. Use Eq. (8.14) to compute the shear force.

Let us look at some examples.

8.17.5 Example: Simple-clamped beams subjected to

temperature gradient
Consider beams of length l with cross-section having a second moment of
area I. It is made of material with Young’s modulus E. It is subjected
to temperature gradient Tg across its thickness. Its boundaries are simple-
clamped (s-c).

1. Derive formulas for solutions, including deflection, rotation, moment, and

shear force in the beam.
2. Plot the distribution of all these solutions, using unit for all its
parameters.

We write the following code to complete the task. First, define all neces-
sary symbolic variables.
 Solution to Bended Beams 341

1 E, I, A, l = symbols('E, I, A, l ', nonnegative=True)

2 x, q = symbols('x, q ') # geometry & force
3
4 # variables for displacement boundary conditions (DBCs):
5 v0, 0, vl, l = symbols('v0, 0, v_l, _l')
6
7 # variables for force boundary conditions (FBCs):
8 V0, M0, Vl, Ml = symbols('V0, M0, V_l, M_l')
9
10 by = [Link]('b_y')(x) # for force function

[Link] Find the general displacement solution for an s-c beam

1 by = 0 # no mechanical loading
2 vM = solver1D4(E, I, by, l,v0, 0,vl, l,V0,M0,Vl,Ml,key='s-c')
3 print(f'Is solution correct? {by/E/I==vM[4]}; The solution u(x) is:')
4 vM[0] # mechanical deflection vM(x)

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? True; The solution u(x) is

xθl 3v0 x 3vl x x3 θl v0 x3 vl x3 M0 lx M0 x2 M0 x3

v0 − − + + 2 + 3 − 3 − + −
2 2l 2l 2l 2l 2l 4EI 2EI 4EIl
This is the general displacement solution for an s-c beam for possible internal
mechanical forces as variables on the boundaries.

[Link] Add in the displacement caused by temperature gradient

1 vT, Tg, = symbols("v_T, T_g, ") # Tg: temperature gradient

2 vT = - *Tg/2*(l-x)**2 # deflection caused by temperature gradient
3 v = vT + vM[0] # total displacement
4 v

Tg α (l − x)2 xθl 3v0 x 3vl x x3 θl v0 x3 vl x3

− + v0 − − + + 2 + 3 − 3
2 2 2l 2l 2l 2l 2l
M0 lx M0 x2 M0 x3
− + −
4EI 2EI 4EIl

[Link] Find the combined displacement using BCs

For our s-c beam, the clamped BCs at the right-end is zero. We use the BCs
to simplify the general solution.
342 Mechanics of Materials: Formulations and Solutions with Python

1 v = [Link]({vl:0, l:0,M0:0}) # set known values for BC variables

2 v # simplified total displacement

Tg α (l − x)2 3v0 x v0 x3
− + v0 − + 3
2 2l 2l

The combined displacement at the left-end must be zero. We thus use

this condition to find variable v0 .

1 sln = [Link]([Link](x,0), v0) # solve for displacement at left-end

2 [Link](sln[0], 'v0 is found as:')

Tg l2 α
v0 is found as :
2

1 v_T = [Link]({v0:sln[0]}) # put the solution for v0 back

2 v_T # final solution for the deflection caused the temperature change

Tg l2 α 3Tg lxα Tg α (l − x)2 Tg x3 α

− − +
2 4 2 4l

This is the final solution for the deflection caused by the temperature
gradient. It can then be used to find the other solutions: rotation, moment,
and shear force. We put this set of solutions in a list:

1 S = [v_T, v_T.diff(x,1), E*I*(v_T.diff(x,2)+ *Tg), -E*I*v_T.diff(x,3)]

2 S = [[Link]() for si in S]
3 S # formulas for the set of solutions

 
Tg xα l2 − 2lx + x2 Tg α l (l − 4x) + 3x2 3EITg xα 3EITg α
, , , −
4l 4l 2l 2l
These are all functions of x, except the shear force that is constant.

[Link] Plot the distribution of solutions

Finally, we plot the solutions using gr.plot2curveS() that finds also some of
the extreme values (maximum, minimum, and values at the boundaries.)
To generate the plots for viewing the variation of the solutions, we set all
the parameters to unit.
 Solution to Bended Beams 343

1 Sx=[S[i].subs({E:1,I:1,l:1,q:1,Tg:1, :1}) for i in range(len(S))]

2 Sx # This is a set functions of only x

 
x x2 − 2x + 1 3x2 1 3x 3
, −x+ , , −
4 4 4 2 2

1 title = [["Deflection", "Rotation"], ["Moment", "Shear force"]]

2 for i in range(len(title)):
3 gr.plot2curveS(Sx[2*i:2*(i+1)], 0., 1., title=title[i])
4
5 [Link]()

Maximum Deflection value=3.703e-02, at x=0.33

Maximum Rotation value=2.500e-01, at x=0.0
Extreme Deflection values=[ 0.000e+00 3.703e-02],
at x=[ 1.000e+00 3.300e-01]
Critical Rotation values=[ 0.000e+00 1.675e-03],
at x=[ 1.000e+00 3.300e-01]
Deflection values at boundary =(0.0, 0.0)
Rotation values at boundary =(0.25, 0.0)

Maximum Moment value=1.500e+00, at x=1.0

Maximum Shear force value=-1.500e+00, at x=0.0
Extreme Moment values=[ 0.000e+00],
at x=[ 0.000e+00]
Critical Shear force values=[-1.500e+00],
at x=[ 0.000e+00]
Moment values at boundary =(0.0, 1.5)
Shear force values at boundary =(-1.5, -1.5)
344 Mechanics of Materials: Formulations and Solutions with Python

It is seen that although there is no mechanical loading on the beam, the

beam deforms with moments and shear force built up along the beam. This
is purely due to the temperature gradient.

8.17.6 Example: Beams simple-simple supported

subjected to temperature gradient
Consider the same problem as the previous one, but change s-c to s-s support.

[Link] Find the general displacement solution for an s-s beam

1 by = 0 # no loading
2 vM = solver1D4(E, I, by, l, v0, 0, vl, l, V0, M0, Vl, Ml, key='s-s')
3 print(f'Is solution correct? {by/E/I==vM[4]}; The solution u(x) is:')
4 vM[0]

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? True; The solution u(x) is

v0 x vl x M0 lx M0 x2 M0 x3 Ml lx Ml x3
v0 − + − + − − +
l l 3EI 2EI 6EIl 6EI 6EIl
This is the general displacement solution for an s-s beam with possible
mechanical forces in the beam as variables on the boundaries.

[Link] Add in the displacement caused by temperature gradient

1 vT, Tg, = symbols("v_T, T_g, ") # Tg: temperature gradient

2 vT = - *Tg/2*(l-x)**2 # disp. caused by temperature gradient
3 v = vT + vM[0] # total displacement
4 v

Tg α (l − x)2 v0 x vl x M0 lx M0 x2 M0 x3 Ml lx Ml x3
− + v0 − + − + − − +
2 l l 3EI 2EI 6EIl 6EI 6EIl

[Link] Find the combined displacement

For s-s beams, we can leave v0 as the only unknown and set the rest zero.

1 v = [Link]({vl:0, l:0,M0:0,Ml:0}) # set known values for BC variables

2 v # simplified total displacement

Tg α (l − x)2 v0 x
− + v0 −
2 l
The combined displacement at the left-end must be zero. We thus use
this condition to find variable v0 .
 Solution to Bended Beams 345

1 sln = [Link]([Link](x,0), v0) # solve for displacement at left-end

2 sln

 
Tg l2 α
2

1 v_T = [Link]({v0:sln[0]}) # put the solution for v0 back

2 v_T

Tg l2 α Tg lxα Tg α (l − x)2
− −
2 2 2

This is the final solution for the displacement. It can then be used to find
the other solutions: rotation, moment, and shear force. We put this set of
solutions in a list:

1 S = [v_T, v_T.diff(x,1), E*I*(v_T.diff(x,2)+ *Tg), -E*I*v_T.diff(x,3)]

2 S = [[Link]() for si in S]
3 S # formulas for the set of solutions

 
Tg xα (l − x) Tg α (l − 2x)
, , 0, 0
2 2

It is seen that although there is no mechanical loading on the beam, the

beam deforms with rotation. This is purely due to the temperature gradient.
The moments and shear force are zero because of the s-s support. The tem-
perature gradient causes deflection, but no stress is built up. The s-s beam is
determinate structure. It does not constraint the thermal rotation and hence
will not create thermal moments. This is the major difference from the s-c
beam that is indeterminate.

[Link] Plot the distribution of solutions

Finally, we plot the solutions using gr.plot2curveS() that finds also some of
the extreme values (maximum, minimum, and values at the boundaries.)
To generate the plots for viewing the variation of the solutions, we set all
the parameters to unit.
346 Mechanics of Materials: Formulations and Solutions with Python

1 Sx=[S[i].subs({E:1,I:1,l:1,q:1,Tg:1, :1}) for i in range(len(S))]

2 Sx # This is a set functions of only x

 
x (1 − x) 1
, − x, 0, 0
2 2

1 title = [["Deflection", "Rotation"], ["Moment", "Shear force"]]

2 for i in range(len(title)):
3 gr.plot2curveS(Sx[2*i:2*(i+1)], 0., 1., title=title[i])
4
5 [Link]()

Maximum Deflection value=1.250e-01, at x=0.5

Maximum Rotation value=5.000e-01, at x=0.0
Extreme Deflection values=[ 1.250e-01],
at x=[ 5.000e-01]
Critical Rotation values=[ 0.000e+00],
at x=[ 5.000e-01]
Deflection values at boundary =(0.0, 0.0)
Rotation values at boundary =(0.5, -0.5)

Maximum Moment value=0.000e+00, at x=0.0

Maximum Shear force value=0.000e+00, at x=0.0
Extreme Moment values=[ 0.000e+00],
at x=[ 0.000e+00]
Critical Shear force values=[ 0.000e+00],
at x=[ 0.000e+00]
Moment values at boundary =(0.0, 0.0)
Shear force values at boundary =(0.0, 0.0)

As shown there is no thermal moment and force in the beam, because it

is determinant. The thermal effects are only on the deformations (deflection
and rotation).
 Solution to Bended Beams 347

8.18 Remarks
The beam solver solver1D4() developed in this chapter is a powerful tool
for thin beams. It is quite easy to use and applicable to a wide range of
problems, including frames and thermal effect analysis.
In the existing textbooks and literature, there is a quite common approach
by drawing shear force and moment diagrams, and then use Eq. (8.17) and/or
(8.9) to obtain the deflection via integration. This may work for statically
determinate beams because it may be easier for drawing the shear force
and moment diagrams. However, use of beam solver solver1D4() is much
more straightforward and effortless. All the diagrams, deflection, rotation,
moment and shear force can all obtained in one run, with much fewer manual
operations.
For large frame structures or frame with truss members, it is much easier
to use a more systematic approach based on the concept of stiffness matrix
assembly, as we often do in the finite element method (FEM) [1]. Using
FEM, the number of members in a structure can be millions and even bil-
lions because no manual operations for setting various types of condition are
needed.
The next chapter deals with another important structural member: bars
subjected to torsion.

Reference
[1] G.R. Liu and S.S. Quek, The Finite Element Method: A Practical Course.
Butterworth-Heinemann, New York, 2013.
This page intentionally left blank
 Chapter 9

Torsion of Bars

1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.

2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 init_printing(use_unicode=True) # For latex-like quality printing
11 #np.set_printoptions(precision=4,suppress=True) # Digits in print-outs
12 np.set_printoptions(formatter={'float': '{: 0.4e}'.format})

Structure members or components are frequently used in power transmission

or similar purposes. Figure 9.1 shows a rotor of a modern steam turbine by
Siemens Pressebild ([Link] as a typical example. The
shaft used in it is one of the most critical components in such a rotational
machine undergoing torsional deformation.
Most shafts are essentially made of bars with circular cross-section, which
are quite effective in many ways and are relatively easier to deal with in
mechanics aspects. However, there are also bar members with other cross-
sectional shapes, as shown in Fig. 9.2. These types of bars require more
intensive stress analysis for proper design.
This chapter discusses torsional bar members with various cross-sectional
shapes. The objective is to find the relationship between the twist angle

349
350 Mechanics of Materials: Formulations and Solutions with Python

Figure 9.1. A modern steam turbine, where shafts are used for power transmission.
The shaft undergoes torsional deformation.
Source: Image from Siemens Pressebild available at Wikimedia Commons under
CC BY-SA 3.0 license.

Figure 9.2. A bar with square cross-section subjected to a torque, resulting in compli-
cated deformations in the cross-section.
Source: Image from Orion-8 at Wikimedia Commons under the CC BY 3.0 license.

(an important measure of deformation) and the applied torque (a typical

force application) and stresses (critical for failure analysis) on the cross-
section of the members. Since our focus is purely on torsional deformation,
the members or components should have a slender geometry, like a bar or
rod. Thus, we will call all these types of torsional members and components
 Torsion of Bars 351

a bar because it captures the major geometric feature, allowing proper

assumptions to be made. The cross-section of the bar can be arbitrary as
long as its cross-sectional dimensions are sufficiently smaller than its length.
The torques are applied about the axial axis of the bar.
The major challenge in dealing with torsional bars is its three-
dimensional (3D) nature: the torque is applied about is axial axis (z-axis).
The stresses, on the other hand, occur on its cross-sectional plane (x–y
plane). The art of the technique lies in converting the 3D problem into a 1D
or at most a 2D one, so that solutions can be found analytically. Detailed
formulations will be provided in this chapter on how this is done.
If the bar is bulky, it needs to be treated in principle as a true 3D problem.
The most effective techniques will be numerical, including the finite element
method (FEM) [1], smoothed FEM [2], and even the meshfree methods [3].
All these techniques require intensive formulation and substantial coding.
They are thus beyond the topic of this book.
This chapter is written in reference to the textbooks in Refs. [4–8]. Our
study begins with the simplest and most fundamental case: bars with a
circular cross-section, which are the most widely used in engineering.

9.1 Bars with circular cross-section

9.1.1 Problem description

Consider a solid cylindrical bar with a circular cross-section. It has a cross-
sectional area A and length L. It is defined with the Cartesian coordinates
(x, y, z). Let z be the axial coordinate, and it is perpendicular to the cross-
section. It is subjected to torque T about the z-axis. An originally straight-
line AB on the lateral surface of the bar becomes a helical curve AB* due to
the twist deformation of the bar, as shown in Fig. 9.3. Our task is to find out
how the twist angle relates to the applied torque and the resulting stresses
in the bar.

9.1.2 Assumptions
Clearly, any cross-section of the bar undergoes shearing, which can result in
shear stresses. The shearing deformation accumulates to a twist angle that
may change along the z-axis. To formulate this problem properly, we assume
the following:

1. The plane of the cross-section of the bar remains perpendicular to the

z-axis.
352 Mechanics of Materials: Formulations and Solutions with Python

Figure 9.3. A bar with circular cross-section subjected to a torque, undergoing a twist
deformation.

2. The rotation is small, and the cross-section rotates as a rigid plane about
the z-axis.
3. The rotation Θ of a given section, relative to the base plane z = 0, depends
linearly on the distance z from the base plane.

Based on these assumptions, we have

Θ = ϑt z (9.1)

where ϑt is the twist angle per unit length and is a constant (does not change
with x, y, or z). It is also sometimes called torsional curvature (the rate of
angular change with respect to z).
These assumptions practically fix the dependence of the problem on z.
The problem becomes 2D with displacements on the cross-section of the bar.

9.1.3 Displacements on the cross-section

Because the cross-section undergoes a rotation about the z-axis and the
rotation is small, the displacements on the cross-section can be shown in
Fig. 9.4.
Because the cross-section rotates rigidly, these displacement components
can be obtained using simple geometry relations. They are written as

u = −yΘ; v = xΘ; w=0 (9.2)

or

u = −yϑt z; v = xϑt z; w=0 (9.3)

 Torsion of Bars 353

Figure 9.4. Displacement components on the cross-section of a circular bar subjected

to a torque.

It appears that the displacements depend on x, y, and z: a 3D problem.

It is controlled by one parameter: ϑt or Θ.

9.1.4 Shear strains and stresses on the cross-section

Using the small strain–displacement relations for a 3D solid given in
Chapter 4 and since ϑt is independent of x, y, and z, we can find the strains
on the bar readily:
 
∂u ∂v ∂w 1 ∂u ∂v
εxx = = 0; εyy = = 0; εzz = = 0; εxy = + =0
∂x ∂y ∂z 2 ∂y ∂x
(9.4)
and
∂u ∂w ∂w ∂v
2εzx = γzx = + = −yϑt ; 2εzy = γzy = + = xϑt (9.5)
∂z ∂x ∂y ∂z

It is clear that there are only two nonzero strain components. All these
can also be obtained using the following codes. First, define the displacement
vector:
354 Mechanics of Materials: Formulations and Solutions with Python

1 x, y, z = symbols("x, y, z") # define symbolic coordinates

2 X = Matrix([x, y, z])
3 t = symbols(" _t") # twist angle per length
4
5 U = Matrix([-y* t*z, x* t*z, 0]) # Assumed displacement vector [u,v,w]
6 U

⎡ ⎤
−yzϑt
⎢ ⎥
⎣ xzϑt ⎦
0

1 # Compute the strains (see Chapter 4):

2 U_g = gr.grad_vf(U, X) # displacement gradient
3 ε = (U_g + U_g.T)/2 # strain tensor,symmetric
4 ε

⎡ ⎤
0 0 − yϑ2 t
⎢ xϑt ⎥
⎣ 0 0 2 ⎦
− yϑ2 t xϑt
2 0
The code finds again that there are only two nonzero strains, and they
are both shear strains.
For isotropic materials, there can only be two shear stresses on the cross-
section of the bar, which relate to shear strains by only the shear modulus
G (see Chapter 5). These are

σzx = Gγzx = −yGϑt ; σzy = Gγzy = xGϑt (9.6)

where the shear modulus G is a material constant. Note that these stresses
are independent of z, implying that these stresses are uniform along the axial
axis of the bar.
With this simple stress state, we can perform some interesting analysis
on it using the following code. The stress tensor at any point in a torsion
bar is given as follows:

1 G = symbols("G")
2 σ_tensor = G*(2*ε) # _ij = 2 _ij
3 σ_tensor

⎡ ⎤
0 0 −Gyϑt
⎢ ⎥
⎣ 0 0 Gxϑt ⎦
−Gyϑt Gxϑt 0
 Torsion of Bars 355

[Link] Divergence of the stress states in a circular bar

This set of stresses satisfy the equilibrium equation for 3D solids given in
Chapter 6 (considering a static problem with no body force). This is because
their divergences (see Chapter 3) are zeros:

1 div_σ_tensor = gr.div_tensorF(σ_tensor, X) # a vector of 3 functions

2 div_σ_tensor # Or use: vector.div_tensor(σ_tensor.T)

⎡ ⎤
0
⎢ ⎥
⎣0⎦
0
Clearly, the stress tensor of torsion is a divergence-free tensor. Therefore,
the equilibrium equations are satisfied for the free body forces in the bar.

[Link] Curl of the stress vector field in a circular bar

The stresses on the z-plane in a circular bar under torsion can form a vector
function with three components: σzx , σzy , σzz . Let us take a look at the curl
(see Chapter 2) of this vector field formed by these torsional stresses.

1 # Define a vector field as symbolic functions of (x, y, z)

2 σ = symbols("σ_ ")
3 σ = σ_tensor[2,:] # get the vector on z-surface
4 σ

−Gyϑt Gxϑt 0

1 curl_F = gr.curl_vf(σ , X)
2 curl_F # Curl of the torsional stress vector field

⎡ ⎤
0
⎢ ⎥
⎣ 0 ⎦
2Gϑt
This means that:

• The rates of rotation of the torsional stress vector field around the x- and
y-axis are both zero at any point on the cross-section. The field has zero
circulation around these two axes.
• It has a positive constant circulation at any point on the cross-section
around the z-axis. The shear stresses are in the x–y plane, and the torque
(curl) is with respect to the z-axis. This is a typical feature of a circular
bar under pure torsion.
356 Mechanics of Materials: Formulations and Solutions with Python

Also, the sum of these two shear stress components is in the circum-
ferential direction on the cross-section of the bar, which is the red arrow
in Fig. 9.4. This can be confirmed by viewing the stress state in the polar
coordinates (r, θ).

9.1.5 Distribution of shear stress on the cross-section

Again, because the cross-section undergoes a rigid body rotation about the
z-axis and the rotation is small, the displacements on the cross-section in
the polar coordinate (r, θ) are

ur = 0; uθ = rϑt z; w=0 (9.7)

The only nonzero displacement is uθ , which is in the circumferential

θ-direction (the red arrow rΘ in Fig. 9.4). This again means that the problem
is 1D. Reader may derive Eq. (9.7) via a coordinate transformation using
Eq. (9.3).
The stress in the polar coordinates can be obtained using the stress tensor
transformation rule:

σ  = TσT (9.8)

where σ  is the stress tensor in the polar coordinate system, σ is the stress
tensor in the Cartesian coordinates with components given in Eq. (9.6), and
T is the transformation matrix defined in Section 2.10.
Let us use the following Sympy code to get it done.

1 # Compute the stress tensor on the cross-section of the bar:

2
3 , = "\u03D1", "\u03B8" # Define symbol in hex code
4 G, t = symbols("G, _t") # Define symbolic variables
5 r, = symbols("r, ") # for polar coordinates
6
7 # write the stresses in terms of r and :
8 σzx = -r*[Link]( )*G* t # y=rsin
9 σzy = r*[Link]( )*G* t # x=rcos
10
11 S = Matrix([[ 0 , 0 , σzx], # stress state for torsion w.r.t z-axis
12 [ 0 , 0 , σzy],
13 [σzx, σzy, 0 ]])
14 S

⎡ ⎤
0 0 −Grϑt sin (θ)
⎢ ⎥
⎣ 0 0 Grϑt cos (θ) ⎦
−Grϑt sin (θ) Grϑt cos (θ) 0
 Torsion of Bars 357

Readers are not to be confused about the symbols θ and ϑt . The former
is a coordinate of the polar system, and the latter is the twist angle per unit
length.
Perform coordinate transformation for the stresses.
1 # coordinate transformation for the stresses:
2 T = [Link]( ) # transformation matrix for rotation w.r.t. z
3 Sp=T@[email protected] # S: stress tensor; @: the shorthand of [Link]
4 gr.simplifyM0(Sp) # simply the terms in the output

⎡ ⎤
0 0 0
⎢ ⎥
⎣0 0 Grϑt ⎦
0 Grϑt 0
We found that the stress components are all zero everywhere in the polar
coordinates:

σrr = σrθ = σrz = σθθ = σzz = 0 (9.9)

The only nonzero component is

σzθ = rGϑt (9.10)

which changes linearly in r and is in the circumferential direction, as shown

in Fig. 9.4. The largest shear stress is registered by the surface of the bar.
Because there is only one shear stress on the cross-section of a bar, the stress
is often denoted as τ , when dealing with torsion:

τ = rGϑt (9.11)

The strain should also have only one component: 2εzθ = rϑt , also chang-
ing linearly in r.
Note that, in the coordinate transformation, we used an arbitrary angle θ.
However, the transformed results do not depend on θ at all. That means the
result obtained is applicable to any θ location on the cross-section. This is
the consequence of the rigid cross-section rotation assumption. The problem
is essentially 1D.

9.1.6 Boundary condition on the lateral surface

of the bar
Because the only nonzero stress is in the circumferential direction, it will not
result in any stress on the lateral surface of the bar. Therefore, the zero-stress
boundary condition on the lateral surface of the bar is also satisfied.
358 Mechanics of Materials: Formulations and Solutions with Python

A rigorous proof is given by the code right above Eq. (9.9) which gives
for any θ: σrr = σrθ = σrz = 0. These are all the possible stress components
on the lateral surface of a circular bar, and they all are zero regardless of θ.
Note also that σrr = σrθ = σrz = 0 is true for any r. This means that
any cylinder in the circular bar rotates together without an inter-surface
stress. In other words, if we put cylinders together to form a circular bar,
the assembly behaves the same as one solid circular bar in terms of pure
torsion, if they all have the same twist angle.
Note that in usual applications, the shear stress distribution on the cross-
section at the two ends of the bar may not be linearly distributed. In those
cases, the solution is still applicable far away from the ends, based on the so-
called Saint-Venant principle. This implies that our formulation applicable
in general for slender bars for locations away from the two ends.

9.1.7 Equilibrium on the cross-section, twist-torque

relation
We found the stresses and know their distribution on the cross-section. The
amplitude of the stresses must depend on the level of the torque applied.
For example, if the torque is zero, the stresses must be zero. We need now to
establish the relationship between the stresses and the applied torque. For
this, naturally, we must invoke equilibrium conditions.
Let us now examine the equilibrium status on the cross-section of the
bar subjected to a torque T . The sum of all moments with respect to the
z-axis should be zero. Thus, the torque T must result from these two shear
stresses.
With the help of Fig. 9.5, we have

Mz = 0 : T = (xσzy − yσzx )dA = (x2 Gϑt + y 2 Gϑt )dA

A A
(9.12)
= Gϑt (x2 + y 2 ) dA = Gϑt r 2 dA = Gϑt J
A   
r2  A  
J

where J is the polar second moment of area. For circular cross-section,

we have
R
πR4
J= r 2 dA = r 2 
rdθ dr = 2π r 3 dr = (9.13)
A A 0 2
dA
 Torsion of Bars 359

Figure 9.5. Shear stresses on the cross-section of a circular bar result in a torque.

9.1.8 On polar second moment of area

The polar second moment of area J is a property of the cross-section of a
bar, and it is the most important one for all bars subject to torsion. The
twist angle and the shear stress all depend on it. For circular cross-section,
we have a very nice Eq. (9.13) for computing it. It is derived using the polar
coordinate system, as shown above. The word “second” stands for r 2 in the
integral in Eq. (9.13).
For other shapes of cross-section, the formulas can be very complicated,
and often the polar coordinate system is not usable in the derivation. We
call such a J equivalent polar second moment of area. We need to use the
word “polar” because the second moment of area in Cartesian coordinate is
a different one. A better unified name might be torsional second moment
of area for bars with any type of cross-section.
Our strategy for dealing with bars with other types of torsional bars is to
find the equivalent polar second moment of area of the cross-section of the
bar. Once this is done, the formulas for computing the twist angle and the
shear stress will be the same.
360 Mechanics of Materials: Formulations and Solutions with Python

9.1.9 Twist-angle and torque relations

Equation (9.12) gives the twist-angle per-unit-length, ϑt :
T
ϑt = (9.14)
GJ
If the bar is fixed at z = 0, the (total) twist-angle at an arbitrary location
z becomes
T
Θ= z (9.15)
GJ
The total twist angle for a bar with length L becomes
L
Θ= T for 1D bars subject to torque T (9.16)
GJ
where GJL is the torsional stiffness of the bar, and GJ the torsional
stiffness per length.
Comparing the solution for a 1D bar subjected to axial force F , which is
given in Chapter 7:
L
uR = F for 1D bars subject to axial force F (9.17)
EA
The formula is exactly the same if we replace
uR by Θ; E by G; A by J (9.18)
Equation (9.16) can also be written as

GJ
T = Θ (9.19)
L
L
In this case, GJ is the torsional flexibility of the bar.
Using Eq. (9.11), we have the shear-stress torque relation:
Tr
τ = rGϑt = (9.20)
J
The maximum shear stress is registered at the furthest point from the
z-axis, r = R (by the lateral surface of the bar):
TR
τmax = (9.21)
J
All these formulas are useful in designing torsional bars. These formulas
are also the base formulas for many other types of bars. The difference will
be on the calculation of J.
 Torsion of Bars 361

9.1.10 Continuous formulation

Consider a small cell dz in the bar, as shown in Fig. 9.3. The twist angle for
it becomes
T (z) dΘ T (z)
dΘ = dz; or = (9.22)
GJ dz GJ
This is the differential equation (DE) for twist-torque relation, which
allows the torque to be a function of z, implying that 
one can apply torque
in a distributed manner. In this case, the equilibrium Mz = 0 is enforced
locally at any location z in the integral manner over the cross-section of
the bar. Equation (9.22) is a simple 1D DE of first order and can be solved
with ease.
In general, even J and/or G can also be a function of z, but the bar must
be “long” to justify being 1D.

9.1.11 Bars with hollow cylindrical cross-section

In practice, bars designed for taking torsional loads are often with a hollow
cylindrical cross-section, as shown in Fig. 9.6. This is because the material
near the axial axis does not take much stress, as shown in Fig. 9.4, and can
be removed.

Figure 9.6. Bar with hollow cylindrical cross-section.

362 Mechanics of Materials: Formulations and Solutions with Python

In such a case, all the formulations derived so far still apply. The only
change is the calculation of the polar second moment of area J. It is calcu-
lated using
RO
π(RO 4 − RI 4 )
J= r 2 dA = 2π r 3 dr = (9.23)
A RI 2

where RI is the inner radius, and RO is the outer radius of the cylinder. The
maximum shear stress is registered at the furthest point from the z-axis,
r = RO (the outer radius):
T RO
τmax = (9.24)
J
The cross-section areas is computed using

A = π(RO 2 − RI 2 ) (9.25)

9.1.12 Bars with thin-wall cross-sections

If the hollow cylinder is very thin with thickness t, we have RI ≈ RO , and
RO − RI = t. Equation (9.23) gives

π(RO 4 − RI 4 ) π(RO 2 + RI 2 )(RO + RI )(RO − RI )

J= = ≈ 2πRm
3
t
2 2
(9.26)
where Rm = RO +R
2
I
is the mean radius of the cylinder.
Equation (9.26) can be reformulated as
2 )2
4t(πRm 4tA2m
J ≈ 2πtRm
3
= = (9.27)
2πRm lm
where Am is the area of the circle with mean radius Rm , and is computed
using

Am = πRm
2
(9.28)

and lm is the length of the mean perimeter of the circle with Rm .

It is found that the following equation works for thin-wall cylinders with
arbitrary cross-section shape, as long as the cross-section is enclosed. We
thus have the general formula:

4tA2m
J= (9.29)
lm
 Torsion of Bars 363

where Am is the area formed by the middle line of the cross-section, lm is the
length of the mean perimeter of the cross-section, and t is the thickness of
the wall. This is because cross-sections regardless of the shape all have these
three geometric features: Am , lm , and t. The J value should be proportional
to t, inversely proportional to lm , and A2m is needed for dimension match,
similar to the thin wall circular cylinder. Detailed rigorous (but lengthy)
derivation leading to Eq. (9.29) can be found in the following section and in
Sections 9.2 and 9.6.
Note that when the cross-section is not closed (have a cut), the J value
is drastically reduced (see examples in Section 9.6).

9.1.13 Example: A shaft with hollow circular

cross-section subject to a torque
Consider a bar made of Grade-304 stainless steel with shear modulus
G = 88 GPa. The cross-section is hollow circular with inner radius of
RO = 30 mm and outer radius of RO = 50 mm. It is fixed at z = 0, and
subjected to a torque of T = 1000 Nm at z = L, as shown in Fig. 9.7.
Calculate:

1. the maximum shear stress on the cross-section of the shaft;

2. the twist angle per unit length;
3. the stress components on the cross-section and by the lateral surface in
the Cartesian coordinates on the x-axis;
4. the stress components on the cross-section and by the lateral surface in
the Cartesian coordinates at θ = 45◦ from the x-axis;
5. without calculation, give the stress components on the cross-section and
by the lateral surface in the Cartesian coordinates on the y-axis;
6. compute the principal stresses and maximum shear stress in the shaft.

Figure 9.7. Stress on a hollow cylindrical cross-section of a uniform bar.

364 Mechanics of Materials: Formulations and Solutions with Python

Solution to 1: The maximum shear stress in the shaft.

1 RI, RO = symbols("R_I, R_O") # Define symbolic variables

2 T = 1000 # Nm, torque
3 RI = 0.03; RO = 0.05 # m, radii
4 J = [Link] * (RO**4 - RI**4)/2
5 print(f"The polar second moment of area J = {J} (m^4)")
6
7 _RO = T*RO/J
8 print(f"The maximum on the cross-section = { _RO*1.0e-6} (MPa)")

The polar second moment of area J = 8.545132017764239e-06 (m^4)

The maximum on the cross-section = 5.851284672496151 (MPa)

Solution to 2: The twist angle per unit length.

1 G = 88.0e9 # Pa
2 _t = T/(G*J)
3
4 print(f"Twist-angle/length={ _t:.4e} rad/m or {np.rad2deg( _t):.4e}∘/m")

Twist-angle/length=1.3298e-03 rad/m or 7.6194e-02 /m

Solution to 3: The stress components on the cross-section and by lateral

surface in the Cartesian coordinates on the x-axis.

1 = 0.0 # On the x-axis, =0

2 x = RO*[Link](np.deg2rad( ))
3 y = RO*[Link](np.deg2rad( ))
4
5 σzx =- _t*G*y
6 σzy = _t*G*x
7
8 print(f"On x-axis: σzx={σzx*1.0e-6:.4e}; σzy={σzy*1.0e-6:.4e} (MPa)")

On x-axis: σzx=-0.0000e+00; σzy=5.8513e+00 (MPa)

Solution to 4: The stress components on the cross-section and by lateral

surface in the Cartesian coordinates at θ = 45◦ from the x-axis.

1 = 45.0
2 x = RO*[Link](np.deg2rad( ))
3 y = RO*[Link](np.deg2rad( ))
4
5 σzx =- _t*G*y
6 σzy = _t*G*x
7
8 print(f"at 45∘: σzx = {σzx*1.0e-6:.4e}; σzy = {σzy*1.0e-6:.4e} (MPa)")

at 45 : σzx = -4.1375e+00; σzy = 4.1375e+00 (MPa)

 Torsion of Bars 365

Solution to 5: On the x-axis, θ = 90 degrees. The solution is found as

follows using the same code for solution 3.
On y-axis: σzx = −5.8513e + 00; σzy = 0.0000e + 00 (MPa)
Solution to 6: Compute the principal stresses and maximum shear stress.

1 = 90.0 # use the stress components on suface-y

2 x = RO*[Link](np.deg2rad( )); y = RO*[Link](np.deg2rad( ))
3
4 σzx =- _t*G*y; σzy = _t*G*x
5 print(f"at 90∘: σzx = {σzx:.4e}; σzy = {σzy:.4e} (Pa)")
6
7 σ_np = [Link]([[ 0 , 0 , σzx], # stress tensor, torsion wrt z-axis
8 [ 0 , 0 , σzy],
9 [σzx, σzy, 0 ]], dtype = float)
10
11 eigs, eigvs = [Link](σ_np)
12 print(f" _max = {(eigs[0]-eigs[-1])/2:.4e}(Pa)")

at 90∘: σzx = -5.8513e+06; σzy = 3.5829e-10 (Pa)

Principal stress/strain (Eigenvalues):
[ 5.8513e+06 1.0074e-74 -5.8513e+06]

Principal stress/strain directions:

[[ 7.0711e-01 6.1232e-17 7.0711e-01]
[-4.3298e-17 1.0000e+00 -4.3298e-17]
[-7.0711e-01 -1.6652e-48 7.0711e-01]]

Possible angles (n1,x)=45.00000000000001 or 135.0

_max = 5.8513e+06(Pa)

1 # Check the angle for the principal axis.

2 T, _ = [Link](45., about = 'y') # rotation axis, angle
3 T@σ[email protected]

array([[ 5.8513e+06, -2.5335e-10, 1.5231e-10],

[-2.5335e-10, 0.0000e+00, 2.5335e-10],
[-1.5231e-10, 2.5335e-10, -5.8513e+06]])

9.1.14 Example: A shaft with two segments subject to

two torques
Consider a shaft consisting of two uniform bars perfectly connected together,
as shown in Fig. 9.8. It is fixed at point A. Both bars have solid circular cross-
section. It is loaded at point B with torque T1 , and at point C with T2 . The
diameters of these two bars are denoted as d1 and d2 , and their length are
denoted as L1 and L2 , respectively, for bar-1 and bar-2. Both bars are made
of Grade-304 stainless steel with shear modulus G = 88 GPa.
366 Mechanics of Materials: Formulations and Solutions with Python

Figure 9.8. A shaft consisting of two uniform bars perfectly connected together.

1. Derive the formula for computing the twist angle at C.

2. Derive the formula for computing the maximum shear stress and its loca-
tion for bar-1.
3. Determine the ratio of these two diameters d1 /d2 , so that the maximum
shear stress these two bars are the same.
4. Let T1 = 100 Nm, T2 = 50 Nm, L1 = 0.5 m and L2 = 1.0 m. Compute
items 1, 2, and 3.

Solution to 1: First, the 2nd polar moment of area for these two bars are:
π(d1 /2)4 πd41 πd42
J1 = = ; J2 = (9.30)
2 32 32
The twist angle at C in reference to B is
T2 L2 32T2 L2
ΘBC = = (9.31)
GJ2 πGd42
In bar-1, the torque is (T1 + T2 ), and hence the twist angle at B in
reference to A is
32(T1 + T2 )L1
ΘAB = (9.32)
πGd41
The total twist angle at C becomes
32(T1 + T2 )L1 32T2 L2
ΘC = ΘAB + ΘBC = +
πGd41 πGd42
  (9.33)
32 (T1 + T2 )L1 T2 L2
= +
πG d41 d42
Solution to 2: Compute the maximum shear stress and its location for
bar-1.
d1
For bar-1, we have R1 = 2 . Thus,
(T1 + T2 )d1 16(T1 + T2 )
τ1max = = (9.34)
2J1 πd31
 Torsion of Bars 367

Solution to 3: Determine the ratio of these two diameters d1 /d2 :

d2
For bar-2, we have R2 = 2 . Hence,

T2 d2 16(T2 )
τ2max = = (9.35)
2J2 πd32

The ratio d1 /d2 is found when τ1max = τ2max :


d1 3 T1 + T2
= (9.36)
d2 T2

Solution to 4: The numerical calculation is done using the following code.

T1 = 100 Nm, T2 = 50 Nm, d1 = 0.03 m, d2 = 0.02 m, L1 = 0.5 m and

L2 = 1.0 m.

1 T1 = 100 ; T2 = 50 # Nm
2 d1 = 0.03; d2 = 0.02 # m
3 L1 = 0.5 ; L2 = 1.0 # m
4 G = 88.0e9 # Pa
5
6 # Question 1:
7 deg = 180./[Link]
8 T12 = T1+T2
9 = 32./([Link]*G)*(T12*L1/d1**4 + T2*L2/d2**4)
10 print(f"Q1: The twist angle at C, = { :.4e} rad/m or { *deg:.4e}∘")
11
12 # Question 2:
13 1max = 16.*G*T12/([Link]*d1**3)
14 print(f"Q2: The max shear stress in bar-1 = { 1max*1.0e-6:.4e} MPa\n"
15 f" on the surface of bar-1 within its entire length")
16
17 # Question 3:
18 Rd12 = (T12/T2)**(1/3)
19 print(f"Q3: The ratio of two diameters 1/ 2 = {Rd12:.4e} \n")

Q1: The twist angle at C, = 4.6889e-02 rad/m or 2.6865e+00∘

Q2: The max shear stress in bar-1 = 2.4899e+12 MPa
on the surface of bar-1 within its entire length
Q3: The ratio of two diameters 1/ 2 = 1.4422e+00

We can check the codes using EXAMPLE 6.2 given in Ref. [4], using the
following code.
368 Mechanics of Materials: Formulations and Solutions with Python

1 L1 = 1.0 ; L2 = 1.27 # m
2 G = 77.0e9 # Pa
3 T1 = 113 ; T2 = 113 # Nm
4 d1 = 0.0224; d2 = 0.0177 # m
5 print(f"The ratio of two diameters 1/ 2 = {d1/d2:.4e}")
6
7 # Question 1:
8 deg = 180./[Link]
9 T12 = T1+T2
10 = 32./([Link]*G)*(T12*L1/d1**4 + T2*L2/d2**4)
11 print(f"Q1: The twist angle at C, = { :.4e} rad/m or { *deg:.4e}∘")
12
13 # Question 2:
14 1max = 16.*T12/([Link]*d1**3)
15 print(f"Q2: The max shear stress in bar-1 = { 1max*1.0e-6:.4e} MPa\n"
16 f" on the surface of bar-1 within its entire length")
17
18 # Question 3:
19 Rd12 = (T12/T2)**(1/3)
20 print(f"Q3: The ratio of two diameters 1/ 2 = {Rd12:.4e} \n")

The ratio of two diameters 1/ 2 = 1.2655e+00

Q1: The twist angle at C, = 3.1217e-01 rad/m or 1.7886e+01∘
Q2: The max shear stress in bar-1 = 1.0241e+02 MPa
on the surface of bar-1 within its entire length
Q3: The ratio of two diameters 1/ 2 = 1.2599e+00

9.2 Bars with enclosed thin-wall cross-sections

Consider a long bar with a uniform cross-section that is a hollow with an

enclosed thin-wall, as shown in Fig. 9.9. The axis of the bar is denoted as z
that is perpendicular to the plane of the cross-section, and passing through
its centroid O. The thickness t of the wall needs not be constant and can
change at different locations, denoted by a curve coordinate s, and local
coordinate r orthogonal to s. Thus, r, s, and z axes form a local orthogonal
coordinate system with the right-hand rule.
The shape of the cross-section can be arbitrary but closed. The bar is
subjected to a torque T with respect to the z-axis, resulting in a twist Θ of
the cross-section by at z = L.
Before our analysis, we make the following assumptions:

1. The wall thickness is very small compared to the other dimensions of the
cross-section.
2. The shear stress direction is along the middle line of the thin-wall
boundary.
 Torsion of Bars 369

Figure 9.9. A long bar with a uniform cross-section that is a hollow with an enclosed
thin-wall. It is fixed at z = 0 and subjected to a torque T with respect to the z-axis of the
bar, and the entire cross-section at the right-end undergoes a rigid rotation Θ.

3. The cross-section rotates rigidly with respect to the axis of the bar under
the action of a torque, so that the displacements in the x–y plane behave
like a rigid disk. The displacement in the z-direction, w, is not constrained.
4. The bar is loaded only with a torque with respect to its axis.

Based on these assumptions, the stresses on the cross-section (the z-plane)

become

σzr = 0, thin-wall, and circumferential surface is free of stresses

(9.37)
σzz = 0, the bar is not loaded in the axial direction

The only nonzero stress on the cross-section is σzs . For convenience, it is

denoted as τ . Its direction is in the tangent direction of the middle line of
the thin-wall, and its positive direction follows the standard right-hand rule
for rotation with respect to the z-axis. This creates a flow of shear stress
along the cross-section of the bar. Because the wall is thin, the shear stress τ
is uniform across the thickness (or τ is treated as the averaged shear stress)
at any fixed location s.

9.2.1 Concept of shear flow

The shear flow is quite an important concept when dealing with shear
stresses in thin-walled structural members, and will be used multiple times
in this book. It is usually denoted as q. To introduce this concept, let us
cut an infinitely small free-body diagram from the thin-walled bar at an
arbitrary location s, shown in Fig. 9.9(a). We can look at the equilibrium of
370 Mechanics of Materials: Formulations and Solutions with Python

all the forces in the z-direction:

Fz = 0 : τs+ds ts+ds dz − τs ts dz = 0 (9.38)

which is

τs+ds ts+ds = τs ts (9.39)

This means that the product τs ts does not change when s changes. It is
constant along s. This conserved quantity is called shear flow, defined as

q = τ s ts (9.40)

The shear stress τs and the wall thickness ts may change with s, but q is
constant for a given bar with thin-walled cross-section subjected to torque T!

9.2.2 Torque and shear-flow relation

Because the shear stress is generated by the torque, we have (see, Fig. 9.9):
 
1
T = τs ts rs ds = 2q rs ds = 2qAm (9.41)

q:constant 2  
 dA 
Am

where Am is the area enclosed by the middle line of the wall on the cross-
section, and r is “radius” at point s which changes with s (which is different
from the circular cross-section). It is the shortest distance from point O in
the r-axis direction. Clearly, this formula holds only if the cross-section is
closed, allowing the closed loop integration. We obtain now a very simple
but important equation, known as the Bredt-Batho formula:

T
T = 2qAm or q= (9.42)
2Am

This means that the torque is twice the shear flow multiplied by the mean
cross-section area. With Eq. (9.42), the shear stress, as a function of ts and
hence of s, can be easily obtained for a given torque:

T
τs = (9.43)
2Am ts

Note here that ts is in general a function of s. It is clear that the maxi-

mum shear stress is registered where ts is smallest.
 Torsion of Bars 371

9.2.3 Twist-angle and torque relation

The final piece of formulation is the relation between the twist-angle and the
torque. This clearly leads to the stiffness of the bar under torsion. Naturally,
it must involve the shear strain and stress relation.
Based on the strain–displacement relations similar to Eq. (9.5), and the
fact that z-axis and s-axis are always orthogonal, we have
γzr = 0 because σzr = 0
 
∂w ∂us (9.44)
γzs = 2εzs = +
∂s ∂z
Because the cross-section rotates rigidly, the displacement us relates the
twist angle Θ as (see Fig. 9.9)
us = rs Θ (9.45)
Note that rs is in general a function of s, but independent of z. Substitut-
ing Eq. (9.45) into Eq. (9.44), and using the constitutive equation for shear
strain and stress, we obtain:
   
τs q ∂w ∂[rs Θ] ∂w ∂Θ
γzs = = = + = + rs (9.46)
G Gts ∂s ∂z ∂s ∂z
which gives
 
q ∂w ∂Θ
= + rs (9.47)
Gts ∂s ∂z
Integration of the foregoing equation along s over the entire cross-section
of the bar gives
  
q ∂w ∂Θ
ds = ds + rs ds (9.48)
Gts ∂s ∂z
Now, because the cross-section rotates rigidly, Θ does not depend on s,
we have  
q s=l ∂Θ
ds = ws=0m + rs ds (9.49)
Gts    ∂z   
=0
2Am
where lm is the perimeter length of the middle line of the cross-section and
Am is the area enclosed by the middle line of the cross-section. Because
the cross-section is enclosed, and s comes back to its original location after
looping around, w|s=l
s=0 = 0. Note also that q is independent of s, Eq. (9.49)
m

becomes

∂Θ q 1
= ds (9.50)
∂z 2Am G ts
∂Θ
where ∂z is the twist angle per unit length ϑt .
372 Mechanics of Materials: Formulations and Solutions with Python

Using Eqs. (9.42) and (9.50) can also be written as


∂Θ T 1 T
= ds = (9.51)
∂z 4Am G
2 ts GJ
where J is the equivalent polar moment of area of the cross-section,
defined as
4A2
J =  1m (9.52)
ts ds

If the thickness ts is piecewise constant: t1 , t2 , . . . , tn with the correspond-

ing lengths: l1 , l2 , . . . , ln , we shall have
 n
1 li
ds = (9.53)
ts ti
i

If the bar has a uniform wall thickness t, the foregoing equation is sim-
plified to

4tA2m
J= (9.54)
lm
where lm is the perimeter of the middle line of the thin wall cross-section.
This is Eq. (9.29) we obtained earlier by simple induction or guessing.
If the cross-section is circular with radius R (that is constant), we have
lm = 2πRm , and Am = πRm 2 , we have Eq. (9.26). Readers may find the

formula for J for a rectangular thin-wall cross-section.

If Eq. (9.1) holds: Θ = ϑt z, Eq. (9.51) leads to Eq. (9.14) and also
Eq. (9.16).

9.2.4 Example: Shear stress in a thin wall bar with

box cross-section
Consider a long bar of length L with box cross-section with dimensions given
in Fig. 9.10. It is fixed at z = 0, and subjected to a torque of T at z = L.
The sides of the box are considered as thin walls made of elastic material
with shear modulus G.

1. Derive the formulas for computing its twist angle, shear flow, and shear
stress in the left-wall on the cross-section.
2. Given data: G = 25.0 GPa, L = 0.5 m, a = 58 mm, b = 38 mm, tT =
2 mm, tB = 4 mm, tL = 3 mm, tR = 4 mm, T = 5 kNm. Compute
the numerical values for the twist angle, shear flow, and the maximum
shear stress.
 Torsion of Bars 373

Figure 9.10. A long bar with a box cross-section of thin-walls subjected to a torque T
with respect to its axial axis.

Solution: We write the following code to complete the task.

1 # 1) Derive formulas using Sympy:

2
3 t, tT, tB, tL, tR = [Link]("t, t_T, t_B, t_L, t_R") # dimensions
4 a, b = [Link]("a, b ")
5 L, G, T, t = [Link]("L, G, T, _t")
6
7 # compute the area enclosed by the middle lines:
8 l = 2*a + 2*b # perimeter of the middle line of the thin wall
9 A = a*b # area
10 inverse_t_ds= a/tT + a/tB + b/tL + b/tR # ∫(1/t)ds
11 J = 4*A**2/inverse_t_ds # equivalent polar moment of area
12 [Link](J,"Formula for J:")

Formula for J:
4a2 b2
a
tT + taB + tbR + b
tL

1 t = (T/(G*J)).simplify() # twist angle per unit length

2 = t*L # total twist angle
3 [Link]( ,"Twist angle by T:")

 
a a b b
LT tT + tB + tR + tL
Twist angle by T:
4Ga2 b2
374 Mechanics of Materials: Formulations and Solutions with Python

1 dic = {tT:t, tB:t, tL:t, tR:t} # if all thickness are the same t
2 t = .subs(dic).factor(t)
3 [Link]( t,"Twist angle for uniform t:")

Twist angle for uniform t:

LT (a + b)
2Ga2 b2 t

1 Jt = [Link](dic).factor(t) # equivalent polar moment of area

2 [Link](Jt,"J for uniform t:")

J for uniform t:
2a2 b2 t
a+b

1 q = T/(2*A)
2 [Link](q,"Shear flow by T:")

Shear flow by T:
T
2ab

1 _L = q/tL
2 [Link]( _L,"Shear stress on the left wall:")

Shear stress on the left wall:

T
2abtL

1 # 2) numerical calculation:
2
3 G_ = 25.0e9 # Pa
4 L_ = 0.5 # m
5 a_ = 0.058; b_=0.038 # m
6
7 tT_= 0.002; tB_=0.004; tL_=0.003; tR_=0.004 # mm,
8
9 T_ = 5.0e3 # kNm.
10 data = {T:T_, G:G_, L:L_, a:a_, b:b_, tT:tT_, tB:tB_, tL:tL_, tR:tR_}
11
12 _data = .subs(data)
13 [Link]( _data, " value (rad) for the given data:")

value (rad) for the given data: 0.338

 Torsion of Bars 375

Figure 9.11. Comparison of major formulas for exact and thin-wall assumptions for
torsional bars.

1 [Link]([Link](data),"shear flow value (m⋅Pa) for the given data:")

shear flow value (m Pa) for the given data: 1.134 · 106

1 _max = [Link](data)/tT_
2 [Link]( _max,"Maximum shear stress (Pa) in the top wall (thinest):")

Maximum shear stress (Pa) in the top wall (thinest): 5.671 · 108

9.2.5 A comparison of exact and thin-wall formulas

Figure 9.11 gives a comparison of major formulas for exact and thin-wall
assumption for bars subjected to torques.

9.3 Bars with non-circular cross-sections*

9.3.1 Setting of the problem

Consider a torsion bar member with non-circular cross-section subjected to
a torque T . It is defined in coordinates (x, y, z) with z-axis being the axial
axis of the bar. The cross-section of the bar is assumed uniform along z-axis.
376 Mechanics of Materials: Formulations and Solutions with Python

Because of the non-circular shape, the cross-section may not remain plane
and can warp, implying that the displacement w may vary with x and y.
The rotation of the bar changes along with z-axis. Therefore, the problem
is three dimensional. The method used to perform analysis is the so-called
Saint-Venant’s semi-inverse method [4]. It approximates the displacements
in such a way that it leads to a stress state (σzx , σzy ) on the cross-section,
under torque T , aiming to convert the problem to a two-dimensional one.
Our formulation and notation follow largely the textbook [4]. Codes are
developed for some of the derivations.

9.3.2 Assumed displacements

Consider a point B at (x, y, z) on a cross-section in an undeformed bar. When
it is deformed by subjecting T , the cross-section rotates by Θ about z-axis
with respect to the origin, and B is displaced by displacements (u, v, w) and
moves to B*, as shown in Fig. 9.12.
The semi-inverse method assumes:

1. The rotation is the principal cause of displacements (u, v), which decou-
ples u and v from w.
2. The twist angle changes linearly with z: Θ = ϑt z, where ϑt is the angle of
twist per unit length and is a constant (does not change with (x, y, z)).

Based on these assumptions, the displacements can be written as

u = −yϑt z; v = xϑt z; w = ϑt ψ(x, y) (9.55)

Figure 9.12. A bar with non-circular cross-section subjected to a torque. Warping may
occur on the cross-section.
 Torsion of Bars 377

where ψ is the warping function representing the off-cross-section deforma-

tion on the bar cross-section. Note that in the circular cross-sectional bars,
ψ is assumed to be zero. In non-circular bars, it may not be zero and we
may need to find the warping function.
We have two unknowns ϑt and ψ that depend on (x, y). We have converted
the 3D problem to 2D.

9.3.3 Shear strains in non-circular bars

Using the small strain–displacement relations given in Chapter 4, and ϑt is
a constant, we can find the strains in the bar:
∂u ∂v ∂w
εxx = = 0; εyy = = 0; εzz = =0
∂x ∂y ∂z
 
1 ∂u ∂v
εxy = + =0
2 ∂y ∂x
  (9.56)
∂u ∂w ∂ψ
2εzx = γzx = + = ϑt −y
∂z ∂x ∂x
 
∂w ∂v ∂ψ
2εzy = γzy = + = ϑt +x
∂y ∂z ∂y
The codes to obtain these equations are as follows:

1 x, y, z = symbols("x, y, z") # define symbolic coordinates

2 = [Link](" ")(x, y)
3 X = Matrix([x, y, z])
4 t = symbols(" _t") # twist angle per length
5
6 U = Matrix([-y* t*z,x* t*z, * t]) # Assumed displacement vector[u,v,w]
7 U

⎡ ⎤
−yzϑt
⎢ ⎥
⎣ xzϑt ⎦
ϑt ψ(x, y)

1 # Compute the strains (see Chapter 4):

2 U_g = gr.grad_vf(U, X) # displacement gradient
3 ε = (U_g + U_g.T)/2 # strain tensor,symmetric
4 2*ε # change to

⎡ ∂ ⎤
0 0 −yϑt + ϑt ∂x ψ(x, y)
⎢ ∂ ⎥
⎣ 0 0 xϑt + ϑt ∂y ψ(x, y) ⎦
∂ ∂
−yϑt + ϑt ∂x ψ(x, y) xϑt + ϑt ∂y ψ(x, y) 0
378 Mechanics of Materials: Formulations and Solutions with Python

It is clear that the only nonzero strains are:

 
∂ψ
γzx = ϑt −y
∂x
  (9.57)
∂ψ
γzy = ϑt +x
∂y

It is easy to confirm that the compatibility conditions given in Chapter 4

are satisfied. Note that if ψ is zero, Eq. (9.57) becomes Eq. (9.5).
Differentiate γzx with respect to y, and γzy with respect to x, and then
subtracting the results of these two, the warping function ψ is eliminated,
leading to

∂γzx ∂γzy
− = 2ϑt (9.58)
∂y ∂x

This means that if γzx and γzy satisfy the foregoing equation, any function
ψ is admissible. Thus, we need not worry about ψ (if we do not want to find it
out). Instead, the strains resulting from the stresses to-be-found must satisfy
Eq. (9.58).

9.3.4 Equilibrium equations at any point in torsion bar

and stress function
Because only γzx and γzy are nonzero, the nonzero stresses are σzx and σzy
(for isotropic materials). The equilibrium equations for 3D solids given in
Chapter 6 are reduced to (without body force):
∂σzx ∂σzy
= 0; =0
∂z ∂z
(9.59)
∂σzy ∂σzx
+ =0
∂y ∂x
The first two equations mean that σzx and σzy are independent of z, and
they shall satisfy the last equation in Eq. (9.59). Such stresses can be written
in terms of the so-called Prandtl stress function φ(x, y) as
∂φ(x, y) ∂φ(x, y)
σzx = ; σzy = − (9.60)
∂y ∂x
Function φ is in fact a stream function of the field of shear stress vector
[σzx σzy ] on a surface-z. It is easy to confirm that the last equation in
Eq. (9.59) is satisfied, by substituting Eq. (9.60) into it.
 Torsion of Bars 379

The problem becomes finding the stress function φ(x, y), subjected to
proper boundary conditions and overall equilibrium on the cross-section of
the bar.

9.3.5 Boundary conditions for the stress function

Because the lateral surface s of a torsion bar is free of stresses, the component
of the shear stress τ at any point on the cross-section by the surface S in its
surface normal direction n must be zero.
Consider an infinitely small curve segment ds for a point P on the bound-
ary of the cross-section shown Fig. 9.13. The corresponding infinitely small
line segments along the Cartesian coordinates are denoted as dx and dy,
which cut off ds on the boundary. Based on the geometry of the infinitely
small triangular cell on the boundary, we shall have on boundary S:
dx dy
sin θ = − ; cos θ = (9.61)
ds ds
The minus sign is due to the increase in ds (counter-clockwise), resulting
in a decrease in dx.
In order to obtain the shear stresses on the cross-section plane z by
boundary surface S, we can simply perform coordinate transformation.
Because these shear stresses are all on the same cross-sectional surface-z,

Figure 9.13. Stress components on the cross-section by the lateral surface S of a bar.
They can be expressed in both the Cartesian and polar coordinates, and the coordinate
transformation rule for vectors apply.
380 Mechanics of Materials: Formulations and Solutions with Python

and the coordinate transformation is with respect to z-axis. This means

that (σzx , σzy ) and (σzr , σzθ ) carry the same first index z. The coordinate
transformation rules for vectors (only one index is transferred) apply, which
gives
      
σzr cos θ sin θ σzx σzx cos θ + σzy sin θ
= = (9.62)
σzθ −sin θ cos θ σzy −σzx sin θ + σzy cos θ
  
T

Here, we used transformation matrix T for 2D, its formation is discussed

in Section 2.10. Note that this transformation angle is between the x-axis
and the normal vector n at a point on the boundary surface S.
We can also just follow the standard tensor transformation rule for stress
tensor, given in Eq. (9.8). The code is as simple as follows:

1 # Define the stress tensor on the cross-section of the bar:

2
3 σzx, σzy, = symbols("σ_zx, σ_zy, ") # Define symbolic variables
4
5 S = Matrix([[ 0, 0, σzx], # stress state for torsion w.r.t z-axis
6 [ 0, 0, σzy],
7 [σzx, σzy, 0]])
8
9 # Create sympy transformation matrix for rotation , w.r.t. z:
10 T = [Link]( )
11
12 # Perform the second order tensor transformation for the stresses:
13 Sp=T@[email protected]
14 gr.simplifyM0(Sp) # simply the terms in the output

⎡ ⎤
0 0 σzx cos (θ) + σzy sin (θ)
⎢ ⎥
⎣ 0 0 −σzx sin (θ) + σzy cos (θ)⎦
σzx cos (θ) + σzy sin (θ) −σzx sin (θ) + σzy cos (θ) 0
Compared the result with that in Eq. (9.62), these two shear stresses
(σzr , σzθ ) in the (r, θ) coordinates are exactly the same.
The first equation in Eq. (9.62) gives

σzr = σzx cos θ + σzy sin θ (9.63)

Because there is no shear stresses on the lateral surface s of the bar, we

must have σzr = 0. Substituting Eq. (9.61) into Eq. (9.63) leads to
dy dx
σzx − σzy =0 (9.64)
ds ds
 Torsion of Bars 381

Next, substituting Eq. (9.60) into Eq. (9.64), we obtain:

∂φ dy ∂φ dx dφ
+ = =0 (9.65)
∂y ds ∂x ds ds
Note the change from partial derivatives of φ to total derivative. This
means that φ must be a constant on the boundary of the cross-section
of the bar. Further because the stresses are given by partial derivatives of
φ, they ϕ can be set to zero on the boundary without affecting the stress
solution. Therefore, the boundary condition for φ is found as

φ=0 at any point on the boundary of the cross-section of a solid bar

(9.66)

9.3.6 Shear stress on the cross-section by the boundary

By the boundary because σzr = 0, the only nonzero stress components in
the polar coordinates is σzθ , and is often denoted as τ . It is given by the 2nd
equation in Eq. (9.62):
τ = σzθ = −σzx sin θ + σzy cos θ (9.67)
In other words, σzx and σzy are the two components of τ , and we have
σzx = −τ sin θ
(9.68)
σzy = τ cos θ
When the foregoing equation is substituted to Eq. (9.67), it justifies
our argument. Readers may observe the foregoing equation together with
Fig. 9.13.
Next, substituting Eq. (9.60) into Eq. (9.67), we obtain
∂φ ∂φ ∂φ
τ =− sin θ − cos θ = − (9.69)
∂y ∂x ∂r
In obtaining Eq. (9.69), we used coordinate transformation again, but for
the gradient vector of φ.
Equation (9.69) means that τ is the negative slope of the stress function
φ with respect to r (in the normal direction of bar’s lateral surface S with
normal n). This finding is important when using the so-called membrane
analogy for analyzing bars with complicated cross-section, which will be
discussed later. Most importantly, it tells us that
the maximum absolute shear stress along the lateral surface is at location
where normal slope of φ is the largest.
382 Mechanics of Materials: Formulations and Solutions with Python

9.3.7 Equilibrium conditions for the stress function

We found the expressions for the stresses on the cross-section (in terms of the
stress function). The amplitude of the stresses must depend on the level of
the torque applied. Therefore, we must examine the equilibrium conditions
for the stresses obtained using the stress function using Eq. (9.60), in relation
to the torque on the cross-section. These conditions are

∂φ
Fx = σzx dxdy = dxdy = 0
A A ∂y
∂φ
Fy = σzy dxdy = − dxdy = 0 (9.70)
A A ∂x
 
∂φ ∂φ
Mz = (xσzy − yσzx )dxdy = − x +y dxdy = T
A A ∂x ∂y

where T is the torque on the cross-section. We need to find the conditions on

stress function φ so that the foregoing equilibrium conditions are satisfied.
First, we evaluate the 2nd equation in Eq. (9.70). Consider the strip dy
across the cross-section along x-direction, as shown in Fig. 9.14.
Because φ does not vary in the y-direction for this strip, we have

φ(xB )
∂φ dφ
− dxdy = −dy dx = −dy dφ
A ∂x A dx φ(xA )

= −dy[φ(xB ) − φ(xA )] = 0 (9.71)

Figure 9.14. Integration along x-direction over dy strip.

 Torsion of Bars 383

This means that the sum of all the forces in the y direction vanishes.
The same can be done for the 1st equation in Eq. (9.70), by considering
a dx strip. Readers may get this done on a piece of paper. The first two
equations in Eq. (9.70) are both satisfied. This makes sense because all the
shear stresses on the cross-section can only be circling around, and the total
divergence must be zero [9].
Consider now the 3rd equation in Eq. (9.70).
 
∂φ ∂φ
Mz = − x +y dxdy
A ∂x ∂y
 ∂φ 
∂x
=− x y ∂φ
dxdy = − x y · ∇φdxdy
A ∂y A
 
x
=− φ x y · n dxdy + φ∇· dxdy
s  A y
=0, on s

=2 φdxdy = T (9.72)
A

In the 2nd to last line we used the Gauss’s formula [9]. We finally found
that the integral of the stress function over the area of the cross-section
should equal to half of the torque on the cross-section. In other words, the
volume under φ(x, y) is half of the torque T:

2 φdxdy = T (9.73)
A

This is an important condition for φ because it relates directly to the

torque applied.

9.3.8 Linear elastic solution

Assume the bar is made of a linear elastic isotropic material. The stresses
relate to the strains by Hooke’s law:

∂φ
σzx = = Gγzx
∂y
(9.74)
∂φ
σzy =− = Gγzy
∂x
384 Mechanics of Materials: Formulations and Solutions with Python

or
1 ∂φ
γzx =
G ∂y
(9.75)
1 ∂φ
γzy =−
G ∂x
Substituting Eq. (9.75) to Eq. (9.58), we obtain:

∂2φ ∂2φ
+ 2 = −2Gϑt (9.76)
∂x2 ∂y

This is a Poisson’s equation or a nonhomogeneous Laplace equation. If

the twist angle ϑt is given and φ satisfies BC Eq. (9.66), then Eq. (9.76)
uniquely determines φ(x, y). The stresses can be found via Eq. (9.74), and
the torque by Eq. (9.73). We thus have a logical and workable path to the
solution.

9.3.9 Techniques for constructing the stress function

Because φ must vanish on the boundary of the cross-section, we can make
use of the equation that describes the boundary of the cross-section to con-
struct φ. Assuming the boundary of the cross-section is given by f (x, y) = 0,
φ can be written as
φ(x, y) = Cf (x, y) (9.77)
where C is a constant. If f (x, y) satisfies
∂2f ∂2f
+ = constant (9.78)
∂x2 ∂y 2
we can then adjust C to have Eq. (9.76) be satisfied.

9.3.10 Example: Bars with elliptical cross-section

Consider a bar with elliptical cross-section, as shown in Fig. 9.15.
Stress function φ can be written as
 2 
x y2
φ=C 2 + 2 −1 (9.79)
a b
Substituting it to Eq. (9.76), we found
a2 b2 Gϑt
C=− (9.80)
a2 + b2
 Torsion of Bars 385

Figure 9.15. Elliptical cross-section for a torsional bar.

The shear stresses can be found via Eq. (9.74):

∂φ 2Cy 2a2 Gϑt y

σzx = = 2 =− 2
∂y b a + b2
(9.81)
∂φ 2Cx 2b2 Gϑt x
σzy =− =− 2 = 2
∂x a a + b2
The maximum shear stress is found as

 2a2 bGϑt
τmax = σzy  = 2 (9.82)
y=b a + b2

Torque T is obtained by substituting Eq. (9.79) into Eq. (9.73):

2C 2C
T =2 φdA = x2 dA + y 2 dA − 2C dA
A a2 A b2 A A
2C 2C
= Iyy + 2 Ixx − 2CA (9.83)
a2 b
where A is the area, Ixx , Iyy are the 2nd moments of area with respect to
x- and y-axis of the cross-section of the bar. For an ellipse, the formulas are
as follows:

πab3 πba3
Ixx = ; Iyy = ; A = πab (9.84)
4 4
Using these formulas given above, we have

T = −πCab (9.85)
386 Mechanics of Materials: Formulations and Solutions with Python

The torque may be expressed either in terms of τmax or ϑt by means of

Eqs. (9.80), (9.82) and (9.85).

2T
τmax = (9.86)
πab2
T (a2 + b2 ) T
ϑt = = (9.87)
πGb3 a3 GJ
where GJ is the torsional stiffness per unit length, with J given by

πb3 a3
J= (9.88)
a2 + b2

This is the equivalent polar moment of area of ellipses. When a = b,

Eq. (9.88) is reduced to Eq. (9.13) for the circular cross-section.

9.3.11 Comparison of torsional stiffness between

elliptic and circular bars
Let us conduct an analysis on torsional stiffness of elliptic and circular solid
bars made of the same material. Assuming the radius of the circular cross-
section is R, we shall set R2 = ab to ensure their areas are the same. The
stiffness ratio of these two bars becomes.
 πb3 a3  b3 a3
Jellipse  
a2 +b2  a2 +b2 2ab
αstiffness = = = = ≤1 (9.89)
Jcircle R2 =ab πR  2
4
R =ab
(ab)2 a2+ b2
2 2

In the last step, we used inequality a2 + b2 ≥ 2ab (because (a − b)2 ≥ 0).

This means that the stiffness of an elliptic bar is always smaller that the
circular bar with the same area. They are equal only when a = b.

9.3.12 Example: Bars with equilateral triangle

cross-section
Consider a bar with equilateral triangle cross-section, as shown in Fig. 9.16.
It is subjected to torque T .
Stress function φ can be written as
   
Gϑt √ 2h √ 2h h
φ= x − 3y − x − 3y + x+ (9.90)
2h 3 3 3
 Torsion of Bars 387

Figure 9.16. Equilateral triangle cross-section for a torsional bar.

Following the same process for the ellipse bar, the maximum shear stress
is found at the center of the edges as
√
15 3T 20T
τmax = = 3 (9.91)
2h 3 a
The twist angle per unit length is found as
√
15 3T T
ϑt = = (9.92)
Gh 4 GJ
where GJ is the torsional stiffness per unit length for the equilateral triangle
cross-section, with J being the equivalent polar 2nd moment of area given
by
√ 4
h4 3a
J= √ = (9.93)
15 3 80
For cross-sections of more complex shape, analytical solutions are difficult
to obtain. Approximate solutions may be obtained by Prandtl’s membrane
analogy for some of those.

9.4 Prandtl’s elastic membrane analogy for torsion*

9.4.1 General formulation

Considering a membrane subjected to pressure p, resulting in a uniform
membrane tension S, and undergoes a small deflection, as shown in Fig. 9.17.
It is fixed on its boundary s. Our formulation and notation follow largely
the textbook [4]. Codes are developed for computations and result plots.
388 Mechanics of Materials: Formulations and Solutions with Python

Figure 9.17. A membrane subjected to pressure and undergoes a small deflection.

An infinitely small cell at a point is taken to examine the equilibrium state. Top: top-
down view. Bottom: side section view.

An infinitely small cell of dxdy at a point is taken to examine the equi-

librium state. Assume the membrane undergoes a tension stress S and its
deflection is denoted as w.
On edge A-D, the force in the membrane is
∂w
−Sdy sin α ≈ −Sdy tan α = −Sdy (9.94)
∂x
On edge B-C, the force is
   
∂α ∂ ∂w
Sdy tan α + dx = Sdy w+ dx (9.95)
∂x ∂x ∂x
Similarly, on edge A-B, the force in the membrane is
∂w
−Sdx (9.96)
∂y
On edge D-C, it is
 
∂ ∂w
Sdx w+ dy (9.97)
∂y ∂y
 Torsion of Bars 389

The sum of all these four forces with the pressure should be zero, and
dividing dxdy for all terms, we obtain the equilibrium equation in terms of
w:
∂2w ∂2w p
+ =− (9.98)
∂x 2 ∂y 2 S
This equation has the same form as Eq. (9.76) for the stress function.
They both are Laplace equation, but with different constants on the right-
hand side. Therefore, the membrane deflection and the stress function differ
by a constant factor c:
w = cφ (9.99)
where c is found by equaling terms on the right-hand side of Eqs. (9.98) and
(9.76), which is
p
c= (9.100)
2Gϑt S
Finding the stress function for torsion bar becomes solving for the deflec-
tion of a pressured membrane with a shape same as the cross-section of the
bar.
Examples are given in the following sections.

9.4.2 A bar with narrow rectangular cross-section*

Consider a bar with a narrow rectangular cross-section subjected to a
torque T, as shown in Fig. 9.18. The torque will create a force couple on
the cross-section that results in shear stress distributed there. Because of its
narrow shape, it is expected that the shear stress will form two major flows:
one near the top surface and another near the bottom surface, with mutually
opposite direction. We shall find a way to quantify it.
Using the membrane analogy, we consider a membrane with a narrow
rectangular shape and subjected to pressure, as shown in Fig. 9.19.
Based on the slender geometry of the deflected membrane, we can assume
membrane deflection is independent of x, and varies in a parabolic way with

Figure 9.18. A bar with narrow rectangular cross-section subjected to a torque.

390 Mechanics of Materials: Formulations and Solutions with Python

Figure 9.19. Narrow rectangular cross-section of a bar subjected to a pressure and

undergoes a small deflection. It has a width of 2b and height of 2h.

respect to y.
 
y2
w = w0 1− 2 (9.101)
h
where w0 is the maximum deflection at y = 0. This deflection satisfies the
boundary condition on y = ±h. Except for the small regions near x = ±b.
Hence, this is a good approximation. The approximation error reduces with
decreasing h/b. The Laplace of the assumed w becomes

∂2w w0
=− 2 (9.102)
∂y 2 h

It is now a 1D problem. The membrane analogy leads to w0 /h2 = 2cGϑt .

The stress function is found as
 
y2
φ = Gϑt h2 1 − 2 (9.103)
h

The shear stresses can be found via Eq. (9.74):

∂φ
σzx = = −2Gϑt y
∂y
(9.104)
∂φ
σzy =− =0
∂x
It is clear that σzx is linearly in y and anti-symmetric with respect to
y = 0.
An actual distribution of the shear stress on a narrow rectangular cross-
section is give in Fig. 9.20. This is produced using our code to be introduced
later.
 Torsion of Bars 391

Figure 9.20. An actual case of shear stress on narrow rectangular cross-section of a bar
subjected to torsion (b = 10h).

As shown, the shear stress flows near the top and bottom surfaces form
a force couple giving the torque on the cross-section of the bar, as shown in
Fig. 9.18. The resulting two forces are located about h/3 away from the top
and bottom surfaces.
The maximum shear stress is found at y = ±h:

τmax = 2Gϑt h (9.105)

Torque T is obtained using Eqs. (9.73) and (9.103):

b h
T =2 φdA = 2 φdxdy = Gϑt 31 (2b)(2h)3 = GJϑt (9.106)
A −b −h   
J

where J is the equivalent polar 2nd moment of area given by

1
J= (2b)(2h)3 (9.107)
3

Note that Eq. (9.107) can be used for any cross-section that has a narrow
channel shape, and it can be curved. This is because of the membrane
analogy, the deformed membrane will have the same volume regardless of
whether it is curved or not, as long as it is a narrow channel. Therefore,
Eq. (9.107) is very useful to calculate the stiffness per length GJ for such
bars.
Using Eqs. (9.105) and (9.106), the maximum shear stress can now be
written in terms of T :
3T 2T h
τmax = = (9.108)
(2b)(2h)2 J
Finally, the twist angle per length can be written for given T :
3T T
ϑt = = (9.109)
G(2b)(2h)3 GJ
We mention that the boundary conditions are not satisfied on x = ±b.
The error is expected to be small if h/b is small.
392 Mechanics of Materials: Formulations and Solutions with Python

9.4.3 Bars with cross-sections consisting of narrow

rectangles*
Many torsion bars are with composite sections consisting of a number of
narrow rectangles. The most typical one is the I-beams, as shown in Fig. 9.21.
This type of bars is analyzed easily by making use of the results from the
previous section.
Assuming the cross-section of a bar is formed with n narrow channels, its
J value can be obtained simply by
n
1
J =c (2bi )(2hi )3 (9.110)
3
i=1

This summation is justified by, again, the membrane analogy because

the total volume of the deflection (hence the stress function) for a composite
section with narrow channels are just simple addition of the individual chan-
nels. Each channel acts alone having a stress distribution like that shown in
Fig. 9.20, forming an individual torque. The sum of the torques of all the
channels of the cross-section gives the total torque T .
In Eq. (9.110), we introduced a correction coefficient c. It can be used
for proper adjustment when a rectangle is not narrow. If bi > 10hi for a
rectangular part, then c ≈ 1. If bi < 10hi , c may be reduced to, say, c ≈ 0.9.

Figure 9.21. A typical I-beam with three narrow rectangles. The image on the left is
from Poutrelle métallique at wikipedia commons under the CC BY 3.0 license.
 Torsion of Bars 393

It is a rough estimation. As will be seen in the next section with computer

codes, we can use more precise formula.
Note also that when computing the maximum shear stress in a composite
section, we shall use
2T hmax
τmax = (9.111)
J
where hmax is maxi (hi ). This is because a wider channel has larger membrane
deflection and hence larger φ values.

9.5 Bars with rectangular cross-section*

9.5.1 Strategy for solution

Consider torsion bars with the rectangular section of width 2h and depth 2b,
as shown in Fig. 9.22. This types of bars require more intensive analysis for
proper design against torsion. Our formulation and notation follow largely
the textbook [4]. Codes are developed for computations and result plots.
We need to find the stress function φ, and then solve the torsion problem
for stresses and twist angle. Because of the symmetry of the cross-section
and the Laplace operator is also symmetric, φ must be an even function of
both x and y. The problem can be mathematically defined by
∇2 φ = −2Gϑt PDE in the domain of the cross-section
(9.112)
φ=0 on the boundary of the cross-section
One effective approach to solve this problem is to use the method of sep-
aration of variables. To do so, we first make the Laplace PDE homogeneous

Figure 9.22. A typical rectangular cross-section for bars subjected to torsion.

394 Mechanics of Materials: Formulations and Solutions with Python

(with zero right-hand side). This is done by writing φ in the following form,
assuming h ≤ b:
φ = Gϑt (h2 − x2 ) + w(x, y) (9.113)
where w is an unknown function to be determined. It is easy to check that
the first term in the foregoing equation satisfies the PDE in Eq. (9.112). It is
thus a particular solution to Eq. (9.112). Now, substituting Eq. (9.113) into
Eq. (9.112) gives
∇2 w = 0 governing PDE

0 for x = ±h (9.114)
w=
Gϑt (h2 − x2 ) for y = ±b
We need next to solve the foregoing equation for w(x, y), using the method
of separation of variables. Note if assuming h > b, we would use Gϑt (b2 − y 2 )
instead as the particular solution to Eq. (9.112). This is just for a better
convergence in the solution to be obtained later.

9.5.2 Method of separation of variables

We write the unknown function as the product of two unknown functions
f (x) and g(x):
w(x, y) = f (x)g(y) (9.115)
Substitute Eq. (9.115) into Eq. (9.114), we obtain
∇2 w = g(y)f  (x) + f (x)g (y) = 0 (9.116)
which can be rewritten as
f  (x) g (y)
=− = −λ2 (9.117)
f (x) g(y)
The first term is a function of only x and the second is a function of
only y. Thus, λ can only be a (unknown) constant independent of both x
and y. Equation (9.117) gives two separate equations:
f  (x) + λ2 f (x) = 0
(9.118)
g  (y) − λ2 g(y) = 0
Each of these equations is a 1D DE. The solutions to them are,
respectively,
f (x) = A cos λx + B sin λx
(9.119)
g(x) = C cosh λy + D sinh λy
 Torsion of Bars 395

The solutions contain four constants. Because the original stress function
φ must be an even function, both f (x) and g(x) must be even. Thus, B =
D = 0. Substitute Eq. (9.119) into Eq. (9.115), we obtain
w(x, y) = A cos λx cosh λy (9.120)
where A (absorbed C) is a to-be-determined constant, in addition to
constant λ. This is done using boundary conditions.
To satisfy the boundary condition at x = ±h, we set
nπ
λ= , n = 1, 3, 5, . . . (9.121)
2h
Equation (9.120) becomes
nπx nπy
w(x, y) = An cos cosh (9.122)
2h 2h
n=1,3,5,...

Using the boundary condition at y = ±b, we have

nπx nπb
An cos cosh = Gϑt (h2 − x2 ) (9.123)
n=1,3,5,...
2h 2h

To solve the foregoing equation for An , we multiply cos nπx

2h on both sides
and integrate along x from −h to h, and use formulas for cosine function,
which gives
h
nπb Gϑt nπx −32Gϑt h2 (−1)(n−1)/2)
An cosh = dx =
(x2 − h2 ) cos
2h h
−h 2h n3 π 3
(9.124)
These above-integrals can be done using the following code.

1 x, h = symbols("x, h")
2 n = symbols('n', integer=True, odd=True)
3
4 # The integral = on the left-hand-side
5 integrate((([Link](n*[Link]*x/(2*h))**2)), (x, -h, h))

 πn
h for h = 0
2h otherwise

1 # The integral = on the right-hand-side

2 integrate((x**2-h**2)*[Link](n*[Link]*x/(2*h)), (x, -h,h))

n 1
32 (−1) 2 − 2 h3
−
π 3 n3
396 Mechanics of Materials: Formulations and Solutions with Python

Finally, An is found as

−32Gϑt h2 (−1)(n−1)/2)
An = , n = odd (9.125)
n3 π 3 cosh nπb
2h

9.5.3 Final solution of the stress function

The stress function becomes

32Gϑt h2 (−1)(n−1)/2 nπx nπy

φ = Gϑt (h2 − x2 ) − nπb
cos cosh
π3 n cosh 2h
3 2h 2h
n=1,3,5,...
(9.126)

Note that in the denominator of the coefficient of the series above, there
is n3 term. Also, the cosh function has the following Taylor expansion:

x2 x4
cosh x = 1 + + + ··· (9.127)
2! 4!
Therefore, the coefficients in Eq. (9.126) goes to zero quite fast when b/h
is large. When b/h goes to infinity, Eq. (9.126) reduces to the one for narrow
rectangles discussed in the previous section.

9.5.4 Shear stress solution

Using Eq. (9.60), the shear stresses can be found as follows:

∂φ 16Gϑt h (−1)(n−1)/2 nπx nπy

σzx = =− nπb
cos sinh (9.128)
∂y π 2
n=1,3,5,...
n2 cosh 2h 2h 2h

∂φ 16Gϑt h (−1)(n−1)/2 nπx nπy

σzy = − = 2Gϑt x − nπb
sin cosh
∂x π2 n2 cosh 2h 2h 2h
n=1,3,5,...
(9.129)

9.5.5 Torque twist-angle solution and equivalent polar

2nd moment of area
The torque is found using Eq. (9.73) as
b h
T =2 φdxdy = GJϑt (9.130)
−b −h
 Torsion of Bars 397

where
⎡ ⎤
 
1 192 h 1 nπ b ⎦
J = (2b)(2h)3 ⎣ − 5 tanh = (2b)(2h)3 k1
3 3π b n5 2 h
n=1,3,5,...
  
k1
(9.131)
where k1 is a coefficient that depends only on b/h. When b/h → ∞, k1 → 13
because the summation term in the foregoing equation approaches to zero.

9.5.6 The maximum shear stress

The maximum slope of the stress function (membrane) on the rectangle
section occurs at x = ±h, y = 0, where σzx = 0 and σzy has the maximum
value (h ≤ b):
⎡ ⎤

 8 1
τmax = σzy  = 2Gϑt h ⎣1 − 2 2 cosh nπb
⎦ = 2Gϑt hk
x=h π n
y=0 n=1,3,5,... 2h
  
k
(9.132)
k1
If we introduce k2 = k , the maximum shear stress can be computed
alternatively using
T
τmax = (9.133)
k2 (2b)(2h)2 )
Coefficient k1 and k can be computed using the code given below. Tabu-
lated k1 and k2 are given in Ref. [4].

9.5.7 Python code for computing k1

1 # Define the function for computing k1:

2 def k1_f(b_h, n_end):
3 sum1 = 0.0
4 for n in range(1, n_end, 2):
5 sum1 += [Link](n*[Link]*b_h/2)/n**5
6 return 1.0/3.0 - 192.0/(3*b_h*[Link]**5)*sum1
7
8 # Define the function for computing k:
9 def k_f(b_h, n_end):
10 sum1 = 0.0
11 for n in range(1, n_end, 2):
12 sum1 += 1./([Link](n*[Link]*b_h/2))/n**2
13 return 1.0 - 8.0/([Link]**2)*sum1
398 Mechanics of Materials: Formulations and Solutions with Python

[Link] An example
First, we compute coefficients k1 , k, and k2 for two special values of b/h = 1
and b/h = 10. These values may be referenced in later analysis.

1 np.set_printoptions(formatter={'float': '{: 0.5e}'.format})

2 n_end = 21
3 b_h = [1, 10] # at special values, b=h, b=10h
4
5 k1 = [Link]([k1_f(bhi, n_end) for bhi in b_h])
6 print(f"Coefficient k1 for J with b/h of {b_h} = {k1}")
7 k = [Link]([ k_f(bhi, n_end) for bhi in b_h])
8 print(f"Coefficient k for _max with b/h of {b_h} = {k}")
9 k2 = k1/k
10 print(f"Coefficient k2 for _max with b/h of {b_h} = {k2}")

Coefficient k1 for J with b/h of [1, 10] = [ 1.40577e-01 3.12325e-01]

Coefficient k for _max with b/h of [1, 10] = [ 6.75314e-01 1.00000e+00]
Coefficient k2 for _max with b/h of [1, 10] = [ 2.08165e-01 3.12325e-01]

[Link] Cut-off terms in the solution series

We write the following code to study the required terms needed for solution
of k1 with desired accuracy. Because we assume h ≤ b, we study the worst
(slowest convergence) case of b = h.
1 np.set_printoptions(precision=7,suppress=True) # Digits in print-outs
2
3 b_h, n_end_max = 1.0, 10
4 x_data = list(range(1, n_end_max, 1))
5 y_data = []
6 for ni in x_data:
7 y_data.append(k1_f(b_h, ni)) # function values
8
9 print(f"Cut-off term n_max: {[Link](x_data)}")
10 print(f"The corresponding k1 values:\n {[Link](y_data)}")

Cut-off term n_max: [1 2 3 4 5 6 7 8 9]

The corresponding k1 values:
[0.3333333 0.141523 0.141523 0.1406625 0.1406625 0.1405955 0.1405955
0.1405831 0.1405831]

As shown when n = 9 is used, the error is at the 6th digit. Let us use the
following code to plot the convergence curve.
1 X_data = [x_data]
2 Y_data = [y_data] # put the function values to a list
3 labels = ['k1'] # define the label
4 gr.plot_fs(X_data, Y_data, labels, 'n', 'k1', 'k1_n')
5 # x-label, y-label, file-name
 Torsion of Bars 399

We observed a fast convergence from the plot above.

Note that these formulas work also for h > b, but the convergence is a
little slower. Since we are using computer code to do it, it is really not a
problem. Even if we take 101 terms, the computation takes less than one
second.
In conclusion, with the formulas derived in this section, one can solve all
torsion bars with a rectangular cross-section.

[Link] Values as function of b/h

We write the following code to compute the values of k1 , k and k2 for given
a list of b/h.
1 n_end = 25
2 dx = 0.01
3 b_h = [Link](0.1, 10.0, dx) # b/h as x_data
4
5 k1_data = []
6 k_data = []
7 for i in range(len(b_h)):
8 k1_data.append(k1_f(b_h[i], n_end)) # function values
9 k_data.append( k_f(b_h[i], n_end))
10
11 k2_data = [k1_data[i]/k_data[i] for i in range(len(b_h))]
12 step = max(int(len(b_h)/5), 1)
13 print(f"Ratio of b/h: {[Link](b_h)[::step]}")
14 print(f"The corresponding k1 values:\n {[Link](k1_data)[::step]}")
15 print(f"The corresponding k values:\n {[Link](k_data)[::step]}")
16 print(f"The corresponding k2 values:\n {[Link](k2_data)[::step]}")
400 Mechanics of Materials: Formulations and Solutions with Python

Ratio of b/h: [0.1 2.08 4.06 6.04 8.02]

The corresponding k1 values:
[0.003124 0.2326236 0.2815891 0.2985514 0.3071385]
The corresponding k values:
[0.074386 0.9382967 0.9972449 0.9998771 0.9999945]
The corresponding k2 values:
[0.0419976 0.2479212 0.282367 0.2985881 0.3071401]

1 k_3 = [k_data[i]/3. for i in range(len(b_h))]

2
3 X_data = [b_h, b_h, b_h] # put the x (b/h) values to a list
4 Y_data = [k1_data, k_3, k2_data] # put the function values to a list
5
6 labels=['k1', 'k/3', 'k2'] # define the label
7 gr.plot_fs(X_data, Y_data, labels, 'b/h', 'k1, k/3, k2', 'ks_bh')
8 # x-label, y-label, file-name

9.5.8 Comparison of equivalent J and polar moment

of area Jz
Consider the bar has a square cross-section. Equation (9.131) becomes

J = (2h)4 k1(b=h) = 0.140577(2h)4 (9.134)

 Torsion of Bars 401

Here, we used k1(b=h) = 0.140577 obtained earlier. Compared to the

polar moment of area of a square bar, using perpendicular axis theorem
4
(see Chapter 10), Jz = 2Ixx = (2h)
6 , the corresponding k1 should be calcu-
lated using

1
k1square = = 0.166667 (9.135)
6

This shows that the equivalent J value of the same square bar subjected
to torsion is about 15.65% lower than Jz value of a square bar of (2h) × (2h).
(2b)(2h)3
For narrow channel of b/h → ∞, Jz = (Ixx + Iyy ) → 12 , the corre-
sponding k1 should be calculated using

1
k1narrow = (9.136)
12

For the same infinitely long channel bar subjected to torsion, we have
k1 = 1/3, which is 4 times k1narrow .

9.5.9 Comparison of torsional stiffness between

rectangular and square bars
Let us conduct an analysis on torsional stiffness of rectangular bar of (2b) ×
(2h) (b > h) and square bars of (2a) × (2a). They both are solid and made of
the same material. We set (2a)2 = (2b)(2h) or a2 = bh to ensure their areas
are the same. The stiffness ratio of these two bars becomes
 
Jrectangle  (2b)(2h)3 k1 
αstiffness = =
Jsquare a2 =bh (2a)4 k1(b=h) a2 =bh
(9.137)
bh3 k1 h k1
= =
(bh) k1(b=h)
2 b k1(b=h)

To find out how αstiffness changes with b/h, we use the following code to
plot the curve.
402 Mechanics of Materials: Formulations and Solutions with Python

1 n_end = 25; dx = 0.01

2 b_h = [Link](1.0, 10.0+dx, dx) # b/h as x_data
3
4 X_data = [b_h] # put the x (b/h) values to a list
5 k1_data = []
6 for i in range(len(b_h)):
7 k1_data.append(k1_f(b_h[i], n_end)) # function values
8
9 _stiffness = [k1_data[i]/b_h[i]/k1_data[0] for i in range(len(b_h))]
10 Y_data = [ _stiffness] # put the function values to a list
11
12 print(f"b/h={b_h[-1]}, _stiffness={ _stiffness[-1]}")
13 labels=['$ $_stiffness'] # define the label
14 gr.plot_fs(X_data, Y_data, labels, 'b/h', '$ $_stiffness', 'a_stiff')
15 # x-label, y-label, file-name

b/h=10.000000000000007, _stiffness=0.22217349780492124

It is clear that αstiffness < 1 for b > h. This means that the torsional
stiffness of a rectangular bar is smaller than that of the square bar with
the same area. They are equal when h = b. When b/h = 10, the stiffness
reduces to only about 22%. This significant reduction can be explained using
Fig. 9.18. When the height is very small, the distance between these force
couple (one near the top and another near the bottom) of the rectangular
channel becomes small. In addition, for the same twist angle per bar length,
the shear stress on both top and bottom surfaces is also small. The double
reduction effect results in significantly reduction in torque, and hence the
stiffness.
 Torsion of Bars 403

This finding is similar to that given in Section 9.3.11, when comparing

the narrow elliptic and circular bars.

9.5.10 Comparison of torsional stiffness between

circular and square bars
Let us also compare the torsional stiffness of a circular bar of radius R with
a square bar of (2a) × (2a). Assume both bars are solid and made of the
same material. We set πR2 = (2a)(2a) or R2 = π4 a2 to ensure their areas are
the same. The stiffness ratio of these two bars becomes
 
Jcircular  πR4 /2 
αstiffness = =
Jsquare R2 = 4 a2 (2a)4 k1(b=h) R2 = 4 a2
π π
(9.138)
π( π a )
4 2 2
1 1
= = = = 1.132
2(2a)4 k1(b=h) 2πk1(b=h) 2π × 0.140577
This means that the torsional stiffness of a circular bar is about 13.2%
higher than that of the square bar with the same area.

9.5.11 Code to compute the stress function on

rectangular cross-section
We write the following code to compute the stress function for torsion bar
with rectangular cross-section, using Eq. (9.126).
It computes effectively a series solution for any function governed by a 2D
Laplace PDE with a constant right-hand side, and fixed boundary condition.

1 def sf_phi(b, h, Gt, x, y, n_end = 21):

2 '''Compute the stress function phi for torsion bar with rectangular
3 cross-section. It is effectively a series solution to 2D Laplace PDE
4 with a constant right-hand-side, and fixed boundary condition.
5 b: half-height of the domain
6 h: half-width of the domain
7 Gt: constant at the right-hand-side of the Laplace.
8 n_end: terms used in the series of the solution
9 x and y: point at which the solution to be found.'''
10
11 sum1 = 0.0
12 for n in range(1, n_end, 2):
13 sum1 += (-1)**((n-1)/2)/(n**3*[Link](n*[Link]*b/(2*h)))*\
14 [Link](n*[Link]*x/(2*h))*[Link](n*[Link]*y/(2*h))
15
16 return Gt*(h**2-x**2) - (32*Gt*h**2/[Link]**3)*sum1
404 Mechanics of Materials: Formulations and Solutions with Python

9.5.12 Plot of the stress function on rectangular

cross-section
1 [Link]['[Link]'] = 500
2 h, b, Gt, n_end = 2., 1., 1., 21 # arbitrarily set values
3
4 xl, xr, dx = -h, h, 0.02
5 yl, yr, dy = -b, b, 0.02
6
7 X = [Link](xl, xr+dx, dx)
8 Y = [Link](yl, yr+dy, dy)
9 X, Y = [Link](X, Y)
10 xn, yn = [Link]
11 W = X*0
12
13 for xk in range(xn):
14 for yk in range(yn):
15 W[xk, yk] = sf_phi(b, h, Gt, X[xk,yk], Y[xk,yk], n_end)
16
17 fig = [Link]() # Create a 3D plot
18 ax = fig.add_subplot(111, projection='3d')
19
20 surf = ax.plot_surface(X,Y,W,rstride=1,cstride=1,cmap='viridis')
21 ax.set_xlim(-4, 2)
22 ax.set_ylim(-3.5, 1.)
23 ax.set_zlim(-5, 2.5)
24 [Link](False);
25 #[Link]()

Figure 9.23 plots the stress function of a rectangular cross-section using

the code given above. It is the deformed membrane under pressure.

Figure 9.23. A schematic view of the stress function (deformed membrane) for a rect-
angular cross-section.
 Torsion of Bars 405

9.5.13 Code to compute the shear stresses on

rectangular cross-section
We write the following code to compute the shear stresses on a rectangular
cross-section bar subjected to torsion, using Eqs. (9.128) and (9.129).
It computes effectively a series solution for the derivatives of the function
governed by any 2D Laplace PDE with a constant right-hand side, and fixed
boundary condition.

1 def stress_rect(b, h, Gt, x, y, n_end = 21):

2 '''compute the shear stresses on a rectangular cross-section
3 bar subjected to torsion. It is effectively a series solution for
4 the derivatives of the function governed by any 2D Laplace PDE
5 with a constant right-hand-side, and fixed boundary condition.
6 b: half-height of the domain
7 h: half-width of the domain
8 Gt: constant at the right-hand-side of the Laplace.
9 n_end: terms used in the series of the solution
10 x and y: point at which the solution to be found.
11
12 Return: sigam_zx, sigam_zy
13 '''
14 factor = 16*Gt*h/[Link]**2
15 sum_x = 0.0; sum_y = 0.0
16 for n in range(1, n_end, 2):
17 coef = (-1)**((n-1)/2)/(n**2*[Link](n*[Link]*b/(2*h)))
18 sum_x += coef*[Link](n*[Link]*x/(2*h))*[Link](n*[Link]*y/(2*h))
19 sum_y += coef*[Link](n*[Link]*x/(2*h))*[Link](n*[Link]*y/(2*h))
20
21 sigam_zx = - factor*sum_x
22 sigam_zy = 2*Gt*x - factor*sum_y
23
24 return sigam_zx, sigam_zy

9.5.14 Plot of the shear stresses on rectangular

cross-section
1 from matplotlib import cm
2 [Link]['[Link]'] = 500
3 h, b, Gt, n_end = 2., 1., 1., 21 # arbitrarily set values
4
5 xl, xr, dx = -h, h, 0.2
6 yl, yr, dy = -b, b, 0.1
7
8 X = [Link](xl, xr+dx, dx)
9 Y = [Link](yl, yr+dy, dy)
10 X, Y = [Link](X, Y)
406 Mechanics of Materials: Formulations and Solutions with Python

11 xn, yn = [Link]
12 Szx = X*0
13 Szy = Y*0
14
15 for xk in range(xn):
16 for yk in range(yn):
17 Szx[xk, yk], Szy[xk, yk] = stress_rect(b,h,Gt,X[xk,yk],Y[xk,yk] )
18
19 magnitude = [Link](Szx**2 + Szy**2)
20 [Link](X, Y, Szx, Szy, magnitude, scale=20, cmap=[Link])#'viridis')
21 [Link]('equal') #[Link]()
22 [Link]('images/[Link]', dpi=500) # save plot to a file
23 [Link]()

It is seen that the shear stress vector is circulating on the cross-section

(the x−y plan) of the bar. The total circulation is the curl of the stress
vector integrated over the cross-section, relating in the torque with respect
to the z-axis. Because of the circulating nature, the total stress along both
x- and y-direction will be zero. The equilibrium conditions are all satisfied.
In addition, the shear stress is always along the edges of the cross-section,
there will be no shear stress on the circumferential surface of the bar. The
traction boundary conditions are also satisfied.
The stress distribution given in Fig. 9.20 is produced using the same codes
given above.
 Torsion of Bars 407

9.5.15 Example: An I-beam

Consider an I-beam (W760 × 220) with basic shape shown in Fig. 9.21. Its
dimensions (mm) are given as follows:

beam height: hb = 779; flange width: wf = 266; flange thickness: tf = 30;

web thickness: tw = 16.5.

It is subjected to a torque T = 5000 Nm. Assume the shear modulus

G = 200 GPa.
1. Determine the maximum shear stress and its location (Ignore the effects
of fillets and stress concentrations). 2. Determine the twist angle of per unit
length.

Solution to 1: There are of 3 rectangles. Their parameters of h, b, and J

are computed using:

1 =.779; =.266; =.030; =.0165 # (m)

2 T, G = 5000, 200e9 # (Nm), (Pa)
3 n_end = 11 # terms to use
4
5 h2 = [tf, tw, tf] # 2h
6 b2 = [wf, (hb-2*tf), wf] # 2b
7
8 b_k= [b2[i]/h2[i] for i in range(len(h2))]
9 print(f" Ratios of bi/hi = {b_k} ")
10
11 k1 = [k1_f(b2[i]/h2[i], n_end) for i in range(len(h2))]
12 k = [k_f( b2[i]/h2[i], n_end) for i in range(len(h2))]
13 Ji = [k1[i]*b2[i]*h2[i]**3 for i in range(len(h2))]
14
15 J = sum(Ji)
16 print(f" 2h values = {h2} ")
17 print(f" k1 values = {k1} ")
18 print(f" k values = {k} ")
19 print(f" J values = {[Link](Ji)} (m^4)")
20 print(f" Total J values = {J} (m^4)")

Ratios of bi/hi = [8.866666666666667, 43.57575757575758, 8.866666666666667]

2h values = [0.03, 0.0165, 0.03]
k1 values = [0.30964005201232053, 0.3285122943107051, 0.30964005201232053]
k values = [0.9999985509443367, 1.0, 0.9999985509443367]
J values = [0.0000022 0.0000011 0.0000022] (m^4)
Total J values = 5.508711157672834e-06 (m^4)
408 Mechanics of Materials: Formulations and Solutions with Python

The maximum shear stress is computed using Eq. (9.132):

1 hi_max = [Link](h2)
2 max = k[hi_max]*T*max(h2)/J
3 print(f"The maximum shear stress = { max*1.0e-6} (MPa)")

The maximum shear stress = 27.229560299730473 (MPa)

Solution to 2: After the effective polar 2nd moment of area J is known,

the twist angle of per unit length is computed using genetic Eq. (9.14):

1 t = T/G/J
2 print(f"The twist angle of per unit length = { t} (rad/m)")

The twist angle of per unit length = 0.004538266626156035 (rad/m)

9.6 Hollow thin-wall torsion members, use of

membrane analogy*

9.6.1 Solution strategy

Consider a bar with a cross-section that is a hollow enclosed thin-wall, as
shown in Fig. 9.24. The thickness of the wall need not be constant, and the
shape of the cross-section can be arbitrary. Our formulation and notation
follow largely the textbook [4]. Codes are developed for computations.
Before our analysis, we make the following assumptions:

1. The wall thickness is small compared to the other dimensions of the cross-
section.
2. The shear stress is constant through the wall thickness (It may not be
constant along the boundary, unless the thickness t is constant).

Based on the membrane analogy studied earlier, we know that the stress
function φ relates to the deformed membrane w just by a constant factor.
Because φ must satisfy Eq. (9.65) on the boundary of the cross-section, it
must be either zero or a constant there. We set φ = 0 on the outer boundary
of the thin-wall. On the inner boundary it must be a constant, and it is
noted as φ1 . Correspondingly, w must also be zero on the outer boundary,
and a constant on the inner boundary and it is noted as w1 . This leads to
the drawing on the bottom of Fig. 9.24.
 Torsion of Bars 409

Figure 9.24. A hollow enclosed thin-wall cross-section for bars subjected to torsion
(top), and the idealized deformed membrane w on the cross-section (bottom).

Using Eq. (9.69), we have τ = − ∂φ

∂n , where n is normal to contour of w
(or φ), we shall have
 
1 ∂z 1 1 w1
τ= − = tan α = (9.139)
c ∂n c c t
  
tan α

Introducing q as follows:
w1
q = τt = = φ1 (9.140)
c
Clearly, q has a unit of [Nm], and is called shear flow. It must be a
constant because φ1 is a constant. Equation (9.139) shows that the shear
stress τ is inversely proportional to the thickness of the wall t, and hence
the maximum shear stress is at the smallest t on the wall.
For a thin-wall member with perimeter segments l1 , l2 , . . . , ln , each of
which with constant thickness t1 , t2 , . . . , tn , the corresponding shear stresses
become: τ1 = q/t1 , τ2 = q/t2 , . . . , τn = q/tn .
410 Mechanics of Materials: Formulations and Solutions with Python

9.6.2 Formula for twist angle

We need to relate the twist angle with the shear stress. For this, we use
the again the membrane analogy. Considering the equilibrium of force in the
z-direction, we have

Fz = 0 : pA − S tan αdl = 0 (9.141)

where dl is the infinitely small length segment of the mean perimeter of the
cross-section. Equation (9.141) is rewritten as

w1
pA = S dl = 0
t

pA 1
= w1 dl
S t

pA w1 1
= dl (9.142)
cS c t

pA 1
= φ1 dl
cS t

1
2GAϑt = q dl
t
We thus have
 
q 1 1
ϑt = dl = τ dl (9.143)
2GA t 2GA

This equation is the same as Eq. (9.51).

If t is piecewise constant, and there are n pieces, we have
 n
1 l1 l2 ln li
dl = + + ··· + = (9.144)
t t1 t2 tn ti
i

The twist angle becomes

n
q li
ϑt = (9.145)
2GA ti
i

Further, if the thickness of the wall is uniformly t, we have

ql τl
ϑt = = (9.146)
2GAt 2GA
where l is the length of mean perimeter of the cross-section.
 Torsion of Bars 411

9.6.3 Torque stress relation

Using now Eq. (9.73), the torque is twice the volume of the stress function.
We obtain torque written in stress:

T =2 φdxdy = 2Aφ1 = 2Aq = 2Aτ t (9.147)

A

where A is the area enclosed by the mean perimeter of the cross-section.

This is Eq. (9.42).
The twist angle can be written in torque:
τl Tl
ϑt = = (9.148)
2GA 4GtA2

9.6.4 Equivalent polar 2nd moment of area

T
This can be written in the conventional form ϑt = GJ , if we write J as
4A2 4tA2
J =  1 = (9.149)
general
t dl
uniform l
This is Eq. (9.54).
Assuming the thin-wall cross-section is a thin hollow cylinder with uni-
form thickness t, Eq. (9.149) becomes
4tA2 4t(πR)2
J= ≈ = 2πR3 t (9.150)
l 2πR
where R is the mean radius. This formula is Eq. (9.26) we obtained earlier.
This means that Eq. (9.149) is the general form for computing the equivalent
polar moment of area for closed thin-wall cross-sections.
If the thin-wall cross-section is a hollow rectangle with uniform thickness
t, Eq. (9.149) becomes
4tA2 4t(hb)2
J= ≈ = 8b3 t (9.151)
l rectangle h+b square

where h is the mean half-height, and b is the mean half-width of the hollow
rectangle.

9.6.5 Example: Bars with thin-wall circular

cross-sections
Consider two bars of thin-wall circular cross-section. Bar-1’s cross-section is
closed, and bar-2’s cross-section has a small cut, as shown in Fig. 9.25.
412 Mechanics of Materials: Formulations and Solutions with Python

Figure 9.25. Two thin-wall circular cross-sections. One is closed (left), and one has a
small cut (right).

Its dimensions are:

outer diameter is 22 mm; inner diameter is 18 mm.

1. Calculate the J values of bar-1 without assuming it is a thin-wall.

2. Calculate the J values of bar-1 assuming it is a thin-wall.
3. Calculate the J values of bar-2 assuming it is a narrow channel.
4. Assume the material shear stress is allowed at 150 MPa. Calculate the
torque, based on the three assumptions in items 1, 2, and 3.
5. Assume the material shear stress is allowed at 150 MPa. Calculate the
twist angle per unit length, based on the three assumptions in items 1, 2,
and 3.

Solutions: We write the following code to solve and answer these questions.

1 d_outer = .022; d_inner = .018

2 t = (d_outer - d_inner)/2 # thickness
3 r = d_outer/2 # radius of the outer circle
4 R_mean = (d_outer + d_inner)/4
5 A_mean = [Link]*R_mean**2
6 l = 2*[Link]*R_mean
7
8 J_exact = [Link]/32*(d_outer**4-d_inner**4) # Q1
9 J_thinW = 4*t*A_mean**2/l # Q2
10 k1 = k1_f(l/t, 21)
11 J_chnnl = k1*l*t**3 # Q3
12
13 print(f"k1 = {k1} ")
14 print(f"J_exact = {J_exact} (mm^4)")
15 print(f"J_thinW = {J_thinW} (mm^4)")
16 print(f"J_chnnl = {J_chnnl} (mm^4)")
17 print(f"J_chnnl/J_thinW = {J_chnnl/J_thinW}")
 Torsion of Bars 413

k1 = 0.32664619022207575
J_exact = 1.2692034320502761e-08 (mm^4)
J_thinW = 1.2566370614359166e-08 (mm^4)
J_chnnl = 1.6419028344396277e-10 (mm^4)
J_chnnl/J_thinW = 0.013065847608883036

It is found that the torsional stiffness of the bar with a cut is only ∼1.3%
of the closed one.

1 # Q4
2 _allow = 70.0e6 # Pa
3 T_exact = _allow * J_exact / r
4 T_thinW = _allow * J_thinW / R_mean # or use: 2*A_mean* _allow*t
5 T_chnnl = _allow * J_chnnl / t
6
7 print(f"T_exact = {T_exact} (Nm)")
8 print(f"T_thinW = {T_thinW} (Nm)")
9 print(f"T_chnnl = {T_chnnl} (Nm)")
10 print(f"T_chnnl/T_thinW = {T_chnnl/T_thinW}")

T_exact = 80.76749113047212 (Nm)

T_thinW = 87.96459430051418 (Nm)
T_chnnl = 5.746659920538697 (Nm)
T_chnnl/T_thinW = 0.06532923804441518

It is found that the loading capacity of the bar with a cut is only ∼6.5%
of the closed one.

1 # Q5
2 G = 77.5e9 # Pa
3 _exact = T_exact / J_exact / G
4 _thinW = T_thinW / J_thinW / G
5 _chnnl = T_chnnl / J_chnnl / G
6
7 print(f" _exact = { _exact} (Nm)")
8 print(f" _thinW = { _thinW} (Nm)")
9 print(f" _chnnl = { _chnnl} (Nm)")
10 print(f" _chnnl/ _thinW = { _chnnl/ _thinW}")

_exact = 0.08211143695014664 (Nm)

_thinW = 0.0903225806451613 (Nm)
_chnnl = 0.45161290322580644 (Nm)
_chnnl/ _thinW = 4.999999999999999

It is found that the twist angle of the bar with a cut is about 5 times the
closed one, for the same allowable stress level.
In conclusion, a cut significantly reduces the loading capacity and the
stiffness of a torsional bar with a thin-wall cross-section.
414 Mechanics of Materials: Formulations and Solutions with Python

9.7 Remarks
This chapter presents theory, formulation, techniques, and codes for comput-
ing twist-angles, shear stress, and stiffness of bars with various cross-section
shapes, subjected to torsion. It is clear that the shape of the cross-section
makes differences.
The following chapter presents theory, formulation, techniques, and codes
for computing various properties of different shapes of cross-sections, which
are useful for stress analysis and design for bars subject to torsion and beams
subject to bending.

References
[1] G.R. Liu and S.S. Quek, The Finite Element Method: A Practical Course.
Butterworth-Heinemann, New York, 2013.
[2] G.R. Liu and T.T. Nguyen, Smoothed Finite Element Methods. Taylor and Francis
Group, New York, 2010.
[3] G.R. Liu, Mesh Free Methods: Moving Beyond the Finite Element Method. Taylor
and Francis Group, New York, 2010.
[4] A.P. Boresi and R.J. Schmidt, Advanced Mechanics of Materials. John Wiley & Sons,
New York, 2002.
[5] J.M. Gere and S.P. Timoshenko, Mechanics of Materials. Van Nostrand Reinhold
Company, New York, 1972.
[6] S. Timoshenko and J.N. Goodier, Theory of Elasticity. McGraw Hill, New York,
1970. [Link]
asticity&#38;ei=ICiKSsr3G4jwkQSbxMyPCg.
[7] Z.L. Xu, Elasticity, Vols. 1&2. People’s Publisher, China, 1979.
[8] T.H.G. Megson, Structural and Stress Analysis, 1996. Available online https://
[Link]/books?id=qZykmEz4nzkC.
[9] G.R. Liu, Vector Calculus: Formulation, Applications, and Python Codes, World
Scientific, New Jersey, 2025.
 Chapter 10

Property of Areas

1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.

2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 init_printing(use_unicode=True) # For latex-like quality printing
11 #np.set_printoptions(precision=6,suppress=True) # Digits in print-outs
12 np.set_printoptions(precision=4,suppress=True,
13 formatter={'float_kind': '{:.4e}'.format})

Area in general refers to any planar region that may have an arbitrary shape.
In mechanics, many analyses involve areas of various shapes resulting from
structural components. A typical case is the cross-section area of beams
undergoing bending, bars under torsion, and bars undergoing tension. This
chapter deals with issues on properties of areas, including area, the first
moment of area, and the second moment of area, which are frequently used
in mechanics.

10.1 Area of areas

10.1.1 Integral formulas

The most essential property of an area (planar region) is its area, often
denoted as A. It measures the extent of the coverage of the area and carry a

415
416 Mechanics of Materials: Formulations and Solutions with Python

unit of [m2 ]. In our study, we discuss regions that cover a nonzero area, and
hence A is always nonzero and positive.
Because an area can have different shapes, its measure needs to be done
generally using integrals [1]:

A= dA (10.1)
A
The subscript A for the area integral stands for the planar region area.
This means that we can cut the region into small pieces of area dA and then
sum all these areas up leading to the area measure A. If a sheet of material
is uniform in thickness and density, it measures the amount of material.
Therefore, the area of an area is the physical property of the area. It does
not depend on the coordinate system. Any coordinate system can be used
to evaluate the area, and we often choose the one that is easy to work with.
If the shape of the area is regular, we would then have more explicit
integrals, with the help of a coordinate system. For example, it can be a
rectangle with base b and height h, as shown in Fig. 10.1.
We can use the 2D Cartesian coordinate system (x , y  ) with the x -axis
along its base b and the origin at its left-bottom corner. Equation (10.1)
becomes
  h  b   h
 
A = dA = dx dy = b dy  = bh (10.2)
general A rectangle 0 0 0
In the foregoing integral process, the area integral becomes a so-called
double integral, which is integrated in two steps: one along the horizontal

Figure 10.1. A rectangular area and the Cartesian coordinates set for evaluating its
area. Two Cartesian coordinate systems are used.
 Property of Areas 417

axis and the other along the vertical axis. We can also use the 2D Cartesian
coordinate system (x, y) with the origin passing through its center and the
x-axis parallel to its base b. Equation (10.1) becomes
  h/2  b/2   h/2
A = dA = dx dy = b dy = bh (10.3)
general A rectangle −h/2 −b/2 −h/2

The use of two different coordinates gives the same value. This coordinate
independence feature of area evaluation is important to note. The code to
get all these done is as follows:
1 x, y, xp, yp, b, h = [Link]("x, y, xp, yp, b, h")
2 Ap = [Link]([Link](1, (yp, 0, h)), (xp, 0, b))
3 A = [Link]([Link](1, (xp, -b/2, b/2)), (yp, -h/2, h/2))
4 print(f"Area of rectangle, in (x',y') = {Ap}; in (x,y) = {A}")

Area of rectangle, in (x',y') = b*h; in (x,y) = b*h

The code can be modified for computing the area of more complicated
shapes.
If the shape is circular with a radius of a, we can use the polar coordinate
system (r, θ) with its origin at the center of the circle, and the integral
becomes
 a  2π   a
A = rdθ dr = 2π rdr = πa2 (10.4)
circular 0 0 0
If the shape is totally irregular, we need to cut it into very small pieces
without gaps and overlaps, called meshing, typically into rectangles and/or
triangles, each of whose area can be computed with ease, and then sum
all these areas up. This is essentially a numerical discrete evaluation of
Eq. (10.1). Assume the area is then divided into n small areas ΔAi . The
total area becomes
n 
A= ΔAi n→∞= dA (10.5)
i=1 ΔAi →0 A

which is essentially Eq. (10.1) under the limits.

It is important to note that A is a global measure of an area. It does not
give any information on the shape of the area. In other words, many areas
with different shapes can have the same A value. It provides only 0th order
of information on a given area.

10.1.2 Area of a composite area

Because of the summation form of the definition, the area calculation can be
done for a complicated area consisting of multiple regular sub-areas for which
418 Mechanics of Materials: Formulations and Solutions with Python

the area can be found with ease. Assuming there are m regular sub-areas,
the general formula is as follows:
m
A= Ai (10.6)
i=1

where Ai is the area of the ith sub-area. A list of formulas for calculating
the areas of various relatively regular areas can be found at the Wikipedia
page on area ([Link]

10.2 First moment of area

10.2.1 Definition
The concept of area and its calculation are quite straightforward and taught
in elementary schools. The concept, calculation, and property of the “first
moment of area” are, however, much less trivial, and their use is quite con-
fusing in the open literature. They are typically introduced in high schools.
The first moment of area is a measure of the spatial distribution of an
area in relation to an axis. Therefore, it depends on the reference axis of a
coordinate system used to evaluate it. It gives the first order of information
about a given area. The shape of it matters. Its definition is given as follows.
Consider a given area A of arbitrary shape.

1. First, a Cartesian coordinate system (x, y) is introduced, and the coordi-

nates for any point in the area can be found.
2. The area is then divided into n small areas ΔAi . The coordinates of the
center within ΔAi are found and denoted as (xi , yi ).

The first moments of area A are defined as

n 
Qx = yi ΔAi =
n→∞
y dA
arm wrt x ΔAi →0 A
i=1
n  (10.7)
Qy = xi ΔAi =
n→∞
x dA
arm wrt y ΔAi →0 A
i=1

The unit for the first moment of area is [m3 ].

We reiterate that the first moment of an area is with respect to (wrt)
the reference axis in a coordinate system. A coordinate transformation,
translational or rotational, will affect its values. The first moment of
an area has two components: Qx and Qy , and it needs one index. It is a
 Property of Areas 419

vector, a first-order tensor, and the coordinate rotational transformation

rule for vectors (discussed in Chapter 2) applies.
The effects of a translation of the coordinate system are revealed in the
following example.

10.2.2 First moment of a composite area

Because of the summation form of the definition, the first moment of area
calculation can be done for a complicated area consisting of multiple regular
sub-areas (for which the center and area can be found with ease). Assuming
there are m regular sub-areas, the general formulas are as follows:
m
Qx = ȳi Ai
i=1
(10.8)
m
Qy = x̄i Ai
i=1

where Ai is the area of the ith sub-area and xi , yi are coordinates of the
center of the ith sub-area with respect to the coordinate system (x, y).

10.2.3 Example: The first moment of area of

a rectangle
Consider a rectangle with base b and height h, as shown in Fig. 10.2. We
use two Cartesian coordinate systems: (x , y  ) that pass through its two
edges, and (x, y) that pass through its two middle lines and the center of the
rectangle.

1. Calculate the first moment of area of the rectangle with respect to x -

and y  -axis.
2. Calculate the same with respect to the x- and y-axis.

Solution to 1: Using the coordinate system (x , y  ) and Eq. (10.7), we obtain
  b  h  h
    bh2
Q x = y dA = dx y dy = b y  dy  =
A 0 0 0 2
    (10.9)
h b b
     hb2
Qy  = x dA = dy x dx = h x dx =
A 0 0 0 2
420 Mechanics of Materials: Formulations and Solutions with Python

Figure 10.2. A rectangular area and the Cartesian coordinates set for evaluating its
first moment of area. Two Cartesian coordinate systems are used.

We see clearly that the shape, the values of a and b, of the rectangle will
affect the values of the first moments. The shape does matter for the first
moments of area.

Solution to 2: Using the coordinate system (x, y) and Eq. (10.7), we obtain
  b/2  h/2  h/2
Qx = y dA = dx y dy = b y dy = 0
A −b/2 −h/2 −h/2
(10.10)
  h/2  b/2  b/2
Qy = x dA = dy x dx = h x dx = 0
A −h/2 −b/2 −b/2

The code to get all these done is as follows:

1 x, y, xp, yp, b, h = [Link]("x, y, xp, yp, b, h")

2 Q_xp = [Link]([Link](yp, (yp, 0, h)), (xp, 0, b))
3 Q_yp = [Link]([Link](xp, (xp, 0, b)), (yp, 0, h))
4 Qp = [Link]([Q_xp, Q_yp])
5 [Link](Qp, "1st moments of area of rectangle: Q_x', Q_y':")

1st moments of area of rectangle: Q_x', Q_y':

 bh2 
2
b2 h
2
 Property of Areas 421

1 Q_x = [Link]([Link](y, (y, -h/2, h/2)), (x, -b/2, b/2))

2 Q_y = [Link]([Link](x, (x, -b/2, b/2)), (y, -h/2, h/2))
3 Q = [Link]([Q_x, Q_y])
4 [Link](Q, "1st moments of area of rectangle: Q_x, Q_y:")

1st moments of area of rectangle: Q_x, Q_y:

 
0
0
We find the following:
1. The moments of an area are coordinate dependent.
2. When the reference axis passes through the center of the area, the first
moment of the area becomes zero. This leads to the definition of the
concept of a centroid.

10.2.4 Centroid of an area

The centroid of an area, denoted as (x̄, ȳ), is defined based on the first
moment of area. The formula is as follows:
A x dA
x̄ =
A dA
(10.11)
y dA
ȳ = A
A dA

From this definition and because the area A is not zero (we are not
interested in discussing any area that is zero), it is clear that
the centroid is zero if and only if the first moments are zero.
This means that if the coordinates are set at the centroid of the area, the
first moment vanishes.
For the rectangular area given in the previous section, in the (x , y  ) coor-
dinate system, the centroid is found as
 dA b2 h
Ax b
x̄ = = 2
=
A dA bh 2
(10.12)
bh 2
y  dA h
y¯ = A = 2 =
A dA bh 2
Therefore, the centroid is the coordinates of the center of the area in the
given reference coordinate system. Using the (x, y) coordinate system, the
422 Mechanics of Materials: Formulations and Solutions with Python

centroid is then found as

A x dA 0
x̄ = = =0
A dA bh
(10.13)
y dA 0
ȳ = A = =0
A dA bh
This confirms that the center of the area is right at the origin of the (x, y)
coordinate system.
In many applications, it is often convenient to set our reference coordinate
system with the origin at the center of the area of interest. In this case,
the first moments become zero, and the formulation can be significantly
simplified. However, we do not usually know the location of the center for a
given arbitrary area. This is what we can do:
First, set up any coordinate system as a reference, and use Eq. (10.11) to
find the centroid. Then, move the coordinate system to have its origin at
the centroid.
Equation (10.11) is thus useful in many applications involving areas.

10.2.5 Centroid for composite areas

Consider a composite area with m regular sub-areas. Its centroid can be
computed using
m
i=1 x̄i Ai
x̄ = m
i=1 Ai
m
(10.14)
i=1 ȳi Ai
ȳ = m
i=1 Ai
where Ai is the area of the ith sub-area, and x̄i and ȳi denote the centroid
of the ith sub-area with respect to the common coordinate system (x, y).

10.2.6 Example: Centroid of an I-beam

Consider an I-beam with a basic shape, as shown in Fig. 10.3.
Its dimensions (mm) are given as follows:
beam height: hb = 200; width of flange on top: wf t = 100; width of flange
at bottom: wf b = 150; thickness of flange on top: tf t = 20; thickness of
flange at bottom: tf b = 18; web thickness: tw = 10.
Determine the centroid of the I-beam.
 Property of Areas 423

Figure 10.3. Schematic view of an I-beam. Left: generic cross-section with rounded
concave corners to avoid stress concentration. Right: idealized cross-section for analysis.
The image on the left is from Poutrelle metallique at Wikimedia Commons under the CC
BY 3.0 license.

Solution: We write the following code to get it done using Eq. (10.14). First,
we write a code function for computing the centroid of an I-beam, which will
be used multiple times in this book.

1 def centroid_I(hb, wft, wfb, tft, tfb, tw):

2 '''Compute the centroid of an I-beam made of rectangles.
3 The reference coordinate is on the bottom edge with the left-bottom
4 corner as the origin. The code can be used for any composite area
5 with up to 3 rectangles. For a T beam, for example, just set tfb=0.
6 The whole area must be left-right symmetric.
7 Inputs:
8 hb: beam height (total)
9 wft: width of flange on top
10 wfb: width of flange at bottom
11 tft: thickness of flange on top
12 tfb: thickness of flange at bottom
13 tw: web thickness
14 Return:
15 x_bar, y_bar: the centroid of the composite area.
424 Mechanics of Materials: Formulations and Solutions with Python

16 A: The total area

17 Ai: areas of each rectangles
18 '''
19 width_ai = [Link]([wft, tw, wfb])
20 height_bi = [Link]([tft, hb-tft-tfb, tfb])
21 xi_bar = [Link]([wfb/2, wfb/2, wfb/2])
22 yi_bar = [Link]([hb-tft/2, (hb-tft-tfb)/2+tfb, tfb/2])
23
24 Ai = width_ai*height_bi
25 A = [Link](Ai)
26 print(f"Areas of sub-areas = {Ai} [L^2]")
27 print(f"Areas of I-beam = {A} [L^2]")
28
29 x_bar = [Link](xi_bar*Ai)/A
30 y_bar = [Link](yi_bar*Ai)/A
31
32 print(f"The centroid [L]):\n x = {x_bar} \n y = {y_bar}")
33 return x_bar, y_bar, A, Ai

Using the code function given above, the centroid can be found with ease.
1 # preparing data:
2 hb =.200; wft=.100; wfb=.150; tft=.020; tfb=.018; tw=.010 # (m)
3
4 # call the code function:
5 x_bar, y_bar, A, Ai = centroid_I(hb, wft, wfb, tft, tfb, tw)

Areas of sub-areas = [2.0000e-03 1.6200e-03 2.7000e-03] [L^2]

Areas of I-beam = 0.00632 [L^2]
The centroid [L]):
x = 0.075
y = 0.0893481012658228

10.2.7 Example: Compute the first moment of an area,

using centroid
Consider again a rectangle with base b and height h shown in Fig. 10.2.
Because of its simple geometry, we know its centroid based on a Cartesian
coordinate system. For example, when we use (x , y  ) as the reference coordi-
nate system, its centroid is at x̄ = b/2, ȳ  = h/2. Calculate the first moment
of the rectangle using the known centroid.
Solution: Using Eq. (10.11), we have


Qx = ȳ dA = ȳ  A
A
 (10.15)
 
Qy = x̄ dA = x̄ A
A
 Property of Areas 425

Substituting x̄ = a/2, ȳ  = b/2, and A = ab for a rectangle into

Eq. (10.15), we obtain

h bh2
Qx = ȳ  A = bh =
2 2
(10.16)
b hb2
Qy = x̄ A = bh =
2 2
which is the same as the results given in Eq. (10.9).
Equation (10.15) is useful for regular areas for which the centroid and
area are readily available.

10.2.8 First moment of area, coordinate rotation

Since the first moment of an area has two components: Qx and Qy with one
index, the coordinate rotational transformation rule follows that for vectors
discussed in Chapter 2. Consider a new Cartesian coordinate system (x , y  )
that is rotated by an angle θ from the current coordinate system (x, y).
Assume the first moments of area are known in (x, y). We would like to find
the first moments of the same area with respect to the rotated (x , y  ). Using
the transformation matrix defined in Chapter 2, the new coordinates are
given by
    
x cos θ sin θ x
= (10.17)
y −sin θ cos θ y
 
T

The first moments of area in (x , y  ) becomes

 

Qx  = y dA = (−x sin θ + y cos θ)dA = −Qy sin θ + Qx cos θ
A A
  (10.18)

Qy  = x dA = (x cos θ + y sin θ)dA = Qy cos θ + Qx sin θ
A A

In matrix form, they become

    
Qx  cos θ −sin θ Qx
= (10.19)
Qy  sin θ cos θ Qy
 
T

This means that the transformed first moments use the transpose of the
transformation matrix T. If both Qx and Qy are zero, both Qx and Qy will
426 Mechanics of Materials: Formulations and Solutions with Python

also be zero. This always remains true because if the origin of the original
coordinate is at the centroid, any rotated coordinate will have the same
origin.

10.3 Second moment of area

10.3.1 Definition
One can increase the order of the integrand by one when performing area
integral. This constructs the second moments of an area. The formula is as
follows:

Ixx = y 2 dA
A

Iyy = x2 dA (10.20)
A

Ixy = xy dA
A

It has three components and carries two indexes. The first moment of
an area extracts the first order of information from the area. One can nat-
urally expect that a second moment of an area shall extract second-order
information from it. We need to find a way to calculate it.

10.3.2 Second moment of area for regular areas

For areas with regular shape, we can derive formulas, and the calculation can
be done analytically. Consider again the rectangular area shown in Fig. 10.2.
We use the Cartesian coordinate system (x, y), and let the x- and y-axis be
on the middle lines of the rectangle. We have
  b/2  h/2  h/2
bh3
Ixx = y dA =
2
dx y dy = b
2
y 2 dy =
A −b/2 −h/2 −h/2 12
  h/2  b/2  b/2
hb3
Iyy = x2 dA = dy x2 dy = h x2 dy = (10.21)
A −h/2 −b/2 −b/2 12
  b/2  h/2
Ixy = xy dA = xdx y dy = 0
A −b/2 −h/2

Here, we see the benefit of having the coordinate going through the cen-
troid of the area. Using a similar integration technique, the second moment
 Property of Areas 427

of area for other regular shapes can be found. The code to get this done is
as follows:

1 Ixx = [Link]([Link](y**2, (y, -h/2, h/2)), (x, -b/2, b/2))

2 Iyy = [Link]([Link](x**2, (x, -b/2, b/2)), (y, -h/2, h/2))
3 Ixy = [Link]([Link]( x*y, (y, -h/2, h/2)), (x, -b/2, b/2))
4 I = [Link]([[Ixx, -Ixy], [-Ixy, Iyy]])
5 [Link](I, "2nd moments of area of rectangle:")

2nd moments of area of rectangle:

 bh3 
12 0
b3 h
0 12

We found in the rectangle case, Ixy = 0. This is because the area is

symmetric. For a general shape, it is usually not.
Consider next a box area that is symmetric in both directions, shown in
Fig. 10.4. We use the Cartesian coordinate system (x, y) with its origin at
the centroid of the box.

Figure 10.4. A rectangular box.

428 Mechanics of Materials: Formulations and Solutions with Python

Because of the additive property of integral operations, the second

moment of the box area can be simply obtained using

 
1
Ixx = y 2 dA − y 2 dA = (bh3 − b h3 )
A A 12
  (10.22)
1
Iyy = x2 dA − x2 dA = (hb3 − h b3 )
A A 12

Similar techniques can be used to calculate the I values of areas of var-

ious shapes. A list of second moments of area for various regular shapes
can be found at the wikipage ([Link] of second
moments of area).

10.3.3 Second moment of area, coordinate translation

Consider a new Cartesian coordinate system (x , y  ) that is created by a
translation of (x, y), and we found the centroid of the area as x̄ and ȳ in
reference to (x , y  ). The second moment of the area in reference to (x , y  )
can be found as
 
2 2
Ix  x  = y dA = (y + ȳ)2 dA = Ixx + y¯ A
A A
 
x2 dA =
2
Iy  y  = (x + x̄)2 dA = Iyy + x̄ A (10.23)
A A
 
2
Ix  y  = y dA = (x + x̄)(y + ȳ) dA = Ixy + x̄ y¯ A
A A

where Ixx , Iyy , and Ixy are the second moments of area with respect to
coordinates (x, y) that are passing through the centroid. In the derivation
above, we used the fact that the first moments of area with respect to x
and y are all zero, leading to the simple coordinate translation rule for these
second moments of area. Equation (10.23) is useful when calculating the
second moment of a composite area consisting of multiple regular sub-areas
for which their second moments of area and centroids are readily available.
 Property of Areas 429

10.3.4 Second moments of composite areas

Consider a composite area with m regular sub-areas, its second moments of
area can be computed using

m
2
Ix  x  = (Ixx,k + y¯k Ak )
k=1
m
2
I y y = (Iyy,k + x¯k Ak ) (10.24)
k=1
m
I x y  = (Ixy,k + x¯k y¯k Ak )
k=1

where Iij,k (i = x, y; j = x, y) is the second moments of the kth sub-area with

respect to its own coordinates (xk , yk ) that go through its own center. And
x¯k and y¯k are the centroid of the kth sub-area with respect to the common
coordinate system (x , y  ) for the whole area.
Note that in practice, we often set the common coordinate system (x , y  )
going through the centroid of the whole area, so that the obtained second
moments are in reference to the center of the whole area. This can be easily
done by shifting the coordinates, after the global centroid is found, using for
example centroid I(). We will demonstrate this in the example section.

10.3.5 Example: Second moments of an I-beam

Consider an I-beam with a basic shape shown in Fig. 10.3.
Its dimensions (mm) are given as follows:

beam height: hb = 200; width of flange on top: wf t = 100; width of flange

at bottom: wf b = 150; ; thickness of flange on top: tf t = 20; thickness of
flange at bottom: tf b = 18; web thickness: tw = 10.

Determine the second moments of the I-beam.

Solution: We write a code to get it done using Eq. (10.24). First, we write
a Python function for computing the second moments of an I-beam using
the centroids of sub-areas.
430 Mechanics of Materials: Formulations and Solutions with Python

1 def secondMoments_I(hb, wft, wfb, tft, tfb, tw, cxi, cyi, sy=True):
2 '''Compute the second moments of I-beam cross-sections made of
3 rectangles. The reference coordinate is on the bottom edge with
4 the left-bottom corner as the origin and x-axis on the bottom edge.
5 The code can be used for any composite area
6 with up to 3 rectangles. For a T beam, for example, just set tfb=0.
7 Inputs:
8 hb: beam height (total)
9 wft: width of flange on top
10 wfb: width of flange at bottom
11 tft: thickness of flange on top
12 tfb: thickness of flange at bottom
13 tw: web thickness
14 cxi, cyi: centers of each sub-area wrt the common coordinates.
15 sy: if True (I-, T-beam), cxi, cyi are not used, computed here in.
16 Otherwise, the provided cxi, cyi will be used here.
17 Return:
18 Ixx, Iyy, Ixy: 2nd moments
19 x_bar, y_bar: the centroid
20 A: The total area
21 Ai: areas of each rectangles
22 '''
23 width_ai = [Link]([wft, tw, wfb])
24 height_bi = [Link]([tft, hb-tft-tfb, tfb])
25
26 # when whole area is symmetric with respect to y-axis:
27 if sy: #compute the centers (cxi, cyi) for each sub-area.
28 cxi = [Link]([wfb/2, wfb/2, wfb/2])
29 cyi = [Link]([hb-tft/2, (hb-tft-tfb)/2+tfb, tfb/2])
30
31 print(f"Center of sub-areas [L]:\n x = {cxi} \n y = {cyi}")
32
33 Ai = width_ai*height_bi
34 A = [Link](Ai)
35 print(f"Areas of sub-areas = {Ai} [L^2]")
36 print(f"Areas of I-beam = {A} [L^2]")
37
38 x_bar = [Link](cxi*Ai)/A
39 y_bar = [Link](cyi*Ai)/A
40
41 print(f"The centroid = ({x_bar}, {y_bar}) [L]")
42
43 # shifts the coordinates to the centroid of the whole area:
44 xi_bar = cxi - x_bar
45 yi_bar = cyi - y_bar
46
47 print(f"New center of sub-areas [L]:\n x = {xi_bar}\n y = {yi_bar}")
48
49 Ixx = [Link](width_ai*height_bi**3/12 + yi_bar**2 * Ai)
50 Iyy = [Link](height_bi*width_ai**3/12 + xi_bar**2 * Ai)
51 Ixy = [Link](xi_bar*yi_bar * Ai)
52
53 print(f"2nd moments [L^4]:\n Ixx = {Ixx}\n Iyy = {Iyy}\n Ixy = {Ixy}")
54
55 return Ixx, Iyy, Ixy, x_bar, y_bar, xi_bar, yi_bar, A, Ai
 Property of Areas 431

1 # preparing data:
2 #hb =.060; wft=.050; wfb=.050; tft=.01; tfb=.0; tw=.01 # T-beam (m)
3 hb =.200; wft=.100; wfb=.150; tft=.02; tfb=.02; tw=.01 # I-beam (m)
4
5 # call the code function for 2nd moment calculation:
6 I_xx, I_yy, I_xy, x_bar, y_bar, xi_bar, yi_bar, A, Ai =\
7 secondMoments_I(hb, wft, wfb, tft, tfb, tw, [0]*2, [0]*2, sy=True)

Center of sub-areas [L]:

x = [7.5000e-02 7.5000e-02 7.5000e-02]
y = [1.9000e-01 1.0000e-01 1.0000e-02]
Areas of sub-areas = [2.0000e-03 1.6000e-03 3.0000e-03] [L^2]
Areas of I-beam = 0.0066 [L^2]
The centroid = (0.075, 0.08636363636363638) [L]
New center of sub-areas [L]:
x = [0.0000e+00 0.0000e+00 0.0000e+00]
y = [1.0364e-01 1.3636e-02 -7.6364e-02]
2nd moments [L^4]:
Ixx = 4.2852727272727287e-05
Iyy = 7.304999999999999e-06
Ixy = 0.0

Because the cross-section of the I-beam is symmetric with respect to the

y-axis, the product moment of area Ixy found is zero. If the cross-section of
the beam is asymmetric, Ixy will not be zero. The flowing code computes
the second moments of an S-beam without left-right symmetry.

10.3.6 Example: Second moments of an S-beam

Consider an S-beam with a basic shape shown in Fig. 10.5.
Its dimensions (mm) are given as follows:

beam height: hb = 200; width of flange on top: wf t = 80; width of flange at

bottom: wf b = 80; thickness of flange on top: tf t = 15; thickness of flange
at bottom: tf b = 15; web thickness: tw = 10.

Determine the centroid and the second moments of the S-beam.

Solution: Because the cross-section is asymmetric, we need to compute the

centroids of these three rectangles of the S-beam, and then feed these to
secondMoments I() to compute the second moments of areas. The code is as
follows:
432 Mechanics of Materials: Formulations and Solutions with Python

Figure 10.5. Schematic view of an S-beam.

1 # preparing data:
2 # hb =.2; wft=.1; wfb=.1; tft=.02; tfb=.02; tw=.02 # S-beam
3 hb =.2; wft=.08; wfb=.08; tft=.015; tfb=.015; tw=.01 # S-beam
4
5 # whole area is asymmetric with respect to y-axis, compute the centroid:
6 cxi = [Link]([(wfb-tw)+wft/2, wfb-tw/2, wfb/2]) # from top to bottom
7 cyi = [Link]([hb-tft/2, (hb-tft-tfb)/2+tfb, tfb/2])
8
9 # call the code function:
10 I_xx, I_yy, I_xy, x_bar, y_bar, xi_bar, yi_bar, A, Ai =\
11 secondMoments_I(hb, wft, wfb, tft, tfb, tw, cxi, cyi, sy=False)

Center of sub-areas [L]:

x = [1.1000e-01 7.5000e-02 4.0000e-02]
y = [1.9250e-01 1.0000e-01 7.5000e-03]
Areas of sub-areas = [1.2000e-03 1.7000e-03 1.2000e-03] [L^2]
Areas of I-beam = 0.0040999999999999995 [L^2]
The centroid = (0.075, 0.1) [L]
New center of sub-areas [L]:
x = [3.5000e-02 0.0000e+00 -3.5000e-02]
y = [9.2500e-02 -1.3878e-17 -9.2500e-02]
2nd moments [L^4]:
Ixx = 2.467416666666666e-05
Iyy = 4.2341666666666675e-06
Ixy = 7.770000000000002e-06

Indeed, the product moment of area of this S-beam is nonzero.

 Property of Areas 433

10.3.7 Second moment of area, coordinate rotation

Consider a new Cartesian coordinate system x = (x , y  ) that is rotated
by an angle θ from the current coordinate x = (x, y) under which all the
second moments of area are known. Suppose (x, y) pass through the centroid
of the whole area. We would like to find the second moment of the same
area with respect to the rotated (x , y  ). The formula can be derived using
the coordinate transformation rule for the coordinate vector discussed in
Chapter 2. The new coordinates are given by Eq. (10.17). Substituting them
into Eq. (10.20) gives
 
2
Ix  x  = y dA = (−x sin θ + y cos θ)2 dA
A A

= Ixx cos θ + Iyy sin2 θ − 2Ixy sin θ cos θ

2
 
2
Iy  y  = x dA = (x cos θ + y sin θ)2 dA
A A (10.25)
= Ixx sin θ + Iyy cos θ + 2Ixy sin θ cos θ
2 2
 
 
Ix  y  = x y dA = (−x sin θ + y cos θ)(x cos θ + y sin θ) dA
A A

= (Ixx − Iyy ) sin θ cos θ + Ixy (cos θ 2 − sin2 θ)

where Ixx , Iyy , and Ixy are the second moments of area with respect to
coordinates (x, y) that are passing through the centroid. Equation (10.25)
is useful when calculating the second moment of an area when coordinates
rotate.
Comparing Eq. (10.25) with the expression for 2D stresses transforma-
tion given in Chapter 3, we found that these are exactly the same if Ixx
corresponds to σxx , Iyy corresponds to σyy , and −Ixy corresponds to σxy . In
other words, if we form a 2D tensor,
 
Ixx −Ixy
(10.26)
−Ixy Iyy
The same coordinate transformation (rotation) rule apply. Therefore, the
double angle formulas for stress transformation given in Eq. (3.74) can also
be used, with the sign flip for Ixy and Ix y , which gives
1 1
Ix x = (Ixx + Iyy ) + (Ixx − Iyy ) cos 2θ − Ixy sin 2θ
2 2
1 1
Iy y = (Ixx + Iyy ) − (Ixx − Iyy ) cos 2θ + Ixy sin 2θ (10.27)
2 2
1
Ix y = (Ixx − Iyy ) sin 2θ + Ixy cos 2θ
2
434 Mechanics of Materials: Formulations and Solutions with Python

Because of the double angle formula, it is clear that if the rotation of

the coordinates is 180◦ , there is no effect on all the second moments. For
example, the x-right, y-up coordinate pair is the same as the x-left and
y-down coordinate pair.

10.3.8 Principal second moments of area

As we did for stress tensor, we can rotate the coordinate to an angle at which
Ix y vanishes. Such an angle is found from the third equation in Eq. (10.27),
which gives

2Ixy
tan 2θ = − (10.28)
Ixx − Iyy

The axis with Ix y = 0 is called the principal axis. The corresponding
second moments of area are called the principal moments of area. This
double angle formula is convenient for calculating the rotation angle needed
to arrive at the principal axis. A list of principal moments of area for various
regular shapes can be found at the wikipage ([Link]
List of second moments of area).

10.3.9 Example: Second moments of area after

coordinate rotation
Using the second moments of area obtained in the previous example, com-
pute the new second moments of area when the coordinates are rotated
by 30◦ .

Solution: We write first the following code function to do this work, which
can be used for future works.

1 def Ixx2IXX(Ixx, Iyy, Ixy, ):

2 '''Coordinate transformation for second moments of area.
3 Ixx, Iyy, Ixy: for original (x,y)
4 IXX, IYY, IXY: for (X,Y) that is -rotated from (x,y)
5 '''
6 c = [Link]( ); s = [Link]( )
7 c2= c**2; s2= s**2; cs = c*s
8 IXX = Ixx*c2 + Iyy*s2 - 2*Ixy*cs
9 IYY = Ixx*s2 + Iyy*c2 + 2*Ixy*cs
10 IXY = (Ixx-Iyy)*cs + Ixy*(c2 - s2)
11
12 return IXX, IYY, IXY
 Property of Areas 435

Find the new second moments of area using Ixx2IXX().

1 #Ixx = 3.936e-05; Iyy = 9.840e-06; Ixy = 1.440e-06 # (m^4)

2 Ixx = I_xx
3 Iyy = I_yy
4 Ixy = I_xy
5
6 = 30.0
7 IXX, IYY, IXY = Ixx2IXX(Ixx, Iyy, Ixy, np.deg2rad( ))
8 print(f"2nd moment of area wrt the principal axis (m^4):\n"
9 f"IXX= {IXX:.4e}, IYY= {IYY:.4e},IXY= {IXY:.4e}")

2nd moment of area wrt the principal axis (m^4):

IXX= 1.2835e-05, IYY= 1.6073e-05,IXY= 1.2736e-05

10.3.10 Example: The principal second moments

of area
Using the second moments of area obtained in the previous example, com-
pute the principal second moments of area and their direction.
Solution:
1 # Compute the angle to principal axis for the 2nd moment:
2 #Ixx = 3.936e-05; Iyy = 9.840e-06; Ixy = 1.440e-06 # (m^4)
3
4 tan2 = -2*Ixy/(Ixx-Iyy)
5 = [Link](tan2 )/2
6 print(f"Angle to the principal axis = { } (rad)")
7 print(f"Angle to the principal axis = { *90/[Link]}∘")

Angle to the principal axis = -0.3250220453336827 (rad)

Angle to the principal axis = -9.311195723164865

1 # Compute the principal 2nd moment of area:

2
3 IXX, IYY, IXY = Ixx2IXX(Ixx, Iyy, Ixy, )
4 print(f"2nd moment of area wrt the principal axis (m^4):\n"
5 f"IXX= {IXX:.4e}, IYY= {IYY:.4e},IXY= {IXY:.4e}")

2nd moment of area wrt the principal axis (m^4):

IXX= 2.7292e-05, IYY= 1.6159e-06,IXY= 0.0000e+00

10.3.11 Use of eigenvalue solver

Although not recommended, we can also use the eigenvalue solver to find the
principal axis for the second moments. However, we need some extra work
436 Mechanics of Materials: Formulations and Solutions with Python

to determine the sign of the angle of the direction cosines. The details are
as follows.
First, form the I tensor, and invoke an eigenvalue solver.

1 #Ixx = 3.936e-05; Iyy = 9.840e-06; Ixy = 1.440e-06 # (m^4)

2
3 I = [Link]([[ Ixx, -Ixy],
4 [-Ixy, Iyy]])
5 e, v = [Link](I)
6 print(f"Eigenvalues of I = {e}") # This gives principal IXX and TYY
7
8 print(f"Eigenvectors of I: \n{v}") # This gives the direction cosines
9 print(f"Principal axis (arccos(x,x') =±{[Link](v[0,0])} (rad)")
10 print(f"Principal axis (arccos(x,y') =±{[Link](v[1,0])} (rad)")

Eigenvalues of I = [2.7292e-05 1.6159e-06]

Eigenvectors of I:
[[9.4764e-01 3.1933e-01]
[-3.1933e-01 9.4764e-01]]
Principal axis (arccos(x,x') =±0.3250220453336826 (rad)
Principal axis (arccos(x,y') =±1.8958183721285793 (rad)

We obtained the same values for the principal second moment of area.
Because the cosine value can correspond to both positive and negative angles,
we need to determine the sign of the angle between x and x . This can be
done by computing IXX , IY Y , and IXY (or just IXY ) from the positive angle
and negative angle using our code function Ixx2IXX(). We should choose the
one that gives IXY = 0. The code is as follows:

1 n1 = [Link](v[0,0]) # arccos(x,x')
2
3 IXX, IYY, IXY = Ixx2IXX(Ixx, Iyy, Ixy, n1) # take positive
4 print(f"2nd moment of area wrt the principal axis (m^4):\n"
5 f"IXX= {IXX:.4e}, IYY= {IYY:.4e},IXY= {IXY:.4e}")
6
7 IXX, IYY, IXY = Ixx2IXX(Ixx, Iyy, Ixy, -n1) # take negative
8 print(f"\n2nd moment of area wrt the principal axis (m^4):\n"
9 f"IXX= {IXX:.4e}, IYY= {IYY:.4e},IXY= {IXY:.4e}")

2nd moment of area wrt the principal axis (m^4):

IXX= 1.7887e-05, IYY= 1.1021e-05,IXY= 1.2371e-05

2nd moment of area wrt the principal axis (m^4):

IXX= 2.7292e-05, IYY= 1.6159e-06,IXY= 8.4703e-22

We shall then choose the negative sign. We can compare the results with
that obtained using Eq. (10.28).
 Property of Areas 437

1 print(f"Principal axis (arccos(x,x') =-{[Link](v[0,0])} (rad)")

2 print(f"Principal axis (arccos(x,x') =-{[Link](v[0,0])*180/[Link]}∘")
3 print(f" 1 to the principal axis={[Link](-2*Ixy/(Ixx-Iyy))/2}(rad)")

Principal axis (arccos(x,x') =-0.3250220453336826 (rad)

Principal axis (arccos(x,x') =-18.622391446329726∘
1 to the principal axis=-0.3250220453336827(rad)

We obtained the same results.

The other principal axis should be orthogonal to θ1 , and we are done. If
interested, readers can also do the same to determine the sign of the second
cosine angle between x and y  :

1 n2 = [Link](v[1,0]) # arccos(x,y')
2
3 IXX, IYY, IXY = Ixx2IXX(Ixx, Iyy, Ixy, n2) # take positive
4 print(f"2nd moment of area wrt the principal axis (m^4):\n"
5 f"IXX= {IXX:.4e}, IYY= {IYY:.4e},IXY= {IXY:.4e}")
6
7 IXX, IYY, IXY = Ixx2IXX(Ixx, Iyy, Ixy, -n2) # take negative
8 print(f"\n2nd moment of area wrt the principal axis (m^4):\n"
9 f"IXX= {IXX:.4e}, IYY= {IYY:.4e},IXY= {IXY:.4e}")

2nd moment of area wrt the principal axis (m^4):

IXX= 1.1021e-05, IYY= 1.7887e-05,IXY= -1.2371e-05

2nd moment of area wrt the principal axis (m^4):

IXX= 1.6159e-06, IYY= 2.7292e-05,IXY= 0.0000e+00

We shall take that negative sign also.

1 print(f"Principal axis (arccos(x,y') =-{n2} (rad)")

2 print(f"Principal axis (arccos(x,y') =-{n2*180/[Link]}∘")

Principal axis (arccos(x,y') =-1.8958183721285793 (rad)

Principal axis (arccos(x,y') =-108.62239144632973

This additional work is because the eigenvalue solver gives only the direc-
tion cosine values. However, the arccos function is multivalued in [−π, π].
Only one of which is the correct answer to our mechanics problem. All these
computations are very fast and takes only milliseconds, but we do need to
figure out how to find the correct one. In contrast, Eq. (10.28) is easier to
use because arctan(x) is single-valued for x in [−∞, ∞].
438 Mechanics of Materials: Formulations and Solutions with Python

10.4 Polar moment of area

We have seen the polar moment of area, J, in Chapter 9 when studying

bars subjected to torsion. It is the stiffness of the bar from the shape and
size of the cross-sectional area. We have derived the formula for compute its
4
value for a circular cross-sectional area: Jz = πR
2 , where R is the radius of
the cross-section. We now present a general approach for computing Jz for
other shapes of areas using the second moments of areas Ixx and Iyy and the
perpendicular axis theorem.

10.4.1 Perpendicular axis theorem

The theorem is expressed as
Jz = Ixx + Iyy (10.29)
Because we know how to compute Ixx and Iyy for a given area, this
theorem gives a straightforward method to compute Jz for the same area.
On the other hand, if we know Jz , it can also help to obtain Ixx and Iyy .

10.4.2 Proof of the perpendicular axis theorem

The proof is straightforward and needs only the use of the Pythagorean
theorem:

Jz = r 2 dA
A

= (x2 + y 2 )dA
A (10.30)
 
= x2 dA + y 2 dA
A A

= Ixx + Iyy
This equation shows also that Jz is an invariant with respect to the
coordinate rotation about the z-axis. This makes sense because the property
of an area in the x–y plane should not be affected by the rotation of the x–y
plane about the z-axis.

10.4.3 Other useful cases

If the area is rotational symmetric, such as circles and hollow circles,
we have
Jz = 2Ixx or Jz = 2Iyy (10.31)
 Property of Areas 439

For an annulus (hollow circles) area with inner radius RI and outer
radius RO , the second polar moment of area J is calculated using
  b
π(RO
4 − R4 )
I
Jz = r dA = 2π
2
r 3 dr = (10.32)
A a 2

Using Eq. (10.31), we have the second moment of area for an annulus
area:
1 π(RO
4 − R4 )
I
Ixx = Iyy = Jz = (10.33)
2 4
For a circular (solid) area with radius R, we have

1 πR4
Ixx = Iyy = Jz = (10.34)
2 4
A list of second moments and polar moments of area of some other shapes
can be found at the wikipage ([Link] of second
moments of area).

10.4.4 Equivalent polar second moment of area

The polar second moment of area is defined as Jz = A r 2 dA. It is directly
used in torsional problems for bars with a circular (solid or hollow) cross-
section and is denoted as J (see Chapter 9). This is because the shear stress
on a circular cross-sectional area is in the polar angle θ direction when the
bar is subjected to torsion. However, for any non-circular bars, Jz cannot
be used, because the shear stress on a non-circular cross-sectional area is
not necessarily in the polar angle θ direction. Instead, the shear stress must
always flow along the edge of the cross-section. Therefore, we need to fol-
low the principles of mechanics to derive a J specifically for a given cross-
section, so that we can still use the same formulas derived for circular bars.
Such a J for a non-circular cross-section is called the **equivalent polar
second moment of area** in this book. It is thus important to distinguish
them. Section 9.5.8 presents a comparison of the equivalent J with the polar
moment of area Jz for the cases of rectangular cross-sections.

10.5 Remarks

This chapter studies the properties of areas related to mechanics of materials.

Area is the most basic property and used in all mechanics problems. The
first and second moments of area are particularly useful in beam members.
For torsional problems, the equivalent polar second moment is used when
440 Mechanics of Materials: Formulations and Solutions with Python

the cross-sectional area is non-circular, which needs to be derived following

the principles of mechanics, as presented in Chapter 9.
For irregular shapes of areas, we do not usually have closed-form for-
mulas for their properties. Because all these properties can be computed via
integration, one can always compute these via numerical integration without
much difficulty.
For standard beams produced in the industry, the cross-sections may
not be regular. Often these properties are provided by the manufactures or
standardized database. Designers can directly make use of those.
Except area, all these properties change when the coordinates transform.
The formulas are essentially the same as those we derived for vectors (first-
order tensors) and stress tensors (second-order tensors). The rule for trans-
formation is provided together with codes.

Reference
[1] G.R. Liu, Calculus: A Practical Course with Python, World Scientific, New Jersey,
2025.
 Chapter 11

Normal Stress by Moments

1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.

2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 init_printing(use_unicode=True) # For latex-like quality printing
11
12 # Digits in print-outs
13 np.set_printoptions(precision=4,suppress=True,
14 formatter={'float_kind': '{:.4e}'.format})

We studied beams undergoing bending in Chapter 8 and developed methods

to compute the moment at any location along the beam axis. This chapter
presents techniques for computing the normal stress and its distribution on a
cross-section of an arbitrary shape resulting from a moment with respect to
an arbitrary direction. This chapter is written with reference to Refs. [1–4].

11.1 Loading plane

This chapter considers beams with arbitrary cross-sectional shapes. We thus
need to set our coordinate system carefully. When a beam is subjected to
transverse forces (concentrated or distributed), the direction of these forces

441
442 Mechanics of Materials: Formulations and Solutions with Python

Figure 11.1. A beam subjected to forces and moments, resulting in a moment on its
cross-section. When a force is applied in the loading plane, it results in a moment without
torque. The moment results in normal stress σzz distributed on the cross-section.

are perpendicular to the beam’s axial z-axis, which results in moments, as

shown in Fig. 11.1. The z-axis is chosen, passing through the centroids of
all the cross-sections along the beam. Using the technique and code given
in Chapter 10, the centroid of a given cross-section area can be found with
ease.
The location of the force in relation to the cross-section of the beam can,
however, also result in a torque on the cross-section, and the torque vector
is in the z-axis. Only when these forces pass through the so-called shear
center on the cross-section no torque is introduced. The location of the
shear center will be discussed in the following chapter. For now, we assume
that it is known. We thus have an axis called shear-center axis that goes
through the shear centers of all the cross-sections of the beam. The shear-
center axis is also called the bending axis. It is parallel with the centroid
axis. All these little complicated positional relations are shown in the busy
drawing Fig. 11.1 [1].
This chapter assumes that the transverse forces pass through the shear
center. The loading plane is a plane formed by the shear-center axis and the
load force vector PA , as shown in Fig. 11.1. We only consider forces that are
in this plane and perpendicular to the z-axis, such as PA , so that the cross-
section of the beam experiences moments without any torque. If a moment is
applied, the loading plane is perpendicular to the moment vector, as shown
also in Fig. 11.1. Essentially, the loading plane determines the direction of
the moment applied on the cross-section resulting from a transverse force.
The loading plane is measured by ϕ counterclockwise from the x–z plane
as viewed from the positive z axis.
 Normal Stress by Moments 443

11.2 Setting of the problem

Consider a cross-section B of the beam at distance z from the left end A.
Assume there are forces and/or moments at the left end A. The equilibrium
of the free-body diagram shall lead to a moment M on section z. The moment
vector M can always be decomposed into two components along the x- and
y-axis, using the direction cosines:
Mx = M cos(ϕ − 90◦ ) = M sin ϕ
(11.1)
My = M cos(180◦ − ϕ) = −M cos ϕ
where ϕ defines the direction of the loading plane and hence the direction of
the moment with respect to the x-axis. Our problem now becomes finding
the normal stress and its distribution on the cross-section for given two
orthogonal moment components Mx and My on the cross-section.

11.3 Symmetrical bending

11.3.1 Superposition of stresses

Consider first simpler cases where the x- and y-axis are the principal axes
for the second moments of the area of the beam’s cross-section, and the
origin is at the centroid. In this case, the bending by each component of
the moment gives a linearly distributed stress that is zero at the neutral axis
that coincides with a principal axis. The final stress results and distributions
are a superposition of these two stress distributions. A typical case is shown
in Fig. 11.2.
For the problem given in Fig. 11.2, assume that the beam has a rectan-
gular cross-section with base b and height h. The formulas for calculating
the normal stress in the z-direction are as follows. For moment component
Mx ,
Mx y
σzz = (11.2)
Ixx
where Ixx is the second moment of area of the rectangular cross-section,
3
which is bh12 . The normal stress varies linearly with respect to y and is
constant along x. On the line of y = 0, the stress is zero, implying that the
y = 0 line is the neutral axis.
Similarly, for moment component My , we have
My x
σzz = − (11.3)
Iyy
3
where Iyy is the second moment of area of the rectangular cross-section: hb
12 .
The normal stress varies linearly with respect to x and is constant along y,
444 Mechanics of Materials: Formulations and Solutions with Python

Figure 11.2. Stress distributions on the cross-section of a beam with a symmetric cross-
section by two orthogonal moment components and the superposed stress distribution.
Images adapted from the version by Cdang on Wikipedia under the CC BY-SA 3.0 license.

and the x = 0 line is the neutral axis, on which the stress are zero. The minus
sign is because positive My results in negative normal stress (at a positive y
in the positive cross-section), as shown in Fig. 11.1 or also in Fig. 11.2.
The superposed stress becomes

Mx y My x
σzz = − (11.4)
Ixx Iyy

which is a plane varying linearly with respect to both x and y, similar to the
case shown in the bottom-right corner of Fig. 11.2.
The maximum stress occurs at y = ±h/2 and x = ±b/2. Depending on
the sign of Mx and My , two of the four vertexes of the rectangular cross-
section will register the maximum stresses, one tension and one compression,
similar to the case shown in the bottom-right corner of Fig. 11.2.
 Normal Stress by Moments 445

If we integrate the stress over the entire cross-section, we obtain

  
My Mx
σzz dA = xdA − ydA = 0 (11.5)
A Iyy A Ixx A
     
Qy =0 Qx =0

These first moments Qx and Qy are zero because the origin of the coordi-
nates (x, y) is set at the centroid of the cross-section. In essence, the integral
of the stress in the z-direction vanishes because all the forces applied on the
beam are perpendicular to the z-axis.

11.3.2 Neutral axis of the combined normal stress

Because the distribution of σzz is a plane in the (x, y) coordinates on the
cross-section and the integral of it is zero, there must be a line on the cross-
section along which σzz = 0. Such a line is called the neutral axis. It can
be easily found by setting σzz = 0 in Eq. (11.4), and the function for the
neutral axis can be written as
My Ixx
y= x = x tan α (11.6)
Mx Iyy
  
tan α

where tan α is the slope of the neutral axis, in which α is the angle measured
from the x-axis, as shown in Fig. 11.1. It is computed using
My Ixx
tan α = (11.7)
Mx Iyy
The blue line on the cross-section at the bottom-right corner of Fig. 11.2
is the neutral axis for the combined stress distribution.

11.3.3 Example: A cantilever beam subjected to a force

and a moment
Consider a cantilever of length L shown in Fig. 11.3. Its cross-section is
rectangular with base b and height h. It is clamped at its left end and sub-
jected to a force PA and a moment MA at its right end marked with A. The
force PA passes through the shear center and hence no torque is generated.
Assuming MA = PA3 L , derive the formula to compute the maximum stress
on the beam.
Solution: First, we need to find out where along the beam axis the moment
is the largest. Here, we use the usual method: free-body diagram and equi-
librium conditions.
446 Mechanics of Materials: Formulations and Solutions with Python

Figure 11.3. A cantilever beam subjected to a force and a moment at its free end A.

Figure 11.4. A free-body diagram of the cross-section at x for a cantilever beam sub-
jected to a force and a moment.

Consider an arbitrary point x on the beam axis. We draw a free-body

diagram in Fig. 11.4.
We put the moment M and shear force V on the cross-section at x,
following the sign convention. Because the direction of the cross-section is
leftward (the negative x-direction), the moment must be clockwise,
 and the
shear force must point downward. The equilibrium equation of Mz = 0
with respect to the point at x gives

−M + MA + PA (L − x) = 0 (11.8)
PA L
Using the given condition MA = 3 , we have

4PA L
M = MA + PA (L − x) = − PA x (11.9)
3
This is a linear function of x. Its maximum can only be at its two bound-
aries: x = 0 or x = L. Clearly, the maximum moment is at x = 0 and is
given as
4PA L
Mmax = (11.10)
3
 Normal Stress by Moments 447

The stress distribution on the cross-section by the maximum moment

becomes
Mmax y 4PA Ly
σxx = − =− (11.11)
Izz 3Izz
This formula works for any cross-sectional shape, as long as it is symmet-
3
ric with respect to x = 0. For a rectangular cross-section, we have Izz = bh
12 ,
and the stress solution becomes
4PA Ly 16PA Ly
σxx = − =− (11.12)
3Izz bh3
It varies linearly along y. The maximum stresses are at both the top and
bottom surfaces, y = ±h/2, of the beam at x = 0:

8PA L
σxx = ± (11.13)
bh2
Note that the stress is inversely proportional to h2 . Therefore, it is much
more effective to reduce the stress by increasing the height of the beam
compared to increasing the base b for a given load. Ideally, we want to design
beams with as large a height as possible.

11.3.4 Use gr.solver1D4() to find the moment

We can also simply use our general solver gr.solver1D4() for beams developed
in Chapter 8 to find the moments.

1 E, Izz, A, b, h, L = symbols('E, I_zz, A, b, h, L', nonnegative=True)

2 x, y, PA, MA = symbols('x,y, P_A, M_A') # geometry & force
3
4 # for displacement boundary conditions (DBCs):
5 v0, 0, vl, l = symbols('v0, 0, v_l, _l') # DBCs
6
7 # for force boundary conditions (FBCs):
8 V0, M0, Vl, Ml = symbols('V0, M0, V_l, M_A') # FBCs
9
10 by = [Link]('b_y')(x) # for force function

1 # Consider constant distributed force

2 by = 0 # [F/L], this is zero for this example problem
3
4 v = gr.solver1D4(E, Izz, by, L, v0, 0, vl, l, V0, M0, Vl, Ml,key='c-f')
5 print(f'Is solution correct? {by/E/Izz==v[4]}; The solution u(x) is:')
6 v[0] # solution of general displacement function
448 Mechanics of Materials: Formulations and Solutions with Python

Outputs form solver1D4(): v, , M, V, qy

Is solution correct? True; The solution u(x) is

LVl x2 MA x2 Vl x3
v0 + xθ0 + + −
2EIzz 2EIzz 6EIzz

Use now all the prescribed displacements and moments given in this
example problem:

1 # Set all the pre-scribed conditions:

2
3 vx=[v[i].subs({v0:0, 0:0, Vl:PA, Ml:MA}) for i in range(len(v))]
4 vx # Outputs are: v, , Mx, Vx, qy


LPA x2 MA x2 PA x3 LPA x MA x PA x2
+ − , + − ,
2EIzz 2EIzz 6EIzz EIzz EIzz 2EIzz

LPA + MA − PA x, PA , 0

This gives the full set of solutions. For our example problem, we need
only the moment.

1 # Get only the solution for the moment:

2 [Link](vx[2], 'Solution of moment distribution along the beam:')

Solution of moment distribution along the beam: LPA + MA − PA x

1 # Substitute the M_A with known value in terms of P_A:

2 [Link](vx[2].subs({MA:PA*L/3}),'Solution of moment along the beam:')

4LPA
Solution of moment along the beam: − PA x
3

1 # Set x=0, for the maximum moment:

2 Mmax = vx[2].subs({MA:PA*L/3, x:0})
3 [Link](Mmax,'Solution at the beam root, Mmax:')

Solution at the beam root, Mmax:

4LPA
3

We have obtained exactly the same solution as we did using the free-
body diagram with equilibrium conditions. Our general solver gr.solver1D4()
is powerful. It works for not only this problem but also many other types
 Normal Stress by Moments 449

of problems for beams. If the problem is indeterminate, using a free-body

diagram alone will not work. In such cases, our general solver gr.solver1D4()
must be used. Note also that gr.solver1D4() gives a full set of solutions,
including deflection, rotation, moment, and shear force! All are in closed-
form formulas. The computation took only a few seconds on the author’s
laptop.
The formulas for the normal stress are finally found using:

1 xx = - Mmax*y/Izz
2 print(' xx=', xx)
3 print(' xx (in b and h) = ', [Link](Izz, b*h**3/12))
4 print(' xx (at upper surface) = ', [Link]({Izz:b*h**3/12, y:h/2}))

xx= -4*L*P_A*y/(3*I_zz)
xx (in b and h) = -16*L*P_A*y/(b*h**3)
xx (at upper surface) = -8*L*P_A/(b*h**2)

11.3.5 Example: A cantilever T-beam subjected to a

force and a moment
Consider again the cantilever of length L shown in Fig. 11.3. All the settings
are the same, except for its cross-section becoming T shape, as shown in
Fig. 11.5.

Figure 11.5. Schematic drawing of the cross-section of a typical T-beams.

450 Mechanics of Materials: Formulations and Solutions with Python

Its dimensions (mm) are given as follows:

beam height: hb = 200; width of flange on top: wf = 100; thickness of

flange on top: tf = 20; web thickness: tw = 10.

1. Find the centroid of the cross-section.

2. Find the second moments of the cross-section.
3. Determine the maximum stress of the beam. Assume PA = 5000 N and
L = 600 mm.

Solution to 1: To find the centroid of the cross-section of the T-beam, we

use the Python code function [Link] I() developed in Chapter 10, which
is also in the grcodes module. The [Link] I() is made for I-beams con-
sisting of three rectangles. The T-beam cross-section has only two rectangles
and can be treated as a special case of the I-beam by setting the thickness
of the flange at the bottom zero.

1 # preparing data: Data from Example 7.2 AMM by Boresi and Schmidt
2 #hb =.060; wft=.050; wfb=.010; tft=.010; tfb=.0; tw=.010 # (m) √
3 # This example:
4 hb =.200; wft=.100; wfb=.010; tft=.020; tfb=.0; tw=.010 # (m)
5
6 # call the code function:
7 x_bar, y_bar, A, Ai = gr.centroid_I(hb, wft, wfb, tft, tfb, tw)

Areas of sub-areas = [2.0000e-03 1.8000e-03 0.0000e+00][L^2]

Areas of I-beam = 0.0038000000000000004[L^2])
The centroid [L]):
x = 0.005
y = 0.14263157894736841

Solution to 2: Find the second moments of the cross-section.

This time we use the Python code function [Link] I() devel-
oped in Chapter 10. It also computes the centroid while computing the sec-
ond moments. A T-beam can be treated as a special case of the I-beam with
zero thickness for the bottom flange.

1 # call the code function for 2nd moment calculation:

2 I_xx, I_yy, I_xy, x_bar, y_bar, xi_bar, yi_bar, A, Ai =\
3 gr.secondMoments_I(hb, wft, wfb, tft, tfb, tw, [0]*2, [0]*2, sy=True)
 Normal Stress by Moments 451

Center of sub-areas [L]:

x = [5.0000e-03 5.0000e-03 5.0000e-03]
y = [1.9000e-01 9.0000e-02 0.0000e+00]
Areas of sub-areas = [2.0000e-03 1.8000e-03 0.0000e+00] [L^2]
Areas of I-beam = 0.0038000000000000004 [L^2]
The centroid = (0.005, 0.14263157894736841) [L]
New center of sub-areas [L]:
x = [0.0000e+00 0.0000e+00 0.0000e+00]
y = [4.7368e-02 -5.2632e-02 -1.4263e-01]
2nd moments [L^4]:
Ixx = 1.4400350877192984e-05
Iyy = 1.6816666666666672e-06
Ixy = 0.0

One needs to ignore the third entry that corresponds to sub-areas with
A = 0. The centroid of the T-beam is also found as x bar, y bar. In addition,
the new centroids of the two sub-rectangles are also found.

Solution to 3: Determine the maximum stress of the beam. Assume PA =

5000 N and L = 600 mm.
We can use Eq. (11.11) derived in the previous example, but we need
to do a little more work to compute the distance from the centroid of the
T-beam cross-section and find the y values at the top and bottom of the
cross-section.

1 y_top = hb - y_bar # y_bar: the centroid of the T-beam

2 y_bot = - y_bar
3 print(f" y_top= {y_top}, y_bot ={y_bot}")
4
5 PA = 5000. # (N)
6 L = 0.600 # (m)
7 Mmax = 4*PA*L/3 # from previous example
8 print(f" Mmax = {Mmax} (Nm)")
9
10 _top = Mmax*y_top /I_xx
11 _bot = Mmax*y_bot /I_xx
12
13 print(f" _top= { _top*1e-6}, _bot ={ _bot*1e-6} (MPa)")

y_top= 0.057368421052631596, y_bot =-0.14263157894736841

Mmax = 4000.0 (Nm)
_top= 15.935284227967157, _bot =-39.61891766769815 (MPa)

It is found that the largest normal stress is located at the bottom edge
of the T-beam.
452 Mechanics of Materials: Formulations and Solutions with Python

11.4 Beams under non-symmetrical bending*

In practical applications, beams with a non-symmetrical cross-section

are used. In addition, the loading plane or the moment on the cross-section
can be arbitrary. We need to find the normal stress on the cross-section and
often the maximum one.
A general procedure is introduced in this section to compute the nor-
mal stresses on the non-symmetrical cross-section of a beam, subjected to
moments in any given direction.

11.4.1 General formula for computing normal stress

distribution
On the cross-section, the coordinates (x, y) are set with the centroid of the
cross-section at the origin. Based on our discussion so far, we know that
the stress σzz is linearly distributed on the cross-section defined in (x, y)
coordinates, as shown in Fig. 11.1. This is true for all beams that obey the
Euler–Bernoulli assumption. Thus, the normal stress can be written in the
following general form:

σzz = a + bx + cy (11.14)

where a, b, and c are three constants to be determined.

Now, on the cross-section, we shall have the following three equilibrium
conditions:

0= σzz dA
A

Mx = yσzz dA (11.15)
A

My = − xσzz dA
A

This minus sign is because positive My results in negative σzz at positive

y on the positive cross-section (see Fig. 11.1).
 Normal Stress by Moments 453

Using these three conditions in Eq. (11.15), we can determine the three
constants a, b, and c:
   
0= (a + bx + cy)dA = a dA +b xdA +c ydA
A A A A
        
A =0 =0
   
2
Mx = (ay + bxy + cy )dA = a ydA +b xydA +c y 2 dA
A A A A
        
=0 Ixy Ixx
   
My = − (ax + bx2 + cxy)dA = −a xdA −b x2 dA −c xydA
A A A A
        
=0 Iyy Ixy
(11.16)

We used the fact that first moments of the cross-section area are zero
because (x, y) is set at the centroid. The foregoing equation simply becomes

0 = aA
Mx = bIxy + cIxx (11.17)
My = −bIyy − cIxy

The first equation in Eq. (11.17) gives a = 0 because A = 0. The last two
equations can be easily solved using the following code.

1 Mx, My, Ixx, Iyy, Ixy, b, c = symbols('M_x, M_y, I_xx, I_yy, I_xy, b,c')
2 sln = [Link]([b*Ixy+c*Ixx-Mx, -b*Iyy-c*Ixy-My], [b, c])
3 sln

−Ixx My − Ixy Mx Ixy My + Iyy Mx

b: 2
, c: 2
Ixx Iyy − Ixy Ixx Iyy − Ixy

We write down the results for later reference:

My Ixx + Mx Ixy
b=− 2
Ixx Iyy − Ixy
(11.18)
Mx Iyy + My Ixy
c= 2
Ixx Iyy − Ixy
454 Mechanics of Materials: Formulations and Solutions with Python

Substituting Eq. (11.18) and a = 0 back into Eq. (11.14), we obtain

My Ixx + Mx Ixy Mx Iyy + My Ixy
σzz = bx + cy = − 2
x+ 2
y (11.19)
Ixx Iyy − Ixy Ixx Iyy − Ixy

This is the general formula to compute the normal stress distributed on

the cross-section of a beam that undergoes non-symmetric bending. It is
a linear function of x and y. If Ixy = 0, the bending becomes symmetric,
and Eq. (11.19) becomes Eq. (11.4). This means that symmetric bending
occurs only when Ixy = 0.

11.4.2 Neutral axis: general formula

The neutral axis is the line on the cross-section on which σzz = 0. Using
Eq. (11.19), the function for the neutral axis can be written as
My Ixx + Mx Ixy
y= x = x tan α (11.20)
Mx Iyy + My Ixy
  
tan α

where tan α is the slope of the linear function and α is the angle that deter-
mines the neutral axis. It is measured from the x-axis, as shown in Fig. 11.1.
It is computed using
My Ixx + Mx Ixy
tan α = (11.21)
Mx Iyy + My Ixy
It is clear that the slope of the neutral axis depends on both the property
of the cross-section and the moment direction. Using Eq. (11.1), we have
My
cot ϕ = − (11.22)
Mx
Substituting Eq. (11.22) into Eq. (11.21), we have, alternatively,
Ixy − Ixx cot ϕ
tan α = (11.23)
Iyy − Ixy cot ϕ
Once α is found, the stress can also be written alternatively as
Mx (y − x tan α)
σzz = (11.24)
Ixx − Ixy tan α
It is clear that the normal stress depends on both the moment M and its
direction and the second moments of area of the cross-section of the beam,
which is an area property depending also on orientation. Its distribution is
 Normal Stress by Moments 455

a linear function of x and y, and hence the maximum stress should register
on the extreme points that are the farthest from the neutral axis.

11.4.3 Example: A cantilever L-beam subjected to

a force and a moment
Consider again the cantilever of length L shown in Fig. 11.3. All the settings
are the same, except for its cross-section becoming an L shape that is not
symmetric, as shown in Fig. 11.6.
Its dimensions (mm) are given as follows:

beam height: hb = 100; width of foot flange: wf = 70; thickness of foot

flange: tf = 10; web thickness: tw = 10.

Assume PA = 5000 N, L = 600 mm, and the angle of the loading plane
is ϕ = 2π/3.

1. Find the centroid of the cross-section.

2. Find the second moments of the cross-section.
3. Determine the maximum stress of the beam using the formulas for non-
symmetric bending.
4. Find the principal axis of the cross-section.
5. Determine the maximum stress of the beam using the formulas for sym-
metric bending.

Figure 11.6. Schematic drawing of the cross-section of a typical L-beams.

456 Mechanics of Materials: Formulations and Solutions with Python

[Link] Solution to 1 and 2

We use the Python code function [Link] I() developed in Chap-

ter 10 to compute the centroids and second moments.

1 # preparing data. Data from Example 7.4 AMM by Boresi and Schmidt
2 #hb =.120; wfb=.080; tfb=.010; tw=.010; wft=.0; tft=.0 # (m) √
3
4 # This example:
5 hb =.100; wfb=.070; tfb=.010; tw=.010; wft=.0; tft=.0; # (m)
6
7 # L-shape area is not symmetric with respect to y-axis, we need to
8 # provide the center of each sub-area of the L-beam
9 cxi = [Link]([wft/2, tw/2, wfb/2]) # from top to bottom
10 cyi = [Link]([hb, (hb-tft-tfb)/2+tfb, tfb/2])
11
12 # call the code function:
13 Ixx, Iyy, Ixy, xbar, ybar, xibar, yibar, A, Ai =\
14 gr.secondMoments_I(hb, wft, wfb, tft, tfb, tw, cxi, cyi, sy=False)

Center of sub-areas [L]:

x = [0.0000e+00 5.0000e-03 3.5000e-02]
y = [1.0000e-01 5.5000e-02 5.0000e-03]
Areas of sub-areas = [0.0000e+00 9.0000e-04 7.0000e-04] [L^2]
Areas of I-beam = 0.0016000000000000003 [L^2]
The centroid = (0.018125000000000002, 0.033125) [L]
New center of sub-areas [L]:
x = [-1.8125e-02 -1.3125e-02 1.6875e-02]
y = [6.6875e-02 2.1875e-02 -2.8125e-02]
2nd moments [L^4]:
Ixx = 1.597708333333334e-06
Iyy = 6.477083333333335e-07
Ixy = -5.906250000000002e-07

One needs to ignore results corresponding to sub-areas with A = 0.

[Link] Solution to 3: Determine the maximum stress of the beam,

using the formulas for non-symmetric bending

Because the cross-section is not symmetric, we must first find the neutral
axis using Eq. (11.21).
 Normal Stress by Moments 457

1 # Mmax=4.8e3; Data from Example 7.4 AMM by Boresi and Schmidt for test √
2
3 PA = 5000. # (N)
4 L = 0.600 # (m)
5 Mmax = 4*PA*L/3
6
7 = 2*[Link]/3
8 Mx = Mmax*[Link]( )
9 My = -Mmax*[Link]( )
10
11 tana = (My*Ixx + Mx*Ixy)/(Mx*Iyy + My*Ixy)
12 = [Link](tana)
13
14 print(f" tana= {tana:.4f}, neutral axis = { :.4f} (rad),"\
15 f"{np.rad2deg( ):.4f}∘")

tana= 1.0818, neutral axis = 0.8247 (rad),47.2512∘

Using the computed tana, the neutral axis is computed using Eq. (11.20).
The stress is given in Eq. (11.24). Points A and B are the possible points
having maximum stresses. To find the stress there, we need to find their
coordinates in terms of (x, y).

1 #xA = wfb-xbar; yA = -ybar # other possibilities

2 #xB = -xbar; yB = hb-ybar
3
4 xA = -wfb+xbar; yA = ybar
5 xB = xbar; yB = -hb+ybar
6
7 dm = Ixx - Ixy*tana
8 zz_A= Mx*(yA - xA*tana)/dm
9 zz_B= Mx*(yB - xB*tana)/dm
10
11 print(f" zz_A= { zz_A*1e-6:.4f}, zz_B= { zz_B*1e-6:.4f} (MPa)")

zz_A= 138.2213, zz_B= -133.9434 (MPa)

The maximum normal stress is found at point A, the lower-right vertex

of the L-beam.

[Link] Solution to 4: Find the principal axis of the cross-section

To use the symmetric bending beam formulas for a non-symmetric beam,
the principal axes X−Y of the beam cross-section should be found first
458 Mechanics of Materials: Formulations and Solutions with Python

(see Fig. 11.6). For this, we use tan 2θ1 = −2Ixy /(Ixx − Iyy ) to find the angle
of rotation θ1 between X and x, and then compute the new second moments
with respect to the principal axis. These can be done using the code function
gr.Ixx2IXX() given in Chapter 10.

1 tan2 = -2*Ixy/(Ixx-Iyy)
2 = [Link](tan2 )/2
3 print(f"Angle to the principal axis 1={ } (rad); {np.rad2deg( )}∘")
4
5 IXX, IYY, IXY = gr.Ixx2IXX(Ixx, Iyy, Ixy, )
6 print(f"2nd moment of area wrt the principal axis (m^4):\n"
7 f"IXX= {IXX:.4e}, IYY= {IYY:.4e},IXY= {IXY:.4e}")

Angle to the principal axis 1=0.4467398651270534 (rad); 25.596308812023786∘

2nd moment of area wrt the principal axis (m^4):
IXX= 1.8806e-06, IYY= 3.6478e-07,IXY= -5.2940e-23

We see that IXY = 0, and hence the formulas for symmetric bending can
be used. Note that after the coordinate rotation, the loading plane angle ϕ
needs to be updated.

1 = -
2 print(f" Angle of the loading plane wrt (X, Y) = { :.4f}, "\
3 f"{np.rad2deg( ):.4f}∘")
4
5 MX = Mmax*[Link]( ) # compute the moment components on (X, Y)
6 MY = -Mmax*[Link]( )
7
8 tanaX = MY*IXX /(MX*IYY)
9 X = [Link](tanaX)
10
11 print(f" tana= {tanaX:.4f}, neutral axis X = { X:.4f} (rad) or "\
12 f"{np.rad2deg( X):.4f}∘")

Angle of the loading plane wrt (X, Y) = 1.6477, 94.4037

tana= 0.3970, neutral axis X = 0.3779 (rad) or 21.6549

Using tanaX, the neutral axis is given by Eq. (11.6). The stress is given in
Eq. (11.4). We need to find the coordinates of points A and B with respect to
the principal (X, Y ). This is done through coordinate transformation studied
in Chapter 2.
 Normal Stress by Moments 459

1 XA = xA*[Link]( ) + yA*[Link]( )
2 YA = -xA*[Link]( ) + yA*[Link]( )
3 print(f" XA= {XA*1e3:.6f}, YA= {YA*1e3:.6f} (mm)")
4
5 XB = xB*[Link]( ) + yB*[Link]( ) # coordinate transformation
6 YB = -xB*[Link]( ) + yB*[Link]( ) # for position vectors
7 print(f" XB= {XB*1e3:.6f}, YB= {YB*1e3:.6f} (mm)")
8
9 zz_AX= MX*YA/IXX - MY*XA/IYY
10 zz_BX= MX*YB/IXX - MY*XB/IYY
11
12 print(f" zz_AX= { zz_AX*1e-6:.4f}, zz_BX= { zz_BX*1e-6:.4f} (MPa)")

XA= -32.473090, YA= 52.285559 (mm)

XB= -12.545630, YB= -68.142413 (mm)
zz_AX= 138.2213, zz_BX= -133.9434 (MPa)

The maximum stress results are exactly the same as those found in item 3.
As an exercise, readers may compute the following:
1. The stresses and their distributions if the L-beam is upside down.
2. Find the neutral axis.

11.5 Effects of error in loading plane angle*

We know that for better performance in terms of higher stiffness and lower
normal stress for the same moment, we would like to design the beam in
such a way that the loading plane is normal to the principal axis that has
the larger second moment of area. For example, if the loading plane is vertical
(its normal is in the x-direction), we would want the beam as tall as possible,
leading to a large Ixx . However, if it is too tall, the performance can be
unstable, implying that it may be sensitive to small errors in the loading
plane angle ϕ. Let us use the following example to demonstrate this effect.

11.5.1 Example: A tall beam with rectangular

cross-section
Consider a beam with rectangular cross-section. It has a height h that is
10 times its base b = 50 mm. The cross-section is defined in (x, y), with x
leftward and y downward. It is designed for transverse loads in the loading
plane of x = 0 and perpendicular to the z-axis (out of the paper), implying
that the angle of the loading plane is ϕ = π/2. Assume the moment created
by a transverse force P is M = 2000 Nm at ϕ = π/2, as shown in Fig. 11.7.
460 Mechanics of Materials: Formulations and Solutions with Python

Figure 11.7. A tall rectangular beam designed for a moment with ϕ = π/2 and a small
error in ϕ in the actual implementation.

Perform this following analysis:

1. Compute the stress σzz at point A (b/2, h/2).

2. Assuming in a service situation, the value of angle ϕ of the loading plane
is reduced by 1% due to error, determine the angle of the neutral axis
resulting from the error.
3. Determine the percentage error in stress σzz at point A.

[Link] Python code for the solution

We write the follow code to complete the task, with proper comment lines.

1 # Compute the properties of the cross-section:

2 b = 0.05 # mm
3 h = 10*b
4
5 Ixx = b*h**3/12
6 Iyy = h*b**3/12
7
8 x_A = b/2; y_A = h/2 # coordinates at point A
9 print(f" x_A= {x_A}, y_A ={y_A}")

x_A= 0.025, y_A =0.25

The following code computes and prints out the requested results for all
these questions. The results for both the perfect and error cases are placed
in an array and computed in one go. The perfect case is for reference.
 Normal Stress by Moments 461

1 M = 2000. # (Nm)
2 0 = [Link]/2 # angle of loading plane by design
3 coef = [Link]([1.0, 0.99]) # reduction rates: perfect, 1% reduction
4
5 = coef* 0 # angles (rad) perfect and 1% reduction
6 cot = [Link]( )/[Link]( )
7
8 Mx = M*[Link]( ) # components of the moment
9 My = -M*[Link]( )
10 print(f"Components of moment:\n Mx= {Mx} (Nm)\n My= {My} (Nm)")
11
12 tana = My*Ixx/(Mx*Iyy)
13 = [Link](tana) # angle of the neutral axis
14 print(f"Angles of the neutral axis:\n = { } (rad) \n"\
15 f" {np.rad2deg( )}∘")
16
17 zz_A0=Mx[0]*y_A/Ixx # when no error in 0
18 print(f"Normal stress at point A (no error), zz_A0={ zz_A0*1e-6}(MPa)")
19
20 zz_A =Mx*y_A/Ixx - My*x_A/Iyy
21 print(f"Normal stress at point A, zz_A= { zz_A*1e-6} (MPa)")
22
23 d R = ( zz_A- zz_A0)/ zz_A0*100
24 print(f"Percentage error in stress d R = {d R} (%)")
25
26 d R = ( 0 - )/ 0*100
27 print(f"Percentage error in angle d R = {d R} (%)")

Components of moment:
Mx= [2.0000e+03 1.9998e+03] (Nm)
My= [-1.2246e-13 -3.1415e+01] (Nm)
Angles of the neutral axis:
= [-6.1232e-15 -1.0039e+00] (rad)
[-3.5084e-13 -5.7520e+01]∘
Normal stress at point A (no error), zz_A0=0.96(MPa)
Normal stress at point A, zz_A= [9.6000e-01 1.1107e+00] (MPa)
Percentage error in stress d R = [6.0633e-14 1.5695e+01] (%)
Percentage error in angle d R = [0.0000e+00 1.0000e+00] (%)

[Link] An important finding

Although the angle error of the loading plane is only 1%, the resultant error
in the neutral axis is significant, and the error in the normal stress is ∼15.7%
for this example. There is about 16 times magnification in error. If the h/b
ratio becomes higher, the magnification will be even bigger. Readers may
test this using the code given above.

11.5.2 Underlying reasons for error magnification

Let us examine in more detail the underlying reasons for the error magnifi-
cation.
462 Mechanics of Materials: Formulations and Solutions with Python

Assume (x, y) is the principal axis, and hence Ixy = 0. Equation (11.23)
becomes

−Ixx
tan α = cot ϕ (11.25)
Iyy

Substituting Eq. (11.25) into Eq. (11.24) and using Ixy = 0, we have

Mx y Mx y xIxx
σzz = + cot ϕ (11.26)
Ixx Ixx yIyy
  
βs

The second term on the right-hand side of Eq. (11.26) is the increased
part of the normal stress, in which βs is the magnification factor (per cot ϕ)
for the increased part of the normal stresses:

xIxx
βs = (11.27)
yIyy

Note that at ϕ close to but less than π/2, cot ϕ ≈ ϕ − π/2.

Equation (11.26) depends on the property of the beam cross-section and
the interested point location on it. If the maximum stress is concerned, say
at point A, x is replaced by xA and y is replaced by yA , and we obtain

xA Ixx
βs = (11.28)
yA Iyy

For beams with rectangular cross-section of height h and base b, we have

b 2 bh3 12 h
βs = 3
= (11.29)
2 h 12 hb b

The stress magnification factor is proportional to the height–base ratio

of a rectangular beam. If h = b, there is no magnification.
This study shows that when the beam is made too high, care is needed
in actual implementations. Almost no error is allowed in the loading plane
angle.
Using the code given above, readers may set different ratios of h/b to see
the results. One may also set h = b to see what the code produces.
For other shapes of cross-section, βs will be different from that given in
Eq. (11.29) and can be quite complicated to derive.
 Normal Stress by Moments 463

11.6 Remarks
1. This chapter focuses on computing the normal stresses on the cross-
section of a beam for a given moment. The beam can have an irregular
cross-section shape, and the loading can be in any direction as long as it
is in the loading plane.
2. The moment can be easily computed using codes given in Chapter 8. If it
is statically determinant, the moment can be obtained using a free-body
diagram and equilibrium conditions.
3. The computation of the normal stress can be done in two ways: using
either the asymmetric bending formula or the symmetric bending for-
mulas. The latter requires finding the principal axis of the cross-section
followed by a coordinate transformation.
4. An analysis is provided to demonstrate that although it is beneficial to
have tall beams, but there is a limit because of stability and robustness
issues.

References
[1] A.P. Boresi and R.J. Schmidt, Advanced Mechanics of Materials. John Wiley & Sons,
New York, 2002.
[2] J.M. Gere and S.P. Timoshenko, Mechanics of Materials. Van Nostrand Reinhold
Company, New York, 1972.
[3] S. Timoshenko and J.N. Goodier, Theory of Elasticity. McGraw-Hill, New York,
1970. [Link]
elasticity&#38;ei=ICiKSsr3G4jwkQSbxMyPCg.
[4] Z.L. Xu, Elasticity, Vols. 1&2. People’s Publisher, China, 1979.
This page intentionally left blank
 Chapter 12

Shear Stress by Shear Force

Let us import the necessary external modules or dependencies for later use
in this chapter.

1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.

2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from functools import partial # create function with single variable
10
11 from continuum_mechanics import vector
12 init_printing(use_unicode=True) # For latex-like quality printing
13
14 # Digits in print-outs
15 np.set_printoptions(precision=4,suppress=True,
16 formatter={'float_kind': '{:.4e}'.format})

When a beam is subjected to transverse loading, there will be shear force

on its cross-section, in addition to the moments. Both moments and shear
forces on the cross-section at any location along the beam axis can be com-
puted using the techniques developed in Chapter 8, and gr.solver1D4() for
various types of loaded thin beams. In the previous chapter, we found the
normal stress on the cross-section caused by the moment.
This chapter zooms-in on a cross-section of a beam and examines the
shear stress distribution on the cross-section resulting from the shear force.
The concept of shear-center and techniques to find it will also be studied. It
is written with reference to [1–4].

465
466 Mechanics of Materials: Formulations and Solutions with Python

12.1 Setting of the problem

Figure 12.1 shows a cantilever beam subjected to a transverse point force

P at its right-end, resulting in both moment and shear force on a cross-
section at any location along the beam. The x-axis starts at the fixed end of
the beam, and goes through the neutral axis of the cross-sections where the
normal stress is zero.
For convenience in analysis, the transverse force is along the y-axis and
passing thorough the shear-center of the cross-sections, so that the force does
not generate any torque. We assume also that the bending is symmetric,
implying that Izy = 0 and My = 0. The z-axis can then be chosen on the
neutral axis on the cross-section.
In this simple case with such a coordinate system, the shear force at
x = L is Vy = P , and moment at x = L is M = 0. As usual, we use our
standard sign convention. Using equilibrium conditions, we find at arbitrary
location x, the shear force on the cross-section is Vy = P , and the moment
becomes M = (L − x)P . It is easy to see that the relation between the shear
force and moment satisfies the following equation:

dMz
= −Vy (12.1)
dx

This equation was derived in Chapter 8 for general cases. Equation (12.1)
means that the change in moment is compensated by the shear force. If
the shear force is zero, the change in moment will be zero, and hence
the moment will be constant, which is the case of pure bending. In this
chapter, we discuss cases that are not pure bending, and hence the shear
force is nonzero. Otherwise, the shear stress will be simply zero, and we
are done.

Figure 12.1. A cantilever beam subjected to a transverse point force at its right-end.
 Shear Stress by Shear Force 467

We use the coordinate system shown in Fig. 12.1, and consider only one
shear force Vy and one moment Mz involved. Thus, there is only one normal
stress σxx and one shear stress σxy . Thus in this chapter, we

drop the subscript z, y, and x for V , M , and σ.

The normal stress σxx is denoted as σ, and the shear stress σxy is denoted
as τ . We put the subscript on when it is needed for avoiding confusion.

12.2 General formulation

12.2.1 Free-body diagram

To derive the formula for computing the shear stress on the beam cross-
section, we must analyze the stress states in more detail, using again the
usual trick: free-body diagram. It is shown in Fig. 12.2. We assume that
the Euler–Bernoulli beam theory holds so that we can still use the moment
shear-force relation and moment normal-stress relation. We require also the
thickness ty does not change too drastically with y, so that the shear stress
boundary conditions on the free surfaces can be reasonably well observed.
The free-body is an infinitely small segment from the beam with length
dx and throughout its entire depth (in the z-axis). It is taken from beam

(a)
(c)

(b)

Figure 12.2. (a) Internal forces on a cross-section of the beam. (b) A free-body diagram,
the shaded area, with an infinitely small length segment. (c) The free-body diagram viewed
on the cross-section.
468 Mechanics of Materials: Formulations and Solutions with Python

top surface (where the shear stress is zero) until y where there is unknown
shear stress denoted as τ . The free-body is the shaded area in Fig. 12.2(b)
and (c). We can then put all the nonzero stresses on the free-body, as shown
on Fig. 12.2(b). All these variables are uniform in the z-direction.

12.2.2 Equilibrium conditions

Consider the equilibrium of free-body for all forces in the x-direction, we
have
  
Fx = 0 : − σdA + (σ + dσ)dA − τ ty dx = 0 (12.2)
A A

where dσ is the incremental normal stress with respect to dx, A is the shaded
area on the cross-section shown on Fig. 12.2(c), and ty is the thickness (in
the depth z-direction) of the beam at y. Equation (12.2) can be simplified to

τ ty dx = dσdA (12.3)
A

12.2.3 Formula for shear stress

Because the normal stress on any given cross-section at x relates moment as
My
σ=− (12.4)
I
Substituting Eq. (12.4) into Eq. (12.3) gives
 
dM y dM
τ ty dx = − dA = − ydA (12.5)
A I I A
  
QA

where QA is the first moment of area A and I is the second moment of
the entire cross-section. They all are with respect to the z-axis. In deriving
Eq. (12.5), we used the fact that dM I is fixed at a given cross-section, does
not relate to dA, and hence can be moved out of the integral. Equation (12.5)
is re-arranged to
dM QA
τ =− (12.6)
dx Ity
Using Eq. (12.1), we finally obtain the formula for computing the shear
stress:
V QA
τ= (12.7)
Ity
 Shear Stress by Shear Force 469

This result shows that the shear stress at location y on any cross-section
of the beam depends on, naturally, the shear force on the cross-section. It
varies in the y-direction through the first moment of area A . The calculation
of the first moment of an area is discussed in Chapter 10. When y changes,
the shaded area changes, and hence the value of QA changes. When y gets
to the top surface (or bottom surface), the shaded area becomes zero (or the
entire cross-section), and QA = 0, leading to correct result: zero shear stress
on the top (and bottom) surface of the beam.

12.2.4 Shear-flow resulted from shear force

Using the shear stress given in Eq. (12.5), we can define the shear-flow q as
V QA
q = τ ty = (12.8)
I
Both V and I are fixed values at a cross-section, but QA changes with
y, the shear-flow resulted from a shear force varies with the vertical location
on the cross-section. Its integration over the cross-section becomes constant,
which is the shear force at that cross-section.
The shear-flow q carries a unit of [N/m]. It does not change with z, but
changes continuously with y because there is no concentrated force in the y-
direction at any point in the cross-section. The shear-flow has no circulation,
and its curl is zero. Due to these properties, we often work with q. Let us
call it q-value, and its distribution on the cross-section a q-map, which
is important to obtain when dealing with cross-sections with complicated
geometry.
Note that the shear-flow caused by shear force and that caused by torsion
behave differently. The shear stress resulting from torsion has a circulation
(a curl) round the torsion axis, as discussed in Chapter 9.

12.3 Shear stress on rectangular cross-section

12.3.1 Parabolic distribution

Consider a beam with rectangular cross-section, as shown in Fig. 12.3.
To evaluate the shear stress at location y, we compute the first moment
with respect to z-axis of the shaded area A with bottom edge located at y.
Using the last part of Eq. (12.5), it is found as
  
 1 h h b h 2
QA = y c A = +y −y b= − y2 (12.9)
2 2 2 2 2
     
yc A
470 Mechanics of Materials: Formulations and Solutions with Python

Figure 12.3. Rectangular cross-section for thin beams, on which the shear stress is
evaluated. The centroid of the shaded area A is denoted by yc .

This equation will be frequently used for computing Q of an area A for

rectangular areas. It is a parabolic function in y.
Substitute Eq. (12.9) into Eq. (12.8), and note that for rectangular cross-
section, ty = b, we obtain the shear-flow as

 2  2
bV h 6V h
q= −y 2
= 3 − y2 (12.10)
2I 2 h 2

The shear stress becomes

 2  2
V h 6V h
τ= −y 2
= 3 − y2 (12.11)
2I 2 bh 2

It is clear that shear stress is zero at y = ± h2 . It has a maximum value at

y = 0:

V h2 3V
τmax = = (12.12)
8I 2A

It is 1.5 times the average shear stress. In the foregoing equation, we used
3
I = bh
12 , and the area A = bh.
The shear stress distributed parabolically along the y-axis and uniform
along the z-axis. It is drawn in Fig. 12.4.
 Shear Stress by Shear Force 471

Figure 12.4. Parabolic distribution of shear stress in the vertical direction on a rectan-
gular cross-section.

We have a useful finding:

the shear stress has a parabolic profile in y, if the cross-section is
rectangular.
This finding can be used to draw the profile if a cross-section consists of
rectangles, such as in I-beams.
If we integrate it over the cross-section, it should give the shear stress V
on the cross-section. Let us use the following Sympy code to confirm this.

1 V, y, h, b = symbols("V, y, h, b")
2 I = b*h**3/12
3 [Link](V*b*(h**2/4 - y**2)/(2*I), (y, -h/2, h/2))

12.3.2 Alternative approach

It may be mentioned here that there is a simple approach to obtain
Eq. (12.11) if the cross-section is rectangular shape. First, from the bound-
ary condition, the shear stress must be zero at y = ± h2 . In addition, it must
be a continuous function of y. Thus, it can be written in the following form:

h 2
τ =c − y2 (12.13)
2

where c is a constant. Next, from the equilibrium condition, the integral of

it must be the shear force V . We can then integrate the foregoing expression
over the cross-section, and force it to be V to determine c. The result is
exactly the same as Eq. (12.11).
472 Mechanics of Materials: Formulations and Solutions with Python

12.4 Shear stress on thin-wall cross-sections

12.4.1 Setting of the problem

The formulation derived in the previous section works well for beams with
cross-sections on which the thickness ty does not change significantly. In engi-
neering applications, however, we often use beams with thickness changing
significantly. A typical example is the I-beams. Its flange width is far larger
than the web thickness. In these cases, the detailed distribution of the shear
stress (in the flange for example) is not properly quantified by Eq. (12.7).
For these type of beams, we need new approaches.
Many widely used beams (such as I-beams, T-beams and L-beams) has a
common significant feature: the cross-section consists of thin walls, as shown
in Fig. 12.5. Such designs are particularly good for efficient use of materials
and the structure can be made light.
For this type of thin-wall beams, the cross-section are typically consist-
ing of narrow channels. Each of the channel has a thickness t and length l
with t << l. For most beams used in engineering, the channels are largely
straight (narrow rectangles), but can also be curved (e.g. circular or elliptic).
Since the thickness is small, we often use the middle line of the channel to
represent it.
At any location in the channel, we can always set up a local orthogonal
Cartesian coordinate system, say (s, r, x) with s along with the length direc-
tion of channel, and x along with the centroid axis of the beam perpendicular
to the cross-section plane, as shown in Fig. 12.5.

Figure 12.5. Shear stress distribution along thin walls of the cross-section of an L beam.
The transverse force P goes through the shear-center.
 Shear Stress by Shear Force 473

12.4.2 Assumptions
We can make the following assumptions to each of the channels:

1. Because shear stress component σxr is zero on surfaces BB and DD , and
the channel thickness t is small with respect to the channel length, σxr
should not have significant value. Hence, σxr is assumed as zero at any
point in the channel.
2. The shear stress component σxs shall be approximately constant through
the channel thickness, and equals to the average shear stress in the
channel. Because there is only one stress component, we denote σxs as τ .
3. In addition, τ at any point in a channel is assumed to be along the
tangential direction of the channel that may be curved.

12.4.3 Free-body diagram and formulation

With these assumptions, we can then have a free-body diagram shown in
Fig. 12.6, which is essentially the same as that given in Fig. 12.2. Therefore,
the derivation and the formulas obtained in Section 12.2 are all applicable
to thin-wall cross-section beams.

(a)
(c)

(b)

Figure 12.6. (a) Internal forces on a cross-section of a beam of thin-wall cross-section

with two channels. (b) A free-body diagram with an infinitely small length segment dx
for a thin-wall cross-section. It is the shaded area starting from beam top surface to y, at
which the shear stress is evaluated. (c) The free-body diagram viewed on the cross-section.
474 Mechanics of Materials: Formulations and Solutions with Python

The difference is that the shear stress in a thin-wall flows along the tan-
gential direction of the wall curve, and not necessarily in the y-direction. In
addition, because the wall is thin, we can use its middle line for measuring
its length, when calculating the first moment of area QA in Eq. (12.7).

12.4.4 Shear-flow on thin wall cross-sections

Because of the geometric complexity of thin wall cross-sections, the shear
stress distribution is expected to be complicated. Shear-flow q defined in
Eq. (12.8) is thus particularly useful for thin wall cross-sections. This is
because a flow map of q, called q-map, determines all the major characteris-
tics of cross-section under a shear force. The major characteristics are listed
as follows:

1. The thicknesses of the channels forming the cross-section are more or

less the same. The shear stress will not change too drastically.
2. The distribution of q follows always the channel length direction, and
varies continuously, even if the thickness of the channels are different.
3. Most importantly, q is conserved at any location in the thin-wall. When
a channel branches to multiple channels at a juncture, the shear-flow out
of a channel at the juncture must be the net-in to all remaining channels
at the juncture. This is useful in computing the shear-flows in each of the
channels.
4. The integral of q for any channel on a cross-section gives a force on the
channel. The sum of the forces of all the channels is the given shear force
V on the cross-section. Because V is in the y-direction, the sum of the
forces projected on the z-direction of all the channels will be zero.
5. The moment about x-axis (in reference to a point on the cross-section)
resulting from all the channel forces forms the torque with respect to the
x-axis. This is useful to find out the location of the shear-center of the
cross-section.
6. Calculation of the shear stress is simple: dividing the shear-flow by the
local thickness of the channel t: τ = q/t. Therefore, it is convenient to
work out the shear-flow first and then compute the shear stress in these
channels.

12.5 Example: Shear stress on the cross-section

of a C-beam

Consider a C-beam with thin-wall cross-section of uniform thickness t sub-

jected to a vertical shear force V that passes through the shear-center C,
as shown in Fig. 12.7(a). The cross-section is symmetric with respect to the
 Shear Stress by Shear Force 475

(a) (b) (c)

Figure 12.7. A thin-wall cross-section with three thin channels. (a) Dimensions and
coordinates; local s coordinate located in the flange and web; (b) local s coordinate located
in the web; (c) shear stress distribution (q-map).

z-axis. It has a base b and height h measured from middle lines. Derive the
formulas for calculating the shear stress distributed on the cross-section.

12.5.1 Shear-flow and stress at any point in the flange

To compute the shear-flow, we need to compute the first moment of the area
A starting from a point where the shear stress is zero. Let us work from
the flange on the top. We know at point T, the shear stress is zero because
of free surface there. Set the local coordinate s starting from T, and move
by distance s in the top flange shown in Fig. 12.7(a). The area A is shaded
darker. Because in the flange, the y coordinate of the middle line does not
change, A shall grow linearly with increase of s. Therefore, for any point on
the flange, the Q value, QA (with respect to z-axis), must be linear in s. It
can be calculated as

th
Qs−flange = tyc s = s (12.14)
2

where yc is the y value of the centroid of A , and t is the thickness, of the

flange. The shear-flow is

V thV
qs−flange = Qs−flange = s (12.15)
I 2I

where I is the second moment of area for the cross-section that can be
computed using the techniques or code given in Chapter 10. Because the
476 Mechanics of Materials: Formulations and Solutions with Python

channels are thin, I can also be approximated using


1 3 h 2 1 1 th2
I = th + 2 tb = th3 + tbh2 = (h + 6b) (12.16)
12  2
  
12 2 12
for web for flange

The shear stress becomes

qs−flange hV
τs−flange = = s (12.17)
t 2I
It is found that the shear stress (or shear-flow) is a linear function of s.
At point D, s = b, we have
bhtV
qD =
2I
(12.18)
bhV
τD =
2I
The distribution is a triangle shown in Fig. 12.7(c).

12.5.2 Shear stress at any point s in the web

Inside the web, area A is the sum of that of the flange and the additional area
in the web: ts, as shown in Fig. 12.7(b). To compute the the first moment
of ts, we shall use the work done for the rectangular cross-section in Section
12.3 because the web is a rectangle. The Q value at y in the web becomes
(see, Eq. (12.9)):

h t h 2
Qs−web = tb + − y2 (12.19)
2
 2 2
Qflange

The shear-flow at y in the web can be found as

 2
V tV h
qweb = Qs−web = bh + − y2 (12.20)
I 2I 2
The shear stress at y becomes
 2
V h
τweb = bh + − y2 (12.21)
2I 2
The foregoing three formulas work for any point in the web, including
point D at the bottom of the web. At point D , we have exactly the same
results given in Eq. (12.18), by substituting y = −h/2 in these three equa-
tions. This is due to the symmetry of the cross-section with respect to the
 Shear Stress by Shear Force 477

z-axis. In addition, the shear stress distribution along the bottom flange
must also be mirror image of that on the top flange. The distribution of
the shear-flow in the web is parabolic in y, as studied in Section 12.3. The
complete results of shear stress and shear-flow are schematically drawn in
Fig. 12.7(c).

12.6 Example: Shear stress on the cross-section of

an L-beam

Consider an L-beam (inclined 45◦ ) with thin-wall cross-section of uniform

thickness t subjected to a vertical shear force V that passes through the
shear-center C, as shown in Fig. 12.8(a). The cross-section is symmetric with
respect to z-axis. The leg length is hL measured from middle lines. Derive
the formulas for calculating the shear stress distributed on the cross-section.
Because the shear force is in y-direction, the first and second moments of
area of the cross-section is with respect to z-axis. Therefore we can project
the area on the y-axis without affecting these area moments, as long as
the area is unchanged. This gives a rectangular cross-section as shown in
Fig. 12.8(b), for which equations for shear-flow given in Section 12.3
√ apply.
To ensure the area unchanged,
√ the base of the rectangle become 2t, and
the height becomes 2hL . The second moment of area of the L-beam can
now be calculated using
√ √
2t( 2hL /)3 th3
IL = = L (12.22)
12 3

(a) (b) (c)

Figure 12.8. A thin-wall L-shaped cross-section with two thin channels of equal length.
(a) Dimensions and coordinates; (b) area projected in y-axis; (c) shear stress shear-flow
distribution (q-map).
478 Mechanics of Materials: Formulations and Solutions with Python

The shear-flow can be calculated using Eq. (12.10):

√  √ 2 
2tV 2hL / 2 3V h2L
q= −y 2
=√ 3 − y2 (12.23)
2I 2 2hL 2

These formula can be derived using the following code.

1 V, y, h, hL, b, t, I, IL = symbols("V, y, h, h_L, b, t, I, I_L")

2 A_Lbeam = 2*(t*hL) # original area
3 A_Rctgl = [Link](2)*t*[Link](2)*hL # projected area
4 print(f'Is A_Lbeam = A_Rctgl? {A_Lbeam==A_Rctgl}') # these must equal
5
6 I_R = b*h**3/12 # formula for I value for rectangles
7 ILp = I_R.subs({b:[Link](2)*t,h:[Link](2)*hL}) # I for projected area
8 print(f'2nd moment of area of L-beam with 45 angle = {ILp}')

Is A_Lbeam = A_Rctgl? True

2nd moment of area of L-beam with 45 angle = h_L**3*t/3

1 q = V*b*(h**2/4 - y**2)/(2*I_R) # formula for q value for rectangles

2 q_L = [Link]({h:[Link](2)*hL})
3 print(f'The maximum shear-flow = {q_L.subs(y,0)}')
4 [Link](q_L, "The formula for calculating shear-flow q:")

The maximum shear-flow = 3*sqrt(2)*V/(4*h_L)

The formula for calculating shear-flow q:

√  h2 
3 2V 2L − y 2
2h3L

The shear stress in the L-beam cross-section becomes

√  √ 2 
q 2V 2hL / 2 3V h2L
τ= = −y 2
=√ 3 − y2 (12.24)
t 2I 2 2hL t 2

Note here that we need to use the original thickness of the L-beam. The
maximum value at y = 0:
√
3 2V 3 V
τmax = =√ (12.25)
4thL 2 AL

It is about 2.12 times the average shear stress. The distribution of the
shear-flow (or shear stress) is schematically plotted in Fig. 12.8(c).
 Shear Stress by Shear Force 479

12.7 Shear stress on an I-beam cross-section*

12.7.1 Problem analysis

Consider an I-beam made of three thin channels of possibly different lengths,

as shown in Fig. 12.9. Because the beam is symmetric with respect to y-axis,
the shear-center C is on the y-axis. We assume that the dimensions of the
I-beam are all given.
Let the Cartesian coordinate have its origin at the cross-section centroid
that can be found using [Link] I() presented in Chapter 10. The cross-
section has a given shear force V in the y-direction and hence passing through
the shear-center. The 2nd moment of area I (with respect to z-axis) can be
computed, using [Link] I() developed in Chapter 10.
The cross-section consists of three rectangles. For convenience in discus-
sion, we divide it into five rectangular channels marked with boxed number
in Fig. 12.9(a). Each channel has a constant thickness. Channel-1 branches
to channel-2 and -3 at juncture point D, and channel-4 and -5 at juncture
point D .

(a) (b) (c)

Figure 12.9. (a) A typical thin-wall cross-section of an I-beam with 5 channels. (b) Area
A for computing the shear flow in the flange. (c) Area A for computing the shear flow in
the web.
480 Mechanics of Materials: Formulations and Solutions with Python

Because the geometry of the I-beam is discontinuous, we shall intro-

duce local coordinate s for each of the channels. For channel-1, for example,
s starts from its left vertical edge where shear stress is zero. For the web,
it starts at point D on the top-most-point of the web. Because the rectan-
gles are assumed thin, local coordinate s stays on the middle line of the
rectangles.

12.7.2 Calculation procedure

[Link] Shear-flow and stress at any point in the flange
Let us work from the flange on the top. We know at points T and T , the
shear stress is zero. Thus, we can form area A starting from either points. Let
us start from T, and move by distance s in channel-2 shown in Fig. 12.9(b).
The area formed is shaded darker. Because on the flange, the y coordinate
does not change, the area grows linearly with increase in s. Therefore, for
any point on the flange, the Q value, QA (with respect to z-axis), must be
linear in s. It can be calculated as
Qs−flange = tf t syf tc (12.26)
where yf tc is the y value of the centroid, and tf t is the thickness of the flange
on the top. The shear-flow is
V V
qs−flange = Qs−flange = tf t syf tc (12.27)
I I
and the shear stress becomes:
qs−flange V
τs−flange = = yf tc s (12.28)
ty I
Here, we used ty = tf t because it is in the flange.

[Link] Shear-flow and stress at point D in the flange

The largest stress value within the flange should be at point D where
s = wf t /2. The Q value there is computed using
wf t
QD−flange = tf t yf tc (12.29)
2
The shear-flow is found as
V V
qD−flange = QD−flange = wf t tf t yf tc (12.30)
I 2I
The shear stress becomes
V V
τD−flange = QD−flange = wf t yf tc (12.31)
Itf t 2I
 Shear Stress by Shear Force 481

The same can be done starting from point T and working leftwards to
point D along channel-3. The shear-flow value will be the same if the I-beam
is symmetric.
The shear stress, and hence the shear-flow is linear in s as shown in
Eq. (12.28). The distribution profile is a right triangle over both channel-2
and -3. It is zero at points T and T . This will be plotted in Fig. 12.10.

[Link] Shear stress at point D in the web

At point D in the web, area A is suddenly doubled because channel-2 and
-3 merge to channel-1. Therefore, QA value is doubled:
QD−web = 2yf tc tf t (wf t /2) = tf t wf t yf tc (12.32)
  
QA at D at top of web

The shear-flow there should be twice of that at point D in the flange:

V V
qD−web = QA = tf t wf t yf tc (12.33)
I I
The stress at point D in the web becomes
qD−web V tf t
τD−web = = wf t yf tc (12.34)
ty I tw
Here, we used ty = tw .

[Link] Shear stress at any point s in the web

Inside the web, channel-1, area A is the sum of that of the flange and the
additional area in the web: tw s. Therefore, QA value is the sum of the Q
value for the flange, both channel-2 and -3, and the Q value of the additional
area in the web. To compute the Q value of the additional area in the web,
we shall use the work done for the rectangular cross-section in Section 12.3
because the web is a rectangle. To do so, we need to find out the height of
the upper part of the web above y = 0, which is denoted as hwu shown in
Fig. 12.9(c). The formula can be found using length relations:
hwu = ywc + (hb − tf t − tf b )/2 (12.35)
where ywc is the y value of the centroid of the web. Now, the Q value at s
in the web becomes (see, Eq. (12.9)):
tf t 1
QA −web = yf tc wf t + tw (hwu − y) (hwu + y) (12.36)
tw    2
s
482 Mechanics of Materials: Formulations and Solutions with Python

The shear-flow at s in the web can be found as

V
qweb = QA −web (12.37)
I
The shear stress at the same location becomes
V V
τweb = QA −web = QA −web (12.38)
Ity Itw
The foregoing three formulas work for any point in the web, including
point D at the bottom of the web. The distribution of the shear-flow in
the web is quadratic in y, as studied in Section 12.3. It will be plotted in
Fig. 12.10.

[Link] Shear stress at D on the top of the bottom flange

Because the shear-flow branches into two flows in channel-4 and -5, when
gets into the flange (or two shear-flows from the flange converge to one flow).
The shear stress at D on the top of the bottom flange can be computed using
the shear-flow at D obtained in Eq. (12.37), by dividing it by two:
1
qDp−flange = qweb (12.39)
2
where qDp−flange is for one of the branches: channel-4 or -5. The stress in the
flange becomes
qDp−flange 1 1
τDp−flange = = qweb /tf b (12.40)
2 tf b 2
Alternatively, we can do this in the same way as in Eqs. (12.29) and
(12.31):
wf b
QDp−flange = yf bc tf b (12.41)
2
Note that yf bc is the y coordinate of the centroid of the bottom flange,
which is negative. And tf t is the thickness of the bottom flange. The shear-
flow is
V V
τDp−flange = QDp−flange = yf bc wf b (12.42)
Itf b 2I
The shear-flow (and the stress) is linear, and the distribution profile is a
right triangle over both channel-4 and -5. It is zero at points B and B . This
will be plotted in Fig. 12.10.
All the complicated tedious formulas given above can be directly coded
in Python.
 Shear Stress by Shear Force 483

12.7.3 Example: Python code for computing the

shear-flow and stress in an I-beam

Consider an I-beam with basic shape shown in Fig. 12.9(a). Its dimensions
(mm) are given as follows:

beam height: hb = 200; width of flange on top: wf t = 100; width of flange

at bottom: wf b = 150; thickness of flange on top: tf t = 20; thickness of
flange at bottom: tf b = 18; web thickness: tw = 10.

Assume the shear force V = 0.1M N . Compute the distribution of the

shear stress in the I-beam.

Solution:

[Link] Data preparation

First, prepare the input data, and compute the area properties.

1 # partial checked with Fig.6.8 in the textbook by Beer using data:

2 #hb =.140; wft=.100; wfb=.100; tft=.02; tfb=.02; tw=.02 # I-beam (m)
3
4 # preparing date:
5 # Symmetric case for test:
6 #hb =.200; wft=.100; wfb=.100; tft=.020; tfb=.020; tw=.010 # I-beam (m)
7
8 # data for this example:
9 hb =.200; wft=.100; wfb=.150; tft=.020; tfb=.018; tw=.010 # I-beam (m)
10
11 # call the code function to compute the sectional properties:
12 Ixx, Iyy, Ixy, x_bar, y_bar, xi_bar, yi_bar, A, Ai =\
13 gr.secondMoments_I(hb, wft, wfb, tft, tfb, tw, [0]*2, [0]*2, sy=True)

Center of sub-areas [L]:

x = [7.5000e-02 7.5000e-02 7.5000e-02]
y = [1.9000e-01 9.9000e-02 9.0000e-03]
Areas of sub-areas = [2.0000e-03 1.6200e-03 2.7000e-03] [L^2]
Areas of I-beam = 0.00632 [L^2]
The centroid = (0.075, 0.0893481012658228) [L]
New center of sub-areas [L]:
x = [0.0000e+00 0.0000e+00 0.0000e+00]
y = [1.0065e-01 9.6519e-03 -8.0348e-02]
2nd moments [L^4]:
Ixx = 4.152574084388186e-05
Iyy = 6.742666666666666e-06
Ixy = 0.0
484 Mechanics of Materials: Formulations and Solutions with Python

[Link] Compute shear-flow and stress at D in the top flange

1 # Compute shear-flow and stress at D in the top flange

2
3 V = 0.1e6 # (N)
4 Izz = Ixx # in our coordinates, z is x in the gr.secondMoments_I()
5
6 Q_D_flange = yi_bar[0]*tft*wft/2
7
8 q_D_flange = V/Izz * Q_D_flange
9 print(f"shear-flow at D in top flange, q = {q_D_flange*1e-6} [MPa/m]")
10
11 _D_flange = V/(Izz*tft) * Q_D_flange
12 print(f"Shear stress at D in top flange, = { _D_flange*1e-6} [MPa]")

shear-flow at D in top flange, q = 0.24238435411082282 [MPa/m]

Shear stress at D in top flange, = 12.119217705541141 [MPa]

[Link] Compute the shear-flow and stress at D on the top of the web

1 # compute the shear-flow and stress at D on the top of the web

2 Q_Dweb = yi_bar[0]*wft*tft
3 q_D_web = V/Izz * Q_Dweb
4 print(f"shear-flow at D in web, q = {q_D_web*1e-6} [MPa/m]")
5
6 _D_web = V/(Izz*tw) * Q_Dweb
7 print(f"Shear stress at D in web, = { _D_web*1e-6} [MPa]")

shear-flow at D in web, q = 0.48476870822164575 [MPa/m]

Shear stress at D in web, = 48.47687082216458 [MPa]

[Link] Compute shear-flow and stress at any s in web

1 # Compute shear-flow and stress at key points in web

2 hw = hb - tft/2 - tfb/2 # the height of the web
3 hwu = yi_bar[1] + hw/2 # height above z=0
4
5 ys = 0 # at the neutral axis, _max
6 Q_sweb = Q_Dweb + tw*(hwu - ys)*(hwu + ys)/2
7
8 q_0_web = V/Izz * Q_sweb
9 print(f"shear-flow at the neural axis, q = {q_0_web*1e-6} [MPa/m]")
10
11 _0_web = V/(Izz*tw) * Q_sweb
12 print(f"Shear stress at the neural axis, = { _0_web*1e-6} [MPa]")

shear-flow at the neural axis, q = 0.6055420239557431 [MPa/m]

Shear stress at the neural axis, = 60.55420239557431 [MPa]
 Shear Stress by Shear Force 485

1 # Compute shear-flow and stress at key points in web

2
3 hwl = yi_bar[1] - hw/2 # height below z=0
4 ys = hwl # D' at the bottom of the web
5 Q_sweb = Q_Dweb + tw*(hwu-ys) * (hwu+ys)/2
6
7 q_Dp_web = V/Izz * Q_sweb
8 print(f"shear-flow at D' in web, q = {q_Dp_web*1e-6} [MPa/m]")
9
10 _Dp_web = V/(Izz*tw) * Q_sweb
11 print(f"Shear stress at D' in web, = { _Dp_web*1e-6} [MPa]")

shear-flow at D' in web, q = 0.5268388467762832 [MPa/m]

Shear stress at D' in web, = 52.68388467762832 [MPa]

[Link] Compute shear stress at D at the top of the bottom flange

1 # Compute shear stress at D' at the top of the bottom flange,

2 # using the shear-flow conservation property:
3
4 q_Dp_flange = q_Dp_web/2 # q branches to two flows
5 print(f"shear-flow at D' in flange, q = {q_Dp_flange*1e-6} [MPa/m]")
6
7 _Dp_flange = q_Dp_flange/tfb
8 print(f"Shear stress at D' in flange, = { _Dp_flange*1e-6} [MPa]")

shear-flow at D' in flange, q = 0.2634194233881416 [MPa/m]

Shear stress at D' in flange, = 14.634412410452313 [MPa]

1 # Alternatively, using the standard VQ/I formula:

2
3 Q_Dp_flange = -yi_bar[2]*tfb*wfb/2
4 q_Dp_flange = V/Izz * Q_Dp_flange
5 print(f"shear-flow at D' in flange, q = {q_Dp_flange*1e-6} [MPa/m]")
6
7 _Dp_flange = V/(Izz*tfb) * Q_Dp_flange
8 print(f"Shear stress at D' in flange, = { _Dp_flange*1e-6} [MPa]")

shear-flow at D' in flange, q = 0.2612113221932849 [MPa/m]

Shear stress at D' in flange, = 14.511740121849163 [MPa]

12.7.4 Plot of shear-flow in the web

We use the following code to plot the shear-flow distribution in the web of
the I-beam.
486 Mechanics of Materials: Formulations and Solutions with Python

1 dx = 0.005
2 y_web = [Link](hwl, hwu+dx, dx) # y_web values as x_data
3
4 Q_sweb = Q_Dweb + tw*(hwu - y_web)*(hwu + y_web)/2
5 q_0_web = V/Izz * Q_sweb
6
7 step = max(int(len(y_web)/5), 1)
8 print(f"y values in the web (m):\n {[Link](y_web)[::step]}")
9 print(f"The shear-flow in the web (MPa/m):\n {q_0_web[::step]*1e-6}")
10
11 x_data = [y_web]
12 y_data = [q_0_web*1e-6] # put the function values to a list
13 labels=['shear-flow (MPa/m)'] # define the label
14 gr.plot_fs(x_data, y_data, labels, 'y', 'shear-flow q (MPa/m)', 'temp')
15 # x-label, y-label, file-name

y values in the web (m):

[-8.0848e-02 -4.5848e-02 -1.0848e-02 2.4152e-02 5.9152e-02 9.4152e-02]
The shear-flow in the web (MPa/m):
[5.2684e-01 5.8023e-01 6.0413e-01 5.9852e-01 5.6341e-01 4.9881e-01]

The final results on the shear-flow map are given in Fig. 12.10.

12.7.5 Compute forces in channels

After shear-flow is obtained, we can then compute the resultant force in each
of the channels in the cross-section. This is done by integrating the shear-flow
over the channel length. The general formula can be written as

F = qds (12.43)
l
 Shear Stress by Shear Force 487

(a) (b)

Figure 12.10. Typical shear-flow in an I-beam. (a) Shear-flow conservation at junctions

of branches; (b) q-map on the channels of the I-beam.

where l is the length of the channel, and q is the shear-flow in the channel.
The direction of the force follows the shear-flow direction. If the channel
is not orientated orthogonal to the coordinates, one may need to project
the resultant force to the coordinate directions to obtain the components
of the force. For most of the beams used in engineering, the channels are
rectangular and are arranged orthogonal to the coordinate. In such cases,
the integration can be done by calculating the area under the shear-flow
profile. For example, in the results given in Fig. 12.10(b), the force on the
flanges is the area of the triangles. The formula for computing the area under
a parabolic function can also be done easily. Following are some Python codes
to do such computations. We use the same I-beam example studied before.

1 # Computing the force in the flange of I-beam.

2 F_flangeTop = -0.5*q_D_flange*wft
3 print(f"Force in the right top flange (kN): {F_flangeTop*1e-3}")
4
5 F_flangeBot = 0.5*q_Dp_flange*wfb
6 print(f"Force in the right bottom flange (kN): {F_flangeBot*1e-3}")

Force in the right top flange (kN): -12.119217705541143

Force in the right bottom flange (kN): 19.59084916449637
488 Mechanics of Materials: Formulations and Solutions with Python

1 # Sometimes it is more convenient using numerical integration:

2
3 qt_s = lambda s: -q_D_flange/wft * s # define a function for q
4 F_topFlange, er = [Link](qt_s, 0, wft) # use Gauss int.
5 print(f"Force in the right top flange={F_topFlange*1e-3}(kN),\
6 int_error= {er:.4e}")
7
8 qb_s = lambda s: q_Dp_flange/wfb * s
9 F_botFlange, er = [Link](qb_s, 0, wfb) # use Gauss int.
10 print(f"Force in the right bottom flange={F_botFlange*1e-3}(kN),\
11 int_,error={er:.4e}")

Force in the right top flange=-12.119217705541141(kN), int_error= 1.3455e-10

Force in the right bottom flange=19.59084916449637(kN), int_,error=2.1750e-10

Let us compute the force in the web of I-beam. We define first a function
for computing the shear-flow at any location in the web.
1 # Define the distribution function:
2 qw_s = lambda y: V/Izz * (Q_Dweb + tw*(hwu - y)*(hwu + y)/2)
3 F_webG, error = [Link](qw_s, hwl, hwu) # use Gauss int.
4
5 print(f"Force in the web (kN): {F_webG*1e-3}, int_error= {error:.4e}")
6 print(f"The true shear force V (kN): {V*1e-3}")

Force in the web (kN): 103.45021322740982, int_error= 1.1485e-09

The true shear force V (kN): 100.0

1 # Computing the force in the web of I-beam. Use simpler trapezoidal rule
2 dx = 0.001
3 y_web = [Link](hwl, hwu+dx, dx) # y_web values as x_data
4
5 step = max(int(len(y_web)/10), 1)
6 print(f"y values in the web (m):\n {[Link](y_web)[::step]}")
7 print(f"The shear-flow in the web (MPa/m):\n {q_0_web[::step]*1e-6}")
8
9 F_web = [Link](qw_s(y_web), y_web)
10 print(f"Force in the web : {F_web*1e-3}(kN)")
11 print(f"Actual shear force V: {V*1e-3}(kN)")

y values in the web (m):

[-8.0848e-02 -6.2848e-02 -4.4848e-02 -2.6848e-02 -8.8481e-03 9.1519e-03
2.7152e-02 4.5152e-02 6.3152e-02 8.1152e-02 9.9152e-02]
The shear-flow in the web (MPa/m):
[5.2684e-01 6.0453e-01 4.8717e-01]
Force in the web : 103.44984999889122(kN)
Actual shear force V: 100.0(kN)

We obtained the force in the web via numerical integration. Compared

to the given true shear force V , we got about 3% error. It is a good approx-
imation. The error is because of the thin wall approximation and use of the
 Shear Stress by Shear Force 489

middle line to compute the Q values for these channels. Readers may con-
firm this by artificially setting the thickness of all the channels at very small
value, and observing the reduction of error.
Readers may modify the codes given above to compute shear-flows and
stresses of beam with other types of cross-sections.

12.8 Shear-center of a cross-section

12.8.1 Setting of the problem

We mentioned several times that transverse forces on beams should go
through the so-called shear-center, so that no torque will be introduced
to the beam. Figure 12.11 shows the location of shear-center for three dif-
ferent cross-sections, and twist deformations when the transverse force does
not go through the shear-center. The major observations are:

1. When a transverse force passes through the shear-center C of a cross-

section of the beam, it causes bending without torsion.
2. If the cross-section has either an axis of symmetry or an axis of antisym-
metry, C is located on that axis. If the cross-section has two or more axes
of symmetry or antisymmetry, the shear-center is located at the intersec-
tion of these axes, such as the cases in Figs. 12.11(a) and (c). In such
cases, locating the shear-center is easy.
3. If the cross-section has only one or no axis of symmetry (or antisymme-
try), as shown typically in Fig. 12.11(b), effort is needed to locate the
shear-center.

We now discuss about the procedure and technique to locate the shear-
center for a given cross-section.
Consider a C-beam shown in Fig. 12.12.

(a) (b) (c)

Figure 12.11. Schematic drawings of shear-center C on three typical shapes of cross-

section of beams.
490 Mechanics of Materials: Formulations and Solutions with Python

Figure 12.12. Cross-section of a C-beam. It does not have any symmetric axis. The
shear-center is located on the left side of the web with an unknown distance e.

The C-beam has three rectangular channels. Its cross-section may not
have any symmetric axis, and hence it is not obvious where the shear-center
is located. By intuition, the shear-center should be on the left side of the
web of the cross-section, but the distance e needs to be determined.

12.8.2 Strategy to find the shear-center

Based on our analysis in the previous section, we know how to calculate the
shear-flow on each of the channels of the cross-section, and we demonstrated
this for a general I-beam. The shear-flow can then be integrated to compute
the shear force on these channels. This is done based on the assumption
that the transverse shear force passing through the shear-center and hence
no torque is generated. Therefore, the shear forces in these channels are the
results of shear force only. We also demonstrated that the shear force in the
vertical web channel equals the shear force V on the cross-section. In other
words, the shear force V creates the shear force on the vertical channel.
For the C-beam cross-section, we can do exactly the same calculation and
find shear-flow and then a force in the vertical web channel for any given
shear force V on the cross-section. The channel force is created by V , under
the assumption that it does not generate any net torque. At the same time,
we will also find the two channel forces: each on a horizontal channel as
 Shear Stress by Shear Force 491

shown in Fig. 12.12. Force Ft is on the top flange and Fb is on the bottom
flange. These two channel forces must be equal in magnitude and opposite in
direction because of the equilibrium condition for all forces in the horizontal
z-direction. These two forces form a couple that gives a torque about the
x-axis, Ft h. Because these two forces are also generated by the given vertical
force V (just like the Fw is created by V ), V must be placed at a distant e
from the middle line of the vertical web, so that the torque of eV can create
the torque of Ft h. Therefore, we have

eV = hFt (12.44)

which gives the formula for e.

hFt
e= (12.45)
V
Note that the value of Ft dependents linearly on V . Therefore, e value
will not depend on the value of V . It is a property of the geometry of the
cross-section, and its orientation in relation to the direction of V . Therefore,
one can use any nonzero amount of V to compute e, such as a unit force,
V = 1.

12.8.3 Procedure to find the shear-center

The procedure to find e is quite straightforward:

1. Compute the shear-flow and the shear forces on each of the channels on
the cross-section, for a unit (or any nonzero) shear force applied on the
cross-section.
2. Compute the torque generated by the channel forces with respect to a
point on a vertical channel.
3. Use Eq. (12.45) to compute e.

12.8.4 Example: Find the formula to compute e

for a C-beam
Consider a C-beam shown in Fig. 12.12. Assume tf t = tf b = tf , and wf t =
wf b . The channel force Ft can be calculated using Eq. (12.43).
 b
1 V V tf b2 h
Ft = qt ds = b tf b h2 = (12.46)
0 2 I   4I
A ȳ 
  
qt
492 Mechanics of Materials: Formulations and Solutions with Python

where b is the length of the top flange measured by middle lines of the
channels, qt is the shear-flow in the top flange, tf is the thickness of the top
flange, and h is the height of the beam measured between the two middle
lines in the two flanges. Because the channels are thin, the second moment
of area of the cross-section I can be approximated using

 2
1 h 1 1
I= tw h3 + 2tf b = tw h3 + tf bh2 (12.47)
12 2 12 2

Substituting Eq. (12.47) into Eq. (12.46), and then into Eq. (12.45), we
obtain:

3b
e= (12.48)
6 + αA

where αA is ratio of the web area and the area of one flange:

tw h
αA = (12.49)
tf b

It is clear that shear force V is not involved in Eq. (12.48) at all. All we
need is the geometric parameters for the cross-section, for the given direction
of V .
The approximate locations of the shear-center for several other thin-wall
sections are given in Table 8.1 in Ref. [1]. Those are all dependent on geo-
metric parameters of the cross-section.

12.8.5 Example: Python code for computing the

shear-center location for a box-beam*
Consider a box-beam with basic shape shown in Fig. 12.13(a). It is from the
textbook [1]. Its dimensions (mm) measured based on the middle lines of the
edges are given as follows:

beam height: h = 500; width of top- and bottom-edge: b = 300; thickness

of top-edge: tT = 10; thickness of bottom-edge: tB = 10; thickness of left-
edge: tL = 20; thickness of right-edge: tR = 10.
 Shear Stress by Shear Force 493

(a) (b)

Figure 12.13. Cross-section of a rectangular box-beam with thin walls. It has four edges.
The shear-center is located inside of the box with an unknown distance e measured from
the middle line of the right-edge. (a) Geometry description with shear force direction given;
(b) two divided sections.

Compute the shear-center of this box-beam, with respect to the middle

line of the right-edge.
Solution: We shall use the following strategy to solve this problem.

[Link] Dividing the box to two open cross-sections

Because the cross-section of the beam is closed, we do not know where is
the point on the edges that has zero shear-flow. The technique given earlier
cannot be used directly to obtain the q-map. Therefore, we perform the
analysis in following steps:

1. Cut the box into two sections: one on the left becomes a C-beam and
another one on the right is a rectangular beam, as shown in Fig. 12.13(b).
Therefore, the shear stress (hence shear-flow) at point A for both beams
is now zero. Formulas given in the previous sections apply.
2. Find the q-map for the “open box” with divided two sections separately.
3. Because the shear-flow at point A is not zero in the original box section,
superpose an unknown shear-flow, say qA , uniformly on all edges, and
then use the following condition to determine qA :

qfinal
dl = 0 (12.50)
box t
494 Mechanics of Materials: Formulations and Solutions with Python

where qfinal is the final superposed q-map, and t is the thickness of the
edges. This condition must be satisfied because the torque on the closed
box is zero, based on our discussion in Chapter 9.

Because qfinal = qopen-box + qA , Eq. (12.50) can be written as

 
qopen-box 1
dl + qA dl = 0 (12.51)
box t box t

where qopen-box is the q-map for the “open box” (with divided two sections).
The foregoing equation gives,
 qopen-box
box t dl
qA = (12.52)
box t dl
1

After qA is found, the superposed q-map is the final q-map for the closed
box section. We can then compute the forces on all edges, and then find the
shear-center, by enforcing the moment equilibrium about the axial x-axis.

[Link] Splitting the shear force

Because the shear-center of any cross-section is a property of the geometry
of the section, once the direction of the shear force is given (in our case,
it is fixed in the y-direction). Therefore, the location of the shear-center on
the box cross-section should not depend on the shear force V value, we can
simply set it to the I value, just for conveniences in evaluation:

V =I (12.53)

In this special setting, the q value can be evaluated simply using

q = QA (12.54)

Note that this equation is only used for convenience in locating the shear-
center. The shear-flow has a special unit of VI .
Using the relation of the shear force with the deflection (Eq. 8.17), we
have

V ∂3v
= −E 3 (12.55)
I ∂x
When the box is divided into two open sections, these two parts must
have the same deflection to represent the original problem, implying that the
 Shear Stress by Shear Force 495

right-hand-side of Eq. (12.55) must be equal for these two divided beams.
Thus, the ratio of VI for both parts must be equal. Because the box is divided
vertically, the second moment of area must satisfy:

I = IL + IR (12.56)

where IL is the I value for the left, and IR is the I value for the right section,
all with respect to the z-axis. Using Eq. (12.53) the shear force should also
be divided as

V = VL + VR (12.57)

where VL is the shear force for the left, and VR is the shear force for the
right section, all in the y-direction.
Note that in computing the q values for the purpose of locating the shear-
center, we can use Eq. (12.53) for both sections:

VL = I L ; VR = IR (12.58)

Also because torsion is involved in the process for the closed section, we
use the standard sign conversion for involved variables: All q vectors along
the counterclockwise direction with respect to x is positive.
With this setting, we are now ready to write a Python code to solve the
problem. We write a Python function for the general purpose of computing
the shear-center of any given rectangular box section, together with the
relative shear-flow values at key positions where values are the largest. We
provide detailed comments in the code, and it should be self-explanatory.
Figure 12.14 can be a good guide, when reading the following codes.

[Link] General purpose code for rectangular box sections

Based on the strategy outlined above, computing q values is to compute
the first moment of area QA (see Eq. (12.54)). This can be easily done for
these two cutoff sections, starting from point A (or B). This gives q-maps
for all these four channels of the box-section. The obtained q-maps are given
in Fig. 12.14(a). The integrals in Eq. (12.52) can be computed using the
following equations. For channel i, we compute first:
 li
qi
τl,open-box,i = dl (12.59)
0 ti

where qi , ti , li (i = top, left, bottom, right) are, respectively, the q-map,

thickness, and length of these four channels of the open-box. The positive
496 Mechanics of Materials: Formulations and Solutions with Python

(a) (b) (c)

Figure 12.14. q-maps on the rectangular box section. (a) q-map for the two divided
sections; (b) q-map for the uniformly distributed qA on the box section, whose value is
to be found in the process locating the shear-center; (c) the superposed final solution of
q-map on the box section. The shear force in the vertical y-direction is set as V = I. The
positive direction is counterclockwise about x-axis.

direction follows the counterclockwise with respect to x-axis. Next, we com-

pute their sum:

τl,box = τl,open-box,i (12.60)
i

Similarly, for the uniform q-map given in Fig. 12.14(b), we set qi = 1 (all
positive) for all channels, and then compute for channel i:
q1
τl,1,i = li (12.61)
ti

and

τl,1 = τl,1,i (12.62)
i

Finally, qA is found using

τl,box
qA = − (12.63)
τl,1

After qA is found, we can the produce the final q-map via superposition
for the closed-box. The result is given in Fig. 12.14(c). Using this final q-map,
we compute the shear forces in each of the channels, and finally the shear
center, as we did in the previous examples.
 Shear Stress by Shear Force 497

The following Python function follows exactly the foregoing equations.

1 def box_shearCenter(h, b, t_L, t_R, t_T):

2 '''Compute the shear-center of a box-section with 4 thin edges wrt
3 the middle line of the right-edge. Its dimensions are measured
4 based on the middle lines of the edges.
5 h: beam height; b: width of top- and bottom-edge;
6 t_T: thickness of top-edge; t_B thickness of bottom-edge: t_B=t_T
7 t_L, t_R: thickness of left-edge and right-edges.
8 Return:
9 e: shear-center; qP, qQ, qS, qA: q values at key points:
10 P(north-west), Q(west), S(east), A(north-east).
11 '''
12 t_B = t_T
13 # compute Izz of whole box section:
14 Izz = (1./12)*(t_L+t_R)*h**3 + b*t_T*(h/2)**2 + b*t_B*(h/2)**2
15 print(f"2nd moment of area of whole box section Izz = {Izz} [L^4]")
16
17 # Compute shear-flow at P and R (north-west, south-west corners)
18 qP = b*t_T*(h/2) # after cutting
19 qR = b*t_B*(h/2) # after cutting
20 # qQ = qP + (h/2)*t_L*(h/4); qS = (h/2)*t_R*(h/4)
21 #print(f"shear-flows at key points [Pa/m]:qP={qP:.4e};qR={qR:.4e}")
22
23 # Define = q/t as functions of s for all four edges:
24 t_s = lambda s:-qP/b/t_T * s # top edge
25 b_s = lambda s:-qR/b/t_B * s # bottom edge
26 l_s = lambda s:-(qP + t_L*(h/2 - s)*(h/2 + s)/2)/t_L # left edge
27 r_s = lambda s: t_R*(h/2 - s)*(h/2 + s)/2/t_R # right edge
28
29 # Compute l_{box,i} via integration for all four edges:
30 l_t, _ = [Link]( t_s, 0, b) # use Gauss integration
31 l_b, _ = [Link]( b_s, 0, b)
32 l_l, _ = [Link]( l_s, -h/2, h/2)
33 l_r, _ = [Link]( r_s, -h/2, h/2)
34
35 l_box = l_t + l_b + l_l + l_r # the sum
36
37 # Compute total area of q/t for the unit q=1 on all four edges:
38 l_q1 = b/t_T + b/t_B + h/t_L + h/t_R
39
40 qA = - l_box/ l_q1
41 #print(f"q value at point A (north-east corner) qA = {qA:.4e}")
42
43 # Compute the channel forces:
44 # Define (q+qA) as functions of s for all four edges:
45 qt_sA = lambda s:-qP/b*s + qA # for top edge
46 qb_sA = lambda s:-qR/b*s + qA # for top edge
47 ql_sA = lambda s:-(qP + t_L*(h/2-s)*(h/2+s)/2) + qA # left edge
48 qr_sA = lambda s: t_R*(h/2 - s)*(h/2 + s)/2 + qA # right edge
498 Mechanics of Materials: Formulations and Solutions with Python

49
50 # Compute shear forces for the left and right channels:
51 F_l = [Link](ql_sA, -h/2, h/2)[0] # use Gauss integration
52 F_r = [Link](qr_sA, -h/2, h/2)[0]
53 #print(f"Shear forces: F_l={F_l:.4e}; F_r={F_r:.4e}")
54 V = -F_l+F_r
55 #print(f"The total shear forces in box-beam (V=Izz): V={V:.4e}")
56
57 # Compute shear-center wrt the middle line of the right edge.
58 F_t, _ = [Link](qt_sA, 0, b) # use Gauss int.
59 F_b, _ = [Link](qb_sA, 0, b)
60 #print(f"Shear forces: F_t={F_t:.4e}; F_b={F_b:.4e}")
61 e = -(F_t*h + F_l*b)/V
62 print(f"e wrt the middle line of the right edge: {e:.4e}")
63 return e, qt_sA(b), ql_sA(0), qr_sA(0), qr_sA(h/2)

[Link] Data preparation

Because the box-beam is top-down symmetric, its centroid is on z-axis. The
input data for this example are given as follows:

1 # Given length of middle lines. Assume the box is top-down symmetric.

2 h =.500 # all in m # height of the beam
3 t_L =.020; t_R =.010 # thickness of the Left and Right edge
4 b =.300 # base (width) of the beam
5 t_T =.010 # thickness of the top (=bottom) edge

Using box shearCenter() to obtain the solution.

1 e, qP, qQ, qS, qA = box_shearCenter(h, b, t_L, t_R, t_T)

2 print(f"q/t value at key points:\n qP = {qP:.4e}; qQ = {qQ:.4e};"
3 f"\n qS = {qS:.4e}; qA = {qA:.4e}")

2nd moment of area of whole box section Izz = 0.0006875 [L^4]

e wrt the middle line of the right edge: 2.0303e-01
q/t value at key points:
qP = -4.4444e-04; qQ = -1.0694e-03;
qS = 6.1806e-04; qA = 3.0556e-04

[Link] Distribution profile of the shear-flow

Finally, the distribution profile of the shear-flow on the box cross-section at
different stages is given in Fig. 12.14.
 Shear Stress by Shear Force 499

12.8.6 Example: The shear-center location

for a C-beam
Consider a C-beam shown in Fig. 12.12. Its dimensions (mm) measured based
on the middle lines of the edges are given as follows:

beam height: hb = 500; width of top- and bottom-edge: b = 300 with

twf b = twf t ; thickness of top-edge: ttf t = 10; thickness of bottom-edge:
ttf b = ttf t ; thickness of web on the left: tw = 20.

Compute the shear-center of this C-beam, with respect to the middle line
of the right-edge.

Solution: One may modify a little the Python function box shearCenter()
to make it for C-beams. The easiest is to use it as it is, by simply setting
the thickness of the right-edge tR to a very small value. The code is given in
the following.

1 # Given length of middle lines. The C-section is top-down symmetric.

2 h =.500 # all in m # height of the beam h_b
3 t_L =.020 # this is same as $t_w$
4 t_R =1.0e-12 # Set to a small value but not 0 to avoid zero division
5 b =.300 # base (width) of the beam
6 t_T =.010 # thickness of the top (=bottom) edge

Using box shearCenter() to obtain the solution.

1 e, qP, qQ, qS, qA = box_shearCenter(h, b, t_L, t_R, t_T)

2 # or: gr.box_shearCenter(h, b, t_L, t_R, t_T)
3
4 # change the reference to the middle line of the web on the left:
5 eL = e - b
6 print(f"e wrt the middle line of the left edge: {eL:.4e}")
7 print(f"q/t value at key points:\n qP = {qP:.4e}; qQ = {qQ:.4e};"
8 f"\n qS = {qS:.4e}; qA = {qA:.4e}")

2nd moment of area of whole box section Izz = 0.0005833333333437501 [L^4]

e wrt the middle line of the right edge: 3.9643e-01
e wrt the middle line of the left edge: 9.6429e-02
q/t value at key points:
qP = -7.5000e-04; qQ = -1.3750e-03;
qS = 1.1375e-13; qA = 8.2500e-14
500 Mechanics of Materials: Formulations and Solutions with Python

We can confirm the solution using Eq. (12.48) derived earlier for C-beams.

1 aA = t_L*h/(t_T*b)
2 e = 3*b/(6+aA)
3 print(f"e wrt the middle line of the right edge: {e:.4e}")

e wrt the middle line of the right edge: 9.6429e-02

As shown, we obtained the same solution for the shear-center. The

code box shearCenter() provides additional results on relative shear-flows
at key points on the C-beam cross-section. It can be useful for other similar
problems.

12.9 Remarks
This chapter focuses on computing the shear stress or shear-flows on the
cross-section of a beam subjected to transverse loading. The value of the
shear stress depends on the first moment of the area formed from the sur-
face free from shear stress to the surface where the shear stress is in question.
Techniques and codes are developed for computing the q-map, the distribu-
tion of the shear stresses, as well as the shear forces. These techniques and
codes work for beams with irregular cross-section, including the section with
multiple narrow channels.
With the techniques and codes for computing the shear-flows and shear
forces, the shear-center can then be determined with ease. This enables
proper application of transverse loading without generating torsion to the
beam.

References
[1] A.P. Boresi and R.J. Schmidt, Advanced Mechanics of Materials. John Wiley & Sons,
New York, 2002.
[2] J.M. Gere and S.P. Timoshenko, Mechanics of Materials, Van Nostrand Reinhold
Company, New York, 1972.
[3] S. Timoshenko and J.N. Goodier, Theory of Elasticity, McGraw-Hill, New York,
1970. [Link]
asticity&#38;ei=ICiKSsr3G4jwkQSbxMyPCg.
[4] Z.L. Xu, Elasticity, Vols. 1&2. People’s Publisher, China, 1979.
 Chapter 13

Energy Methods

1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.

2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 init_printing(use_unicode=True) # For latex-like quality printing
11 np.set_printoptions(precision=4,suppress=True) # Digits in print-outs
12 np.set_printoptions(formatter={'float_kind': '{:.4e}'.format})

A stressed solid stores energy. The level of the energy stored shall naturally
depend on the externally applied loads, as well as the states of the loaded
solid, including the displacement and/or stress field in it. Therefore, there
should be possible methods that make use of the energy to find the states of
the solid in response to the external loads. This type of technique is called
energy methods, which are the focus of this chapter.
Because such methods are useful in dealing with large-scale structural
systems, especially in powerful numerical methods for approximated solu-
tions, we thus place emphasis on understanding the principles and concepts
of energy methods. This chapter is written with reference to [1–8].

13.1 Energy principles and fundamental concepts

Energy methods are used widely to obtain solutions to elasticity problems.
When a loaded solid is in a stable or equilibrium state, the energy should be

501
502 Mechanics of Materials: Formulations and Solutions with Python

at stationary. Otherwise, the solid shall still be in a deformation process.

The solution or important equations can thus be obtained via essentially a
minimization process. Energy methods discussed in this chapter focuses on
linearly elastic solids with small displacements, unless specified explicitly.
We will study two energy principles that are most fundamentally essential:

1. Potential energy principle (PEP); and

2. Complementary energy principle (CEP).

Both can be derived from the principle of virtual work, which uses virtual
displacements or virtual forces. The former leads to PEP and the latter to
CEP. This chapter will elaborate both in detail.
Using energy method requires a conceptual shift in dealing with mechan-
ics problems. Thus, we will first use a one degree of freedom system (1-DOF)
to elaborate the fundamental concepts in great detail. Once these concepts
are established, we will then naturally extend the formulation to complex
solid structure systems.

13.2 Principle of virtual work by virtual displacement

Consider a mechanically loaded solid constrained without rigid body move-

ment. Assume there are no other effects involved, such as electromagnetic,
thermal, etc. The principle of virtual work states as follows.
The work done by the external forces to a solid that undergoes an arbitrary
infinitely small virtual displacement equals the variation in strain energy
in the solid caused by the virtual displacement.
Mathematically, it is expressed as
δWd = δU (13.1)
where W denotes the work done by external forces applied on the solid, U
stands for the strain (also called potential) energy stored inside the solid,
and δ stands for small variation.
Equation (13.1) can be expressed as
δ (W − U ) =0 (13.2)
 d 
total potential energy

where (Wd − U ) is often called the total potential energy. Its stationary
point corresponds to an equilibrium state of the solid.
 Energy Methods 503

The principle of virtual work is an energy conservation law for solids that
are loaded only mechanically. This principle is the basis for deriving PEP.
The best way to discuss the concepts and formulations is to use a simple
example that is already familiar to us.

13.2.1 Example: A loaded linear elastic spring

Let us consider the simplest possible 1-DOF system: a linear elastic spring
fixed at point A and loaded at point B. It has a stiffness K known as spring
constant. The force f is (externally) applied at point B in the longitudinal
direction of the spring, resulting in displacement d at the same point B.
On the other hand, the spring experiences an internal force denoted as N ,
and shall have an elongation denoted as e.
Because the principle of virtual work relates two energies: δWd that is
the work-done by the external force, and δU potential energy change inside
the solid/structure. The best way to clearly discuss about this relation is to
draw two graphs: f ∼ d giving the work-done by the external force shown in
Fig. 13.1(a), and N ∼ e which gives the internal potential energy shown in
Fig. 13.1(b). The former is the external action, and the latter is the internal
reaction.
Let us use this simple system to study in detail these energy principles.

13.2.2 Evaluation of potential energy

At such a loaded state, inside the spring, there must be an internal force
N and an elongation e. The constitutive equation for the linear elastic
spring can be written as

N = Ke (13.3)

Figure 13.1. A spring loaded by a force f at point B resulting in a displacement d also

at point B.
504 Mechanics of Materials: Formulations and Solutions with Python

(a) (b)

(c)

Figure 13.2. (a) The force displacement relation at point B of the spring system that
is subjected to small virtual displacement δd; The external action: work done by the
force f under δd is evaluated. (b) The relationship of internal-force N and elongation e
of the spring body that stores strain energy; The internal reaction: variation in strain
energy δU resulting from δd is evaluated. (c) The outcome form the PEP.

This implies that the internal force in the spring is proportional to its
elongation with a constant K. This N ∼ e relationship can be plotted in
Fig. 13.2(b). Note that we need to carefully distinguish the internal and
external forces because of the need for proper energy evaluation. The external
work done is by the external force f , but the potential energy should be
evaluated using the internal force N .
In the loaded spring, some energy is stored within the spring. One measure
is the potential energy denoted as U . It is the area under the black line in
 Energy Methods 505

Fig. 13.2(b), which is the triangular area below the straight line. Thus, it
can easily be calculated using
1 1
U = N e = Ke2 (13.4)
2 2
Equation (13.4) is the formula for computing the potential energy in a
spring in terms of its elongation e. Also, we need to carefully distinguish
the displacement d at point B, and the elongation of e of the spring. The
displacement d goes with the external force f , resulting in the work done.

13.2.3 Imposition of infinitely small virtual

displacement
Next, at the actually loaded state, we impose (imaginarily) an infinitely small
virtual displacement δd at point B on the spring, at which f is on. The work
done by f should be

δWd = f δd (13.5)

where δWd is the virtual work done by f as a result of the imposed virtual
displacement δd. It is the area of the shaded strip shown in Fig. 13.2(a).

13.2.4 Compatibility condition

In the process of imposing δd, if the virtual displacement (at point B) is
compatible with the elongation e (of the spring body), implying that the
spring is not broken, we must have

e=d (13.6)

This is known as the compatibility condition. As will see in later

examples, the compatibility condition can get complicated for general solids
and structures. It is simple for this spring system because there is only one
displacement (1-DOF).
Equation (13.4) can now be re-written as
1
U = Kd2 (13.7)
2
Note here that U is a function of the elongation d (in fact, quadratic in d),
and hence U is differentiable with respect to d. We are essentially treating
d as an independent variable of function U . This is a shift of concept of
functions, and d is called a generalized coordinate.
506 Mechanics of Materials: Formulations and Solutions with Python

13.2.5 Potential energy principle for 1-DOF systems

Inside the spring, in response to the imposed virtual displacement δd, the
strain energy must have a variation, which is denoted as δU . Because U is
differentiable with respect to d, δU can be expressed using its gradient ∂U
∂d as

∂U
δU = δd (13.8)
∂d
We see now how δU relates δd: through the gradient of U with respect
to d.
Substituting Eqs. (13.5) and (13.8) into the virtual work principle
Eq. (13.2), we obtain
 
∂U ∂U
f δd = δd or f− (δd) = 0 (13.9)
∂d ∂d
Now, because δd is arbitrary, for the foregoing equation to hold, the term
in the square bracket must vanish, which gives

∂U
f= (13.10)
∂d
This is the PEP for the spring, which is derived based on the principle of
virtual work, using virtual displacement. Equation (13.10) means that if we
can write the potential energy U in terms of displacement d, we can obtain
the force f that causes the displacement d.

13.2.6 Results of potential energy principle:

Equilibrium equation
Substituting U given in Eq. (13.7) into Eq. (13.10), we obtain:

f = Kd (13.11)

This is the relation between the external force f and the displacement d,
obtained from the potential energy principle. Using Eqs. (13.6) and (13.3),
Eq. (13.11) can be rewritten as

f = Kd = Ke = N (13.12)

This is the equilibrium equation for the spring system. This means that
the potential energy principle is equivalent to the the equilibrium equation,
as long as the compatibility equation is satisfied. The results from the PEP
is shown in Fig. 13.2(c).
 Energy Methods 507

Note that Eq. (13.12) is the same as that we would obtain via the usual
means of using a free-body diagram.
Remark: The above derivation process can be confusing to beginners.
However, one should be convinced that the entire process never used the
equilibrium equation. We used only the PEP, constitutive equation for
the spring, and the compatibility Eq. (13.6). The outcome is indeed the
equilibrium Eq. (13.12) or its equivalent Eq. (13.11).
Using Eqs. (13.4) and (13.3), U can also be written as
1 2
U= N (13.13)
2K
This is the formula for computing the potential energy in a spring in
terms of internal force N .

13.2.7 On compatibility equation

We see also that the compatibility equation is a prerequisite to apply the
PEP (or the virtual work principle using virtual displacements). An intuitive
way to understand this is as follows. Because the virtual work needs to
impose an arbitrary virtual displacement δd, if the internal elongation e is
not responding to δd, such an imposition will have no effect on the spring,
and hence the virtual work principle cannot work!

13.2.8 Clapeyron’s theorem

Equation (13.13) can be re-written as
1 1 1 1
U= N N = Ne = fd (13.14)
2 K  2 2 
work done by f over constant d
e

Here, we used again Eqs. (13.6) and (13.12). Equation (13.14) is known as
Clapeyron’s theorem, which states that “the potential energy of deformation
of a body, which is in equilibrium under a given load, is equal to half the work
done by the external forces computed assuming these forces had remained
constant from the initial state to the final state” [9]. This is practically saying
that the orange triangular area is half the square area shown in Fig. 13.2(c).
We can now extend our findings to all 1-DOF systems.

13.2.9 Formulas of potential energy for 1-DOF systems

As seen in the previous chapters, any one-dimensional (1D) structural mem-
ber can have its relationship of the internal-force with the displacement
508 Mechanics of Materials: Formulations and Solutions with Python

written in the form of Eq. (13.3). For all these linear elastic structure mem-
bers, the strain energy has the following generalized form:
⎧
⎨ K e2 in terms of elongation
2
U= (13.15)
⎩
2K N
1 2 in terms of internal force

where e stands for generalized deformation (including strain, elongation,

twist angle, cross-section rotation, etc.), and N stands for generalized inter-
nal force (including stress, axial force, shear force, moment, torque, etc.). K is
the stiffness and N = Ke. At the material level, K is Young’s modulus E.
The relationships like N = Ke are summarized as follows:

N = Ke, K = string const. Spring with force N , e: elongation

EA
N = Ke, K = Bar (truss) with axial force N , e: elongation
L
GJ
T = KΘ, K = Bar with torque T , Θ: twist angle
L
EI
M = Kθ, K = Beam with moment M , θ: cross-section rotation
L
GA
V = Kdy , K = Beam with shear force V , dy : deflection in y-direction
kL
(13.16)

Note that for beams under shear forces, the stiffness K is with a correction
coefficient k. This is because when a beam is with a shear force on its cross-
section, the resulting shear stress is not uniformly distributed, as discussed
in Chapter 12. It depends on the cross-section shape of the beam. Exact
values of k are not generally available, and hence estimation is needed. Some
often used empirical values for thin beam with different cross-sections are
given as follows [1]:

Rectangular cross-section: k = 1.2

Solid circular cross-section: k = 1.33

(13.17)
Thin-wall circular cross-section: k = 2.0

I-beam, box-beam, channel: k = 1.0

The shear correction coefficient k will be used in this book.

 Energy Methods 509

13.2.10 Formulas of potential energy for various

members
In general, the internal forces in a 1D structure member can be a function of
axial coordinate. In addition, the area properties (such as area, moments of
area, etc.) can also change with the coordinate. Evaluation of the potential
energy needs to be in integral forms. This section presents these formulas for
various 1D structure members, including tension bar, torsion bar, and beams.

13.2.11 Strain energies UN for bars with varying

cross-section
Consider a bar with varying cross-section subjected to axial loads as shown
in Fig. 13.3. It has length L, and a varying cross-section A. Young’s modulus
of the material is E. The loads generate an internal force N in the bar that
can be a function of x. The potential energy can be evaluated as follows.
Consider a small cell dx taken from the bar. Using Eqs. (13.15) and
(13.16), the potential energy for this dx cell can be given as

1 dx 2
dUN = × Internal-force2 = N (13.18)
2 × Stiffness for bars 2EA
The potential energy for the entire bar can then be obtained by integrat-
ing over the length of the bar:
L L
N2
U= dUN = dx
0 0 2EA
(13.19)
N 2L N2 Ke2
= = =
if uniform 2EA 2K or 2
Note that in this equation, E can be a function of x too, such as function-
ally graded materials. Also, if the tapered bar has multiple segments with
different A, the integral should just be piecewise.

Figure 13.3. A tapered bar subject to axial forces in its axis x-direction.
510 Mechanics of Materials: Formulations and Solutions with Python

Figure 13.4. A tapered beam subject to transverse forces in y-direction.

13.2.12 Strain energies for beams with varying

cross-section
Consider a beam with possibly varying cross-section subjected to transverse
loads as shown in Fig. 13.4. It has length L, varying cross-section A, and
2nd moment of area I. The material has Young’s modulus E and shear
modulus G. The loads result in internal moments M and shear force V in
the beam. The potential energy can be evaluated, by considering a small cell
dx taken from the beam.

[Link] From moment, UM

Using Eqs. (13.15) and (13.16), the potential energy for this dx cell resulting
from the moment M on the cross-section of the beam is given as
1 dx
dUM = × Internal-force2 = M2 (13.20)
2 × Stiffness for beams 2EI
The potential energy for the entire beam is obtained by integrating over
the length of the beam:
L L
M2
UM = dUM = dx
0 0 2EI
(13.21)
M 2L M2 Kθ 2
= = =
if uniform 2EI 2K or 2

[Link] From shear force, UV

Using also Eqs. (13.15) and (13.16), the potential energy for this dx cell
resulting from the shear force V on the cross-section of the beam can be
given as
1 kdx 2
dUV = × Internal-force2 = V (13.22)
2 × Stiffness for beams 2GA
 Energy Methods 511

The potential energy for the entire beam can then be obtained by integrating
over the beam length:
L L
kV 2
UV = dUV = dx
0 0 2GA
(13.23)
kV2L V2 Ke2y
= = =
if uniform 2GA 2K or 2
Note that in these equations, E and G can be a function of x. Also, if the
tapered beam has multiple segments with different A, the integral should
just be piecewise.

13.2.13 Strain energies UT for bars under torsion

Consider a bar with possibly varying cross-section subjected to torques in
the x-direction as shown in Fig. 13.5. It has length L, possibly varying cross-
section and a 2nd polar moment of area J. The shear modulus of the material
is G. The loads generate an internal force T in the bar that could be a
function of x. The potential energy can be evaluated as follows.
Consider a small disk cell dx taken from the bar. Using Eqs. (13.15) and
(13.16), the potential energy for this cell can be given as
dx 2
dUT = T (13.24)
2GJ
The potential energy for the entire bar is obtained by integrating over
the length of the bar:
L L
T2
UT = dUT = dx
0 0 2GJ
(13.25)
T 2L T2 KΘ2
= = =
if uniform 2GJ 2K or 2

Figure 13.5. A bar subject to torsional forces in x-direction.

512 Mechanics of Materials: Formulations and Solutions with Python

Note that in this equation, G can be a function of x. If the tapered bar

has multiple segments with different GJ, the integral should be piecewise.

13.2.14 Summary of formulas for U for various

structural members
We can now summarize the formulas as follows:

Struc. member Stiffness U (N ) : Int., uniform U (e) : Int., uniform

N2 Ke2
spring, N − e K UN = 2K 2

L L
EA N2 LN 2 EAe2 EAe2
truss, N − e UN = 2EA dx, 2EA , 2L2
dx,
L 0 0
2L

L L
GJ T2 LT 2 GJΘ2 GJΘ2
Tor. bar, T − Θ L UT = 2GJ dx, 2GJ , 2L2
dx, 2L
0 0
L L
EI M2 LM 2 EIθ 2 EIθ 2
Beam, M − θ L UM = 2EI dx, 2EI , 2L2
dx, 2L
0 0
L L
GA kV 2 kLV 2 GAe2y GAe2y
Beam, V − ey kL UV = 2GA dx, 2GA , 2kL2
dx, 2kL
0 0
(13.26)

Note that if the structure member is uniform (the stiffness and the inter-
nal force do not change with the coordinate), it behaves simply like a spring.
One can get its formulas for U simply by replacing the K in the formulas
for spring. If the stiffness or internal force change with the coordinate, we
need to do an integration, and one of the length L in the uniform formula
should be changed to dx. Once this is understood, the foregoing tabulated
equations are easy to write out.

13.3 Principle of virtual work by virtual force

Consider a mechanically loaded solid constrained without rigid body move-

ment. Assume there are no other effects involved. The principle of virtual
work by virtual force states as follows:

The work done by arbitrary infinitely small virtual forces over the displace-
ment field in a loaded solid in an equilibrium state equals the variation in
complementary energy in the solid caused by the virtual forces.
 Energy Methods 513

Mathematically, it is expressed as
δWf = δC (13.27)
where Wf denotes the work done by the virtual forces δf imposed on the
solid over the displacement field, C stands for the complementary energy
stored in the solid, and δ stands for small variation.
Equation (13.27) can be expressed as
δ (Wf − C) =0 (13.28)
  
total complementary energy

As will be seen in later examples, the stationary point of the total com-
plementary energy gives the compatibility equations of the solid.
Consider the same spring system shown in Fig. 13.1. Assume the spring
is loaded by f at point B resulting in a displacement d also at point B.
The spring system is at state (d, f ). We shall now use the similar trick and
process that led to the PEP in the previous section.
The complementary energy stored in the spring is denoted as C, which
is the shaded area above the black line shown in Fig. 13.6(b).
Let us examine all these energies.

13.3.1 Application of an infinitely small virtual force

Next, at the actually loaded state, we apply (imaginarily) an infinitely small
virtual force δf at point B of the spring system, at which d is measured. The
work done by δf should be
δWf = d δf (13.29)
where δWf is the virtual work done as a result of the virtual force δf . It is
the area of the shaded strip shown in Fig. 13.6(a).

13.3.2 Evaluation of complementary energy

At such a loaded state, inside the spring, there must be an internal force N
and an elongation e. This N ∼ e relationship can be plotted in Fig. 13.6(b).
Thus, some energy is stored within the spring. This time we use the com-
plementary energy denoted as C. It is the shaded triangular area above the
black line in Fig. 13.6(b), and computed using:
1 1 2
C = Ne = N (13.30)
2 2K
Here, we used the constitutive equation for the spring, Eq. (13.3).
514 Mechanics of Materials: Formulations and Solutions with Python

(a) (b)

(c)

Figure 13.6. (a) The force displacement relation at point B of the spring system that is
subjected to small virtual force δf ; The external action: work done by the force δf under
d is evaluated. (b) The relationship of internal-force N and elongation e of the spring body
that stores complementary energy; The internal reaction: variation in strain energy δC
resulting from δf is evaluated. (c) The result from the complementary energy principle.

Equation (13.30) is the formula for computing the complementary energy

in a spring in terms of internal force N . The expression is the same as
Eq. (13.4) because this spring system is linear elastic and the N ∼ e relation
is a straight line. The area of the two triangles below and above the straight
line in Fig. 13.6(b) is the same.
If the system is not linear, Eqs. (13.30) and (13.4) can differ.
 Energy Methods 515

13.3.3 Equilibrium condition

When the virtual force δf is applied, we require the external force f and the
internal force N in the spring satisfying:
f =N (13.31)
This is known as the equilibrium condition. For this spring system, it
is simple. As we will see in later examples, the equilibrium condition can be
complicated for general solids and structures.
Using Eqs. (13.31) and (13.30) can now be re-written as
1 2
C= f (13.32)
2K
Note here that C is a function of the external force f (in fact, quadratic
in f ), and hence C is differentiable with respect to f . We treat f as an
independent variable of function C. Hence, f is also called a generalized
coordinate.

13.3.4 Complementary energy principle for 1-DOF

systems
Inside the spring, in response to the applied virtual force δf , the comple-
mentary energy must have a variation, which is denoted as δC. Because C
is differentiable with respect to f , δC can be expressed using its gradient ∂C
∂f
as
∂C
δC = δf (13.33)
∂f
We see how δC relates δf : through the gradient of C with respect to f .
Substituting Eqs. (13.29) and (13.33) into the virtual work principle
Eq. (13.28), we obtain
 
∂C ∂C
d δf = δf or d− δf = 0 (13.34)
∂f ∂f
Now because δf is arbitrary, for the foregoing equation to hold, the term
in the square bracket must vanish, which gives
∂C
d= (13.35)
∂f
This is the CEP for the spring, which is derived based on the principle of
virtual work using virtual force. Equation (13.35) means that if we can write
the potential energy C in terms of force f , we can obtain the displacement
d that are caused by force f .
516 Mechanics of Materials: Formulations and Solutions with Python

Substituting C given in Eq. (13.32) into Eq. (13.35), we obtain:

f
d= (13.36)
K
This is the formula for computing the displacement at point B, when
the spring is loaded there with f . We obtained this using the principle of
complementary energy.

13.3.5 Compatibility equation

Using the equilibrium condition Eq. (13.31), the foregoing equation becomes
N
d= =e (13.37)
K
This is the compatibility equation for the spring system: the displace-
ment at point B equals the elongation of the spring. This means that the
CEP is equivalent to the compatibility equation, as long as the equilibrium
equation is satisfied. The outcome of the CEP is shown in Fig. 13.6(c).
Remark: Note that the process leading to Eq. (3.37) never used the com-
patibility equation. We used only the complementary energy principle, con-
stitutive equation for the spring, and the equilibrium Eq. (13.31). The
outcome is the compatibility Eq. (13.37).

13.3.6 On equilibrium equation

We see also that the equilibrium equation is a prerequisite to apply the
CEP (or the virtual work principle using virtual forces). An intuitive way
to understand this is as follows. Since the virtual work needs to apply an
arbitrary virtual force δf , if the internal force N is not responding to δf ,
such an application will have no effect on the spring, and hence the virtual
work principle cannot work!

13.3.7 General formulas for displacements of 1-DOF

members
Note that if the problem is linear elastic, C = U , Eq. (13.35) can also be
written as
∂U
d= (13.38)
∂f
This means that if we can write the potential energy in terms of the exter-
nal force on a structure member of 1-DOF, we can obtain the displacement
using Eq. (13.38).
 Energy Methods 517

Equation (13.38) is generalized as follows:

∂U
dgeneralized = (13.39)
∂fgeneralized

where dgeneralized is the generalized displacement including translational

displacements, rotation, twist angle, etc., and fgeneralized is the general-
ized force including forces, moment, torque, etc. The condition is that
dgeneralized and fgeneralized must be an energy conjugate pair, implying that
dgeneralized δfgeneralized produces work done δWf , so that the corresponding
potential energy U can be used. The often used formulas are listed as follows:
∂UN
d= for translational displacements
∂f
∂UT
Θ= for twist angle of torsional bars
∂T
(13.40)
∂UM
θ= for cross-sectional rotation of thin beams
∂M
∂UV
dy = for vertical displacement of thin beams by shear
∂f
In Eq. (13.40), the potential energy U is given in Eq. (13.26).
Equation (13.40) is more frequent used because writing potential energy
in terms of external forces can often be more convenient. We will demonstrate
this with examples.

13.3.8 Example: Cantilever beam subjected to

a concentrated force at its tip
Consider a 1D bar/beam with uniform cross-section and subjected to point
vertical load P and point horizontal load F at its tip, as shown in Fig. 13.7.
The bar/beam has a cross-section area A, and the length is L that is much
larger than its height, and hence can be treated as a thin beam. The 2nd
moment of area of the cross-section is I. Young’s modulus of the material is
E, Poisson’s ratio is ν, and all are given. Use the energy method to complete
the following tasks:

1. Determine the tip displacement in the horizontal direction of the bar.

2. Determine the deflection of the beam at its tip, considering only the
moment in the beam.
3. Determine the rotation of the beam cross-section at its tip, considering
only the moment in the beam.
518 Mechanics of Materials: Formulations and Solutions with Python

Figure 13.7. A 1D bar/beam subjected to a concentrated forces at its tip.

4. Determine the deflection of the beam at its tip, considering only the shear
force in the beam. Assume the shear force correction coefficient k is given.
5. Determine the ratio of the DP from the internal shear force and that from
the internal moment on the cross-section of the beam. Assume the beam
has a rectangular cross-section with height h known. Briefly discuss the
results obtained.
6. Determine the rotation of the beam cross-section at its tip, considering
only the shear force in the beam.

Solution to 1: Determine the tip displacement in the horizontal direction of

the beam. Since only the horizontal force F is applied at the tip of the beam.
We need only consider F . Using the 1st equation in Eq. (13.40) and the 2nd
equation in Eq. (13.26). The formula for computing the tip displacement DF
in the x-direction resulting from the force F becomes

∂U ∂ LN 2 ∂ LN 2 ∂N LN ∂N
DF = = = = (13.41)
∂F ∂F 2EA ∂N 2EA ∂F EA ∂F

Here, we used the chain rule of differentiation [11]. Using now the equi-
librium equation,

N =F (13.42)
∂N
which gives ∂F = 1. Equation (13.41) becomes

LF
DF = (13.43)
EA
This result is exactly that obtained via solving differential equation for
bars (see Chapter 7). Note that if EA is a function of x, we shall use the inte-
gral form of the equation in Eq. (13.26) and carried out with the integration.

Solution to 2: Since only moment is considered, we shall use the 1st equa-
tion in Eq. (13.40) and the 4th integral equation in Eq. (13.26). The formula
 Energy Methods 519

for computing the deflection at tip DP resulting from the force P becomes
L L
∂U ∂ M2 M ∂M
DP = = dx = dx (13.44)
∂P ∂P 0 2EI 0 EI ∂P
Here, we used the chain rule of differentiation. The internal moment M
in the beam is a function of x, and can be computed using the equilibrium
equation:
M = P (L − x) (13.45)
Its derivative with respect to P becomes
∂M
= (L − x) (13.46)
∂P
Substituting Eqs. (13.45) and (13.46) into (13.44), we obtain the solution:
L
L
P P −1  P L3
3
DP = (L − x) dx =
2
(L − x)  = (13.47)
0 EI EI 3  3EI
0

This result is exactly that obtained via solving differential equation for
thin beams (see Chapter 8). Note that if EI are function of x, we shall keep
it in the integrand and carried out with the integration.
Solution to 3: This time we shall also use the 3rd equation in Eq. (13.40)
and the 4th integral equation in Eq. (13.26), but the differentiation should be
with respect to M because we need to compute θ. The process is as follows:
L L
∂U ∂ M2 M ∂M
θP = = dx = dx (13.48)
∂M ∂M 0 2EI 0 EI ∂M
The moment M within the beam is already given in Eq. (13.45). Its
derivative with respect to M itself is simply 1.
Equation (13.48) gives
L
P (L − x) P L2
θP = dx = (13.49)
0 EI 2EI
This solution is also exact, which can be obtained using the codes given
in Chapter 8.
Solution to 4: Since only shear force is considered, we shall use the 4th
equation in Eq. (13.40) and the 5th integral equation in Eq. (13.26). The
formula for computing the deflection at tip DP becomes
L L
∂U ∂ kV 2 kV ∂V
DP = = dx = dx (13.50)
∂P ∂P 0 2GA 0 GA ∂P
520 Mechanics of Materials: Formulations and Solutions with Python

Due to equilibrium, the shear force V = P within the beam, and is

uniform. Its derivative with respect to P is simply 1. Using Eq. (13.50), we
obtain the solution:
L
kP kP L
DP = dx = (13.51)
0 GA GA

where k is the shear force correction coefficient given in Eq. (13.17). Note
for thin beams, L is much larger compared to its cross-sectional dimension
(that affects I and A), the deflection resulting from the shear force is much
smaller than that from the moment given in Eq. (13.47).

Solution to 5: The ratio of DP from the internal shear force and that
from the internal moment on the cross-section of the beam can be computed
using:

kP L P L3 kEI kI
RV M = ÷ =3 = 6(1 + ν) (13.52)
GA 3EI GAL2 AL2

where ν is Poisson’s ratio, and 2G = E/(1 + ν). Assuming the beam cross-
section is rectangular, we have I = bh3 /12, A = bh, we obtain

2
(1 + ν)k h
RV M = (13.53)
2 L

Further, let k = 1.2 and ν = 0.3, we obtain

2
h
RV M = 0.78 (13.54)
L

The result shows clearly that when L  h, the ratio is very small, imply-
ing that the deflection resulting from the shear force is very small because of
the quadratic ratio. If L = 5h (which is considered very thick), the ratio is
about 3%. Therefore, for thin beams, we can often ignore the shear forces and
consider only the moment, when computing the deflection. For thick beams,
the shear force effect becomes more important. Also, the thin beam theory
may not work properly, and alternative theory and method are needed. Or
the problem should be treated as a 2D one, which becomes an advanced
topic in solid mechanics.

Solution to 6: Since the shear force does not result in any change in the
rotation of the beam cross-section, based on the thin beam theory. The
answer to this question is 0.
 Energy Methods 521

13.3.9 Comparison of PEP and CEP, linear elastic

systems
For linear systems, the f ∼ d relation is a straight line, and the N ∼ e
relation is also a straight line, we have U = C: the potential energy and the
complementary energy are the same. We state:

1. If the solid structure system satisfies the compatibility conditions, imply-

ing that the internal deformation and external displacement have no con-
flict (all parts and members staying together, no gap and overlap) when
it is loaded, we use the PEP. The outcome will be automatically the equi-
librium equations. This is the base of numerical methods [6, 7], and very
powerful in dealing with large-scale solid structural systems for approxi-
mated solutions.
2. If the solid structure system satisfies the equilibrium conditions, imply-
ing that all the externally applied forces are balanced by the internal
forces/stresses at a stable state, we use the CEP, and the outcome will
be automatically the compatibility equations. This is particularly useful
in numerical methods for obtaining upper-bound solutions [7].
3. If the solid structure system satisfies both the compatibility and the equi-
librium conditions. We have already the solution, and it is exact.

The comparison is shown schematically in Fig. 13.8.

Figure 13.8. Comparison of the potential energy and the complementary energy prin-
ciples. Linear elastic systems.
522 Mechanics of Materials: Formulations and Solutions with Python

Figure 13.9. Comparison of the potential energy and the complementary energy prin-
ciples. Nonlinear elastic systems.

13.3.10 Comparison of PEP and CEP, nonlinear

systems
For nonlinear systems, the f ∼ d relation or N ∼ e relation or both is curved,
we have in general U = C: the potential energy and the complementary
energy are not the same. In general, these three points given above still
holds, but only locally. Solutions can only be found via iterations [7]. The
comparison is shown schematically in Fig. 13.9.
In the discussions for PEP, it is clear that how to find the potential energy
is the key in applying PEP. This can be a challenge when the solid structure
is higher dimensional and more complicated. The next section derives the
general expression for strain (potential) energy and its density at any point
in a stressed 3D solids. As one can expect, formulation will be a lot more
complicated for 3D solids. Einstein summation will be used, to avoid over
3-page long derivation otherwise.

13.4 Strain energy formulation for 3D solids*

13.4.1 Mathematical derivation

Consider a solid with volume V and an enclosed surfaces S. A Cartesian
coordinate system (xi , i = 1, 2, 3) is used. On the surface, there are external
traction applied on it. The solid is also subjected to body forces over its
 Energy Methods 523

volume, and is in an equilibrium state. We impose a small variation of dis-

placement field (δui , i = 1, 2, 3) that is differentiable over the volume of the
solid. The displacement variation is infinitely small so that the equilibrium
state of the solid is not affected.

[Link] Work done by external forces

The work done by the external forces under (δui ) shall consist of two parts:
the work done by the surface stresses, and the work done by the body
forces. The former, δWds , is evaluated using

δWds = (tni δui )dS (13.55)

S

where tni is the traction on surface S with normal n defined by the direction
cosine ni . Using the traction–stress relation Eq. (3.38), Eq. (13.55) can be
rewritten as

δWds = (nj σji δui )dS (13.56)

S

The above integral is on the surface of volume. It can be converted to

the volume integration using the well-known Green’s divergence theorem
[10], which gives
∂
δWds = (nj σji δui )dS = (σji δui )dV
S V ∂xj
∂σji ∂δui (13.57)
= (δui )dV + σji dV
V ∂xj V ∂xj
  
σji δεij

Here, we used the the symmetry of the stresses, and the strain–
displacement relation Eq. (4.14) because δui is assumed differentiable.
The work done by body forces, δWdb , is evaluated using

δWdb = (bi δui )dV (13.58)

V

where bi is the body forces applied over volume V . Adding up Eqs. (13.57)
and (13.58), we obtain total work done:
⎛ ⎞
⎜ ⎟
δWd = δWds + δWdb = ⎝ (σji,j + bi ) δui + σji δεij ⎠ dV = (σji δεij )dV
V    V
equilibrium=0
(13.59)
524 Mechanics of Materials: Formulations and Solutions with Python

Here, we used the equilibrium equation at any point inside the volume.
This is because of our assumption that the solid is in equilibrium when δui
is imposed.

[Link] Strain energy expression

Using the virtual work principle given in Eq. (13.1), we found the expression
for the variation of the strain energy as
δU = (σij δεij )dV
V
  
δWd

= (σxx δεxx + σyy δεyy + σzz δεzz + 2σyz δεyz + 2σxz δεxz + 2σxy δεxy ) dV
V
(13.60)
Here, we used the symmetry of the stress and strain components.
Finally, we obtained the widely used important expression for computing
the variation of the strain energy density (the strain energy per unit
volume):
δU0 = σij δεij
= σxx δεxx + σyy δεyy + σzz δεzz + 2σyz δεyz + 2σxz δεxz + 2σxy δεxy
  δ
=σ ε
(13.61)
where σ  and ε are the stress and strain vectors defined in Eq. (5.60). The
variation of the strain energy density is now written as the dot product of
the stress vector and the variation of the strain vector.
It is clear that δU0 is a sum of the energy of each of the 9 stress and strain
components. Each contribution is evaluated with individual term σij δεij
(i, j = 1, 2, 3, no summation). This proves that σij and εij are energy con-
jugate: σij and εij are in the same direction, and hence their product gives
the energy.
Note that if the stress-strain pair is not conjugate, implying that they are
not in the same direction. One needs to project one of them onto the other
and use the projected component to compute the energy. If the stress and
strain are orthogonal, the energy by them is zero, regardless of the values of
the stress and strain components.
Also because our stresses and strains are all expressed in orthogonal coor-
dinates, the stress component in one axis direction is orthogonal to those
in other axis-directions. Therefore, the energy can be written in individual
σij δεij terms, and then summed up. One stress component is not coupled
with any other ones in energy evaluation, as shown in Eq. (13.60).
 Energy Methods 525

[Link] Compatibility conditions

Note that the strains are subjected to variation. This implies that these
strains must be compatible with the displacements, implying that these com-
patibility equations defined in the last section of Chapter 4 must be satisfied
by these strains. This is similar to the discussion given in Section 13.2.7 for
1-DOF systems.
It turns out that in numerical methods based on domain discretization,
we can find ways to construct displacement fields for 2D and 3D solids and
various structures, while ensuring the compatibility (known as the admissible
displacements). PEP can then be applied to establish equilibrium equations
and then find approximated solutions in terms of displacements as well as
in stresses for large-scale solid structure systems. Detailed discussion is out
of the scope of this book. Interest readers may refer to [6–8].

13.4.2 Pictorial derivation

An alternative and easier derivation of Eq. (13.61) is a pictorial approach to
show that σij and εij are in the same direction. This is done as follows.
First, for all the normal stresses and strains, it is obvious that σii and εii
are in the same direction because εii gives displacement in the i-direction
which is the same as the direction of σii .
For the shear stress and strain, let us consider σxy ∼ εxy pair and σyx ∼
εyx pair. We use a picture shown in Fig. 13.10, where a small square 2D
solid abdc has a small deformation becoming ab d c . The displacement by
εxy is εxy dx and is in the y-direction, which is the same as that of stress
component σxy . Therefore, σxy and εxy are energy conjugate, and the energy
can be evaluated by their product.
The displacement by εyx is εyx dy, and is in the x-direction, which is the
same as that of σyx . Therefore, σyx and εyx are also energy conjugate, and
the energy can be evaluated by its product.
The same arguments can be made for other shear components. This
proves that the strain energy density can be directly written out in the
form of Eq. (13.61).

13.4.3 Strain energy density for 1D problem

With the expression for 3D solid, we can re-visit the 1D problems. This leads
to some familiar formulas given earlier, and also some general formulas.
526 Mechanics of Materials: Formulations and Solutions with Python

Figure 13.10. A small square solid abdc is deformed to ab d c , by pure shear stresses.

[Link] General formulas

For 1D problems, there is only one stress or strain. Note that variation and
differentiation carry the same meaning if both treat ε as the independent
variable. Equation (13.61) can be expressed as

dU0 = σdε (13.62)

or
dU0
σ= (13.63)
dε
This means that if the strain energy density is given as a function of the
strain, one can easily obtain the stress via differentiation with respect to the
strain. This is in fact the potential energy principle at material level.
Also, Eq. (13.62) gives
ε
U0 = σ(ε )dε (13.64)
0

where the prime denotes an integral variable, which disappears after the
integration is done [11]. The strain energy density is the shaded orange area
in Fig. 13.11(a).
Consider a 1D stress problem for bars with uniform cross-section of area
A subject to axial loads. The strain (potential) energy of the bar at the
 Energy Methods 527

(a) (b)

Figure 13.11. Strain energy density U0 and complementary energy density C0 at a point
in a stressed solid. (a) Nonlinear stress-strain relation; (b) linear stress-strain relation.

structural member level can be evaluated using Eq. (13.64):

L ε
U= U0 dV = σdε Adx

V 0 0
dV
⎛ ⎞ (13.65)
L ε L ε
= ⎝ Aσ dε ⎠ dx = N dε dx

0 0 0 0
N

where N is the internal force in the bar, L is the length, V is the volume.
If N does not change along the bar, implying that there is no body force on
the bar, the integral over the bar length can move inside, which gives
⎛ ⎞
ε ⎜ L ⎟  ε
U= U0 dV = N⎜
⎝ dx⎟⎠ dε = N dLε
 (13.66)
V 0
  
0 0
de
L

or
e
U= N de (13.67)
0

where e is the elongation of the bar.

Note that the condition for Eq. (13.67) is that the internal force N does
not change along the bar. If the internal force N changes along the bar,
Eq. (13.65) should be used.
528 Mechanics of Materials: Formulations and Solutions with Python

When the material is linear elastic, U0 is a linear stress-strain curve.

Equation (13.64) gives

1
U0 = σε (13.68)
2

which is the shaded orange area in Fig. 13.11(b).

In this case, the strain energy for a 1D linear elastic bar with Young’s
modulus E becomes
L L
1 1
U= U0 dV = σε Adx
 = σA εdx
V 0 2 0 2 
dV N
(13.69)
L L
1 N2
= N εdx = dx
0 2 0 2EA

In this equation, E, A, and E can all be a function of x. This is Eq. (13.19)

obtained earlier.

[Link] Formulas for linear elastic materials

Consider first 1D stress problem for uniform bars with Young’s modulus E,
we have σ = Eε, and thus Eq. (13.64) becomes

ε
1 2
U0 = Eε dε = Eε (13.70)
0 2

where E is the stiffness of the material. Using σ = Eε, Eq. (13.70) can also
be written in stress.

1 1 2
U0 = σε = σ (13.71)
2 2E

Using the first part of the foregoing equation, it is trivial to confirm

Eq. (13.63).
For 1D uniform bars made of linear elastic materials with uniform internal
force N , we have

EA
N= e = Ke (13.72)
L

K
 Energy Methods 529

where K is the stiffness of the bar. The bar practically behaves like a spring.
Equation (13.67) becomes
e
EA EA 2 K 2
U= e de = e = e (13.73)
0 L 2L 2
or in terms of force:
1 2
U= N (13.74)
2K
This is also the expression of strain energy stored in an elastic spring,
when it is subjected to force N . This is the same as Eq. (13.15).

13.4.4 Various formulas for strain energy density for

3D solids
For 3D linear elastic solids, the strain energy density can be written in various
forms. We write all these different formulas together for easy comparison and
future reference:
ε
U0 = σ(ε )dε General definition. Prime stands for integration variable.
0
1 1
= σij εij = (σxx εxx + σyy εyy + σzz εzz + σyz γyz + σxz γxz + σxy γxy )
2 2
1  2
= σ + σyy2
+ σzz
2
− 2ν(σyy σzz + σxx σzz + σxx σyy )
2E xx

− 2(1 + ν)(σyz
2
+ σxz
2
+ σxy
2
)
1  2 
= σ1 + σ22 + σ32 − 2ν(σ2 σ3 + σ1 σ3 + σ1 σ2 )
2E
(σ1 + σ2 + σ3 )2 (σ2 − σ3 )2 + (σ1 − σ3 )2 + (σ1 − σ2 )2
= + (13.75)
18K 12G
1  2 
= I1 − 2(1 + ν)I2
2E
1  
= λ (εxx + εyy + εzz )2 + G ε2xx + ε2yy + ε2zz + 2ε2yz + 2ε2xz + 2ε2xy
2
1 
= c11 ε21 + c22 ε22 + c33 ε23 + (c23 ε2 ε3 + c13 ε1 ε3 + c12 ε1 ε2 )
2
1  
= λ (ε1 + ε2 + ε3 )2 + G ε21 + ε22 + ε23
2
1
= λ + G I¯12 − 2GI¯2
2
530 Mechanics of Materials: Formulations and Solutions with Python

13.5 Potential energy principle for structures

13.5.1 Virtual work via virtual displacement for

structures
Consider a structure member that is properly constrained and loaded,
and is in equilibrium. The generalized displacements are denoted as
(D1 , D2 , . . . , Dn ). These can be translational and rotational. For beams, for
example, it is the deflection of the beam axis or the rotation of the cross-
section of the beam. The number of the generalized displacements is the
number of degrees of freedom (DoFs). All these displacements must be com-
patible, ensuring the integrity of the structure. In addition, the loads on the
structure are also treated as generalized loads denoted as (P1 , P2 , . . . , Pn )
that result in displacements Di in an equilibrium state. For beams, the gen-
eralized loads include transverse load and moments.
Principle of virtual work is also a most general form of energy principle
for solids and structures. It can be used to derive many formulations using
energies in the system, including the principle of potential energy and the
principle of complementary energy, as we did for the 1D systems. When
virtual displacements are used, it leads to the principle of total potential
energy. The process is as follows.
Consider a set of (infinitesimal) arbitrary independent virtual displace-
ments (δD1 , δD2 , . . . , δDn ), subjected to the boundary and compatibility
conditions. The virtual displacements are then imposed on the structure.
Since the structure is loaded with Pi , these forces will do work under the
virtual displacements, which is denoted as δWP and can be evaluated as
δWp = P1 δD1 + P2 δD2 + · · · + Pn δDn = Pi δDi (13.76)
The virtual displacements δDi will result in a virtual strain field. Also,
within the loaded structure, there exists a stress field. Such a stress field
will do virtual work under the virtual strain field resulting from δDi , which
should be the strain energy change caused by δDi . Therefore, we have
∂U ∂U ∂U ∂U
δU = δD1 + δD2 + · · · + δDn = δDi (13.77)
∂D1 ∂D2 ∂Dn ∂Di

13.5.2 Formulas of potential energy principle for

structures
It is also called PEP.
Substituting Eqs. (13.76) and (13.77) into (13.2) gives,
∂U ∂U ∂U
P1 δD1 + P2 δD2 + · · · + Pn δDn = δD1 + δD2 + · · · + δDn
∂D1 ∂D2 ∂Dn
(13.78)
 Energy Methods 531

which is re-grouped to give

∂U ∂U ∂U
P1 − δD1 + P2 − δD2 + · · · + Pn − δDn = 0
∂D1 ∂D2 ∂Dn
(13.79)

Now because each of the virtual displacements δDi is arbitrary and δDi
are independent of each other. The foregoing equation holds, only if

∂U
Pi = for all i = 1, 2, . . . , n (13.80)
∂Di

This states that if the potential energy U can be written as a function

of generalized displacements (D1 , D2 , . . . , Dn ) that are compatible in the
structure, the load Pi at the same location and direction of Di can be found
using Eq. (13.80).
Equation (13.80) is known also as Castigliano’s theorem. It is in fact
the potential energy principle at the structural level, in comparison with
Eq. (13.63) that is for material level. For 1DoF problems, Eq. (13.80) is
Eq. (13.10).
All these loads Pi found using Eq. (13.80) will be in equilibrium, as will
be demonstrated in the example sections.

13.5.3 Example: A two-bar truss subjected to two loads

[Link] Setting of the problem

Consider a truss structure with two bars: bar-1 (AB) and bar-2 (AC) of
lengths l1 and l2 , respectively. The truss is fixed to a rigid wall at points
A and C, and is loaded with horizontal (x-direction) force P1 and vertical
(y-direction) force P2 at point B, as shown in Fig. 13.12. Thus, point B
undergoes finite displacements with components u1 and u2 , respectively, in
x- and y-directions. The lengths of these two bars become L1 and L2 .
The cross-section area of bar AB is A1 and that of bar AC is A2 . Their cor-
responding Young’s moduli are E1 and E2 . Assume these two bars remaining
linearly elastic during the deformation.

(a) Derive formulas for forces P1 and P2 in terms of displacements u1 and

u2 , based on the potential energy principle.
(b) Using the formulas derived in part (a), determine the values of P1 and P2 .
Assume E1l1A1 = k1 = 3000 N/m and E2l2A2 = k2 = 4300 N/m, b = 0.4 m,
h = 0.4 m, and set u1 = 0.05 m and u2 = 0.03 m.
532 Mechanics of Materials: Formulations and Solutions with Python

Figure 13.12. A truss structure with two bars AB and AC of initial lengths l1 and l2 ,
and deformed lengths L1 and L2 , which can all be computed using the geometric relations,
ensuring compatibility (two bars remain supported and connected).

(c) Consider the equilibrium at the displaced position B* and verify the
results of part (b).
(d) Considering small displacements (u, v  b, h), linearize the formulas for
P1 and P2 derived in part (a).

[Link] Solutions to the problem

The initial lengths of these two bars l1 and l2 , is calculated as

l1 = b; l2 = b2 + h2 (13.81)
Solution to question (a): Denote the elongations of these two bars, respec-
tively, as e1 and e2 . Using the geometry of the truss shown in Fig. 13.12, the
lengths of these bars after deformation become

L1 = l1 + e1 = (b + u1 )2 + u22
 (13.82)
L2 = l2 + e2 = (b + u1 ) + (h − u2 )
2 2

Solving the foregoing equation for e1 and e2 gives,


e1 = L1 − l1 = (b + u1 )2 + u22 − l1
 (13.83)
e2 = L2 − l2 = (b + u1 )2 + (h − u2 )2 − l2
These relations between the displacements and elongations ensure that
the compatibility of the displacements so that these bar members of the
 Energy Methods 533

structures are always connected during the deformation. We can thus apply
the potential energy principle.
Since the bars remain linearly elastic, the elongations of these two bars
can be calculated using
Ni li
ei = for i = 1, 2 (13.84)
Ei Ai
where N1 and N2 are the internal forces in these two bars. The strain energy
can then be evaluated using
1 Ei Ai 2
Ui = Ni ei = e for i = 1, 2 (13.85)
2 2li i
where U1 and U2 are the strain energies for bar-1 and bar-2. The total strain
energy U for the whole truss structure becomes
E1 A1 2 E2 A2 2
U = U1 + U2 = e + e (13.86)
2l1 1 2l2 2
We have written the potential energy in terms of the elongations.
Using the PEP Eq. (13.80), we obtain:
∂U E1 A1 ∂e1 E2 A2 ∂e2
P1 = = e1 + e2
∂u1 l1 ∂u1 l2 ∂u1
(13.87)
∂U E1 A1 ∂e1 E2 A2 ∂e2
P2 = = e1 + e2
∂u2 l1 ∂u2 l2 ∂u2
in which ei is given in Eq. (13.83). Using this, the partial derivatives of ei
with respect to ui can be obtained as
∂e1 (b + u1 ) (b + u1 )
= =
∂u1 L1 (b + u1 )2 + u22
∂e2 (b + u1 ) (b + u1 )
= =
∂u1 L2 (b + u1 )2 + (h − u2 )2
(13.88)
∂e1 u2 u2
= =
∂u2 L1 (b + u1 )2 + u22
∂e2 h − u2 (h − u2 )
=− = −
∂u2 L2 (b + u1 )2 + (h − u2 )2
Solution to question (b): Using the formulas derived in part (a), deter-
mine the values of P1 and P2 . Assume E1l1A1 = K1 = 3000 N/m and
E 2 A2
l2 = K2 = 4300 N/m, b = 0.4 m, h = 0.4 m, and set u1 = 0.05 m
and u2 = 0.03 m. We write the following code to do this.
534 Mechanics of Materials: Formulations and Solutions with Python

1 # External forces
2 K1 = 2.0e3; K2 = 3.0e3
3 b = 0.4; h = 0.4
4 u1 = 0.05e-3; u2 = 0.03e-3
5
6 l1 = b; l2 = [Link](b**2 + h**2) # initial lengths
7
8 L1 = [Link]((b+u1)**2 + u2**2) # deformed lengths
9 L2 = [Link]((b+u1)**2 + (h-u2)**2)
10 print(f'Initial lengths of two bars: {l1:0.6e},{l2:0.6e}(m)')
11 print(f'Deformed lengths of two bars: {L1:0.6e},{L2:0.6e}(m)')
12
13 e1 = L1 - l1 # elongations
14 e2 = L2 - l2
15 print(f'Elongations of two bars: {e1:0.6e},{e2:0.6e}(m)')
16
17 pe1u1 = (b+u1)/L1 # partial derivatives of elongations
18 pe1u2 = u2 /L1
19 pe2u1 = (b+u1)/L2
20 pe2u2 =-(h-u2)/L2
21
22 P1 = K1*e1*pe1u1 + K2*e2*pe2u1
23 P2 = K1*e1*pe1u2 + K2*e2*pe2u2
24
25 print(f'Two forces at point B: {P1:0.6e},{P2:0.6e}(N)')

Initial lengths of two bars: 4.000000e-01,5.656854e-01(m)

Deformed lengths of two bars: 4.000500e-01,5.656996e-01(m)
Elongations of two bars: 5.000112e-05,1.414496e-05(m)
Two forces at point B: 1.300112e-01,-2.999550e-02(N)

Note that when using PEP, the displacements can be large as shown.
Solution to question (c): Use the equilibrium of at the displaced position
B* and verify the results of part (b).
Using the values of P1 and P2 obtained in (b), the values of the internal
normal forces N1 and N2 in these two bars, as shown in Fig. 13.12, can be
found using the following code.
1 N1 = e1*K1
2 N2 = e2*K2
3 print(f'The internal forces in two bars: {N1:0.4e}, {N2:0.4e}(N)')

The internal forces in two bars: 1.0000e-01, 4.2435e-02(N)

The sine and cosine values of θ1 and θ2 (see, Fig. 13.12) can be found
using:
1 sin 1 = u2 /L1; cos 1 = (b+u1)/L1
2 sin 2 = (h-u2)/L2; cos 2 = (b+u1)/L2
 Energy Methods 535

Using the free-body diagram for point B*, shown in Fig. 13.12, we can
now establish the equilibrium equations.

P1 − N1 cos(θ1 ) − N2 cos(θ2 ) = 0 all forces in x-direction

(13.89)
P2 − N1 sin(θ1 ) + N2 sin(θ2 ) = 0 all forces in y-direction

or
 
P1 − N1 b+u
L1
1
+ N b+u1
2 L2 = 0 all forces in x-direction
  
Nx
  (13.90)
P2 − N1 Lu21 − N2 h−u
L2
2
= 0 all forces in y-direction
  
Ny

The total force from N1 and N2 in the x-direction becomes

1 N_x = N1*cos 1 + N2*cos 2

2 print(f'Total forces of N1 & N2 in x-direction: {N_x:0.6e}(N)')

Total forces of N1 & N2 in x-direction: 1.300112e-01(N)

The total force from N1 and N2 in the y-direction becomes

1 N_y = N1*sin 1 - N2*sin 2

2 print(sin 1, sin 2)
3 print(f'Total force of N1 & N2 in y-direction: {N_y:0.6e}(N)')

7.499062596087012e-05 0.707036068741148
Total force of N1 & N2 in y-direction: -2.999550e-02(N)

We found that Nx = P1 and Ny = P2 . This confirms that the PEP results

satisfy automatically the equilibrium equations. The total potential energy
principle is in place of the equilibrium equations. This finding is extremely
useful for complicated structural systems, for which equilibrium equations
are difficult to establish. One can then use the total PEP to obtain (approx-
imate) equilibrium equations. The widely used finite element method [6] is,
in fact, established based on the total potential energy principle.
Solution to question (d): Considering small displacements (u, v  b, h),
linearize the formulas for P1 and P2 derived in part (a).
Using the binomial expansion, for an arbitrary small variable α, we have
1 1 1
(1 + α) 2 ≈ 1 + α − α2 + · · · (13.91)
2 8
536 Mechanics of Materials: Formulations and Solutions with Python

Equations in (13.82) can be linearized as



u2 + u22 bu1
L1 = (b + u1 )2 + u22 = l1 1 + 2bu1 + 1 ≈ l1 +
l1 l1

 bu1 hu2 bu1 hu2
L2 = (b + u1 )2 + (h − u2 )2 ≈ l2 1 + 2 −2 = l2 + −
l2 l2 l2 l2
(13.92)

In the foregoing derivation, we first remove the second-order terms in the

square root, and then apply the binomial formula keeping only up to the
first order.
Using Eqs. (13.92) and (13.83), the elongations are linearized:

bu1
e1 = L1 − l1 = (b + u1 )2 + u22 − l1 = (13.93)
l1
 bu1 − hu2
e2 = L2 − l2 = (b + u1 )2 + (h − u2 )2 − l2 =
l2
Further, the partial derivatives of ei with respect to ui , given in
Eq. (13.88) are linearized as
∂e1 b ∂e2 b
= , =
∂u1 L1 ∂u1 L2
(13.94)
∂e1 ∂e2 h
= 0, =−
∂u2 ∂u2 L2
Substituting Eqs. (13.93) and (13.94) into Eq. (13.87), we obtain the
linearized expression for forces as follows:

E1 A1 b2 E2 A2 b
P1 = u1 + (bu1 − hu2 )
l1
3 l23
(13.95)
E2 A2 h
P2 = − (bu1 − hu2 )
l23
This is the set of two equilibrium equations for this truss structure.
It works for small displacements compared to the length of the bars. The
following code computes linearized forces for given small displacements.

1 P1_l = K1*b**2*u1/l1**2 + K2*b*(b*u1-h*u2)/l2**2

2 P2_l =-K2*h*(b*u1-h*u2)/l2**2
3 print(f'Two forces at point B (linearized): {P1_l:0.6e},{P2_l:0.6e}(N)')

Two forces at point B (linearized): 1.300000e-01,-3.000000e-02(N)

 Energy Methods 537

It is found that the forces are very close to those using the non-linearized
formulas, which were (1.300112e-01, −2.999550e-02).
Readers may set larger displacements u1 and u2 , and re-compute the
forces using the foregoing codes. One should find that when the displace-
ments get larger, the results from the linearized and non-linearized formulas
depart.

13.6 Complementary energy formulation for 3D solids*

13.6.1 General formulations

Using the proven fact that the stress and strain components in orthogonal
coordinates are energy conjugate, the variation of complementary energy C
can be written out directly as
δC = (εij δσij ) dV
V

= (εxx δσxx + εyy δσyy + εzz δσzz + 2εyz δσyz + 2εxz δσxz + 2εxy δσxy ) dV
V
(13.96)
The variation of the complementary energy density (the complementary
energy per unit volume) is given as
δC0 = εxx δσxx + εyy δσyy + εzzδσzz + 2εyz δσyz + 2εxz δσxz + 2εxy δσxy
= ε δ
σ
(13.97)
where σ and ε are the stress and strain vectors defined in Eq. (5.60). The
variation of the strain energy density is now written as the dot product of
the strain vector and the variation of the stress vector.

13.6.2 Equilibrium conditions

Note that the stresses are subjected to variation. This implies that these
stresses must be in equilibrium with the externally applied forces over the
solid body and over the boundary surface. These equilibrium equations and
the stress boundary conditions derived in Chapter 6 for solids must be sat-
isfied by these stresses, when the CEP is applied. This is similar to the
discussion given in Section 13.3.6 for 1-DOF systems.
It turns out finding stresses that satisfy equilibrium equations is usu-
ally quite difficult for higher-dimensional solids and complex structures.
Therefore, using CEP for such systems is much more challenging. On the
other hand, methods based on CEP are particularly useful for obtaining
upper-bound solutions. Therefore, methods based on PEP with so-called
strain smoothing operations can have equilibrium equations satisfied locally,
538 Mechanics of Materials: Formulations and Solutions with Python

so as to soften the model and produce upper bound solutions. Detailed dis-
cussion is out of the scope of this book. Interested readers may refer to
refs. [7, 8].

13.6.3 Complementary energy 1D problems

For 1D problems, there is only one stress or strain. The variation and differen-
tiation carry the same meaning if both treat the stress σ as the independent
variable. Equation (13.97) can be expressed as

dC0 = εdσ (13.98)

or
dC0
ε= (13.99)
dσ
This means that if the complementary energy density is given as a func-
tion of the stress, one can easily obtain the strain via differentiation. This is
in fact the complementary energy principle at material level.
Also, Eq. (13.98) gives
σ
C0 = ε(σ  )dσ  (13.100)
0

The complementary energy density is the shaded blue area in

Fig. 13.11(a).
Consider a 1D stress problem for bars with uniform cross-section subject
to an axial load. The complementary energy of the bar should be its density
times its volume. Thus, at the structural member level, Eq. (13.100) becomes
L σ
C= C0 dV = εdσ  Adx

V 0 0
dV
⎛ ⎞
⎛ ⎞
L N N ⎜ L ⎟
⎝ ε dAσ ⎠ ⎜ εdx⎟ 
=   dx = ⎝ ⎠ dN (13.101)
0 0
dN 
0
 0  
e

or
N
C= edN  (13.102)
0
 Energy Methods 539

Note that in deriving this equation, we did not impose any condition
to the material. This means that Eq. (13.102) is applicable to nonlinear
materials. We will use it in the example section for bars with nonlinear
stress-strain relations.
If the material is linear elastic, we shall have
σ
1
C0 = ε(σ  )dσ  = σε (13.103)
0 2
which is the shaded blue area in Fig. 13.11(b). Also, in this case we have
C0 = U0 . The shaded orange and blue areas are the same, as shown in
Fig. 13.11(b).

13.7 Complementary energy principle for structures*

13.7.1 Virtual work via virtual forces for structures

When virtual forces are used in the principle of virtual work, it leads to
the principle of complementary energy, similar to the 1D problems discussed
earlier. The process for a structure is as follows.
Consider a set of (infinitesimal) arbitrary independent virtual forces
(δF1 , δF2 , . . . , δFn ), subjected to the equilibrium conditions. These virtual
forces are then applied to the loaded structure, which has a displacement
field Di . These virtual forces δFi will do work under Di , which is denoted as
δWF and can be expressed as
δWF = D1 δF1 + D2 δF2 + · · · + Dn δFn = Di δFi (13.104)
The virtual forces δFi will result in a virtual stress field. Also, within
the loaded structure there exists a strain field. The virtual stresses will do
virtual work under the strain field, which is the change in the complementary
energy caused by δFi . Therefore, we have
∂C ∂C ∂C ∂C
δC = δF1 + δF2 + · · · + δFn = δFi (13.105)
∂F1 ∂F2 ∂Fn ∂Fi

13.7.2 Formulas of complementary energy principle for

structures
Substituting Eqs. (13.104) and (13.105) into the foregoing equation gives,
∂C ∂C ∂C
D1 δF1 + D2 δF2 + · · · + Dn δFn = δF1 + δF2 + · · · + δFn
∂F1 ∂F2 ∂Fn
(13.106)
540 Mechanics of Materials: Formulations and Solutions with Python

which is re-grouped to give

∂C ∂C ∂C
D1 − δF1 + D2 − δF2 + · · · + Dn − δFn = 0
∂F1 ∂F2 ∂Fn
(13.107)
Now because each of the virtual force δFi is arbitrary and δFi are inde-
pendent of each other. The foregoing equation holds only if
∂C
Di = for all i = 1, 2, . . . , n (13.108)
∂Fi
This states that if the complementary energy C can be written as a
function of generalized forces (F1 , F2 , . . . , Fn ) acting on the structure, the
displacement Di at the location and direction of Fi can be found using
Eq. (13.108).
Equation (13.108) is in fact the CEP at the structural level, in compari-
son with Eq. (13.99) that is for material level. For 1DoF structural members,
Eq. (13.108) is Eq. (13.35).
We reiterate that all these forces Fi must be in equilibrium when using
Eq. (13.108).
As the results of CEP, all these displacement Di obtained using
Eq. (13.108) will be compatible, as will be demonstrated in example sec-
tion.
Note also that when the load is a moment M , the corresponding defor-
mation will be rotation, and we shall have
∂C
θi = for all i = 1, 2, . . . , p (13.109)
∂Mi
where p is the number of moments on the structure.
Let us look at some examples.

13.7.3 Example: Linear spring–mass system

Consider a spring–mass system with two linear elastic springs connecting
sequentially two rigid masses. The left end of the first spring is fixed to a
wall. The spring constants are k1 and k2 , respectively, for the two springs.
Two forces F1 and F2 are applied, respectively, on these two masses, as
shown in Fig. 13.13. These two masses will have displacements, respectively,
denoted as D1 and D2 .
1. Using the PEP, determine the forces F1 and F2 , assuming k1 , k2 , D1 , and
D2 are given.
 Energy Methods 541

Figure 13.13. A springmass system with two springs and two rigid masses.

2. Using the CEP, determine the displacements D1 and D2 , assuming k1 ,

k2 , F1 , and F2 are given.

Solution to 1: For the PEP, Eq. (13.80), strain (potential) energy must
be written in terms of displacements. The expression for the strain energy
can be found using the equation for spring given in Eq. (13.26), for which
we need to find the elongations for these two springs: e1 and e2 . From the
geometry given in Fig. 13.13, we have the following compatibility equations:

e1 = D1 ; e2 = D2 − D1 (13.110)

The strain energy stored in the system of two springs can now be written
in terms of displacements, as follows:
k1 e21 k2 e22
U = U1 + U2 = +
2 2 (13.111)
k1 k2
= D12 + (D2 − D1 )2
2 2
Next, using Eq. (13.80), we obtain:
∂U
F1 = ⇒ F1 = k1 D1 − k2 (D2 − D1 )
∂D1
(13.112)
∂U
F2 = ⇒ F2 = k2 (D2 − D1 )
∂D2
Substituting Eq. (13.110) into (13.112), we obtain:

F1 = k1 D1 = k1 e1 − k2 e2 = N1 − N2
(13.113)
F2 = k2 (D2 − D1 ) = k2 e2 = N2
where Ni is the internal force in the ith spring. This is in fact the equilibrium
equations that one can obtain using free-body diagram for each of the masses.
It is seen again that the potential energy principle produces the equilibrium
equations. In the process of obtaining the results Eq. (13.112), we must use
the compatibility equations Eq. (13.110).
542 Mechanics of Materials: Formulations and Solutions with Python

Solution to 2: To use the CEP, Eq. (13.108), we need to write the comple-
mentary energy principle in terms of forces in the springs. Since the strings
are linear elastic, the expression for the complementary energy is the same
as the potential energy. However, we need to write it in terms of the internal
forces. This can be done using the equation for spring given in Eq. (13.26),
which gives,
N12 N2
C = U = U1 + U2 = + 2 (13.114)
2k1 2k2
Note here that N1 and N2 are the internal forces in these two springs.
These forces should be in equilibrium with the externally applied forces F1
and F2 . These equilibrium equations should be:

N1 = F1 + F2 ; N2 = F2 (13.115)

Substituting Eq. (13.115) into (13.114), we obtain:

(F1 + F2 )2 F2
C= + 2 (13.116)
2k1 2k2
With the help of the equilibrium equations, the complementary energy
is now in terms of external forces. We are ready to apply the CEP. Using
Eq. (13.108), we arrive at:
∂C F1 + F2
D1 = ⇒ D1 = (13.117)
∂F1 k1
∂C F1 + F2 F2
D2 = ⇒ D2 = +
∂F2 k1 k2
Careful examination should find that the foregoing equation is identical
to Eq. (13.112).
Substituting Eq. (13.115) into (13.117), we obtain:
N1
D1 = = e1
k1
(13.118)
N1 N2
D2 = + = e1 + e2
k1 k2
These are compatibility equations: the displacement at mass-1 is the elon-
gation of spring-1, and the displacement at mass-2 is the sum of the elon-
gations of spring-1 and spring-2. The deformations of the springs agree with
the displacement solutions. It is seen again that the complementary energy
principle produces the compatibility equations. In the process of obtaining
the results Eq. (13.117), we must use the equilibrium equations Eq. (13.115).
 Energy Methods 543

13.7.4 Example: Nonlinear spring–mass system

Consider the same spring–mass system studied in the previous example, but
assume the two spring constants becoming nonlinear. The force-elongation
relation or the constitutive equation is given by [1]:

N = ki e2 for i = 1, 2 (13.119)

where k1 and k2 are given constants.

Determine the displacements D1 and D2 , assuming k1 , k2 , F1 , and F2 are
given.
Solution: Since the force-elongation relationship is nonlinear, the comple-
mentary energy principle should be used.
Using Eq. (13.119), we found e in terms of the internal force N for each
of the string:
1/2
N
e= for i = 1, 2 (13.120)
ki
Using integral Eq. (13.102), the complementary energy becomes
N1 N2
C = C1 + C2 = edN + edN
0 0
 (13.121)
N1 1/2 N2 1/2 3/2 3/2
N N 2 N1 N2
= dN + dN = +
0 k1 0 k2 3 k1
1/2 1/2
k2
Next, substituting Eq. (13.115) into (13.121), we obtain:

3/2
2 (F1 + F2 )3/2 F2
C= 1/2
+ 1/2 (13.122)
3 k k 1 2

The complementary energy is now in terms of external forces. Using

Eq. (13.108), we arrive at:
1/2
∂C F1 + F2
D1 = ⇒ D1 =
∂F1 k1
(13.123)
1/2 1/2
∂C F1 + F2 F2
D2 = ⇒ D2 = +
∂F2 k1 k2
We see from this example that the CEP works well for nonlinear consti-
tutive equations. Using the equilibrium equations Eq. (13.115), it is easy to
see that Eq. (13.123) is the compatibility equation.
544 Mechanics of Materials: Formulations and Solutions with Python

13.7.5 Example: A two-bar truss subjected to two loads

[Link] Setting of the problem
Consider the same setting of the previous example shown in Fig. 13.12. In
this example, we assume the displacement is small. We would like to complete
the following tasks, using the complementary energy principle.

(a) Derive formulas for displacements u1 and u2 in terms of forces P1 and P2 .

(b) Given the same values for b, h, K1 , K2 , P1 , and P2 as the previous
example, compute the two displacement components u1 and u2 using
the formulas obtained in part (a).

[Link] Solution to the problem

Solution to question (a): To use the CEP, we need derive the formulas
for computing C in terms of the forces for the structure shown in Fig. 13.12.
Using Eq. (13.101), the complementary energy Ci of the ith bar should be
it volume Vi times C0 . We thus have (see, Eq. (13.102)):
Ni
Ci = ei dNi (13.124)
0

Since the bars remain linearly elastic, the elongations of these two bars
can be calculated using Eq. (13.84). Thus, Eq. (13.124) becomes
Ni
Ni li li 1
Ci = dNi = Ni2 = N2 (13.125)
0 Ei Ai 2Ei Ai 2Ki i
where Ki is the stiffness of the ith bar.
The complementary energy for the structure becomes
1 1
C = C1 + C2 = N12 + N2 (13.126)
2K1 2K2 2
Next, we need to find formulas for Ni in terms of Pi , so that we can
use Eq. (13.108). This requires the use of equilibrium conditions. Since the
displacement is small, the two equations in Eq. (13.90) become

P1 − N1 Lb1 − N2 Lb2 = 0
(13.127)
P2 + N2 Lh2 = 0

Let us check how well the equilibrium is satisfied.

 Energy Methods 545

1 Nx = -N1*b/l1 - N2*b/l2 # sum of the internal forces in bars

2 Ny = N2*h/l2 # sum of the internal forces in bars
3 Fx = P1_l + Nx # this should be close to zero
4 Fy = P2_l + Ny # this should be close to zero
5
6 print(f'The internal forces in two bars: {N1:0.4e}, {N2:0.4e}(N)')
7 print(f'Two forces at point B (linearized): {P1_l:0.4e},{P2_l:0.4e}(N)')
8 print(f'Sum of all forces at point B (linear): {Fx:0.4e},{Fy:0.4e}(m)')
9 #1.3001e-01,-2.9995e-02(N) u1 = 0.05e-3; u2 = 0.03e-3

The internal forces in two bars: 1.0000e-01, 4.2435e-02(N)

Two forces at point B (linearized): 1.3000e-01,-3.0000e-02(N)
Sum of all forces at point B (linear): -8.2496e-06,5.9998e-06(m)

It is seen that the equilibrium is reasonably well satisfied.

Equations (13.127) can be solved for N1 and N2 , which gives

P1 L1 P2 L1
N1 = +
b h
(13.128)
P2 L2
N2 = −
h

Now use Eq. (13.108), we obtain the two displacement components at

point B:

∂C ∂C ∂N1 ∂C ∂N2 N1 L1 N2
u1 = = + = + ×0
∂P1 ∂N1 ∂P1 ∂N2 ∂P1 k1 b k2
(13.129)
L2 P1 P2
= 1 +
bk1 b h

∂C ∂C ∂N1 ∂C ∂N2 N1 L1 N2 L2
u2 = = + = −
∂P2 ∂N1 ∂P2 ∂N2 ∂P2 k1 h k2 h
(13.130)
L2 P1 P2 L2
= 1 + + 2 2 P2
k1 h b h k2 h

Solution to question (b): Given the same values for b, h, K1 , K2 , P1 , and

P2 , compute the two displacement components u1 and u2 using the formulas
obtained in part (a).
This can be done by running the following code, with the results from
the previous example. Note that the displacements must be small.
546 Mechanics of Materials: Formulations and Solutions with Python

1 # print out the Displacements given in the previous example:

2 print(f'Displacements at point B (linear): {u1:0.4e},{u2:0.4e}(m)')
3
4 u1_ = l1**2/(b*K1)*(P1_l/b + P2_l/h)
5 u2_ = l1**2/(h*K1)*(P1_l/b + P2_l/h) + l2**2/(h**2*K2)*P2_l
6 print(f'Displacements at point B (by CEP): {u1_:0.4e},{u2_:0.4e}(m)')

Displacements at point B (linear): 5.0000e-05,3.0000e-05(m)

Displacements at point B (by CEP): 5.0000e-05,3.0000e-05(m)

As shown, the results obtained using the CEP are the same as those by
PEP that are compatible. This confirms that the CEP results satisfy auto-
matically the compatibility condition. This means that the complimentary
energy principle is in place of the compatibility equations.

13.8 Displacements of statically determinate structures*

13.8.1 Principles for linear elastic structure members

Consider structures undergoes linear elastic deformation:

1. the material is linear elastic; and

2. the deformation is small.

For such a structure, the stress-strain curve for its material is shown in
Fig. 13.11(b), which is a straight line. Hence, the strain energy density U0
is equal to the complementary energy density C0 . When the structure under-
goes small deformation, the displacement-load relation for such a structure
is also linear. The strain energy U is equal to the complementary energy C.
Therefore, U and C are exchangeable. This can be made of a good use.
Since the energy evaluation is additive, the energy in a structure is the
sum of the energies in each of the members in the structure. In addition, the
principle of superposition applies for linear problems. Therefore, formulas
can be derived for computing the displacements for various structures with
multiple different members subject to multiple loads.
This section presents some of the widely used formulas.
Since C = U , Eqs. (13.108) and (13.109) can be rewritten as

∂U
Di = for i = 1, 2, . . . , n (13.131)
∂Fi
 Energy Methods 547

∂U
θi = for i = 1, 2, . . . , p (13.132)
∂Mi

These two equations allow the use of U for computing the displacements,
which are of interest. In many cases, formulating U in terms of forces is
more convenient than formulating C, and hence the foregoing equation is
more often used.

13.8.2 Principles for structures with multiple

multiple-type members
Many structures can be solved for internal forces using only the static equi-
librium equations, and are said to be statically determinate structures. Con-
sider a structure with m members of multiple types that may be coupled
with each other. Assume it behaves linear elastic, so that C = U .
The strain energy U should be the sum of these of all members. For
the jth member, its strain energy is denoted as Uj . In general, it can be
expressed as

Uj = UN j + UM j + USj + UT j (13.133)

Using Eqs. (13.26) and (13.108), C = U , and the additive property of the
potential energy, we shall have
m
!
∂U ∂U
Di = ∂Fi = j
∂Fi
j=1
m 
! 
∂UNj ∂Nj ∂UT j ∂Tj ∂UM j ∂Mj ∂USj ∂Vj
= ∂Nj ∂Fi + ∂Tj ∂Fi + ∂Mj ∂Fi + ∂Vj ∂Fi
j=1
(13.134)
m
! Lj Lj
Nj ∂Nj Tj ∂Tj
= Ej Aj ∂Fi dx + Gj Jj ∂Fi dx
j=1 0 0

Lj Lj
Mj ∂Mj kVj ∂Vj
+ Ej Ij ∂Fi dx + Gj Aj ∂Fi dx
0 0

This computes the displacement at the same degree, location and direc-
tion where Fi is applied.
548 Mechanics of Materials: Formulations and Solutions with Python

Similarly, to compute the rotation at the location where Mi is applied,

we use:
m
!
∂U ∂Uj
θi = ∂Mi = ∂Mi
j=1

m 
! 
∂U Nj ∂Nj ∂UT j ∂Tj ∂UM j ∂Mj ∂USj ∂Vj
= ∂Nj ∂Mi + ∂Tj ∂Mi + ∂Mj ∂Mi + ∂Vj ∂Mi
j=1
(13.135)
m
! Lj Lj
Nj ∂Nj Tj ∂Tj
= Ej Aj ∂Mi dx + Gj Jj ∂Mi dx
j=1 0 0

Lj Lj
Mj ∂Mj kVj ∂Vj
+ Ej Ij ∂Mi dx + Gj Aj ∂Mi dx
0 0

A similar derivation can also be done to compute the shear defection dy

at the location when Fi is applied, and the twist angle Θ at the location and
direction where Ti is applied. We omit these formulas because they are very
similar.
If the structure member does not have a type of internal force, the corre-
sponding term in Eqs. (13.134) and (13.135) should be ignored. For example,
for a pure truss structure, any member can only have internal normal force
Nj . We only need to use the first integral term in these equations.

13.8.3 Examples: Cantilever beam subjected to two

concentrated forces
Consider a cantilever beam with uniform cross-section and subjected to two
point vertical forces, F1 and F2 , as shown in Fig. 13.14. The length of the
beam is given by L and is assumed as a thin beam. The 2nd moment of area
of the cross-section is denoted as I, Young’s modulus of the material is E,
shear modulus is G, and all are given.
Determine the deflection of the beam at point C, consider both effects
from the moment and shear forces in the beam.

Solution: We shall use the 3rd and 4th integral in Eq. (13.134). Due to the
point force F1 at the middle point, both the moment and shear force can
only be expressed in two pieces of functions.
 Energy Methods 549

Figure 13.14. Cantilever beam subjected to two concentrated forces: one at its free tip,
and one at the middle span.

In the A-B segment, we have

∂V
V = F1 + F2 , =1
∂F2
(13.136)
L ∂M
M = F1 − x + F2 (L − x), =L−x
2 ∂F2

In the B-C segment, we get

∂V
V = F2 , =1 (13.137)
∂F2
∂M
M = F2 (L − x), =L−x
∂F2

Considering both effects from the moment and shear force, the deflection
at point C is computed using

∂U ∂U ∂M ∂U ∂V
D2 = = +
∂F2 ∂M ∂F2 ∂V ∂F2
L/2 L
F1 ( L −x)+F2 (L−x) F2 (L−x)
= 2
EI (L − x)dx + EI (L − x)dx (13.138)
0 L/2

L/2 L
k(F1 +F2 ) kF2
+ GA dx + GA dx
0 L/2

Note the integrals are over two segments. These integrations can be done
easily by hand, but it is even easier using the following Sympy codes.
550 Mechanics of Materials: Formulations and Solutions with Python

1 EI, L, x, F1, F2 = symbols('EI, L, x, F1, F2') # define variables

2 f1 = Function('f1')(x) # define functions
3 f2 = Function('f2')(x)
4 f1 = (F1*(L/2-x)+ F2*(L-x))*(L-x)/EI
5
6 int_f1 = [Link](f1,(x, 0,L/2)) # integration over 1st segment
7
8 f2 = F2*(L-x)*(L-x)/EI
9 int_f2 = [Link](f2,(x, L/2, L)) # integration over 2nd segment
10
11 D_M =(int_f1.simplify()+int_f2.subs(F1,F2)).simplify() # print out
12 [Link](D_M, "Deflection resulted from internal moments:")

Deflection resulted from internal moments:

L3 · (5F1 + 16F2 )
48EI

1 D_M = D_M.subs(F2, F1)

2 [Link](D_M, "Deflection resulted from internal moments, if F1=F2:")

Deflection resulted from internal moments, if F1=F2:

7F1 L3
16EI

Codes for computing the deflection resulting from internal shear force are
computed using:

1 GA, k = symbols('GA, k') # define variables

2 sf1 = Function('f1')(x) # define functions
3 sf2 = Function('f2')(x)
4
5 sf1 = k*(F1 + F2)/GA
6 int_sf1 = [Link](sf1,(x, 0,L/2)) # integration over 1st segment
7
8 sf2 = k*F2/GA
9 int_sf2 = [Link](sf2,(x, L/2,L)) # integration over 2nd segment
10
11 D_S =(int_sf1.simplify()+int_sf2.subs(F1,F2)).simplify() # print out
12 [Link](D_S, "Deflection resulted from internal shear force:")

Deflection resulted from internal shear force:

Lk (F1 + 2F2 )
2GA
 Energy Methods 551

1 D_S = D_S.subs(F2, F1)

2 [Link](D_S, "Deflection resulted from internal shear, if F1=F2:")

3F1 Lk
Deflection resulted from internal shear, if F1=F2:
2GA
The final result adds up these two deflections from moment and shear
force:
L3 (5F1 + 16F2 ) kL(F1 + 2F2 )
D2 = + (13.139)
48EI 2GA
If F1 = F2 , we have
7L3 F1 3kLF1
D2 = + (13.140)
16EI 2GA
The first part is from the internal moment and the second from the inter-
nal shear force on the cross-section resulting from the externally applied two
forces.

13.8.4 Examples: Cantilever beam subjected to

distributed force
Consider a cantilever beam with uniform cross-section and subjected to uni-
formly distributed force, p, as shown in Fig. 13.15. The length of the beam
is given by L and is assumed as a thin beam. The 2nd moment of area of
the cross-section is denoted as I, Young’s modulus of the material is E, and
all are given. Consider only the moment in the beam.

1. Determine the deflection at the beam’s free end.

2. Determine the rotation of the cross-section at the beam’s free end.

Figure 13.15. Cantilever beam subjected to uniformly distributed force. A fictitious

force or moment is applied at the beam’s free end when calculating the displacement
there.
552 Mechanics of Materials: Formulations and Solutions with Python

Solution to 1: Since only moment is considered, we shall use the 3rd integral
in Eq. (13.134). However, at the free end, there is no concentrated force. We
can apply a fictitious force F there to obtain the formula for the deflection.
After that, we can set F = 0 to obtain the final solution.
The formula for computing the rotation at tip DF resulting from the force
F becomes
L
∂U Mx ∂Mx
DF = = dx (13.141)
∂F 0 EI ∂F
The moment Mx is the internal moment within the beam, and is a func-
tion of x, which has contributions from both the originally applied dis-
tributed force and the fictitious one. The subscript x is not for rotation
direction. The moment Mx can be computed by equilibrium condition using,
1
Mx = F (L − x) + p(L − x)2 (13.142)
2
Its derivative with respect to F becomes
∂Mx
= (L − x) (13.143)
∂F
Substituting Eqs. (13.142) and (13.143) into (13.141), we obtain the solu-
tion:

L F (L − x) + 1 p(L − x)2 

DF = 2
 (L − x)dx
0 EI 
F =0
L (13.144)
L 1 p(L − x)3
p −1  pL 4

= 2
dx = (L − x)4  =
0 EI 2EI 4  8EI
0

Readers may use gr.solver1D4() developed in Chapter 8 to compute the

same.
Solution to 2: Since the rotation from the internal moment is requested, we
shall use the 3rd integral in Eq. (13.135). However, at the free end, there is
no concentrated moment. We need to apply a fictitious moment M there to
obtain the formula for the rotation. After that, we can set M = 0 to obtain
the final solution.
The formula for computing the rotation at tip θM resulting from the
internal moment Mx becomes
L
∂U Mx ∂Mx
θM = = dx (13.145)
∂Mx 0 EI ∂M
 Energy Methods 553

The moment Mx within the beam is a function of x, which has contri-

butions from both the originally applied distributed force and the fictitious
one. It can be computed by equilibrium condition using

1
Mx = M + p(L − x)2 (13.146)
2
Its derivative with respect to M is simply 1.
Substituting Eq. (13.146) into (13.145), we obtain the solution:

L M + 1 p(L − x)2 

θM = 2
 × 1dx
0 EI 
M =0
L (13.147)
L 1 p(L − x)2
p −1  pL 3

= 2
dx = (L − x)3  =
0 EI 2EI 3  6EI
0

Readers may also use gr.solver1D4() developed in Chapter 8 to compute

the same. Such a comparison can enhance the understanding, and it is easy
to do.

13.8.5 Fictitious unit load method

In the previous example we used fictitious load to obtain displacement at a
location where there is no load. We then need to set back the fictitious load
to zero. This means in computing U , we do not really need to include the
fictitious load, but we need it for differentials for the internal forces with
respect to the corresponding external forces. The improved approach is the
so-called fictitious unit load method [1]. The detailed procedure as follows.

1. Equilibrium: Write an expression for each of the internal actions (axial

force, shear, moment, and torque) in each member of the structure in
terms of only the actually applied external forces (and moments, etc.).
2. Apply a unit fictitious force Fi = 1 (or moment Mi = 1. etc.) at the
location and in the same direction as the displacement Di (or rotation θi )
to be determined.
3. Evaluate coefficients nFji , mFji , vji
F , tF , which are defined as
ji

∂Nj ∂Tj ∂Mj ∂Vj

nFji = , tFji = , mFji = , F
vji =
∂Fi ∂Fi ∂Fi ∂Fi
(13.148)
∂Nj ∂Tj ∂Mj ∂Vj
nM
ji = , tM
ji = , mM
ji = , M
vji =
∂Mi ∂Mi ∂Mi ∂Mi
554 Mechanics of Materials: Formulations and Solutions with Python

A coefficient given above is essentially the internal force in mem-

ber j resulting from a unit fictitious force at location i, obtained via
equilibrium.
4. Substitute these expressions into Eqs. (13.134) and (13.135), which gives
m
! Lj Lj Lj Lj
Nj n F Tj tF Mj mF F
kVj vji
Di = Ej Aj dx
ji
+ Gj Jj dx
ji
+ Ej Ij dx
ji
+ Gj Aj dx
j=1 0 0 0 0
(13.149)
and
m
! Lj Lj Lj Lj
Nj n M Tj tM Mj mM M
kVj vji
θi = Ej Aj dx
ji
+ ji
Gj J j dx + Ej Ij dx
ji
+ Gj A j dx
j=1 0 0 0 0
(13.150)
5. Perform the integrations, which gives Di (or θi ).

13.8.6 Example: Cantilever beam subjected to

distributed force
We repeat the previous example, but using the fictitious unit load method.

Solution to 1: The internal moment Mx is computed by equilibrium con-

dition using only the actual force.

1
Mx = p(L − x)2 (13.151)
2

We set F = 1, and calculate the coefficient mFji that is the moment at

j = x resulting from the unit force, and is obtained via equilibrium:

mFxi = 1 × (L − x) (13.152)

Substituting Eqs. (13.151) and (13.152) into (13.149), and using only the
integral for moment, we obtain the solution:
Lj L 1 p(L
− x)2
Mx mF
DF = EI dx
xi
= 2
(L − x)dx
0 0 EI
L (13.153)
L 1
p(L − x)3
p −1  pL4

= 2
dx = (L − x)4  =
0 EI 2EI 4  8EI
0

Solution to 2: The internal moment Mx is computed by equilibrium con-

dition using only the actual force. Thus, it is the same given in Eq. (13.151).
 Energy Methods 555

Set now M = 1, and calculate the coefficient mM ji that is the moment at

j = x resulting from the unit moment, and is obtained via equilibrium:
mM
xi = 1 (13.154)
Substituting Eq. (13.151) and (13.151) into (13.150), we obtain the
solution:
Lj L 1 p(L
− x)2
Mx mM
θM = EI
xi
dx = 2
× 1dx
0 0 EI
L (13.155)
L 
2 p(L − x)
1 2
p −1 3 pL3
= dx = (L − x)  =
0 EI 2EI 3  6EI
0
As shown, we obtained the same solution. The fictitious unit load method
is more convenient to use because 1. it eliminates the differentiations, and
2. it simplifies the evaluation of the equilibrium equations. Essentially, one
needs only to get the internal force via equilibrium considering only the
actual external forces, and the internal force via equilibrium considering
only the fictitious unit force.

13.8.7 Example: A planar truss structure with

three bars
Consider a simple planar truss structure with three bars (numbered in small
box), as shown in Fig. 13.16. These bars all have uniform cross-section with
area A, and all made of the same material with E. The structure has three
nodes (numbered in small circles). The structure is subjected to a verti-
cal downward force of F2y at node 2. Assume the structure behaves linear
elastically. Using the energy method, determine:

1. the vertical displacement at node 2, D2y ;

2. the horizontal displacement at node 2, D2x ;
3. the values of D2x and D2y , letting F2y = 1000 N, E = 70.0 GPa, and
A = 0.01 m2 .

Solution to 1: We shall use the 1st integral in Eq. (13.134). Therefore,

we need to find out the internal axial forces for each of the bars. Since this
structure is statically determinate, these internal forces can be found using
equilibrium conditions.
First, at node 2, we can easily find:
F √
N2 = − sin 2y
45◦ = −F2y 2
(13.156)
N1 = −N2 cos 45◦ = F2y
556 Mechanics of Materials: Formulations and Solutions with Python

Figure 13.16. A planar truss structure with three bars hinged together. It is subjected
to a vertical force.

Next, at node 3, we found

N3 = −N2 cos 45◦ = F2y (13.157)

The derivatives of these internal forces with respect to F2y become

∂N2 √ ∂N1 ∂N3

= − 2; = 1; =1 (13.158)
∂F2y ∂F2y ∂F2y

Substitute Eqs. (13.156), (13.157), and (13.158) to the 1st integral in

Eq. (13.134), we obtain

L1
N1 L2
N2 √ L3
N3
D2y = × 1dx + (− 2)dx + × 1dx
0 EA 0 EA 0 EA
√
F2y L1 −F2y L2 2 √ F2y L3 (13.159)
= + (− 2) +
EA EA EA
F2y 2F2y √
= (L1 + 2L2 + L3 ) = (1 + 2)
EA EA
 Energy Methods 557

Solution to 2: Since there is no force in the x-direction at node 2, we shall

use the fictitious unit load method. We apply F2x = 1. The internal forces
have already been found and given in Eqs. (13.156) and (13.157). We need
only to compute the coefficients, each of which is the internal force in a bar
resulting from F2x = 1. This can again be obtained via equilibrium.
First, at node 2, we shall find:

nF2x2x = 0; nF1x2x = 1; nF3x2x = 0 (13.160)

Substitute Eqs. (13.160) and (13.157) to the 1st integral in Eq. (13.149),
we obtain
L1 L2 L3
N1 × 1 N2 × 0 N3 × 0
D2x = dx + dx + dx
0 EA 0 EA 0 EA
F2y L1
= (13.161)
EA
Solution to 3: Substitute the given values for all these variables, we com-
pute the numerical values for F2x and F2y using the following codes.

1 E = 70.0e9; A = 0.01; L1 = 1.0

2 F2y = 1000.
3
4 D2y = 2*F2y*L1/E/A*(1.+[Link](2))
5 D2x = F2y*L1/E/A
6 print(f"Displacements at node 2: ({D2x}, {D2y})")

Displacements at node 2: (1.4285714285714286e-06, 6.8977530353517e-06)

13.9 Statically indeterminate structures*

13.9.1 Issues in dealing with indeterminate structures

In using the CEP, we need to find internal forces in a structure using equi-
librium equations. This means that the structure needs to be statically
determinate. However, many structures that are statically indeterminate.
In fact, most of the structures used in practical engineering are indetermi-
nate because of the important requirement on the structural stiffness. For an
indeterminate structure, we need to find additional equations or constrain
conditions to find the solution, when using CEP.
In this section, we use an example to demonstrate how an indeterminate
structure can be solved.
558 Mechanics of Materials: Formulations and Solutions with Python

13.9.2 Example: Cantilever beam with a simple support

at the other end
Consider a cantilever thin beam with a support at its right-end, as shown
in Fig. 13.17(a). It is thus an indeterminate structure, implying that the
internal forces cannot be found via only the use of equilibrium conditions.
The beam is subject to an uniformly distributed force p. Assume Young’s
modulus of the material is E, and the 2nd moment of area is I. Use energy
method.
1. Determine the reaction R at the support at the right-end.
2. Determine the rotation of the cross-section of the beam at the right-end.
Solution to 1: Due to the redundant support to the beam, we cannot find
the internal forces directly. Thus, we remove the support at the right-end,
and replace it with an unknown force R, called reaction force. The new
setting is given in Fig. 13.17(b). We now have a cantilever beam subjected
to two forces: one distributed p, and one concentrated force R. This problem
is solved in previous examples and the solutions are given in Eqs. (13.47)
and (13.144). The deflection at the reaction force location DR can be written
as
RL3 pL4
DR = + (13.162)
3EI 8EI

(a)

(b)

Figure 13.17. A cantilever beam with an additional support at its right-end. It is subject
to a distributed force. (a) The beam with redundant support; (b) the support at the right-
end is replaced by a reaction force.
 Energy Methods 559

Since the right-end of the beam should be zero, DR = 0, we thus obtain

a constraint equation:
RL3 pL4
+ =0 (13.163)
3EI 8EI
This gives the reaction force:
3pL
R=− (13.164)
8
Solution to 2: Since the reaction force has now been found, computing the
rotation of the cross-section of the beam at the right-end is straightforward.
The solution is, in fact, found in our earlier examples, given in Eq. (13.49)
for a concentrated load at the right-end, and in Eq. (13.147) for a distributed
loading. The solution is the sum of those:
RL2 pL3 −3pL3 pL3 pL3
θR = + = + =− (13.165)
2EI 6EI 16EI 6EI 48EI
Readers may compare the results with that obtained in Section 8.13.

13.10 Remarks
This chapter studies two major energy principles:
1. PEP, which is derived from the principle of virtual work using virtual
displacement. It expresses the external force in terms of the partial dif-
ferential of the potential (strain) energy with respect to displacement.
Compatibility conditions for the displacement and elongation (or strains)
are required to use PEP. The result is the equilibrium equation.
2. CEP, which is derived from the principle of virtual work using virtual
force. It expresses the displacement in terms of the partial differential of
the complementary energy with respect to the external force. Equilibrium
conditions for the external and internal forces are required to use CEP.
The result is the compatibility equation.
These energy methods are demonstrated using a number of examples with
Python codes. The energy principles, especially the PEP, form a theoretical
foundation for extremely powerful numerical methods, such as FEM [6],
S-FEM [7] and meshfree methods [8] for large-scale solid mechanics problems.

References
[1] A.P. Boresi, R.J. Schmidt and O.M. Sidebottom, Advanced Mechanics of Materials.
John Wiley & Sons, New York, 1985.
[2] A.P. Boresi, K. Chong and J.D. Lee, Elasticity in Engineering Mechanics. John
Wiley & Sons, New York, 2010.
560 Mechanics of Materials: Formulations and Solutions with Python

[3] J.M. Gere and S.P. Timoshenko, Mechanics of Materials, Van Nostrand Reinhold
Company, New York, 1972.
[4] S. Timoshenko and J.N. Goodier, Theory of Elasticity. McGraw-Hill, New
York, 1970. [Link]
+of+elasticity&#38;ei=ICiKSsr3G4jwkQSbxMyPCg.
[5] Z.L. Xu, Elasticity, Vols. 1&2. People’s Publisher, China, 1979.
[6] G.R. Liu and S.S. Quek, The Finite Element Method: A Practical Course.
Butterworth-Heinemann, New York, 2013.
[7] G.R. Liu and T.T. Nguyen, Smoothed Finite Element Methods. Taylor and Francis
Group, New York, 2010.
[8] G.R. Liu, Mesh Free Methods: Moving Beyond the Finite Element Method. Taylor
and Francis Group, New York, 2010.
[9] A.E.H. Love, A Treatise on the Mathematical Theory of Elasticity, 1944, http://
[Link]/isbn/0486601749.
[10] G.R. Liu, Vector Calculus: Formulation, Applications, and Python Codes, World
Scientific, New Jersey, 2025.
[11] G.R. Liu, Calculus: A Practical Course with Python, World Scientific, New Jersey,
2025.
 Chapter 14

Failure Criteria and Failure Analysis

Let us import the necessary external modules or dependencies for later use
in this chapter:
1 # Place cursor in this cell, and press Ctrl+Enter to import dependences.
2 import sys # for accessing the computer system
3 [Link]('../grbin/') # Change to the directory in your system
4
5 from commonImports import * # Import dependences from '../grbin/'
6 import grcodes as gr # Import the module of the author
7 [Link](gr) # When grcodes is modified, reload it
8
9 from continuum_mechanics import vector
10 init_printing(use_unicode=True) # For latex-like quality printing
11 np.set_printoptions(precision=4,suppress=True) # Digits in print-outs

The major purpose of stress analysis is to assess whether the stressed

material will fail. Such an assessment is called failure analysis. The most
important stage in material failure process is yielding, which can often be
clearly observed during experiments of many types of materials. When the
stressed material gets to the yielding point, it starts to deform drastically
without much increase in loading, and the load level is often very close to
the ultimate failure load.
In addition, in many conservative designs of structure, material yielding
is regarded as failure. Such yielding-based designs apply to all structures and
devices that are subjected to multiple repetitive uses and often with large
safety factors, implying that the service stress should be far below the yield
stress. Therefore, the yield criterion of a material is particularly important.
This chapter focuses on various types of yield criteria used for different types
of materials. The chapter is written with reference to Refs. [1–3].

561
562 Mechanics of Materials: Formulations and Solutions with Python

14.1 Basic concept

As discussed in previous chapters, a structural component can have multiple
(as much as six) stress components at a point in the material. When a
yielding criterion is defined, it is thus convenient to define a single equivalent
or effective stress σe that is a combination of all the stress components. Since
material yielding/failing is a physical property of the material and it should
not depend on the coordinates used to compute the stresses, the effective
stress should be a proper combination of the principal stresses or other stress
invariants.
Clearly, different forms of yield criteria are needed for different types
of materials. We define a yield function f (σij , σY ), where σij represent the
stress states and σY is the yield strength of the material in uniaxial tension
(or compression) obtained from an experiment. A stress-based yield criterion
can be expressed as [4]
f (σij , σY ) = 0 (14.1)
at which the material is said to be on the yield surface. When f (σij , σY ) < 0,
the stress state is elastic, and f (σij , σY ) > 0 is not allowed.

14.2 A case for this study: Yield analysis of a steel shaft

We use a typical case for studying various yielding criteria. The case chosen
has a combined stress state resulting from bending, pressure, and torsion at
the same time. The strategy is to introduce different yielding criteria one by
one, and use it to assess whether or not the material will yield for the same
given stress state at a point. We compute also the safety factor based on the
yielding criterion used.

14.2.1 Setting of the problem

Consider a solid cylindrical shaft with a diameter of d = 100 mm. It is loaded
by a torque T = 30 kNm, a bending moment M = 15 kNm, and a pressure
p = 50 MPa on its outer surface. The directions of these loads are shown
in Fig. 14.1. The shaft is made of steel treated as a linear elastic material
before yielding. Its Young’s modulus is E = 200 GPa, and Poisson’s ratio is
ν = 0.3. Determine whether or not the shaft will yield at point A, assuming
the safety factor as Sf = 2.5, and the material yield stress σY = 750 MPa.

14.2.2 Formulas for computing the stresses

From the setting given in Fig. 14.1, the contributions to the stresses on the
lateral surface of the shaft are from the torque and the uniform pressure.
 Failure Criteria and Failure Analysis 563

(a)

(b)

Figure 14.1. (a) A solid cylindrical shaft subjected to a torque T, moment M, and a
pressure p on the outer surface. (b) Top-down view on the upper surface of the bar. All
stress components shown are positive following the standard sign convention.

The normal stress in the x-direction caused by the moment will be largest
at points A or B. Therefore, the critical points on this shaft should be points
A or B. Let us examine the stress states at point A.
The normal stress in the x-direction at point A (z = d/2) caused by the
moment is computed using

zSF M SF M d
σ= =− (14.2)
I 2I

Here, we use the same safety factor SF to scale up the moment applied
to the bar, and I is the second moment of area of the cross-section of the
shaft, computed using

πd4
I= (14.3)
64

The shear stress caused by the torque is computed by

SF T d
τ= (14.4)
2J
564 Mechanics of Materials: Formulations and Solutions with Python

Similarly, we use the safety factor SF to scale up the torque applied to

the bar, and J is the second polar moment of the cross-section of the shaft,
computed using

πd4
J= (14.5)
32
The stress components at point A are then given by

σxx = σ; σxy = τ ; σzx = 0; σyy = 0; σzy = 0; σzz = −SF p

(14.6)

14.2.3 Python code to compute the stresses

We write the following code to compute the stress tensor in matrix form,
and then compute the principal stresses.

1 # External forces
2 Torque = 30.0e3 # Nm
3 Moment = 15.0e3 # Nm
4 p = 50.0e6 # Pa, pressure on the surface
5 SF = 2.5 # safety factor
6
7 # Material property
8 E = 200.0e9 # Pa, Young's modulus, not used in this example
9 = 0.3 # Poisson's ratio
10 sY= 750.0e6 # Pa, Yield stress
11
12 # Dimensions of the cylindrical shaft
13 d = 0.100 # m, diameter
14 #A = [Link]*((0.1/2)**2-(0.08/2)**2)
15
16 # Compute structure/member properties
17 I = ([Link] * d**4)/64.0 # 2nd moment of circular area
18 J = ([Link] * d**4)/32.0 # 2nd polar moment of circular area
19
20 # Compute the stresses
21 =-SF*Moment*(d/2.)/I # stress by the moment at point A
22 = SF*Torque*(d/2.)/J # stress by the Torque
23
24 xx= ; xy= ; xz=0; yy=0; yz=0; zz=-SF*p
25
26 Sij = [Link]([[ xx, xy, xz],
27 [ xy, yy, yz],
28 [ xz, yz, zz]])
29 print(f'Stress state:\n{Sij}\n')
30
31 eigens, eigenVectors = [Link](Sij) # principal stresses
 Failure Criteria and Failure Analysis 565

Stress state:
[[-3.8197e+08 3.8197e+08 0.0000e+00]
[ 3.8197e+08 0.0000e+00 0.0000e+00]
[ 0.0000e+00 0.0000e+00 -1.2500e+08]]

Principal stress/strain (Eigenvalues):

[ 2.3607e+08 -1.2500e+08 -6.1804e+08]

Principal stress/strain directions:

[[-0.5257 0. -0.8507]
[-0.8507 0. 0.5257]
[ 0. 1. 0. ]]

Possible angles (n1,x)=121.71747441146103 or 58.282525588538974

We got two angles because of the non-uniqueness of the eigenvectors and

the multivalued feature of cosine functions. We use the following code to find
out which one is correct.
1 # Form the transformation matrix using the first angle:
2 T = [Link](58.282525588538974, about = 'z')[0]
3
4 # Perform the transformation to the stress tensor
5 T @Sij@T .T

array([[ 2.3607e+08, 4.1106e-07, 0.0000e+00],

[ 4.1692e-07, -6.1804e+08, 0.0000e+00],
[ 0.0000e+00, 0.0000e+00, -1.2500e+08]])

This angle of 58.28 degree is not correct. The angle of 121.72 or −58.28
degree is correct.
We now discuss various yield criteria to assess whether or not the material
will yield at point A and the actual safety factor.

14.3 Maximum principal stress criterion

The maximum principal stress criterion states that yielding begins at a point
where the maximum absolute value of the principal stress reaches the abso-
lute yield stress σY that may be obtained from a tensile or compressive test.
For given a stress state with σij at a point in solid, we compute the
principal stresses and rank them in the order of |σ1 | ≥ |σ2 | ≥ |σ3 |. The
effective stress σe is chosen as
σe = |σ1 | (14.7)
and the yield criterion is
σe − σY = 0 (14.8)
566 Mechanics of Materials: Formulations and Solutions with Python

This equation is convenient to use for assessing whether or not the mate-
rial is yielding at a point, and hence useful in design of structures.

14.3.1 Yield surface

We can also compute a surface, called the yield surface, in the (σ1 , σ2 , σ3 )
coordinate system. It is an envelope. If the principal stresses of any given
stress state fall on the envelope, the material will yield. If it is within the
envelope, it is still elastic and safe. Such a yield surface is useful to determine
the bounds on which the material will yield. The function for the yield surface
is called the yield function. For the maximum principal stress criterion, the
yield function can be written as

f1 or 2 = |σ1 − σY | = 0 or f1 or 2 = σ1 − ±σY = 0
f3 or 4 = |σ2 − σY | = 0 or f3 or 4 = σ2 − ±σY = 0 (14.9)
f5 or 6 = |σ3 − σY | = 0 or f5 or 6 = σ3 − ±σY = 0

It has a total of six surfaces, and the yield function is the enclosed piece-
wise function formed by these six surfaces, denoted as

f = f1 ∪ f2 ∪ f3 ∪ f4 ∪ f5 ∪ f6 (14.10)

It is the six surfaces of a cube in the (σ1 , σ2 , σ3 ) coordinate system. For

plane stress problems, it is a square in the (σ1 , σ2 ) coordinate system.
Figure 14.2 shows a yield surface with four straight lines in the (σ1 , σ2 )
coordinates for plane stress problems.
This yield criterion may not be good for ductile materials but can be
good for brittle materials.

Figure 14.2. Schematic view of the yield surface of the maximum principal stress for
plane stress problems. It consists of four straight lines.
 Failure Criteria and Failure Analysis 567

14.3.2 Application to the case study problem

The following codes compute the maximum principal stress using the prin-
cipal stresses obtained in Section 14.2.3. This gives the material yield stress
value needed to have point A stay below yielding, which determines the yield
state.

1 # maximum principal stress criterion

2 ps_max = max(abs(eigens))
3 print(f'The maximum principal stress = {ps_max:0.5e}')
4
5 print(f'The material yield stress needs to be > {ps_max:0.5e}')
6 print(f'The yield stress of the material = {sY:0.5e}')
7 print(f'Is yield? {ps_max>sY}')
8 print(f'Actual safety factor = {SF*sY/ps_max}')

The maximum principal stress = 6.18043e+08

The material yield stress needs to be > 6.18043e+08
The yield stress of the material = 7.50000e+08
Is yield? False
Actual safety factor = 3.0337672480085427

14.4 Maximum principal strain criterion

It states that yielding begins when the maximum absolute value of the prin-
cipal strains at a point reaches the absolute value of the yield strain εY that
may be obtained from a tensile or compressive test.
For given a strain state with εij , we compute the principal strains and
rank them in the order of |ε1 | ≥ |ε2 | ≥ |ε3 |. The effective stress εe is chosen
as

εe = |ε1 | (14.11)

and the yield criterion is

εe − εY = 0 (14.12)

This yield function is strain-based. It may not be good for ductile mate-
rials but can fit better for brittle materials.
Using the constitutive equations for materials (see Chapter 5), one can
express the principal strains in terms of principal stresses, and then the
maximum principal strain criterion can be written in terms of stresses.
568 Mechanics of Materials: Formulations and Solutions with Python

For example, using Eq. (5.73) for isotropic materials, we have

1
ε1 = [σ1 − νσ2 − νσ3 ]
E
1
ε2 = [σ2 − νσ1 − νσ3 ] (14.13)
E
1
ε3 = [σ3 − νσ1 − νσ2 ]
E
From the uniaxial load at the yield point (σ1 = σY , σ2 = σ3 = 0), we
have
σY
εY = (14.14)
E

14.4.1 Yield Surface

Using Eqs. (14.12), (14.13), and (14.14), we shall have a yield surface with
six flat planes in the (σ1 , σ2 , σ3 ) coordinates:

f1 or 2 = σ1 − νσ2 − νσ3 ± σY = 0
f3 or 4 = σ2 − νσ1 − νσ3 ± σY = 0 (14.15)
f5 or 6 = σ3 − νσ1 − νσ2 ± σY = 0

The yield surface is the union of the six surfaces of a parallelepiped.

The effective yield stress can be written as a single expression as

σe = max (σi − νσj − νσk ) (14.16)

i=j=k

The yield criterion can also be written as

σe − σY (14.17)

Figure 14.3 shows a yield surface with four straight lines in the (σ1 , σ2 )
coordinates for plane stress problems.

14.4.2 Application to the case study problem

The following codes compute the maximum principal strain using the prin-
cipal stresses obtained in Section 14.2.3. This gives the material yield stress
 Failure Criteria and Failure Analysis 569

Figure 14.3. Schematic view of the yield surface of the maximum principal strain for
plane stress problems. It is a parallelogram of four straight lines.

value needed to have point A stay below yielding and determines the
yield state.

1 # maximum principal strain criterion

2 se1 = eigens[0]- *eigens[1]- *eigens[2] # principal strains E ε1
3 se2 = eigens[1]- *eigens[0]- *eigens[2] # principal strains E ε2
4 se3 = eigens[2]- *eigens[0]- *eigens[1] # principal strains E ε3
5 e_max =max(abs([Link]([se1, se2, se3])))
6
7 print(f'The maximum principal strain is {e_max:0.5e}')
8 print(f'The material yield stress needs to be > {e_max:0.5e}')
9 print(f'The yield stress of the material = {sY:0.5e}')
10 print(f'Is yield? {e_max>sY}')
11 print(f'Actual safety factor = {SF*sY/e_max}')

The maximum principal strain is 6.51365e+08

The material yield stress needs to be > 6.51365e+08
The yield stress of the material = 7.50000e+08
Is yield? False
Actual safety factor = 2.878570669361039

14.5 Strain-energy density criterion

The strain-energy density is one of the coordinate-independent quantities

and hence is often used for establishing failure criteria. This criterion states
that yielding occurs at a point where the strain-energy density equals the
strain energy density at yield in uniaxial tension (or compression). It is also
known as the total strain energy criterion.
570 Mechanics of Materials: Formulations and Solutions with Python

14.5.1 Strain-energy density

The strain-energy density can be defined in various forms, which are given
in Eq. (13.75). The strain energy in principal stresses is given by
1  2 
U0 = σ1 + σ22 + σ32 − 2ν(σ2 σ3 + σ1 σ3 + σ1 σ2 ) (14.18)
2E
For a uniaxial load at the yield point (σ1 = σY , σ2 = σ3 = 0), U0 at
failure is
σY2
U0Y = (14.19)
2E
The failure criterion becomes

σ12 + σ22 + σ32 − 2ν(σ2 σ3 + σ1 σ3 + σ1 σ2 ) − σY2 = 0 (14.20)

The effective stress σe becomes


σe = σ12 + σ22 + σ32 − 2ν(σ2 σ3 + σ1 σ3 + σ1 σ2 ) (14.21)

and the yield function is

f = σe2 − σY2 = 0 (14.22)

This yield criterion is an ellipsoid in the principal stress space. It is a

circle when ν = 0. The yield function is one piece.

14.5.2 Application to the case study problem

The following codes compute the strain-energy density using the principal
stresses obtained in Section 14.2.3. This gives the material yield stress value
needed to have point A stay below yielding, which determines the yield state.

1 # Strain-energy density criterion

2 U_stress = gr.energy_stress(eigens, )
3 print(f'The strain-energy stress is {U_stress:0.5e}')
4 print(f'The material yield stress needs to be > {U_stress:0.5e}')
5 print(f'The yield stress of the material = {sY:0.5e}')
6 print(f'Is yield? {U_stress>sY}')
7 print(f'Actual safety factor = {SF*sY/U_stress}')

The strain-energy stress is 7.15700e+08

The material yield stress needs to be > 7.15700e+08
The yield stress of the material = 7.50000e+08
Is yield? False
Actual safety factor = 2.619813700936882
 Failure Criteria and Failure Analysis 571

14.6 Yielding criteria for ductile materials

Many materials used in engineering applications are purposely made ductile.

Yielding criteria for ductile materials are thus well studied. This section
introduces these type of criteria.

14.6.1 Maximum shear-stress (Tresca) criterion

Many crystals have slip planes with relatively weak against shearing. Yield
criteria based on shear stress make sense. The Tresca criterion states that
the maximum shear stress equals the yield shear stress under the uniaxial
load σY .
With ranked principal stresses obtained, the maximum shear stress τmax
at a point in a solid can be computed using

τmax = (σ1 − σ3 )/2 (14.23)

Note that because the principal stresses are invariant, the maximum shear
stress is also invariant. The effective stress σe is chosen as

σe = τmax (14.24)

For uniaxial stress states, we have

σ1 = σY , σ3 = 0 =⇒ τY = σ1 /2 = σY /2 (14.25)

Thus, the Tresca yield criterion becomes

σe − σY /2 = 0 (14.26)

14.6.2 Tresca yield surface

For an arbitrary state of stresses, we can compute all these principal stresses,
and then draw a Tresca yield surface. This is done by computing all three
possible maximum shear stresses in terms of these principal stresses. Con-
sidering each of the maximum shear stresses can be possibly either positive
or negative, the entire yield surface shall consist of six planes. The detailed
formulation is given as follows:
σ2 − σ3 σ1 − σ3 σ1 − σ2
τmax1 = ± ; τmax2 = ± ; τmax3 = ± (14.27)
2 2 2
The Tresca yield surface on which the material can yield becomes

σ2 − σ3 ± σY = 0; σ1 − σ3 ± σY = 0; σ1 − σ2 ± σY = 0 (14.28)
572 Mechanics of Materials: Formulations and Solutions with Python

Figure 14.4. Schematic view of the Tresca yield surface. Left: cylinder with uniform
hexagonal cross-section in 3D space of the principal stresses (σ1 , σ2 , σ3 ); right: elongated
hexagon in 2D space (σ1 , σ2 ). Transferred from [Link] by User: Jaqen, original
uploader was Janek Kozicki at [Link] under the CC BY-SA 3.0 license.

These six planes in 3D space of the principal stresses are shown on the left
image of Fig. 14.4. It can be better viewed in 2D space by setting σ3 = 0,
as shown on the right. In such a 2D case, the surface consists of six line
segments. The expressions for the corresponding six lines are given by
σ2 = ±σY ; σ1 = ±σY ; σ1 − σ2 = ±σY (14.29)
These are four vertical lines and two 45-degree inclined lines.
This maximum shear stress yield criterion is one of the most widely used
criteria for ductile materials.

14.6.3 Application to the case study problem

The following codes compute the maximum shear stress using the principal
stresses obtained in Section 14.2.3. This gives the material yield stress value
needed to have point A stay below yielding and determines the yield state.

1 # Maximum shear stress (Tresca) criterion

2 Tau_max =0.5*(eigens[0]-eigens[2])
3 print(f'The maximum shear stress is {Tau_max:0.5e}')
4 print(f'The material yield stress needs to be > {(2*Tau_max):0.5e}')
5 print(f'The yield stress of the material = {sY:0.5e}')
6 print(f'Is yield? {(2*Tau_max)>sY}')
7 print(f'Actual safety factor = {SF*sY/(2*Tau_max)}')

The maximum shear stress is 4.27058e+08

The material yield stress needs to be > 8.54115e+08
The yield stress of the material = 7.50000e+08
Is yield? True
Actual safety factor = 2.195254603450228
 Failure Criteria and Failure Analysis 573

14.6.4 Distortional energy density criteria

Yielding begins when the distortional strain-energy density is at that of the
uniaxial limit. The distortional strain-energy density is associated with the
shape distortion in a solid. The total strain-energy density U0 can be written
in two parts:

(σ1 + σ2 + σ3 )2 (σ2 − σ3 )2 + (σ1 − σ3 )2 + (σ1 − σ2 )2

U0 = + (14.30)
 18K
   12G
 
UV UD

where UV and UD are energies related to the volumetric and deviatoric

deformation. In the distortional strain-energy density criterion, we discard
UV and assume that only the deviatoric energy density UD controls the
yielding.
Using the von Mises stress defined in Chapter 3, we have
2
σvm
UD = (14.31)
6G

[Link] von Mises criterion

The effective stress σe is chosen as

1
σe = σvm = [(σ2 − σ3 )2 + (σ1 − σ3 )2 + (σ1 − σ2 )2 ] (14.32)
2

For 2D stress cases, say σ3 = 0, we have


σe = σvm = σ12 + σ22 − σ1 σ2 (14.33)

Consider uniaxial loading (σ1 = σY , σ2 = σ3 = 0), the corresponding

2
σY
deviatoric energy density becomes UDY = 6G . Equaling UD and UDY gives
the yield function:

f = σe2 − σY2 = 0 (14.34)

This is the well-known von Mises yield criterion.

It is clear that the von Mises yield surface is an ellipsoid in the principal
stress space, as shown in Fig. 14.5.
The von Mises yield criterion is the most widely used criterion for ductile
materials.
574 Mechanics of Materials: Formulations and Solutions with Python

Figure 14.5. Schematic view of the von Mises yield surface together with the Tresca
yield surface. Left: the von Mises yield surface is the cylindrical surface in 3D space of the
principal stresses (σ1 , σ2 , σ3 ); The Tresca yield surface is a cylinder with hexagonal cross-
section. Right: in 2D space (σ1 , σ2 ). Images modified based on the one from [Link]
by Rswarbrick under the CC BY-SA 3.0 license (left); and the one by Melchoir to public
domain (right).

[Link] Application to the case study problem

The following codes compute the von Mises stress using the principal stresses
obtained in Section 14.2.3. This gives the material yield stress value needed
to have point A stay below yielding and determines the yield state.

1 # von Mises criterion

2 svm = gr.von_Mises(eigens)
3 print(f'The von Mises stress is {svm:0.5e}')
4 print(f'The material yield stress needs to be > {svm:0.5e}')
5 print(f'The yield stress of the material = {sY:0.5e}')
6 print(f'Is yield? {svm>sY}')
7 print(f'Actual safety factor = {SF*sY/svm}')

The von Mises stress is 7.42623e+08

The material yield stress needs to be > 7.42623e+08
The yield stress of the material = 7.50000e+08
Is yield? False
Actual safety factor = 2.524835151818577

[Link] Octahedral shear-stress criterion

The von Mises stress relates to the octahedral shear-stress as discussed in
Chapter 3, which is
3
σvm = √ τoct (14.35)
2
 Failure Criteria and Failure Analysis 575

The yield criterion can also be written as

√
2
τoct − σY = 0 (14.36)
3
This is known as the octahedral shear-stress criterion. It has a
slightly different form, but it is essentially the same as the von Mises yield
criterion.
Since octahedral shear-stress relates to the second invariant of the devi-
atoric stress tensor, J2 , in the form of
3 2
J2 = − τoct (14.37)
2
the octahedral shear-stress function (or the von Mises function) can also be
written as
σY
f = |J2 | − √ (14.38)
3

[Link] Application to the case study problem

The following codes compute the octahedral shear stress using the principal
stresses obtained in Section 14.2.3. This gives the material yield stress value
needed to have point A stay below yielding and determines the yield state.

1 # Octahedral shear stress criterion

2 _, tau_oct=gr.Oct_stress(eigens)
3 print(f'The Octahedral shear stress is {tau_oct:0.5e}')
4 Y = tau_oct*3/[Link](2)
5 print(f'The material yield stress needs to be > {Y:0.5e}')
6
7 print(f'The yield stress of the material = {sY:0.5e}')
8 print(f'Is yield? {Y>sY}')
9 print(f'Actual safety factor = {SF*sY/Y}')

The Octahedral shear stress is 3.50076e+08

The material yield stress needs to be > 7.42623e+08
The yield stress of the material = 7.50000e+08
Is yield? False
Actual safety factor = 2.524835151818577

[Link] Difference between the von Mises and Tresca criteria

As shown in Fig. 14.5, the von Mises criterion is the upper bound of the
Tresca criterion. In the 2D case, the von Mises criterion is a smooth ellipse,
but the Tresca is piecewise straight lines. Their difference at points T and
M shown in the right image of Fig. 14.5 can be quantified as follows.
576 Mechanics of Materials: Formulations and Solutions with Python

Considering a state of pure shear (σ1 = −σ2 = σ, σ3 = 0), which is the

blue line shown in Fig. 14.5, we have
σ1 − σ2
τmax = =σ (14.39)
2
The failure τY predicted by the Tresca criterion is
σY
τY = τmax = = 0.5σY (14.40)
2
The von Mises criterion Eq. (14.32) gives
σY
σ = τY = √ = 0.5774σY (14.41)
3
In this case, the von Mises criterion prediction is approximately 15%
greater than that by the Tresca criterion.
The von Mises criterion is often found to be more accurate for some
materials than the Tresca criterion in predicting yield under pure shear. The
Tresca criterion is on the conservative side.

14.6.5 π-plane
The π-plane is a plane on which the hydrostatic stress (or the means stress),
σm = (σ1 + σ2 + σ3 )/3 = 0. It is the blue plane shown in the left image of
Fig. 14.5. It passes through the origin of the coordinate
√ √ of (σ1 , σ2 , σ3 ).
√ system
The direction cosines of the π-plane is (1/ 3, 1/ 3, 1/ 3), which is the
same direction as the octahedral plane. For 2D cases, the π-plane is a line
on which σm = (σ1 + σ2 )/2 = 0. It is the blue line shown in the right image
of Fig. 14.5. It passes through the origin of the coordinate system of (σ1 , σ2 ).
It is a special case of the so-called deviatoric plane.

14.6.6 Deviatoric planes

A deviatoric plane is a plane on which the hydrostatic stress or the means
stress, σm = (σ1 + σ2√+ σ3 )/3
√ = √constant. The direction cosines of the
deviatoric plane is (1/ 3, 1/ 3, 1/ 3), same as the octahedral plane. It is
a plane parallel to the π-plane, and thus there are infinite number of them.
One of them is shown as a light blue ellipse in the left image of Fig. 14.5.
For 2D cases, is a line on a deviatoric plane σm = (σ1 + σ2 )/2 = constant.
In other words, the projection on the hydrostatic axis of an arbitrary stress
state σa in a deviatoric plane is a constant.
Such a deviatoric plane is important because a distortional energy density
criterion (such as the nov Mises) applies to stress states on any of these
planes because the level of the hydrostatic stress does not affect the failure.
 Failure Criteria and Failure Analysis 577

Therefore, the yielding surface is a cylinder as shown in the left image of

Fig. 14.5. Due to this, we often draw yield surfaces on the π-plane, as often
seen in the literature.
Note that because the direction of a deviatoric plane is the same as the
octahedral plane, the von Mises criterion is the same as the octahedral shear
stress criterion.

14.7 Yield criterion for cohesive materials*

In many applications in engineering and nature, we frequently encounter
cohesive materials, including concrete and rock. It is found that:
1. Yielding of many such cohesive materials depends on hydrostatic stress.
2. An increase in hydrostatic compressive stress can increase the resistance
to yield.
3. Such materials may exhibit different behaviors in tension and
compression.
We discuss two widely used yield criteria for cohesive materials.

14.7.1 Mohr–Coulomb yield function

The Mohr–Coulomb yield criterion can be used for cohesive materials. It is a
generalization of the Tresca criterion. It accounts for the effects of hydrostatic
stress and uses two material parameters: the cohesion c and the angle of
internal friction φ. For ordered principal stresses of σ1 > σ2 > σ3 , the Mohr–
Coulomb yield function is [4]
f = σ1 − σ3 + (σ1 + σ3 ) sin φ − 2c cos φ (14.42)

14.7.2 Determination of material parameters

The material parameters c and φ can be determined via experiments.
First, we can set a uniaxial tension failure test, which gives σ1 =
σY T , σ2 = σ3 = 0, and the yield stress σY T can be found using Eq. (14.42):
2c cos φ
σY T = (14.43)
1 + sin φ
Next, we set a uniaxial compression failure test, for which we have this
time σ1 = σ2 = 0, σ3 = −σY C , and the yield stress σY C can be found using
Eq. (14.42):
2c cos φ
σY C = (14.44)
1 − sin φ
578 Mechanics of Materials: Formulations and Solutions with Python

Solving now Eqs. (14.43) and (14.44) simultaneously gives

1√
c= σY T σY C (14.45)
2
and
σY C − σY T
φ = sin−1 (14.46)
σY C + σY T

For a frictionless material (φ = 0), the Mohr–Coulomb yield criterion

reduces to the Tresca criterion.

14.7.3 Mohr–Coulomb yield surface

The Mohr–Coulomb yield surface is an irregular hexagonal pyramid. The
axis of the pyramid is still the hydrostatic axis, as shown in Fig. 14.6.

14.7.4 Drucker–Prager yield criterion

The Drucker–Prager yield criterion is a generalization of the von Mises cri-
terion, taking the hydrostatic stress into consideration [5]. In reference to
Eq. (14.38), the Drucker–Prager yield function can be written as

f = αI1 + |J2 | − K (14.47)

Figure 14.6. Schematic view of the Mohr–Coulomb yield surfaces. Left: an irregular
hexagonal pyramid in 3D space of the principal stresses (σ1 , σ2 , σ3 ); right: irregular hexagon
in 2D space (σ1 , σ2 ). Image modified base on the one from Wikimedia Commons by Janek
Kozicki under the CC BY-SA 3.0 license.
 Failure Criteria and Failure Analysis 579

Figure 14.7. Schematic view of the Drucker–Prager yield surfaces. Left: the Drucker–
Prager yield surfaces is a circular cone in 3D space of the principal stresses (σ1 , σ2 , σ3 );
right: the Drucker–Prager yield surfaces is an ellipse (black) in 2D space (σ1 , σ2 ); the irreg-
ular hexagon is for the Mohr–Coulomb yield surface. Image on the left is from Wikimedia
Commons by Janek Kozicki under the CC BY-SA 3.0 license. On the right is by Vtzslav
tembera under the CC BY-SA 4.0 license.

where α and K are coefficients that depend on the cohesion c and the angle
of internal friction φ:

2 sin φ
α= √ (14.48)
3(3 − sin φ)

and

6c cos φ
K=√ (14.49)
3(3 − sin φ)

For a frictionless material (φ = 0), the Drucker–Prager criterion reduces

to the von Mises criterion.
The Drucker–Prager yield surface is a circular cone, as shown in Fig. 14.7.

14.8 Comparison of different criteria

We finally use the case study problem defined in Section 14.2, and the prin-
cipal stresses obtained in Section 14.2.3, to compute the material yield stress
value needed to have point A stay below yielding and determines the yield
state. This time we put all the results from the different criteria together for
easy comparison.
580 Mechanics of Materials: Formulations and Solutions with Python

1 print(f'\nThe yield stress of the material: {sY:0.5e}\n')

2
3 # maximum principal stress criterion
4 ps_max = max(abs(eigens))
5 print(f'The maximum principal stress is {ps_max:0.5e}')
6 print(f'The material yield stress needs to be > {ps_max:0.5e}')
7 print(f'Is yield? {ps_max>sY}\n')
8
9 # maximum principal strain criterion
10 se1 = eigens[0]- *eigens[1]- *eigens[2] # principal strains E ε1
11 se2 = eigens[1]- *eigens[0]- *eigens[2] # principal strains E ε2
12 se3 = eigens[2]- *eigens[0]- *eigens[1] # principal strains E ε3
13 e_max =max(abs([Link]([se1, se2, se3])))
14 print(f'The maximum principal strain is {e_max:0.5e}')
15 print(f'The material yield stress needs to be > {e_max:0.5e}')
16 print(f'Is yield? {e_max>sY}\n')
17
18 # Maximum shear stress (Tresca) criterion
19 Tau_max =0.5*(eigens[0]-eigens[2])
20 print(f'The maximum shear stress is {Tau_max:0.5e}')
21 print(f'The material yield stress needs to be > {(2*Tau_max):0.5e}')
22 print(f'Is yield? {(2*Tau_max)>sY}\n')
23
24 # von Mises criterion
25 svm = gr.von_Mises(eigens)
26 print(f'The von Mises stress is {svm:0.5e}')
27 print(f'The material yield stress needs to be > {svm:0.5e}')
28 print(f'Is yield? {svm>sY}\n')
29
30 # Octahedral shear stress criterion
31 _, tau_oct=gr.Oct_stress(eigens)
32 print(f'The Octahedral shear stress is {tau_oct:0.5e}')
33 Y = tau_oct*3/[Link](2)
34 print(f'The material yield stress needs to be > {Y:0.5e}')
35 print(f'Is yield? {Y>sY}\n')
36
37 # Strain-energy density criterion
38 U_stress = gr.energy_stress(eigens, )
39 print(f'The strain-energy stress is {U_stress:0.5e}')
40 print(f'The material yield stress needs to be > {U_stress:0.5e}')
41 print(f'Is yield? {U_stress>sY}')
 Failure Criteria and Failure Analysis 581

The yield stress of the material: 7.50000e+08

The maximum principal stress is 6.18043e+08

The material yield stress needs to be > 6.18043e+08
Is yield? False

The maximum principal strain is 6.51365e+08

The material yield stress needs to be > 6.51365e+08
Is yield? False

The maximum shear stress is 4.27058e+08

The material yield stress needs to be > 8.54115e+08
Is yield? True

The von Mises stress is 7.42623e+08

The material yield stress needs to be > 7.42623e+08
Is yield? False

The Octahedral shear stress is 3.50076e+08

The material yield stress needs to be > 7.42623e+08
Is yield? False

The strain-energy stress is 7.15700e+08

The material yield stress needs to be > 7.15700e+08
Is yield? False

Readers my do the same analysis for point B on the shaft using the codes
given and examine the differences.

14.9 Remarks

A number of failure criteria have been introduced in this chapter with

detailed formulations and Python codes. The choice of the criteria shall
depend on the material. Therefore, experiences on the material to be used
and the databases of material properties for different types of materials are
of importance.
The author was involved in one project that used a newly designed com-
posite material. We did not have sufficient data on the material, so we
decided to write a code to assess the safety using a number of criteria together
to ensure the safety for possible scenarios we can think of.
There is a large class of materials called hyper-elastic materials, such as
rubbers widely used in industry and soft issues involving biological systems.
For these types of materials, the analyses are more on the large geometrical
deformation, and yielding is not usually a concern. This is an important and
582 Mechanics of Materials: Formulations and Solutions with Python

currently active area of study. Examples of study on these types of materials

from the author’s group can be found in Refs. [6–11].
When the structure is designed for one-time use, we may stretch to use
the full capacity of the material and allow the material to go beyond the
yield point, into plastic deformation, until fracture. This involves plastic-
ity analyses. It requires yield criterion together with: 1. a proper flow rule
governing the plastic strain and stress evolution after yielding, and 2. a
hardening rule that predicts changes in the yield surface. Examples for such
analysis can be found in textbooks. Some studies on these types of materi-
als in the authors; group can be found in Refs. [12–14]. Interested readers
may pursue this topic, starting from the wikipage on flow plasticity theory
([Link] wiki/Flow plasticity theory), and the references
therein.

References
[1] A.P. Boresi and R.J. Schmidt, Advanced Mechanics of Materials. John Wiley &
Sons, New York, 2002.
[2] J.M. Gere and S.P. Timoshenko, Mechanics of Materials. Van Nostrand Reinhold
Company, New York, 1972.
[3] S. Timoshenko and J.N. Goodier. Theory of Elasticity. McGraw Hill, New York,
1970. [Link]
elasticity&#38;ei=ICiKSsr3G4jwkQSbxMyPCg.
[4] J. Lubliner, Plasticity Theory. Macmillan, 1990. [Link]
d=LUqYPwAACAAJ.
[5] W.F. Chen and D.J. Han, Plasticity for Structural Engineers. Gau Lih Company,
Limited, 1995. [Link]
[6] G.R. Liu and T.T. Nguyen, Smoothed Finite Element Methods. Taylor and Francis
Group, New York, 2010.
[7] G.R. Liu, Mesh Free Methods: Moving Beyond the Finite Element Method. Taylor
and Francis Group, New York, 2010.
[8] Wu Shao-Wei, Jiang Chen, Jiang Chao et al., A selective smoothed finite element
method with visco-hyperelastic constitutive model for analysis of biomechanical
responses of brain tissues. International Journal for Numerical Methods in Engi-
neering, 121(22): 5123–5149, 2020.
[9] S. Wu, C. Jiang, C. Jiang et al., A selective S-FEM with visco-hyperelastic model
for analysis of biomechanical responses of brain tissues. In The 11th International
Conference on Computational Methods (ICCM2020), 2020.
[10] Wu Shao-Wei, Jiang Chao, Liu GR et al., An n-sided polygonal selective smoothed
finite element method for nearly incompressible visco-hyperelastic soft materials.
Applied Mathematical Modelling, 107: 398–428, 2022.
[11] Li Yang, Sang Jianbing, Wei Xinyu et al., A novel constitutive parameters identi-
fication procedure for hyperelastic skeletal muscles using two-way neural networks.
International Journal of Computational Methods, 19(02): 2150060, 2022.
[12] Zeng W., Larsen J.M. and Liu G.R., Smoothing technique based crystal plasticity
finite element modeling of crystalline materials. International Journal of Plasticity,
65: 250–268, 2015.
 Failure Criteria and Failure Analysis 583

[13] Cui X.Y., Liu G.R., Li G.Y. et al., Elasto-plasticity analysis using edge-based
smoothed finite element method (ES-FEM). International Journal of Pressure Ves-
sels and Piping, 86: 711–718, 2009.
[14] Zeng W., Liu G.R., Li D. et al., A smoothing technique based beta finite ele-
ment method (βFEM) for crystal plasticity modeling. Computers & Structures, 162:
48–67, 2016.
This page intentionally left blank
 Index

A box shearCenter, 497

actual safety factor, 565 bulk modulus, 195–196
admissible BCs for beams, 292
algebraically closed, 11 C
angle of the principal axis, 118 C-beam, 474, 491, 499
angular change, 164 C2toC4, 191
anti-commutativity, 31 C4toC2, 192
anti-symmetric, 32 cantilever beam, 292
antisymmetric matrix, 152 carbon/epoxy composite, 228
apply symmetry, 192 Cauchy–Green deformation tensor, 158
area of a composite area, 417 centroid for composite areas, 422
area projection, 88 centroid of an area, 421
axial flexibility of a bar, 251 centroid of an I-beam, 422
axial stiffness of a bar, 251 characteristic equation, 107
characteristic value, 107
B clamped-clamped beam, 297, 317
bar with non-circular cross-section, 376 Clapeyron’s theorem, 507
bars with varying cross-section, 509 closure, 10
basic property preservation, 72 code for computing various stresses,
beam equation, 290 112
beam solver, 294–295 cohesive materials, 577
beam with two spans, 321 collinearity, 31
bended beams, 283 comparison of different criteria, 579
bending deformation, 284 compatibility condition, 505
bending stiffness, 314 compatibility equation, 177–178, 507,
bi-linear function, 212 516, 542, 546
body force, 74, 77 complementary energy, 513, 537
boundary conditions for beams, 291 complementary energy 1D problems,
box cross-section, 372 538
box cross-section of thin-walls, 373 complementary energy density, 527,
box-beam, 492 537

585
586 Mechanics of Materials: Formulations and Solutions with Python

complementary energy in 3D, 537 curl of the stress vector, 355

complementary energy principle curl vf, 70
(CEP), 502, 542
complementary energy principle for D
structures, 539 1D bar problems, 244
complete and unique, 141 1D beam member, 284
compliance equations for 2D problems, 2D plane strain, 198
208 2D plane stress, 197
compliance matrix, 200 2D stress states, 116
compliance matrix for isotropic 3D free-body diagram, 234–235
materials, 207 deflection, 284
complimentary, 295 deflection in its third derivative, 290
complimentary energy principle, 546 deformation gradient, 155, 157
components of a force, 78 derivative of deflection, 286
composite bar, 275 derivative of the moment, 290
computation of large strains, 155 derivatives of displacement functions,
computation of various strains, 175 148
computing cross products, 33 description of vectors, 42
concentrated force, 73 deviatoric plane, 576
conditions for the stress function, 382 deviatoric stress, 200
conservation law, 85 deviatoric stress tensor, 575
constitution equations for stress and diagonalization of a stress tensor in
strain invariants, 197 3D, 133
constitutive equation, 184, 513 differential equation solver, 299
constructing the stress function, 384 dimensionality and stresses, 85
continuity and equilibrium conditions, direction cosine, 19–20, 24, 48, 87, 163
332 direction cosines of the principal axes,
continuity equation, 276 107
continuity in rotation, 324 direction of the force, 75
contraction of index, 67 direction of the surface, 75
contraction operations, 96 direction-independent, 197
coordinate rotation, 425, 433 discontinuity of the solutions, 321, 328
coordinate systems, 42 discontinuous body forces, 264
coordinate transformation, 41, 46, 171 displacement boundary conditions, 246
coordinate transformation in 3D, 50 displacement components on the
coordinate transformation rule of cross-section, 353
strain tensor, 172 displacement functions, 142
coordinate translation, 428 displacement gradient, 148–150
cross product, 27–29, 31, 70, 107 displacement gradient caused by
cross-section, 75 temperature, 270
cross-section of beams, 287 displacement gradient matrix, 150
cross-section rotation assumption, displacement method, 247
357 displacement vector, 140
cross-section surface, 74 displacement-driven, 261
cubic materials, 194 displacements derivatives, 154
curl of a vector function, 69 distortional energy density, 111
 Index 587

distortional energy density criteria, 573 equilibrium, 68

distortional energy density criterion, equilibrium condition, 515, 537
576 equilibrium equation, 76–77, 233, 236,
distortional strain-energy density, 573 506
distributed force, 74, 264 equilibrium equation for beams, 289
distribution of solutions, 342, 345 equilibrium equation in terms of
div tensorF, 239 stresses, 234
divergence-free tensor, 355 equilibrium equations, 378, 535–536,
divergence of a vector function, 67 541
divergence of stress tensor function, equilibrium equations for 2D plane
238 stress, 244
divergence of the stress, 77 equilibrium equations in
divergence of the stress states, 355 displacements, 240
divergence of the stress tensor, 240 equilibrium in mechanics, 72
divergence of the stress vector, 237 equilibrium in moment, 324
divergence of the tensor function, 239 equivalence of force vector, 78
divergence theorem, 523 equivalent polar 2nd moment of area,
division by zero, 11 396
dot product, 30, 37, 67 error magnification, 461
dot product of vectors, 18 Euler–Bernoulli assumptions, 285
double angle formula, 116 Euler–Bernoulli beam theory, 284
Drucker–Prager yield criterion, 578 external action, 504, 514
Drucker–Prager yield surfaces, 579 external force, 237
dummy index, 52 external modules, 5
dynamic equilibrium equation, 290 external packages, 4
externally applied force, 76
E extreme values, 302
E SnC2D strain, 206
E SnC2D stress, 203 F
E SnC3Dep, 202 features of a vector, 13
effective stress, 562 fictitious unit load method, 553–555
eigenvalue problem, 106 first axis of the stress tensor, 89
eigenvalue solver, 435 first moment of a composite area, 419
elasticity tensor, 181 first moment of area, 418, 425
elasticity tensor transformation, 221 first-order tensor, 43
element-wise division, 18 force vector in transformed
element-wise multiplication, 18, 37 coordinates, 79
elliptical cross-section, 384–385 force vectors in 3D space, 78
enclosed thin-wall, 369 force-driven problem, 255
enclosed thin-wall cross-sections, 368 forces in channels, 486
energy conjugate, 524–525 formula for shear stress, 468
energy methods, 501 forward transformation, 141
energy principles, 501 fourth-order differential equation, 284
energy stress, 112 fourth-order tensor, 184, 241
engineering constants, 195 free-body diagram, 76, 276, 324, 467
engineering strain, 161 free-clamp supported beams, 312
588 Mechanics of Materials: Formulations and Solutions with Python

free-indexes, 52 indeterminate beams in temperature

frictionless material, 578 gradient, 340
from stress tensor symmetry, 186 indeterminate structures, 557
function concept, 46 inner product, 18, 22
function of coordinates, 60 inner product associated norm, 20
internal, 295
G internal force, 76
internal reaction, 504, 514
general definition of stress, 80
inverse transformation, 97, 105, 142,
general solution, 251, 293
185
generalization of the Tresca criterion,
isotropic materials, 194
577
Ixx2IXX, 434
generalized deformation, 508
generalized displacement, 517, 531 K
generalized Hooke’s Law, 184
generalized internal force, 508 k1 f, 397
generating 2D Mohr circles, 121 k f, 397
geometric representation, 115 kinematic relations, 145
Kronecker delta, 52
geometric representation of tensors,
115 Kronecker delta function, 21
governing equation, 233
L
[Link] vf(), 68
[Link] vf(), 66 L-beam, 455, 477
grad vf, 149 Laplace equation, 389
gradient of a scalar function, 61, 64 large deformation, 160, 164
gradient of the displacement vector, large strains, 153
149 law of cosines, 21
gradient of vector functions, 66 linear body force, 260
gradient operator, 61 linear combination, 15–16
Green–Lagrange Strain, 153 linearized expression, 536
Green–Lagrange strain tensor, 158 linearly distributed normal stress, 286
linearly independent, 15, 28
H loading plane, 442
loading plane angle, 459
Hadamard product, 18
hollow cylindrical cross-section, 361, M
363
M eign, 112
hollow enclosed thin-wall, 408
M invariant, 112
hydrostatic axis, 576
magnification factor, 160–161
hydrostatic stress, 199, 576
magnitude of the shear stress, 91
hyper-elastic materials, 581
matrix-vector multiplication rule, 89
maximum moment, 446
I
maximum principal strain, 569
I-beam, 392, 407, 479, 483 maximum principal strain criterion,
idealized deformed membrane, 409 567
in-plane strains, 242 maximum principal stress, 566
 Index 589

maximum principal stress criterion, O

565 Oct stress, 112
maximum shear stress, 397, 571 octahedral shear-stress, 574
maximum shear stress yield criterion, octahedral shear-stress criterion, 575
572 octahedral stress, 110
maxminS, 301 operations for vectors, 17
mean stress, 199 orthogonal coordinate, 46
meanDeviator, 112 orthogonal coordinate system, 43, 79
means stress, 576 orthogonal fibers, 164
membrane analogy, 387 orthogonal moment components, 443
membrane deflection, 389 orthogonal projection, 20, 43–44, 72,
modern steam turbine, 350 88, 90–91, 141
Mohr circle for 2D, 119–120 orthogonal projection rule, 42, 45, 48,
Mohr circle for 3D, 126, 128 78, 140
Mohr circle of a uniaxial stress state, orthogonal vectors, 39
122 orthogonality of vectors, 40
Mohr circles, 115 orthotropic materials, 193
Mohr–Coulomb yield criterion, 577 outer product, 25–26, 65, 149
Mohr–Coulomb yield function, 577
Mohr–Coulomb yield surface, 578–579 P
Mohr circle2D, 121
π-plane, 576
Mohr circle3D, 128
P-wave modulus, 195
moment equilibrium, 84–85
parabolic distribution, 469
multi-valued, 56
parabolic distribution of shear stress,
multiple multiple-type members,
471
547
partial differential equations (PDEs),
233
N
perpendicular axis theorem, 438
narrow rectangular cross-section, 389 piecewise solutions, 319, 326, 335
neutral axis, 285–286, 445, 454 planar truss structure, 555–556
non-circular cross-sections, 375 plane strain, 204
non-symmetrical bending, 452 plane strain problems, 242
non-uniqueness of the eigenvectors, 109 plane stress, 203
nonlinear spring–mass system, 543 plane stress problems, 243
normal direction of the surface, 90 plastic deformation, 582
normal strain, 147, 286 plot2curveS, 256
normal strain along a fiber, 160 point force, 74
normal stress, 75, 286, 294 Poisson’s ratio, 183, 246
normal stress by moments, 441 polar moment of area, 438
normal stress distribution, 452 polar second moment of area, 358–359
norms of a vector, 24 possible BCs for beams, 291
notation convention for a stress possible boundary conditions, 291
component, 83 potential energy, 503
Numpy code for transformation potential energy for 1-DOF systems,
matrices, 54 507
590 Mechanics of Materials: Formulations and Solutions with Python

potential energy for various members, representation of a vector, 141

509 resultant moment, 287
potential energy principle (PEP), 502, right-hand rule, 28, 46, 91, 107
506, 541 rotation angle, 285
potential energy principle for rotation matrix, 58–60
structures, 530 rotation-displacement relations, 152
Prandtl stress function, 378 roundS, 301
principal axis, 106, 434
principal axis angle, 131 S
principal directions, 108 same contraction effects, 97
principal moments of area, 434 scalar product, 19
principal plane, 106 scalar triplet product, 29
principal second moments of area, scaling the vector, 16
434–435 second derivative of deflection, 286
principal strains, 173 second moment of area, 287, 426, 428,
principal stresses, 105, 107, 564 433
principal stresses for 2D stress states, second moment of area for regular
117 areas, 426
principal stresses of 2D stress states, second moments of an I-beam, 429
118 second moments of an S-beam, 431
principal2angle, 118 second moments of composite areas,
principalS, 108 429
principals ypr, 134 secondMoments I, 430
principle of equilibrium, 44–45 separation of variables, 394
principle of superposition, 546 sf phi, 403
principle of virtual work, 512, 530, 539 shaft undergoes torsional deformation,
problem well defined, 253 350
property of an area, 415 shaft with two segments, 365
pure bending, 316 shear center, 442
shear correction coefficient, 508
Q shear flow, 369, 409
shear force, 294
q-maps, 496
shear force and moment, 466
shear force correction, 520
R
shear strains, 147
randomly generated matrix, 57 shear strains in non-circular bars,
randomly generated vectors, 34 377
rank-one matrix, 26 shear stress by shear force, 465
rate of change in length, 146 shear stress distribution (q-map),
rate of circulation, 70 475
rate of rotation, 70 shear stress equivalence, 84–85
reaction force, 266, 269 shear stresses on a surface, 81, 91
reactions with the rigid block, 276 shear stresses on rectangular
rectangular box-beam, 493 cross-section, 405, 469
rectangular cross-section, 393 shear-center, 489–490
redundant support, 558 shear-center location, 499
 Index 591

shear-flow, 469, 474 stress boundary conditions, 89, 246

shear-flow in the web, 485 stress components, 82
sign convention, 76 stress components on surfaces, 83
sign convention for 2D Mohr circles, stress divergence, 77
120 stress function, 379
simple-clamped beams, 310, 340 stress function on rectangular
simple-simple supported beams, 304 cross-section, 403
slender geometry, 350 stress invariants, 110
slope of the deflection, 285 stress method, 247
smoothness of the strain energy stress tensor transformation rule, 96
density, 187 stress transformation, 185
solid cylindrical shaft, 562–563 stress transformation via Yaw, Pitch,
solution exists, 11 and Roll, 132
solution for the deflection, 293 stress vector projection, 87, 90
solution for the rotation, 294 stress-strain-temperature relation, 221
solutions at specific locations, 303 stress rect, 405
solutions for bars, 252 stresses on an arbitrary surface, 88
solver1D2, 252 stresses on the cross-section, 353
solver1D4, 296 stressOnN, 91
spatial rate of loss of the stresses, 239 superposed stress distribution, 444
spring–mass system, 540 superposition of stresses, 443
square cross-section, 350 swap the order of the two indexes, 80
standard arithmetic operations, 10 swapping the indexes, 54
standard basis vectors, 15, 35 symmetric bending, 454, 457
standard convention, 76 symmetric frame, 328–329
standard sign convention, 235, 295 symmetric relations, 201
static problem, 237 symmetric tensor, 85
statically determinate structures, 546 system of equations, 247
statically indeterminate, 266
stiffness matrix of materials, 200 T
strain energy, 524 T-beam, 449
strain energy density, 188, 527 tall beam, 459
strain energy density for 1D, 525 tall rectangular beam, 460
strain energy density for 3D, 529 temperature gradient, 338
strain energy in 3D, 522 tensile spacemen, 182
strain energy stress, 112 tensor concept, 44
strain invariants, 174 tensor operation for transformation,
strain tensor symmetry, 186 144
strain tensor transformation, 172 tensor vs. matrix, 190
strain transformation, 185 Tensor1 transfer, 144
strain–displacement relation, 148, 151, Tensor2 transfer, 99
159, 177, 353, 377 Tensor4 transfer, 186
strain-energy density, 570 tensorT2angle, 117
strain-energy density criterion, 569 thermal deflection, 340
strain-stress-temperature relation, 221 thermal displacement, 267
stress and strain symmetry, 186 thermal expansion coefficients, 279
592 Mechanics of Materials: Formulations and Solutions with Python

thermal expansion effects, 220 twist-torque relation, 358

thermal forces, 268 two indexes for a stress component,
thermal strains, 220 82
thermal stress, 220, 267–268, 273, two-bar truss, 531, 544
337 two-side application, 58
thermal stress in beams, 337 twobars bf, 270
thermal stresses, 269 types of forces, 73
thick beams, 520
thin beam theory, 284 U
thin beam under bending, 285 U for various structural members,
thin-wall assumptions, 375 512
thin-wall cross-sections, 362, 472 under- or overflow, 11
torque and shear-flow relation, uniaxial compression failure test,
370 577
torque on the cross-section, 383 uniaxial force, 101
torque stress relation, 411 uniaxial stress state, 102
torque twist-angle solution, 396 uniaxial stress tensor, 101
torsional bar members, 349 uniaxial tension failure test, 577
torsional flexibility of the bar, 360 unit vectors, 14
torsional stiffness, 386 unitary matrix, 52–53
torsional stiffness of the bar, 360 unitary property, 55
total angular change, 147 universal testing machine, 182
total derivative, 381 use of membrane analogy, 408
total potential energy, 502
total potential energy principle, V
535
tractions on an arbitrary surface, 86 variation, 506
tractions on the surface, 87 vector operations, 34
transformation matrices, 131 vector property preservation, 53
transformation matrix, 47–48 vector space, 17
transformation matrix for 3D, 49 vectors, 12
transformation of fourth-order tensor, vectors in mechanics, 42
185 version mismatch, 4
transformation of strain tensor, 171 virtual displacement, 502, 505
transformation of stress tensor in 2D, virtual displacement for structures,
99 530
transformation rules, 45 virtual force, 512–513, 539
transversely isotropic materials, Voigt notation, 190
193 volume of a parallelepiped, 29
Tresca criterion, 571, 576 von Mises, 112
Tresca yield surface, 571–572, 574 von Mises criterion, 573, 577
triangle cross-section, 386 von Mises stress, 111, 573
twist angle, 349, 410 von Mises yield surface, 574
twist deformation, 352
twist-angle and torque relation, 360, W
371 well posed, 11
 Index 593

work done by external forces, 523 yield surface, 566

work-done, 22 yielding criterion, 562
Young’s modulus, 182
Y
Yaw, Pitch, and Roll, 131 Z
yield function, 562, 566 zero vector, 16