|Cone ae Agoheations

in Ge5omechanics

fer aa a” Dales Deb

ae sues
se
 Te
ae
_ ap

a
|
is

tp
pam TIA
— _ ' "TR AY i
ey
oe . a
_co
.

ue
ae

=
ree)
a
TRUE

Hoadley

Hey naa
.
ye
a

a
ti

GEE

PHI Learning (iivctio Limitiec

New Delhi-110001
2009
Rs. 250.00

FINITE ELEMENT METHOD—Concepts and Applications in Geomechanics

Debasis Deb

© 2006 by PHI Learning Private Limited, New Delhi. All rights reserved. No part of this book may
be reproduced in any form, by mimeograph or any other means, without permission in writing from
the publisher.

ISBN-978-81-203-2983-6

The export rights of this book are vested solely with the publisher.

Second Printing te te August, 2009

Published by Asoke K. Ghosh, PHI Learning Private Limited, M-97, Connaught Circus,
New Delhi-110001 and Printed by Mudrak, 30-A, Patparganj, Delhi-110091.
 To
my wife Samhita
and our angels Jayee and Ricky
 Digitized by the Internet Archive
in 2022 with funding from
Kahle/Austin Foundation

https://archive.org/details/finiteelementmetO000debd
Preface xi

Chapter 1 Analysis of Stresses and Strains 1-33

ea Introduction 1
1.2 Stress, Strain and Strain Energy |
1.2.1 Definition of Stress I
1.2.2 Definition of Strain Z
1.2.3 Stress and Strain Tensor 4
1.2.4 Strain Energy A
1.3 Transformation of Displacements, Strains and Stresses 7.
1.3.1 Transformation of Displacement Vector &
1.3.2 Transformation of Strain Tensors 8
1.3.3. Transformation of Stress Tensors 9
1.3.4 Transformations in Three Dimensions 11
1.4 Stress Analysis res
1.4.1. Principal Stresses in Two Dimensions Lo
1.4.2 Principal Stresses in Three Dimensions He)
1.4.3 Maximum Shear Stress 18
1.4.4 Stresses on Octahedral Plane 18
1) Mohr’s Circle of Stress in Two Dimensions 20
1.5.1 Steps for Constructing Mohr’s Circle 20
1.5.2 Inference from the Mohr’s Circle of Stress 20
1.5.3. Mohr’s Circle of Special Stress Conditions oz
1.6 Analysis of Strain Pi}
1.6.1 Principal Strains in Two Dimensions 24
1.6.2 Principal Strains in Three Dimensions 24
1.6.3 Maximum Shear Strain Zo
1.6.4 Mohr’s Circle of Strain in Two Dimensions Be)
1.6.5 Volumetric Strain i.
Xe Re
1.7 Stress Equilibrium Equations 28
1.8 Strain Compatibility Conditions 30
Summary 30 ;
Exercises 31

Chapter 2 Stress-Strain Relationships 34-50

2.1 Introduction 34
2.2 Linear Isotropic Elasticity oD)
2.2.1 Relationship between Principal Stresses and Principal Strains oo
2.2.2 Stress-Strain Relationship in Two Dimensions 40
2.3 Linear Transverse Isotropic Elasticity ad
2.4 Degree of Anisotropy 46
2.5 Non-Linear Elasticity 46
2.6 Plasticity 47
2.6.1 Perfect-Plasticity 48
_ 2.6.2 Elastic-Plasticity 48
Summary 49
Exercises 50

Chapter 3 Finite Element Method in Elasticity: Isoparametric

Triangular Elements 51-91
3.1 Introduction oF
3.2 Discretization of Physical System with Finite Elements Sys
3.2.1 Concept of Continuum Mechanics in One Dimension a
3.2.2 Discretization with Finite Element in One Dimension 54
3.2.3 Discretization with Finite Element in Two Dimensions 5)
3.3. Nodal Shape Function 56
3.3.1 Natural Coordinate System for Triangular Element Sige
3.3.2 Shape Functions of Triangular Element 58
3.4 Elemental Contributions or Element Equations 61
3.4.1 Constant Strain Triangle (CST) 65
3.4.2 Linear Strain Triangle (LST) 66
3.5 Variational Method of Finite Element Formulation 66
3.6 Equilibrium Equations for Multi-Degree of Freedom 68
3.7 Numerical Integration Scheme 70
3.7.1 Stiffness Matrix 71
3.7.2 Body Force 72
3.7.3. Traction Force for Straight-Sided Triangle Hiei
3.7.4 General Formulation of Traction Force for Curved/Straight-Sided
Triangle 75
3.7.5 PomtsForce IGE
3.8 Assembly into Global Simultaneous Unreduced Equations iis
3.9 Application of Essential Boundary Conditions 79
3.9.1 Penalty Method 7?
3.9.2 Row and Column Adjustment Method 80
3.10 Solutions of Primary Variables &1
3.11 Determinationof Strains, Stresses and Reaction Forces &3
3.11.1 Calculation of Strain and Stress Vector for Each Element 83
3.11.2 Calculation of Reaction Forces &3
3.12 Stresses at Nodes 5&4
3.13 Interpretation of Results 8&5
3.14 Tunnel Example &6
3.14.1 Closed-form Solution 56
3.14.2 Finite Element Solution &8
Summary &9
Exercises 90

Chapter 4 Quadrilateral Finite Elements 92-109

4.1 Introduction 92
4.2 Discretization with Quadrilateral Elements 92
4.3 Natural Coordinates for Quadrilateral Element 93
4.4 Determination of Shape Functions UB
4.4.1 Direct Method 9
4.4.2 Serendipity Method 95
4.4.3 Lagrange’s Interpolation Method 96
4.5 Strain-Displacement Matrix 98
4.6 Stiffness Matrix and Numerical Integration Scheme 99
4.7 Body Force 100
4.8 Traction Force 101
4.9 Stresses at Nodes 104
4.10 Tunnel Example 106
Summary 107
Exercises 108

Chapter 5 Axisymmetric and 3D Finite Element Method 110-134

5.1 Introduction 110
5.2 Axisymmetric Analysis 110
5.2.1 Formulation of B and K Matrices for Axisymmetric Analysis 1/1
224i taction Force Tis.
352.3% Body Force LID
5.2.4 Comments on Stiffness Matrix and Stress Computations 115
5.3. Tunnel Example 115
5.4 Finite Elements in Three Dimensions 116
5.4.1 Discretization in Three Dimensions LIZ
5.4.2 Shape Functions of Tetrahedral Elements 118
<a ee
CT |
5.4.3. Hexahedral or Brick-Type Element 119
5.4.4 Formulation of B matrix 120
5.4.5. Formulation of Stiffness Matrix 124
5.4.6 Traction Force 125
5.4.7. Body Force 126
5.4.8 Determination of Strains, Stresses and Reaction Forces 127,
5.5 Example of Shaft Pillar Design in Three Dimensions D27
5.5.1. Problem and Model Description 127
5.5.2 Results and Discussions 129
Summary 133,
Exercises ISB!

Chapter 6 Rock and Rock Mass Failure Criteria 135-157

6.1 Intreduction 135
6.2 Invariants of Deviatoric Stress 138
6.3. Mohr-Coulomb (MC) Yield Criterion 142
6.3.1 MC Criterion in Principal Stress Space—Two Dimensions 142
6.3.2 MC Criterion in Deviatoric Plane 144
6.4 Drucker-Prager Criterion 146
6.5 Hoek-Brown (HB) Criterion 147
6.5.1 HB Criterion in Principal Plane—Two Dimensions 149
6.5.2 HB Criterion in Deviatoric Plane 192
6.6 Alternative Criteria Based on Hoek-Brown Yield Surface 153
6.7 Tensile Yield Criterion (ee
Summary 156
Exercises 156

Chapter 7 Elastic-Plastic Finite Element Analysis 158-191

7.1 Introduction 158
7.2. Non-Linear Solution Methods 158
7.2.1 Newton—Raphson Method 159
7.2.2 Modified Newton—Raphson Method 160
7.2.3 Incremental Method or Euler’s Method 161
7.2.4 N-R Scheme for Multiple Degrees of Freedom 163
7.3 Convergence Criteria for Non-Linear Solutions 165
7.4 Elastic-Plastic Analysis Method 166
7.4.1 Yield Criteria 166
7.4.2 Hardening/Softening Laws 167
7.4.3. Flow Rule 169
7.5 Elastic-Plastic Stress-Strain Relationship 170
7.5.1 Implementation of Shear Yield Criteria for Rock and Rock Mass 171
7.5.2 Implementation of Tensile Yield Criterion 176
7.5.3. Improvement of Solution Accuracy: Numerical Procedure 178
wer Contents | ix

7.6 Tunnel Example—Elastic-Plastic Analysis 180

7.6.1. Analytical Solution—Mohr-Coulomb Media 180
7.6.2 Alalytical Solution—Hoek-Brown Media 184
7.6.3, Non-Linear Finite Element Solutions and Comparison with Analytical
Results 184 .
Summary 189
Exercises 190

Chapter 8 Strength and Deformability of Jointed Rock Mass 192-220

8.1 Introduction 192
8.2 Fracture Strength of Jointed Rock Mass oz
8.2.1 Fracture Strength with Single Joint 193
8.2.2 Fracture Strength with Two Joints 19%
8.2.3 Fracture Strength with Multiple Joints 196
8.3 Shear Strength of Rock Joint 197
8.3.1 Barton’s Empirical Shear Strength Envelop 199
8.3.2 Scale Effect on JRC and JCS 203
8.3.3. Ladanyi and Archambault’s Shear-Strength Envelop 204
8.3.4 Mohr-Coulomb Shear Strength Envelop 204
8.3.5 Instantaneous Cohesion and Friction Angle from Barton’s Envelop 205
8.3.6 Instantaneous Cohesion and Friction Angle from Hoek—Brown Criterion 206
8.4 Deformability of Rock Joint 206
8.4.1 Joint Normal Deformation 206
8.4.2 Joint Shear Deformation and Dilatancy 208
8.5 Concept of Joint Compliance 210
8.5.1 Joint Compliance Matrices 212
8.5.2 Equivalent Compliance for Joints Sets 214
8.5.3 Transformation of Joint Compliance 214
8.5.4 Joint Compliances for Multiple Joint Sets 214
8.6 Elastic-Plastic Analysis of Rock Joints 216
8.6.1 Mohr-Coulomb Shear Yield Criterion 216
8.6.2 Barton’s Shear Yield Criterion PAW
Summary 217,
Exercises 218

Chapter 9 Finite Element Procedures of Rock Joints 221-244

9.1 Introduction 22]
9.2 The Discrete Joint Elements es
9.2.1 Goodman’s Joint Element: Direct Formulation 223
9.2.2 Numerically Integrated Joint Element 226
9.3 Analysis Procedure for Non-Linear (Elastic-Plastic) Joints ass
9.3.1 Tensile Yield Function ike:
9.3.2 Shear Yield Function JLRS
aa Contents

9.3.3. Plastic Corrections 229

9.3.4 Comments on Non-Linear Analysis of Joint 231
9.3.5 Implementation of Mohr-Coulomb and Barton’s Yield Criteria x8 |
9.4 The Equivalent Material Approach for Joints 233
9.4.1 Equivalent Material Model: Method I 234
9.4.2 Equivalent Material Model: Method II 238
9.4.3. Comparison between the Two Methods 239
9.5 Tunnel Example 239
9.6 Non-Linear Analysis with Equivalent Material 24]
Summary 243
Exercises 243

Appendix A: Galerkin Finite Element Method 245-249

Appendix B: Skyline Storage of Stiffness Matrix 250-252
References 253-257
Answers to Exercises 259-265
Index 267-269
Rock mechanics is an interdisciplinary subject dealing with field of applied researches and
activities comprising mining, civil constructions, dams, tunnelling, underground caverns and
excavations, slope stability, underground storage facilities for foods, fuels and waste materials
and others. For the last three decades, numerical methods, especially the finite element method,
have imbedded firmly and popularly into the analysis of rock engineering problems. The
theoretical developments in the field of non-linear behaviour of rock mass, the concept of
equivalent material approach to comprehend joint behaviour and modelling of discrete joints
have made it possible to analyse complex characteristics of rock mass using the finite element
method. Today, numerical methods stand first in designing, analyzing and optimizing structures
or excavations made in rock mass.
The finite element literature is very large to list all publications or even to discuss all
possible numerical procedures. This text is intended to address the application of the finite
element method in the area of rock mechanics. The material presented in this book is the
elaboration of a course of lectures originally prepared for the postgraduate mining engineering
students at Indian Institute of Technology (IIT), Kharagpur. The author believes that the
contents of this book are sufficient for one semester course work for senior undergraduate and
postgraduate students of civil, mining and geotechnical engineering. The professional engineers
involving with rock mechanics applications and design may also use this book profitably as a
reference or to learn about state-of-the-art techniques for the application of the finite element
method in their specialization.
This text emphasizes on the fundamental concepts of stress-strain analysis, isoparametric
finite element method, rock mass yielding/failure behaviour and their formulation in the finite
element procedure, joint behaviour as equivalent material and discrete system. Many solved
problems are given in the text so that the students find it easy to grasp the concepts and then
apply them to real-world problems. Particularly, tunnel problems are solved to emphasize the
application of the finite element method in geomechanics. Chapter | gives an introduction of
stress and strain analysis. It highlights the basic concept of transforming stress and strain tensors
and displacement vector from one coordinate system to another. Fundamentals of principal
xi
 19 SSS
stresses and strains, Mohr circle of stress and strains are explained in detail with worked-out
problems wherever they are appropriate. Chapter 2 deals with the constitutive relations of
isotropic and anisotropic elastic rocks. Detailed description of plane stress, plane strain and
axisymmetric conditions are given with examples. The concept of elastic-plastic properties of
rocks is also presented in this chapter. The basic concepts of isoparametric finite element
method are elaborated in Chapter 3 with linear and quadratic triangular elements. This chapter
introduces the concept of elemental (local) coordinate system for triangular elements. With the
help of worked-out problems, the concept of discretization, derivation of shape functions,
elemental equations and numerical integration scheme are explained in detail. Applications of
these techniques are also focused in this chapter. Chapter 4 elaborates the isoparametric finite
element formulation of linear and quadratic quadrilateral elements. Students will learn the
procedure to derive shape functions for serendipity and Lagrange’s elements. With suitable
examples, numerically integrated element is also” formulated emphasizing the calculation
procedure of stiffness matrix, ‘traction and body force vectors. Chapter 5 deals with the
axisymmetric and three-dimensional finite element procedures.
The fundamentals of rock mass yield/failure criteria such as Mohr-Coulomb, Drucker-
Prager, Hoek—Brown and alternative criteria are explained in Chapter 6. This chapter also
introduces the concept of deviatoric plane followed by derivation of yield criteria using
invariants of deviatoric stresses. Chapter 7 mainly deals with the procedure to implement yield
criteria into finite element method using classical plasticity theory. Non-linear solution methods
such as Newton—Raphson, modified NR and Euler’s method are explained giving suitable
examples. In this chapter, step-by-step procedures are given elaborate the concept of elastic-
plastic finite element analysis in geomechanics. This chapter helps students to learn the complex
procedures of plastic stress correction. Chapters 8 and 9 are devoted to the fundamental
concepts and implementations of rock joints into finite element procedure. Chapter 8 highlights
the strength and deformability of characteristics of joint sets and elaborates shear strength
envelops given by Mohr-Coulomb, Barton, and Ladanyi and Archambault. Chapter 9 includes
the implementation procedure of joint as a discrete plane and modelled with Goodman’s joint
element. Formulation of ubiquitous joints into finite element analysis is also discussed
considering the equivalent material approach. Suitable examples are worked out to elaborate
step-by-step procedures for elastic and elastic-plastic analysis.
The author has developed a two-dimensional fInite element code called FEMROCK using
C language and can be applied for elastic, elastic-plastic and jointed rock mass. This code can
be obtained by sending a request to the author at [email protected].
The prime motivation for writing this book came from students at IIT Kharagpur who
needed a text where both finite element method and rock mechanics principles are combined.
The author is greatly thankful for their support and encouragement. The author is especially
thankful to three of his students Abhiram Kumar Venna, Hemant Kumar and Sama Mayank
for helping him in checking some of the exercises. The author is greatly indebted to Dr. Y.P.
Chugh, Professor, Southern Illinois University, Carbondale (USA) and Dr. D.W. Park,
Professor, Civil and Environmental Engineering Department, The University of Alabama,
(USA) for their valuable guidance throughout his professional carreer. The author expresses his
deep gratitude to all of his colleagues in the Department of Mining Engineering, IIT Kharagpur
for their support during writing of this book. The author is also thankful to Dr. H.S. Shin,
i 4 Preface 3 on xiii

Director, Dr. S.O. Choi, and Dr. C. Sunwoo, Principal Researchers—al! of Korea Institute of
Geoscience and Mineral Resources (KIGAM), South Korea, for their help, support and
encouragement during writing of this book. Special thanks go to the publisher, Prentice-Hall
of India, for making necessary measures for publishing this book.
This book could not be completed without encouragement and loving support from the
author’s wife, Samhita, and their children, Jayee and Ricky. Finally, the author expresses his
deepest gratitude to his late father as well as his mother, brothers and in-laws for their
encouragement and support.

DEBASIS DEB
 = eens -

- elesacraal Minions and pone pair

these lechahyues arc-aigs featusedimite chegtt C
i Of bnear aed gumeedte-s
- preestace Serve a funcuers.fir wigtty aif
i neers
= < heen of pian 2 Segue and? body force ecicien,
Ps axis} munbtric and three-dimensional. finite element procedurts.
ay
:a © The lundameculs of tock mass yirid/feilure extasin sat as Moke
Prager, Hock—irewo 2nd altetnative certterta wre eeplained ih Crayne
7
. tnteduces: the consent of deviatoric plane follawedg by derivation
of
> fie atlants of Spviataric tresses, Chapter? muinty desis wily
the provediare’ >it cameagite
. e
ete hg Tine clement method using classical plasidity theory; Non-tinear
s¢; i ih
' itis:Neoxaa- tapheon, - rivdifted NR and. Filters mothed
are: lia] ‘
sages. In Unc chapter, etepdectep procericres are pivet: elaburalie the
<1
a _ phase Tate atenes suatyzis in geomechaties. Pha chapter helpe sider
tohe
= pocdures of fe site lee conection, Chapters 8 and 9 are dewotnd ic. ‘vide
- -Sonsepts aninplempmafions of mek Jot
| inte fire rion prooddure hegre
_ .
fhe Seeyth ond deformation od characteyistics Of pot iets and elatiinet
a 7 SMPEtope tea hy Moh: ietcanees Garton, wa} Ladaviyi
es shea a
one
and Anchamb eutt Chaps 9 i oe if
|< Bx siplementation otocetige of jolat.as a discrac plone wad ravidel ied with ar an's joi
.. teeieeat Fae: muito. oF UbbGuirous Gti: = >.
mtn- finite Gord anhivsis ts lee 4
_ airttebetias Dem, eqValent material ‘o:opicach Sy sable cxatonids ape worked out.
fo8
Saeb sep wecedutes for clastic Sad ehisric- piss arialy ais,
: "Stes de Bi slo & tee etencirers! +Tate Merent code Galles!
io oe: ‘ a
=
a oe at mnt fit elostic. edactepcas tec amd intel trek cos ‘Tie cade
ae oe 2) an ee i co ida Soe ah -

en < 7 ‘ +. ove Sines ie sae 2

feted di i -gati Ta wae Coste een! Sah
Ta .
Wie
5 erally Mine Ee Ao
ange; Professce”
cede (SA) ger Be fiWa :
Prot=tsor. Civil spd 4 : = -peeent.
Cio AV) Te ee wale
Te. Oe) copy Gf ae us i
Sigs, ts ia Vides ate mavreer, The 2abor #8
Ange ca rr
Os i SiggTs Seaaes i ie ~pement v8 Waring Siising,
ee oe acta
CY Khaeage
eeents ite Dy ThGes
> - > :
 Analysis of Stresses
and Strains |

1.1. Introduction
One of the major objectives of a mining, civil and geotechnical engineer is to assess
deformations and stresses in rock or rock mass. This understanding leads to the determination
of failure characteristics of an underground excavation and/or failure of rock slopes. There are
mainly two types of stresses which exist in a geofield—in situ stresses and induced stresses.
The pre-excavation or pre-mining stresses are termed in situ stresses and can be classified into
gravitational, tectonic and locked-in stresses based on their origin. Once an excavation is made
in rock mass, deformations occur around the excavation boundaries and stresses are
redistributed based on the geometry of the excavation and rock properties. This new stress
pattern is termed induced stress.
We are mainly interested in the computational aspects of the induced deformations and
stresses which are responsible for destabilization or collapse of an excavation. In this chapter,
we will overview the analysis of stresses and strains considering equilibrium of a rock element.
Transformation of displacements, strains and stresses will be illustrated for two-dimensional
(2D) and three-dimensional (3D) cases. Graphical representation of stresses and strains will
also be explained using Mohr’s circles. Concepts of principal stresses and determination of
their magnitudes and directions will be explained in detail. Numerous examples are solved for
better understanding of these concepts. Finally, the strain compatibility and stress equilibrium
conditions will be established for a differential element. Note that in the entire text, bold
character symbols are used for denoting vectors and matrices and italicized symbols are used
for expressing equations.

1.2 Stress, Strain and Strain Energy

1.2.1 Definition of Stress
Let us assume a body Q in which the traction force F,, point forces F; and body forces F,
are applied as shown in Figure 1.1(a). The body is constrained at boundary B. An arbitrary
1
pega Finite Element Method: Concepts and Applications in Geomechanics

(a) A solid body with loading and (b) Normal and shear load components
boundary conditions

Figure 1.1 Components of a load vector in Cartesian coordinate system and in shear plane.

plane P has bisected the body into two parts denoted by Q* and Q.. Figure 1.1(b) shows the
plane P and the resultant force F, acting on an incremental area AA of the 2" lying on the plane
P. We assume a Cartesian coordinate system to represent the force vector. Using this system,
the force vector F can be resolved into x, y and z directions as F,, F, and F, respectively and
can be represented as
F = Fi + Fj + Fk (1.1)
Again the force vector F can also be resolved along the directions of unit normal n and
unit shear s on the plane P as shown in Figure 1.1(b). The resolved forces F,, and F, are called
normal (perpendicular) and shearing (tangential) forces respectively on the area AA. The
magnitudes F/A4, F,/AA and F,/AA are called average stress, average normal stress and
average shear stress respectively acting on the area AA.
The concept of stress on a point is obtained by letting AA become infinitesimal. The
forces F, F, and F, also approach to zero but the limits of F/A4, F,/AA and F,/AA, when
AA — 0, are finite. Hence the stress vector o on a point is defined as

= AA->0
dimeAA (1.2):
Similarly, normal stress vector 6, and shear stress vector, T on a point can be defined
as

o, = lim — and 4.= lim a (1.3)

AA>0 AA AA>0 AA :

Normal stress can be either compressive or tensile. In rock mechanics applications,

compressive stress is generally considered as positive stress while the tensile stress is assumed
to be negative. An example of normal stress will be the gravitational stress in vertical and
 Chapter 1 Analysis of Stresses and Strains =m

horizontal directions beneath the surface of the earth. However, in this text, we will maintain
the engineering mechanics sign convention having positive normal stress as tensile stress and
negative normal stress as compression. This sign convention is followed to preserve the
continuity with finite element analysis. Likewise, positive shear stress is defined when it tries
to rotate an element or plane in clockwise direction and negative shear stress develops when
it tries to rotate an element or plane in counterclockwise direction. Shear strength of intact
rock, rock joints or fracture planes is of major concern for determination of overall strength
of intact rock or rock mass. For joints, the magnitude of the shear strength will depend on
the nature of contact surfaces (wavy, zaggard, plane, slickensided) and also on the type of
gauge material present.

1.2.2 Definition of Strain

When a body is subjected to external loading, it undergoes deformation or movement. If the
body is not constrained, unbalanced force creates rigid body motion based on Newton’s second
law of motion. On the contrary, if the body is constrained at some part of the boundary, it
deforms due to the applied loading and thus strains occur in the body. As for example, during
uniaxial compression loading of rock sample, the load is applied axially from one end of the
sample and the other end is constrained for any axial movement. As a result, deformation or
strain occurs in the sample. Consider a small element of axial length A/ inside the sample before
loading and let o/ be the change in length Al after the load is applied. Then the infinitesimal
axial strain in the sample is defined as

€= lim ot (1.4)
Al>0 Al

Here € is termed infinitesimal normal strain and it occurs when the body is deformed
along a straight line. The length of the body Al is considered infinitesimal so that the variation
of strain over the axial length can be ignored and the superposition principles of strains due
to various stresses can also be applied (Obert and Duval, 1967).
The shear strain occurs in a body when the angle between two straight lines changes.
Figure 1.2 shows that a right angle AOB is strained in such a way that its shape changes to

A@
AC

Ad;
x
- A
Figure 1.2 Definition of shear strain.
| 34 7] Finite Element Method: Concepts and Applications in Geomechanics

A’OB’. Thus by definition, the shear strain will be the sum of the angles A@, and AO.
Mathematically, ¥,, = AQ, + A@. In this case, %, will be considered as negative shear strain
because the right angle has decreased to A’OB’. On the contrary, if the right angle increases,
a positive shear strain occurs.

1.2.3 Stress and Strain Tensor

A tensor is a quantity with magnitude, direction and a plane under consideration (Hudson
et al., 1997). Stress and strain are considered tensor quantities since they have all three properties.
Figures 1.3(a) and 1.3(b) show stress tensors in 2D and 3D respectively. Here 0,, signifies the
normal stress acting in x direction while t,, defines the shear stress acting on the plane

(a) Stress tensors in 2D (b) Stress tensors in 3D

Figure 1.3 Stress tensors.

perpendicular to x axis and is directed towards y direction. Similar nomenclature is applied for
other tensors. In matrix form, the stress tensors are expressed as

oOXX
T=| — ; for 2D (1.5a)
Tyx yy

for 3D (1.5b)

From the force equilibrium of the area or volume element, it can easily be shown that
Tx = Ty, Ty = Tz, and T,, = T, (Obert and Duval, 1967). Thus for two-dimensional stress
analysis, O,,, Oy, and T,, are in-plane stress tensors. In this text, the tensor 0,, is considered to
be the out-of-plane stress tensor in 2D. For three-dimensional problems, six stress tensors—o,,,
Oy, Oz:, Ty, Tz, and T,,—are used to define a state of stress condition. From the above
discussion, it can be noted that the tensor matrices given in Eqs. (1.5a) and (1.5b) are
symmetric and can be rewritten as
| Chapter 1 Analysis of Stresses and Strains LedSinnd

oO 6.
r-|me »| for 2D (1.6a)
Ty Oy,

On TW Ty
T=| Ty, py tothe tor 3D (1.6b)
Tx anty, |) Og

Strain tensors are obtained from the gradient of the displacements. Consider an original
(undeformed) body in two dimensions in Cartesian coordinate system xy as shown in
Figure 1.4. When the body is deformed denoted by Q*, every particle in the body takes up
a new position, which is described with new coordinate system x*y*. Let us consider the

yy
* + dx’y" i. dy’)

Q (x\+ dx,y + dy)

Figure 1.4 Displacement vector.

particle P which is located in the original body with coordinates x and y. After deformation,
this particle is moved to location P* in the deformed body with coordinates x* and y*. The
vector denoted by PP* represents the displacement vector of the particle. Mathematically, the
component of displacement vector can be written as
fae ae Ow (1.7a)
and
yr = Vo (1.7b)

where u and v are the displacement fields in x and y directions respectively and are functions
of both the coordinates x and y. Note that displacement fields are also functions of coordinates
<-and y™:
Now consider an infinitesimal line element PQ in the original body Q where the
coordinates of the particle Q are defined by x + dx and y + dy. The square of the length ds
of the line element PQ in its original configuration is given by

(ds)” = (dx)? + (dy) (1.8)

io | Finite Element Method: Concepts and Applications in Geomechanics

After the body is deformed, the square of the length ds* of the deformed line element
P*Q* will be
(ds*)? = (dx*)? + (dy*)? (1.9)
From Eq. (1.7), the differentials dx* and dy* are computed based on the differentials dx
and dy as

* _ ox. ox” Ou ou
dx ae—dx+ —ap dy = 1+—
+a)|ldx+—d
ee y ;
(1.10a)

and

alfyoys oy” Ov ov
dy == an dx + —d
ay ly =x v41+—
4 ldly >
(1.10b)

From Eqs. (1.8), (1.9) and (1.10), the change in square length is found as

(ds’)” — (ds)? = 2e,,(dx)” + 2eyy(dy)” +2€y, dudy (1.11)

3-2-8 8)
where the strain tensors e,,, ae and exy are

Cx eg Ox ox ax 21 Lax Ox (1.12a)

2 a1 (aeY(a) _]_ ay (au) |(avy

YW 21 dy oy Oy 2| Lay dy (1.12b)

3 -($+2 }+[BS.22
lax oy ox dy ox doy =

Then Cauchy’s infinitesimal strain tensors in two dimensions are expressed by ignoring
the quadratic terms in Eqs. (1.12a), (1.12b) and (1.12c) as

etou esov dv ou
SO (1.13)
a Rema? oc
For three-dimensional displacement fields with u = u(x, y, z), v = v(x, y, z) and
w = w(x, y, z) in x, y and z directions respectively, the additional strain tensors apart from that
given in Eq. (1.13) will be
_ ow _ ow ov du dw
aR ori Wisse ae.
Vo a SS PSS 5) eae —_—_— (1.14)

Just as stress tensors in this book—é&,,, & and Yoy—are considered in-plane strain tensors
for two-dimensional problems. The out-of-plane strain tensor is €... For three-dimensional
| Chapter 1 Analysis of Stresses and Strains |

problems, &, Ey, Ex Mey» Ae and y., represent six strain tensors. The strain tensor matrix is
also symmetric and of similar form as given in Eq. (1.6) with the exception that stress notations
will be replaced by strain notations.

Example 1.1 Show that for a line element of length dx parallel to x axis, the infinitesimal

strain will be E,, =

dx —dx
dk Nid
Solution: Since the line element is parallel to x axis, dy = dz = 0. Hence Eq. (1.11)
becomes

(dx")? — (dx)? =2e,, (dx)’

or

iy : ey eae
2e,. eal
(dx)?
dx +dx

In this case, dx* differs from dx by a small quantity of the second order if we assume
the displacement fields u, v and w and the strain components to be infinitesimal. Hence in the
right hand side of the equation, we can assume dx* = dx and thus

_ dxdx
fe dx

1.2.4 Strain Energy

Strain energy is developed due to elastic distortion of a body. During loading, strain energy
is contained in the body and during unloading, it dissipates. Hence for loading and unloading
cycle, the net strain energy is zero. Strain energy density is defined as strain energy per unit
volume of the body and is expressed as

1 (1.15a)
Uo Ge oO Spe PCy OpyateeneO
Ki KK
et Vay lyy + Vyeyz t Vex lee)

Considering the volume of the body as V, the total strain energy will be given by

U = [Upav (1.15b)
V

1.3 Transformation of Displacements, Strains and

Stresses
In this section, we will overview the techniques for transforming displacement vector,
strain and stress tensors from one coordinate system to another. Firstly, we will elaborate
 8 Finite Element Method: Concepts and Applications in Geomechanics

coordinate transformation in two dimensions and then extension will be given in three
dimensions.

1.3.1. Transformation of Displacement Vector

Let us assume that x’y’ represents a coordinate system where angle between x and x’ axes is 0
degrees as shown in Figure 1.5. Let us also assume that the displacement fields based on
xy’ coordinate system are represented as wu’ and v’ respectively. Now, we would like to represent
u’ and v’ in terms of u and v.

CBO) S]am

Figure 1.5 Transformation of displacement in 2D.

A displacement vector A can be represented by both the coordinate system as shown in

Figure 1.5. Hence we can find displacements u’ and v’ in terms of u and v as follows:

u’ = A-i’ =(ui+ vj)-i’ =ucos@ + vsin@ (1.16a)

and
VW =A-j =(ui+vj)-j/ =—usin@ + vcosé (1.16b)

Thus the transformation matrix R, for transforming displacement from xy coordinate

system to x’y’ coordinate system, can be written as

Caeele (1.17)

1.3.2 Transformation of Strain Tensors

Transformation of strain tensors from one coordinate system to another is not that
straightforward as shown for displacement vector. For strain tensors, we would like to represent
strain vector €7 = (ey Eyy Vey €z) in terms of strain vector ea (ce. Evia Vaan) aoNote, that
for two-dimensional problems the z axis does not change during transformation. Considering
Evy as the gradient of u’ with respect to x’, and using chain rule, we find

du’ du’ ox du’ oy

coc laa = a (1.18)
Late s bettapo Oe owe
| Chapter 1 Analysis of Stresses and Strains [ 9 |

The direction cosines are expressed as vo and oy =s. From Eq. (1.16), we can
. ox’ Ox’
write

ox Oxted 0x Say
and

ou =u, ay
dy ay ay (1.19b)

Combining Eqs. (1.18) and (1.19), we find

Eq'y! = C Egy +S Ey, + CSVxy (1.20)

Similarly, other two strain tensors €,, and ¥yy expressed in x’y’ coordinate system can be
obtained using the strain vector €7 = (€,, Ey Vey €,)- After combining all the results, we
find the strain transformation matrix as

co s° cs 0
2 2
Re § G —cS 0

—J65 =2cs\ Ga sam) (1.21)

0 0 0 1

e’=Re

The last row and the last column are added to consider the strain transformation in z
direction. It is clear that strain in z direction does not change due to transformation of
coordinate system.

1.3.3. Transformation of Stress Tensors

Transformation of stress tensors from xy coordinate system to x’y’ coordinate system
is performed based on the transformation matrix given in Eq. (1.21). We know that
for a given loading condition, the strain energy of a body will be the same irrespective of
the coordinate system based on which calculations are made. Considering a virtual strain
vector d€, we can equate internal strain energies estimated using the coordinate systems xy
and x’y’ as
de'o = de’!0’ = 5e'R'0’ (1.22)
Since for any virtual strain vector 6€ # 0, we find that

o’=(R')'o=To (i283)
where transformation matrix T can be obtained from matrix R as
 Finite Element Method: Concepts and Applications in Geomechanics

C S PFE 0
T= fo OC —2cs 0
a<tucvosmct stot (1.24)
0 0 0 I

Using the matrix given in Eq. (1.24), normal and shear stresses in x’y’ coordinate system
can be explicitly written if the x’ axis is rotated @ degree counterclockwise with respect to x
axis (Figure 1.6) as given below

On + Oy, Ox —Oyy
C= ae + —e cos26 + T,, sin20 (1.25a)

OZ iC>, Ox TO yy
Oyy = rear Lane cos2@—T,, sin2 (1.25b)

Sporn 6:
Exiy = — aca sin
20 +T,, cos20 (1.25c)

Figure 1.6 New stress tensors after rotating the element @ degree in counterclockwise direction.

Example 1.2 The vertical and horizontal stresses at the middle of a coal pillar are found to
be 15 MPa and 3 MPa respectively. The shear stress is 2.5 MPa as shown in the figure below.
Determine the stresses in xy’ coordinate system as shown in the figure below.
Veet, ES:
} 2S v
3.0 => = 3.0

30°
: OS
- a
Solution: Here 0 = —30°. Hence c = 0.866 and s = —0.5. Using Eqs. (1.23) and (1.24),
we find

Ong 0.75 0.25 —0.866 || —3 —8.165

Oyy ¢=| 0.25 0.75 0.866 |,—-15;=4-9.835+ MPa
Tel 0.433 —0.433 0.5 25 6.446
| Chapter 1 Analysis of Stresses and Strains |[ 11 |

1.3.4 Transformations in Three Dimensions

Transformation or rotation matrices in 3D are straightforward extension of 2D. Illustration of
3D rotation matrices are given here based on the reference (Cook et al., 1989). Let us consider
a displacement vector A in 3D space represented by both xyz and x’y’z’ coordinate systems as
shown in Figure 1.7. Direction cosines are defined below:

I, = cos(x’, x), cosine of angle between the axes x’ and x

m, = cos(x’, y), cosine of angle between the axes x’ and y
n, = cos(x’, z), cosine of angle between the axes x’ and z

Direction cosines

Figure 1.7 Transformation of displacement vector in 3D.

Similarly, direction cosines of y’ and z’ axes with x, y and z axes can also be defined as
shown in Figure 1.7. However, all 9 direction cosines are not independent of each other. It
can be easily shown that there are relationships between the direction cosines as given in
Eqs. (1.26a) and (1.26b).

I? +m? +n? =1, Bm, +m 1, B+ m? +n? =1 (1.26a)

LL +mymy, + NN, = 0, Lk +m7mM3 ar Nyj7Nn3 = OF Ll +m3mM, +N3n = 0 (1.26b)

Using similar approach as shown in Eq. (1.17), a three-dimensional displacement vector

expressed in xyz coordinate system can be transformed into a displacement vector in xyz
UPA

coordinate system by transformation matrix R as given below:

u’ L, my n, |{u
vr=|b m ny |, Vv
w’ L, m3 ny ||w (1.27)
R

Strain tensors are transformed from xyz coordinate system to x’y’z’ system based on the
concepts outlined in sub-section 1.3.2. In three-dimensional cases, 6 strain tensors are to be
 12 Finite Element Method: Concepts and Applications in Geomechanics

transformed instead of 3 tensors for 2D cases. In 3D cases, the transformation matrix R is in

the form of

R-|R 11 R a (1.28)

Rx, Ry»
where

yom om
2 2 2}
Ry =| m nm
p2 2 2
L m3 n3

Lm, mn nh
Ry =|bm mn nl
hm, M3N3 N3l,

2h1, 2mm, 2nn

R>; = 2hl, 2mm, 2n7N;

2131, 2m3m, 2n3n

lm athm m Ny air Mj7N ny L + Nl;

R5> = Lm + l,m m7N3 ar M3Ny Nl, a6 Nl,

km, SF 1m m3n ae mn nl; ae nl

The components of strain vector are in order e! = (€,,, eres rtunay(ited Aye mi)
must be followed while transforming from xyz to x’y’z’ coordinate system.
The procedure for transforming stress tensors is also straightforward extension of two-
dimensional transformation matrix T. As before, the matrix T is expressed as

=I
T=(R™) (1.29)
where the matrix R is given in Eq. (1.28) and thus the matrix T can be obtained as

Ry, 2Rj2
{t=
1.30
(1/2)R5, R5> ( )
 Ca
1.4 Stress Analysis

1.4.1. Principal Stresses in Two Dimensions

The principal stresses are normal stresses that act in planes where shear stress does not exist.
In two-dimensional problems, principal stresses also signify the maximum and minimum
normal stresses that can exist at a particular plane and generally termed as major and minor
principal stresses. Note that the magnitude of principal stresses is independent of the coordinate
system. The direction of principle stresses, however, will depend on the reference axis.
Consider that a principal stress 0; acts on the oblique plane AB which has the normal
vector ON as shown in Figure 1.8. The direction cosines of the normal vector are represented
by / = cos@ and m = sin@. The components of o; in Cartesian x and y directions are o;, and
Ojy, respectively. Hence we can write
0;,=l0; and Oj =mo; (1.31)
Applying force equilibrium relationships in x and y directions, it can be easily shown that
(Figure 1.8)
105, + mt, —O,, =10; (1.32a)
and
[Ty +MOyy, = Oy = mo; (1.32b)
The above relationship can be written in matrix form as

(1.33)

Oxx
Figure 1.8 Estimation of principal stress.

If this system of equations has a non-zero solution, then the determinant of the left hand
side matrix must be zero. Thus we arrive at the following equation.
G Nic —N,=0 (1.34)
where
N, =0,, + Oyy (1.35a)
 Finite Element Method: Concepts and Applications in Geomechanics

and
Ny =O. Oyy + Try (1.35b)
Here N, and Nj are called invariants of stress tensor which means that these values are
not dependent on the coordinate system. By solving Eq. (1.34), the magnitude of major (0;)
and minor (03) principal stresses can be obtained considering 0; < 63 as

(1.36)

In the above equation, we assume that normal stresses are compressive. Once the
magnitude of principal stresses are known, Eq. (1.33) can be used to find the direction cosines
using the additional relationship that states 7 + m? = 1.
In solid and rock mechanics, principal stresses have a special significance since their
magnitudes do not alter with the coordinate systems. Thus it is convenient to represent a state
of stress condition using principal stresses. Let us assume that the principal axes are defined
by numerals | and 3 representing major and minor principal stress directions respectively. We
are now interested to find the normal and shear stresses as a function of principal stresses.
Consider Figure 1.9 where an oblique plane AB makes an angle @ with the axis 1. The line
ON represents the normal vector of the plane AB. The direction cosines of the normal vector
are represented by / = cos@ and m = sin@ with respect to the axes 1 and 3 respectively. Consider
that the normal stress 6, acts on the oblique plane parallel to the line ON and shear stress T
acts tangential to the plane AB. The resultant stress of 0, and T is Op. If we resolve Op into
the directions 1 and 3, we find Op, and Op3 respectively. Hence from Figure 1.9, we can write
OR] =lo, and Opn3 = M03 (1.37)

The magnitude of o, can be obtained as

On =I0p +MOR3 = Fo, +m’o; (1.38)

The magnitude of shear stress T is then computed as

1? =0R —02 =o? +m’o3 - (Po, + m*o3)" = I’m? (6, - 03)" (1.39)
Therefore
(1.40)

Figure 1.9 Estimation of normal and shear stress from principal stresses.
| Chapter 1 Analysis of Stresses and Strains |[ 15 |

1.4.2 Principal Stresses in Three Dimensions

Principal stresses in three dimensions can also be found in the similar fashion as given in two-
dimensional cases. As before, assume that a principal stress 0; acts on the oblique plane ABC
whose normal vector is designated by ON as shown in Figure 1.10. Consider that J, m and n
represent direction cosines of the normal vector with respect to x, y and z axes respectively.
It is obvious from Figure 1.10 that the components of o; in x, y and z directions are

Oj, =10;, Fiy =MO;, Gj, =NO; (1.41)

24

Figure 1.10 Concept of principal stress in 3D.

Equating force equilibrium conditions in directions of three coordinate axes, we find

O30; Ty Tey,
fy eel OF Ty, |=0 (1.42)

Tx T yz O02, —0;

The above determinant gives the following cubical equation:

6; —N,o7 — No; — Nz =0 (1.43)

where N; (i = 1, 3) are the invariants of stress tensors and are expressed as

Nj =O, +Oy +O, (1.44a)

Ce Neat Mia (1.44b)
Ny =-(0,,0yy +O yO, +0704) + Ty + Ty, + Tz
2 2 p (1.44c)
OyyOz, + 27 yyTyb, — OxxTyz— OyyTx — Ox Try
N3 = Ox,
 Finite Element Method: Concepts and Applications in Geomechanics

The solution of Eq. (1.43) will give three principal stresses — major principal stress 0},
intermediate principal stress 0, and minor principal stress 03. This is a typical eigenvalues
(principal stresses) and eigenvectors (direction cosines) problem. Once the magnitudes of the
principal stresses are known, Eq. (1.42) can be used to calculate the direction cosines. In
general, the magnitudes of principal stresses are such that 0; 2 0) 2 03. However, in rock
engineering problems, we mainly deal with compressive or negative stress and thus it is
convenient to assume 6, < GO» < 63. This convention will be followed in the subsequent
chapters. The directions of these principal stresses can also be obtained in the similar fashion
as described for 2D case. The best way to obtain the solution of Eq. (1.43) is by trial and error
method such as Newton-Raphson method using a computer program or with a programmable
calculator.
Once the magnitudes of three principal stresses are known, the invariants of stresses can
be obtained based on principal stresses,as follows:

N, =0, +0) +03 (1.45a)

Ny =—(0102 +0703 + 6304) (1.45b)

N3 = 010703 (1.45c)

As before, let us denote the principal axes by numerals 1, 2 and 3 to represent the
directions of major, intermediate and minor principal stresses respectively as shown in
Figure 1.11. Again let us consider the oblique plane ABC which has the normal vector ON
having direction cosines J, m and n. The normal stress 6,, acts parallel to the normal vector ON
3

Figure 1.11 Normal and shear stress in 3D.

and shear stress T acts tangential to the plane ABC. As shown for two-dimensional cases, let
us assume that Og denotes the resultant stress of o, and t The resolved stresses of Op in
directions 1, 2 and 3 are Opy, Op2 and Op3 respectively. Hence we can write the following
relationships:
OR] =lo, > Ona —MmO>, Ops — Os (1.46)
| Chapter 1 Analysis of Stresses and Strains |

From Figure 1.11, it can be shown that the magnitude of normal stress will be

0, =10, +moy +no3 =1°o, +m’, +n’o3 (1.47)

The magnitude of shear stress can be obtained from the following expression:

T” 2 =0R2 — 0,2 =(l0,)” + (mon) + (no3)” — (?o, + m*o, +1703)

2 2,2 (1.48)
=! m (0; =0,)° +m>n* (07 = 0s): +n (03 Or ie

Example 1.3 Find the magnitudes and directions of principal stresses for Example 1.2.

Solution: Using Eq. (1.36), we find o, = -15.5 MPa and o; = —2.5 MPa. Now for the
direction of 0; with x axis, we find

m=—1=5I and (1+57)i* =1, 1=0.196116

Hence the angle will be 78.69° counterclockwise direction from x axis.

Example 1.4 The six stress components are given below. Estimate the values of principal
stresses and their directions.
O,, = —20.5 MPa, O,y = —34.8 MPa, O,, = —8.1 MPa
Ty = 2.4 MPa, T,, = 8.0 MPa, T, = —5.2 MPa

Solution: The stress invariants are:

INF pSALI =2 Oh Otte Oe)

Ne—= (COC j0n 0,0, 10,07.)

Le ee
tz, + t,, +t, =—1004.53

Oe07, la ly 2 20 yl2 07,15) —

yz Orxx Og xx ye 3018.91
2 =__
N3 = 0,0), “yz Z

Hence the cubic equation of principal stress becomes

6; — 63.407 +1064.530; + 3678.57 =0

Solution of this equation can be obtained using trial and error (or Newton—Raphson)
method and the results are given below:
Major: 0, = -37.8315 MPa
Intermediate: 0, = —20.9206 MPa
Minor: 0; = —4.6478 MPa

Let us assume that the direction cosines of the major principal stress 0, with respect
to x, y and z axes are /, m and n respectively. Then the system of equations as given in
Eq. (1.42) becomes
 18 | Finite Element Method: Concepts and Applications in Geomechanics

17.3321 + 2.4m — 5.2n = 0

2.41 — 3.032m + 8n = 0
—5.21 + 8m + 29.732n = 0
From these equations, the following relationships can be obtained:
l= —4.3225m
v= 123382
Using the equality condition ? + m* + n* = 1, the direction cosines are calculated as
f=, 0.2157;7°), wm, =2-0.9327 (138°) i= 0.2351 aG
Similarly, the direction cosines for intermediate and minor principal stresses are:
Intermediate: 1, = 0.9377, my = 0.2805, n, = 0.2053
Minor: emma bRil ws}. mz = —0.2264, nz = —0.9351

1.4.3 Maximum Shear Stress

The magnitude and direction of the maximum shear stress are important for the analysis of rock
failure mechanism. In 2D, the magnitude of the maximum shear stress can be obtained by
maximizing Eq. (1.40) with respect to the angle @. For this case, the maximum shear stress
1
occurs when / = m = —= and hence
V2
Ope G3

For 3D cases, the maximum shear stress at xy plane is found by letting n = 0 and then
maximizing Eq. (1.48) as

of OC
(imax)sy= (1.50)
Similarly, for yz and zx planes, considering / = 0 and m = 0 respectively, the maximum
shear stresses can be computed from

07 = 03 = 03 = O07}
(Tmax )yz = and Pe: = )
Cied4)

Note that the direction of the maximum shear stress is 45° from any principal stress
direction in a given plane.

1.4.4 Stresses on Octahedral Plane

In solid mechanics, normal and shear stresses acting on an octahedral plane has special meaning
as far as plastic yielding and creep phenomena are concerned. An octahedral plane is defined
when the normal vector to that plane makes equal angle with the 1, 2 and 3 axes. Consider
that /, m and n represent the direction cosines of the normal vector of an octahedral plane with
 | Chapter 1 Analysis of Stresses and Strains |

respect to coordinate axes 1, 2 and 3 respectively. By definition, direction cosines of the normal
vector of an octahedral plane must be equal, i.e. / = m = n. Hence we can estimate the equal
angle as

3/2 =] (1.52a)

1
DUE liea or cos 54.7° (cos 125.3°) (1.52b)

Since there are eight quadrants, there will be eight possible planes whose normals make
‘ equal direction cosines with the coordinate axes. The stress normal to an octahedral plane is
termed the octahedral normal stress. The magnitude of octahedral normal stress can be obtained
by substituting direction cosine values in Eq. (1.47):

Cal Op em Or 4n-O, (1.53)

Again, 1=m=n= ey and thus

V3
1 N,
ey ong(ae ed)hi a ate (1.54)

Since the octahedral normal stress is the average of principal stresses, the octahedral
normal stress is also called the mean normal stress. The shear stress on octahedral plane is of
more interest in terms of plastic yielding, creep phenomena and criterion of rock failure.
Although the maximum shear stress is slightly higher than that of octahedral shear stress, later
has more significance since it occurs on eight planes through a ‘point’ (Herget, 1988). The
octahedral shear stress T,,, can be obtained from the following relationship:

OF = ny + Toot
or

2 2 2
Toct =OR — Foct

= (Io? + m’o3 +n’03)- (0, +m*o> +103)

1
=3(7 ne +03)-5(01 +07 +03)

=—(N,
rer) +2N,)-——N.
Mets
ra 1 2) 9!

2 2
==N? +=N,
9 3
Therefore

T oct —
V2(y2 +3N>)°> (1.55)
3
 20 Finite Element Method: Concepts and Applications in Geomechanics

1.5 Mohr’s Circle of Stress in Two Dimensions

Mohr’s circle of stress was first presented by Otto Mohr in 1882. This method applies graphical
technique to represent two-dimensional stress fields in order to evaluate the orientation and
magnitude of normal stress, shear stress and principal stresses.

1.5.1 Steps for Constructing Mohr’s Circle

Step 1. Normal stress (6,) is plotted on the horizontal axis and the vertical axis refers
to the shear stress (t). The scale of the vertical axis is the same as that of the
normal stress axis.
Step 2. Shear stress is plotted with negative sign if it tends to rotate the rock element
in counterclockwise direction as shown in Figure 1.6. In this figure, the shear
stress acting on the horizontal plane will have negative sign. On the contrary,
if a shear stress rotates the element in clockwise direction, it will be designated
with a positive sign while constructing the Mohr’s circle. In the same figure,
shear stress acting on the vertical plane will be designated with positive sign.
Step 3. Plot points A(O,x, Ty) and A’ (0,,, —T,,.) on a graph paper with proper scale and
sign convention as mentioned above. In this case, the point A refers to the
stresses acting on the horizontal plane and the point A’ signifies the same on the
vertical plane (Figure 1.12).
Step 4. Join the line AA’ and the intersection point between that line and the horizontal
axis is the centre of the Mohr’s circle. It can also be shown that the coordinate
of this centre point will be
ORO y
OC = (1.56)
wa
Step 5. Draw the circle with the centre at point C and diameter as AA’. This completes
the construction of the Mohr’s circle with any state of stress condition in two
dimensions.
Step 6. If the principal stresses are used to draw the Mohr’s circle, then the points A
and A’ will lie on the horizontal axis. Repeat steps 4 and 5 to draw the Mohr’s
circle.

1.5.2 Inference from the Mohr’s Circle of Stress

1. Each point of the circumference of Mohr’s circle represents a state of stress condition
with normal and shear stresses.
2. The Mohr’s circle intersects the horizontal axis or zero shear stress plane in two
points, P and P’. The magnitude of stresses represented by OP and OP’ equals to the
major and minor principal stresses, respectively, considering compressive as negative
stress.
3. The direction of the principal stresses can also be estimated by measuring the angle
ACP and dividing it by 2 as shown in Figure 1.12.
| Chapter 1 Analysis of Stresses and Strains 2-4

_ ZACP
o, (1257,
2
4. The magnitude of the maximum shear stress, T,,, is the radius of this circle and can
be given by CB. It can also be shown that the magnitude of Tyax is

OTUs
Tmax (1.58)
2
5. The normal of the maximum shear plane makes 45° angle with the normal of the
principal plane. The normal stress on the maximum shear plane is non-zero except for
“pure shear” stress condition and the value is given by

0, +o 0,+0
Ormax =—> rat sie
: (1.59)
6. The normal and shear stresses at any given plane measuring an angle 9 with the
current state of stress condition as given by AA’ line can be obtained as follows:
(a) Draw a line passing through the point C and making an angle of 20 degrees
clockwise with the line CA as shown in Figure 1.12. If @ is negative, rotate 20
degrees counterclockwise direction. Let us assume that this line intersects the
- circle at D and E points. Let us also assume that the projections of the points D
and E on the horizontal axis are D’ and E’ respectively.
(b) The magnitudes of OD’ and OE’ represent the values of normal stress of Og and
O9499. The magnitude of shear stress, T, is equal to the line DD’ or EE’.

09+ 90

E plane

Figure 1.12 Mohr’s circle of stress.

 22 Finite Element Method: Concepts and Applications in Geomechanics

1.5.3. Mohr’s Circle of Special Stress Conditions

Uniaxial compression. In uniaxial compression test, core sample is loaded axially and the
failure stress 0, is recorded. No external lateral stress 03 and shear stress act on the sample.
Figure 1.13(a) shows the Mohr’s circle of stress for uniaxial compression loading condition.
The maximum shear stress developed in the sample will be Tax = 0;/2 and acts 45° from the
horizontal plane of the specimen.
Uniaxial tension. In direct uniaxial tension test, the rock sample is pulled with an axial load.
The failure stress 6, due to tensile loading is recorded. The Mohr’s circle is drawn with
0; = 0 and o, = F, where F is the failure stress as shown in Figure 1.13(b).

Triaxial compression. In this test, the rock specimen is loaded both in axial and lateral
directions. The failure stress oO; is recorded for a given lateral stress 6; generally applied by
placing the sample in an oil chamber. The Mohr’s circle is drawn with centre as —(0; + 03)/2
and the radius as (0, + 03)/2 [Figure 1.13(c)].

Pure shear. Pure shear condition occurs when the rock sample is twisted with equal moment
from the two ends. A small rectangular element on the sample will experience compressive
stress in one plane and equal tensile stress from the orthogonal direction of the compressive
plane. Hence 03 = —o,. In this case, normal stress vanishes in the maximum shear plane.
Figure 1.13(d) shows the Mohr’s circle of stress of pure shear condition. Here Tpax = OF.
Hydrostatic stress condition. Uydrostatic stress field means that the equal compression
loading acts on the rock sample from all directions. Hence 03 = 0; and thus Mohr’s circle
vanishes to a point as shown in Figure 1.13(e).
“Al T O; -

on
(a) Uniaxial compression

(e) Hydrostatic static condition

Figure 1.13 Mohr’s circle of special stress conditions.
 Chapter 1 Analysis of Stresses and Strains am | 23 =|

Example 1.5 Determine the principal stresses and their directions using Mohr’s circle of
Example 1.2.

Solution: Solution with Mohr’s circle is shown below. The magnitu¢e and direction of
principal stresses are also shown. The dotted line indicates the state of stress after rotation
of —30°.
8
6
4
(2155) 2
6, (MPa) - é
(215.25) :

Example 1.6 The tangential stress acting on a small rock element of —_500 m3 500
the roof of a circular opening is found to be 500 kPa as shown in the
adjoining figure. Estimate the principal stresses and the maximum
shear stress with their directions.

Solution: In this example, 0; = 0 and o, = 500 kPa. The maximum shear stress is
250 kPa and directed 45 degrees from the principal axis. Hence the Mohr’s circle of stress will
be as shown below:
T (kPa)
300

200

100

0, (kPa)

1.6 Analysis of Strain

The procedure of strain analysis is similar to that of stress analysis. In sub-section 1.3.2,
transformation of strain tensors is explained. In this section, we will briefly illustrate the
principal strains and graphical representation of strains using Mohr’s circle.
 24 Finite Element Method: Concepts and Applications in Geomechanics |

1.6.1. Principal Strains in Two Dimensions

The principal strains are normal strains that act on the planes where shear strains are absent.
In this respect, the computational procedure of principal strains is the same as that of principal
stresses. The procedures described in sub-section 1.4.] are directly applicable for the estimation
of the principal strains. Thus principal strains can be computed by solving the following
equation:

eo Miee MeO (1.60)

where M, and Mj are invariants of strains and are computed based on

M, =€,, + &y, (1.61a)

and
y
My =-€,,Eyy Sgr (1.61b)

E+E l
E13 ar omar etc = Ey) +75 (1.62)

Comparison between Eqs. (1.36) and (1.62) shows that for computation of principal
strains, normal stresses are replaced by normal strains but shear stress is replaced by half of
the shear strain. It can also be shown easily that the direction of principal stresses and respective
principal strains are the same.

1.6.2 Principal Strains in Three Dimensions

The procedures outlined in sub-section 1.4.2 are directly applicable for the estimation of
principal strains in 3D. As before, for equations mentioned in that section, normal stresses can
be replaced by normal strains but all shear stresses must be replaced by half of the shear strains.
For example, the cubical equation for estimating principal strains will be

€? — Me; — M>€; - M3 =0 (1.63)

where invariants of strains are:

Mi= 2 &,, +e, (1.64a)

7 QD D
My = (Gye) +2, emtbe ec, ) 4 y ji
ee .
(1.64b)
- 4 4 4
and

Be EV 2 NeeEee 2 me 2
M3 = Ex Ey Ex y Tota Saye Ewa Fx Vay (1.64c)
8 4 4 4
| Chapter 1 Analysis of Stresses and Strains
__} L-25_]
1.6.3 Maximum Shear Strain
The magnitude of one-half of the maximum shear strain in xy, yz and zx planes can be obtained
in terms of principal strains as:

(Ze) fe E; — Eg
4 - 2 (1.65a)

(1.65b)

[tas = E3 — &|
7 |eee: (1.65c)

Again note that the direction of the maximum shear strain will be 45° from the principal
strain axes.

1.6.4 Mohr’s Circle of Strain in Two Dimensions

The Mohr’s circle can also be constructed for graphical representation of two dimensional strain
tensors €,, &, and ¥,,. The construction procedure of the Mohr’s circle for strains is similar to
that of stresses with the following exceptions:
1. The horizontal axis is represented by the normal strain and the vertical axis is denoted
by half of the shear strain. The scale of the vertical axis will be the same as of the
horizontal or normal strain axis.
2. The location of shear strain in the plot, i.e. above or below the normal strain axis,
depends on the sign of the shear strain. As mentioned earlier, if the right angle
decreases, the sign of shear strain is negative, then half of the shear strain is plotted
below the horizontal axis. On the contrary, if the right angle increases, the sign of
the shear strain is positive. Then the same is plotted above the horizontal strain axis
(Figure 1.14).
yl2
Original Deformed B{(& + €3)/2, Ynax/2}

ee
wee
eK

Figure 1.14 Mohr’s circle of strain.

 Finite Element Method: Concepts and Applications in Geomechanics

Consider the schematic drawing shown in Figure 1.14. Assume that the normal strains €,,
and €,, are both compressive strains or negative quantities and act in A and A’ planes
respectively. The shear strain y,, is positive in A plane but negative in A’ plane. Thus, Mohr’s
circle of strains can be plotted with points A and A’ as shown in the figure. Note that one half
of the magnitude of shear strains has to be considered while plotting the Mohr’s circle.

Example 1.7 The rectangle shown in the figure below is deformed into the shape indicated
by dashed lines. The displacements are found as u = Mxy and v = Nxy.

1.0m

(a) Determine the state of strain at point C when coordinates of point C* for the
deformed body are (0.502, 0.996) m.
(b) Determine the normal strain at C in the direction of line CA.
(c) Determine the shearing strain at C for the undeformed line CA and a line
perpendicular to line CA.

Solution: (a) The magnitudes of M and N are obtained from the known displacement
at point C. We have
M x0.5=0.002

Hence, M = 0.004 and u=0.004xy

N x0.5 = —0.004

Hence, N = -0.008 and v=-0.008xy

Thus strain components are

P
Ex|-=—| =0.004y=0.004 «,, =) __6.008x =-0.004
Ox|o ae ONO Ye

eno
Yy|.=|—+= | =0.004x —0.008y = -0.006
OC dyerocs»
 a Chapter 1 Analysis of Stresses and Strains

(b) Let x’ denotes the line from C to A. Hence the direction cosines of the line CA will
be
c = -0.4472, s = -0.8944
Using Eq. (1.21), the magnitude of normal strain in x’ direction is computed as

Eyyr = Ce, +5°Ey +C5Yxy = —0.0048

(c) The shearing strain Yx’y’ will be

Vx = Oconee 2CSEyy + (Cc? = s*) Yxy = —0.0028

1.6.5 Volumetric Strain

Volumetric strain refers to the change of volume to the original volume of the body due to
- orthogonal straining. For example, let us consider a parallelogram with sides Al,, Al, and Al,
as shown in Figure 1.15. The original volume of this parallelogram is given as

AV =Al,Al, Al, (1.66)

Figure 1.15 Definition of volumetric strain.

Let us also assume that due to normal or orthogonal strains in three coordinate directions,
the volume changes to AV’. Then the volumetric strain will be

eS AV =AVZ AV’ 1.67

“ AV AV og)
; él
Now, AV’ = (Al, — ol,)(Al, - Ob) ol.) and €,, Sia and so on. Thus Eq. (1.67)
=
can be rewritten as

=
(Ale = 81, (Aly = 5ly(Ale = Sle) _ iis d= ele) (1.68)
Ee Al, Al, Al, 2 aad
 Finite Element Method: Concepts and Applications in Geomechanics

By multiplying all strain components and neglecting the product of strains, we find that
Ea Grp ale aera ey (1.69)

Thus the volumetric strain of a body is equal to the sum of the normal or orthogonal
strains.

1.7 Stress Equilibrium Equations

Equilibrium of an element either in 2D or in 3D requires that at least the resultant forces in
two or three perpendicular directions vanish respectively. Let us consider a rectangular element
in two dimensions comprising of edges dx and dy. The stresses acting on the surfaces are shown
in Figure 1.16. For example, the force o,,dy acts on the left side of the rectangle while the
00 ' ae ; 19k ed!
force [owa ee ax Jayacts on the right side. This is required to maintain the continuity of
Pp
the stresses. In other words, the continuity of the stresses satisfies the increment of stresses from
one surface to the opposite surface. The body forces are B,dxdy and B,dxdy in x and y directions
respectively. The equilibrium of the body requires that the resultant force acting in x and y
directions are zero. Hence we can write that

Dy =F =0 (1.70)

dy

Figure 1.16 Equilibrating forces in 2D.

| Chapter 1 Analysis of Stresses and Strains

Thus summation of all the forces in x and y directions are:

ue 00 OT
Suki = ody -(O + ap ts) tyde-[ty ala —B,dxdy=0 (1.71a)
0

SF,Fae doy,
=9,,dx-| 0, + Fy
OT,
JAH tay —|Bay+5 de |dy—Bydady=O (1.71b)

Dividing each equation with dxdy, we can obtain two equilibrium conditions as follows:

alee, sin le ads

qeiules +B, =0 (1.72a)
and

OT xy 00 yy
ee eet. +B, =0 (1.72b)

The equilibrium conditions expressed in Eqs. (1.72a) and (1.72b) mean that the stress
functions 6,,, O,, and T,, cannot be selected independently. Rather they are generally defined
based on the problem geometry and boundary conditions.
The stress equilibrium conditions are also obtained for volume element in 3D using stress
tensors O,,, Oy, O, Ty, T, and 1, as

d0,, 9 OT
By ippeit
gow oer ssa
OT 00 OT
Bertin Perea By 20 (1.73b)

—=+B,=0 (1.73c)

where B,dxdydz is the body force in z direction. Note that for six stress functions, there are
three equilibrium equations. Hence stress functions (tensors) are not independent of each other
if force equilibrium conditions are to satisfy.

Example 1.8 A steel rope of length L is hung from the ceiling. If the density of the rope
is p, find the longitudinal stress in the rope. Neglect shear stresses.

Solution: Let us assume that the vertical direction is denoted by y axis from the ceiling.
We assume that only o,, acts on the rope and the shear stresses 7,, and 7,, are zero. Hence
applying stress equilibrium equation, we find that
00 yy + pg=0
dy
 30 Finite Element Method: Concepts and Applications in Geomechanics

Hence, by integrating, we get

Oyy = —P8&y =u (6

Now the stress at y = L will be zero and hence c = pgL. Therefore, the expression of
longitudinal stress will be
Oyy = pg(L—y)

1.8 Strain Compatibility Conditions

As mentioned earlier, displacement fields u and v are functions of coordinates x and y and strain
tensors are the gradients of displacement fields. At every point of a two dimensional body,
displacements u and v and three in-plane strain tensors &€,, €, and %,—derived from these
displacement fields—will exist. Now, the dilemma is that three strain tensors are derived from
two displacement fields. Thus strain tensors cannot be independent of each other. There must
exist an interrelation between the strain tensors so that continuity of strains remains at every
point of the body. For any admissible displacement fields u and v, we find that

deere Hee neee? E mY) OYxy

Se = (1.74)
oy" dx? oxdy\dy ox) odxdy

Hence Eq. (1.74) must be satisfied at every point in the body and is generally termed as
strain compatibility condition in two dimensions.
In a three dimensional body, u, v and w represent the displacement fields in x, y and z
directions respectively. In 3D problems, there must exist three strain compatibility conditions
to maintain the continuity of strains at every point of the body. Equation (1.74) can be
extended for other dimensions as

fe eee (1.75a)

mee (1.75b)

nie (1.75¢)

SUMMARY
In this chapter, definitions of stress and strain are given. Transformation matrices of
displacements, strains and stresses are also formulated for any arbitrary rotation of coordinate
axes. Computational aspect of principal stresses, principal strains are also postulated for both
| Chapter 1 Analysis of Stresses and Strains [ 31 |

2D and 3D conditions. The concept of Mohr’s circle for stresses and strains is given in detail
in this chapter. Various numerical examples are solved to provide the idea of stress and strain
calculations. Finally, stress equilibrium equations and strain compatibility conditions are also
explained. These concepts will be useful for the subsequent chapters and especially for finite
element computations.

EXERCISES
1
1.1 Show that T,4, = 51 —0;) and that the plane at which 1,,,, acts makes an angle of
45° with the direction of the largest and smallest principal stresses.
1.2 The component of a stress tensor at a certain place in a 2D body is represented by

==) 1
1 _7| MPa

If the plane 2x + y = 5 passes through the place in question, what is the stress vector
acting on the outside (the side away from the origin) of the plane? What are the normal
and tangential components of the stress vector on this plane?
1.3 The state of stress at a point (a, b, c) is given by the following tensors:

=o =| 2
=i —6 is MPa
Z 1 —8

(a) Determine the magnitude and direction of the principal stresses at the point
(ad, 40. G).
(b) Based on the principal stresses obtain in part (a), find the stress vector and the
magnitude of the normal and shear stress acting on the plane (x — a) + (y — b)
az) =)

1.4 The state of stress at a point is given by the following tensors:

=98 —1.6 ZO
-1.6 2.6 —2.8 MPa
2.6 2-8 =3.6

Determine the normal and shearing stresses acting on the plane perpendicular to x’ axis
1
whose direction cosines are /; = m, = n, = Ei Also assume that 1, = mp).

is For the stress tensor given in Problem 1.4, determine the magnitude and direction of
the principal stresses and the magnitude of the maximum shearing stress.
SQ Finite Element Method: Concepts and Applications in Geomechanics

1.6 Solve Problem 1.2 using Mohr’s circle of stress.

1.7 Using the data given in Problem 1.4, compute the ratio of magnitude of octahedral
shear stress to the maximum shear stress.
1.8 The stress at a point in a body has the following components:

—2 0 2
0 2 0 MPa
Ds 0 —2
Find the value of stress invariants N,, N, and N3 and also determine the magnitudes
of principal stresses.
1.9 The rectangular plate given in Figure P1.1 is in the state of plane strain, i.e. €, = ¥.,
= %- = 0. Determine the displacements for the plate given the deformations shown in
the figure and strain components for the (x, y) coordinate axes. Also determine the
strain component in (x’, y’) axes.
Straight line

Figure P1.1

1.10 Solve Problem 1.9 for the deformed body shown in Figure P1.2. Also estimate the
magnitudes of principal strains at the point (0.5, 0.25) and their directions with respect
to x axis.
y ’ Straight lines

Figure P1.2
| Chapter 1 Analysis of Stresses and Strains __ [ 33°]

1.11 For the given stress function ® = a,x* + ayx*y + ayxy” + ayxy’ + asy* in xy coordinate
system, stress tensors are obtained as

vo O° o°@
Se a Re Oma
= ——_ = ——_ d 1: =—

(a) Show that the stress equilibrium equations are satisfied (ignore body forces).
(b) If only a4 is a non-zero quantity, draw the o,, and 0,, on the side AB of the
element shown below:

Figure P1.3
 Stress-Strain
Relationships

2.1 Introduction
In this chapter, we will review the relationship between stress and strain tensors based on
material properties. The magnitude of induced strain in a body depends on the amount of the
applied stress, boundary conditions and the properties of the material. The relationship between
stress and strain tensors is known as constitutive relation of the material.
Rocks are mostly elastic-plastic material. The term elasticity refers to a particular
engineering property of rock having mostly linear variations between strains and stresses. An
elastic rock must also exhibit that upon removal of applied loads, the deformed body returns
precisely to its original geometry. In general, elasticity is of two types—linear and non-linear.
In linear elasticity, the magnitude of strains linearly depends on the amount of applied stresses.
The concept of linear elasticity is explained in this chapter for linear isotropic and transverse
isotropic behaviour of rocks. In non-linear elasticity, strain is a polynomial function of stress.
However, it is not a common elastic property of rocks.
Like most materials, rock also exhibits yielding characteristics and achieves ultimate
strength before failure occurs. Figure 2.1 shows stress-strain relationships under uniaxial
compressive loading of rock sample. Linear portion of the curve represents the elastic response
of the rock until its yield strength (0,) is reached. If the stress is increased beyond that point,
permanent or irreversible deformation occurs in the sample. In some rocks, failure occurs
suddenly after yielding having little or no permanent (plastic) deformation. After that, residual
stress condition is attained in the sample. This type of stress-strain behaviour is termed as
elastic-brittle plastic behaviour of rock. Examples of such rock type are granite, gneiss, hard
sandstone, slate, etc. Ductile rocks can sustain permanent deformation beyond yielding. Stress
increases with further straining of the rock sample. This property signifies strain-hardening
characteristics of rock. Weak rocks such as shale, mudstone, coal, etc. exhibit such behaviour
upon large confining stress. Ideally, elastic-perfectly plastic behaviour is defined when stress
does not increase with strain increment in the plastic regime. In other words, the stress remains
the same but the plastic strain continues to increase. However, in the plastic domain, most rocks
i Chapter 2. Stress-Strain Relationships [ 35 |

show strain-softening characteristics where the stress decreases upon further strain increment.
Shale, limestone, siltstones are some of the examples of this type of stress-strain behaviour.
In the plastic domain, increment of stress depends on the current state of stress and plastic
Strain increments. Thus, the constitutive relationship is non-linear and formulated based on the
current stress, plastic strain increment and elastic properties of rock material.

Strain-harding
——
any

— Perfect plastic

\ : ;
i Strain-softening
VS ae SESS

Brittle plastic
es

Figure 2.1 Stress-strain relationship of rock.

2.2 Linear Isotropic Elasticity

The term isotropic refers to uniform material properties in all directions. Thus linear isotropic
elasticity signifies uniform linear stress-strain behaviour of a material in all directions and upon
unloading, the deformed body returns to its original shape. Consider a rock specimen under
uniaxial compression loading as shown in Figure 2.2(a). Load is applied in the axial (y)
direction of the rock sample. The axial strain and stress are plotted for every load increment
as shown in Figure 2.2(b). In the elastic domain, the linear portion of the curve, stress is
directly proportional to the strain and can be expressed mathematically as:

Oy ay (2.1)

One— 2 EOL gEPer

a jae)
(2.2)
Sy
where
Oy = axial normal stress, N/m?
Ey = axial normal strain, dimensionless
E = proportionality constant or modulus of elasticity, N/m?

The above stress-strain relationship is known as Hooke’s law for one-dimensional loading
condition. Modulus of elasticity can be computed in three different ways. Average modulus of
elasticity is the slope of the stress-strain curve measuring from point A to B as shown in
Figure 2.2(b). Secant modulus is defined by the slope of the curve from point A to C, where
the point C denotes the 50% of the compressive strength. Tangent modulus is measured as the
slope of the stress-strain curve at point C. In general, the slope of the linear portion of the curve
is also denoted as the tangent modulus.
 Finite Element Method: Concepts and Applications in Geomechanics

© elastic
Plastic domain

Figure 2.2 Stress-strain relationship in uniaxial compression loading.

Due to axial compressive loading (y-direction) the diameter of the rock sample increases
causing lateral or circumferential strain. This phenomenon is termed Poisson’s effect.
Mathematically, this relationship can be written as

Oy
eas (2.3)
Exx E

where the parameter V is called Poisson’s ratio and defined as the ratio of lateral strain to axial
or longitudinal strain. The negative sign signifies that lateral strain has the opposite direction
that of axial strain. In the above example, axial strain is compressive or negative while lateral
strain 1s positive or tensile.
Applying the similar principle in three-dimensional stress conditions, where all three
normal stress tensors O,,, om and o,, act on x, y and z directions respectively, the infinitesimal
strains in three directions can be estimated as follows:
: (On. Oo Oo
Strains developed
Pp
due to 0,,:
XX
€,5B § =—“.,
E
¢,=-v—@,
yy E
€é,=-v—~=
ZZ E (
2.4a )

Onn Oo Oo
Strains developed due to 0,,: €&. =— v— | Eyy = , €&, =v (2.4b)
=f E E; E

Cx 2 O07, eumz Ox
Strains developed due to 0;,: €,, ealere: ey at ieee ie (2.4c)

Note that in the above equations, E and v are kept the same in all directions due to
isotropic conditions. Using superposition principle of strain, we can write the expressions for
the three normal strains as given below:

1
eee = alk: Vi GS iO) (2.5a)

1
Ey = ow —V( Ona) (2.5b)
| Chapter 2 Stress-Strain Relationships a

1
ae Sele VAT nat obeal (Qu5C)

In the similar fashion, the shear strain y also depends linearly on the shear stress T as

if “G6 (2.6)

where the parameter G is proportionality constant referred to as shear modulus or modulus

of rigidity. The shear modulus G is related to elastic modulus E by the following
relationship:

G= s Co)
2(1+ Vv)

For isotropic material, shear strains in three planes—xy, yz and zx—can be computed from
the shear stresses as given below:

G T
Vy = Ny Vx a (2.8)

Now combining Eqs. (2.5) and (2.8), the general Hooke’s law relating strain and stress
tensors can be written in the matrix form as

Exx 1 Vai V 0 0 0 O xx
Evy =. 1 —v 0 0 0 Oyy
E 1|-v -v 1 0 0 0 Oo
Payet iid (2.9)

pe 0 0 0 0 2(1+ Vv) 0 oy
Vex 0 0 0 0 0 21+v)]iz

or
sc Co
Here, the matrix C is called the compliance matrix of the material and contains only two
material constants as E and v. The compliance matrix is symmetric and contains 9 non-zero
positions in the upper-half of the diagonal. The interaction between normal strain tensors and
shear stress tensors or between shear strain tensors and normal stress tensors are neglected.
Apart from, E, v and G, another two elastic constants, bulk modulus K and the lame
constant A, are also used in theory of elasticity. These two constants are related to E and v
as follows:

E VE

"30-2" “OF yd=29) eee

 Finite Element Method: Concepts and Applications in Geomechanics

The bulk modulus is the ratio of hydrostatic stress to volumetric strain. The above
expression of bulk modulus can be derived computing volumetric strain and hydrostatic stress
in terms of normal strain and normal stress tensors, respectively. From Eqs. (2.5a), (2.5b) and
(2.5c), the volumetric strain is obtained as

1
Sy Seto yea ee ete aU Toya o..)] (2.11)

Considering hydrostatic stress condition, O,, = Oy, = O,, = p, the bulk modulus K is
computed as
Pp EE
Ej tL—2y) Cz)

In finite element method, we estimate strains from displacements and then compute
stresses in terms of strains. To obtain the stresses in terms of strains, we need to invert the
matrix C as given in Eq. (2.9), and then the relationship between stress and strain tensors
becomes
l-v V V 0 0 0
On Vv l—v v 0 0 0 ee
Oy v V l-v 0 0 0 ay

AGE E 0 0 OF sere eae 0 |J% 1 043)

Tip (oe Glenvid Sav) - ae
Tye 0 0 0 aie sh = 0 |l%,
aa 0 0 0 0 0 ap pee cee
D
or

o=De
The matrix D is called the elastic or constitutive matrix of the material. This constitutive
matrix is also symmetrical and has 9 non-zero terms.
Based on the above discussion, the complete stress-strain relationships can be
expressed as
6 = D(é- £9) + Gg (2.14a)
€=C(O - 69) + Eg (2.14b)
where Oo and €p are vectors comprising initial stress and strain tensors. Unless specified, for
most of the cases, they are considered to be zero or null vector.

Example 2.1 The following figure shows two rods connected to each other. Dimensions of
rods and material property are given. If 1 MN force is applied at point A, calculate the stresses
and strains in the rods. What will be the displacement at point A?
 Chapter 2 Stress-Strain Relationships |

E=4~x 10! Pa E=2~x

10! Pa
0am if =30. 55m
d= 05 d=0.25m

Solution: Since it is one-dimensional stress field, stresses in the rods can be estimated
as follows:

6 Fr 10°
eos) Foy —20.27 MPa
A; 10.57/4 Ay 70.257 /4
Using Eq. (2.2), strains can be estimated as

| Cie =:
g, = Tb= 2 81.0725 x10, €) =—2 === =1.0185x 10°
E, 4x10 FE, 2x10

Deflection at point A will be the summation of deflections of rods 1 and 2 as

O4 = €,L, aia €>L> = 0.573 mm

2.2.1 Relationship between Principal Stresses and Principal Strains

Since the material constants do not change with the direction, the relationship between principal
stresses and principal strains is straightforward. Recalling that shear stresses do not exist in
principal plane, the compliance matrix and constitutive matrix can be obtained by truncating
Egs. (2.9) and (2.13) respectively, as

E} j i =—V -V || 0;
E> ee -V 1 -V |40> (2.15)
& =V —-V 1 | [03

0; =v v Ve;
Or; = a v iy, Vole
AVE sila 2 (2.16)
03 Vv Vv 1-v|[&

Example 2.2. Refer to Example 1.3 and calculate the principal strains. Assume the isotropic
rock with E = 20 GPa and v = 0.25.

Solution: Since in the principal stress planes, no shear stress exists, Eq. (2.15) can be
written as:
[40 | [ Finite Element Method: Concepts and Applications in Geomechanics

Ey , 1 Ve =Vi On
E> avy V4 1 —V 1505

&3 =V ae 1 | [03
or
E| ; 1 —0.25 —0.25 | |-37.83
E> eT 0:25 1 —0.25 |4—20.64
3 0.25 —0.25 1 —4.64
Hence
€, = -1.575e3 (m/m), —-&) = —5.01le* (m/m), —-& = 4.988e* (m/m)
Here the major and intermediate principal strains are compressive but minor principal
strain is tensile.

2.2.2 Stress-Strain Relationship in Two Dimensions

Stress analysis in 2D is an approximation of 3D by taking suitable slices of a 3D problem.
In mining, civil or geotechnical engineering problems, three types of approximations—plane
stress, plane strain and axisymmetric conditions—are assumed to represent a 3D problem in a
plane.

Plane Stress Conditions

In this case, loads are only applied to xy plane and z direction is not restrained so that
O., = Ty, = T., = 0 (Figure 2.3). Examples of this type of analysis are thin pipe, hole in a thin
plate and others. For the analysis of underground stresses, plane stress analysis is not generally
considered since the length of excavation is much greater as compared to the excavation width
or height. Normal strain in out-of-plane or z direction does exist and is computed using stresses
O,, and Oy,.

Figure 2.3 Plane stress condition.

 Chapter 2 Stress-Strain Relationships
The compliance matrix can be obtained by simply omitting appropriate locations of Eq. (2.9)
as given below:

Exx 1 -V 0 —V | |Ox
yall lV. 1 0 -V | |%y
tore 04) 0. 20 4V) 2 both, (2.17)
Ex Via iV 0 1 0
Note that in the above equation, the out-of-plane tensor is placed at the last. The
constitutive matrix for plane stress condition is then obtained as:
l-v Vv 0 Vv
Ornx Exx

Oy, E Vv wv. 0 v Ey
= AM aya
1-2 1-2v
tay | el VAL ZV). 0 0 eer 0 |1Yxy (2.18)
0 Vv V 0 fy ee

From Eq. (2.17), it can be easily shown that the normal strain in the out-of-plane
direction is:
Vv Vv
Cea meoy) = aes + Oyy) (2.19)

Plane Strain Conditions

For the stress analysis of a long tunnel or mine gallery, deformation along z direction is
assumed to be zero except in the end boundaries (Figure 2.4). Displacement fields u and v in
xy-plane are dependent on x and y coordinates and not on z coordinate. Since it is assumed that
no deformation exists in the out-of-plane direction, i.e. z axis, €,, = %, = Ye = 0. Then the
constitutive matrix is obtained by truncating Eq. (2.13) as:

oe —
l-v Vv 0 \.
Oy z Vv l-v 0 eye
oe 2 4 1-2v (2.20)
(eo (l+v)ad-—2v)| 0 0) rere 0 Vxy

OY 0

Figure 2.4 Plane strain condition.

 Finite Element Method: Concepts and Applications in Geomechanics

As before, note that out-of-plane component is placed at the last row and column. The
compliance matrix is then obtained as follows:

re i -—V 0 —V | |Ox
yi 1{-v 1 0 -V ||Oyy
yop HOC. 0 dy), Uae (2.21)
0 -—V -V 0 1 On

From Eq. (2.20), it can be easily shown that the out of plane stress, which develops in
the z direction, is

Ox, =V(Oxx + Dy) = ACE, + Evy) (2.22)

Axisymmetric Conditions
Axisymmetric analysis is performed for structures that are rotationally symmetric about an axis.
Examples of axisymmetric structures in geotechnical engineering are circular tunnels and shafts,
tunnel lining, supports and others. If the loading on the structure is symmetric as shown in
Figure 2.5(a), a two-dimensional analysis can be performed considering a unit radian of the
structure. Here, y axis is considered to be the axis of symmetry. Stress notations are given in
polar coordinate system as 0” = (0, Og Te Teo O,,), where 6,, is radial stress, Ogg is
tangential or hoop stress, 7,9 is shear stress and O,, is longitudinal stress. Symmetrical loads are
applied in radial or longitudinal directions. If the loading on the structure is non-symmetrical,
then the choice of analysis lies between fully three-dimensional, in which substructuring or
cyclic symmetry and a Fourier decomposition of the loads with a Fourier-axisymmetric solution
(Bathe, 1982) are used. In this text, we will discuss axisymmetric analysis with symmetric
loading condition only.

Axis of symmetry

(b)
Figure 2.5 Axisymmetric condition.

Consider the stress conditions of a small element as shown in Figure 2.5(b). Let us assume
that the radial and longitudinal displacement fields are u and v respectively. Then the strain
tensors can be obtained as

du _2a(r 4u)— lar eu du ov _ ov

Sp a?
ae ie £00 r
=
ee
5 S16
yy jp =)
a
=
oy G23}
' Chapter 2 Stress-Strain Relationships || 43

The constitutive relations between stress and strain are similar to that of plane strain
condition and given as

On es
Soe E v LV, 0 Vv Lop

tof (+vy—2v)| 0 0 := 0 |) (2.24)

we v Vv 0 toy i

From Eq. (2.24), compliance matrix is obtained as

Err 1 eV 0 ZV O+,

E99 a -V 1 0 —V | |Og@ (2.25)

Yre| E| 0 0). oe Ee
Evy -V =V 0 1 | {Sy

Example 2.3 A 45° rossette is used to estimate in situ stresses in a rock. The normal strains
obtained are as shown in the figure below. If the modulus of elasticity and Poisson’s ratio of
that rock is 5 GPa and 0.25, determine the major and minor principal stresses. Assume plane
strain condition.
-300 uw
y
-500

BLS)”
x, -300

Solution: Given data are

&j5= 00H = &y & =-l00n &,=-300H E=5 GPa v= 0.25

We know that

epee bei
yy = + Bio cos20 + sin 20
2

Putting the given values in the above equation, we find

B56 _ L004, _ BO a0 “2sue

Yay =~ 600
 Finite Element Method: Concepts and Applications in Geomechanics

Considering plane strain condition and using Eq. (2.20),

on l-v Vv 0 Vv on
Vv l-v 0 2
Oy, E yy
Gl | a ee 1+2v
i) CF v)dee2y) }. 0 0 5 0 Y xy

Ore V V 0 l-v 0

Or 0.75 On25 0 0.25 | |-—300 —2.0

oOoi ee 8000 0.25 0.75
: 0 0).25 11-100 x —1.2 “vith
(Se 0 0 0.25 0 —600 —1.2
on 0225 0.25 0 0.75 0 —0.8

Be aleet nD ; (2.0 + 1.2) + 4x(-1.2)°

2
= — 2.865 MPa

Similarly
03 =—0.336 MPa
Assuming direction cosines of the maximum principal stress as / and m,
m=0.721 and I = 0.8115
The angle with x axis will be 35.71° counterclockwise.

2.3 Linear Transverse Isotropic Elasticity

The modulus of elasticity, Poisson’s ratio and other material constants can be different in
different directions, especially for stratified rock layers such as shale, mudstone, siltstones, etc.
Transverse isotropy is a special case of generalized relationships between stress and strain
tensors. In this case, a plane of isotropy denoted by xy plane is assumed as shown in
Figure 2.6. In this plane, the modulus of elasticity and Poisson’s ratio are constant in all
directions. Along the outer axis (z axis), normal to the plane of isotropy can possess different
elastic parameters than the plane of isotropy. Since two different elastic properties can exist in
perpendicular directions, this property of material is termed transverse isotropic material.
Assuming E,, v; and E>, v2 as the elastic moduli and Poisson’s ratios in the plane of isotropy
and normal to that plane respectively, the compliance matrix can be expressed as
| Chapter 2 Stress-Strain Relationships

oe 1 -V; -V> 0 0 0 On
Eyy -V, 1 -V> 0 0 0 Oyy
ee lve | —Viaeink,/E, 0 0 Ouse) Oval: 5h
V xy Ey 0 0 0 2(1+V,) 0 0 Ts
as 0 0 0 0 E,/G, 0 Liss
Vex 0 0 0 0 0 E,/G, Tis

where G, is defined as the shear modulus along perpendicular direction and in general, it is
independent of E;, v;, E, and v3. However, Batugin and Nirenburg (1972) found that G, is
dependent on the elastic parameters E,, E, and v> for different rock types such as limestone,
sandstone, granite, slatestone, phyllite and granodiorite and the relationship is given as
(Kwasniewski, 1984):

E\E>
Reh (lov ES zee)

It can be noted that the compliance matrix is symmetric just as in the case of linear
isotropic elasticity. The constitutive matrix is reproduced from Brady and Brown (1985) and
given as follows:
2
On n(l—nvz) ny, + nv>) nvz(l+v,) 0 0 0 Ev

oe n(v, t+nvs) n(l—nv3) nv2(l+v,) 0 0 0 ee

0., E, |m(ity) mtv) (d-v) 0 0 0 Ex
To |, tv JA 0 0 0 nA 0 0 Voy
1e
g) Yyz
e 0 0 0 0 m(l+v,)*A 0 4
Tox 0 0 0 0 0 meee ALCS
(2.28)
where A = (1-V, nVA n = E,/E, and m = E,/G). It can be noted that E,, E>, v, and Vv,
are supposed to be independent of each other. However, the values of n and m can be restricted
due to constraints of positive definiteness of the constitutive matrix (Brady and Brown, 1985).
Zz
E, Vz

Figure 2.6 Transverse isotropic condition for stratified rock layers.

 | Finite Element Method: Concepts and Applications in Geomechanics

2.4 Degree of Anisotropy

The degree of anisotropy of transversely isotropic material is generally denoted by the ratio of
elastic moduli as:

kr _ Fy (2:29)
E
However, stresses and displacements occurred in anisotropic rock mass are functions of
coefficients of elastic moduli and Poisson’s ratios determined in various directions. Hence the
knowledge of only kz may not be sufficient to define degree of anisotropy. Considering two-
dimensional problems, Lekhnitskiy (1962, 1968) suggested the parameters of anisotropy for
transversely isotropic material as (Kwasniewski, 1984):

n=2k+m (2.30)

_ E/G, -—2v,0.+V;)
where and m
lve

For problems related to state of stresses of rock mass around cross-headings, diagonal
headings, roadway openings, an acute angle in relation to the strike of rock strata occurs. Due
to this reason, an additional parameter / is defined by the following equation:

l= fied Mises
~ ¥20+V4)G, (2.31)
The degree of anisotropy of transversely isotropic material is then estimated by the
deviation of values of the above-mentioned parameters, namely k, n and J, from their isotropic
values of k = 1, n = 2 and / = 1. The classification of transverse isotropy is given based on
the value of parameter n and is shown in Table 2.1 (Kwasniewski, 1984).

Table 2.1 Classification of anisotropy

Parameter, n Class of anisotropy

<M Quasi-isotropy
Between 2.1 and 2.5 Poor anisotropy
Between 2.5 and 3.0 Moderate anisotropy
> 3.0 Strong anisotropy

2.5 Non-Linear Elasticity

The major difference between linear and non-linear elasticity is that for the latter case, the
relationship between stresses and strains is not linear in the elastic regime. Non-linear elasticity
can be best represented by the schematic diagram as shown in Figure 2.7. The loading curve
| Chapter 2. Stress-Strain Relationships

Loading
Stress

Unloading

Dissipated energy

Strain

Figure 2.7 Non-linear elasticity.

is non-linear and depicting stress is a polynomial function of strain. The unloading curve may
not coincide with the loading curve and thus exhibits hysteresis effect which corresponds to the
imperfect elastic behaviour. In general, non-linear elastic behaviour is not a common property
for rocks but certain hard rocks may show this type of behaviour (Jeremic, 1985).
Determination of elastic modulus and other constants is performed by dividing the
hysteresis of the stress-strain curve into several piecewise linear divisions and then for each
division, linear elastic principles are applied. However, superposition strain theory cannot be
applied in such cases.

2.6 Plasticity
Plasticity results in permanent deformation or strain in material causing micro- to macro-
cracks. Let us revisit the case of uniaxial compression loading on a rock specimen as described
in Figure 2.1. At the initial stages of compressive loading test, most rocks show linear elastic
constitutive behaviour at least up to 50% of yield load. As the loading increases, material
departs from elastic property once the applied stress exceeds the yield strength of the material.
Micro- or macro-cracks develop in the rock specimen causing irrecoverable strains. Gramberg
commented that under uniaxial compressive loading, micro-cracks initiate in rock samples at
about 50% of yield load and secondary fracture phenomena develops after 75% of yield load
(Gramberg, 1989).
In the plastic domain, stress 6 and strain € are no longer related by matrix of constants
such as the constitutive matrix D. The relationship is based on rather strain-dependent
constitutive matrix. The computational difficulty in plastic domain is that equilibrium equations
must be formulated using material properties that depend on strains, but strains are not known
in advance (Cook et al., 1989). For this reason, incremental theory of stress and strain is
adopted. The strain increment is divided into two parts —elastic and plastic strain
increments—as given below:
de’ =de° + de? (2.32)
 in Geomechanics |
[ 48 || Finite Element Method: Concepts and Applications

and plastic cases respectively.

Here the superscripts t, e and p represent total, elastic
the stress and thus the relationship
Elastic strain component is responsible for incrementing
between stress and strain increments can be expressed as:

do = Dde* = D(de' — de”) (2:33)

strain increment is equal

In elastic domain, plastic strain increment is zero and thus total
both elastic and plastic strain
to the elastic strain increment only. In the plastic domain,
component of strain increment is
increments exist except for ideal plastic case where elastic
d.
zero. In Chapter 7, details of elastic-plastic analysis are presente

2.6.1 Perfect Plasticity

for one dimensional
Figure 2.8 shows the diagram of idealized plastic behaviour of the material
ts occur in the
loading condition. Once the stress exceeds the yield strength, no stress incremen
strain increment
material while strain continues to develop. For ideal plasticity problem, elastic
the applied load is
is zero and hence no stress increment occurs in the plastic domain. If
removed at point A, the original geometry cannot be recovere d due to permanent strain
developed on the specimen.

Perfect-Plastic A

Figure 2.8 Elastic-perfectly plastic behaviour.

2.6.2 Elastic Plasticity

Material behaviour in the plastic regime is shown in Figure 2.1. In this section, we will
illustrate the relationship between stress and strain for strain hardening behaviour due to one-
dimensional loading condition as shown in Figure 2.9. For simplicity, linear stress-strain
relationship is considered in plastic region. Let us assume that in the plastic region, stress is
increased from point A to point B giving stress increment, do. Then do can be expressed in
different forms as follows:
do =fdé =Ede —Hde (2.34)
Here, E, and H are called tangent modulus and strain-hardening parameter. By manipulating
Eq. (2.34), we can write that
 Chapter 2 Stress-Strain Relationships

ips
_&
or

E
B= e(1-5e) (2.35)

Plastic

Figure 2.9 Elastic-plastic stress strain behaviour in one dimension.

For ideal plasticity, E, = 0 and hence H = 0. For strain softening behaviour, the hardening
parameter H is assumed to be negative. The tangent modulus can be obtained from uniaxial
test of rock samples.
For multi-dimensional cases with 3 or 6 stress and strain tensors, similar approach is
applied. However, determination of increment of plastic strain requires special assumption on
the behaviour of the material. In Chapter 7, concept of elastic plastic analysis is given in detail
with rock mass yield criteria.

Example 2.4 Considering Figure 2.10, if the stress increment from A to B is 0.1 MPa,
determine the increment of plastic strain. Assume E = 5 GPa and E, = 1.8 GPa.

Solution: From Eq. (2.30), we find that e* = 0.1/5000 = 2e> and & = 0.1/1800
= 5.55e>. Hence, plastic strain increment will be €? = é& — & =. a 550°

SUMMARY
This chapter mainly deals with the relationships between stresses and strains in a body. The
concept of linear isotropic elasticity is introduced mentioning the essential two material
properties viz. modulus of elasticity and Poisson’s ratio. The compositions of compliance and
constitutive matrices are also discussed here. The properties of transverse isotropic material are
illustrated with indicators for measuring degree of anisotropy of rock mass. The general
concept of elastic-plastic behaviour of rocks is also portrayed clearly mentioning elastic-
perfectly plastic, strain softening and strain hardening materials.
 EXERCISES
ting strain gauges in axial
2.1 A rock sample is tested in uniaxial compression by connec
0. = 40 MPa, E = 5.7 GPa and
and lateral directions. The results of the test are:
of failure of the sample.
v = 0.28. Determine the volumetric strain at the time
tensors if the modulus
2.2 Considering the results of Problem 1.9, determine the stress
and 0.25 respectively.
of elasticity and Poisson’s ratio of the plate are 5 GPa
tensors if the modulus
2.3 Considering the results of Problem 1.10, determine the stress
5 GPa and 0.25 respectively.
of elasticity and Poisson’s ratio of the plate are
2 and | =1 as given
2.4 Show that for linear isotropic material, the parameters k = 1, n =
in Eqs. (2.30) and (2.31).
mentioned in
2.5 If G, = 2G, and v, = 0.5, v2 = 0.2, determine the parameters as
Problem 2.4.
rosette.
2.6 Measurement of principal stresses is conducted along a vertical plane with 60°
magnitude
The normal strains are obtained as given in Figure P2.1. Determine the
direction of major principal stress with the x axis.
of the principal stresses and the
ratio of the rock are 5 GPa and
Assume that the modulus of elasticity and Poisson’s
0.25 respectively. Consider the plane strain condition .

—500 u 200 1

60°
60°
—800

Figure P2.1

2.7 In Figure 2.9, assume Oy = 41.9 MPa, Og = 42.2 MPa and o, = 42 MPa. Determine
the elastic and plastic strains when the load is increased from A to B. Assume
= 4. GPa and E25) GPa:
3.1. Introduction
The finite element method (FEM) is an alternative procedure to analytical methods for
analyzing structures, heat transfer, fluid flow and other problems related to science and
engineering. It is an approximate numerical solution technique in which continuous systems are
discretized into many small and simple pieces called finite elements. For each element, it is
necessary to make an assumption as to how the primary variables, such as displacement, are
distributed in terms of geometric position. This assumption is the basis for the development of
finite element analysis procedure. Then a set of simultaneous equations is developed for
describing the constitutive or other behaviour of each element in terms of discrete nodal point
values of the primary variable. Each of these elements is then combined using proper
compatibility relations between them and global set of simultaneous equations is obtained. Then
the application of loads and boundary conditions are imposed to the global set of simultaneous
equations. These equations are solved simultaneously or implicitly in a personal computer,
mainframe, workstation or supercomputer. Solutions of these equations provide the
approximate results or the prediction of behaviour of the physical system that has been
modelled. The solution obtained from finite element analysis is not exactly like the closed form
solutions.
In geotechnical engineering, finite element method is being applied in design and analysis
of underground structures such as tunnels, caverns, powerhouse, mine pillars and galleries and
stopes. Besides, analysis of slope stability, dam structure, subsidence and others can also be
effectively performed using this technique.
The major advantages of finite element method reside on its versatility to apply in various
problems having irregular shapes and sizes, multiple loading and boundary conditions. As for
example, stresses and deformations around any size and shape of underground opening can be
analyzed under different loading and support conditions. Modelling is also carried out with
multiple rock layers having different mechanical properties. The strength of FEM also lies on
the resemblance of the physical size of the problem with the FEM model. There is no need
51
 | = Finite Element Method: Concepts and Applications in Geomechanics
[ 52

the stress and

to scale down or up the actual structure being analyzed in order to obtain
deformation behaviour. The actual engineering properties of rock or other material are used in
t material properties as
finite element analysis. On the contrary, it is required to use equivalen
in the case of physical modelling techniques.
Recent improvement in computing speed and digital storage capacity resulted in the
development of versatile FEM softwares. The introduction of pre- and post-processing modules
with most of the commercial FEM software packages enables the user to create models using
graphic user interface (GUI) and visualize the stresses and deformation of structures on the
computer screen. These options allow the user to interpret the results more effectively. If
required, the model parameters can be modified instantaneously using those softwares and the
results of the modified models are verified subsequently for its accuracy. Thus the user has now
a powerful tool to design and analyze the stability conditions of any structure whether it is
located on the surface or underground. Optimization of the design parameters are also
performed before the structures are built or excavated. The development and analysis of
numerical models has saved tremendous amount of time and money since the entire structure
is designed and simulated in the computer before it can be put into practice.
The disadvantage of finite element method is that it does not provide closed-form solution
and thus minimization of error has to be done judiciously by increasing the number of elements
or independent variables or by changing the model parameters. The FEM also provides
problem-specific results. In other words, if a FEM model is created for a 2 m diameter circular
tunnel, then the results obtained from this model cannot be extrapolated for a 3 m diameter
tunnel. For the later case, a separate model is developed and analyzed. Sometimes, limitations
of computer capacity can also obstruct the FEM analysis procedure. Lastly, the numerical
voluminous output may have to be sorted out before any meaningful information or result can
be obtained.

3.2 Discretization of Physical System with Finite Elements

In this section, we will explain how a physical system is represented with finite elements to
solve primary variables such as displacement. The term physical system means a solid finite
continuum which is fixed at boundary I, with displacement vector q = q* and is loaded at
boundary I, with surface traction vector F as shown in Figure 3.1. Note that the figure
represents a solid continuum in 2D Euclidian space xy. The body force components, B, and B,,
may also be present in the system. FE [

| ¥ Solid body, Q

Figure 3.1 A typical solid body under loading condition in 2D.

|Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements [ 53 |

One-dimensional discretization of solid body with finite elements will be discussed first.
Then we will proceed into discretization of two-dimensional solid bodies.

3.2.1 Concept of Continuum Mechanics in One Dimension

Figure 3.2(a) shows a solid bar in one dimension. The length and cross-sectional area of the
bar are L and A respectively. The bar is constrained at one end with vz = 0 and a load F is
applied in negative y-direction at the other end, C. Applying the principle of continuum
mechanics for one-dimensional problem, we can define a continuous polynomial as the
displacement field v = v(y). In this case, we assume a linear displacement field except the
intercept, since vg = 0, as explained below:
v(y) = ay (3.1)
where a is a constant and will be determined from the loading and boundary conditions of the
problem.

D
©)

y L L

| gtw B
(a) Solid 1D bar (b) Discretization (c) Discretization with
with single element multiple elements
Figure 3.2 Discretization in one dimension.

Hence the strain and stress in y-direction can be estimated as given below:
es Cheon , 3
y dy ( .2)

0, =EE, = Ea (3.3)

Then the total potential energy of the system is

I, =5Jeoav
1
sy ==A |,rl Oyeydy ~ Fal > = AEL
a" eo.—aFL 34
(3.4)
V

Applying the stationary principle of potential energy to obtain the equilibrium equation
(see sub-section 3.5.1), we find that

dll
OI, — FL) da =0
=" 6a =(AELa (3.5)
on anlG

From Eg. (3.5), since da # 0, we find the value of a by solving the equation
AELa — FL = 0 as
 Finite Element Method: Concepts and Applications in Geomechanics

F
seit =—
93 (3.6)
3.6

Thus the displacement field is obtained as

By
v(y)(y) =——
4E 37
(3.7)

Once the displacement field is known, the strain and stress at every point in the solid body
can be determined.

3.2.2 Discretization with Finite Element in One Dimension

In order to analyze the problem outlined in sub-section 3.2.1 and Figure 3.2(a), first
discretize the bar into small piece(s) or element(s). In this case, for simplicity, let us divide
the bar with only one element having two nodes as shown in Figure 3.2(b). A bar element
having two nodes at each end is called linear bar element. Note that the same bar can also be
divided with multiple elements connected in series as shown in Figure 3.2(c). In this case, the
elements 1 and 2 are connected at the node 2 and the elements 2 and 3 are connected at node
3. If a bar element has an additional node at the mid-length, then it is called quadratic bar
element.
Finite element method seeks for displacements at nodes of an element whereas continuum
mechanics method finds it at every point of the body. For linear bar element oriented parallel
to y axis as shown in Figure 3.2(b), we assume a linear displacement field based on
y-coordinate only as

v(y) = a + by (3.8)

The constants a and b are obtained by letting the nodal displacements v, at y = 0 and
v, at y = L. Then Eq. (3.8) can be rewritten in terms of nodal displacements as

v(y) = [1
— an + (2).

or

v(y)
= Nyy + Nov (3.9)

Here N, and N, are the functions of the coordinate y and are termed shape functions of
the nodes 1 and 2 respectively. Details of the derivation of shape functions and their
properties will be explained later. Equation (3.9) refers a displacement field in the element as
a function of coordinate (in this case y) and nodal point displacements. Then axial strain and
stress are obtained from the displacement field v(y) and constitutive relations. After that,
external forces and boundary conditions are applied in the system of equations and nodal
displacements are solved applying stationary principle of potential energy as explained in
Section 3.10.
 Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements [ 55 |

3.2.3 Discretization with Finite Element in Two Dimensions

Now, let us assume a general body Q in two dimensions defined in Cartesian coordinate system
as shown in Figure 3.1. The body is constrained at boundary I, and an external traction is
applied at the surface I. Let us assume that the displacement field in the body can be defined
with a continuous polynomial function of x and y having p degree as

u(x,y) = dg + a,x? + ayxP ly 4---ta. yxy) +a,y? (3.10a)

V(x,
y) =by + Bx? + by xP ly $0 +b, yxy? | + byy?> (3.106)
where a; and b; are constants. The constants a; and b; are evaluated based on the nodal point
displacements and corresponding interpolation (shape) functions of element. For triangular
elements, a complete polynomial can be defined in Cartesian coordinates using all terms of a
Pascal triangle as shown in Figure 3.3. Based on the degree of polynomial, number of nodes
is defined as

na let dlp +2) CMW

Me

Pascal triangle

ee ae
neama Saiftantabou
a eee
Figure 3.3 Pascal triangle for estimating polynomial of triangular element.

Thus if the degree of the polynomial is 1, there are three nodes in the vertices of the
triangle and it is termed as linear triangle. For a polynomial of degree 2 or a quadratic
triangle, node at each vertex and at the mid-point of each side will be needed. The triangular
elements of degrees 1, 2 and 3 are called linear, quadratic and cubic triangular element
respectively, as shown in Figure 3.4. Quadratic and higher order triangular element can be used
with curved sides. For cubic and higher order triangular elements, internal node(s) will be

ak
Linear triangle Quadratic triangle Cubic triangle

Figure 3.4 Different triangular element with location of nodes.

 in Geomechanics
Finite Element Method: Concepts and Applications

y with 6-noded triangular

present in the element. Figure 3.5 shows a discretized tunnel boundar
y or curved surfaces and
elements. In general, smaller elements are formed near the boundar
the boundary. If a triangular
relatively bigger elements are used to model surface away from
element is too skewed, i.e. any internal angle is less than 30° or more than 120°, the element
avoid such element in the
is considered to be a bad-shaped triangle. Effort must be made to
finite element analysis.

DANES
VAVAVAN
DS4
2k!
ee
ra
(WANE

Na avave®: ZYDALY
VAVay RABE
Ay
ava. LY?
ZN
OAV AVAVAVINO AU AWA Pa
qPLLA AZ
VAVAVAV DAAALAAAK

Figure 3.5 Discretized tunnel boundary with quadratic triangle.

3.3 Nodal Shape Function

After the element type and corresponding displacement fields are defined, relationships
between displacement fields and nodal point displacement vector are established similar to
Eq. (3.9). In general, the displacement fields can be expressed with respect to the nodal point
displacements and interpolation functions of each node as

u(x.y)=
> Ni(xyuj, — vx.y)
= DNC y)y; (3.12)
i=\ i=l

where u; and v; are the displacements in x and y directions for ith node. The interpolation
function, N,(x,y), is called the shape function of ith node. A shape function describes how the
primary variable is distributed over an element. This function is estimated from the assumed
polynomial for each element type and is the basis for developing finite element procedure.
Shape functions can be derived directly using global coordinate system. However, it is easier
to formulate them using ‘local or natural coordinate’ system which will be described further.
The isoparametric finite element formulations consider the derivation of elemental equations
using natural coordinate system and then integrate them into the global coordinate system. In
other words, the principal idea is to establish a direct relationship of displacement fields at any
point within the element based on nodal point displacements and nodal shape functions derived
with local or natural coordinate system. Thus isoparametric formulation of finite element
|Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements

provides better understanding of finite element method. The following sub-sections illustrate
the concept of natural coordinate system for triangular elements.

3.3.1. Natural Coordinate System for Triangular Element

Natural coordinates are dimensionless and defined with respect to the element rather than global
coordinate system. Using natural coordinates, formulation of element matrices becomes simple.
For a triangle having vertex nodes 1, 2 and 3, the natural coordinate system is defined as
follows and is shown in Figure 3.6.

A
[Og OE See
Ay a RS eae
A; (3.13)
A A A
where P, having coordinates x and y, is any point inside the triangle and A is the area of the
triangle defined as

' 1 : 1
ae) 6 iia Ghee Ne (3.14)
. yj y2 ¥3

The values of r, s and ¢ at nodes 1, 2 and 3 are (1, 0, 0), (0, 1, 0) and (0, 0, 1) respectively.
yy

(x3, 3)
y2)
2,

1
(1, yi)

Figure 3.6 Definition of natural coordinate system for a triangle.

Table 3.1 shows the values of natural coordinates for 3-noded and 6-noded elements. From the
definition of natural coordinates, it is clear that they satisfy the following constraint:
rest t= 1 (3.15)
Table 3.1 Natural coordinates at different nodes

Node no. (1) rj Sj tj

We) = i.e) ro)

See
a
DR
=
nPWN 1/2 0 1/2
 Finite Element Method: Concepts and Applications in Geomechanics

3.3.2 Shape Functions of Triangular Element

Nodal shape functions are derived based on natural coordinates and are functions of r, s and
t, rather than global coordinates x and y. In order to generate finite element from a triangle,
shape functions at each node are defined as a function of natural coordinates as

N; = N,(7,5,01) = Nj) (3.16)

where N, is the shape function of ith node. Here the coordinate ¢ is omitted since
t=l1—rT-—'s.
Let us assume that @ represents a primary variable or field quantity such as displacements
u or v, u being displacement variable in x direction and v that in y direction. Then for
isoparametric elements @ will relate the shape function as

P=) Ng; (3.17)

i=l

where n is the number of nodes and @, represents field quantity at node i. The shape function
has two necessary properties as:
1. The value of N; is unity at node i and vanishes in all other nodes.
2. Summation of all shape functions should be unity at any point inside the element.

Derivation of Shape Function: Direct Method

Shape functions can be derived in different ways. The direct approach is to start with the
polynomial equation of functions r, s and ¢t (note that coordinates x and y are replaced by
natural coordinates) of order p having constants a;. The general expression of the polynomial
in natural coordinates is
n

p= > ar’st4 (3.18)

=|

where b, c and d are non-negative integers and ranges over the possible combination such that
b +c+dz=p. The constant a; must be estimated from the condition of function at each nodal
point. For example, a polynomial of degree | or displacement field of linear triangle is written
as
Q = ayr + ans + azt G19)
Now the value of this polynomial at node | with r = 1, s =t=0 will be @, = a;. Similarly
replacing the values of natural coordinates of nodes 2 and 3 in Eq. (3.19), we can find that
(2 = ay and @; = a; respectively. Thus the field variable can be expressed using nodal values
and shape functions as

Q,
= N,Q, + N22 +N303 =[N; Ny N3]49,+=N'o@ (3.20)
3
 Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements | 59

where N; = r, Ny = s and N; = t. For higher degree polynomial, the direct method for
evaluating shape functions is tedious. Thus direct method is not so popular technique if the
degree of polynomial is more than 2.

Serendipity Method
Serendipity element contains nodes only at the boundary where Lagrange element may contain
nodes inside the element. However, since the polynomial function of field quantities of a
triangular element is defined based on all terms of a Pascal triangle, all triangular elements are
from both the serendipity and Lagrange family of elements. The difference between serendipity
and Lagrange element will be more apparent for quadrilateral elements and is given in detail
in Chapter 4. Here, derivation of shape functions of a 6-noded quadratic triangular element is
given based on serendipity method. Let us consider the equation of lines given in Figure 3.7
with natural coordinates.

Figure 3.7 Serendipity method for deriving shape functions.

The shape function for node i of a serendipity elements is estimated by multiplying the
equation(s) of minimum number of lines that pass(es) through all other nodes except the node
i and a constant, a and then equating this product to unity. As for example, for 3-noded linear
triangle, the line r = 0 passes through both nodes 2 and 3 and thus shape function of node 1
will be
Nero A (3.21)

Now the value of r at node 1 is unity and hence a = 1 at node 1. Thus N, = r. Readers
can easily find that with the similar concept Nj = s and N; = t.
In the similar fashion, for a 6-noded quadratic triangular element, the lines r = 0 and
r — 1/2 = 0 pass through nodes 2-5-3 and 4-6 respectively and thus the shape function of
node 1 will be
N, = ar(r — 1/2) = 1 (322)

Here, the value of a at node 1 [letting r = 1 in Eq. (3.22)] will be 2 and thus
Nrp=r@r=_1) (B28)
 60 Finite Element Method: Concepts and Applications in Geomechanics

= 1. At node 4, the value of

For mid-side node 4, the shape function will be Ny = ars
s of each node
rand 5s is 1/2 and hence a = 4. Therefore, Ny = 4rs. This way shape function
for linear and quadratic triangles can be obtained as given in Table 3.2.

Table 3.2 Shape functions of linear and quadratic triangles

e
i eee ee ee
e ee e

Node no. 3-noded linear triangle 6-noded quadratic triangle

Nise N, = r(2r —- 1)

Ny =S N, = s(2s — 1)
N55 aes N3 = (l-—.n—s) (l= 2r — 25)
| Ng = 4rs

- N,; = 4s(1 — r — 5)
=
DAunkrwn No = 4r(l — r—- 5)

Lagrange’s Interpolation Method

Lagrange’s interpolation formula for 1D problem is defined as if a polynomial function

y = Q(x) of degree n—1 has the form

Q= YN; (3.24)
Sl!

where Q; is the corresponding value of the function at abscissa x;, then the shape functions
N; are
(Saenx (Kae X) EX)
Ni (x)=
(x2 — x1 )(%3 Sore W ageh ca eee
: : BEES)
gaat (=) Gas 2) Oe ae)

(4 — Xp (x2 — Xn or X at ad Xn)

. Here we recalculate the shape function at node | of a quadratic triangular element. From
Figure 3.8, the abscissa r; is considered as r, = 1, r2 = 1/2 and r3 = 0 at nodes 1, 4 and 6 and
nodes 2, 5 and 3 respectively.
y

Figure 3.8 Lagrange’s method for deriving shape function.

|Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements | | 61 fi

Hence from Eq. (3.25), N, can be obtained using variable r as

_@=r\2=r) |
PROS 11/2.) er) (3.26)
Similarly, for mid-side node 4, the shape function can be obtained as

ee oo) Og ee *
N= G_qay0=1/) ® ean
At node 4, the value of t = 0 and thus from Eq. (3.15), we can write that s = 1 — r. Hence
Eq. (3.27) becomes N, = 4rs. Similarly, shape functions of other nodes can be obtained.

3.4 Elemental Contributions or Element Equations

For isoparametric finite elements, interpolation of field variables such as coordinates or
displacements is evaluated by linear combination of field values at nodal points and
corresponding shape functions. In mathematical term, these relationships can be written as

n n

x(7,8)= > Njx;, y(r,8) =) Ny; X= (x yr =Ne (3.28a)

i=] i=]

n n

u(r,s)= Niu, v(r,s)= > Njv;,u={u v} =Nq (3.28b)

i=1 t=

where x; and y; are the coordinates of node i, and u; and v; are the displacements at node i along
x and y directions respectively. The coordinate and displacement vector are ec? = {x y;
X> Yr -** Xm Yn} and q’ = {u, v; uy v2 -- Uy v,} respectively. The shape function matrix is
evaluated from

Now let us assume a field variable @ which represents displacement variables, u and v.
The gradient of @ with respect to variables r and 's can be written with chain rule as

Op 199 ox | 0pdy (3.30a)

Or ox or oy or

dM 0 ox i d@ oy
(3.30b)
ds ox ds oy ds
 cs
Finite Element Method: Concepts and Applications in Geomechani

) can be written as
In matrix terminology, Equations (3.30a) and (3.30b

dp dx dy 0p
dr |_| or or ||ox
dp{ |ax ody||9P (3.3 1a)
Os Os ds ||dy

or

ag] {2
onmiis J ox
ag| ~ \9e (3.31b)
Os oy

the gradients of ~ with respect

where J is called the Jacobian matrix. We are more interested in
n coordinate system.
to variables x and y since they represent strains in Cartesia
relationship evaluates the strains and can be
From Eg. (3.31b), it is clear that the inverse
written as

99 oo
Ox & iia or

ap ap (3.32a)
dy Os
or
a9 a le a
ox Leos or |} or =
Sor dN ts)
og]
=——

detJ]ox ax ||0p , (3.32b)

dy Os dr jlds Pn
where
ox Oy dy Ox
J=det J=z——-=—
a dr Os or Os (3.33)

is termed as the Jacobian. Using Eq. (3.28), we can expand derivatives as

dx WOON; dy WON; Ox_WON; dy _QONi

or 2 ar Os 2 as? Os Sas a ear (3.34)
Hence J can be obtained as

“ ON;oe “ON; ON;

7 Reo y
n ON:Eh ior
= ) eee a ee
J=d

iS »
i=l or | i=l ds 7 » dr »| ds | Se
|Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements

The matrix dN is expressed as

aN aN, aN,
BN aor baie Or or (3.36)
Pty eae Ses sant.
Os Os Os
The parameter J is regarded as a scale factor which estimates area dxdy from drds. In
general, Jacobian is a function of r and s but for triangle it is constant and is equal to 2A. The
value of J will be positive if nodes are arranged counterclockwise as shown in Figures 3.7 and
3.8. The matrix dN contains derivatives of shape functions with respect to r and s.
From Eq. (3.32b), strains with respect to global coordinate system can be obtained as

au
ox a4 0)
m) fav
ox =
"
v2
auf =J 7 dN
Lila d Tonle?
=J dN; :
: (3.37)
dy : oy Vn

du ov du a
Recalling the strain vector ef = {E35 -Eyy rol=|a ~ art we can combine

the terms given in Eq. (3.37) and find that

Uy
oe by B2 by Dion || 1
E=7Ey ¢=| by, dyn bog barn = Bq —
Vxy bs; bz. b33 b32n Un
Vn

where

ba 5 ; = ON ;
i nosce @ is ONiee ole 5 aN, |ONG TORgh ls 2s Sones, Tt
12D 711 Os “1 | Or Ty: | Os
=| i=l

bi(2j)
=9

bo j=-1)
=9
1] |9N; JON, | ehan, |ON; Rod
by(2) aaaPea = ais a 3s for J =o fl, 2. Sie caeeny) (3.39)

i= j=]

b3(2j-1) = by(2;) (8 2p ee Pee a eA

b3(27)= 25-1)
| 64 Finite Element Method: Concepts and Applications in Geomechanics

The matrix B is called strain-displacement matrix which contains information of geometry

of the element. The matrix B is a function of natural coordinates, r and s, but for linear
triangle, it is constant. Hence 3-noded linear triangular element is called constant strain
element. The derivatives of shape functions for 3 and 6-noded elements are given in Table 33:
Table 3.3 Derivatives of shape functions for linear and quadratic
triangular elements

3-noded triangular 6-noded triangular

element element

. ON, ON, ON, ON,

or os or os
1 1 0 4r-1 0
2 0 1 0 4s—1
3 -] -l 4(r +s) -—3 4(r
+ s) —3
4 - - 4s 4r
5 = = —4s 4(1 —r—-2s)
6 = - 4(1 - 2r - s) —4r

The stress vector is then estimated from the strain vector by multiplying with constitutive
matrix as

6 =;,0, >=DBq (3.40)

Example 3.1 Find the physical coordinates x and y at r = 2/3 and s = 1/6 for the element
shown below.
My

3 £00, 2)
6
(2, 0)
(0, 0 is -

Solution: For r = 2/3 and s = 1/6, we can calculate the shape functions as:

r = 0.6667 s = 0.1667 t = 0.1667 .

Node is y N Nx Ny

1 0 0 0222 0.000 0.000

2 2 0 -~0.111 0.222 0.000
3 0 2 Ott 0.000 =
4 | 0 0.444 0.444 0.000
5 1 1 0.111 0.111 0.111
6 0 1 0.444 0.000 0.444
Sum 0.333 0.333
|Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements 65

Using Eq. (3.28a), we can find the coordinates as:

6 6
x=) N;x; =0.333, y=) N,y; = 0.333
i=] =A

Example 3.2 Find the Jacobian J for the quadratic triangle given in Example 3.1 and show
that J = 2A.

Solution:

- aN ON ON ON ON ON
os “ y or Os or # Os : or ? Os
0 1.6667 0.0000 0.0000 0.0000 0.0000 0.000
0 0.0000 0.3333 0.0000 0.0000 0.0000 0.6667
2, 1.0000 1.0000 0.0000 2.0000 2.0000 0.0000
0 13333 2.6667 123333 0.0000 0.0000 2.6667
1 —1.3333 —1.3333 -1.3333 —1.3333 —1.3333 -1.3333
DHAnkWN
OS
ON
orrK1 —2.6667 —2.6667 0.0000 —2.6667 —2.6667 0.0000

Sum 0.0000 -2.0000 -2.0000 2.0000

Thus Jacobian will be

6 6 6 6
pee els ON; yes
= U x U . _ peel . I . =4.0

Pas |
i=l]
or ap) Os | |
i=]
or ap Os s]

The area of the triangle, A = ;KOK. 2!

HenCe-ds— 2A.

3.4.1. Constant Strain Triangle (CST)

Strain inside a linear triangle having 3 nodes at its apices is constant. In other words, the
matrix B is independent of natural coordinates r and s. It can be shown from Table 3.3 and
Eqs. (3.35) and (3.39) that
Fey (3.41a)

1 1 ]
by pay 2 — y3), by =5 Sys bys = — y2) (3.41b)

1 | 1
by age? Xa), by4 SS — +4) pee er x9) (3.41c)

Hence the matrix B is constant and is only dependent on the nodal point coordinates. Thus
strain vector € = Bq is constant. This element is easy to implement in the computer procedure
 66 Finite Element Method: Concepts and Applications in Geomechanics

but care should be taken while modelling a physical system with CST elements. The accuracy
of results can only be enhanced with large number of elements.

3.4.2 Linear Strain Triangle (LST)

Quadratic triangular element is an excellent element for stress analysis and popularly used in
rock engineering problems. Strain varies linearly inside the quadratic triangle and thus it is
known as linear strain triangle or LST. It can be easily shown that the components of strain-
displacement matrix linearly vary with the natural coordinates r and s. Based on Eq. (3.39),
the component b,, is evaluated as given below:

_LJ[ RON; [OM | ON; (OM,

by 15% = ps= | = (3.42)
i=l i=l

Here, the Jacobian J = 2A and it is constant for a triangle. Using Table 3.3, we find that

6 6
ON; ON,
Dies Se aerand Seer Vera tae (3.43)
i=] i=l

Hence b,, =S-ll4r Dor — y;)] and it varies linearly with coordinate r. Similarly, it
can be shown that all other non-zero components of the B matrix are linearly dependent of
coordinates r or s.

3.5 Variational Method of Finite Element Formulation

Finite elements obtained by discretization of a continuum are formulated in general and
systematic way. The stiffness matrix and load vectors of an element can be formulated by
Rayleigh-Ritz or variational principle and Galerkin weighted residual methods. In this section,
brief description of variational principle method is presented. Galerkin weighted residual
method is explained in Appendix A.
A functional is defined by the integral expression of the differential equation that
describes the problem. In solid mechanics, total potential energy of a system expresses the
integral form of the differential equation. The total potential energy of a elastic body is the
sum of stored strain energy and work done by the external forces. For example, again consider
an axial load applied on a rock sample as shown in Figure 3.9. For simplicity, only axial
displacement and strain will be considered for this analysis. We assume that the sample is
discretized with a linear bar element having two nodes | and 2. The nodal displacements and
forces are symbolized by v,; and v3, and P, and P respectively. Then the total potential energy
of the bar element is expressed as

L\
Nig =Al 5 Oy, Ey ay — By, — Pyv> (3.44)
 Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements |[ 67 |

where A and E represent cross-sectional area and elastic modulus of the bar material,
respectively. The gradient of the displacement field expressed in Eq. (3.9) will provide the
Strain in axial direction. The axial stress O,, 18 evaluated by multiplying the strain with the
elastic modulus E of the rock. Hence, Ey anid O,, are:

poten
Ya evol
wea (3.45)

0, — EE, ——(vy —)

1 a
12
: Ji
Figure 3.9 Derivation of elemental equations in one dimension.

Replacing Eq. (3.45) into Eq. (3.44), we find

II, sais _ vi) — Pv, — Prv2 (3.46)

oT
The first expression of the right-hand side of Eq. (3.46) is the strain energy and the
negative terms denote the work done by the external force. The essence of the variational
approach is to express total potential energy of the system in terms of field quantities such as
displacements and then invoke the principle of stationary potential energy, which states that
Among all admissible configurations of a conservative system, those that satisfy the
equations of equilibrium make the potential energy stationary with respect to small
admissible variations of displacement.
In other words, the above principle can also be represented as the principle of virtual
displacements which states that the equilibrium of the body requires that for any compatible,
small virtual displacements which satisfy the essential boundary conditions imposed onto the
body, the total internal virtual work is equal to the total external virtual work. According to
this principle, @; being the ith field quantity or displacement, the virtual change in the potential
energy due to the virtual displacement 69,, will be zero. Mathematically, this relationship can
be written as

oll
A DD eed ea (3.47)
 68 4 Finite Element Method: Concepts and Applications in Geomechanics

where n = total number of degree of freedom. Since dg; # 0, then each quantity

OL hem fOl.? =i, 2 3. aeaTt (3.48)

OQ;

Equation (3.48) represents n equilibrium equations or elemental equations. Applying

stationary principle of potential energy in Eq. (3.46), we can write two equilibrium equations
as
oll, — EA
i -=——(y, Vi —v2)-F
Vo) ar =0 3.49
(3.49)

oll, EA
aePe1 @i=%,)
(si 1,) 2 P=
21> 0

The above equation can be expressed in matrix form as

FAll —1 ||see
Tew 1 |v ee NPs ow
or

Kq =F
Equation (3.50) denotes the characteristic equation of finite element. The first term of the
left hand side of the equation represents the stiffness of the element comprising geometry and
material properties. The second term denotes the displacement variables which are to be solved
simultaneously. The right hand side term signifies the nodal fofce vector which is the
summation of traction, body and point forces. In this case, only point forces are applied in the
structure. Now applying boundary condition v, = 0 in Eq. (3.50), we find that

vy,2A
=——
PyL and FP1 =-P. 2 (3-50)

3.6 Equilibrium Equations for Multi-Degree of Freedom

Figure 3.10 shows a general two-dimensional body where external forces, e.g. body force
vector f’, surface traction force vector f° and concentrated forces f', are applied. These forces
can be represented in global coordinates as

in is coe
fo = . f=, 1 ands shoe ee
4 yi bi One
Using strain energy density function as given in Eq. (1.15), the total strain energy stored
in a finite element can be estimated for the entire body of volume V as
 Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements

Lye iz odV
U=—a (3.53)

In the above equation, strain and stress vectors can be replaced by Egs. (3.38) and (3.40)
respectively, and then we find

ey
ok ety
B'DBqdV (3.54)

Figure 3.10 Loading in a two-dimensional body.

Work done by the external forces is the summation of work accumulated by body forces,
surface traction and concentrated forces as given below:

M1. = fa"N’ fav + [asNGf* ds + Sait’ bee

4 Ss I

Work done by body forces is estimated for the entire volume of the body. Work done
due to surface traction is assigned to those nodes which reside on the surface S$. The point force
f' contributes to external work only at node i.
Now, combining Eqs. (3.54) and (3.55) and applying the stationary principle of potential
energy, we can find

d11, =5(U -I1,) =0 (3.56)

The above equation is extended for a discretized finite element as

NT fav + [oatnte’ dS+ > sq t! (3.57)

; J5q™B'DBq dV + Jq'B' DBdq dV |= |Sqi
V V V S i

Since the matrix D is symmetric, both the terms in left hand side are identical and the
above expression is reduced to the following:
 Finite Element Method: Concepts and Applications in Geomechanics

6q"| [B/DBAV |q=5q" [N'faV + di [NSfds + Sidart (3.58)

V y :
Considering unit virtual displacements, 1.e. dq! #0,dq5 #0 and dq, #0, in turn at all
displacement components, we get the equilibrium equations for the finite element

[B”Bav |q = [ntebav + [Notas + Se (3.59a)

V V S i
or

Kq=F?+F°+F' =F (3.59b)
where the matrix K defines the stiffness of the element and is dependent on the natural
coordinates r and s and expressed as

K = [B'DBav (3.60)
4
The vector F is the summation of nodal body forces, surface traction forces and point
loads. Note that there are 2n number of equilibrium equations where n is the number of nodes.
Thus, the dimension of stiffness matrix and load vector will be 2n x 2n and 2n x | respectively.
The 2n equations are to be solved simultaneously using Gauss-Jordon solution scheme and this
is discussed below.

3.7 Numerical Integration Scheme

One of the important aspects of isoparametric finite element formulation is the necessity
for accurate integration of stiffness matrix, body force and surface tractions as given in
Eq. (3.59a). For isoparametric finite elements, numerical integration scheme is adopted
accurately to calculate these quantities. In general, there are two issues which arise while
selecting numerical integration scheme:
(i) What kind of integration scheme to use?
(11) What order of integration to select?
The Gauss quadrature rule is an important rule and is mostly used for finite element
analysis (Bathe, 1982). Selection of order of integration is important since higher order
integration evaluates elements more accurately but with more cost of analysis. On the other
hand, low order integration may not evaluate the elemental equations properly giving large
number of zero eigenvalues in the stiffness matrix as compared to the number of physical rigid
body modes. Hence the results will be erroneous. Thus a balance is sought to use optimum
order of integration for achieving a desired accuracy and at the same time, cost of analysis
remains moderately low.
|Chapter 3. Finite Element Method in Elasticity: Isoparametric Triangular Elements 71

3.7.1. Stiffness Matrix

In this text, the Gauss quadrature numerical integration rules are used for triangular,
quadrilateral and three-dimensional elements. Table 3.4 shows the Gauss quadrature rule for
linear and quadratic triangles.
Table 3.4 Numerical integration scheme for linear and quadratic triangles

Element type Integration Degree of _ Location of integration rands © weights

order (number precision points coordinates
of sampling
points)
3-noded 1-point (1) 1 pS We w= 1.0
triangle Ss ir
(Linear)

6-noded 3-point (3) 2 r,= 1/6 w, = 1/3

triangle ey IP) W2= Wy
(Quadratic) R=" W3= W
SS i
Sy
Sep 1B

Considering h as the uniform thickness of elements, the component of stiffness matrix

given in Eq. (3.60) can be rewritten as
P
K={B feDB dV =h|B £ DBdA=—hy
1
B! T (7;,5;)DB(j,5;)J
(7.5; W; (3.61)
V A i=l

In the above equation, p signifies the order of integration or number of sampling points.
W; represents the weight of ith sampling (Gauss) point adopted by Gauss-Legendre integration
scheme. The degree of precision expressed in Table 3.4 means the highest order of polynomial
that can be integrated precisely using the corresponding integration order. For more detailed
may refer
explanation on numerical integration scheme using Gauss-Legendre scheme, reader
to any text on numerical analysis.
elements, a multiplication factor of 1/2 is included in
Note that for triangular
the area of the reference triangle in natural coordinates is half of the
Eq. (3.61) because
J. In other words, it can be easily shown that for an undistorted triangle of unit
Jacobian
J will be 2. Using
area measured in Cartesian coordinate system, the value of Jacobian
Eq. (3.61), area of a triangle can be computed as

a=fa
- “TsYuefhs,)W,
a=l pid W,i =|
227 (3.62)
 72 | Finite Element Method: Concepts and Applications in Geomechanics |

Pp
Since
yw, =1, the area A is found as A = J/2.
i=]
The following equation shows the expanded form of Eq. (3.61) having plane strain
condition:

by (7 5;) 0 by9(7;,5;)
0 bo (%;,8;) by 1 (%,5;) l-v 4 0

ETS : ee a
Spe (1+ v)(1—2v)
FN Bint) %>5;) 0 by2n (%+5;) ream times
0 bran (%»5;) Oyan—1) G>5;) 2

11%, 5) 0 oe by2n-1) %>5;) 0

x 0 Boo (%,5;) = 0 dyn (G5) |IG5)W; (3.63)
bx(Sj) G5) Bran5i) — Panay (ri)

For linear and quadratic triangular elements with straight sides, the direct integration can
also be performed using the following relationship:

'p!
[res?aa ay (3.64)
, (a+b+2)!

where A is the area of the triangle. The coefficients a and b are any arbitrary non-negative
integers.

3.7.2 Body Force

The components of body force vector are determined in the similar fashion as mentioned in
the previous sub-section. The body force vector as given in Eq. (3.59a) can also be rewritten
for two-dimensional elements as

F,eels
= [N'fav =h[ Mink
NTEd a8
V A
(3.65)

In rock engineering problems, we mostly deal with gravity load which acts in the
vertically downward direction. If only gravity loading is assumed in the Opposite direction of
y-axis, then the vector f? becomes

gy (3.66)
|Chapter 3° Finite Element Method in Elasticity: Isoparametric Triangular Elements | [_ 73

where g is acceleration due to gravity and y is the density of the rock. Then the body force
vector expressed in Eq. (3.63) becomes

Ni Gj, 8;) 0

Poe Re ae . ae 0 teed
FPR ShYIN sO S IW = hy) : 0
| |senso
=Ey
(3.67)
a i=l) N (1; 5;) 0
0 N,,(55;)

The above equation shows that F? is a vector with 2n x 1 components.

Example 3.3 Show that body force is equally divided into nodes irrespective of the shape of
3-noded triangular element.

Solution: For 3-noded triangular element, Eq. (3.65) can be expanded as

b
Pie ;
Fely ;
Or
b
F-a5 =1 S ) yf

Foy A si ly
ns t O
is, 0
re
3y

Now, applying the direct integration rule as given in Eq. (3.64), the first row of the above
equation becomes
ZAhjg Any,
Fe h|thydA =
fo ar

(+2)! 3
A

Similarly, it can be shown that igs = fee eS Alt Ahf.

and Fi, = Fy = Fs, = ee Thus, the
body forces are distributed equally at the nodes. Similar results are also obtained with the
numerical integration scheme.

3.7.3. Traction Force for Straight-Sided Triangle

Traction or surface pressure is applied at sides of a triangular or quadrilateral element. In rock
engineering, traction pressure is applied to simulate additional gravity pressure, horizontal
stress, internal support pressure and others. Figure 3.11 shows a tunnel example for the applica-
tion of traction pressure. In the figure, traction or surface pressure is applied vertically to simulate
additional pressure. The horizontal pressure is applied on the outer boundary. For providing
the horizontal stress and internal support, pressure is given along the tunnel boundary.
 Finite Element Method: Concepts and Applications in Geomechanics

4
5C|
m
ue
{\ TAF
4
a AAS
AS
AV
, {NZ

KA
KOO
PKA
IN is
ale.

Figure 3.11 Application of traction loading.

The applied traction pressure on the element boundaries is converted into nodal forces
before solutions of primary variables are obtained using Eq. (3.59b). Let us consider the
following linear triangular element, where x-directional traction pressure acts on the surface
2-3 as shown in Figure 3.12. The value of traction pressure at nodes 2 and 3 are P and P3
respectively.

Figure 3.12 Distribution of surface loading into nodes.

We denote the surface or side where traction pressure is applied as $2—3. Thus the traction
pressure vector f° on the surface S2-3 can be written as
: 52-3

f -|is; | (3.68)

The interpolation function fae can be obtained from the nodal point pressures as

52-3 = NP» qe N3P3 = SP, ote ad — s)P3 = P; at 5(P, a P;)

x (3.69)

The traction force term in Eq. (3.59a) can be now expanded using Eg. (3.69) as
|Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements

S
Fy Kadi
Fiy 0 A,
P= ee
ss =[N Tasf°dS= | N>cae0 {P, + s(P, — P;)}ds (3.70)
2y S $2-3

Ss 0 N;

Since the traction pressure is applied only on the surface 2-3 (r = 0), the value of N,
= r = 0. Thus there will be no traction force component at node 1 and hence FE =(. In this
case, we will propose for direct integration scheme to obtain the values of integrals expressed
in Eq. (3.70). For straight-sided triangles, direction line integrals are obtained using the
following equation:

ab a'b!
r¢s°dS = § ———_
J (a+b+})! (3.71)
S

where S is the length of the side. The parameters a and b are non-negative integers. Using the
above equation, we can estimate traction force at node 2 from Eq. (3.70) as

Foe } N, fs22dS= } s{P, +(P, — P;)s}dS

S2-3 S2-3

Ps!) (P,—P,)2!| hb
tiles (42)+ acd - meray + Ps) (3.72)

The term /,_3 is referred to the side length from nodes 2 to 3. The parameter h denotes
the uniform thickness of the element. Similarly, other components of traction force vector can
be evaluated. Again, note that the dimension of surface traction vector is 2n x 1.

3.7.4 General Formulation of Traction Force for Curved/Straight-

Sided Triangle
If the side of the triangle, where traction pressure is applied, is not straight, then we cannot
apply the direct integration scheme to evaluate nodal traction force vector. In such case, nodal
traction force vector is formulated based on numerical integration scheme. Consider that
traction pressures are applied on surface S of a triangular element. The interpolation functions
of traction pressure in x and y directions on the surface S$ will be

fe ee yN (PL) fi = VN (P,); (3.73)

i= i=]
 anics
| 76 Pi i Finite Element Method: Concepts and Applications in Geomech

denote the nodal values

where n is the number of nodes lying on the surface S. (P,); and (P,);
on the surface S is
of traction pressure at ith node. Interpolation function of coordinates
obtained based on
n n
SPS
we Nr y* =r Vj (3.74)
=I t=

Jacobian J is
where ce and y> are ith nodal point coordinates lying on the surface S. The
then estimated from

s ax> ; oy? site

det J~ = (= (2 (3.79)

where k represents natural coordinate variables r or s depending on the surface S. For two
dimensional elements, the surface S represents a curve (or line) segment having a uniform
thickness h.
The components of force vector at jth node is then estimated from

I< ahh
5 and R; = Ne
2S det J;
S (3.76)
F? =

t=!

where m = number of sampling (Gauss) points taken along the surface S, a; = weight of
sampling point at ith location,f° = force function expressed in Eq. (3.73), N/ = value of shape
function of jth node on the surface § at ith sampling point. The multiplication factor
1/2 is applied only for triangular elements, the reason of which was mentioned before. For
quadrilateral elements (Chapter 4), this multiplication factor will be omitted.
In this case, numerical integration is performed along a curved line with suitable number
of sampling (Gauss) points. However, for quadratic elements two Gauss points are sufficient
for accurate estimation of nodal traction forces. Consider a curve segment ranging from 0 to
1 in natural coordinate system as shown in Figure 3.13. The value of sampling points are
mapped from (—1, 1) coordinates to (0, 1) coordinates as shown in the figure. The weights of
the sampling points a; = 1 (for i = 1, 2).

Curve § 1
1
(J3 +1)/2V3
WV3

x V3 2
(J3 -1)/2V3
x 0
—1
Figure 3.13 Numerical integration for evaluating traction force vector.
| Chapter 3_ Finite Element Method in Elasticity: Isoparametric Triangular Elements | 77

Let us again consider the example as given in Figure 3.12 in the previous sub-section.
The surface S is denoted by nodes 2 and 3 having shape functions N, = s and N3 = 1 -—s. The
Jacobian J is obtained from

ays .
ox = x,
— =y, —y3; and J= (x. — x3)? + (y, —*y,)*)'?=
Os econ. h_3 (3.77)

Thus, the traction force component at node 2 can be estimated using Eq. (3.76) as

eahb || J3=1 P; + (Py - P. V3-1 + V3 +1 P; +(P) —P. V3 +1

as =|(Selave- Seder re)fa aejata) eabebn Panis Oafs
(3.78)

Note that the value is exactly same as given in Eq. (3.72). The versatility of this method
lies in handling curved surfaces and application of numerical integration scheme which is
already programmed for determining stiffness matrix and body forces. Thus matrices and
vectors necessary for Eq. (3.59b) are evaluated using a similar method.

3.7.5 Point Force

The concentrated force is directly applied to nodes and thus added to the global force vector
after decomposing it into x and y directions. The unit of the x and y components of the vector
f' in Eq. (3.59a) is Newton or kgf or Ib for the ith node.

3.8 Assembly into Global Simultaneous Unreduced

Equations
Once the elemental stiffness matrices are formulated using Eq. (3.61), they are assembled into
a global stiffness matrix to generate a set of simultaneous equations of all degrees of freedom.
For two-dimensional analysis, the dimension of global stiffness matrix is 2n x 2n, n being the
total number of nodes in the model. In mathematical term, the global simultaneous equations
are written as
Kq° =F (ao79)
where
m

K® is the global stiffness matrix (= SKS; m being the number of elements)

iat
q® is the global displacement vector = {u, Vv; -*- Uy vet

F® is the global force vector = {Fi, Fiy «Fax Fny}"

Consider Figure 3.14 having two 3-noded elements with the following elemental
equations:
 Finite Element Method: Concepts and Applications in Geomechanics

Fiy kop ka3 ag kas kn | [M1

Fy k33 k3q kgs hg ||2
1: = (3.80a)
Element
Foy kag kas Kae ||¥2
Fax Sym. ks5ks6 ||4a
Fay kee | (Ya

Foy bo b3 ba bs b6 || 2
P3, bh; by bs be | |43
Element 2: = (3.80b)
Fy lyg I45 Ugg ||¥3
Fay Sym. l55 l56 U4

Fay l66 | (Ya

Figure 3.14 Assembly of elemental stiffness matrices into global matrix.

Here, each node has 2 degrees of freedom, i.e. displacements along x and y directions at
each node. Then the global unreduced simultaneous equations will be in the form of

as ki Ky ky3 kig 0 60 kis ki6 uy

Ay kon kg ko4 0 0 kos k6 yy
Fox k33+hy keg tho 43 a kes ths ko the |)
3 Wie kag thy bg log Kas ths kag +126 |}v2
= (3.81)
F3; bz |g hs lg U3
Fy Sym Iyg gs l46 ¥3
Fi, ksstlss ks6 +156 ||Mg
 Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements

Or

FS = K%q%
Note that since nodes2 and 4 are connected to both the elements
, stiffness components
are added into the global stiffness matrix. The matrix K® is also symmetri
c with non-negative
values at the diagonals. This matrix is known as global unreduced stiffness
matrix since the
boundary conditions are yet to apply into the system of equations.

3.9 Application of Essential Boundary Conditions

Essential boundary conditions are generally refers to displacement boundary condition
at the
boundary of the model. As mentioned in Figure 3.1, q = q* at boundary I, represents
the
essential boundary condition. There are also non-essential or natural boundary conditions
mostly dealing with force or stress boundary conditions at designated part of the model. For
example, radial stress or force at excavation boundary must be zero. However, finite element
equations do not generally require enforcing non-essential boundary conditions. These
boundary conditions are imbedded into the equilibrium equations. On the contrary, essential
boundary conditions must be enforced for proper convergence of the solution.
In finite element models, we apply displacement boundary conditions at some specified
nodes depending on the problem. Nodal displacements in the x- and/or y-directions can be set
to zero or any specified value. Again revisit Figure 3.14 and see that node 1 is constrained in
both x- and y-directions and node 4 is fixed only in the x-direction. In order to apply these
boundary conditions in the global system of equations [Eq. (3.81)], the following two methods
can be adopted for easy implementation in computer programs.

3.9.1 Penalty Method

In this method, a large number, @ (at least 10 orders higher than the highest value of the
stiffness matrix) is added to the diagonal position of the constrained degree of freedom of the
global unreduced stiffness matrix. The force value of the specified constrained degree of
freedom is set to au*, u* being the specified displacement at the node. The idea is to make
an equation like (B+a@)u+6 = au’, where B and 6 are insignificant as compared to a. Thus,
after applying boundary conditions at nodes 1 and 4 for the above example, the global stiffness
matrix and force vector will be as follows:

ou ky ta ky ky3 kg Py kis ki6 uy

avy kop +O kg koq 0 0 kos ky¢ yy
Foy Kez thy hea tho fy ha bs ths hp the ||m
Foy |_ Kay tly7 yx lng kas ths kag the |}v2
je bz bya bs 6 U3
Fyy 44 l45 lag ¥3
ou, kss tls5 +O ksg +156 ||ug
Fy kee + loo | \%4
 cs
| Finite Element Method: Concepts and Applications in Geomechani

a large number © is added to the

Since displacements uj, v, and ug are constrained, ed
ess matrix. The force vector is also chang
corresponding diagonal terms of the stiffn
accordingly to equate the following:
*
* *

to the enforcement of boundary conditions,

In the above example, uy = V1 = U4 =(). Due
.
in this case, another 5 equations are to be solved
the number of solvable equations is reduced; as
of boundary conditions, is termed
That’s why the stiffness matrix, after the application
is one of the easiest techniques for the
global reduced stiffness matrix. The penalty method
ter programs. The accuracy of the solution
application of essential boundary condition in compu
depends on the value of a and computer capacity.

3.9.2 Row and Column Adjustment Method

directly imposed in the global stiffness
In this technique, displacement boundary conditions are
n u = u. by adjusting the rows
matrix and force vector. The idea is to make a trivial equatio
above example, imposing u, = Uy
and columns of stiffness matrix and force vector. For the
can be performed in the following way:

[5 , : 1 0 0 0 0 0 0 0 uy
es u
O ky = ka Kyq Carry ks kx6 vy
Lane? :
h3 lig k35 thes k36 the uz
Fy, — ky 0 k33 +1 k34 +l

kya tha 3 log kas ths kao + he |}¥2 (3.84)

Fo, hay |ol
Fi; 0 bz bg bs ls6 U3
Sym. 44 I45 l46 V3
F3y 0
0) kee55 +l BS) kee56 +1 56 u4
: ,
Fyy Pa kis

ys kee + loo | \%4

Fyy — kyo

second
Note that since u; = u,, the stiffness component k3, = kj2 has to be adjusted for the
force vectors are
equation in the force vector as Fj, — Riou Similarly, other components of
boundary condition v; = v;, the stiffness and force
adjusted. Now, if we have to impose the
vector will be adjusted again as given below:

; io. OC 0 0 0 0 0 uy
a: 0 4 0 0 0 0 0 0 v1
Fo, — ky —koa) 10 0 keg thy Kea tho ba ba kas ths hse the ||
Fyy —kygth —kay | _]O 0 kag tho to3 bog kas +s kao +6 |)¥2 |(3.85)
F;, 0 0 bz by4 ls lo ie
Fyy 0 O Sym. l44 las l46 V3

2a ae 0 0 ks5tls5 ks6 +156 ||ua

0 0 ke + l66 || V4
+ * *
Fay = kyo — ko6Y
 | Chapter 3 Finite Element Method in Elasticity: Isoparametri
c Triangular Elements | [ 81 f
Similarly, boundary condition U4 = uz can also be impose
d. The stiffness matrix remains
Symmetric and banded with zeros in several rows
and columns except 1’s in the diagonal
positions of the prescribed degrees of freedom. This
method is also easy for computer
implementation.

3.10 Solutions of Primary Variables

The most effective and direct solution technique of linear
simultaneous equations is Gauss
elimination and back substitution method. However, applica
tion of this method in solving
primary variables such as displacements mainly depends on
the properties of reduced stiffness
matrix—positive definiteness, symmetry, and bandedness. The
global reduced stiffness matrix
possesses all of these three properties. For detailed explanation
of Gauss elimination and back
Substitution method, reader can refer to excellent texts in numeric
al methods (Kreyszig, 2002;
Press et al., 2002).
The most effective way of storing a large symmetrical stiffness
matrix is by Skyline
method. In this method, stiffness matrix is stored in a single array
based on the connectivity
of elements. Explanation of this method is given in Appendix B.
For more information
regarding computer storage and finite element solver, reader can refer
to the excellent
references (Bathe, 1982: Zeinkiewicz, 1979; Krishnamurthy, 1987; Cook et al., 1989:
Chandrupatla et al., 2002). A C code developed by the author for solving
displacements from
single array stiffness matrix generated using Skyline method and force vector
is shown below:
A C code for solving displacements using Gauss elimination method
void SolvePrimary (double al[],double rhs[])
/*Getting the Solutions for Displacement */
{
sLighe jr,jd,is,jh,ih,id,ie,ir,k,kkk,kk1,neq, flagl, flag2,p;
double agency,c,d;
/*rhs=Force Vector on Calling, Solution Vector on Return.
pt-Clobal Stiffness) Matrix) «om (Calling; Upper diagonal matrix on
Return. neq=Number of Equations. node _num=number of nodes.
dof=degree of freedom=2. jdiag(j)=Array of the position number
of jth column diagonal in global K matrix formed as al[]
ety me eee draOy g=1) "4 aiag(T)=3'r" ebat! */
num;
neq=dof*node _
<0);
agenicy=0
jr=0;
Lor aij=lyj<=sneq;j++) {
jd=jdiag{[j-1); jh=jd-jr; is=3}-7jh+2;
Det jh=2) >0)}-4
peal. kadrs2. id=jdiagilis-1-1];
FOR Gate -ie=ie tits) if
Tra. 0= diag [i-1] ;
iad te ote (t-is41)) {ih=id-ir-ay
else {ih=i-is+1;}
Pies cihso) <4
 and Applications in Geomechanics
Finite Element Method: Concepts

kkk=k-ih; kkisid-ih; c=0;

; p++) c+=al [kkk+p-1] *al [kk1+p-1];
for (p=0 ;p<= ih-1
aul [ikea =e;
}
ke,
}// end i loop
flagl=1;
}
if ((jh-2) ==0||f£lagl==1) {
Lr=jrtl; 16=j0-1; k=j -jd;
for +4)
(i=if;i<=16,1 {
kkk=k+i; id=jdiag[kkk-1];
if (al [id-1] !=0.0) {
deal[i-1]; at(i-1)=a1[i-1]/alfid-1];
ai [jd-1] -=d*all[i-1];

flag2=1;
}
ve (£lagi==1|'|flag2==1)"4
C=0F
c+=al [jr+1+p-1] *rhs [is-1+p-1];
for (p=0;p<=jh-2;p++)
rhs [j-1] -=c;
}
jr=jd;
t1// Pend j 7 (to0op
for i++)
(i=1;i<=neq; {
id=jdiag[i-1];
if Cattiid-=1) 107.0) rhs [i-1]=rhs[i-1]/a1[id-1];
agency+=rhs [i-1]*rhs[i-1]*a1[id-1];
} //end i loop
jd=jdiag[j-1]; d=rhs[j-1]; j-=1;
J =nedy
while (j>0) {
jr=jdiag[j-1];
vEeC(jd-jyrr or) {
is=j-jd+jr+2; k=jr-is+1;
for (i=is;i<=j;it+) {
kkk=i+k;
]=
[i-1]
rhs [i-1 s *d;
[kkk-1]
-alrh
}
}
jd=jr;
d=rhs[j-1];
a=
} // end while loop
} // end subroutine
|Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements | | 83 |

3.11 Determination of Strains, Stresses and Reaction

Forces
Solution of secondary variables such as reaction forces, strains and stresses at each element is
calculated after primary variables such as displacements are solved. In the following sub-
sections, methods are explained for calculation of elemental strains and stresses and nodal
reaction force vector.

3.11.1 Calculation of Strain and Stress Vector for Each Element

Once the global displacement vector is evaluated, nodal displacement vector for each element
is extracted separately. Let us denote a nodal displacement vector as q¢ for eth element. Initially
strains and stresses are calculated at each sampling point of an element. It is found that
stresses especially the shear stress are more accurate at the sampling points rather than at the
element centroid. Strain vector is calculated from the displacement vector q® and the matrix
B of the element calculated at a sampling point. In mathematical term, this relationship can
be expressed as:

ef =Béq’, i=l,...,m (3.86)

where the subscript i represents sampling point and m denotes the number of sampling points
within the element. Stress vector at each sampling point is estimated by simply multiplying the
Strain vector with constitutive matrix as given below:

oO; =D°¢;, eee rry (3.87)

The constitute matrix D® is selected based on the plane stress or plane strain analysis. In
this fashion, strain and stress vector at each sampling point of every element can be obtained.
It is a common practice to make average of strain and stress vectors of sampling points to
provide a single vector of strains and stresses for the element. For some elements such as linear
and quadratic triangular, linear quadrilateral (Chapter 4) and other elements, the average strain
and stress values represent the strains and stresses at the element centroid.

3.11.2 Calculation of Reaction Forces

Once the displacements of all nodes are obtained, reaction forces at each node can also be
estimated. In finite element analysis, accuracy of estimation is found to be the highest for
displacement calculations followed by reaction forces, strains and then stresses. Global reaction
force vector R* can be calculated from the global unreduced stiffness matrix and displacement
vector as given below:

R’ = Ke eaucea (3.88)

The above approach is economical probably only for linear triangle with fewer degrees
of freedom. For higher degrees of freedom, the reaction force vector is calculated for each
element as given below:

R SI h m

ava Derma 8) (3.89)

V =
FF 84 | Finite Element Method: Concepts and Applications in Geomechanics

where m denotes the number of sampling points within the element. Once all individual force
vectors are obtained from all the elements, they are added together based on degree of freedom
for evaluating the global reaction force vector R°.

3.12 Stresses at Nodes

Stresses at any point inside an element are estimated by interpolating the stresses obtained at
the sampling points. For linear triangle, the stresses are constant at every point including
nodal points and they are equal to the values obtained at the centroid, i.e. r= 5s = ¢ = 1/3. For
quadratic triangle, stresses are obtained in three sampling points denoted by 1, 2 and 3 as shown
in Figure 3.15. Points denoted by A, B, C, D, E and F represent nodes. Let us assume that
the coordinate system defined by r-s represents the triangular element and the same defined by
n—-& signifies the coordinate system of the triangle formed by points 1, 2, and 3. From Figure
3.15, the relationship between these two coordinate systems will be
1 1
Ea vat (3.90)

Figure 3.15 Extrapolation of stresses into nodes from sampling points.

The stress vector at any, point P inside the element can be interpolated using the stress
vectors obtained at sampling points as
3

Op = NG; (3.91)
i=)

where N; = r, Ny = s, and N; = 1 — r—s and o; represents the stress vector with components
Oxx, Gy, and T,, at ith sampling point. For example, x-directional stress (6,,) at point A can be
obtained by replacing € = 1 and 7 = 0 or r = 5/3 and s = -1/3 in Eq. (3.91) and so on. Hence
the nodal stress of node A will be:

1 ]
Orr ~ 3 Ox ~ 302 ~ 3 O03 (3.92)

The interpolation matrix representing transformation of stress vector from sampling

points to nodal points for 3-noded and 6-noded triangular elements are given in Eqs’ (3.93)
and (3.94) respectively.

(3,95)
 Chapter 3_ Finite Element Method in Elasticity: Isoparametric Triangular Elements | | 85

where © represents the stress vector with components 0,,, O,, and T,, at the centroid and oy,
denote stress vector at node A and so on.

te ennai !
3 3 3
ay 5 I
Oa 2 S 3
oO
ve at ies 5 B
SEN eal= ae 3 3 O07
Op 42 aZ Sere1 pe
Or 3 3 3 (3.94)
Or ae 2 2
3 3 3
a ek 2
2 3 3

In Eq. (3.94), stress vectors at sampling points are represented by numerical subscript and
that of nodes are denoted by alphabets. If a node is connected with more than one element then
nodal stress vector is the average of all the stress vectors estimated at the node from the
interpolation results of each element.

3.13 Interpretation of Results

Comparison of finite element results with measurements from physical system is required to
validate the model being analyzed. Validation of results can be made with analytical or closed
form solutions or actual measurements of displacements, loads and stresses in the field. If the
results from finite element models are not satisfactory, model parameters such as number of
elements, types of element, degree of freedom and even elemental equations can be changed.
In general, the following considerations must be made before interpreting results:
Geometry of the physical system and that of finite element model is identical.
2. Material properties used in the constitutive matrix are equivalent to that of field
parameters.
3. External loads applied to the model are identical in magnitude and direction with the
physical system.
Boundary effects of finite element model are marginalized.
5. Sufficient elements are generated to analyze displacements and stresses near the
excavation boundaries.
 Finite Element Method: Concepts and Applications in Geomechanics |

3.14 Tunnel Example

In this section, application of finite element method is demonstrated for analyzing a circular
tunnel section. This example is repeated in most of the subsequent chapters to compare the
results. Apart from tunnel problems, applications of finite element method are widespread in
geotechnical engineering problems such as slope stability, mine pillar design, dam design,
subsidence, pile design and others.

Statement of the Tunnel Problem

A circular tunnel of diameter 4 m is excavated in sandstone rock strata having elastic modulus
of 5.5 GPa and Poisson’s ratio of 0.25. The uniform far field stress in the area is measured
to be 8 MPa. Combination of bolting and shotcreting is applied along the boundary of the
tunnel for providing an internal pressure of 2 MPa. Under these conditions, it is required to
estimate the stresses and deformation along the boundary of the tunnel.

3.14.1 Closed-form Solution

The analytical solutions for stresses and deformation of circular tunnel excavated in rock strata
were given by Kirsh (1898) (Brady and Brown, 1985). For the far field stresses of p, and kp,,
the solutions of tangential (Ogg), radial (0,,) and shear stresses (0,9) are reproduced in
Eq. (3.95) (Figure 3.16). The radius of the runnel is a and radial distance is denoted by r.

2 2 4
CO, “2 va) “1 45 4a (3.95a)
r r

Pe +

es ———
nd ————
— Ss
re Pin ats

ipa > + hp,

a gee
—EE ——

ert yak

—— ees

Se Petes

Figure 3.16 Stress distribution around a circular opening.

 |Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements | [ 87 |

D>
Leia (1+k)| 1+a’ ]+(1-&) 13a‘ cos 20 (3.95b)
= ie r

Di 2a wea, |
Crt aes (1-k) a sin 20 (3.95c)
io OP |
The radial and tangential displacements are obtained for plane strain condition
s as

2 2
ie_ Po
em c+k)-(1- peu—2v)+ {|
2 Cos 2 (3.96a)

Ug = Dot
aes (1-k) a lana
201-20)045 =
(3.96b)

where

G = Shear modulus =
2(1+v)
Vv = Poisson’s ratio

In the tunnel problem discussed above, the value of horizontal stress to vertical stress ratio
(k) is assumed .to be unity. Since internal pressure (p,) is applied on the tunnel boundary, the
equations of tangential and radial stresses are
a

Or, = Pi + (Po — Pi)| 1-> (3.97a)

i

D
a
O99 = Pi + (Po - raS) (3.97b)
r

The solution of radial displacement becomes

u, =— kppeipan
ype (3.98)

The results of Eqs. (3.97) are plotted with respect to radial distance in Figure 3.17 and
termed “analytical”. It can be seen that tangential stress at the boundary of the tunnel
is 1.75 times of far field stress and gradually decreases with radial distance away from the
tunnel. The radial stress on the boundary is the same as internal pressure of 2 MPa. The plot
of radial deformation is shown in Figure 3.18. This figure shows the maximum radial
deformation occurs at the tunnel boundary.
 Finite Element Method: Concepts and Applications in Geomechanics

oO@/P,
or/P,
or 6& Radial stress
0.4 A,
0.2 | Analytical A Linear © Quadratic
0 T 7 =a = T =
] cs 2 BS, 3 35 4
rla
Figure 3.17 Tangential and radial stresses with respect to radial distance.

0.003

0.0025

0.002

(m)
—u 0.015

0.001

0.0005 |
| Analytical a Linear oO Quadratic |
0 San Ti oe

] eS) 2 PBS) 3 3.5 4

rla
Figure 3.18 Radial deformation with respect to radial distance.

3.14.2 Finite Element Solution

In order to solve the tunnel problem mentioned above, a quarter of tunnel section is analysed
using both linear and quadratic triangular elements. Since the tunnel geometry is symmetrical,
a quarter of tunnel section is sufficient for the analysis. The boundary conditions and external
loading are shown in the model mesh (Figure 3.19). Internal pressure of 2 MPa is applied all
along the tunnel boundary. The total radial distance of 40 m is taken from the centre of the
tunnel. The models are developed with 264 and 987 nodes for linear and quadratic triangular
elements respectively. Both the models consist of 460 elements. Near the boundary of the
tunnel, small size elements are developed to obtain more accurate stress and displacement
distributions. Large size elements are developed away from the boundary of the tunnel.
| Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements | [ 89 |

RNY
NYWays (
4
AL
VVAVAVAVA
Figure 3.19 A quarter section of tunnel section.

The numerical modelling results are also plotted in Figures 3.17 and 3.18. It can be seen
that for quadratic triangles, numerical results closely match with the analytical solution
mentioned above. For linear or 3-noded triangular elements, there is little discrepancy with the
analytical solution. However, if more number of elements is generated, this difference would
have been overcome. This example proves that finite element method is applicable for stress
and displacement analysis of circular tunnel. It can be inferred that this technique can be used
and is being used for any shape and size of the tunnels and other geo-structures.

SUMMARY
In this chapter, general concept of isoparametric finite element formulation is given for
triangular elements. The concept of local coordinate system is given to evaluate shape functions
at nodal points. Development of elemental equations based on Rayleigh-Ritz or Variational
Principle is explained with suitable example. Numerical integration scheme for linear as well
as quadratic triangular elements is illustrated giving examples for estimation of traction and
body force vector. The concept of global assembly of elemental equations is shown along with
application of essential boundary conditions. Stress transformation matrix from sampling points
to nodal points is also illustrated for both linear and quadratic elements. Finally, a tunnel
problem is modelled with linear and quadratic triangular elements and comparison of results
is made with the closed-form Kirsh’s solutions.
90 Finite Element Method: Concepts and Applications in Geomechanics

EXERCISES

al (a) Compute the integral / = h] N{NaN3dA, where N; are the shape functions for a
A
linear triangular element. hay
(b) Determine the similar integral considering shape function of a quadratic triangular
element.
ee Determine the shape function of a quadratic bar element as shown in Figure P3.1.
QD -———_—__{> |_

] 2, 3
<> <.—_____—_>
L/2 L/2
Figure P3.1

ATS) Derive shape function of cubic triangular element based on serendipity and
Lagrange’s methods.
3.4 For a linear triangular element shown in Figure P3.2, obtain matrix B and also
determine the strain vector €.
ey
gq’ = {0 0 0.008 -0.01 0 -0.005}

3 G,4)

Figure P3.2

3.5 Shape functions N and N> for a point P located inside the linear triangle (Figure P3.2)
are 0.2 and 0.3 respectively. Determine the x and y coordinates of point P.
3.6 Assemble the body force vector (F), , , for f° = x and ies = y” for the two linear
triangular elements as shown in Figure P3.3.

y 4 (1,4) 3 (4,4)

Figure P3.3
| Chapter 3 Finite Element Method in Elasticity: Isoparametric Triangular Elements | | 91 |

37 For 3-noded triangular elements, determine the nodal forces at nodes 3, 6 and 9 for
the linearly distributed pressure load as shown in Figure P3.4.
3.8 Determine nodal forces at nodes 3, 12, 6, 15 and 9 for quadratic triangular elements
for configuration as shown in Figure P3.4.

Figure P3.4

8) Show that the gravity load component at corner node of a quadratic triangular
element is zero.

3.10 Solve Problem 3.5 considering quadratic triangular element.

3.11 For a linear isosceles triangle of side a, show that if one side is parallel to any of
the axis x or y, the strain displacement matrix will be the only function of a.

3.12 A quarter section of a circular tunnel is analyzed using linear triangular element as
shown in Figure P3.5. The element 1 connected by nodes 1, 12, and 13 whose
coordinates are as follows:
Node x (m) y (m) u (mm) v (mm)
\ Z, 0 —4.482 0
12 2225 0 4.370 0
13 Ze2250 0.35198 4.303 —0.691

The displacements at each nodes are also given in the above table. Determine the
stress concentration factor at element 1. Assume E = 5.5 GPa, and v = 0.25. A
uniform pre-mining stress (p,) of 8 MPa is applied on the outer boundary of the
model (not shown in the figure). Ignore internal pressure p;.

Figure P3.5
4.1. Introduction
The basic concept of isoparametric finite element formulation is described in Chapter 3. In that
chapter, general finite element formulation is explained in detail giving emphasis on triangular
elements. In this chapter, we will apply similar concepts to formulate quadrilateral elements.
The major differences in the formulation of triangular and quadrilateral elements are number
of nodes per element, nodal shape functions and application of numerical integration scheme.
The polynomial functions that define field quantities such as displacement and coordinates are
assumed based on Pascal triangle with suitable adjustments. Hence shape functions for
quadrilateral elements are to be different from those of triangular elements. This chapter mainly
deals with the isoparametric finite element formulation of quadrilateral elements emphasizing
the differences with triangular elements.
Here we will formulate quadrilateral elements having 4 (linear) or 8 (quadratic) nodes
located at boundaries. Quadratic Lagrange element having 8 boundary nodes and 1 inside node
will also be formulated. Natural coordinate system will be assumed to evaluate shape functions
for quadrilateral elements. Determination of stiffness matrix, traction and body force vector
will be explained with the help of numerical integration scheme. Finally, interpolation matrix
for transforming stress vector from sampling points to nodal points will be elaborated for both
linear and quadratic quadrilateral elements.

4.2 Discretization with Quadrilateral Elements

A two-dimensional body has to be discretized into small quadrilateral regions before finite
element analysis can be performed. These small quadrilateral regions one termed as
quadrilateral elements which contain nodes at corners, mid-side and may even inside the
element. In this text, we will mainly focus on quadrilateral elements with 4, 8 or.9 nodes as
shown in Figure 4.1. A linear quadrilateral element contains four nodes at each corner.
The
elements, whose nodes are located at corners and may be on sides, are called serendipity
elements. These elements are widely used for two-dimensional stress analysis in rock
| Chapter 4 Quadrilateral Finite Elements | 93

engineering problems. For quadratic elements, node on the side is located at the mid-point
between the two vertices. The Lagrange elements are defined when at least one of the nodes
is located inside the element as shown for a 9-noded quadratic element in Figure 4.1. In this

ae
case, the inside node is located at the centroid of the quadrilateral.

4-noded linear 8-noded quadratic 9-noded quadratic

(Serendipity) (Serendipity) (Lagrange)
Figure 4.1 Commonly used quadrilateral elements.

Cubic serendipity and Lagrange elements contain 12 and 16 nodes respectively as shown
in Figure 4.2. Except corner nodes, two additional side nodes are located at the 25 and 75
quartile locations of each side. Accordingly, for a Lagrange element, four additional nodes are
located inside the element at 25 and 75 quartile locations from sides.

12-noded cubic 16-noded cubic

(Serendipity) (Lagrange)

Figure 4.2 Cubic quadrilateral elements.

A quadrilateral element will be considered bad-shaped if any internal angle is less that
30° and more than 120°. In finite element analysis, we would like to avoid using such elements
since calculations of strains and stresses at sampling points are inaccurate.

4.3 Natural Coordinates for Quadrilateral Element

For quadrilateral elements, natural coordinate system is defined using a square-shaped “parent”
element. The sides of the square range from —1 to 1 in both r and s directions as shown in
Figure 4.3. Table 4.1 describes the natural coordinates of nodal points in parent element for
4-, 8- and 9-noded elements. Derivation of elemental equations is performed based on the

1 5 2
Parent element

Physical element
Figure 4.3 Comparison between physical and parent elements.
 Finite Element Method: Concepts and Applications in Geomechanics

parent element or using natural coordinates r and s. Then physical element is mapped with the
parent element and numerical integration is performed for evaluating stiffness matrix and force
vector.
Table 4.1 Natural coordinates of nodes in parent element

Node no. (i) I; Sj

1] —l —1
2 ] =I
5 1 1
4 -l 1
5 0 -1
6 1 0
7 0 1
8 —1 0
9 0 0

For serendipity elements, displacement field is expressed using a polynomial function of

order p leaving out the middle of Pascal triangle as shown in Figure 4.4. But for Lagrange
elements, square box terms as shown by dotted lines must also be added to the polynomial
equation. Thus Lagrange elements are more accurate than serendipity elements. For example,
the polynomial functions for quadratic serendipity and Lagrange elements will be

Qs (7,5) =a, + ar +azs+ayrst+ asr? + agr’s 23ayrs* r ays? (4.1a)

QL (7,8) =Qs(r,5) + agr?s* (4.1b)

where a; are the constants to be determined from the nodal point values and the subscripts S
and L refer to serendipity and Lagrange elements respectively.

Linear

Quadratic

N < * F 2 a

Seas
an &
hae
as a“

Serendipity

RMagtanges == So ee

Figure 4.4 Pascal triangle for quadrilateral elements.

 Chapter 4 Quadrilateral Finite Elements =i [ 95

4.4 Determination of Shape Functions

The definition of shape function is given in Chapter 3. The properties of shape functions as
described in that chapter remains valid for quadrilateral elements too. In the following sub-
sections, mathematical derivation of shape functions for quadrilateral elements are given using
direct, serendipity and Lagrange methods.

4.4.1. Direct Method

For linear quadrilateral element, the polynomial function of field quantity such as displacement
is expanded as
QO =a, +ayr+aj3s
+ agrs (4.2)

Imposing nodal point values of natural coordinates into the equation, we get

je a a 1] (a,
Q| |1 i pet Eli la,
“Vt 1 1 1} |a, (4.3)
D4 1 —l 1 —] ag

By inverting the matrix, we can rewrite the above expression as

QQ 1 1 1 1 QD

a) Wel 1 1 -1]|@
Aen Hy —109 7 1|] 0; (4.4)
a4 1 —l 1 —] D4

The expressions of a; (i = 1, ..., 4) are inserted into Eq. (4.2) and the polynomial
function is expressed using nodal point values as

P= Ni + N2Q2 + N33 + N4Gq (4.5)

where N; =F(+ and +55), values of r; and s; for ith node is given in Table 4.1

Application of direct method for higher order elements is cumbersome and hence we will
apply other methods to evaluate their shape functions.

4.4.2 Serendipity Method

This method is very useful for higher order serendipity elements. Using the similar concept as
described in sub-section 3.3.2 under serendipity method, shape functions for quadratic
quadrilateral element can be evaluated with a few steps. Figure 4.5 shows the equations of lines
for an 8-noded quadrilateral element. From the figure, the shape function of node 1 (Nj) can
be obtained as
 96 | | Finite Element Method: Concepts and Applications in Geomechanics

N; ='a(1—r)d
='s)0 +r +'s) (4.6)

Replacing r = s = —1 for node | and equating Eq. (4.6) to unity, we find a = —1/4. Hence
the shape function becomes

N, == 2(==) rs) (4.7)

Similarly shape functions for all other nodes can be evaluated as summarized below:

N,t Ser
4
aera
l U 2
for corner nodes 1, 2, 3 and 4 —_ (4.8a)

2
N; =F A+ gd - s+
? + 5-17) for mid-side nodes 5, 6, 7 and 8 (4.8b)

l+r+s=0

l+r—-s=0

ce 7 0 aya (),

Figure 4.5 Derivation of shape function.

4.4.3 Lagrange’s Interpolation Method

One-dimensional Lagrange’s interpolation function as described in sub-section 3.3.2 under
Lagrange’s interpolation method is extended into two dimensions for the evaluation of shape
functions for quadratic quadrilateral element. The idea is to first apply Lagrange’s interpolation
method to obtain functions along s-direction, keeping the coordinate r constant. Then we again
apply Lagrange’s interpolation method on the new functions along r-direction. Similar
techniques can also be applied by reversing the order of axis. In both cases, similar results are
obtained. We demonstrate this technique for quadratic Lagrange element.
From Figure 4.6, we find that the nodes 1, 8 and 4 have the value r = —1. Then we can
apply Lagrange’s interpolation method [Eq. (3.25)] considering 9; (i = 1, 8, 4) for which the
ordinate s; are —1, 0 and 1 respectively. Hence the interpolation function becomes

P1184 => M9; (4.9)

 Chapter 4 Quadrilateral Finite Elements

(O=s\l—s) % s(1-s)
where l== (+bDa+) =
M, 5 and d similarly,
simi M, and M; are evaluated. Thus Eq. (4.9)

becomes
s(l—s) s(l+s
Pig4 = — 1 +(1-57)@g + li
5 rip (4.10)

Similarly, for r = 0 and r = 1, the interpolation functions are

s(l—s) 1+
DS cae Qs + (1-57) @y a 5 > 6, (4.11)

s(l—-s) sal +
Os So Qy + (1-57) Qe + sta
5 iy (4n12)

1 5 Z
Figure 4.6 Lagrange’s interpolation method in two dimensions.

Now three new functions—94, @ @so7, and @6;—have r coordinates —1, 0 and 1
respectively. Thus we can apply Lagrange’s interpolation method, considering abscissae as —
1, 0 and 1 for these functions, as follows:
rd-r) r(l+r)
PB ian Pits +(1=r°) 597 afer Ie (4.13)

Replacing Eqs. (4.10)-(4.12) into Eq. (4.13), we find

g

Q= SN: (4.14)

where
for corner nodes i = 1, 2, 3 and 4 (4.15a)
N; =<(1+ All + sidings

d-s7) +d t5s-7") , for mid-side nodes i = 5, 6, 7 and 8 (4.15b)

N; = Patan
2,

Ng =(1-r*)(1-s7), for inside node (4.15c)

can be obtained.
Using the similar approach, shape functions of any Lagrange element
 Finite Element Method: Concepts and Applications in Geomechanics |

4.5 Strain-Displacement Matrix

The procedure outlined for the derivation of the matrix B in Section 3.4 [Eq. (3.38)] is directly
applicable to isoparametric quadrilateral elements also. However, the dimension of the matrix
B is different since it varies with total number of degrees of freedom. For example, the
dimension of the matrix B for 4, 8 and 9 noded quadrilateral elements is 3 x 2n, n being the
number of nodes. In order to estimate the Jacobian [Eq. (3.35)] and components of the matrix
B [Eq. (3.39)], the derivatives of shape functions with respect to variables r and s are to be
evaluated as given below:

For 4-noded quadrilateral element:

Nie po)
or 1-4 I 1 (4.16
. a)

St at (lt nn) (4.16b)

TOf ical eee and

For 8-noded quadrilateral element:

ON; 1
+ =—7,(1+5;5)(27r
+ 5,5)
Gn 4 (4.17a)
ON;
se 7 (1+7,r)(2s; s+rr)
(4.17b)
fori = 1.2: °3 and 4,

So ToL d-s?)- 2 275701 +5,5)|

ON net olf as
(4.18a)
ON; 17 3 7) 2
man = 287: 2 (1+77) |
(4.18b)
LOE te 9, 0, 7 and ak.

For 9-noded quadrilateral element:

ON;
Sc
1
55,51 + 5;5)(27,7r + 1)
(4.19a)

ON; 1
eeeeee: ;1+7r)(2s,5 +1)
(4.19b)
for? =" 1, 2 '3"and=4.

ON; 1
ea r(l—s 2 (2rr +1) ~2rs,s(1 + 535) (4.20a)
ON; 1
>, a5 8Gears )(2s;s + 1) - 2nrs(1 + rr)
(4.20b)
for i = 5, 6, 7 and 8.
 = Chapter 4 Quadrilateral Finite Elements
| 99 |

ON.
=e =-2r(1- 5”) (4.21a)

~ = ~2s(1—-r7) (4.21b)
Once the derivatives are evaluated at a Gauss point,
Eqs. (3.35) and (3.39) are used to
estimate the Jacobian J and the matrix B at that point.

4.6 Stiffness Matrix and Numerical Integration Scheme

The stiffness matrix of quadrilateral elements is also evaluated
using Eq. (3.60). The Gauss’
quadrature numerical integration rule for quadrilateral elements
is different from that for
triangular elements. Equation (4.22) gives the formula for calculat
ing stiffness matrix using the
numerical integration scheme.

P q
K= [B"pB dV = h| B™DB dA=h>)'B"(,,5;)DBG;.s,)J(;.5,)W)W, (4.22)
V A i=] j=l
In the above equation, p and q are the numbers of integration or sampling columns
and
rows in r and s directions respectively. Hence there will be pq number
of sampling or
integration points. Table 4.2 describes the location and Gauss—Legendre integration coefficien
ts
Table 4.2 Gauss—Legendre integration points for quadrilateral elements

Element type Integration Degree of __ Location of integration points rand s weights

order precision coordinates
(number of
sampling
points)

4-noded 2x 2 (4) 3 n=-/1/3 W,=1.0

quadrilateral
ora NIUE Wo= 1.0
Si Vy

So = FQ

& or 9-noded "3x3 (9) s) n=-/06 W=

quadrilateral an W) =

rz = 0.6 W3=a
Sy = r)

Sp aly

Sj Lis}

a = 0.55555 55555 55556, B = 0.88888 88888 88889

 Geomechanics
| 100 | = Finite Element Method: Concepts and Applications in

s and rows and columns

of sampling points. Note that sampling points are denoted by alphabet
are expressed with numerals.
eral shown below:
Example 4.1 Determine the Jacobian at Gauss points of the linear quadrilat
y 3 (2, 4)

4 (-1, 3)

2 (4, 1)

1 (0, 0)
1
Solution: First, we calculate the Jacobian for Gauss point r=s = F ige Using Eq. (3.35),

we find the Jacobian as given in the following table:

r = -0.5774 s = -0.5774

wn
or
NN,
Os or
aN)
Os”
NBN,
or ? os
Mies 4 y

1 0 0 —0.3943 0.3943 0.0000 0.0000 0.0000 0.0000

2, 4 1 0.3943 —0.1057 LEST74 —0.1057 0.3943 0.4226

3 2 4 0.1057 0.1057 0.2113 0.4226 0.4226 0.2113

4 -l 3 —0.1057 0.3943 0.1057 1.1830 —0.3170 0.3943

det J = 3.1443 Sum: 1.8943 1.5000 0.5000 —0.6057

=— a detelineene 8557 eaudaton

Similarly, for ,=—22 ,det J = 3.2887. for yay
B B
—ja=— 5.=
1 . det J = PETIA)
V3

4.7 Body Force

Nodal body force vector is derived from the body function (foo =a fe ie } as defined in
Chapter 3. For quadrilateral elements, this force vector is evaluated using numerical integration
at Gauss points as given below:

eed
b= TA b

i=l =

Assuming that only gravity loading acts in negative y direction, the expanded form of the
above equation gives rise to
 | Chapter 4 Quadrilateral Finite Elements
|{ 101

Fix Ni (7,5;) 0
5 Oe Sees Ny(755;)
=h> > : : |renspym
ro} ll, (r,5,) G2 |SaeF (4.24)
Fa 0 Nn(5;)

Example 4.2 Determine the nodal body force vector for Example 4.1 with ape =x and
p03
4
Solution: Here the body force per unit volume is given as ie Shige lhe Hence
i=!

we can find the body force Ee for node | for Gauss point 1, ie. r=s Seber as
V3

r = -0.5774 s = —0.5774 det J = 3.1443

Node Re y N Nx
1 0 0 0.6220 0.0000
2 4 1 0.1667 0.6667
3 2, 4 0.0447 0.0893
4 -1 3 0.1667 —0.1667

Sum 0.5893

For Gauss point 1, the body force is calculated based on Eq. (4.24), F’, = 0.6220
x 0.5893 x 3.1443 = 1.1526. Similarly for Gauss points 2, 3 and 4, Fr are 15220, 0:2224,
—0.0497. Since the weights in all Gauss points are 1.0, we find the body force at node 1 with
Peas = 11526415200 £ 0229400497 12.8473. Similarly, body force in other
nodes can be compued.

4.8 Traction Force

In this section, nodal traction force vector is formulated from surface traction pressure applied
to the side (s) of a quadrilateral element. Consider that surface pressure is applied in positive
x direction on the surface (side) 2—3 of a linear quadrilateral element as shown in Figure 4.7.
The value of surface pressure at nodes 2 and 3 are P, and P3, respectively. Hence the equation
of traction pressure function (pi can be formulated using natural coordinates

F523 = NP, + NGP, (4.25)

| 102 | Finite Element Method: Concepts and Applications in Geomechanics

P; N/m?

,aN P, N/m?
ase :
see element Parent element
Figure 4.7 Surface loading on quadrilateral element.

Since the value of r is unity on the surface 2-3, by replacing shape functions in
Eq. (4.25), we find that

Tat= Fo5Pla bat ao5 2s

S2Sies
x. 4.26
(4.26)

Now, the components of surface traction on each node of the element can be obtained
by expanding Eq. (4.25) as follows:

s
is
nS N, 0
ly 0 N|
FS
2x N> 0

ES 0 N $2-3
Piz ‘ 2 [N7#%as = | ‘9 i é |e (4.27)
P3y1 5 Cyst ims 0

FS Nihor 60
4x 0 N4
5
Fay

Note that the value of shape functions N; and Ny, on the surface 2-3 is zero since
r = 1. In other words, there will be no contribution of the surface pressure acting on the surface
S2-3 at nodes 1 and 4. The component f,°”-> is set to zero since no traction pressure is applied
in y direction. From Eq. (4.27), each of the force components are obtained either by integrating
in closed form or by integrating numerically as explained in Chapter 3. For example, the
component of traction force vector at node 2 will be

(Ga i Ney
= as 4.28
S2-3 )
If the side 2-3 is a straight line, the differential dS is expressed as

dS=hids and f= 22 (4.29)

2
 [ Chapter 4 Quadrilateral Finite Elements
i [_103 |
Hence Eq. (4.28) can be rewritten in closed
form as

FS.Sq= hl,ose
:
1
il

J(it.5 flw ofP,oT —

+P, foun P,-P
£3 __ hl 2-3
: sJasee (2P, + Ps) (4.30)

Similarly, other components of traction force vector

can be estimated. Again, the surface
traction vector is of 2n x 1 dimension, n being the
number of nodes. For general formulation
of traction force vector, either for curved or straight-sided
quadrilateral, the procedure outlined
in sub-section 3.7.4 can directly be applied.

Example 4.3 Determine the nodal traction force vector

on the surface 3-7-4 considering the
coordinates of mid-side node 7 are (0.5, 3.5) as shown
in Example 4.1. The vertical traction
pressures are P; = —5 MPa, P, = —1 MPa and P, = —3 MPa.

Solution: The pressure function on the surface 3-7-4 can be written

as

fy" *= N3Ps +N7Py + N4P,

Since on the surface 3-7-4, 5 = 1, the shape functions become

eae aae a5pee lg Oty Nia" os Sa | hes and N3>-7-

é
4 = 4-2
The interpolation of coordinates in the surface 3-4—7 will be accomplished by

>
3 3
S3—]—4 meee) S3-7-4
z =) 0; Xj, y =), 4;Se)
yj
i=] tI

The determinant of Jacobian is calculated using Eq. (3.73). The following table estimates
'
the shape functions and pressure functions for Gauss points 1
7 =—n = Pe

r, = —0.57735 Tr = 057135

Node x(m) —-y(m) Mn) Mr) P(MPa) f(r) f (ry)

3 2.0 4.0 -—0.12201 0.45534 —5.0 0.6100 —2.2767
4 -1.0 3.0 0.45534 0.12201 -1.0 —0.4553 051220
7 0.5 3 0.66667 0.66667 -3.0 —2.0000 —2.0000

Equation (3.76) is applied to calculate the nodal forces as follows:

 | Finite Element Method: Concepts and Applications in Geomechanics

ie N, fO*N3*J N> fs*¥Ng*J N3 foANe*

r J = det J
.-0.12201 0.35598 0.45534 —1.32854 0.66667 —1.94512
0.57735 1.58114 | -1.84530
0.45534 2.99121 -0.12201 0.80149 0.66667 —4.37944
0.57735 1.58114. 4.15470

Sum = —2.63523 —0.52705 6.32456

The nodal forces are: Fy = -2.635 MN, Fi, = —0.5271 MN and F;, =—6.3246 MN

4.9 Stresses at Nodes

procedure
Strains and stresses are estimated at every Gauss point of each element based on the
stresses are then interpolat ed to obtain nodal point
outlined in Section 3.11. Gauss point
adopted for a 4-noded quadrilate ral
stresses. Figure 4.8 shows the interpolation method
x 2 order
element. Points A, B, C and D denote nodes and points 1, 2, 3, and 4 represent 2
of sampling points. From this figure, the relationship between r-s and 7-& coordinate systems
can be written as
ee ee er (4.31)

Figure 4.8 Coordinate system based on linear element and sampling points.

Stress at any point, P in the element is interpolated using the following equation:

Op => NG; (4.32)

I=

1
where N; = ae +rr)(1+ s,s) and 6; represents stress vector with components 6,,, O,, and T,,.

For example, o,, at node A is obtained by interpolation function as defined in Eq. (4.32) with
rand s values as r=s=—V3. In the similar fashion, nodal stresses at every node can be
interpolated from sampling point stresses. Thus the nodal stress of 4-noded quadrilateral
element is obtained by multiplying the transformation matrix with sampling point stresses as
given below:
| Chapter 4 Quadrilateral Finite Elements
‘| 105

ee a poe me
; 2 2 0 Ds
yA as oe 1 HE a
DE lee wy 2
rales
Z OS A REZ)
1 beBeROM
(4.33)

Op 2) 2} 2
SH,
2 O4
1
=< pai nal ieee.
2 DS 2 2,
In the above equation, o, and 0; denote stress vector for node and sampling point
respectively.
For quadratic element having 8 and 9 nodes, the transformation matrix
is evaluated in the
similar fashion as described above. Denoting nodes with alphabets
such as A, B, C, D, E, F,G
H and I and representing sampling points by numerals 1-9 (Figure 4.9), we can define the
transformation matrix as given below:

“eae axpena eet primi wen any alleen ree Bacar

OB b a b Cc —d —d —e —e ‘i 2
CeCe SO Go hee se eo) ad ne fae)
op b c b a —e ~e —d —d FAVS
Orr =|0 0 0 0 g 0 h 0 wat ly 06 (4.34)
Or 0 0) 0 0 0 h 0 g =) O6
oG 0 0 0 0 h 0 g 0 “tla,
Sten Oke eed (0,
ig] SD RO SOOO hs a |

A E B
Figure 4.9 Coordinate system based on quadratic element and sampling points.
 Geomechanics |
106 | ‘2 Finite Element Method: Concepts and Applications in

where

a= ay(1+ 06), b==,S c= Z-(1-V06) . d==(1+ 06), e=5 (1-06), f= 5°

25 2 5 5 4

06)and i==
06),h=3(1+
¢=2(1-
For quadratic serendipity element, the last row will not be included.

4.10 Tunnel Example

quadratic
The tunnel problem mentioned in Section 3.14 is again solved with linear and
4.10. The
quadrilateral elements. The finite element mesh of the problem is shown in Figure
models are developed with 264 and 757 nodes for linear and quadratic quadrilateral elements
respectively. Both the models consist of 230 elements. Near the boundary of the tunnel, small
size elements are developed for more accurate stress and displacement distributions. Large size
elements are developed away from the boundary of the tunnel. The dimension of the model,
rock properties, boundary conditions and loading conditions are exactly the same as mentioned
in Chapter 3. The results of finite element solution are also compared with those of the Kirsh’s
closed form solution which are mentioned in sub-section 3.14.1.

Figure 4.10 Mesh of a quarter section of a tunnel with quadrilateral element.

Figure 4.11 shows the distribution of tangential stress and radial stress from the boundary
of the tunnel up to 4 times the distance of the radius. It can be seen that stresses obtained from
the model with 8-noded quadrilateral elements are perfect match for those obtained from the
analytical solution. The radial stress obtained using linear quadrilateral elements differs slightly
 a Chapter 4 Quadrilateral Finite Elements _| [107 ]
2) i [ae oo So ee ae

Tangential stress

Radial stress
0,/P,
6,/P,
or
ie —— Analytical e 4-noded A 8-noded
1 15 2 2:5 3 oh ed
iS
Jee
rla
Figure 4.11 Distribution of tangential and radial stresses.

with that of analytical solution near the boundary of the tunnel. The distributio
n of radial
displacement is shown in Figure 4.12. It can be seen that both the models predicted radial
displacements with considerable accuracy when compared with the analytical solution.

0.003
— Analytical e 4-noded
ai
A 8-noded |
0.0025
i
}
0.002

0.0015
(m)
—u
0.001

0.0005
}

0 oc oe eg a a

I JES) Z Dee) 3 3.5 4

rla
Figure 4.12 Distribution of radial displacements.

SUMMARY
Isoparametric finite element formulation of quadrilateral elements is illustrated in this chapter.
Derivation of shape functions for linear, quadratic serendipity and Lagrange elements is given
using the concept of parent element. A parent element is expressed with a square of side
2 (-1 to +1). The coordinates of a physical element are then mapped based on the parent
element. The formulation procedure of stiffness matrix and load vectors for quadrilateral
elements is partrayed with the help of numerical integration procedures. However, Gauss’
quadrature integration rule is different for quadrilateral elements as compared to that for
triangular elements. For linear and quadratic quadrilateral elements, 2 x 2 and 3 x 3 integration
rules are sufficient for the determination of stiffness matrix and load vectors. The
transformation matrices for the extrapolation of Gauss point stresses to nodal point stresses are
also explained in this chapter. Finally, a tunnel example is solved with linear and quadrilateral
elements and the results are compared with those of the analytical closed-form solutions.
108 | | Finite Element Method: Concepts and Applications in Geomechanics

EXERCISES
4.1 A 4-noded quadrilateral element is shown in Figure P4.1. The nodal point coordinates
are also shown in the figure. Determine the following:
(a) The x, y coordinates of a point P whose coordinates in parent element are
r = -0.5 and s = —0.5.
(b) Repeat the problem given in (a) for a point Q with r = 0.5 and s = 0.5
ay
An(25)

3 (4, 2)

1, (0,1) 2 Oph)

Figure P4.1

4.2 If the displacement vector of the element given in Figure P4.1 is q’ ={0 0.03
—4.2 0.5 —5.7 0 -1.4}, find the following:
(a) Displacement vector at the points P and Q
(b) Strain vector at a point with r=1/V3 and s =1//3

4.3 Show that for a 4-noded rectangular element, the value of Jacobian at any Gauss point
is one-fourth of the area of the rectangle.
4.4 Using 2 x 2 integration rule, evaluate the integral J = |(x* + y*) dxdy, where area
A is defined in the elements shown in Figure P4.1. 4
4.5 Determine the components of body forces at inside node (no. 9) of a Lagrange quadratic
rectangular element having corners (0, 0) and (2, 2). assume f? = 0 and un = jy.
4.6 The nodal point coordinates x and y are to be determined for a quarter section of a
circular tunnel having inner and outer radius of 4 m and 10 m respectively as shown
in Figure P4.2. The parent element is divided into 16 parts to map the nodal points
as shown in the figure. Estimate the coordinates of nodes in the physical model.
y 5S

—!—5= 0

—1/2-s=0
aSa—a10)

—1/2+s=0

4m orcs bar ae
4m l+r=1/2+r=01/2-r=01-r=0
10 m :
Parent model
Physical model
Figure P4.2
ee Chapter 4 Quadrilateral Finite Elements | [109|
4.7 For a 4-noded quadrilateral element, determine the nodal forces at nodes 3,
6 and 9
for the linearly distributed pressure load as shown in Figure P4.3.

@ For 4-noded element

O For 8-noded element

Figure P4.3

4.8 Solve Problem 4.7 considering 8-noded quadrilateral elements.

4.9 A quarter section of a circular tunnel is analyzed using linear quadrilateral element
as shown in Figure P4.4. The element 1 is connected by nodes 1, 12,13 and 2 whose
coordinates are as follows:

Node x(m) y(m) u(mm) v(mm)

1 2 0 —4,535 0
12 Ns 0 —4.461 0
13 2.22230 0.35198 -4.406 —0.698
Z 197538 .090.31287 4 =4:479 —0.709

The displacements at each node are also given in the above table. Determine the stress
concentration factor at element 1. Assume E = 5.5 GPa and v = 0.25. A uniform pre-
mining stress (p,) of 8 MPa is applied on the outer boundary of the model (not shown
in the figure). Ignore internal pressure pj.

Figure P4.4
 Axisy panier mid 3D
Finite Element Method

5.1. Introduction
Axisymmetric models are used to simplify a symmetrical three-dimensional structure. The
results obtained from axisymmetric finite element analysis are similar to those from a three-
dimensional structure. In this chapter, axisymmetric finite element analysis will be illustrated
for symmetric loading conditions only. This type of analysis is widely applied for circular
tunnels, nine shafts and other structures where axial symmetry is maintained. The
computational burden of axisymmetric models is greatly reduced, especially if non-linear
analysis is performed.
This chapter also elaborates the isoparametric finite element formulation for three-
dimensional (3D) solid structures. Finite element procedure for three-dimensional structures is
straightforward extension of two-dimensional finite element procedure as mentioned in
Chapters 3 and 4. The procedure of discretization with solid volume elements, determination
of shape functions, formulation of B and K matrices are similar to those in two-dimensional
cases. In 3D analysis, additional components of strain and stress tensors such as €,,, ¥,, Yz, and
O.-, Tyz T, respectively are considered. Thus the dimension of D matrix becomes 6 x 6.
Accordingly, the dimension of elemental B and K matrix also increase many fold depending
on the type of 3D elements. The computational expense also increases many fold depending
on the applications.

5.2 Axisymmetric Analysis

The choice of elements for axisymmetric analysis can be of any two-dimensional solid elements
as mentioned in Chapters 3 and 4. Thus, a solid area is discretized with 3- and 6-noded
triangular elements or 4- 8- or 9-noded quadrilateral elements. The strain displacement
relationship for axisyemmtric problems was given in sub-section 2.2.2. In the subsequent
sections, we will mainly illustrate the formulation of B matrix and how to apply symmetrical
external traction pressure into finite element models.
110
 | Chapter 5 Axisymmetric and 3D Finite Eleme
nt Method | | 111 |
9.2.1 Formulation of B and K Matrices for Axis
ymmetric Analysis
As mentioned before, the procedure of formulating
B matrix is same as 2D solid elements.
However, the components and size of the Strain-displ
acement matrix are different than the 2D
solid elements. Referring to Eq. (2.23), strain compo
nents of axisymmetric analysis, it is clear
that the size of the B matrix will be 4 by 2n, where
n is the number of nodes in the element.
In this section, natural coordinate system will be
represented by 7 and € instead of r and s
respectively. We will illustrate the formulation of B matrix
using 4-noded quadrilateral element
as shown in Figure 5.1. In this figure, y axis is assumed
to be symmetrical axis and r is referred
as radial axis. The shape functions of the nodes are the
same as explained in Chapter 4. The
radial distances r; is measured for ith node from the
Symmetrical axis.

1)
; r3

ay! 1 T4 3 4 3

1 4
1 ¢
ruy
A 1 wy,
i
mer I 5)
Parent element
1 1h)

Symmetry axis
Figure 5.1 Concept of axisymmetrical finite element analysis.

,
Recalling strain vector {e} 7 =18, Ega Exo ae
Ou
Ho
u ou
at ov
or ov
55 , we can
write

a|
E, Dy ebis = ban Sy
gpa by, bay +++ Baan ay (5.1)
bore) b3) b39 ae) b3on fi

Ey Page 23 5D yo, ||”

Vn

where

ing oes spar oN ys en; for joao ween

ae TNS On wali de— |S08 |on

by2j) =0

Doj-n'= ae foigy = 1, 2, 3,...., 7

Im
 Concepts and Applications in Geomechanics
[112 | Finite Element Method:

by(2;) =9
bya j-1) =9
D
1 at ON; ON ; = ON; ON;
ll se ~ se N wa
b4(2/) Sed Pea for J — = —

i=]

b3¢2 j-1) = 94(2/)

(Olsja abs 2s ona ele
b3(2 }) = by2 j-1)
n

and /m— SY Nir

i=!

Equation (5.2) is valid for other elements such as linear and quadratic triangular elements
and higher order quadrilateral elements. Once the B matrix is determined, elemental stiffness
matrix K for triangular and quadrilateral elements is evaluated using the following equations:
Triangular:

(5.3a)
Pp

K = [B™DB AV => 2m) BT G.n)DBE,.n Eon Eom

V 1!

Quadrilateral:
P 4
K = [B™DB dV = (22) ))9B". DBE. n(Gin IGn Wim G3)
V i=l j=l

Here dV =r,, d0 J dé dn. The numerical integration scheme is followed according

to Table 3.4 and Table 4.2 based on element types. The constitutive matrix D is given
in Eq. (2.24). For triangular elements, direct integration scheme can also be applied using
Eq. (3.64).

Example 5.1 Determine the strain-displacement matrix for 3-noded triangular element for
axisymmetric analysis.

}
re 3
'
ee

Axis of symmetry
 Chapter 5 Axisymmetric and 3D Finite Eleme
nt Method se E 113 |
Solution: For 3-noded triangular element (n = 3), the
shape functions are N; = 6,
No = 7, Nz3=1-—E- 7 and Jacobian J = 2A, where
A is the area of the triangle. Hence using
Eq. (5.2), we find that

= 8 eh, =
2A iain 4 in DA
bi2;) =0 for: j= 11256) dK tn

by, ==> by3 =—> pees il

m Tn Tm
by(2;) =0 fory*= NSO! 3) wom;

b4(2 j-1) =0 LOL =e Oma ae 7

b3(2 j-1) = ba2) LOC e=tHls. DO te i

b302j) = by(2j-1)
Here x; coordinates denote the distance of ith node measured from the axis of symmetry.

5.2.2 Traction Force

As mentioned before, only uniform or symmetrical loading conditions will be discussed here.
For implementation of unsymmetrical loading on axisymmetrical problem, reader can refer to
references (Bathe, 1984, Cook et al., 1989). Symmetrical loading can be applied on two
surfaces as radial or longitudinal loads as shown in Figure 5.2.

Longitudinal load

Figure 5.2. Application of symmetrical load.

 Finite Element Method: Concepts and Applications in Geomechanics

is similar as mentioned in sub-

The. implementation procedure for traction load vector
en as
sections 3.7.3 and 3.7.4. In this case, Eq. (3.76) is rewritt

(5.4)
m

PS =2nr, ¥a,R; and R, = Njj f° det JjS

i=)

surface S, a; = weight of sampling

where m = number of sampling (Gauss) points along the
N/ = shape function value
point at ith location, f> = force function expressed in Eq. (3.73),
res as mentioned in sub-
of jth node on the surface S at ith sampling point. All other procedu
A multiplication factor of
section 3.7.4 are directly applicable for axisymmetrical elements.
are used. Note that due to a
1/2 will be included in the above equation if triangular elements
22r;f for a nodal radius
uniform line, the load f having unit N/m will produce the nodal load
.
of r;. Radial load is generally responsible for deformations and stresses

e the traction
Example 5.2. Consider again the example given in sub-section 3.7.3. Determin
axisymme tric analysis as shown in Figure 5.3.
load vector of linear triangular element used for

'
'

r3 3 P3 N/m?
'

os
'

i]
'
'

'

=: P, N/m?
1 Lo) 2
Axis of symmetry
Figure 5.3 Traction load for linear triangular element.

Solution: In this case, the pressure function fe can be written as

£573 = N>P, +.N3P; = sP, +(1—5)P3 = P3 + s(P) — Py)

x

The determinant of Jacobian as given in Eq. (3.73) can be obtained as

s ax® : dy? aia 7 1/2

Hence the force components in radial direction at nodes 1 and 2 can be obtained in the
similar fashion as given in Eq. (3.77) and using Eq. (5.4) as
ey 2s
Fao 23 (Ope)
271
Fa OP P,
 Chapter 5 Axisymmetric and 3D Finite Element Method | | 115 |

5.2.3. Body Force

Nodal body force vector is evaluated by numerical integration at sampling points. In this case,
thickness of the element h is omitted and a new parameter, Ty», Called mean radial distance from
the axis of symmetry is introduced. The body force vectors for triangular and quadrilateral
elements are estimated using the following formulae:
For triangular element:

2 P
F? = —— N*(G.1;)£?I(E.1;) Wit (E;.7;) (5.5a)
i=]

For quadrilateral element:

Pq
b_ T b
F =o yy ON (6. f° SG: )WjWitn (C:.7;) (5.5b)
Age
For the calculation of stiffness matrix and load vectors, a multiplication factor 27 is
included in all the equations. In order to avoid this unnecessary multiplication, 27 can be
replaced by one radian in all the equations.

5.2.4 Comments on Stiffness Matrix and Stress Computations

It can be seen that strain-displacement matrix, B, contains 1/r as the multiplier and so
does the stiffness matrix. Since there is no Gauss point at r = 0, the integral given in
Eqs. (5.3) and (5.5) are finite even if core element is present in the model. However, numerical
accuracy can be improved by letting more number of integration points for the core element.
For distant elements away from the axis of symmetry, the general Gauss’ quadrature integration
rule is sufficient.
For core elements, the indeterminate term €gg = u/r = 0/0 arises at the axis of symmetry.
Thus it may seem that stress at the axis of symmetry is indeterminate. To resolve this, we
estimate the strain and stress at Gauss points which are away from the axis of revolution or
symmetry. Then standard extrapolation techniques as given in Chapters 3 and 4 are used to
estimate the stress at the axis of revolution. It is also noted that radial displacement u = 0
should be prescribed at all nodal points that lie on the axis of symmetry.

5.3 Tunnel Example

The tunnel problem mentioned in Section 3.13 is again solved with axisymmetric finite element
analysis. Since the tunnel geometry is circular and the applied loading is symmetrical, this
problem can be solved with axisymmetric analysis. The finite element mesh of the problem is
shown in Figure 5.4. In this case, y axis is assumed to be the axis of symmetry. The tunnel
surface is 2 m away from the axis of symmetry. From the boundary of the tunnel, the radial
extension (x or r direction) of the model is 38 m. The height of the model is 0.5 m (in the
figure it is exaggerated by 4 times). The models are developed with 48 and 118 nodes for linear
and quadratic quarilateral elements respectively. Both the models consist of 23 elements. The
 Finite Element Method: Concepts and Applications in Geomechanics

Figure 5.4 Axisymmetric finite element mesh of the tunnel problem.

model is constrained vertically as shown in the figure. Near the boundary of the tunnel, small
size elements are developed for more accurate stress and displacement distributions. Large size
elements are developed away from the boundary of the tunnel. The results of finite element
are also compared with the Kirsh closed form solutions which are mentioned in Section 3.13.
Figure 5.5 shows the distribution of tangential stress and radial stress from the boundary
of the tunnel up to 4 times the distance of the radius. It can be seen that stresses obtained from

|
ae

Tangential stress

Og/po
6,/py
OF
Tangential stress
0.4

0.2 — Analytical © 4-noded elements 48-noded elements)

0 a ae T T a

1 ibs) 2 ORS 3 Bh) 4

rla
Figure 5.5 Tangential and radial stress distribution.

model with 8-noded quadrilateral elements are perfect match for those obtained from the
analytical solution. It is also noticed that these results are the replica of the results shown in
Figure 4.11. In this case also, we find that the radial stress obtained by using linear quadrilateral
elements differs little with that of analytical solution near the boundary of the tunnel.
The distribution of radial displacement is shown in Figure 5.6. As before, it can be seen
that both the models predicted radial displacements well when compared with the analytical
solution.

5.4 Finite Elements in Three Dimensions

Finite element procedure in three dimensions is a straightforward extension of two-dimensional
cases. In this case, additional three strain and stress tensors are added to the strain and stress
vectors as given below:
li
€ = Ey Ex Vxy Yyz ies (5.6a)
| Chapter 5 Axisymmetric and 3D Finite Element Method a 117

0.003 — eee Se eorpescteastnenesinasoone ee esas cam essernperen tse sbeseteermeyers onsen teememetaneewesscomssse

4 Analytical © 4-noded elements |

4 8-noded elements

—u(m)

1 es 2) Des) 2 BS) 4
rla
Figure 5.6 Distribution of radial displacement.

0 20, Oy, GO, Ty Ty ee (5.6b)

The constitutive relations between stress and strain are given in Eq. (2.13). In this section,
we will discuss the choice of elements in three dimensions and their implementation procedure
in finite element analysis.

5.4.1 Discretization in Three Dimensions

A volume in 3D is discretized using tetrahedral or hexahedral (brick-type) elements. A
tetrahedron can be linear with 4 nodes at its vertices or quadratic with 10 nodes as shown in
Figures 5.7(a) and 5.7(b). For quadratic tetrahedron, apart from apices, nodes are also placed
at mid side of edges. A brick-type element is vertical extension of quadrilateral element having
8 or 20 nodes [Figures 5.7(c) and 5.7(d)]. A 8-noded brick element is called linear brick
element and 20-noded brick element is termed as quadratic brick element. Lagrange element
can also be formulated in 3D. However, in this text, it will not be discussed.

A>GpLinear tetrahedron
(a)
Quadratic terahedron

Figure 5.7
(b)
8-noded brick
(c)
Element types in three dimensions.
20-noded brick
(d)

In three dimensions, each node has 3 degrees of freedom having u, v and w displacement
fields in x, y and z directions respectively.
 118 | = Finite Element Method: Concepts and Applications in Geomechanics

5.4.2 Shape Functions of Tetrahedral Elements

Volume Coordinates

In order to formulate shape functions for tetrahedral elements, natural coordinate system is
assumed as in the case of triangular elements. In this case, natural coordinates are the ratios
of volumes rather than areas. If a point P(x, y, z) is inside the tetrahedron defined by nodes
1, 2, 3 and 4, then natural coordinate or volume coordinates corresponding to node | can be
obtained as follows (Figure 5.8):
eVIlGrds4-Rieie
ted emus (5.7)
dee ich (lates
where V, is the volume of the shaded portion in Figure 5.8. The volume V is referred to as
the total volume of the tetrahedron and is given by the following equation:

{I and reg boty | Pere Cet fre,

a ee aie 2 el (5.8)
Oly, "Ys ¥3 Ya G| pio yam Sty
ig ee FE Lae ee CA
My

The magnitude of the volume V will be positive if the coordinates 1, 2 and 3 are considered
in counterclockwise direction with respect to the node 4 as given in the above equation.
Similarly, the volume coordinates s, t and p can be estimated as

Alias coe (5.9a)

“vol (1=2—3—4)
I aie (5.9b)
~'Vol(1-2=3=4)
got
eee)
OM (5.9c)
i pVol A234)
Thus we can write that r+s+r+p=1. Figure 5.8 shows the volume coordinate values
at each node for a tetrahedral element.

IS

3 (0,0,1,0)

Figure 5.8 Volume coordinate values at each node for a tetrahedral element.
| Chapter 5 Axisymmetric and 3D Finite Element Method 119

Linear Tetrahedron

In order to derive the shape function, we again assume the polynomial of order 1 as @ =
a, + G>x + azy + a4z and solve for constant terms a; using the field values of nodal points.
Using the direct, serendipity and/or Lagrange methods, the shape functions of linear tetrahedron
is found as N,; = r, N, = s, N3 = t and Ny = 1—r—s-—t. Note that in shape functions are
function of volume coordinates r, s and t.

Quadratic Tetrahedron

Shape functions of quadratic tetrahedron are direct extension of quadratic triangles. For a
quadratic tetrahedron shown in Figure 5.9, the shape functions are:

N, =r(2r—1), Nz =s(2s —1), Nz =t(2t-1D), Ng =A —r—-s—-td-—2r—2s—2t) (5.10a)

Ns =4rs, No = 4st, N7 =4tr, Ng =4r(l—r—s-—t), No =4s(l—r—s-—t) (5.10b)

Nip = 4t(1—r—s-t) (5.10c)

4

g 10

3
]
6
> 2

Figure 5.9 Node numbering scheme of quadratic tetrahedron.

5.4.3. Hexahedral or Brick-Type Element

Linear Hexahedral Element

Figure 5.10 shows a quadratic 20-noded element having parent element in natural coordinate
system denoted by r, s and ¢. In parent element, only corner nodes are shown. The natural

8 , 10
x

16 2
yy,
18
5 13 6
Parent element
Physical element
Figure 5.10 Node numbering of brick-type element.

can
coordinates of each node are listed in Table 5.1. The shape functions of brick-type element
l elements. Thus
also be considered as the direct extension of shape functions of quadrilatera
for 8-noded brick elements, the shape functions are given in Eq. (5.11).
 Finite Element Method: Concepts and Applications in Geomechanics |

Table 5.1 Natural coordinates of nodes of brick-type element

Node no. (i) i; S; LF Node no. (i) I; S; 1;

] -1 —] -l 11 0 -1 ]
D 1 —] —1 12 -1 -] 0
3 1 -l ] 13 0 1 -1
4 -] -1 1 14 1 1 0
5 -1 ] -| 15 0 ] ]
6 1 ] -l 16 -l 1 0
7 1 1 1 17 -1 0 —l
8 -l 1 1 18 1 0 -]
9 0 -l -l 19 1 0 1
10 ] -1 0 20 -] 0 1

N; =-( +rr)1+s;s)1+t;t) for? 71 tes (S-ig

Quadratic Hexahedral Element

The shape functions of 20-noded brick elements are given in Eq. (5.12). Note that these shape
functions are obtained by applying direct or serendipity methods. For serendipity method,
equation of plane is assumed instead of line as mentioned in Chapter 3. The similarities of
shape functions between quadrilateral and brick type elements are remarkable.
]
N; = git nnd + ss) + air t 5; +t — 2) for i = 1 to 8 (5.12a)

522
Np =P +5) +42) -r 2 )+-S
I
tre
E +420 + yr) - s*) +E
rsh
+ Ar) +5; - re)
for i = 9 to 20 (5.12b)

5.4.4 Formulation of B Matrix

The strain displacement matrix for isoparametric 3D elements is formulated in the same fashion
as mentioned in Chapter 3 for 2D element. However, in this case, one additional dimension
(z axis) increases the computational expense in many folds. Just as Eq. (3.28) for
2D elements,
Eq. (5.13) is expressed for isoparametric 3D elements as

x={x y z} =Ne (5.13a)

u={u v wh’ =Nq (5.13b)
where the matrix N is the collection of shape functions of 3D element. Extension of
Eq. (5.13b) is given in Eq. (5.14).
 Chapter 5 Axisymmetric and 3D Finite Element Method tats)

0
0 = Nq (5.14)

Assuming a field quantity g, we reproduced Eq. (3.31a) for 3D elements as given in

Eq. (5.15).

00 dx dy dz|\|d@
or dr or or|lox
ap|
ds|
_|ax ay a |}ag
|as ds ds] dy (5.15)
Oe) Ce (82 0p
ot ot ot ot}l dz

The Jacobian J (det J) is then the determinant of the above matrix and expressed as

j-2 (2% ze) 2 (2% ar 2) (oo 2 os)

Ray aap tovion) ards oFtosOF)lor osor Os or (5.16)
where

Or ONE Ose STON. 5 4Ot a SON: (5.17a)

iC ania ale A
Se teil) ee
oN; dy ON,
dy WON; dy
or Nh ar yi (47D)
eps or A os = os ae

MeN eB Ny a yz, (5.17c)

{=

system as
Equation (5.15) is inverted to express the strains in Cartesian coordinate
 122 | Finite Element Method: Concepts and Applications in Geomechanics

dp 0p
oy 4, 42 %3 oF
dg} 1 dp
er (umes ee a
ao 43; 2 33 dp

se Ot
where the coefficients a;; are given below

a =O_ OY --[22-2%) _ oydz_ dz dy (5.19a)

1 Ogi) dsict’ 1°) land: or or Sereumorus

i --($2-2%) ayy woe _9Z OX --(3S 22) (5.19b)

ere sie Breyo sey we dros oras

_ Ox dy _ dy dx a [RS 2) ae, = OE OY _ OY OX
a ds ot Or Ot’ gat By or ot Or odt A een
SRP oak (5.19c)
Based on the above equation, strain tensors can be related with nodal displacement
vector as

€ 4
22 bi b> bi3 ies bi3n ,
1
Eyy by, Dyn bag +++ bay *
= Eealee
tA bom31 bere
2 33 ae 33n : = Bq

Yay| |541 Dan B43 ++ Dgay > (5.20)

Vel \Oste Us cya 0c Dene a
Vi} 126i Pen. Yes Bea, Ss
n

where coefficients of B matrix is calculated as

: If ON; aN, ne aN,

ee
j-2) yole Ce Wee
1(3 ed OD ov
ae aT omer)
a

by3;-1) =0 for) Sd P83, GIG]

by3j) =9
by(3;-2) =0

1
ON,
ue
aN; ON;
2G /-) Fa ae ata sata {ory = 1, See nT
bx(3 j) =0
| Chapter 5 Axisymmetric and 3D Finite Element Method [123 |

0 ee 3, ne 210)
33 j-2) = 9
me ONG CON ON
b3(3 j) ty ELE i ie SEES Gas aS

ba(3 j-2) = 523j-1)

ba3j-1) = 13 j-2) foreel yo, id)

bya =0 |
bs j-2) =9
bs3j-1 = 5303) for Pa? 3A (S:21e)

bs(3j) = 523-1)
bg(3 j-2) = 9303)
for j= 1, 2, 3, ...,. (5.21f)
b63j-1) =9
b63j) = 1j-2)
are 6
It can be noted that for 4- and 10-noded tetrahedron, the dimension of B matrix
the dimension of
by 12 and 6 by 30 respectively. For 8-noded and 20-noded brick element,
expense increases
B matrix becomes 6 by 24 and 6 by 60 respectively. Thus, the computational
up to 10 folds from 2D elements dependin g on the element type.
estimated using the
Once the strain-displacement matrix is estimated, the stress vector is
following equation:
o = DBq (5.22)

where D is the constitutive matrix given in Equation 2.13.

5.3 Determine the coordinates of point P as shown in Figure 5.8 itera, Ole
Example
the coordinates Gtrnodes as 1G, 1, 1), 2. (2, 037 0)
s = 0.2 and t = 0.3. Consider
eaiaecie 1) and 4°(, 0:5, 3).
point P are
Solution: Using Eq. (5.13a), we can write the coordinates of
4
xp = )Njxj =0.1X140.2x2+0.3x44+04x2=2.5
i=1

4
x0.5 =0.43
yp = > Nii =0.1X1+0.2xX0.5+0.3x14+0.4
i=l
 124 Finite Element Method: Concepts and Applications in Geomechanics

zp =) Njzj =0.1X140.2x0.5+0.3x14+0.4%3=1.7
i=]

5.4.5 Formulation of Stiffness Matrix

The stiffness matrix is formed using direct polynomial integration formula or using numerical
integration with the Gauss’ quadrature numerical integration rules. For linear and quadratic
tetrahedron, polynomial integral formula is used as given in Eq. (5.23).

a_b.c_d vo alb!c!d!

fei Rape A SEETREC - (5.23)

where a, b, c, and d are non-negative integers and V represents the volume of the tetrahedron
.
Hence stiffness matrix K is obtained using the following equation:

K=[B DB dv =[B (r,s,1)DB(r, s,)dV

V V
(5.24)
It can be seen that direct integration using polynomial integral formula requires
large
number of algebraic equations to be formulated and thus it is cumbersome.
For brick type elements, Gauss’ quadrature rule is extended in third
dimension to
integrate the volume of the element. Figure 5.11 shows the location of
sampling points for
linear brick element. In the figure, numerals represent nodes and alphabets
(a—h) signify the
sampling point numbers. It can be easily compared that location of sampling
points for linear
brick elements are just the extension of linear quadrilateral elements
with r=s5 Steals):
and weight equal to 1.0. Similarly, 27 sampling points (9 points
in each layer; t=-J0.6,
t=0and t= 0.6) are required for quadratic brick-type elements.
The weights at corner and
mid-sampling points are same as quadrilateral elements. In
this case also, the location and
weight of these sampling points are the same as quadratic
quadrilateral element with addition
of points along the third dimension t. Equation (5.25) illustrates the numerical integration
process for brick-type element. For linear and quadratic
brick-type elements, we consider a =
b=c=2anda=b=c=3 respectively.

K= |B' DB dV = jjjB' DBJdrdsdt

V —l-1-]
ab G

=) > >B T (7555. )DB(7.5;,44)I(j.8).t,) WW, W, (5:23)

i=l j=l k=
| Chapter 5 Axisymmetric and 3D Finite Element Method [ 125 “i

Figure 5.11 Location of sampling points in linear brick elements.

5.4.6 Traction Force

The procedure for estimating nodal force vector from traction pressure is similar as 2D cases.
The only exception is that in 3D, traction pressure is applied on a surface rather than on a line.
Figure 5.12 shows a linear brick element with a traction pressure applied on the surface (S)
denoted by 2—3—7-6 in x direction.

NX

Figure 5.12 Estimation of traction load vector.

The nodal traction force vector at nodes 2, 3, 6 and 7 is then estimated using
Eq. (5.25) as given below:
m

FS =)\a,R, and Rj =Nj f° detJ; (5.26)

i=]

where @; is the weight of sampling point at ith location. The function f 5 denotes the pressure
function associated with the surface S. In Figure 5.12,

f° =N>P) + N3P3 + NoPo + N7Py (5.27a)

The Jacobian of the surface is obtained from Eq. (5.27b) and numerical integration is
performed with the sampling points similar to that of quadrilateral element.
ax> ay* dz>
ur 4 or» Or
axe dy> woz" (5.27b)
Ost “ds ds
i 1 1
 Finite Element Method: Concepts and Applications in Geomechanics

Example 5.4 If the surface 2—3-7-6 in Figure 5.9 is rectangular with area A, show that
A

Solution: Considering Figure 5.9, we find that on the surface S, the value of r is 1.
Hence the shape functions can be obtained from Eq. (5.11) as

Np = -syl-9, Ns = (1-5) +2), Ng == +=, Ny = +9) +2)

The Gauss points on that surface will be s =+1/ 3, t=+1/V3 with weight 1.0. Since
the surface is rectangular, det J at each Gauss point will be A/4. Then using Eq. (5.25), we
can find

ise =FA (No (N2P + N3P; + NoPs +N7Py))

4

i=]

aotUh Pe Pee Tp)peep

lg 7 eiO RINE GLO UEG |
A

5.4.7. Body Force

If only gravity loading is assumed in the opposite direction of z axis, then the vector f? becomes

(5.28)

Then body force vector distributed at every node can be estimated as given below:

F = N’ feav
(5.29)
V

where matrix N is given in Eq. (5.14).

Example 5.5 Find the body force component of a quadratic tetrahedral element at nodes 1
and 5 as shown in the figure below:

6
| Chapter 5 Axisymmetric and 3D Finite Element Method

Solution: Since ms = ie =(, components of gravity load in the x and y directions are
zero for all nodes. The component of gravity load vector in z direction at node 1 is calculated
using Eq. (5.29) as follows: .

gy EE = a meee
Fz = -sy| N\dV = -2y| r(2r—l)dV = -gy V|Q2r? —r)dV =-6V i !
4 V

Similarly, the components of gravity load vector in z direction at node 5 will be

|- Vey
Fy, = -gy |NsaV = -gy |4rsdV = -6ve7|4 A 2
Vv V
a
For brick type elements, body force vector is estimated using numerical integration as
given in Eq. (5.30). In this case, Gauss’ quadrature integration scheme is applied just as
illustrated for calculating stiffness matrix.

fe {a _<e
Fo = YY IN" (75,55.t EI G5j.t) WWM (5.30)
i=l j=1k=1

5.4.8 Determination of Strains, Stresses and Reaction Forces

outlined in
Elemental strains, stresses and reaction forces are calculated based on the concept
sub-sections 3.11.1 and 3.11.2. For three-dimensional elements, appropriate strain-displacement
is
matrix and constitutive matrices are to be considered. Since the number of sampling points
elements, calculations for strains, stresses and reaction forces are
higher than that of 2D
at
expensive and time consuming. In general, elemental strain and stress vectors are estimated
each sampling point using the following equations:

€; = Bq’ te) beter 0 (5.31a)

of =D’es i= 1, ...,m (5.31b)

of sampling points.
where the subscript i represents sampling point and m denotes the number

5.5 Example of Shaft Pillar Design in Three Dimensions

5.5.1. Problem and Model Description

a 6 m diameter return airshaft of
At a mine, operated by a coal company in Southern India,
sinking work was proposed
190 m depth was developed up to coal Seam III. Since the shaft
was planned in these seams. The
after the development of Seam I and Seam III, no shaft pillar
 Finite Element Method: Concepts and Applications in Geomechanics

proposed shaft passes through the corner of a pillar location in Seam III. For this reason, two
entries were developed for accessing the shaft from the existing galleries by cutting the pillar.
The isometric view of the shaft at Seam III level is shown in Figure 5.13 depicting coal pillars
and shaft entry areas. The proposed width of these new entries was 4.2 m. The most

Figure 5.13 Isometric view of the Seam III near shaft area.

predominant rock in the overburden is white sandstone comprising of about

44% of
overburden. A thin carbonaceous sandy shale layer exists in the immediate
floor. The
immediate roof is composed of coal layer having a thickness of 1.27 m. The strata
is composed
of four different rock material which is given in Table 5.2
The main objective of this study was to analyse the stability condition of the
galleries and
coal pillars when 1 m thick concrete walls built around the pillars (Deb
et al., 2004).

Table 5.2 Lithology of the rock strata modeled

Rock strata E (GPa) V Y c (MPa) )

White sandstone S15) 0.33 2250 1.78 20
Carbonaceous sandstone 1.4 0.36 1890 Lats 15)
Carbonaceous sandy shale 29 0.36 1890 PS 15
Coal 2.9 0.35 1500 2.35 30

where FE = modulus of elasticity, v = Poisson’s ratio, Y =

density, c = cohesion, ¢ = angle of
internal friction.
Three-dimensional finite element models of the Seam III and
overburden rock layers was
developed using ANSYS FEM software package. Figure
5.14 shows the FE model of the
studied area. The model depth is 100 m and an additional
pressure of 2.21 MPa is applied on
top of the model to simulate a total depth of 190 m. The cross
sectional area of the model is
50 m by 50 m comprising of the shaft and pillar area
in Seam III as shown in Figure 5.14.
The finite element model consists of 25,544 quadratic tetrahed
ral elements.
One meter thick concrete walls around the shaft openin
g are designed to protect the
stability condition around the shaft opening at Seam
III level. As displacement and stress
concentration factors (SCF) in the roof around the
shaft opening depends on the physico-
 Chapter 5 Axisymmetric and 3D Finite Element Method =) (geo

i,

;
ay,
PES

Figure 5.14 Three-dimensional finite element model of the study area.

mechanical properties of the concrete wall, so it is absolutely necessary to maintain an optimum

and consistent material property of the concrete wall during the construction of walls. In order
to simulate the effect of different material properties of concrete wall on shaft-pillars, five
different elastic modulus values are assumed. From various literatures (Biron, 1983 and
Hartman, 1993), it is obtained that the average elastic modulus of concrete composed with stone
values 10,
chips and cement is around 20 GPa. For this study, another four elastic modulus
“20 GPa
15, 25, and 30 GPa are also assumed for the analysis. From now onwards in the text
so on.
model” signifies the finite element model with concrete elastic modulus of 20 GPa and

5.5.2 Results and Discussions

are shown
Plan view of Seam III is shown in Figure 5.15. The results of finite element analysis
C- C and D — D’. Figures 5.16 and 5.17 show the
along four line sections A — A’, B — B’,
of concrete wall
principal stress distributions at the roof above Seam III having elastic modulus
tion (factor of
as 20 GPa (or 20 GPa model). As shown in Figure 5.16, high stress concentra
m inside the concrete
12 ~ 15) occurs at the corners of the concrete walls. However, about 0.25
The spalling of the coal pillars may occur
walls. the stress concentration factor drops to 3 ~ 5.
Figure 5.17 shows the minor
without causing any problem to the stability near the shaft area.
same model. It clearly
principal stress distribution in muck fill and coal pillars in detail of the
the shaft opening. The
shows the tensile zone created in the middle of the gallery area near
 Finite Element Method: Concepts and Applications in Geomechanics

Cc

e All dimensions are in m

Figure 5.15 Plan view of Seam III around the shaft opening.

concrete
pillar

= *

Pillar stress distribution Ii

Figure 5.17 Distribution of minor principal stress on pillars and concrete walls
for 20 GPa model.
| Chapter 5 Axisymmetric and 3D Finite Element Method [131 |

stress concentration factor around the pillar ranges from 1.5 to 3.5. The stress concentration
factor on the coal pillar hosting the shaft ranges from 1.0 to 2.0 which signifies that the support
given by the concrete wall is adequate. From this result, it is clear that additional support in
the form of roof bolting or truss is required to reduce the tensile zone in the roof.

Along the Line AA’

Figures 5.18 show the vertical displacement and the maximum stress concentration factor in
the roof inside the pillar along the line AA’. In this case five different curves are plotted to
compare the displacements depending on different elastic modulus of concrete wall. The
maximum average vertical displacement (20 GPa model) of 1.5 cm occurred in the middle of
the gallery. There is a positive displacement of the roof i.e. it means the penetration of the

= 08;

i =" a hi aT =i an T
= 6 aw =j

Vertical Or
(cm
displacement Sa 010) or 20-25 350.33 40° 45-50
Distance (m)

— 10 GPa — — 15 GPa- - - - 20 GPa— - - 25 GPa 30 GPa

(a) Vertical displacement of the roof along the line AA’

2| os
1
factor
concentration
principal
Maximum
stress jase oeeise 2025 —30-35-40" 45) |50
Distance (m)

—_10 GPa — —15 GPa- - - -20 GPa— - -25 GPa 30 GPa

(b) Maximum stress concentration factor in the roof along the line AA’
Figure 5.18

te wall. The average maximum stress

pillar into the roof along the corner of the concre
had occurred on the corner of the concrete
concentration factor (compression) is 4.48 which
ry of the shaft where the pillar ends.
pillar then the stress factor diminishes to 1.8 at the bounda
is around 1.88 and it becomes in-
The concentration factor in the other side of the shaft
stiffness or elastic modulus of concrete
situ value after about 34 m from the shaft edge. As the
around the corner.
walls increases higher concentration of stress occurs
 132 Finite Element Method: Concepts and Applications in Geomechanics

Along the Line BB’

Figures 5.19(a) and 5.19(b) show the vertical displacements and the maximum principal SCF
along the line BB’ respectively. The break in the curves shows the region of the shaft through
which this line passes. The maximum vertical displacement of 1.11 cm is at the junction of
the concrete pillars of both the galleries. For 20 GPa model, the maximum stress concentration
factor of about 2.9 and 2.4 occurred at both the corners of the concrete wall at a distance
between 26 to 30 m from the edge of the pillar. Since the study line BB’ is inside the pillars,
the vertical displacement and stress concentration factors are less than that of the study line
oe ~

oo
=|

S$

Vertical
displacement
(c
m)
0.5 10 15 20°25 30 35°40 45 50
Distance (m)
——10 GPa — — 15 GPa- - - :20 GPa— - «25 GPa— - -30 GPa
(a) Vertical displacement of the roof along the line BB’

34

NO

factor
concentration
Maximum
principal
stress 0+ T =) T T T ia os —T 4
0 So 1041520 5 0 es eAO 45 50
Distance (m)
10 GPa—-—--15 GPa
— —-20
-GPa~ - ~ -25 GPa 30 GPa
(b) Maximum stress concentration factor in the ro. f along the line BB’

Figure 5.19

By comparing results of the lines CC’ and BB’, it is clear

that the displacement and major
principal stress concentration factors have decreased from
the edge of the pillar to the inside.
Inthis case, it is also found that very
nominal stress concentration has occurred around
the shaft
lining. The horizontal displacements around the shaft
lining are found to be about 0.5 mm
which does not warren instability conditions.
| Chapter 5 Axisymmetric and 3D Finite Element Method [
133 |

SUMMARY
In this chapter, two different topics are discussed. Firstly, finite element formulation of
axisymmetric problems was explained giving suitable examples. Finite element formulations of
stiffness matrix, traction forces and body forces are discussed in detail. The tunnel example is
again solved using linear and quadratic quadrilateral elements assuming axissymmetric
condition and it is shown that the results are exact match for the plane strain elements as
mentioned in Chapter 4. In the second part, the concept of volume coordinates is elaborated
for three dimensional tetrahedral elements. Isoparametric formulation of linear and quadratic
tetrahedral element and that of hexahedral or brick type elements is also discussed in detail.
Determination of nodal traction forces and body forces is also illustrated with examples.
Finally, a real example of shaft pillar design is illustrated mentioning stability of coal pillars
near the shaft area.

EXERCISES
5.1 Based on Example 5.1 given in the text, evaluate the stiffness matrix. K of the
triangular element. Consider the coordinates of nodes as 1 (2, 1), 2 (4, 0) and 3 (3,
2). Also assume that E = 6 GPa and v = 0.

5.2 A 4-noded quadrilateral element is used for axisymmetric analysis as shown in

Figure P5.1. Determine the nodal force vector due to the surface traction pressure as
shown in the figure.
eeeol
same 3
i]

P, P,
i uy

I r 1 yy
| T?

Symmetry axis
Figure P5.1

5.3 An axisymmetric model of a circular tunnel is analysed using linear quadrilateral

element as shown in Figure P5.2. The element | is connected by nodes 1, 2, 3 and
4 whose coordinates are as follows:

Node x (m) y (m) u (mm) v (mm)

1 2 0 — 4,535 0
2 225 0 —4.461 0
3 DIS) 0.5 —4.461 0
4 2 0.5 — 4,535 0
 Finite Element Method: Concepts and Applications in Geomechanics

The displacements at each node are also given in the above table. Determine the stress
concentration factor at element 1. Assume E = 5.5 GPa and v = 0.25. A uniform pre-
mining stress (p,) of 8 MPa is applied on the outer boundary of the model (not shown
in the figure). Ignore internal pressure p;.

Figure P5.2

5.4 Find the integral / = |N,N2N3NadV for the linear tetrahedron whose nodal point
V
coordinates are given in Example 5.3.
5.5 Compute the integral [Ns dV considering the quadratic tetrahedron for the nodal
V
coordinates given in Example 5.3.
5.6 Determine the surface load vector based on traction pressure distribution on surface
1-2-3 of a linear tetrahedron as shown in Figure P5.3.
4 (3, 0, 3)

Figure P5.3
5.7 Compute the surface load vector of the quadratic tetrahedron considering the surface
pressure load as shown in Figure P5.3.
5.8 Determine the body force vector of a linear hexahedral element due to its own weight.
Assume the necessary parameters.
5.9 Determine b;;, by, and b33 at Gauss point r = -s = t = 1/V3 of a linear hexahedral
element as shown in Figure PS.4.

Ys (0;3.0) (3,3,0)
Figure P5.4
6.1. Introduction
properties of rock or
Load bearing capacity or limit of stresses is dependent on the material
uity. The limit
rock mass. Here, the term rock signifies the intact rock free from any discontin
laborator y in uniaxial, biaxial, triaxial, shear
of stress of a rock specimen can be tested in the
cases, rocks generally fail if the
and tensile mode. For uniaxial compression or tension loading
along the principal stress
applied stress exceeds its corresponding strengths. If the failure occurs
acts on that plane. The fracture
plane, it is called the extension of fractures since no shear stress
stress plane is develope d due to shear stresses and is called
plane(s) oriented other than principal
loading.
shear fracture(s). It is generally occurs in rocks during triaxial
ly or violentl y depending on the brittle
As mentioned in Chapter 2, rocks may fail gradual
the strain energy absorbed in the
or ductile nature, type of testing machine (soft or stiff) and
is rather complex and many times,
sample before failure. The failure mechanism of rock mass
Based on the displacements
it is almost impossible to identify the root cause of a failure.
d for possible failure of rock mass
induced by excavation, four conceptual models were propose
field in Figure 6.1(a) can be considered
as given in Figure 6.1 (Brady, 1987). The displacement
applied for such conditions. Figure 6.1(b)
continuous and thus elastic-plastic analysis can be
within the two such joints, the rock can
shows large discrete joints or plane of weakness but
near-field stress conditions can also
be considered continuous. Hence for this type of model,
frequently jointed around an excavation,
be regarded as elastic-plastic. If rock mass is
on the condition of joints and rigid body
displacements near the opening will be dependent stress
and rotation) [Figure 6.1(c)]. Application of elastic-plastic
movements (translation
Heavily-jointed rock mass can be regarded
analysis in such conditions may not be appropriate. joint
condition is termed as ubiquitous
as pseudo-continuous as shown in Figure 6.1(d). This
to define near-field stress conditions.
model and elastic-plastic analysis is appropriate
developed over the years and they
Various rock and rock mass failure criteria have been
a are developed based on the peak triaxial
are given in Tables 6.1 and 6.2. Most of these criteri the strength
on the failure surface. Thus, most of
compressive strength or peak shear strength

135
 136 Finite Element Method: Concepts and Applications in Geomechanics

Continuous
plane of
weakness

Joint
sets

Figure 6.1 Schematic models of rock structures (Brady, 1987).

Table 6.1 Failure criteria for intact rock (Edelbro, 2004)

Failure equation Development/comments Author, criterion

first published
(Of= 03)" =a+b(o; +03) An empirical generalization of Fairhurst (1964)
Griffith theory of intact rock
0] =0, +03 + Fo{ Empirical test data fitting for intact Hobbs (1964)
rock

Murrtel (1965)

Empirical curve fitting for intact rock Hoek (1968)

0] = 0, +a03 Triaxial tests on soft rock Bodonyi (1970)

0) =03+ c. G, a3)? Empirical curve fitting for 500 rock Franklin (1971)
specimens
0] = 03 +(mo,03 + sozyl/2 Application of Griffith theory and Hoek and Brown (1980)
empirical curve fitting. Both for intact
and heavily jointed rock masses

Empirical curve fitting for 700 rock Bieniawski (1974),

specimens. Both for intact and heavily modified by Yudhbir
jointed rock masses et al. (1983)
| Chapter 6 Rock and Rock Mass Failure Criteria 137

b
a 63
OT — 03" aos [<2 Applied to 80 rock samples Ramamurthy et al. (1985)

li}
cI Reem
0 30 =a Empirical curve fitting for both soil Johnston (1985)
and rock specimens
b
ze 1e3
PS o-(23) Both for intact and heavily jointed rock Balmer (1952),
t masses Sheorey et al. (1989)

1/B
= A 93
0] = 03 + AO, & rs ) A, B and S are strength parameters Yoshida (1990)
(&

: -——(bo7
COA+ 03)=9,
It reflects the effect of intermediate Yu et al. (1991),
principal stress on the strength of modified by Yu et al.
ioe 0) + Bo3 intact rock specimens. It is called (2002)
1+B Unified Strength criterion. The
B parameter b ranges from 0 to 1 Note: This criterion was
——(boy +07) =—Bo3 =0; depending on the effect of not reported in the
1+b intermediate stress. reference mentioned
es 6) + Bo3 in the title of the
1+B table.

B eile
om

where
Tm =(01 —03)/2 and o,, =(0) +03)/2

O, = major principal stress 03 = minor principal stress

01, = major normalized effective principal stress 6, = uniaxial tensile strength
0%, = minor normalized effective principal stress a ]= uniaxial compressive strength
and a, b, B, f, F, C, D, M, and @ are constants

criteria are expressed in terms of 0; and 63 and considered to be independent of intermediate

principal stress, Oy. In general, failure criteria are expressed using stress invariants so that
transformation of axes does not alter the criteria. Since principal stresses are one such stress
invariants, most of these criteria are expressed using principal stresses. However, computation
of magnitude and direction of principal stresses in three dimensions is cumbersome and can be
avoided if the yield or failure criteria are expressed using invariants of deviatoric stress. The
relationship between principal stresses and invariants of deviatoric stress is given in the
following In this chapter, the most common shear failure criteria such as Mohr-
section.
Coulomb, Drucker-Prager, Hoek-Brown, alternative criteria based on Hoek-Brown and Tensile
failure criteria will be discussed in detail.
 Finite Element Method: Concepts and Applications in Geomechanics

Table 6.2 Rock mass failure criteria (Edelbro, 2004)

Failure equation Development/comments. Author, criterion

first published

o a

Oj =03+0,; fm 2 4 2002 version Hoek-Brown (1980)

ci
ee
01 = AO, + Bog; (z A is dimensionless parameter and B is
ct a rock material constant, @ 1s
suggested = 0.65 Yudhbir (1983)

o; \™
01 =Ocm i— Use RMR7, value Sheorey et al. (1989)
tm
a
, , Og F
0} = 03 + 03B; ties 2001 version Ramamurthy (1995)
3

where
Oo, = major effective principal stress Oy, = tensile strength of rock mass
©; = minor effective principal stress Ocm = compressive strength of rock mass
and aj, b,,, m, are B; are constants O,; = uniaxial compressive strength of intact rock

6.2 Invariants of Deviatoric Stress

Invariants of stresses are scalar quantity and do not change with coordinate transformation.
These invariants are given in Eq. (1.44). However, it is convenient to use deviatoric stress
tensor for the formulation of rock or rock mass yield functions. Deviatoric stress tensor
measures the deviation of a tensor from the mean hydrostatic stress and is given as

ij = Oj — Om Oyj (6.1)
where the symbol s denotes deviatoric stress tensor, 6,; is the Kronecker delta which takes a
value of unity when i = j and zero when i # j, and o,, is the mean hydrostatic stress and
1
expressed as O,, = qe SOR ee ae Fant B
In Cartesian coordinate system, deviatoric stress tensor and principal deviatoric stress can
be expressed as

SY GOES 0.
Szz =Oz, —On,

Sy ae Say (6.2)
Syme ye
Seale
 Chapter 6 Rock and Rock Mass Failure Criteria
Then the invariants of deviatoric stress can be obtained by replacing stress tensor by
deviatoric stress tensor in Eq. (1.44) as

IT = Sixx TF Syy + Sez =0 (6.3a)

ag glee eee
J>5 =5 (si Sit 52, |+ 83 ise tis
(6.3b)

St1 aga sean 73 2 2 2 (6.3c)

AP ee 5 (sm + Syy +53,)- SxxSyz — SyySzx — SzzSzy + 28 yySye5zx

From the above relationship, invariants of principal deviatoric stresses can also be
obtained as
J, =8, +52 + $3 =) (6.4a)

1
p= s(t +55 +53) (6.4b)

Dies
J; = 3(si + $5 +53] (6.4c)

Now let us represent a state of stress in three perpendicular axes with the principal stresses
taken as rectangular coordinates as shown in Figure 6.2. Assume that the state of stress at any
point P in a body can be represented by a vector originating from the origin O and denoted

P(Oj, 02, 03)

Figure 6.2 Deviatoric stress.

by the line OP. Now imagine a line OA inclined to the three axes with equal direction cosines
as 1/,/3 . The stress vector OP can be resolved into vector OM on the line OA and another
vector ON perpendicular to the line OA. The magnitude of the vector OM can be estimated
considering projected length of OP along OA as
1 1
JOM| = (01 +02 +03) = 33m) = N30 (6.5)
 Finite Element Method: Concepts and Applications in Geomechanics

The vector ON represents the deviatoric stress from the mean stress o,, with components
(S}, 52, 53) and its magnitude can be estimated as

r=|ON|=4)5, +53 +8 =/2J, (6.6)

For any given state of stress, deviatoric stress vector will lie on the plane passing through
origin and perpendicular to the line OA. This plane is called deviatoric plane or x plane and
its equation is given as 0; + 06) + 63; = 0 in the principal stress space.
The component vector of OP along 6), G6) and o3 axes can be projected to deviatoric
plane to obtain components of deviatoric stress vector ON as shown in Figure 6.3. Now each
stress axis is inclined to the deviatoric plane by an angle sin !(1//3) in the original stress
space. Hence the projected length of 0; on the deviatoric plane axis can be computed as

O{ =0; cos(sin7!1/
2
V3)= [2or (6.7)

Similarly, projected length of 0, and 03 on the deviatoric plane are also found as

$= |?0
2302 o5=
3 [20
303 6.8
(6.8)

For example, if the value of principal stresses are —10.0, —5.0 and —1.0 MPa in the
original stress space (vector OP), then the components of the vector OP in deviatoric stress
space will be —8.165, —4.082, and —0.816 MPa respectively. We would like to quantify the
angle @ between the pure shear line and the deviatoric stress vector ON. Pure shear condition
is achieved when oO) = 1/2(0; + 03) and then o,, = GO). For this condition, the deviatoric stress
vector in the original stress space is expressed by {0, — o 0 0; — 07}. The uniaxial
compression is defined by the stress condition with o = 0, 0; = 0, and oO, < 0. Thus for
uniaxial compression condition, we find that o,, = 6,/3 and the deviatoric stress vector in the
original stress space is 1/3{20, — o, — o,}. Note that the principal stress axes can
be
interchanged to define pure shear and uniaxial compression conditions in other quadrants.
For
example, we may also define uniaxial compression condition along —O> axis considering
o; =
0, 03 = 0 and so on. In the present configuration (Figure 6.3), the deviatoric stress vector
for

-6;

Pure shear line

Figure 6.3 Projection of principal stresses in deviatoric plane.
| Chapter 6 Rock and Rock Mass Failure Criteria

uniaxial compression will be oriented along —o, axis and the pure shear condition will be
inclined by 30° from the —o, axis. Hence the lengths OB and NB can be computed as

0; — 03
OB = 0; cos30° — 03 cos 30° = =rcosé (6.9a)
a
Ley Oy =
NB = G4 —6/sin30°
—04sin30° = 2-1 “3 = rsin@ (6.9b)
V6
By manipulating Eqs. (6.9a) and (6.9b), we find the lode parameter which ranges from
0 to +1 while @ varies from 0 to $ 30 degrees and is expressed as

207 — 0; — 03 VB
x = ___=
— 4—— = —/3 tan0 (6.10)

The angle @ is called lode angle and is considered as another invariant of deviatoric stress
relating second and third invariants of deviatoric stress, namely Jz and J3 respectively, as
(Chakrabarty, 1987)

sin30 ==
33J3
7 B /2 (6.11)

For pt = 0, pure shear condition is achieved. A state of stress is equivalent to uniaxial

compression for # = 1 or 6 = —30°. Using deviatoric stress notations, Eqs. (6.9a) and (6.9b)
become
S| ~ 53 = V2r-cosé (6.12a)

pe
S$, +83 =—S2 =- iY (6.12b)

Hence replacing s; with o; and r with /2J> , Eqs. (6.12a) and (6.12b) are solved to
g
obtain the relationship between principal stresses and invariants of deviatoric stress. Considerin
0, < 0 < 03, we find that

07} Lene, - Sl /6+0 ) 1

O27= B —sin@ eres | (6.13)
03 cos(z/6 —@) 1

l stresses in MPa as
Example 6.1 Plot the deviatoric stress vector ON on z-plot for principa
0; = -8, 0) = —4 and Oo; = -2.
three
Solution: In a white paper, draw horizontal (x) and vertical (y) axes. Draw
will be aligned with positive directio n
principal axes as shown in Figure 6.3. Note that —o3 axis
 = Finite Element Method: Concepts and Applications in Geomechanics |

of y axis, i.e. vertical upward direction. The —o; axis is inclined 150° and 30° clockwise
direction from the positive x axis. Mark each principal axis with equal grading, e.g. 1, 2, 3,
... MPa. Now, draw a line from the origin to a distance of /2/30, along the —o, axis. From
the end point, draw another line of distance J2/305 parallel to the —o, axis. Draw another
line from the end point of last line to a distance J2/303 parallel to the —03 axis. Join the origin
and the last point by a line vector. This is the deviatoric stress vector on the z-plane.
Alternatively, Excel software can also be used to plot the same as shown below:

Cohesion = 2 MPa, Friction angle = 30°

—03

Pure shear line

6.3 Mohr-Coulomb (MC) Yield Criterion

Mohr-Coulomb shear failure criterion was first proposed by Coulomb in 1773. It was suggested
that shear stress developed in rocks is related to cohesion, angle of internal friction of the
material and is also dependent on the applied normal stress. In the limiting case, when the
magnitude of shear stress is equal to the shear strength, the criterion reduces to

|t|=c-o, tang (6.14)

where
T = shear stress, N/m’, absolute value will be considered
c = cohesion in N/m?
O,,= normal stress in N/m? and negative value is considered (compressive)
@ = angle of internal friction, degrees
Equation (6.14) suggests that shear stress at failure is linearly related with
the normal
stress. In 1900, Mohr proposed that shear failure takes place across a plane where
shear stress
is a function of normal stress. He added that the shear stress at failure could be
linearly related
with normal stress or could be any polynomial function.

6.3.1 MC Criterion in Principal Stress Space—Two Dimensions

Consider that a rock sample is failed at stress o; for a confining stress
of 03. Using the concept
| Chapter 6 Rock and Rock Mass Failure Criteria 143

outlined in Chapter 1, we can plot a Mohr circle of stress in T—o, plane as shown in
Figure 6.4. It is postulated that at failure, Coulomb criterion makes a tangent with the Mohr’s
circle at the failure stress point (6, 7). Thus, Coulomb’s failure criterion is drawn to show the
shear and normal stresses at failure. The definition of cohesion and angle of internal friction
is also given in the figure. From the figure, it can be found that normal and shear stress at
failure is related to the principal stresses as given below

0, +0
o, =(142
5 * |-reos2e (6.15a)

Tt =rsin2a (6.15b)
where r is the radius of the Mohr circle and absolute value is considered given in the following
equation:

03-0 (6.16)
pense

Failed zone
Mohr-Coulomb
envelop ~__»

Safe zone

Figure 6.4 Mohr-Coulomb yield criterion.

At failure, the angle 2@ equals to 90° + and thus Eqs. (6.15a) and (6.15b) become

O, =(S2%),( 252: ling (6.17a)

= [23-21 Joos (6.17b)

Substituting Eq. (6.17) into Eq. (6.14), we get Mohr—Coulomb failure criterion in
6, —6; plane as given in Eq. (6.18) and shown in Figure 6.5.
 |
Finite Element Method: Concepts and Applications in Geomechanics

/ nek IF sings 0 y)
where Ng is called triaxial factor and computed as Ng = ay, = taniaiee chs

03

0; = O3

Yield/failed zone

Tensile cut-off

Safe zone Infeasible zone

Figure 6.5 Mohr-Coulomb failure envelop in 0; — 0; plane.

If the value of F is positive or zero, rock yields/fails under the specified stress condition.
For a negative value of F, rock is safe. From Eq. (6.18), the theoretical uniaxial compressive
strength of rock can be obtained by considering 03 = 0 as

O, =0, =-2c INg (6.19)

The theoretical maximum tensile strength is estimated by letting 0, = 0 in Eq. (6.18) and
the magnitude is given as

max 2c
O; =
JNo (6.20)

However, the experimental or measured tensile strength o, of rock is in general less than
the theoretical tensile strength. In this case, 0, will be considered as tensile cut-off and thus
the failure envelop will be truncated as shown in Figure 6.5.

6.3.2 MC Criterion in Deviatoric Plane

In this section, Mohr-Coulomb shear yielding envelop will be reformulated using invariants
of deviatoric stress. In Eq. (6.18), we can replace 0, and 0; based on Eq. (6.13) as given
below:

2, J>5 TT 2 J>5 1
F= ti co(#0) 05 4]ss[ Evo} ta |My2 Ng =0 (6.21)
 Chapter 6 Rock and Rock Mass Failure Criteria ae 145

Using trigonometric relationship and then by rearranging the above equation, we get

F = JFa|(1+1Ng)eos0 + (N, -1) 9"|

+0 (Ny -1)-2c [N, =0 (6.22)
The above equation represents Mohr-Coulomb yield/failure criterion in deviatoric stress
space. This criterion is dependent on all three invariants of deviatoric stress. In this case also,
if the value of F in Eq. (6.22) is positive or zero, rock yields/fails under the specified stress
condition. Else, the rock is safe. Figure 6.6 shows a typical plot of this yield/failure surface
in principal stress space having 0, < o> < 03. It can be seen that Mohr-Coulomb criterion gives

—O3 Drucker-Prager
(outer)
Hydrostatic axis

Deviatoric plane
(6,+0>+03=const) -

3c cot ¢ Been oiy Sie (oe

v3 Mohr-Coulomb

Figure 6.6 Mohr-Coulomb yield surface in principal stress space.

a conical yield surface whose normal section at any point is an irregular hexagon The conical
shape (other than cylindrical shape) of yield criterion is the manifestation of the influence of
hydrostatic stress on yielding. This is clearly seen by the second term of Eq. (6.22). On the
hydrostatic axis, that is for 0, = 0) = 03 = O,,, second invariant of deviatoric stress,
J, = 0. For this case, Eq. (6.22) gives 0; = c cot @, which signifies that the apex of the conical
yield surface lies at a point where 0, = 0> = 03 = c cot @. Then the distance of the apex from
the origin will be ./3c cot @.
Figure 6.7 shows the two-dimensional plot of Mohr-Coulomb yield surface in deviatoric
plane. For the similar stress level, the shape and size of hexagonal yield surface depends on
cohesion and angle of internal friction of rock.

Example 6.3 Referring to Example 1.4, decide whether the rock fails based on
Mohr-Coulomb criterion. Assume that cohesion and friction angle of rock are 1.5 MPa and
25 degrees respectively.
 Concepts and Applications in Geomechanics
Finite Element Method:

Mohr-Coulomb

Drucker-Prager (outer)

Drucker-Prager (inner)

Figure 6.7. Mohr-Coulomb yield surface in deviatoric plane.

—20.92
Solution: The major, intermediate and minor principal stresses are —37.83 MPa,
MPa and —4.65 MPa respectively. From the data, the following parameters can be estimated:

1+sin25
3 areare ey Oy = -21.13, Jp =16.59, J; = -58.71, 8 = -0.0107 rad

Hence using Eq. (6.22), we find that

F =r (+ Ng )cos6 + (Ng -1) 2)+o (Ng -1)-2c,[Ng =21.68

sin @

Since the value is positive, rock yields/fails under this stress condition.

6.4 Drucker-Prager Criterion

Drucker-Prager yield criterion (1952) is an approximation of Mohr-Coulomb criterion and is
derived using invariants of deviatoric stress. Drucker-Prager yield surfaces resembles circle in
deviatoric plane and are obtained by coinciding outer and inner apices of Mohr-Coulomb
hexagon as shown in Figure 6.7. Drucker-Prager yield surfaces are circular conical pyramid and
the distance of apex from the origin is also V3ccot@ . The loci of the circles, coinciding outer
and inner apices of Mohr-Coulomb hexagon, can be obtained by replacing @ with —30° or
~30° respectively in Eq. (6.22) and are given below:

[J
Fe a (2+Ng ee (Ne — 1)—2c Noa 0, coinciding with outer apices (6.23a)

[J mers oe ae
F= (1425) + on (Ng -1)-2c Ng =0; coinciding with inner apices (6.23b)
| Chapter 6 Rock and Rock Mass Failure Criteria | 147

The above equations can be written in compact form as

F = 01/J + Bon, —2c,[Ny =0 (6.24)

where q@ and f are given in Table 6.3.

Table 6.3 Parameters of Drucker-Prager criterion

Drucker-Prager criterion a B

2+Nyg
Outer Ry Ng -1

1+2Ny
Inner eg N.oo ]

Example 6.4 Recalculate Example 6.3 based on Drucker-Prager yield criterion. In this case,
consider the cohesion of rock as 4 MPa with the same friction angle.

Solution: Taking the calculated stress-dependent parameters from Example 6.1, we can
estimate the functional values given in Eqs. (6.23a) and (6.23b) as

J2(0+%)+o. (V4 -1)-2c. [Ng =-0.68

and

Zt +2Ng)+Om (Ng -1)—2c,[Ng = 13.30

Hence the rock fails if we consider the yield surface which coincide inner apices and it
is safe for the yield surface coinciding outer apices.

6.5 Hoek-Brown (HB) Criterion

The Hoek-Brown yield criterion was first introduced in early eighties. Hoek-Brown (1980)
found that the peak triaxial compressive strength of isotropic rock is related to confining stress
based on rock type designated by m; (Brady and Brown, 1985). This relationship is given
below:

0) = 03 +(m;030,; + a2, (6.25)

where 0; and 0; are the peak triaxial compressive strength and confining stress respectively and
0,; is the uniaxial compressive strength of intact rock. The parameter m; depends on the rock
type. This criterion was then extended for rock mass and another parameter namely s was
 Finite Element Method: Concepts and Applications in Geomechanics

introduced in the equation along with reduced value of m; designated by m. The modified
failure criterion for rock mass is given as:

0, =03 +/mo30,; + 50%; (6.26)

The parameters m and s are constants depending on rock type and the extent to which
it had broken before being subjected to failure stresses. It was suggested that m and s are related
to Bieniawski’s Rock Mass Rating (RMR) and equations for estimating m and s are proposed
(Hoek and Brown, 1980). Since then this criterion was widely used in geotechnical and rock
engineering problems. The strengths and limitations of this criterion were again proposed in
an update in 1988 (Hoek and Brown, 1988). It was then proposed that the correct use of this
criterion should be decided based on the nature of rock mass which does not contain any
structural discontinuities and the volume of rock under consideration might contain four or
more closely spaced and almost uniform discontinuity sets. In 1997, a generalized Hoek-Brown
failure criterion was published for accommodating hard and soft rock masses (Hoek and Brown,
1997). In that paper, the use of Geological Strength Index (GSI) was proposed instead of
Bieniawski’s Rock Mass Rating (RMR) since the estimation of RMR for very weak rock masses
is cumbersome and in some cases, it is too difficult to evaluate. The GSI lies in the range of
0 to 100 as in the case of RMR and is calculated from charts based on the quality of the rock
structure and the condition of the rock surfaces (Hoek and Brown, 1997) as shown in
Figure 6.8. There are mainly two parameters in GSI system—surface conditions and structure
of rock mass. By observing the rock mass in the field, one can circle around the observation
in the chart as shown in the figure. Then average value of GSI within the circle is evaluated.
For the example shown in the figure, the GSI will be 35. The generalized criterion is again revised
in the year 2002 and is the latest update of Hoek-Brown failure criterion (Hoek et al., 2002).
The generalized Hoek-Brown rock mass failure criterion for drained condition is proposed
as follows (Hoek et al., 2002):
a
oO
Poo, 0; + Or, [myS| =0 (6.27)
Oci

where
O|,3 are principal stresses and should be arranged as 0; S 0> S 3 (considering compressive
stress negative)
O,; = uniaxial compressive strength of intact rock (negative value will be considered)
m, is reduced material constant based on intact property of rock, m; and s and a are rock
mass properties based on GSI as given below:

Ro ba eaTet a |
area (6.28a)

pa he ere
ply oben (6.28b)

eaime¢lexP(-GSI/I5) — exp(—20/3)] (6.28c)

 Chapter 6 Rock and Rock Mass Failure Criteria we]

-| Intact or massive: intact Pet

Y
rock specimens or massive
in situ rock with few widely
_| spaced discontinuities

*] Block: well interlocked

“| undisturbed rock mass consisting
of cubical blocks formed by three Ny
“| intersecting discontinuity sets as

‘| Very blocky: interlocked,

partially disturbed mass with
‘| multi-faceted angular blocks
‘| formed by 4 or more joint sets
<a] Blocky/disturbed/seamy: folded
/ with angular blocks formed by
many intersecting discontinuity
sets. Persistence of peeding. Hy
Hifb
_planes_or-schistosi ee
“¢e1 Disintegrated: poorly inter- aA
: locked, heavily broken rock
=< mass with mixture of angular rock
of
interlocking
Decreasing
ST
eee] and rounded rock pieces

~=| Laminated/sheared: lack of

=~ | blockiness due to close spacing
~ "| of weak schistosity or shear
(LL {{4] planes

Figure 6.8 Chart for estimating GSI.

The range of m; is between 5 to 40. Table 6.4 provides some typical values of m; for
different rock types. In Eqs. (6.28a) and (6.28b), D is a factor ranging from 0 to | and depends
on the degree of disturbances due to blasting and stress relaxation (Hoek et al., 1997). For hard
rock and tunnelling without blasting, the value of D is assumed as 0. For tunnelling in poor
rock with blasting D = 0.8 and for large open pit slopes with huge blasting, D can be
considered as 1.0.

6.5.1 HB Criterion in Principal Plane—Two Dimensions

Equation (6.27) is plotted in 0; — 0; plane assuming three different rock mass strength
properties as shown in Figure 6.9. In this figure, three rock mass types are assumed resembling:
(i) Strong: GSI = 100, m; = 25 ando,; = 100 MPa
(ii) Moderately strong: GSI = 50, m; = 10 and o,; = 30 MPa
(iii) Weak: GSI = 25, m; = 7 and o,; = 10 MPa
| 150 Finite Element Method: Concepts and Applications in Geomechanics

Table 6.4 Typical m; values of various rock types

Rock type mj; Rock type m; Rock type m;

Sedimentary Igneous Metamorphic

Anhydrite iVeey Agglomerate 19+3 Amphybolites 26+6
Breccias 20+2 Andesite 25+5 Gneiss 28+5
Chalk 7+2 Basalt ASSES, Hornfels 19+4
Claystones 4+2 Breccia 19+5 Marbles 9+3
Conglomerates 21+3 Dacite 25+3 Metasandstone 19+3
Crystalline Limestones 1243 Diabase fo=5 Migmatite DOES
Dolomite 9+3 Diorite Uae) Phyllites =D
Greywackes 18+3 Dolerite 16+5 Quartzite 20+3
Gypsums 12+2 Gabbro DES Schists 10+3
Marls 7+2 Granite 32+3 Slates 7+4
Micritic Limestones 843 Granodiorite 29+3
Sandstones 17+4 Norite 20+5
Shales 6+2 Obsidian 19+3
Siltstones 7+2 Peridotite Dict
Sparitic Limestones 10+5 Porphyries 20+5
Rhyolite 25+5
Tuff BEES)

O07; (MPa)

400 -350 -300 -250 -200 -150 -100 -50 0

tL 1 aa 5
enews! ie — 1 a =e 10

5)
0
ot Ss

-10
Paice
x * stl. =
GSI = 100, m; = 25, ee Mit lige Adydona dl 8
O,; = -100 MPa Dake AON ES ep bose
eteoiaedl at
/ ; -35
GSL= 25..m ja) 4 eas
Oc; = -10 MPa L—45

Figure 6.9 Hoek-Brown shear yield envelop.

In general, the relationship between 0, and @; is non-linear. However, for stronger rock
mass, it is almost linear for lower confining stress 03.
The compressive strength of rock mass is evaluated by replacing 03 = 0 in Eq. (6.27) as
given below:

Om
a
> Ou; (s )
a
(6.29)
Similarly, tensile strength of rock mass 6,,, is estimated by letting o, = 0 in Eq. (6.27)
and then solving 63 iteratively as given in the following equation:
 Chapter 6 Rock and Rock Mass Failure Criteria [ 151 |

(GO
Mp
(Gs) [2 os—s=0 (6.30)
Cl

Alternatively, for biaxial stress field having 0, = 03, the tensile strength can also be
evaluated from

Om = L633)

6.5.2 HB Criterion in Deviatoric Plane

As shown for Mohr-Coulomb criterion, Hoek-Brown yield surface [Eq. (6.27)] can also be
derived using invariants of deviatoric stress as given in the following equation:

F =a fJn)"" + BYJy +m, Om +50q3 =0 (6.32)

where the expression for parameters @ and B are given in Table 6.5. These two parameters are
dependent on the type of rock mass and lode angle. It is clear from Eq. (6.32) that the

Table 6.5 Parameters of Hoek-Brown yield functions

a B
Yield criterion

-1)/, r) sin@
Hoek-Brown (2cos6)'1/“(-o,,;) my |COS Peas

mp
:
Alternative 1 G3)? Ga? is

aisy (-0; \G4w ack

Alternative 2

generalized Hoek-Brown yield function is dependent on all three invariants of deviatoric stress.
The yield function is plotted in three mutually perpendicular principal stress axes (0; < 07 S 03)
as shown in Figure 6.10. In principal stress space, the yield locus is conical whose normal
section at any point is an irregular hexagon just as Mohr-Coulomb yield surface. Again the
conical shape of this yield surface is the evidence that hydrostatic stress does influence the
yielding as shown in the third term of Eq. (6.32). In addition, the apex of the hexagonal
pyramid can be obtained when J, = 0 with a mean hydrostatic stress, 0,, = —s0,,/m, and hence
the distance of the apex from the origin will be V3s0,;/my.
Figure 6.11 shows the two-dimensional, deviatoric or 7 plane (0; + 02 + 63) = 0),
representing of the generalized Hoek-Brown yield surface. The yield surface is convex and has
six sharp corners or points. In all other points, the yield surface is regular. The shape and size
 152 re Finite Element Method: Concepts and Applications in Geomechanics

=O; Alternative |

Hydrostatic axis
(0, = 02 = 03)
Deviatoric plane
+ 0) + 03 = const.)

Figure 6.10 Generalized Hoek-Brown yield surface in principal stress space.

Generalized Hoek and Brown

Alternative 2

Pure shear line

Alternative 1

Figure 6.11 Yield surfaces in deviatoric plane.

of the yield surface varies with rock mass parameters, GSI, m; and o,;. Figure 6.12 shows two
yield surfaces for moderately weak rock mass with GSI = 50, m; = 10, o,; = —30 MPa and
strong rock mass with GSI = 100, m; = 25, o,; = —100 MPa for the same state of stress. It
can be seen that in case of strong rock mass, the yield surface become flatter when @ = 30°
for relatively low stress regime. This confirms the fact that for stronger rock mass, the
relationship between o; and 03 in Eq. (6.27) is almost linear for lower range of 03.
 Chapter 6 Rock and Rock Mass Failure Criteria a [ 153 |

— GSI
= 50 —— GSI = 100

Figure 6.12 Yield surfaces for different parameters of rock mass.

Example 6.5 Determine the rock mass strength of limestone for GSI = 56, o,; = 90 MPa and
m; = TZ.

Solution: Rock mass strength is estimated from Eq. (6.29) (compressive) and
Eq. (6.31) (tensile). The Hoek and Brown parameters are:

m, =12 X exp a = 2.49

s= exp|S100?) = 0.00753

ae exp -%2|-exp 4 = 0.995

6 is 3

Hence the rock mass strength (compressive) will be o,, = 90 x 0.00753°>% = 4.9 MPa
and the magnitude of tensile strength is o,, = 0.00753 x 90/2.49 = 0.272 MPa.

6.6 Alternative Criteria Based on Hoek-Brown Yield

Surface
the
Hoek-Brown yield surface can be approximated using a circular conical pyramid where
wn hexagon (Figure 6.11) just as
circle can coincide with outer or inner apices of Hoek-Bro
 154 Finite Element Method: Concepts and Applications in Geomechanics

Drucker-Prager criterion is for Mohr-Coulomb yield surface. From now onwards, the yield
surfaces coinciding outer and inner apices will be termed as Alternative | and Alternative 2
yield surfaces respectively. These alternative yield surfaces are developed based on Hoek-
Brown strength parameters (,, s, @ and o;,;) and can also be used as independent yield
functions (Deb et al., 2005). The approximated yield surfaces for alternatives can be written
using Eq. (6.33), where the parameters @ and B are given in Table 6.4.

PeGoy. (ou) Ss (Jo) ae +B Ia tm Fn +9 =Q when Ox—7 (outer)

(6.33a)

F=(y3)/4 Cot (Jo) l/a 2p

Ia tO n +so,; =0 when @ = 3 (inner)

(6.33b)
It can be seen that alternative yield criteria are functions of J, and o,, only. Hence these
yield functions can be plotted in o,,-\/J, plane. Figures 6.13(a) and 6.13(b) show’ the
graphical representation of these functions plotted using three different rock mass strength
parameters. This relationship is comparable with Eq. (6.27) if plotted in o)-03 plane
(Figure 6.9). For Alternative 2 yield criterion, the deviatoric stress is smaller as compared to
Alternative 1 yield surface for the same mean hydrostatic stress since the safe zone covered by
the former is smaller than the latter. The difference between deviatoric stresses for Alternative
1 and Alternative 2 is significant (almost twice as much as Alternative 2), especially in case
of stronger rock mass.

80 -
GSI = 100, m; = 25 r
0; = -100 MPa 70+ §

60 | -=
S
3
50 Se
ce
+ ES
5 >

3005 te
GSI = 50, m, = 10, z
0,; = -30 MPa S
YN

GSIg— 255 1a — a
0,; = -—10 MPa Pe ee a

-50 —40 -30 -20 -10 0 10

Mean hydrostatic stress, o,, (MPa)

(a) Alternative 1 yield function in o,,-VI5 plane

| Chapter 6 Rock and Rock Mass Failure Criteria [ 155 |

80|
GSE=" 100) m, = 25
0,; = —100 MPa 70 |

60|

50 [
GSI = 50, m, = 10,
On; = -30 MPa 40 VJ,
(MPa)

30|
GSh 25, m7:
Og = -10 MPa Second
deviatoric
of
invariant
stress,
—

f T eee)

-50 —40 -30 —20 -10 0 10

Mean hydrostatic stress, o,, (MPa)
(b) Alternative 2 yield function in o,, AUT plane

Figure 6.13

6.7 Tensile Yield Criterion

Rocks are in general weak in tension. Tensile stress can extend crack growth or cause complete
failure. Griffith’s crack extension theory postulates that the elliptical crack will propagate from
the point of maximum tensile stress concentration (Brady and Brown, 1985). The following
criterion was given for crack extension in plane compression:

F! =(0, -03)° +80,(6, +03)=0 when 0) +303 <0 (6.34a)

F' =03-0,=0 when 0; +303 >0 (6.34b)
where 0, is the tensile strength of rock. In rock engineering problems, Eq. (6.34b) is
extensively applied with numerical modelling for both tension and no-tension (0, = 0) material.
Hence rocks will be considered in yielded or failed state when minor principal stress exceeds
tensile strength of the rock. Expanding 03 using invariants of deviatoric stress, we can rewrite
Eq. (6.34b) to generalize tensile failure criterion as

sin@
Fi = J [cos +0? |4.0 -0; =0 (6.35)
| 156 | Finite Element Method: Concepts and Applications in Geomechanics

SUMMARY
Over the years, several rock and rock mass criteria are developed to characterize the onset of
yielding or failure. For applications in rock or soil medium, Mohr-Coulomb, Drucker-Prager,
and Hoek-Brown criteria are popularly used. In this chapter, it is shown that these criteria can
be expressed in principal stress plane as well as with invariants of deviatoric stresses. For three-
dimensional analysis, it is cumbersome to estimate the magnitude and direction of principal
stresses and hence yielding criterion expressed with invariants of deviatoric stresses are more
practical. Concepts and derivations of above mentioned yield criteria are explained using
principal stresses and invariants of deviatoric stresses wherever applicable. The relationships
between principal stresses and invariants of deviatoric stresses are also established. Alternative
criteria based on Hoek-Brown yield surfaces are developed and illustrated with example.
Finally, the concept of tensile yield criterion of rock material is explained.

EXERCISES
6.1 A series of triaxial test of sandstone rock reveals the following relationship 0, — 303
+ 75 = 0 (all stress units are in MPa).
(a) Determine the cohesion and angle of internal friction of sandstone.
(b) If the state of stress at a point is expressed as 0; = —1.5 MPa and o, = —80 MPa,
determine whether the point is safe.
6.2 Stress tensors at a point are given by

3 =e 0
1-60 | MPa
00 e"=14

If the cohesion and angle of internal friction of rock are 3 MPa and 35° respectively,
using Mohr-Coulomb criterion, determine whether the point is safe.
6.3 Solve Problem 6.2 using Drucker-Prager criterion for both inner and outer circles.
6.4 Draw the deviatoric stress vector on the z-plane and compute the Lode angle using the
data given in Problem 6.2.
6.5 Show that for no-friction material, Mohr-coulamb yield surface becomes a regular
hexagon.
6.6 Data obtained from a series of triaxial test of shale rock is given below:

03 (MPa) 0, (MPa)
0 —25
—0.5 —28
—1.0 —33
—1.5 —37
—2.0 —40
—3.0 —49
4.0 2)
| Chapter 6 Rock and Rock Mass Failure Criteria =a 157

(a) Establish Mohr-Coulomb criterion based on the above data.

(b) Establish Drucker-Prager envelop based on the above data.
(c) Determine cohesion and angle of internal friction of the shale.

6.7 Show that ratio of radius of outer circle to inner circle of Drucker-Prager yield surface
I 2N5
1S ‘
2+Nz

6.8 The GSI of a limestone strata is estimated to be 56. Measurement of pillar stresses in
the limestone bed shows o; = —-8.0 MPa, 0) = -5.0 MPa, and o; = 1.0 MPa.
Considering m; and uniaxial compressive strength of limestone to be 20 and 40 MPa
respectively, determine whether the pillar is safe.
6.9 Solve Problem 6.8 considering alternative criteria.
6.10 Solve Problem 6.2, considering Hoek and Brown criterion. Assume m; and GSI of
shale as 12 and 42 respectively.
 Dele Place bande
Element Analysis

7.1. Introduction
The finite element analysis of linear elastic material is presented in Chapters 3-5. In Chapter 6,
rock and rock mass yield/failure criteria are illustrated based on principal stress and invariants
of deviatoric stresses. Rocks subjected to large loads can yield and/or fail causing irreversible/
permanent plastic strain. In such condition, rocks behave non-linearly. This is popularly termed
as “material non-linearity” condition in theory of plasticity. For rocks and rock mass, material
non-linearity is too large to be ignored. Using the concept of material non-linearity, post-failure
characteristics of rock mass, opening, sticking or sliping of joint plane(s), dilation behaviour
of rock are analyzed. In origin, classical plasticity theory was developed based on continuum
theory to predict the behaviour of metals under loads exceeding elastic limit. Similar concepts
are being used to model irreversible deformation in concrete, soils, rocks, ceramics and other
materials. In this chapter, we will describe the concept of plasticity and implement the yield
criteria described in the previous chapter into the finite element procedures using plasticity
theory.
In plasticity, behaviour of solids can be analysed in two ways—time-independent
plasticity and time-dependent plasticity. For time-independent problems, F = Kq denotes linear
problem when stiffness matrix K and/or force vector F are independent of displacement vector
q and non-linear when K and/or F are dependent on q. In this chapter, time-independent non-
linearity will be discussed. Discussion on time-dependent plasticity is beyond the scope of this
text. We will first discuss various solution methods for solving non-linear set of equations.
Then concepts and applications of general plasticity theory will be illustrated with respect to
rock-like material followed by the implementation of rock mass yield criteria in finite element
procedure.

7.2 Non-Linear Solution Methods

There are various techniques for solving a set of non-linear equations. The most popular
[ Chapter 7 Elastic-Plastic Finite Element Analysis
| 159

schemes which are used in finite element analysis are Newton—Raphson

(N-R), modified
N-R methods and incremental or Euler’s method. These schemes are illustrated further.

7.2.1. Newton-Raphson Method.

This is the most frequently used method for solving non-linear system
of equations. This
method uses iterative technique to achieve the required solution. Consider a simple
example of
a loaded spring in one dimension as shown in Figure 7.1. The spring is attached
at one end
and load P is applied at the other.

je \)V\- P
Figure 7.1 A non-linear loaded spring.

Here, we assume that the spring constant k is an arbitrary function of the displacement
u and thus it is dependent on the displacement u. The load P = P(w) can be written with the
expression as
P(u)=k(u)u (71)
Thus the spring constant k at any arbitrary displacement u* can be written as

Palak wil
ku )= rae (7.2)
Let us assume that the spring is displaced to u, when the load P, is applied. Now, we seek
a solution of ug when the load is increased from P, to Pg as shown in Figure 7.2. A Taylor

Pp
Py
P

Pa

u
UA uy Un Up

Figure 7.2 N-R solution scheme.

 in Geomechanics
[ 160 || Finite Element Method: Concepts and Applications

= ua can be written by neglecting

series expansion of force function P around the point wu
higher order terms as
dP
P(u) = Plug) + =| (u-Uug) (7.3a)
u WA

Pe SP hy) Au (7.3b)
Au = u, — Ua, Where uy
From the above equations, we can solve the displacement increment,
is the new displacement once the stiffness at u, is known. Now, the force at uw, will be
P, = P(u;). We recalculate spring constant at displacement u; using Eq. (7.2) and then seek
for displacement uy as Pg — P, = k(u;)Au, where Au = uz — uy. This process continues until
the difference between Pp and P,, (n = 1, 2, ...) is very small.

on
Example 7.1 From the test of a non-linear spring, it is found that load P depends
ent u, = 0.01 unit, find the
displacement u as P = 5000Vu. Assuming initial displacem
displacement if load increment is 500 units.

Solution: Here, we seek for the displacement uz for a load increment of 500 units. Using
Eq. (7.3), we can estimate displacement uj, =u, +500,.fu, /2500 = 0.03 unit. Then P, = 866.025
units. Using similar approach, u. = 0.03928, P; = 990.959 units, u3 = 0.03999, P3 = 999.875
units and uy = 0.04 units, Py = 1000 units. Thus the solution converges within
4 iterations.

7.2.2 Modified Newton—Raphson Method

In the Newton—Raphson method, the major computational expense lies in updating stiffness
parameter after every iterative cycle. As shown in the above example, the spring constant is
updated after every iteration. For multidimensional problems with large number of degrees of
freedoms, the stiffnes parameter is replaced by stiffness matrix. Thus, computational burden for
updating stiffness matrix becomes expensive and time consuming. In order to avoid repeated
updating of stiffness «matrix, modified Newton—Raphson method can be adopted. In this
method, stiffness matrix is updated after each load step and it is kept same for every iteration
within the load step as shown in Figure 7.3. It is evident that in this case more number of
iterations is needed to achieve the desired solution.
One extreme condition of modified N—R scheme is to use the initial stiffness parameter,
denoted by k° = k(u = 0), in Eq. (7.3) for spring example as

Py ~ Py =k Au; (7.4)
All other conditions as mentioned in sub-section 7.2.1 remain the same. Thus, stiffness
value is formulated only once. Force and displacement are updated after each iteration for every
load steps. This method is called initial stress method and is computationally inexpensive for
problems having multi-degree of freedom. However, accuracy of the desired solution can only
be achieved with large number of load steps. Since this method corresponds to the linearization
 | Chapter 7 Elastic-Plastic Finite Element Analysis
onl (2161 3
P

u
UA uy to) ug

Figure 7.3 Modified N-R solution scheme.

of the response about initial configuration of the finite element system, it may results in very
slow and divergent solution (Bathe, 1982).

Example 7.2 Solve Example 7.1 using modified N-R method.

Solution:
Using modified Newton—Raphson scheme, the above example is solved by
keeping the same spring stiffness of 2500/ wie as at displacement u, for every iteration. In this
Case, since stiffness values are not updated, 17 iterations are needed to achieve the desired
solution as shown in the following table:

u; P; U; P; Ui; P; U; P;

0.01000 500.000 0.03945 993.098 0.03998 999.786 0.039999 999.993

0.03000 866.025 0.03973 996.567 0.03999 999.893 0.040000 999.997
0.03536 940.199 0.03986 998.288 0.04000 999.947
0.03775 971.481 0.03993 999.145 0.04000 SOS:
0.03889 986.050 0.03997 999.573 0.04000 999.987

7.2.3 Incremental Method or Euler’s Method

Incremental solution method is used to solve first order differential equations. The major
difference with iterative solution schemes such as N—R scheme is that in incremental method,
the load is incremented after each computational cycle and new displacement is sought. In other
words, after load increment, displacement is calculated and updated based on the tangent
stiffness matrix. On the contrary, in N-R method, after each load increment iterative procedure
is applied to estimate updated displacement after each iteration. Then the load is incremented
 Finite Element Method: Concepts and Applications in Geomechanics

the non-linear
again and this process continues until total load is applied. Again assuming
spring as shown in Figure 7.1, we can write tangent stiffness, k, from Eq. (7.5) as

P= f(u) (7.5)
dP
k, = = (7.6)

Assuming the starting point as P = 0 at u = 0, we can write the displacement after first
load increment AP, as
AR
uy =0+(k,)> (7.7)
In general, at ith load increment, displacement is obtained using the following equation:

u; = Uj;_} ata (k, ae AP, (7.8)

Note that the tangent stiffness value is updated after each load increment. However, using
this method, the approximated solution progressively departs from the exact solution as shown
in Figure 7.4. To avoid this progressive drift, load imbalance after each increment is adjusted
or corrected as follows:

u; = uj + (k, i [AP; + (Pa — Ri-1)] (7.9)

i-l
where P._, Ar , the total external load up to the load step i-1. The quantity R;_; is the
k=l
reaction force of the spring and can be obtained as R,_; = f(uj_,)- The drawback of
incremental method lies in expensive update of stiffness value after each load increment. The
accuracy of solution can only be increased by allowing small increment of load steps.

Figure 7.4 Incremental solution method.

 Chapter 7. Elastic-Plastic Finite Element Analysis
Example 7.3 Solve the problem given in Example 7.1 by incremental method.

Solution: In this case, load increment is assumed to be 100 units for 5 increments.
Considering initial condition as ug = 0.01 and AP, = 100 units, we can find u, as

=
2500
uy =o.o1+[ 100
= 0.014
v0.01

Hence R, = 5000V0.014 = 591.6079 and thus wu, can be obtained as

-1
2500
Uy =o014+( [100 + (600 — 591.6079)] = 0.01913
V0.014

Similarly, other three load increments results in u3 = 0.025129, ug = 0.03194 and

us = 0.03954. If the load increment would have been smaller, more accurate results was
obtained.

7.2.4 N—-R Scheme for Multiple Degrees of Freedom

For multiple degrees of freedom, N-R or modified N-R methods of one-dimensional cases are
expanded using vector and matrix notations. Let us assume that the displacement vector q
contains 2n components, where n is the total number of nodes. The force vector f is dependent
on q. Again Taylor series expansion of the force vector about displacement vector q’_; can be
expressed as

f(a‘) = f (ai atEA)t (q-4j-1) = f(a‘ x EA Aq; (7.10)

i=l el

where subscript i denotes the iteration number within the load step 7. The left hand side quantity
represents the externally applied load at load step 1, F’. The first quantity of the right hand side
us
equation is the reaction force developed at nodes after iteration i—1 for load step ¢. Let
designate it by Rj_;. The derivatives of force vector with respect to displacement vector
produce the stiffness matrix, K‘,_, at iteration i—1 for load step ¢ as

Ue igen <15
0q, 0q2 092n

Cia alos Wee

Ki, =| 9q, 992 042n (ell)

fn frm Fan
Oq 0q2 092n
 164 Finite Element Method: Concepts and Applications in Geomechanics

Assuming that externally applied load vector is independent of displacement vector, we

can rewrite this equation as

F’ —Rj_, =Ki_,Aq; (7212)

The displacement vector for ith iteration is then updated as

qi =i + Aq; a
Using updated displacements, new reaction force vector R‘, is calculated based on
Eq. (3.89) and iterative process continues until displacement or force convergence criteria is
satisfied. The initial conditions for force vector, displacement vector and stiffness matrix at
time (load) step t are assumed from the last iteration of previous load step as
qi =q'!
(7.14a)
Kj = FU!
(7.14b)

Ko = Kr!
(7.14c)

Example 7.4 Let us assume that two springs are connected in-series and are loaded as shown
in the figure below. The spring constants, k; and ky are non-linear and dependent on the
displacements q, and qp. If the initial displacements g; = 0.01 and q> = 0.02 units, using the
force relations as specified in Example 7.1, determine the displacements if the load increments
AF, = 250 and AF, = 500 units.

nN q2
1

ANNA ;
Solution: Considering linear spring constants, the total potential energy can be written
as
| eal 2
po teats AEGON: —H) —Fiq — Fog

Using stationary principle of minimum potential energy, we find that 5II, =0. Hence,
the equilibrium equations are qj
Fy = kiqr — (qa — qi)
Fy = kx(qQx - 4)
Using the non-linear relationship as expressed in Example 7.1, the above
force equations
are rewritten as °

F, = 5000/4, — 5000.9 — q;
F, = 5000./q, - qq
| Chapter 7 Elastic-Plastic Finite Element Analysis ae | | 165 |

The initial conditions are g; = 0.01 and q, = 0.02 units having F; = 0 and F, = 500.
The stiffness matrix for 2 degrees of freedom is estimated using Eq. (7.11).

' eats iM
Vs
~| 25000 25000
Iteration I:
The force vector AF!’ = {250 500}, solution of displacement increments is obtained as

Au,| [ 50000 -25000]' (250) _ {0.03

Au,| |-25000 25000 | |500J {0.05
Hence after iteration 1, new displacements are u, = 0.04 and u, = 0.07. and F, = 133.975,
F, = 866.025. The new stiffness matrix will be

: aes Ran
|
© hase 75714433 757
Iteration 2:
With the similar approach as shown in iteration 1, we can find u; = 0.06, u, = 0.0993
and. Fy, =233./6,. F2 = 990.98.
Similarly, iterations 3 and 4 can be conducted. The solution converges at iteration 4 with
u, = 0.0625, u. = 0.1025 and F, = 250 and F, = 1000.

7.3 Convergence Criteria for Non-Linear Solutions

Realistic terminations criteria of an iterative solution procedure are vital for achieving desired
accuracy and expense of the solution. After each iteration, solution is checked for preset
For
tolerances and decides on the termination of the solution or declares as divergent solution.
structural problems, mainly two types of convergence criteria are used:
iteration
1. Displacement convergence criteria: n this scheme, displacement increment at ith
is compared with the total displacement after ith displaceme nt as

(7.15)

This criterion is especially

where n is total number of equations and ¢, is the tolerance.
increment is very small as
applicable for small strain-hardening modulus in which force
compared to displacement increment.
 Finite Element Method: Concepts and Applications in Geomechanics

2. Force convergence criteria: It’s a similar criteria as above with the exception that out-of-
balance force is compared with a preset tolerance and the original load increment given below:

(7.16)

where F; and R; denote the applied force and reaction force at jth degree of freedom at load
step t and t— | respectively. Rj is the reaction force at jth degree of freedom after ith iteration
for load step t. This criterion is useful when displacement increment is rather small but
continues to change for many iterations.
In many problems, both of these two criteria are simultaneously checked for solution
convergence.

7.4 Elastic-Plastic Analysis Method

Apart from elastic properties such as elastic modulus and Poisson’s ratio, some additional rock
properties related to yielding criterion are required for elastic-plastic finite element analysis.
These properties, such as cohesion, angle of internal friction and others, are used to define the
yield criterion. In general, plasticity theory comprises of
(1) Yield criterion: Stress-based criterion to determine the onset of yielding
(ii) Hardening/softening law: A law which states that how the yield surface changes
with accumulated plastic strain
(iii) Flow rule: A tule which states the direction and amount of plastic
strain increment
Each of these items are discussed in the following sub-sections.

7.4.1. Yield Criteria

Detailed description of yield criteria for isotropic rock and rock mass is given in Chapter 6.
In this text, classical elastic-plastic finite element formulation will be described using
Mohr-
Coulomb, Drucker-Prager, Hoek-Brown and Alternative criteria. These criteria are dependent
on mean hydrostatic stress and invariants of deviatoric stresses. In addition, yield loci
of Mohr-
Coulomb and Hoek-Brown hexagon are also dependent on direction of principal
stresses and
thus we cannot freely interchange oj, o>, and 03 axes in 7-plane since this will
affect the yield
surface. Hence Mohir-Coulomb and Hock-Brown criteria have three axes of symmetry
as
opposed to six axes of symmetry for Tresca yield criterion in-metal plasticity. In general,
yield
criteria predict the onset of inelastic deformation for a given state of stress.
For each of these
criteria, an elastic response is declared if F < 0 and elastic-plastic response
is predicted for
F 2 0. However, in reality, F > 0 cannot exist and thus proper plastic
correction is applied
to bring back the stress point on the yield surface.
 Chapter 7 Elastic-Plastic Finite Element Analysis 167

7.4.2 Hardening/Softening Laws

The hardening or softening rule specifies that how the yield surface changes with accumulated
plastic strain. If the yield surface does not change with plastic strain, then we have ideal or
perfect plasticity problem. For metals, experiments show that if a solid is deformed plastically,
then unload it, and then reload it so as to induce further plastic flow, its resistance to plastic
flow increases. This is termed as strain hardening. Similar behaviour is also observed for
triaxial compression of rocks. Under high confining stress, the rock sample continues to yield
until failure occurs. In brittle rock, however, strain hardening is a rare event since it fails
violently once yielding/failure load is attained.
In general, for rocks and soils, yield surface does not strain harden but strain soften
instead so that the yield stress level at a point decreases with increasing plastic deformation
(Owen et al., 1980). As a result, the original yield surface contracts progressively. In other
words, yielding implies local failure in terms of local fracture development in weak rock or
shear band formation in compacted soil or sand and the yield surface becomes failure criterion.
In rock mechanics, the adjectives ‘ductile’ and ‘brittle’ are used for hardening and
perfectly plastic behaviour, and for softening behaviour respectively, though they usually refer
to the ability to sustain significant plastic deformations (Maier et al., 1979). In that paper, the
authors mathematically proved that based on continuum plasticity theory, the unique solution
does not exist for material with negative hardening modulus or for strain soften rock-like
material. The modeling of strain-softening behaviour using plasticity theory presents numerical
difficulties and instabilities (Brown, 1987). Quite often, strain-softening behaviour is associated
to the formation of zones of limited thickness within the sample under compressive loading
where strain localizes, eventually leading to failure through the formation of collapse
mechanism (Sterpi, 1999). In addition, plasticity is based on continuum theory but strain
softening can only exist in heterogeneous material. Thus analysis of strain-softening behaviour
using classical plasticity theory becomes ill-posed. Many authors have presented the concept and
analysis procedure of strain-softening plasticity using regularization of non-local averaging of
plastic strain considering an intrinsic parameter with the dimension of length (Rolshoven et al.,
2002). Nonetheless, strain softening analysis in plasticity is an active research area. Illustration
of srain localization and non-classical theory of plasticity is beyond the scope of this book.
Instead in this text, the concepts of strain softening analysis are presented by changing of
material state parameters (MSP) based on accumulated plastic strain. In other words, the basic
assumption for this analysis is that as the plastic strain accumulates, certain rock parameters
changes causing alteration of yield surface.
The above discussion suggests that in order to model either strain hardening or softening,
the yield function must also be a function of plastic strain tensor. The yield function is
thus a function of both stress tensor and plastic strain tensor and given by the following
equation:

F(o,e”) =0 C7217)

To get a suitable scalar measure of accumulated plastic strain, effective, generalized or

equivalent plastic strain magnitude can be defined as (Cook et al., 1989)
 Finite Element Method: Concepts and Applications in Geomechanics

3 3 3
1/2

eof-F|(e —eP)? + (eh, -e2)” + (eR - eR +5 (1) +5 (1h) 3a80')| (7.18)

The positive root of the right-hand side is intended. For uniaxial condition Et =|ex.|
can be obtained by considering ee =e? =-0.5e?, for Poisson’s ratio 0.5 in plastic regime and
all plastic shear strains are taken to be zero.
For isotropic hardening or softening, each of the material state parameter like compressive
strength, cohesion (c), angle of internal friction (@), Hoek and Brown rock mass parameters,
m,, S and a characterize a material state and govern the size and shape of the yield surface.
In practice, piecewise linear functions are used to relate these material state parameters with
accumulated effective plastic strain, €%. For example, for simulating strain softening
phenomenon, cohesion, angle of internal friction, dilation angle and rock mass parameters can
be plotted as monotonically decreasing function with accumulated plastic strain as shown in
Figure 7.5. In the figure, the subscript o signifies the original value of the parameter and the
subscript r refers to the residual value of the same.

MSP

MSP. oO

MSP,

ee

Figure 7.5 Schematic diagram showing relationship between MSP and accumulated plastic strain.

Now by taking differential of Eq. (7.17) and equating it to zero, we have the consistency
equation which states that stress point always lies on the yield surface:

ar
oF |"
=} do
OF |08%,
+——,——} de? =0 (7.19)
06 det, |de?
Note that the above equation is a scalar function with one equality condition. The
hardening or softening law must specify a relationship between material state parameters and
cumulative effective plastic strain €/ which must be established by experiments or defined by
the user.
| Chapter 7 Elastic-Plastic Finite Element Analysis |

7.4.3 Flow Rule

In order to derive the relationship between plastic strain components and the stress increments,
an assumption must be made on material behaviour. The most general assumption is that the
plastic strain increments are proportional to the stress gradient of a quantity termed as plastic
potential function, Q, and represented as

eer p 99 acl (7.20)

where d7) is a non-negative proportionality constant and is called the plastic multiplier. This
quantity must be estimated from the consistency condition such that stress point must lie on
the yield surface at all time.
Equation (7.20) is termed as flow rule and it governs the plastic flow after yielding. Note
that although magnitude of the plastic strain depends on the stress increment, do, the components
of plastic strain increment depend on the current stress and not on the stress increment.
The structure of potential function Q is evidently very similar to the yield function and
must be a function of 6,,, Jz and J; and is represented as

O =O(0,,,J2,0) = const. (7.21)

The plastic potential functions used with different yield criteria are illustrated in
sub-section 7.5.1. The relation Q = F has a special significance in the theory of plasticity
especially for metal plasticity. A flow rule derived from the yield surface is known as
associated plastic flow rule. In this case, Eq. (7.20) becomes
F
de? CLE or de? =——dn (7.22)
00
and is termed as normality condition since OF/do; is a vector directed normal to the yield
surface. Figure 7.6(a) shows that for any given state of stress point in the deviatoric plane, the
direction of plastic strain vector is normal to the yield surface. Theory of plasticity that use

Initial yield ~% ie?

surface I

@
dsj;

we
Final yield surface

(a) Associated plastic flow plastic flow (normality condition)

 Finite Element Method: Concepts and Applications in Geomechanics

a separate plastic flow potential function is termed as non-associative plasticity model and is
more appropriate for materials like soils and rocks [Figure 7.6(b)]. In this text, non-associative
plasticity models are described in sub-section 7.5.1 with Mohr-Coulomb, Drucker-Pragger and
Hoek-Brown yield surfaces.

Initial yield
surface

Final yield surface

(b) Non-associated plastic flow

Figure 7.6

7.5 Elastic-Plastic Stress-Strain Relationship

The elastic-plastic constitutive matrix, which relates increment of stress and strain vectors,
depends on the current stress level elastic constitutive relations and plastic properties of rock
material. In this section, formulation of elastic-plastic constitutive matrix is discussed based on
the incremental strain theory. In plastic regime, total strain increment dé is assumed to be
divisible into elastic and plastic strain increment components as given below:
de =de° +de? €LZ3)
The elastic component of strain increment is responsible for stress increment and is related
with elastic constitutive matrix as given in Eq. (7.24).

do = Dde* = D(de - de”) (7.24)

Equation (7.24) can also be expressed in terms of total strain increment and elastic-plastic
constitutive matrix in the form of
do = D,dé (i255
where D,, represent elastic-plastic constitutive matrix. In order to derive an expression for
elastic-plastic constitutive matrix, first we have to estimate the plastic multiplier dr given in
Eq. (7.20). From the flow rule and Eq. (7.24), we have

id dQ
 Chapter 7 Elastic-Plastic Finite Element Analysis 171

Replacing Eqs. (7.20) and (7.26) into the consistency equation, Eq. (7.19), we find that

dn = Ajde (7.27)

mere ian eC (e T
ee (7.28)
{ae p {22} QF |e, {22
00 00 del, de? 00

Note that the denominator of Eq. (7.28) is a scalar quantity. The dimension of the vector
‘aeis the same as
A;, ,
the stress vector 6. The matrix D,, can then be evaluated from stress
increment do as

do =“ D,,d€ = p{ae- dQ
2 an) (7.29)

Replacing Eq. (7.27) into Eq. (7.29), we find

D., = p-p| 22}as

00] +
(7.30)

The elastic-plastic constitutive matrix is non-symmetrical for plasticity model derived

with non-associative flow rule. Element stiffness matrix formed using this constitutive matrix
will also be non-symmetrical and thus computational efficiency reduces drastically. However,
for associative flow rule when Q = F, elastic-plastic constitutive matrix is symmetrical.
For elastic-perfectly plastic material behaviour, yield function is independent of
cumulative plastic strain and thus dF /de?, is zero. Then Ai, can be represented by

fee lis}° 00

toe} 20
Leys 122 (7.31)

7.5.1 Implementation of Shear Yield Criteria for Rock and Rock Mass
In elastic-plastic analysis, behaviour of an element is dependent on the number of sampling
points assigned as quadradture points of a numerical integration rule (Cook et al., 1989).
During iterative solution process some sampling points of an element may yield while others
may not. Thus for each sampling point, we must keep record of stresses, strains and update
them after each iteration. Before executing elastic-plastic analysis, one must select a yield
criterion, flow rule and hardening/softening law. For example, if Mohr-Coulomb criterion is
to be chosen with non-associative flow rule, then data regarding elastic modulus, Poisson’s ratio,
 172 [_ Finite Element Method: Concepts and Applications in Geomechanics

these
original material state parameters such as c, 9, YW, 9, and the relationship between
the
parameters with 2, are required. Then during computation, we must record and update
value of €%, at each sampling point to obtain the current value of these parameters .
As described in the previous section, we need to estimate the plastic multiplier, dn and
components of elastic-plastic constitutive matrix based on selected yield criterion and plastic
potential function. First, we evaluate the derivatives of yield function with respect to stress as
given below:

‘=
—?p
r=Cyay T +Cyazi +C3a3T (752)
00

Parameters C; are the derivatives of yield function with respect to O,,, a and J; and
have to be obtained from the assumed yield function. Expressions of C; are given in Table 7.1
for different yield functions. For three-dimensional stress field the components of a;
(i = 1,...,3) vectors are expressed as

eST ah ois 07 (7.33a)

Hi at
dC eae

cMie2 0 Sl (7.33b)
aya ta)
T
2G 2, oe whee

J J
oJ: SyySzz —Tyz + RAE Oe SE On nee
ae a [ov fi A [shu i m [say "y 2 (7.33¢)
Q(T yxy —SxxTyz) WyTy, —SyT,) AT yt, —Sz,T
yy)

For plane strain condition, a; can be obtained in Cartesian coordinate system assuming
stress vector as of = {o fey v O,,} as
xX yy xy

T OO m peed,
Dae ae (7.34a)

T ts dJ3 ry J>5 Jy 2 J
a3 at ={(sp% eo S577 Ley = 2827 F xy Sy Syy — Txy +22) (7.34c)

eeee See eee

where J = 5 ee Ye ee Pra
 Chapter 7 Elastic-Plastic Finite Element Analysis 173

Table 7.1 Estimation of derivatives C;

Yield Criterion Gy Cy C3

7 37;
Mohr-Coulomb Neo rey meal comalita
Vay 2cos30(/Jo ir
1
Drucker-Prager (outer) Ng - 1 —=(2+N 0

: 1
Drucker-Prager (inner) Ng - 1 —(1+2N, 0

Hoek-Brown m, Rp - He Ry a
V2 2cos30 (Jo )

Alternative | m, R> 0
Alternative 2 Mp R 0

where
OF
it ob [+ n9)sino +N, = |

Ty = (1+ Ng) cos@ +(Ng —1) sin9

3
oF (2sin@ : é
R Sayre (722) 2cos0y" aye Cas Dime Qe —m, [-sine 25?) 73

Ry = 2h" +p
a =

For non-associative flow rule, a suitable plastic potential function for the specified yield
surface need to be defined. In general, for Mohr-Coulomb yield surface, a linear function is
defined in principal stress plane based on dilation angle, y, as
Q=-0, + 03N,, = const. 735)

1+siny
where Ny, = and y = dilation angle
l-siny
Figure 7.7 shows linear potential lines in 0; — 03 plane. Equation (7.35) is modified to
obtain the expression for plastic potential function using invariants of deviatoric stress as given
in Eq. (7.36). In deviatoric plane, this plastic potential surface is convex and resembles an
irregular hexagon. However, if no dilation is assumed, the shape of hexagon is regular as shown
in Figure 7.8. For any case, six points of singularity exists when 6 approaches to +30 degrees.

i= Jila+ Ny)cos6 + (Ny — pee + (Ny —1) 6, = const. (7.36)

 174 | Finite Element Method: Concepts and Applications in Geomechanics

Q' = const.

Os
Figure 7.7 Linear plastic potential function with Hoek-Brown yield criterion.

Potential function associated with Drucker-Prager yield surface is obtained by replacing

lode angle by +7/6 in Eq. (7.36) depending on the outer or inner circle and the equations are:

O= 20 PNA )On, (to be used with outer yield surface) (7.37a)

Q= 2 2Ny +1)-(-N,)o,, (to be used with inner yield surface) (7.37b)

The linear plastic potential function in 0; — 03 plane as given in Eq. (7.35) has been
considered in conjunction with Hoek-Brown yield criterion for the analysis of ground reaction
curve and elastic-plastic analysis of circular tunnel (Alonoso et al., 2003, Carranza et al., 1999,
Carranza et al., 2004, Deb et al., 2005). Alternatively, plastic potential function based on a
dilation parameter mg; ranging from 0 to m,/4 is also applied with Hoek-Brown media
(Rockscience Inc., 1999). In this case, plastic potential function is formulated as given in
Eq. (7.37) and then m, is replaced with mj. These potential functions are plotted with
GSI = 50, m; = 10, o,; = -30 MPa, o,, = —-30 MPa and a constant value of 50 as shown in
Figure 7.8. It can be seen that as the dilation angle increases, the shape and size of the plastic
potential surface changes. On the contrary, plastic potential surface expressed in Eq. (7.38) is
circular in deviatoric plane without any singular points.

_ [F@=1/6)+ F(@=-2/6)| ie
Q a Ths aeaiie Sareea ae = const. (7.38)

DULG Hl

The plastic potential function for alternative yield surfaces are expressed by
Eqs. (7.37a) and (7.37b) for outer and inner yield surfaces respectively.
| Chapter 7 Elastic-Plastic Finite Element Analysis
175

03

Figure 7.8 Plastic potential functions in deviatoric plane

The derivatives of potential function with respect to stress are obtained as

dQ

where a; are defined in Cartesian coordinate system and D, are the derivatives of plastic
potential function with respect to o,,,./J, and J; and given in Table 7.2. In this table, linear
potential function is as expressed in Eq. (7.36).

Table 7.2 Estimation of derivatives D;

Yield criterion D, D, Ds;

tan 30 r 37;
Mohr-Coulomb or Hoek-Brown Ny ~1 Ty a5 1 2c0836( {hay

2+ Ny
DP (outer) or Alternative 1 Ny 1 aan 0

1+ 2Ny
DP (inner) or Alternative 2 Ny ~1 RHR 0

where
5
cos@

f _ sin®@
Tr =| (1+ Ny)oos@
+(Ny 1)
 Geomechanics
176 | Finite Element Method: Concepts and Applications in

ar element is
Example 7.5 The stress vector in MPa at tl load step of a 3-noded triangul
found to be of = {-1.2 -5.5 0.23. -1.675}. Let us assume that the increment of strain
vector at load step t is Ae™ = {3.34 -3.63 0.328 O} x 10+. Assuming the plane strain
criterion with E = 5 GPa, n = 0.25, c = 2.MPa, and @= 257,
condition and Mohr-Coulomb
strain vector assuming
determine whether the element has yielded. If so, estimate the plastic
the associative flow and elastic-perfectly plastic conditio n.

Solution: The stress vector at load step t will be o’ = o’! + DAe’ as given below:

1.2 6000 2000 0 2000) 3.34 0.078

—5.) fs 2000 6000 0 2000 ||-3.63 10-4 = -7.010
0.23 0 OH PALIN 0.328 0.296
—1.675 2000 2000 0 6000 0 —1.733

The following table shows the various parameters based on this stress vector.

o,,, MPa Jy, MPa o,, MPa 03, MPa 6, deg

—2.888 3.694 —7.022 0.090 -15.71

Using Mohr-Coulomb criterion, F = 0.64 [Eq. (6.22)]. Since the value of F is positive,
the element has yielded at load step t.
In order to find the increment of plastic strains, vectors a; and constants C; are estimated
to obtain derivative vector with respect to stress as given in the following table:

|_1.7475

Since Q = F and elastic-perfectly plastic condition, using Eqs. (7.26) and (7.30) and
applying matrix multiplications we can find plastic multiplier, dn = 1.248 x 10+. Thus, plastic
strain increment vector can be obtained from Eq. (7.24) as given below:

Ae? ={2.773 -2.041 0.402 1.450$x10~

7.5.2 Implementation of Tensile Yield Criterion

Tensile yielding occurs when 63 exceeds the tensile strength o, (taken as positive) as given in
| Chapter 7 Elastic-Plastic Finite Element Analysis aa

Eq. (6.35). For tensile yield surface, the associative flow rule is assumed. Using the similar
concept as given in Eq. (7.32) for shear yielding, we can obtain the derivatives as

OF’ ‘ T T
| = Cay +Cyaz +C3a3 (7.40)

where C; are estimated from the following equations:

C, =1

tan 30 1
C, =cos8| 1— +sin@| —=+tan30@ 7.41
3 [ V3 ee ee
C, =,
es ey [Sse
0
24, cos30\ V3

The a; vectors are same as given before. Assuming the associative flow rule, O = F’,
plastic multiplier and elastic-plastic constitutive matrix can be obtained using Eqs. (7.27) and
(7.30) respectively.
In general, for elastic-plastic analysis, the tensile yield criterion is checked first at each
sampling point of an element. If this criterion is not satisfied for the current stress condition,
the shear yield criterion is checked. If a sampling point yields due to tension the tensile stress
is set to be zero for that point and the load is redistributed to nodes. A “‘no-tension” material is such
that the tensile strength, 0, is assumed to be zero. Thus the yield criterion becomes F, = 03.

Example 7.6 Using the data given in Example 7.5, estimate the updated stress vector at load
step f.

Solution: The elastic-plastic constitutive matrix is estimated based on Eq. (7.29) as

230
le 1596 2893-91 Os 189,012 pe=392,112
289 3.9 10'557/80.370 46.439 =2587.732
Pe —189.012 46.439 1990.181 -—124.272
—392.112 2587.732 —-124.272 4427.225

Hence the increment in stress vector can be calculated using Eq. (7.29) as

~0.2679
_|-1.1302
Ao’ Mpa
~ 1-0.0147
-1.0744
 Geomechanics |
178 Finite Element Method: Concepts and Applications in

The stress at time step ¢ is 6! = 0”! + Ao’ and given as

—1.4679
t
_ |-6.6302
pa
0.2153
~2.7494
in Example 7.5.
Reader should compare the stress vectors at load step ¢ given above and

7.5.3 Improvement of Solution Accuracy: Numerical Procedure

by
Numerical procedure involving iterative solution for elastic plastic analysis is performed
updating tangent stiffness matrix after each load step. This method is termed as tangent stiffness
method. If the initial stiffness matrix does not update in the entire solution process then this
method is called initial stiffness method. In this section, we will discuss the numerical procedure
using the tangent stiffness method.

Tangent Stiffness Method

In this method, global stiffness matrix is updated frequently (at least after each load step) with
the new elastic-plastic constitutive matrix. The total loads F on the structure are applied in
k
increments as AF,, AF, ..., AF,, where F = >)AF, . Load increments can be taken as equal

steps meaning AF, = AF, = --- = AF,. ae

Step 1. At the outset, € = o = 0 and D,, = D for all sampling points. If hardening and
softening rule is applied, then €%, is also set to 0 for all sampling points. Apply
the first load increment AF, at load step i = 1 and also set Aqo = 0.
Step 2. Compute structure displacement increment Aq, and strain increments A€, at
element sampling points from the following equations:

AF, = K,yAq, and Ae} = BAq; (7.42)

where Ko is the global initial stiffness matrix formed by usual assembly of n

elements as shown below:

Ky = ki,kf= [B/D,BjdV where i = 1 ton. (7.43)

V:
I

Thus, in general, for jth load step for the sampling points that are yet to be
yielded, the above equations can be rewritten as

Step 3. For each sampling point, update trial stress as

Oj = O:1 + D.pAg;.
ae Chapter 7 Elastic-Plastic Finite Element Analysis 179

Step 4. Determine the value of yield function using updated stress. If F(6;) < 0, elastic
material behaviour is still valid and hence the trial stress is the current stress. If
F(;) > 0, plastic correction is required and so go to the next step.
Step 5. If the previous state of stress was in plastic domain, then set ratio of elastic
portion, m = 0 and so go to the step 6. Otherwise, there is a transition from
elastic to plastic regime and the value of m has to be estimated as follows:

m= eo jae
F(6;)- F(G;-1) they
Note that the yield function has to be evaluated as F(o) = 0. The meaning of
parameter m is illustrated using Figure 7.9 for one dimensional case. Inclusion
of m in the numerical procedure is optional but it improves the accuracy of the
solution.

Og
Oy Point A: Before yielding
OA Point B: After yielding
Stress

Strain
Figure 7.9 Evaluation of parameter m.

Step 6. Redefine trial stress as 6; = 0j_; + mDAg; and the elastic-plastic strain increment
will be Ag; = (1 7, m)Ag;.

Step 7. Now plastic corrections have to be applied to find the final stress based on the
updated strain increment. Stress increment is evaluated using Eq. (7.29). Then
strain, stress and displacement vectors after any iteration are updated as follows:

qj =9j-1
+ Aq;
0,=0.,+Ao,
J ip J

Cer
Jag
oe
apedize
Ac.J dats

gjPairap
=e% , + AesPp

Reaction force vector after kth iteration in jth load step for any sampling point is then
obtained from the updated stress vector as given below:
 hanics i
| 180 | | Finite Element Method: Concepts and Applications in Geomec

ka ek
Rj =a: ojdVv (7.47)
V

obtain global
For every element (yielded or not), reaction force vectors are assembled to
reaction force vector
reaction force vector based on nodal degree of freedom. The cumulative
‘:
from the cumulative load vector F; =>) AF; as shown in
thus formed is then subtracted
i=l
Eq. (7.12) and the difference in loads is applied for k + 1 iteration.

Optional: Since the material behaviour is dependent on the current stresses, better
accuracy can be obtained if the strain increment is divided into number of subincrements and
then stress is updated for each subinterval. Hence in non-linear analysis for each load step, there
are number of iterations and within each iteration, there may be several subincremental steps.
If the load increment is low, subincremental step can be omitted. Within each subincrement,
the plastic multiplier d7 and elastic-plastic matrix are estimated and then stress is updated as
follows:

o7-o, + Di, Ae} (7.48)

The superscript s signifies the subincrement number. This process continues for all elastic-
plastic strain subincrements.
Step 8. The iterative solution continues until the convergence criterion is satisfied.

7.6 Tunnel Example—Elastic-Plastic Analysis

7.6.1 Analytical Solution—Mohr-Coulomb Media

As shown in Chapter 3, high stress concentration occurs at the boundary of the circular tunnel,
especially on the side walls. Rocks surrounding the boundary of the tunnel may yield/fail due
to high tangential (failure) stress concentration and negligible radial (confining) stress. As a
result, a plastic zone may develop around excavation boundary as shown in Figure 7.10. In the
plastic zone, stress concentration is relaxed. For a geotechnical and mining engineer, it is
imperative to estimate the extent of the plastic zone around an excavation boundary so that
judicious decision can be made for designing the support requirement. In this section, the extent
of plastic zone (radius of elastic-plastic boundary r,) and stress conditions are evaluated
considering Mohr-Coulomb yield criterion. For this purpose, uniform hydrostatic far field
stress (p,) condition is assumed. A uniform internal support pressure of p; is also applied at
the boundary of the tunnel.
| Chapter 7 Elastic-Plastic Finite Element Analysis
181
Po

ste
Elastic region Orr

—> a

—> <a

————_> ——
Po

—— <——_

Plastic region

Figure 7.10
feel»
[ef
Elastic-plastic analysis of a circular tunnel.

Stresses in the Elastic and Plastic Regions

For hydrostatic condition, Tg = 0. Hence the stress equilibrium condition in polar coordinates
ignoring body force component in radial direction becomes

do O69 — Orr
>i ai ad (7.49)
In the absence of shear stress, normal stresses acting in an element in the plastic zone are
principal stresses. Assuming Mohr-Coulomb yield criterion, the following relationship can be
obtained:

O99 = Or +N 26,, (7.50)

where
1+sing, ik evinc Sil 2cpCOSo¢
=——__——,, @ = angle of internal friction of fractured rock, 0,, =-—~——_—
f 1—sin@¢ 1—sin 9,
and cy = cohesion of fractured rock
Replacing Eq. (7.50) in Eq. (7.49) and then rearranging, we find
Opp tO, Nel) a
es i Kaa \. "tar
NET i} Tete eo ie Cheat)
f Oop tpi (N ¢-1) x a

where 7 = O,¢ + 0, (Ny - 1)

O,. = radial stres at the elastic-plastic boundary
The solution of the above integral equation finds the radial stress at elastic-plastic
boundary (0,.) as
Ne-l
Ore =(—Cp Cot he + P;) “a +c, cot (i252)
/
Ne
 Concepts and Applications in Geomechanics
Finite Element Method:

in the elastic zone can be

Considering superposition concept, tangential and radial stresses
estimated as given below:

2 D 2
pa 0. i = y Te
cic Caen [2{1-5 Jeon=r 14%)-en[ 5
fOtsTa ey. (7.53)
Po —O ip2 ir ie2 I, 2
oO; = 5 ve 21-8) ro.=n) +00[

r = 7,
From Eq. (7.53), we find that tangential stress at the elastic-plastic boundary,

Soe =2Do — Gre (7.54)

At the elastic-plastic boundary, stress condition is such that a rock element is in
equilibrium and no more yielding is imminent. Hence we can write that Mohr-Coulomb
criterion is just satisfied as
Oge =Ocr + N-Gre C755)

1+sin@, Louk
where N, a aia G,, = uniaxial compressive strength of rock and @, = angle of
— sing,
internal friction of rock. Substituting Eq. (7.54) into Eq. (7.55), we find

2Po is Te 7.56
0,. =———*
to Dieting joni
Replacing Eq. (7.56) into Eq. (7.52), elastic-plastic radius is obtained as given in the
following equation:

_ |2P0 = Ser = ey Cot)N,) Sle

1/(N -¢-1
(7.57)
Te
(pj — cp cot)
+N,)
Radial and tangential stresses in the plastic zone are estimated using the equations given
below:

Ps Nes 2
Ce =(—c¢ cot dr +a(=) + Cf cot fore Sy. (7.58)

Ogg = Of a Noi

Strains and Displacements in the Elastic and Plastic Regions

In the elastic region, the radial displacement is obtained as given in Eq. (3.96a). In this case,
k = 1| and hence the radial displacement in the elastic region becomes
2»
SO
u, =P Gece for i> 1 (7.59)
 Chapter 7 Elastic-Plastic Finite Element Analysis

In the plastic region, a < r < r,, the radial (€,,) and circumferential (€
99) strains are
decomposed into elastic and plastic parts as given below:

e
Ep, aad Err teey (7 60)

3 P ;
E99 = Egg + Egg
Considering radially upward displacements as positive, radial and circumferential strains
at any point in the plastic zone can be expressed based on the radial displacement u as

“ _ du
icpoate Pee

Se u
(7.61)
&99 = T7
Assuming non-associative flow rule given in Eq. (7.35), and replacing 6; with Ogg and
03 with o,,, we can find

dQ
ep. = Shee Hl = Nyan

0) (7.62)
ef, = ——dn=-d
te 00¢¢9 : i

From Egs. (7.60)-(7.62), we can write the following differential equation:

u
Sa eayah eer + Ny €99) =0 (7.63)

The third term of Eq. (7.63) is a function of r only and is given in the following equation
(Sharan, 2003):

i! D
(ee + Myth) =| CO-2000 + Ny) 2 My -»| (7.64)
where

2 2
= (Oe =e — (Pp; 7 Po)a
© %
202
—a

71)
D (D; Or )a Te
ie)
i424

Equation (7.63) is solved after applying the boundary condition which states that at the
elastic-plastic boundary, the radial displacement is equal to the elastic displacement obtained
from Eg. (7.59) for r = r,. Hence the solution of the differential equation given in Eq. (7.63)
is obtained as
 |
184 | Finite Element Method: Concepts and Applications in Geomechanics

u=-r ae sie -20+ My) 3 My -p}jare®

Net ‘ [“)Ms
0 — Ore Me e

R of r
2G
(7.65)
(a S 1 Ny - ‘I+(DO
Do te ce
Ore [4
ee “Ny a Ny +1 _ wat)
r Dir, r Smee
ac [ca 2v)(r

7.6.2 Analytical Solution—Hoek-Brown Media

on made in rock mass
The procedure for obtaining elastic-plastic analytical solutions of excavati
that of Mohr-Co ulomb media as mentioned
based on Hoek-Brown yield criterion is similar to
is applied to calculate stresses and
above. In this case, the yield criterion given in Eq. (6.27)
of r, and o,, becomes
displacements in the plastic region. However, if a # 0.5, exact solution
Fairhurst, 1999). In
tedius. In such cases, iterative process has to be adopted (Correnza and
and the radius
Eqs. (7.66) and (7.67), the expressions for the radial stress in the plastic region
stress,
of elastic-plastic boundary are given considering compressive or negative far field
of rock mass for a = 0.5 based on the
internal stress and uniaxial compressive strength
reference (Sharan, 2003).
2
Ce = "Sill(£) -|n() \Yovou +5,0%; +p; forr<r, (7.66)

(-4 Oc; (mM, Pj + SO ¢j)

Tr, = aexp I(—2m,6;;)
+J2,J0,;(85,0,; + m,O,jm, + 8m,p, +m, O,; (MEO; + 16m,py +16s0;;))

(7.67)
where m, and s, are the residual values (for fractured rock mass) of Hoek-Brown rock mass
parameters. The tangential or circumferential stress in the plastic region is calculated from
Eq. (6.27) considering a = 0.5 and for residual rock mass parameters m, and s,. The radial
displacement in the plastic region is estimated using Eq. (7.65). In the elastic region, radial
displacements and stresses are calculated based on Eqs. (7.59) and (7.53) respectively.

7.6.3 Non-Linear Finite Element Solutions and Comparison with

Analytical Results
The tunnel problem mentioned in Section 5.3 is again revisited. In this case, we intend to apply
analytical as well as finite element elastic-plastic analysis procedures as outlined in this chapter.
The finite element mesh of the problem is shown in Figure 7.11 and it is exactly the same as
Figure 5.4. The y axis is assumed to be the axis of symmetry. The tunnel boundary is 2 m
away from the axis of symmetry. From the boundary of the tunnel, the radial extension (x or
r direction) of the model is 38 m. The height of the model is 0.5 m (in the figure it is
exaggerated by 4 times). Recall that the model is developed 118 nodes with quadratic
quarilateral elements. The model consists of 23 elements. The model is constrained vertically
 Chapter 7 Elastic-Plastic Finite Element Analysis |} [185 |

'

eee <<

a eee
'
3

38m
2m
Figure 7.11 Axisymmetric tunnel model for elastic-plastic analysis (support condition), for no support condition,
Pp; = 1 MPa is removed.

as shown in the figure. This model is analyzed considering both Mohr-Coulomb and Hoek-
Brown rock media and results of finite element results are compared with that of analytical
solutions.

Tunnel in Mohr-Coulomb Rock Media

In this case, it is assumed that the rock mass surrounding the tunnel boundary can be analyzed
considering Mohr-Coulomb criterion as yield function. The compressive strength and angle of
internal friction of rock mass (0,,, @,) and that of yielded (fractured) rock mass (Og 9) are
considered to be 8 MPa and 25° respectively. The uniform external loading (p,) of 8 MPa is
applied in the model boundary. In this case, no internal pressure is applied on the tunnel
boundary. This loading condition is termed as “no-support condition”. The dilation angle is also
varied as 0° and 15° to compare the stresses and radial displacements for this condition. The
elastic-perfectly plastic rock mass is assumed for both analytical and finite element procedures.
The finite element program, FEMROCK developed by the author is used for the numerical
modeling and analysis of this problem.
The calculated results from analytical solutions given in sub-section 7.6.1 and finite
element results obtained from the axisymmetric model are compared and listed in Table 7.3.

Table 7.3 Comparison of results between analytical and finite element method

Parameter Analytical FEM Error (%)

y = 0° w= 15° y= -0° y= 15° Tae CRSes
rda 1272 1.272 1.270% 12 7ky —0.157 0.157
—100u/a (%) atr=a_ 0.223 0.243 0.228 0.247 2.242 1.646
Og/P, atr=a 1.000 1.000 1.015 1.027 1.499 2.699
Ogd/P, atr =r, gis 1.711 33" 174s 1.285 1 7/58

Error =100(FEM-Analytical)/Analytical (%)

*Estimated from the finite element results

Figures 7.12(a) and (b) show the distribution of radial displacement and stresses as a
function of r/a. The clear demarcation of elastic and plastic boundary is evident from the
tangential stress distribution [Figure 7.12(b)]. It can be easily verified from Table 7.3 and the
figures that results obtained from the finite element analysis are in close proximity with the
 186 Finite Element Method: Concepts and Applications in Geomechanics

the zee soe

0.006 ~
'

0.004

E
rf
0.002

0.000
1 ds: 22 POS) 8} SS 4

rla

Analytical dil = 0 deg Analytical dil = 15 deg

* FEMdil=0 deg o 5
FEMdil=1deg

Figure 7.12(a) Distributions of radial displacements (Mohr-Coulomb rock media).

2.00
4.75
4.50
4.25
1.00#
0.75
OY
OP.
Po
Or 0.50
0.254 .
0.00
1 1.5 2 2.5 3 3.5 4
r/a

\——Analyical O FEMDIi=15deg X FEMdil=0 deg |

Figure 7.12(b) Stress distributions in rock around the tunnel boundary (Mohr-Coulomb rock media).

analytical results. The errors of calculation using non-linear FEM procedures are within the
acceptable limit. As expected, for non-dilatant (yw = 0°) rock mass, the radial displacement in
the plastic zone is lower as compared to the dilatant (yw = 15°) rock mass.

Tunnel in Hoek-Brown Rock Media

In this section, non-linear solutions are obtained using analytical as well as finite element
procedures for Hoek-Brown rock media. The rock mass properties are assumed as m; = 10.0,
GSI = 40, and o,; = —30 MPa with dilation angle of 0° and 15°. The Hoek-Brown rock mass
parameter a is assumed to be 0.5 for finite element analysis for comparing results with the
analytical solutions. Apart from “no-support” condition, the study is also extended for ‘support’
condition which describes an internal pressure of 1 MPa applied to the inner boundary of the
tunnel. As before, the elastic-perfectly plastic rock mass is assumed.
| Chapter 7 Elastic-Plastic Finite Element Analysis ]

No-Support Condition
As mentioned earlier, for no-support model, traction loading is applied only on the boundary
of the model. Figure 7.13(a) plots the variation in radial displacements with respect to radial
distance from the boundary of the tunnel. Radial displacements obtained from the analytical
solution are also plotted for comparison. It can be seen that the plastic zone has developed
around the tunnel boundary and the radius of the plastic zone from the centre of tunnel is 3.35
m [Figure 7.13(b)]. Table 7.4 lists and compares the results obtained from both the techniques.
The percent error of estimation for tangential stress at the tunnel boundary is higher as
compared to other parameters. However, the amount of error reduces drastically as the r/a ratio
is greater than 1. As before, radial displacement in the plastic zone is higher for dilatant rock
mass as compared to non-dilatant rock mass.

0.013 a

ria

Analytical dil = 0 deg Analytical dil = 15 deg

* FEMdil=0deg oO FEMdil=15 deg

Figure 7.13(a) Distributions of radial displacements (Hoek-Brown rock media, no-support condition).

StS: — ee |

1200
1.25
1.00

1 1.9 2 2.5 3 3.5 4

ria

Analyical o FEMDiI=15deg xX FEM dil = 0 deg

Figure 7.13(b) Stress distributions in rock around the tunnel boundary (Hoek-Brown rock media, no-support
condition).
 Finite Element Method: Concepts and Applications in Geomechanics

Table 7.4 Comparison of results between analytical and finite element method (no-support condition).

Parameter Analytical FEM Error (%)

y= 0 y= 15° y = 0° y = 15° y= 0° y = 15°

rJa 1.677 1.677 1.675* Gian —0.119 —0.119

—100u/a (%) at r=a 0.397 0.542 0.405 0.527 D&A bil —2.767
Og/P, atr=a 0.135 0.135 0.181 0.221 34.074 63.703
Oop, at r=r. 1.636 1.636 1.668* 1.698* 1.955 3.789

Error =100(FEM-Analytical)/Analytical (%)

*Estimated from the finite element results

Support Condition
In this case, internal support pressure of 1 MPa is applied on the boundary wall of the tunnel.
Figure 7.14(a) shows that due to the support pressure, radial displacements at the tunnel
boundary are reduced by 100% and 32.46% from no support case for non-dilatant and dilatant
rock mass respectively. The radius of elastic-plastic boundary is 2.53 m signifying the reduction
in plastic zone around the tunnel boundary [Figure 7.14(b)]. Table 7.5 lists the results obtained
from analytical and finite element solutions. It can be seen that the effect of dilation is
negligible in terms of stresses and displacements if adequate support pressure is applied.
Results obtained from finite element analysis are in close proximity with that of analytical
solutions.

0.006

0.004

(m)
-u
‘0.002

0.000

Analytical dil = 0 deg Analytical dil = 15 deg

x FEMdil=0 deg Oo FEMdil = 15 deg

Figure 7.14(a) Distributions of radial displacements (Hoek-Brown rock media, support condition).
 Chapter 7 Elastic-Plastic Finite Element Analysis — [ 189

#275

1-50

US

1.00

0.75 |
Or/Po
Oh/Po
OF 0.50

0.254 &

0.00 4
1 1.5 2 25 3 3.5 4

: Analyical oO
ria

FEMDi=15deg x FEM dil = 0 deg|

Figure 7.14(b) Stress distributions in rock around the tunnel boundary (Hoek-Brown rock media, support
condition).

Table 7.5 Comparison of results between analytical and finite element method (support condition)

Parameter Analytical FEM Error (%)

Wi OF y= 15° y= 0° y = 15° y = 0° wy = 15°

rJa 1.266 1.266 1.260* 1.260* -0.473 -0.473

-100u/a (%) atr=a 0.198 0.217 0.204 0.223 3.030 -2.765
Og/P,o atr=a 0.879 0.879 0.896 0.907 1.934 3.185
OedPo at r = Te 1.636 1.636 126557 1.658* 1.161 1.345

Error =100(FEM-Analytical)/Analytical (%)

*Estimated from the finite element results

SUMMARY
This chapter has dealt with the plastic response of rock material. In the plastic regime, response
(reaction) of structures depends on the displacements. As a result, the stiffness (matrix) of the
structure also becomes dependent on the displacements. Thus, solutions of primary variables
such as displacements are not straight forward. In this text, three non-linear solution methods,
Newton-Raphson (NR), modified NR and incremental method are illustrated with examples.
The solution method for multi-degree-of-freedoms non-linear system of equations is also
explained with example. The concept of plasticity theory is explained clearly mentioning the
various items (i) yield criteria (ii) flow rule (iii) hardening/softening rule. The implementation
procedures of shear yielding criteria based on Mohr-Coulomb, Drucker-Prager, Hoek-Brown
and Alternative criteria and tensile criterion are also explained with examples. Analytical
solutions of elastic-plastic rock mass media is explained in detail for both Mohr-Coulomb and
the
Hoek-Brown yield functions. The dilatant behaviour of rock mass is also incorporated in
 Finite Element Method: Concepts and Applications in Geomechanics

derivation of radial displacement equation in the plastic region. Finally, the tunnel example is
analysed again using both of these yield criteria considering elastic-perfectly plastic conditions.
For Hoek-Brown rock mass media, two loading conditions are assumed (i) no-support condition
(ii) support condition while for Mohr-Coulomb rock mass media, only no-support condition is
considered. The finite element analysis is performed using axisymmetric models considering
dilatant and non-dilatant rock media. Results are compared with the analytical solutions and it
is found that they are in close proximity to each other.

EXERCISES
7.1 The relationship between load (P) and displacement (u) of a non-linear spring can
be approximated using the following equation:
P(u) = 100 In(u) + 250Vu — 150
When the load in the spring is increased by 400 units from the initial condition
(point A), having u, = 1 unit and P, = 100 units, the spring reaches to the new
point B as shown in the figure below. Determine the displacement at point B (ug)
using Newton-Raphson method (show results after 3 iterations).

K( u)

J
WADA BY
Fig. P7.1

7.2 In the above problem, determine the displacement ug using Modified Newton-
Raphson method (show results after 6 iterations).
7.3 Solve Problem 7.1 using incremental method or Euler’s method.
7.4 Three identical non-linear springs are connected in series as shown in Figure
Problem 7.2. Considering the load displacement relationship of each spring to be

P(u) =1200u"/> +100

determine the displacements g;, q2 and q3 if the loads increments as AF, = 100,
AF, = 50 and AF; = 500 units. Assume the initial condition as q, = 0.01, gz = 0.02,
ga.= 0.03:

2 Po k, > % ky b> a
FNAL
Fig. P7.2

7.5 (a) Determine whether following trial stress condition satisfies the Mohr-Coulomb
criterion with c = 5 MPa and @ = 30°.
| Chapter 7 Elastic-Plastic Finite Element Analysis
191

=) l 0
] —16 —2| MPa
0 2 Zz
(b) If part (a) is satisfied, determine the elastic and plastic strain increment vectors
assuming associative flow rule and elastic-perfectly plastic rock material if the
total strain increment vector is
Ae’ ={-0.88 -6.16 6.48 0.96 1.92 0.00}x10~. Assume elastic rock
properties as E = 5 GPa and v = 0.25.
(c) Determine the updated stress vector.

7.6 Recalculate Problem 7.5 assuming non-associative flow with dilation anglel5° and
elastic-perfectly plastic condition.
Det Compute Problem 7.6 replacing Mohr-Coulomb criterion with Drucker-Prager
criteria (inner circle).
7.8 Estimate the ratio of elastic portion, m given in Eq. (7.45) for the Problem Problem
TeJamlt (i= m)Ae! is divided into 3 sub-increments, determine the updated stress
vector of Problem 7.5 based on the procedure outlined in sub-section 7.5.3. Show
that the updated stress vector obtained in this method is more accurate as compared
to that of calculated in Problem 7.5.
da, Assuming uniaxial stress 0,, = p (p < 0 and p > yield stress) and all other stresses

are zero, determine (<=) given in Eq. (7.32) for Drucker-Prager yield surface
oO
(outer) with Ng = gq.
7.10 In the Problem 7.9, determine the elastic-plastic constitutive matrix [Eq. (7.30)],
assuming modulus of elasticity of rock is E and v = 0 for associative flow and
elastic-perfectly plastic rock material.
Told. The stress vector in MPa at 10th load step of a 3-noded triangular element is found
to be of = {-2 —8« 1 3}, Assuming plane strain condition and Hoek-Brown
criterion with E = 6 GPa, v = 0.3, GSI = 60, m; = 10 and o,; = -60 MPa, determine
whether the element has yielded at 10" load step.
Let us assume that the increment of strain vector at 11th load step is

Ae* =125 -45 0.3 0}x10~

Determine the following:
(a) Whether the element yields at 11™ load step.
(b) Ratio of elastic portion m, as given in Equation 7.44.
(c) Plastic strain vector assuming associative flow and elastic-perfectly plastic
condition. Use the strain components considering the result of part (b).
(d) Increment of stress vector.

7.12 Compute Problem 7.9 replacing with Alternative 1 criterion.

7.13 Compute Problem 7.10 replacing with Alternative | criterion.
 Strength and
Deformability of Jointed
Rock Mass

8.1. Introduction
Up till now, discussions were limited to isotropic and homogeneous material such as intact
rock. The finite element methods were described for mainly dealing with elastic and elastic-
plastic behaviour of rock and other material. However, due to the presence of bedding or
cleavage planes, joints and fractures, behaviour of rock becomes highly anisotropic. In Section 2.3,
we have defined the simplest form of anisotropy, i.e linear transverse isotropy and this concept
is often used for design in geotechnical problems. Several texts have illustrated in detail about
the concept and mechanical behaviour of anisotropic rocks (Obert and Duval, 1967, Jaeger and
Cook, 1979, Brady and Brown, 1985, Goodman, 1989). In this chapter, we will mainly
concentrate on the strength and deformability aspects of jointed rock mass.
Mechanical properties and behaviour of jointed rock mass were studied by many
researchers using natural and artificial samples. Some of the researchers who have significantly
contributed in understanding of mechanical behaviour of jointed rock mass are Patton (1966),
Goodman (1970), Barton and Choubey (1977), Bandis et al. (1981) and others. A detailed
review on rock joints and their shear strength was presented by Barton (1976). Gerarrd (1985)
and Sun et al. (1985) presented a detailed study on the formulation of compliance matrices of
jointed rock mass which have contributed to the understanding of deformability of jointed rock
mass. Pande et al. (1990) commented that joints in rock masses can vary widely in their
physical state and mechanical behaviour since joints can be fresh or weathered, asperities
matching or mismatching, filled or unfilled with gauge material. Thus mechanical properties
such as strength (fracture and frictional) and deformability of jointed rock mass depends on the
roughness of joint plane, interlocking properties of asperities, applied normal stress and residual
properties of the joint surface.

8.2 Fracture Strength of Jointed Rock Mass

Strength of a jointed rock mass can be defined as fracture strength or frictional (shear) strength.
41am
 Chapter 8 Strength and Deformability of Jointed Rock Mass
The fracture strength is measured by either uniaxial or triaxial tests of jointed rock
sample. The
Shear strength is estimated by direct shear tests or triaxial tests of rock sample under
constant
normal stress. In the following sections, discussions will focus on the fracture strength
and
shear strength properties of jointed rock mass.

8.2.1 Fracture Strength with Single Joint

Jaeger (1960) presented an instructive example of fracture strength of rock element having a
single joint where the joint plane is inclined at an angle B with the minor principal stress
direction as shown in Figure 8.1.

O71

Joint

O3
Figure 8.1 A single joint plane in rock sample.

The shear strength of the joint is defined by Mohr-Coulomb criterion as

|t|=c; —o tang, (8.1)

where c; and @; are the cohesion intercept and angle of internal friction respectively of the joint
plane. Considering compressive stress as negative value and then applying the transformation
of stresses as defined in Chapter 1, we find

el
ES
0, +o 0; —O
Se ao cos2B for 0° < B < 90° (8.2a)

Iz| =— eae sin2B for —90° =.B < 0° (8.2b)

Inserting Eqs. (8.2a) and (8.2b) into Eq. (8.1) and rearranging, we can write
2(03 tan @; ~g cj)
OV 03
iow.
oe ee
(1—tan@, cot B)sin 2B for 0° s B < 90° (8.3a)

2(03 tan; —C;)

0; = 0. - for -90° < B < 0° (8.3b)
aaa (1+tan@; cot B) sin2B
 Finite Element Method: Concepts and Applications in Geomechanics
194

Thus, if B = 90° or B = @;, the stress required to produce the slip along the joint plane
is infinity. Let us define these two limits as Bnox = 90° and Bin = )- Then slip would occur
if Buin < B < Bmax. Figure 8.2 shows a typical plot of difference in principal stresses 0; — 03
vs. angle of inclination B taking @; = 30°, 0; = -2MPa and c; = 2 MPa. It can be seen that
the peak strength of the sample is affected by the joint plane if angle of inclination falls within
g; to 90°. If the angle of inclination is less than @;, the strength of rock sample is determined
by the property of intact rock. Similar analysis can be performed considering the inclination
of the joint plane as —B degree with the 03 axis. In this case, we find that Byin = —90° and
Bmax = —%;- Then the slip would occur if Buin < —~B < Bmax. Hence the rock sample will fail
due to its intact strength for angle of inclination as —@; < B < @j. For all other cases, fracture
strength of the sample is affected by the existence of the joint plane.

Angle, B (deg)
-90 SUT 10 00 i eee 10 20 30 40 50 60 70 80 90

Slip of joint

Fracture
strength of
rock

differential
(MPa)
Maximum
stress
6,
03
—

Figure 8.2 Variation of peak fracture (slip) strength with inclination of the joint.

By differentiating Eq. (8.3a) with respect to B and then equating it to zero, we find the
value of 8 for which the minimum strength occurs.
tan 2B =—cot d; (8.4a)
or
n 9;
Big
=—+—
ties (8.4b)
Similarly, for B ranging between —90° to 0°, the value of B for which the minimum
strength occurs will be

ee
pa { 4 (8.4c)
Thus the minimum strength is obtained by replacing B in Eq. (8.3a) or (8.3b) as given
in Bo*(8.5).
 Chapter 8 Strength and Deformability of Jointed Rock Mass

‘es 1+sing; 2c; cosd;

en fe asi G.j (8.5)
j
Brady and Brown (1985) commented that when B approaches to 90° or within 0 to
g;, the slip on the joint plane cannot occur and thus the peak strength of the rock specimen
for a given value of 03 must be governed by some other mechanism, probably shear fracture
through the rock material in a direction not controlled by the joint plane. Figures 8.3(a) and
8.3(b) show the variation of the peak strength of phyllite and slate rock samples with angle
of inclination of the minor principal stress to the plane of joint. Comparing Figures 8.2, 8.3(a)
and 8.3(b), it can be found that the shape of the peak strength curve with angle B, does not
match with actual test data. In particular, the peak strength value for B between 0 to @,, is not
a plateau as given in Figure 8.2. Based on similar concept, Brady and Brown (1985)
commented that theoretical fracture strength of jointed rock sample is the oversimplified
representation of peak strength variation in anisotropic rocks. It is proposed that cohesion
intercept, c; and friction angle @; are not constants and they vary with B.

B, deg B, deg
al 0 15 8045 860 is 90 a 0 10 20 30 40 50 60 70 80 90
6 ; 6 0 ee
{ | i
6 6 —100 Oe aoe
a a ° oO
6 & —200 One tae tae Oe pas ie ae
Da Da Cy fa ee al :
x 3 -300 oO A af a hd Oo ;

z = aie anou
2 2 400 a @ @

5 cee eee gee

2 5 -600| ‘
£ = ®
I 52100 == =}
2 [© 50 MPa 4 180MPa @® 200 MPa 2 035 MPa 070MPa & 140 MPa @ 210MPa @ 275 MPa|

(a) Moretown phyllite (b) Slate

Figure 8.3 Variation of peak principal stress with angle of inclination of single joint for the confining mention
in the legend (Data source: Brady and Brown, 1985).

8.2.2 Fracture Strength with Two Joints

Figure 8.4 represents a rock specimen with two joints with angles of inclination of B, and B,
degrees with the horizontal axis, Zhu et al. (1993) suggested that the most significant factor
for two joints case will be the plane of joint where sliding begins. They have also suggested
that the sliding will depend on the geometry of the combination of the joint sets with the rock
mass, the stress distribution, the shear strength and the anisotropy. In general, once the sliding
path is determined, sliding continues along that path. Thus it is important to identify which
joint plane is the dominant for sliding.
 Finite Element Method: Concepts and Applications in Geomechanics

OF;

Joint plane |
Joint plane 2 |

OF;
Figure 8.4 A rock sample having two joints under triaxial loading.

Let A, and A, are such that

ie eens
ees Sexe (8.6a)
(1— tan 9; cot B,) sin 2B,
and

A : (8.6b)
~ (I= tan 9,3 cot By) sin 2B,
If B varies between —90° to 0°, A; and A, will be expressed based on Eq. (8.3b). Hence
the peak strength formula expressed in Eq. (8.3) can be rewritten for both the
joints as
For Joint 1: 0) = 63 + 2(63 tang; — cj) Ay (8.7a)
For Joint 2: 0} = 03 + 2(63 tan jy —cj7) Ag (8.7b)

If the magnitude of the right hand side expression of Eq. (8.7a) is more than that of
Eq. (8.7b), then sliding will occur along the joint 2 and vice-versa. If the cohesion intercepts
and friction angles are the same for both the joints, then same analogy is applied just by
comparing the values of A, and A). For this special case, sliding occurs along the joint plane
1 if A, < A, and thus the peak strength for the specified 03 will be expressed by Eq. (8.7a).
Similarly, sliding occurs along joint plane 2 if A; > A> and the peak strength of the rock
specimen will be computed using Eq. (8.7b).

8.2.3 Fracture Strength with Multiple Joints

For three or more joints sets, the concept outlined in Section 6.1 is applicable. Hoek and Brown
(1980) pointed that the deformation and strength properties of rock mass having four or more
joint sets can be considered as isotropic. Brady (1987) also suggested that for heavily jointed
rock mass, we can again consider it to be pseudo-continuous [Figure 6.1(d)]. Thus for multiple
joint sets, we can apply Hoek and Brown strength criterion as given in Eq. (6.27)
for
| Chapter 8 Strength and Deformability of Jointed Rock Mass | 197

determination of rock mass strength. In this case, geological strength index (GSI), m, value,
and uniaxial compressive strength of intact rock must be estimated.

Example 8.1 In an artificial rock sample, two joints are made with orientation of 30° and
ao) with 03 axis respectively. Considering cohesion and angle of internal friction of both
joints to be identical, determine the sliding path once axial load is applied on the sample.

Solution: Since cohension and friction angle are same for both the joints, we can only
compare the magnitude of A, and A, as given below:
1
—————
(1 — tan @ ;, cot 30°) sin 60°

1
A=
(1 — tan 9, cot (—45")) sin(-90")

The figure below shows the plot of A; and A, with respect to friction angle up to 28°.
It is clear that magnitude of A, < A, and hence, joint 2 will slide first.

0 10 20 30
o; (deg)

8.3 Shear Strength of Rock Joint

Shear strength of a rock joint can be measured using direct shear testing apparatus or using
triaxial compression test. In direct shear test, the joint surface is oriented parallel to the
direction of applied shear load and the two halves of the sample are fixed inside a shear box,
using Cycap, concrete, plaster or epoxy (Goodman, 1989). In general, lower block is fixed for
any movement while shear load is applied to the upper block with constant normal stress as
shown in Figure 8.5(a). Figure 8.5(b) depicts a 50 kN capacity direct shear testing apparatus.
A typical shear stress-shear displacement curve is shown in Figure 8.6(a), clearly marking the
peak and residual strengths and the corresponding displacements. Shear displacement across the
joint is designated by u and is the measure of the difference in displacements (parallel to the
joint plane) between upper and lower blocks. Shear stress is measured as the ratio of shear load
to the average contact area between the upper and lower samples. From direct shear tests, for
sliding of joints, the normal and shear displacements during shearing are measured easily and
 Finite Element Method: Concepts and Applications in Geomechanics

Upper half
fo}n of the

Cement, plaster sample

OF epoxy
material
(a) Arrangement of direct shear testing (b) Direct shear testing apparatus
Figure 8.5 Direct shear test of jointed rock sample.

after the peak shear strength (1,) is attained, the minimum value of shear strength termed as
residual shear strength (T,) is also estimated. From the shear stress-shear displacement curve,
shear stiffness, k, can also be estimated as shown in Figure 8.6(a). Normal stiffness, k,, of
joint surface is measured from the normal stress-normal displacement curve as shown in
Figure 8.6(b). Note that this curve is non-linear and hence &k, is a function of normal
displacement.
If the joint plane is rough or undulating, the joint will tend to contract or dilate during
the shearing. The term dilatancy, v(7), is the difference between normal displacements of upper
and lower blocks as a result of shearing or shear displacement, u. In general, the rate of
dilatancy is the highest at peak shear strength [Figure 8.6(c)].
T

(a) (b)

v (Tt)

Dilation
(+)

Contraction
(—)
(c)
Figure 8.6 Relationship between (a) shear stress-shear displacement, (b) normal stress-normal displacement and
(c) normal displacement (t)-shear displacement.
| Chapter 8 Strength and Deformability of Jointed Rock Mass
| 199

For similar type of joint samples, the direct shear tests are conducted separately
for each
sample by applying different normal loads. From laboratory test data, it is establishe
d that the
peak shear strength increases with applied normal stress. Over the years, several shear
strength
envelops are developed based on theoretical and empirical studies. A few significan
t shear
strength envelops are described below.

8.3.1 Barton’s Empirical Shear Strength Envelop

Barton (1976) proposed an empirical peak shear strength criterion of rock joints based on direct
shear test results of natural and artificial specimen. The envelop of peak shear strength is
dependent of normal stress, roughness of joint, joint wall strength and residual friction angle
as given below:

1 =~, tan FRclogio S04] (8.8)

n

where
Q |=
peak shear strength is only the magnitude considered
0, = effective normal stress, for compressive normal stress, negative value is used
% = basic or residual friction angle. It is the slope of residual shear strength and
normal stress curve and is estimated from direct shear tests conducted on

A
perfectly flat (smooth) joint surface.
HT Joint roughness coefficient ranging from 0 to 20 (0 being flat smooth surface
and 20 being undulating rough surface). The first angle inside the bracket in
left hand side of Eq. (8.8) is contributed due to inclination and asperities of
joint. In general, JRC is determined by back calculating using Eq. (8.8), Figure
8.7 or procedure outlined in ISRM (1978).
JCS = Joint wall compressive strength. For unweathered joint it can be equal to
unaxial compressive strength, o,;. Determination of JCS from Schmidt Hammer
tests is given in ISRM (1978). .
Equation (8.8) is used with low level of normal stress (|0,,| << o,;) which means that the
ratio of JCS to o, may vary from about 10 to 10° and is appropriate for geotechnical
problems. However, Barton also proposed modified criterion for high normal stress and will
be discussed later. The major features of Eq. (8.8) are summarized below:
1. Barton (1976) commented that for very low effective normal stress, shear strength
envelop for rough joints resembles vertical tangent at or close to the shear stress axis
and a cohesion intercept is unlikely unless asperities are inclined at more that 60° to
the mean joint plane. He has also proposed that cohesion intercept for peak shear
strength is inherently dangerous even if the extrapolation is made from the mean
effective normal stress level appropriate to the particular engineering problem.
2. JRC is the measures of joint roughness as mentioned before.
3. For unweathered and dry joint surface, JCS is replaced by the uniaxial compressive
strength of rock. Since weathering is more prominent along the joint surface than the
 Finite Element Method: Concepts and Applications in Geomechanics

JRC = 18-20

(Figure not
to scale)

Figure 8.7 Roughness profile with corresponding JRC values (Source: Hoek et al., 1998).

body of rock, the JCS is much lower than the uniaxial compressive strength. It can
be noted that for rough undulating joints, reduction in JCS value due to weathering
will be higher in comparison to flat and smooth joints. The reduction factor of JCS
can be as low as 25% of its original value (dry condition).
4. The peak dilation angle is the maximum dilation angle for a given level of normal
stress and JRC. This angle refers to the instantaneous inclination of the shearing path
at peak strength, relative to the mean plane of the joint or fracture (Barton, 1976).
It is determined from the instantaneous slope of the normal-shear displacement curve
at peak shear strength point. Barton et al. (1985) proposed that peak dilation angle (6,)
can be approximated using the following equation:
 | Chapter 8 Strength and Deformability of Jointed Rock Mass
ij [ 201 |

1 JCS
63 = gd RE (peak) logio
(8.9)
n

5. The basic (residual) friction angle ranges from 28.5° to 31.5° for different
rock types
having flat non-dilatant joint surfaces. However, for unweathered undulating
rock
surfaces this value can be as high as 25° to 35°. In case of strong weathered
joint
surfaces, the value of ¢, may be as low as 15°, even in the absence of clay
filling
(ISRM, 1978). Table 8.1 shows the values of @ for different rock types and level of
normal stress.

Table 8.1 Basic friction angle for various rocks (Source: Barton, 1976)
SS ee ee eee eee eee
Rock Moisture 6, (MPa) @, (deg)
Amphibolite dry 0.1-4.2 82
Basalt dry 0.1-8.5 35-38
wet 0.1-7.9 31-36
Conglomerate dry 0.3-3.4 35)
Chalk wet 0-0.4 30
Dolomite dry 0.1-7.2 31-37
wet 0.1-7.2 27-35
Gneiss (Schistose) dry 0.1-8.1 26-29
wet 0.1-7.9 23-26
Gneiss (f.g.) dry 0.1-7.5 31-35
wet 0.1-7.4 29-3]
Gneiss (c.g) dry 0.1-7.3 31-35
wet 0.1-7.5 31-33
Limestone dry 0-0.5 33-39
wet 0-0.5 33-36
dry 0.1-7.1 37-40
wet 0.1-7.1 35-38
dry 0.1-8.3 37-39
wet 0.1-8.3 35
Porphyry dry 0-1.0 Si
dry 4.1-13.3 331
Sandstone dry 0-0.5 26-35
wet 0-0.5 25-33
wet 0-0.3 29
dry 0.3-3.0 31-33
dry 0.1-7.0 32-34
wet 0.1-7.3 31-34
Shale wet 0-0.3 27
Siltstone wet 0-0.3 31
dry 0.1-7.5 31-33
wet 0.1-7.2 27-31
Slate dry 0-1.1 25-30

Figure 8.8 reproduces peak shear stress-normal stress plots for the four sets of model joint
tested by Bandis et al. (1981). Here, normal stress is considered as to have negative values. This
 anics |
202 Finite Element Method: Concepts and Applications in Geomech

32° and uniaxial

figure shows the peak shear strength envelop using Eq. (8.8) for @ =
are back calculated as the
compressive strength of model material as 2 MPa. The JRC values
shear strength
best fit curve for JCS = 2 MPa. It can be seen that for higher JRC value, peak
of (7/o,,) for
is higher for a given normal stress. Barton (1976) summarized the arctan values
_ some common rock type and is reproduced in Table 8.2.

- 140

120

100

80

60

40 (kPa)
shear
Peak
stress

p AD

Normal stress (kPa)

JRG=16:6 IRC=10.6 «AIRC =7.5_ JRC=65

Figure 8.8 Peak shear strength envelops for 4 sets of model joints (Data source: Bandis et. als) LOSI:

Table 8.2 Maximum values of peak (7/o,,) measured on joints during tests at low normal stress
(Source: Barton, 1976)

(1/6,,) Arctan (1/6,,)

Description of discontinuity (MPa/MPa) deg
Limestone: slightly rough bedding surface 0.68/0.16 77
0.66/0.21 72
1.68/0.60 71
Limestone: rough bedding surface 0.68/0.31 66
2.07/0.68 JZ
Shale: closely jointed seam in limestone 0.06/0.02 71
0.06/0.02 70
Quartzite, gneiss and amphibilite
discontinuities beneath natural slopes - 80
discontinuities beneath excavated slopes - ip
Granite: rough, undulating,
artificial extension fractures 0.45/0.15 2
0.92/0.35 69
 Chapter 8 Strength and Deformability of Jointed Rock Mass 203

For high normal stress level i.e. 6, > 6,;, JCS in Eq. 8.8 is replaced by the confined
compressive strength (0; — 03) and the Barton’s peak shear strength envelop becomes

0; -o
£= oy] IRCiogo( = +) +44 (8.10)
n

8.3.2 Scale Effect on JRC and JCS

Effect of sample size on JRC and JCS were investigated by Barton et al. (1980) and Bandis
et al. (1981). It was found that peak shear strength and JRC reduce with larger profile of the
same joint. The JCS also reduces with increasing size of the sample. Details on scale effect were
given in those two references. On the basis of extensive laboratory testing of natural and
artificial joints and literature reviews, Barton and Bandis (1982) proposed two relationships for
scale effect of JRC and JCS as given below:
~0.02JRCp
IRC. = IRC o (2) (8.11)
Lo
-0.03JRCy
NO antes (Fe)
Lo
where JRCp refers to the value of JRC for a laboratory sample of 100 mm length (9). JRC,,
JCS,, refer to the insitu block sample of length L,,.

Example 8.2 Estimate the value of JRC for the data obtained from direct shear test of 10 cm
square samples as given below:

50... 88

Assume compressive strength and residual angle of the model material are 4 MPa and 30°,
respectively.

Solution: By trial and error method, we can calculate the error of estimation as given
in the table as follows:
 nics
Finite Element Method: Concepts and Applications in Geomecha

JRC values are changed gradually to find the minimum error. In this case, we find that
approximately error is minimum for JRC = 15.8.
ee SSS... SS.

Normal _ Exp. peak — Peak

stress shear strength shearstress —_—Error
(kPa) (kPa), T. (kPa) G = (@ = GS
0 0 0.0 0
—10 25 29.2 17.880
—20 43 45.7 7199
—30 64 60.4 13.207
—40 74 74.0 0.000
—50 88 86.8 1.338
Sum error = 39.625

8.3.3 Ladanyi and Archambault’s Shear-Strength Envelop

Ladanyi and Archambault proposed the shear strength envelop as given below (Goodman
et al., 1977):
o,(1—a,)(tani+ tang;)+a,Sp
1—(1—a,)tanitan 9; (8.12)
where tan i is the dilatancy and varies with normal stress as given in Eq. (8.25a). The area
of contact, a,, is given as follows:
3/2
On .
as =[1-&] LO me Oy: (8.13)
Or

where Oy can be replaced by uniaxial compressive strength of rock. The shear strength Sp of
rock comprises of asperities and varies with o,.

8.3.4 Mohr-Coulomb Shear Strength Envelop

Mohr-Coulomb shear strength criterion as described for intact or isotropic rocks in Chapter 6
is directly applicable for rock joints. In this case, the criterion is expressed in Eq. (8.1). The
cohesion intercept and friction angles of joint surface or gauge material are used in the
equation. A few differences are noticeable between Mohr-Coulomb and Barton’s peak shear
strength envelops as given below:
(i) Cohesion intercept is present.
(ii) Base or residual friction angle is absent.
(iii) Measures of joint roughness is imbedded in the friction angle.
(iv) The friction angle in Barton’s envelop is dependent on the magnitude of normal
stress and thus the envelop becomes non-linear.
For elastic-plastic analysis of joint surface, Mohr-Coulomb criterion is being popularly
applied due to its simplicity as compared to Barton’s envelop.
if Chapter 8 Strength and Deformability of Jointed Rock Mass | 205

8.3.5 Instantaneous Cohesion and Friction Angle from Barton’s

Envelop
Hoek (2000, Practical Rock Engineering) has derived relationships for estimating instantaneous
cohesion intercept and friction angle from Barton’s peak shear strength envelop. Instantaneous
cohesion intercept and friction angle are obtained by drawing a tangent to the Barton’s shear
strength envelop for a specified normal stress as shown in Figure 8.9. Mathematically, from
Barton’s peak shear strength envelop, instantaneous friction angle is obtained as given in
Eq. (8.14).
tT

Mohr-Coulomb
envelop

Barton’s envelop
%;

= On

= OT
g; = tan [= ) (8.14)

From Barton’s envelop and using scale effect for both JRC and JCS [Eq. (8.11)], we find

a) JCS mIRC, JCS

2 = -ian(Rc, logio a +4} n?[ me,
logio a +} (8.15)
00, -O,, 180 1n10 —O,,

Once the instantaneous friction angle is found, cohesion intercept can be obtained as

c; =T +0, tan@; (8.16)

Hence using c; and @¢, values obtained in Eqs. (8.14) and (8.16), we can apply Mohr-
Coulomb criterion as given in Eq. (8.1).

Example 8.3 Based on the Barton’s envelop obtained in Example 8.2, determine the
instantaneous cohesion and friction angle for o,, = —-35 kPa. Assume that the size of the sample
ise Ovem:

Solution: Since the size of the sample is 10 cm, scale effect of JRC and JCS will
not be applied. Using Eqs. (8.14) and (8.15), we find the instantaneous friction angle
iO) = 3.07-.
 206 Finite Element Method: Concepts and Applications in Geomechanics

Shear stress for o, = —35 kPa is calculated using Eq. (8.8) and found to be tT = 67.28
kPa. Hence, using Eq. (6.16), instantaneous cohesion will be c¢; = 19.68 kPa

8.3.6 Instantaneous Cohesion and Friction Angle from Hoek-Brown

Criterion
For multiple joints, similar relationships can be derived using generalized Hoek and Brown
criterion (Deb et al., 2005). Estimation of Mohr-Coulomb instantaneous c¢; and @; of rock
masses begins with the fundamental assumption that for a given confining stress 63, the tangent
of Eq. (6.27) will represent the equivalent Mohr-Coulomb surface as given below:
0,-0,N, +0; =0 (8.17)
ree NILth Prone (z+2). The tangent of both Eqs. (8.17) and (6.27) can be
% 1-sing zy)
written as

AC igeng sper acetal

tt te
00, Ny 1+am,(m,0;/6,+5)*" Gi
Substituting 0, expressed in Eq. (8.17) into Eq. (6.27), o”™ and instantaneous cohesion
intercept, c; can be estimated as given in Eqs. (8.19a) and (8.19b):

of =-03(1- Ng) — 0. (m,03/0, +5)" (8.19a)

c=
i Oe!
2.[Ng (8.19b)

From Eq. (8.18) and Ng equation, we find the instantaneous 9; as

: 1
g; =2tan JN — > (8.20)

8.4 Deformability of Rock Joint

Deformability analysis of joint surface reveals two important parameters—joint normal stiffness
and joint shear stiffness. These parameters are responsible for joint deformation in case of
normal and shear loads. In the following sections, the concept and application of joint normal
and shear stiffness along with compliance characteristics of joint surface will be discussed.

8.4.1. Joint Normal Deformation

Rock joints tend to close upon compressive normal loading. It is found that the relationship
between normal stress and normal displacement is non-linear and resembles hyperbolic curve
as given below [Goodman et al. (1977) and Bandis et al. (1983)]:

Vv

0, -| =, to (8.21)
Vin
| Chapter 8 Strength and Deformability of Jointed Rock Mass
5 [ 207 |
where V,,, is the closure of the joint measured from the axis 0,, = 0 as shown
by a typical normal
stress and normal displacement curve in Figure 8.10. The initial normal stress
0,9 1S measured
when v = 0. In the figure, sg denotes the initial sitting load. The maximum closure
V,,,,. (positive
value) is related to V,, as given in Eq. (8.22) below:

Vin VERS
mc?0
= Pod (8.22)

For joint opening, normal displacement is considered as positive and v 2 (V,,. + V,,). On
the other hand, joint closing occurs when normal displacement V,, < v < (Vn. + Vy). It is
assumed that joint has negligible tensile strength and thus for all practical purposes, tensile
(positive) normal stress is neglected. Saeb et al. (1992) suggested that since a joint traverses
a range of unmated conditions during shearing, its normal stiffness which is the tangent of
normal stress-deformation curve, also depends on the shear displacement. Thus we can express
that infinitesimal change in normal stress is contributed due to

oy sported een: dv = ky du + kyydv (8.23)

ou ov
where k,,, is the normal stiffness and k,, denotes the slope of the normal stress-shear
displacement curve. However, here we will restrict our discussion on normal stiffness only and
assume that k,, = 0. From Eq. (8.21), normal stiffness can be formulated as given in
Eq. (8.24). It can be noticed that normal stiffness increases with normal stress and will reach
to infinity when normal displacement approaches to V,,,.

x On geige Sain 8.24

ahiea a atieedhaat Soe
; 6
Hence, the normal stiffness at v = 0 and 0, = Oyo is (Kuno = Ono
=
m

=v

4--——- |——--
Closing |Opening

tr
mre
He
cere
me
ee
ee
ee,

Figure 8.10 Relationship between normal stress and normal displacement.

 in Geomechanics |
[ 208 | Finite Element Method: Concepts and Applications

if the magnitudes of initial

Example 8.4 Plot the normal stress vs. normal displacement curve
that the maximum closure
sitting stress and initial normal stress are equal to 0.1 MPa. Assume
is 4 mm.

plot as
Solution: Using Eq. (8.21), we can plot the normal stress-normal displacement
shown below:
Normal displacement (m)

-0.004 -0.003 —0.002 -0.001 0 »

b 4 1 —

L n

2-5
(MPa)
Normal
stress
V,, = 0.008 m |;
O49 = —0.1 MPa
-3.5

8.4.2 Joint Shear Deformation and Dilatancy

Consider again the shear stress-shear displacement curve shown in Figures 8.6(a) and 8.6(c).
It can be seen that shear stress-shear displacement curve has three distinct phases—pre-peak,
peak and post-peak. At the onset of shearing, the joint contracts first and then dilates having
a maximum rate of dilation at the peak shear strength. The peak shear strength, shear stiffness
and displacement at peak shear strength are functions of normal stress. It means that peak shear
strength increases with applied normal stress but dilation tends to decrease with increasing
normal stress. According to Goodman (1977), there are two types of shear stress-shear
displacement models with (i) constant stiffness, where peak and residual displacement changes
with normal stress (ii) constant displacement, stiffness during pre-peak shearing phase changes
with normal stress. Schematic diagrams are plotted to elaborate these two models as shown in
Figures 8.11(a) and 8.11(b). For both the models, the following relationship between shear
stress and shear displacement is applicable:

cay for u < Uy (8.25a)

Bp S18 Bo Then, tH
P Jaa Sete) for u, Su Su, (8.25b)
u,—u
Pp r
u,—u
Pp rc

jngends r SOP rte. (8.25c)

 iF Chapter 8 Strength and Deformability of Jointed Rock
Mass a [ 209 z

T T

Increasing o,, Increasing o,,

up uE Up U,
(a) (b)
Figure 8.11 Conceptual diagrams of shear stress-shear displacement with (a) constant stiffness and
(b) constant
displacement (Source: Goodman et al., 1977).

In order to quantify the rate of dilation, Ladanyi and Archambault proposed the secant
rate of dilatancy which is represented by
k
) o
Meeteanige pee talrog for u < u, and O, < oO; (8.26a)
ou Or

dv
Sry LOLS 0r0,6 >) Or (8.26b)

where Or is transitional stress, an independent material constant and can be obtained from the
experimental results (Seab et al., 1992). However, o7 is generally assumed as the uniaxial
compressive strength of intact rock. The constant k is assumed as 4. The angle & is the dilation
angle at peak shear strength for applied normal stress, o,, = 0 and is given as

Ov
69 = an (=)
au) |,
<0 (8.27)
Now, integrating Eq. (8.26a), we find that normal displacement is the function of shear
displacement and normal stress
k
va 22) tanig + f(O,,) (8.28)
256

The constant of integration in Eq. (8.28) must be the function of normal stress and
can be obtained by letting the initial condition u = 0 which is given in Eq. (8.21). Hence
Eq. (8.28) is rewritten as

oO ‘ oO
vau(1-2 tani + V,, [1-20 (8.29)
OT (OF,
 Finite Element Method: Concepts and Applications in Geomechanics

ent not
The above equation is a significant relationship which shows that normal displacem
also depends on the shear displacem ent, material and
only depends on the normal stress but
be found that normal stress is a function
joint parameters. From the above equation, it can also
stress,
of both normal and shear displacements. Since the shear stress depends on the normal
mal change _
it will be also be function of normal and shear displacements. Hence an infinitesi
in
in shear stress is contributed by the change in normal and shear displacements as given
Eq: (8.30):

dt = dus )Ea = k,du + kv (8.30)

For constant stiffness model, k,, for pre-peak phase is estimated by Eq. (8.25) with respect
to u as

ky eae k, (8.31)

kim is the slope of the curve between shear stress and normal displacement and is found to be
dependent on k,,,. However, here we will assume k,, = 0.
Now, combining Eqs. (8.23) and (8.30), the relationship between incremental stresses and
displacements will be
dt | | ky Kin ||du
do, i kn Kin dv Coaa

or

Ao = kAq (8.32b)

8.5 Concept of Joint Compliance

Consider a rock mass having n number of horizontal joints of negligible thickness as shown
in Figure 8.12(a). The joint frequency
f is defined as the number of joints per unit length of
rock mass; in this case f = n/L, where L denotes the length of rock mass. The intact rock is
the rock matrix between two consecutive joint planes. Let us assume that normal stress oO acts
on the rock mass as shown in the figure. It is obvious that the normal stress on the joint plane
and on the intact rock is the same. Then the deformation of intact rock, d' and that of joints,
d can be obtained as

di _oL
aE (8.33)

di = oe a ie
(8.33b)
k, k, Ky, m=l kin

where £; is the modulus of elasticity of the intact rock and k,, is the normal stiffness of mth
 Chapter 8 Strength and Deformability of Jointed Rock Mass ie2at|
joint. If all normal stiffness are equal as k, = k, = --- = k, =k, then the total displacement
including joints and intact rocks is given as follows:

=q' J= Seaee
l

Intact rock

(a) (b)
Figure 8.12 Concept of equivalent material (a) rock mass (b) equivalent material.

Now consider the rock mass can be replaced by an equivalent isotropic and homogeneous
rock having equal dimensions and elastic modulus of E,, so that the displacement incurred by
the equivalent rock is the same as given in Eq. (8.34) (Figure 8.12b). Hence, displacement
occurred in the equivalent material due to normal stress o can be expressed by

d _oL
Em
(8.35)

Then the equivalent elastic modulus is obtained by equating Eqs. (8.34) and (8.35) as

1 Lape
ya m
eel (8.36)

Equation (8.36) can also be written in the form of compliances as

Sin a Si + Sj (8.37)

where
Sm = compliance of equivalent isotropic and homogeneous rock
S; = compliance of intact rock
S; = combined compliance of joints

Equation (8.36) reveals an important property of jointed rock mass that the summation
of compliances of intact rock and joints is the compliance of equivalent rock. This property
has been used to model jointed rock mass numerically and will be explained in Chapter o
 [ 212 | Finite Element Method: Concepts and Applications in Geomechanics

8.5.1 Joint Compliance Matrices

Gerrard (1982) presented the generalized model of rock mass with multiple joints and also
showed examples of special cases of rock mass properties based on compliance matrix of joints.
Based on this study and concept given in Pande et al. (1990), we will present joint compliance
matrices for both two and three dimensional joints. First, consider a joint plane in two
dimensions as shown in Figure 8.13. The joint plane is aligned along the x’-axis and
y-axis represents the normal direction to that plane. We assume that displacement and stress
fields along shear and normal directions are u’ and v’, and tyy and Oy, respectively. Thus the
relationship between displacements and stresses can be written with joint compliance matrix in
the local coordinate system as

ae a Or Rett
em 38a
u S91 529 Tx'y’

q’=C’o’ (8.38b)
The compliance matrix, C’ can be non-symmetric. However, in practice, the off-diagonal
terms (57 and 5>;) are difficult to estimate and thus normally assumed to be zero. In this case,
the unit of compliance components s,, and s > is m/Pa and they are expressed as

1 1
St Pa ou ashy ce, aes (8.39)
nn tt

However, components of the compliance matrix having relationship between strain and
stress will be

1 1
nn = #) oes of dk, (8.40)
nn

where d = I/f = average distance between two consecutive joint planes. The complianc
e matrix
is then generalized by adding another two rows and columns with zeros
to represent
two-dimensional compliance matrix for plane stress or plane strain condition
s as given in
Eq. (8.41).

O05 O70
C’ = 0 Chn 0 0

0 GeO (8.41)
0 Oe20

For three-dimensional problems, the joint plane is represented by two

perpendicular axes
x and y’ and the z’ axis denotes the normal direction to the joint
plane. Hence oy, represents
the normal stress and t,,, and Tyy Signify the shear stresses in
x’ and y’ directions respectively.
Now considering the displacements in the normal directio
n (z’ axis) and shear directions
| Chapter 8 Strength and Deformability of Jointed Rock Mass [ 213 |

Joint line

Vil

Figure 8.13 Transformation of joint compliance matrix.

(x’ and y’ axes) as w’, uv’ and v’ respectively, we can relate displacement and stress vector with
compliance matrix as given below:

v’ =| 0 S99 0 Ty/z! (8.42)

u’ 0) 0 S33 Toy!

1 1
where ee, sense eR (8.43)
kan Ky fay

The parameters k,y and k,y are the shear stiffness along y’ and x’ directions respectively.
For general, three-dimensional stress-strain relations in Cartesian coordinate system, the
dimension of compliance matrix is 6 x 6 as illustrated in Chapter 2. In order to keep the parity
with that notion, we add another 3 rows and 3 columns filled with zeros in the compliance
matrix as given in Eq. (8.44).

0 OR50T -0°%F07F-0
Oa De 0.020
cr |9 9 om 9 9 0 ALD
OsjO mo SOOO
00 0 0uw 0
07 0,90 =<0F 0c,

Note that in Eq. (8.44), Cin ak Cty ay and ¢,, Big

 214 | [ Finite Element Method: Concepts and Applications in Geomechanics

8.5.2 Equivalent Compliance for Joints Sets

If the local (joint) coordinate system is not aligned with the global coordinate system, we need
to transform the joint compliance matrix into global coordinate system. If multiple joint sets
are present with different angles of inclination, we need to transform individual joint
compliance matrix of each joint set into global coordinate directions and then add all
compliance matrices together to evaluate the effect of all joint sets.

8.5.3 Transformation of Joint Compliance

First, consider the transformation in two dimensions. Refer to Figure 8.13, where x and y axes
are defined as the global coordinate system. We need to transform joint compliance matrix, C’
[Eq. (8.41)] from x’y’ plane to compliance matrix C in xy plane. As discussed in
sub-section 1.3.3, the strain energy of the body will be the same, no matter hat coordinate
system is used. Thus applying similar method as given in sub-section 1.3.3 and recalling € =
Co, we find that

C=T'C’T (8.45)
where T is transformation matrix of joint compliance in 2D as given below:

c s 26S 0)
2 2 5)
ra|5 c Res

cs =cH se (8.46)
OBO 0 1

For 3D cases, the transformation matrix T is used as given in Eq. (1.30).

8.5.4 Joint Compliances for Multiple Joint Sets

For ith joint set, the compliance matrix can be obtained by Eq. (8.45) in global coordinate
system as

Ce RA;
TICE (8.47)
Here T; represents the transformation matrix of ith joint set having compliance matrix
C’
in local coordinate system. Thus for n number of joint sets, the global compliance matrix
will
be the summation of all individual compliance matrices as given below:

c=), (8.48)
If the constitutive matrix of intact rock is denoted by D;, then equivalent complian
ce of
the rock mass will be

Ci = D;! +E (8.49)
[_ Chapter 8 Strength and Deformability of Jointed Roch Mass
| | 215 |
Hence the equivalent stiffness matrix, D,, can be obtained by inverting the above
matrix as

D,, = Se aa [D;" 3 Ch (8.50)

Example 8.5 Determine the equivalent compliance matrix for rock

mass having two joint sets
oriented 20° Goint 1) and 70° (joint 2) with the horizontal axis. Assume
the normal and shear
stiffnesses of both the joints as 10 GPa/m. The joint spacings of joint 1
and joint 2 are 5 m
and 2 m respectively. The modulus of elasticity and Poisson’s ratio of intact
rock are 6 GPa
and 0.25 respectively.

Solution: Using Eq. (8.47), we find the compliance for each joint set as

2.340E-06 -4.235E-22 -6.428E-06

_|-4.235E-22 1.766E-05 -6.428E-06
7!C’T, ~ |-6.428E-06 -6.428E-06 2.000E-05
0 0 0 ee
ey
Sa

4.415E-05 0 —1.607E-05
0 5.849E-06 —1.607E-05
i ls
—1.607E-05 -1.607E-05 5.000E-05
0 0 0 SKS
SL)

For intact rock, the compliance can be obtained 2.21 as

1.667E-04 —4.167E-05 0 ~4,167E-05

-4.167E-05 1.667E-04 0 ~4.167E-05
Cniaho neg 0 4.167E-04 0
4.167E-05 —4.167E-05 0 1.667E-04

Hence using Eq. (8.49), we find the equivalent compliance as

2.132E-04 -4.167E-05 -2.250E-05 -4.167E-05

-4.167E-05 1.902E-04 -2.250E-05 —4.167E-05
™~|9250E-05 ~2.250E-05 4.867E-04 0
-4.167E-05 -4.167E-05 0 1.667E-04
 | Finite Element Method: Concepts and Applications in Geomechanics |

8.6 Elastic-Plastic Analysis of Rock Joints

Above discussions on strength and deformability of joint mainly concentrated on elastic
properties of the joint surface. However, joints do yield and/or fail if applied shear stress
exceeds their shear strength. As shown in Figure 8.6(a), the initial shear response of joint
surface is elastic until the peak shear strength is achieved. Once the shear stress is higher than
the peak shear strength, joint yields and the load bearing capacity reduces. Thus for analytical
or numerical analysis, the peak shear strength envelops are considered as the yielding criteria.
In the plastic regime, shear stress reduces gradually with further shear displacement and finally
residual shear stress is attained. In this case, dilatancy represents the plastic flow rule.
For rock joints, the most widely used yield criterion is Mohr-Coulomb yield surface.
However, the peak strength criterion given by Barton and Choubey is also used as yield
criterion with suitable plastic potential function. In the following sections, we will discuss the
procedure for applying Mohr-Coulomb and Barton-Choubey peak strength criterion for rock
joints.

8.6.1 Mohr-Coulomb Shear Yield Criterion

For rock joints, Mohr-Coulomb yield surface is represented using the local shear and normal
stresses as

F* =|t|-c;
+0, tand;=0 (8.51)
Here, T is shear stress along the joint plane and o,, represents normal compressive stress.
The parameters c; and @; are the cohesion intercept and friction angle of the jth joint. The non-
associative flow rule is considered for shear yield surface with a plastic potential function
represented with variable dilation angle, y; as

Q=|t|+0, tan y ; = const. (8.52)

The dilation angle can be obtained from the experiment as given in Eq. (8.9). Pande
et al. (1990) found that the average inclination and height of asperities play a dominating role
in determining the rate of dilation and the maximum dilation that can occur on a joint. Dilation
angle at times significantly affect the peak shear strength ofjoints and also is highly dependent
on the normal stress. For low normal stress, dilation occurs once the joint opening is greater
than the average height of asperities. For high normal stress, dilation can be inhibited due to
shearing-off of the asperities.
For rock joints, no-tension cut-off is considered in tensile yield criterion since rock joints
are regarded week in tension. The tensile yield criterion thus given as

ff = oe g (8.53)
The associative flow rule is assumed with tensile yield criterion and the plastic potential
function Q = F" = const. Figure 8.14 depicts the Mohr-Coulomb yield criterion in T-s, plane
with no-tension cut-off. In this figure, plastic potential lines are also drawn for both the yield
surfaces.
 Chapter 8 Strength and Deformability of Jointed Rock Mass

—0),

Figure 8.14 Mohr-Coulomb criterion for rock joints.

8.6.2 Barton’s Shear Yield Criterion

The peak shear strength envelop given in Eq. (8.8) is converted as shear yield criterion and
is given in Eq. (8.54).

F° =||+0, an[IRCoso|IC4 |=
0 (8.54)
n

If the non-associated flow rule is assumed, Eq. (8.52) can be considered as linear plastic
potential function. However, Pande and Xiong (1982) have adopted a plastic potential function
for specific use with the above yield criterion as given below:

O, tana, JRC o
Se CONST: (8.55)
oat ky 264 JCS
where

ivy=IRClogio{ > mon

k, =1-tanA, tang,

Both of these yield criteria are being applied for the analysis of elastic-plastic behaviour
of joints.

SUMMARY
Existence of joints in rock mass changes the behaviour of stress pattern and deformation
characteristics. In this chapter, descriptions of rock mass fracture strength and deformability are
can
illustrated for single, double and multiple joints. It is shown that shear strength of joints
be obtained either by triaxial tests or by direct shear testing method. Based on laboratory tested
data of peak shear strength, Barton et al. had developed a practical peak shear strength envelop
 218 Finite Element Method: Concepts and Applications in Geomechanics

and is elaborated in this chapter giving numerical example. It can be noticed that Barton’s shear
strength envelop is non-linear as opposed to Mohr-Coulomb criterion and depends on roughness
(JRC), strength of joint wall (JCS) and residual angle. The concepts of normal and shear
stiffness of joint surfaces are illustrated relating incremental stresses and incremental
displacements in normal and shear directions. Equivalent compliance matrix of rock mass is
determined by first transforming compliances of joint surfaces into global coordinate system
and then combining with compliance matrix of intact rock. By inverting equivalent compliance
matrix, equivalent stiffness matrix of rock mass is evaluated and is used for the determination
of elemental stiffness matrix of finite elements intersected by joints planes. Finally, joint yield
criteria are expressed with Barton’s and Mohr-Coulomb envelops and is applied for elastic-
plastic finite element analysis of joint surfaces.

EXERCISES
8.1 A series of artificial rock samples having single joint is tested in triaxial
compression. The minimum fracture strength is recorded as 6 MPa for a confining
stress of 1.5 MPa. If the angle of inclination ofl the joint plane for the minimum
fracture strength is 60°, determine the cohesion and friction angle of the joint plane.
8.2 Show that for hydrostatic stress condition (6, = 63 = p), the internal friction angle
C;

of a joint plane is approximated byfj= tan! [S, where c; is cohesion of the joint

plane.
8.3 Determine the minimum fracture strength and the direction of slip of a doubly
jointed rock sample with the following conditions:
Joint plane 1: ¢ = 1.5 MPa, ¢ = 35°, B = -60°
Joint plane 2: c = 2.5 MPa, @ = 20°, B = 30°
Assume uniaxial loading condition.
8.4 A series of direct shear tests is conducted on similar type of joint surface (roughness
,
strength, and other properties are same). The peak shear stress, residual shear stress
are recorded for different applied normal stress as given in the table below:

On T te
—0.50 0.80 O25
—1.00 1.20 0.54
~1.50 1.40 0.90
—2.00 1.90 1.20
~2.50 2.10 1.41
—3.00 2.50 1.70
—3.50 3.00 2.00
 Chapter 8 Strength and Deformability of Jointed Rock Mass

Stress units are in MPa:

(a) Determine the Mohr-Coulomb peak shear strength envelop of the joint.
(b) Determine the Barton’s Peak shear strength envelop if uniaxial compressive
strength of the joint wall is 30 MPa.
(c) Determine the peak dilation angle for each case.
(d) Plot the peak shear strength with respect to peak dilation angle.
8.5 Estimate the instantaneous cohesion and friction angle from the Barton’s envelop
obtained in Problem 8.4 for each normal stress. Plot instantaneous cohesion and
friction angle with respect to normal stress.

8.6 (a) Determine the equivalent stiffness matrix in x y coordinate system of the rock
mass shown below:

8 MPa

Joint set 1: k,, = 50 GPa/m

ky = 15 GPa/m

Joint set 2: k,, = 100 GPa/m,

ky = 35 GPa/m

Rock matrix: E = 6 GPa,

De OES

Figure. P8.1

(b) Estimate the strain vector at any point on the AA’ line. Assume plane strain
condition.
8.7 In Figure P8.2, if the normal stress is increased from 0 to 0.1 MPa, determine the
normal displacement at the joint surface. Compare this displacement with equivalent
material model.

fn —() C
om
Figure P8.2
220 Finite Element Method: Concepts and Applications in Geomechanics

8.8 If the displacements along x and y directions at point P (Figure P8.1) are 0.05 mm
and 0.15 mm respectively, determine whether the joint slips at point P. Assume
JRC ="10.5,; JCS*='50' MPa‘ and ¢; = 30°.
8.9 Assuming k,,, = k, =k GPa/m, E = k GPa and v = 0, determine the equivalent elastic
stress-strain matrix of the rock mass given in Figure P8.3 below. If O,, = 0,, =k
MPa and 1,, = 0 at a point inside the equivalent rock mass, determine the strain
vector at that point considering plane strain condition.

k MPa

a/m
= k, = k GP
Joint set: kp,

k MPa
ee
oe Nw Rock matrix: E = k GPa
V=0

a N
Figure P8.3
9.1. Introduction
Strength and deformation characteristics of rock mass depend on the number of joint sets
present in the rock mass and also on their structural and engineering properties. For example,
the cause of a roof fall in a crossing of mine roadway or in tunnel may solely be attributed
to the conditions of joint surfaces in the roof even if the quality of intact rock is good. The
properties of joints, i.e. filled with gauge material, mated or unmated, average height of
asperities and others have great influence on the normal and shear stiffness of the joints. The
strength and stiffness of rock mass are greatly reduced from that of intact rock once the affect
of joint sets are included and are shown mathematically in the previous chapter.
Numerically, joints are modeled in two ways—as discrete joint element and as equivalent
rock mass material. If a joint is filled with gauge material and its thickness is at least twice
the average height of asperities, we can model such joints by solid elements having the
properties of gauge material. In this case, the joint is modelled considering solid thin element
composed of gauge material and located between the upper block (element) and lower block
(element) as shown in Figure 9.1. However, the slenderness ratio, i.e. length to height ratio,
of the element is too high to achieve desired numerical accuracy. Thus application of such
elements in numerical modelling is limited. Figure 9.2(a) shows a horizontal joint plane
bisecting a block into two halves. If a force is applied on the top block, displacement at point
A has two values—one for the top block and another for the bottom one [Figure 9.2(b)]. Since
the finite element analysis always seeks for unique solution at a nodal point, using continuum
mechanics, this problem cannot be solved (Pande et al., 1990). In the physical sense, the
displacement must be different at the contact nodes for upper and lower blocks under shear
loading. For analysis of this condition, joint or interface elements are introduced between two
solid elements. The purpose of interface or joint elements is to make smooth interpolations of
a discontinuous displacement function between top and bottom blocks [Figure 9.2(c)].
[ 222 | Finite Element Method: Concepts and Applications in Geomechanics

Upper layer

Weak thin layer

as joint
element

element

Lower layer
Figure 9.1 Weak and thin layers as joint.

Goodman et al. (1968) had proposed a special “joint element” suitable for implementing
joints especially for discrete plane of weakness such as fault, shear zone and others. This
element is connected in the inter-element boundaries and its stiffness matrix is formulated
based on shear and normal stiffness of joint surface and also on element geometry. In this
chapter, formulation of this element is given in detail. Further, the similar concept of joint
element was extended for 3D applications. Subsequently, Ghaboussi et al. (1973) had
introduced an improved joint element based on theory of elasticity. Beer (1985) developed
isoparametric joint element for 2D and 3D applications in the similar concept of Goodman’s
joint element.

cl

Ale
(a) (b)

displacement
Horizontal
Distance along BC
(Cc)
Figure 9.2 Interpretation of joint element (a) block before sliding (b) discontinuous displacement field
(c) discontinuity smoothed over the thickness of a joint element (Source:
Pande et al., 1990)
 Chapter 9 Finite Element Procedures of Rock Joints ial ic 223

If the number of joints in the representative volume of rock mass is relatively large or
a number of joint sets are to be modelled using finite elements, it will be almost impossible
to model all joints with the special joint elements as mentioned above. In such cases, a
equivalent constitutive law is developed for rock mass based on joint and intact rock properties
as given in the previous chapter. Pande et al. (1985, 1990) had implemented this technique
using viscoplastic material model having numbers of joint set and supports. Zhu et al. (1985)
used similar method using quadrilateral finite elements considering various intercept length of
joints in the element. In this chapter, the concept of equivalent constitutive law will be
implemented in the finite element procedures for elastic as well as elastic-plastic analysis.

9.2 The Discrete Joint Elements

Goodman et al. (1968) first proposed an independent joint element to be used with conventional
finite elements. This element was first developed for two dimensional finite elements and later
extended for three dimensional finite elements. Figuratively, joint element is one dimension less
than the conventional finite element and hence a joint element used with 2D solid elements will
be one-dimensional element and so on. As mentioned before discrete joint element is generally
used to model fault, shear zone or other discrete plane of weakness. If joints are ubiquitous
in the rock mass, this approach of modeling will not be appropriate with finite elements.

9.2.1 Goodman’s Joint Element: Direct Formulation

Figure 9.3 shows a 4-noded joint element of length L. The local coordinate axes are represented
by x’ and y’, where x’ implies shear direction and y’ signifies the normal direction. The joint
thickness (f) prior to deformation is assumed to be zero meaning that the nodes 1, 4 and 2,
3 are pairs or having same coordinates. The top and bottom solid elements are connected with
the nodes 3, 4 and 1, 2 respectively.

Figure 9.3 Joint element in two dimensions.

The displacement fields along shear and normal directions are represented by u’ and v’
respectively. Hence relative displacements between top and bottom elements can be expressed
as
Ur — Uy Ww,

Vis aap Wn
 Finite Element Method: Concepts and Applications in Geomechanics

where w, and w, represent the shear and normal displacements. The subscripts T and B signify
the top and bottom elements respectively. Since the thickness of the element is zero, we can
find that the shape function N, = N, and N3; = N>. Considering a line element of length L
between nodes | and 2, we find that

ee meetre
Linear interpolation for displacement fields results the following:
0
up = Nu +N, vp = Nyy + N23
ur = Nyuz,+Nyu4, vr = Nnv3 + Ni v4 (9.3)

Replacing Eq. (9.3) into Eq. (9.1), we find

, (9.4)

Equation (9.4) can also be written in the form of B matrix having the matrix of shape
functions as
w = Bq’ (9.5)
Here, q’ is the vector of nodal displacements and B matrix relates displacement field with
nodal displacements.
Now, consider that the shear traction and normal contact pressure on joint surface are
expressed as P, and P,, (in Pa). Then the traction stress vector and total potential energy (II)
can be written as

p={7| (9.6)
FF

h (Gj Pe

N= | w! Pdx’ — q’TF’ (9.7)

=e

where h is the thickness of joint element in the out-of-plane direction and force vector
F’
contains the nodal forces in x’ and y’ directions. Again the traction stress vector P is related
to the displacement vector w as
 Chapter 9 Finite Element Procedures of Rock Joints [ 225 |

l |=n
T Rae Ww 5)
P= =
om O k nn Wr
(9.8)

Replacing Eqs. (9.8) and (9.5) into Eq. (9.7), we find that

peel? ete pee

——
5 J wTPax qFT ae /T a
5 J wikwas
ae T Va
qF /T a
i Ja B'kBq‘dx’—q’"F’
*T pT , ie ae ‘Try
(9.9)
I —L/2

Applying the stationary principal of potential energy with respect to displacement vector
q’, we find that

an = c| BT kBdx’ Jvore
2

(9.10a)
oq -L/2
K’q’ = F’
(9.10b)

It is noticed that the dimension of the stiffness matrix K’ is 8 x 8 and is estimated

by integrating over the joint length L. In expanded form, the stiffness matrix is expressed as
given below:

| gps (9.11)
6h, 0 |. 228
Cee it 70
So) oe ool ES
‘i O™ 2k nn

In order to obtain the stiffness matrix in the global coordinate system (xy system),
stiffness matrix, K’ is transformed using transformation matrix, T as follows:
K = T'K’T (932)

where the transformation matrix T is expressed with direction cosines c = cos@ and
s = sin@ (Figure 9.3) as
Ben in Sel) 0
acurcis 0 0
Oph ae C RY

T= 0-0 -s 7S
> ©
©)
era
Soa].
(9:13)
ORO, 10
bo tbe al& oS
S&S
LES
SS] S&S]
SS
&

Da 0 Ral
O20" 0 eS
er
ee
&
[226 | Finite Element Method: Concepts and Applications in Geomechanics

The performance of this joint element and convergence of numerical solution highly
depend on the stiffness values and complexity of the problem including loading and boundary
conditions. In practice, high values of k,,, and k,, (atleast three orders higher than the estimated
values) must be used to ensure the convergence of the solution. It is noted that the properties
of joint introduced in this joint elements are only shear and normal stiffness, i.e. k,, and k,,,).
The off-diagonal terms are omitted. Hence the dilation ofjoint is inhibited in this joint element.
Dilation can be introduced if the off diagonal terms i.e. k,, and k,, are added to k matrix.
However, if k,, # k,,, the stiffness matrix will be unsymmetrical and thus numerical
computation will be cumbersome. In addition, recall that the normal stiffness k,,,, is not constant
and it depends on the normal displacement. Thus, a procedure should be adopted to adjust
normal stiffness value based on normal displacement. Specially, k,,, value should very large as
the joint closes.

Example 9.1 Determine the shear and normal stresses in a joint element if the displacements
at nodes in global xy coordinate system are as follows:
gq += {11404-2093 1.20 ~= 15.625 >11:79. 58.125. 141394 44.593} x 10° mm
The inclination of joint plane is 60° with the x axis, and assume k,,, = 50 GPa/m and ky
= 20 GPa/m.

Solution: First the displacements at each node are transformed into local coordinate
system and found as follows:

Local or ‘i/
o 3 . nodes
ie Ps 2473 —0.487
-10.200 | -11.897
49.437 -32.764
—41.710 S154

Let us calculate the shear and normal displacements at x’ = 0 using Eq. (9.5).
Hence

{w’ w"}! ={-39.237 20.867}

The shear and normal stresses are then -7.847 MPa and —10.433 MPa. Readers
can verify
that these values are the same for any point in the element.

9.2.2 Numerically Integrated Joint Element

Beer (1985) has derived similar joint element using isoparametric
formulation considering a
2 x 2 integration scheme as given for quadrilateral elements. In
this case, shape functions are
derived in the natural coordinate system as mentioned in Chapter
4. Formulation of interface
| Chapter 9 Finite Element Procedures of Rock Joints = [ 227

or joint element for three-dimensional modelling was also mentioned in that paper. However,
since joint element is essentially a line element in two dimensions, it can also be integrated
using two Gauss ‘points (-1//3 , 1/,/3 ) having weights of unity considering parent element
ranging from —1 to 1 as shown in Figure 9.4. Here, the shape functions have to be defined
using natural coordinates such as r instead of x’ and Jacobian J has to be estimated as in
Eqs. (9.14) and (9.15) respectively.

Physical element -1 Parent element

Figure 9.4 Joint element with numerical integration.

N =5il-n), Np =5(l+r) (9.14)

1/2

1-{(Z) (¥)
age aa ae (9.15)
Hence for isoparametric formulation, the coordinates are linear functions of shape
functions and are expressed as
x’ =Nixj +N 4x5 (9.16a)
y= Nyy + Naya (9.16b)
The stiffness matrix is then obtained by integrating numerically using Gauss quadrature
tule as
Z
K’ =) BY (7) kBG;)J(7) (9.17)
i=l

where the components of B matrix in Eq. (9.4) is replaced by Eq. (9.14).

Example 9.2 Estimate the uniaxial strength of a jointed rock sample with the following data:
The joint properties are: c; = 3 MPa and @; = 15°. Assume Mohr-Coulomb criterion and verity
Eq. (8.3a).

Solution: Using Eq. (8.3a), we can calculate the fracture strength based on angle
of joint inclination as shown in the following figure (a). Finite element model of jointed
rock sample is prepared with 4 linear quadrilateral elements and 2 joint elements as shown
in the following figure (b). Joint elements are connected by nodes 4, 5, 8, 7 and 5, 6, 9, 8.
The angle of inclination is varied from 20° to 60°. Using the finite element code developed
 228 Finite Element Method: Concepts and Applications in Geomechanics |

by the author, the fracture strength for each inclination of joint plane is obtained and plotted
in figure (a). It can be seen that finite element results matched closely with the theoretical
values.

Inclination angle (f), deg

OF 10> 207 307-40 =50260 +70 80-90 100

Strength
sample,
of
MPa
@ FEM —— Theoretical

(a) (b)

9.3 Analysis Procedure for Non-Linear (Elastic-Plastic)

Joints
The non-linear behaviour of joints is characterized by slip and separation (opening) that
undergoes in a joint. The interface or joint element mentioned above is used for non-linear or
elastic-plastic analysis.

9.3.1 Tensile Yield Function

Normally, tensile strength of joint is assumed to be negligible and thus opening ofjoint planes
is characterized by tensile (positive) normal stress. Thus the yield function for this case is
Fi =, (9.18)
9.3.2 Shear Yield Function
If the shear stress of the joint exceeds the peak shear strength envelop given either
by Mohr-
Coulomb or Barton (Chapter 8), irreversible slips occur. In general, shear yield
criterion refers
the onset of slip between the upper and lower joint surfaces or elements. The
general shear
yield criterion can be expressed as
F* =|t|+@ (9.19)
where

@=0, tang; —c; if 6, < 0 (Mohr-Coulomb)

| Chapter 9 Finite Element Procedures of Rock Joints | 229

JCS
@=6, tan [Rclogio (25)a | if 6, < 0 (Barton)
Roo

9.3.3 Plastic Corrections

Non-linear analysis proceeds in load steps and within each load step, several iterations are
performed. The step wise procedures for non-linear analysis of joint elements are outlined
further.
Step I. After solving the displacements in global coordinate system, displacements in
x’ y’ coordinate system are transformed using Eq. (1.17). Hence displacements w’, and v; for i
= 1 to 4 can be obtained from u; and v,.
Step 2. Using Eq. (9.8), trial shear traction tT and normal contact pressure o,, are evaluated
at every Gauss point. Then, using Eq. (9.18) or (9.19), tensile or shear yielding criteria are
evaluated. Onset of slip is imminent if F* > 0. In this case, the increment of relative
displacements between the upper and lower joint surfaces after each iteration is divided into
elastic and plastic parts as given below:

dw = dw* + dw? (9.20)

The increment of plastic displacement is estimated considering the flow rule with a plastic
potential function Q and non-negative plastic multiplier d7 as

dQ
dw? =——d
oe 7 (9.21)

In this case, the stress vector consists of o” = {t ,}. For joint, non-associative flow
is assumed such that Q # F’. For simplicity, a constant dilation angle can be assumed for
joint surface and based on that, linear plastic potential function is formulated as given in
Eq. (9.22). However, for joint, the dilation angle changes with normal stress [Eq. (8.26)]. In
numerical modelling, the variability of dilation angle can also be included based on normal
stress and will be discussed later.

Q= |is|+0, tany; = const. (9.22)

For tensile yield criterion, associative flow rule is assumed with Q = F". The procedure
for stress correction is the same as shear yield criterion and is discussed below.
Step 3. A consistency equation is expressed in Eq. (9.23) which states that the stress point
must lie on the yield surface.

ar =|
ar |" tisOFSs =U (9.23)
5

where w® represents the plastic component of relative shear displacement. It is a reasonable

assumption that only shear component of plastic strain tensor will be responsible for the
 Finite Element Method: Concepts and Applications:in Geomechanics

determination of residual shear strength which is an important parameter in rock mechanics.

The cohesion and friction angle are then dependent of the cumulative parameter w?. Following
the concept outlined in Figure 7.5, we can update the cohesion and friction angle of joint
surfaces based on cumulative plastic shear displacement, if the relationships between the
cumulative parameter w/’, the cohesion and the friction angle are established.
As given in Chapter 7, the increment in stresses is contributed due to elastic component
of the displacement increments and thus we can write

do =k(dw — dw?) =K{aw- 2an) (9.24)

Replacing Eq. (9.24) into Eq. (9.23) and then rearranging, we find the plastic multiplier
as
T
se kdw
00
dn = Peo a

set Kjset- OF dQ (9.25)

00 00 dw? OT

Once the plastic multiplier is estimated, we can evaluate the increment of stress (Ao)
using Eq. (9.24) and then stress vector after ith iteration at load step t can be obtained as

6; =0)_, + Ao} (9.26)

The increment of residual force vector is estimated from the increment of plastic
displacement (Aw?) of each yielded element and applied to next iterative cycle within a load
step. The force vector in x’y’ coordinate system is obtained using Eq. (9.27) for each yielded
joint element.

AR’ = [BT kaw?av

(9.27)
V

The force vector AR’ is transformed into global coordinate system (x and y
axes) using
Eq. (9.28) and then applied for the next iteration.

AR = TAR’ (9.28)
Step 4. Normal and shear stiffness of joint at a Gauss point can also
be updated based
on the slip and separation of the joint after any iteration. The general
conditions for updating
stiffness are:
kin = 9, ky = 0 ey ESS) (9.29a)
ky = 0 if F° > 0 (9.29b)
This updated stiffness values are applied to estimate element
stiffness matrix for the next
iteration. However, numerical instability may arise if too
small stiffness values are incorporated
into the model.
| Chapter 9 Finite Element Procedures of Rock Joints 231

9.3.4 Comments on Non-Linear Analysis of Joint

If 6, is negative or compressive in nature, then plastic corrections will be applied if shear stress
T exceeds the peak shear strength failure envelop. In general, peak shear strength envelop must
be applied for contacting joints. If a joint is open, compressive normal stress cannot be
transmitted across that. Pande et al. (1990) found that due to incorporation of tensile cut-off
and in the presence of normal tensile stress, joints are relaxed causing accumulation of plastic
Strains signifying joint opening. However, once the stress regime changes in subsequent
iterations, the normal stress can become compressive and thus a compressive stress is
transmitted across an open joint. This condition may not satisfy a realistic situation. Under such
conditions, a simple criterion based on cumulative normal displacements can be adopted to
check the joint opening and closure.
As mentioned before, dilation angle is dependent on the normal stress. Using
Eq. (8.26), dilation angle can be adjusted after each iteration for every Gauss points.
Alternatively, dilation angle can also be adjusted based on JCS and JRC values as given in
Bae (9.30):

1 JCS
Wj ag BGs oe :) (9.30)
me

Dilation will inhibit if normal stress is higher than JCS, or relative shear displacement
is higher than the residual shear displacement. The residual shear displacement can
approximately be estimated based on Eq. (9.31) (Barton et al., 1985)

L, [JRC, 0.33
u, =10
500] L, (9.31)

where L, is the joint length with unit in m. The joint length can be an input to the finite
element procedure or it can be considered as the length of the element.

9.3.5 Implementation of Mohr-Coulomb and Barton’s Yield Criteria

For non-linear analysis of joints, Mohr-Coulomb shear yield criterion is widely applied in finite
element analysis. For each joint element, shear/tensile yield criterion is verified at each Gauss
point with current shear and normal stress levels. If at any Gauss point, yield criterion is
satisfied, the plastic corrections are applied as mentioned in the previous section. Table 9.1
depicts the derivatives of yield functions with respect to positive shear stress and normal stress.
Table 9.2 lists the derivatives of plastic potential function with variable dilation angle
[Eq. (8.26)] with respect to positive shear stress and normal stress. Readers can also derive the
derivatives if dilation angle of Eq. (9.30) is assumed.
[232 | Finite Element Method: Concepts and Applications in Geomechanics |

Table 9.1 Estimation of derivatives of yield functions

Yield criterion

Mohr-Coulomb
Barton
Tensile

where

JCS IRC 2 JCS,

M =tan| JRC, n'0810
lo TS ny és) orc |
——| tan“ | JRC, p lo Cela! + Pp +1

Table 9.2 Estimation of derivatives of plastic potential function

Yield criterion

Shear
Tensile

where

ch vane o
N ~tend [1-22 [i-a+H Ze)
OT OT

Once the derivatives are known, plastic multiplier can be obtained using Eq. (9.25) and
then using Eq. (9.24), increment of shear and normal stresses after iteration i at load step ¢ are
estimated as

dt=k, a -2an)
OT

=~ n dQ
do, = kmlw 72.an) (9.32)
n

Hence the updated shear and normal stress after iteration i at load step t will be

t t t
T; = GS te (dT); (9.33)

(O,, di = (0,))-1 + (do, )j

The increment of residual force vector is summed for the entire element based on the
number of Gauss points, m, and given below:
| Chapter 9 Finite Element Procedures of Rock Joints [ 233 |

; m A)

AR = SBT on{ Pan see

(9.34)
j=l

Example 9.3 Stress vector, o at a Gauss point of a joint element at iteration i of load
step ft is found to be oj ={t o,}={-3.46 —2.00} MPa. At iteration i + 1, displacement
increment vector is {-1.08 -0.25} x 10-7 mm. Assuming Mohr-Coulomb yield criterion with
cj = 3 MPa, 9; = 1S)° ail Y= 5°, determine the stress vector after iteration i + 1. Consider
k, = 20 GPa/m and k,,, = 50 GPa/m.

Solution: Applying Mohr-Coulomb yield criterion at iteration i of load step t, we find

that the stress condition does not satisfy the criterion and hence F* < 0. For i + 1 iteration,
the trial stress vector becomes

T —3.46 20000 0 —1.08 WEMROI- 3,08

= + <I = MPa
GC, —2.0 0 50000 | |—0.25 —2.13

With the trial stress vector, Eq. (9.19) is satisfied for the given cohesion and friction
angle and hence F* > 0. Based on Table 9.1, derivatives are estimated as

di

A SRL {-1 0.268}

cymee lon

Here, plastic potential function is given in the form of Eq. (9.22) and thus respective
derivatives are also estimated. From Eq. (9.25), plastic multiplier is calculated as 8.644 x 107
©. Thus increment of stresses is

AT 20000 0 —1.08 + 0.8644 _5 |—0.0436

= x10~ = MPa
AO, 0 50000 | |-0.25 — 0.8644 x 0.087 —0.1628

Hence the new stress vector after plastic correction is

fal, —3.3.46 " —0.04

’ B —3.51 in
0, —2.0 —0.16 —2.16

9.4 The Equivalent Material Approach for Joints

As mentioned in Chapter 8, when joints in rock mass are assumed to be ubiquitous having
multiple joint sets and directions, it is almost impossible to consider each joint separately.
Hence they are modelled with equivalent material approach. Figure 9.5 depicts the jointed rock
mass for which this approach is appropriate. It is impossible to consider each and every joint
 Finite Element Method: Concepts and Applications in Geomechanics

EEL
KARE Excavation

LOS
Joint set #2

Figure 9.5 Rock mass for which the equivalent material approach is preferred.

as discrete joint element as mentioned above since numerical stability of the finite element
model will be seriously compromised. At times, it will be even difficult to build such a
numerical model with proper connectivity of nodes with elements. In this section, we will
discuss two approaches for determining equivalent stress-strain relations based on stiffness, joint
inclination angle and average spacing of joints in a joint set.

9.4.1 Equivalent Material Model: Method |

In this method, mainly rectangular or square shaped elements are better suited for developing
finite element models. Since joints are assumed to be everywhere, we will consider that a single
or multiple joints can intercept an element. Here, we first restrict our discussion on a single
joint intercepted an element completely as shown in Figure 9.6(a). The figure depicts a
rectangular element with stress tensors. The intercepted joint has the direction cosines
c = cos@ and s = sin@. It can be noted that for a joint set, individual joints can be parallel or
oblique to each other based on elements size and shape.

y my£3

| Oyy
: d |

(a) A rectangular element with joint (b) Equivalent material element

Figure 9.6
| Chapter 9 Finite Element Procedures of Rock Joints [ 235 |

The principle of deformation equivalence assumes that displacement in x and y directions

is similar in the equivalent continuous body [Figure 9.6(b)] and the jointed rock element under
same loading conditions (Zhu et al., 1993). Based on this principle, a new stress-strain relation
can be determined. In this case, the equivalent body or rock mass can be assumed as isotropic,
orthogonally anisotropic, or transversely isotropic depending on circumstances and complexity
of the problem. Let us revisit Figure 9.6(a) where a rectangular rock element with diagonal
joint crossing is shown. Joint can also pass through one side of the element to the opposite side
and not necessarily through the corner points. However, we will confine the derivation to the
joint passing from one corner to the opposite one. The strike of the joint is directed along the
Z axis or perpendicular to the plane. Three in-plane stress tensors O,x, Oy and T,, are acting on
the element as shown in the figure. The in-plane strain tensors are denoted Dynes Ey and Jy.
The out-of-plane normal stress is denoted by o,,. Let us assume that the local coordinate axes
of the joint are x’ and y’ as shown in Figure 9.6(a). If the rock material is assumed to be
transversely isotropic with special consideration of orthotropy, the plane of isotropy is assumed
parallel to the x’ axis. In this case, considering E,, v, are the material constants in shear direction
and E>, V2 in normal direction, the local compliance matrix of intact rock can be given as

sumone Os peal
Ey E, E,

oy Ss wes
Ch = es ee 4 (9.35)
1
0 o> =. *0
G
vb Das eag See
E Ey Ey
Ey me ;
where G, mae e The local compliance matrix is then transformed to the global
sav
coordinates using the following relationship:

Cp =T'CeT (9.36)
where transformation matrix T is given in Eq. (8.46). However, considering isotropic rock
material having elastic modulus, E, and Poisson’s ration v,, the global compliance matrix will
be

1 =a 0 ih
Cue Ve 1 0 =v,
1a a 0 2(1+V,) 0 (GSh)
=V, =Vi 0 1

The following discussion will be based on the isotropic rock material. For transversely
isotropic rock, the compliance matrix will be replaced by Eq. (9.36).
 236 Pe Finite Element Method: Concepts and Applications in Geomechanics

The shear and normal stresses on the joint plane can be obtained by coordinate
transformation and given as

0, = cao +C°Oyy = 2CSTxy (9.38a)

T= CSO CSO (c? - Goalies (9.38b)

where c = cos@ and s = sin@. In this case, we assume that the joint has no thickness and thus
the volume occupied by the joint is negligible as compared to intact rock.

Deformation Equivalence with No Dilation

If no joint dilation is allowed then deformation superposition in x and y directions for intact
rock and the joint lead to:
1 oO T
on = Oy V-Oyy — V-Oz, eae i BC (9.39a)
E, nn tt

1 oO f.
Syy ms==|—-V,-0 xy a3 Oyy — V,027] bate Crt aes (9.39b)
E, nn “tt

Replacing Eq. (9.37) into Eq. (9.39), we find that

l gees V, c’s c*s V, Jes” (6 SNe

Oy =||—+ +—— |o,,+|--+——
- —— |o,, +|-—-|o, + saat ear Sea
E TKyy dk, E, Lia dk, i E, knn dky

V, cs’ cs? l a Thee V, ~2¢*s (c? age )s

0, = Se taeeee ee | SeesSS Oyy + Fa OS a Ty I
a E, linn lky, E, [Kn lky, : E, in lky, j

(9.40)
We now consider an element of equal dimension with equivalent isotropic material having
similar displacements 6,, and 6,, under the same stress condition as shown in Figure 9.6(b).
Hence we can write the compliance matrix of the equivalent material as

Exx Ch ap) C13 mY; /E, O xx

Piya oe |e On ap Vp ee tee
Voy C3} C3. C33 0 Tay (9.41)
° V- {Ep Vp /EO” REE Ore
where

— 1 s Cea ea Gas. es e, ~2cs? a s*)c

Ee Gk eles Eo diy ead dk yn dk
| Chapter 9 Finite Element Procedures of Rock Joints 237

A PREIS cs’ e ‘| Cec of 2c*s (c? - s*)s

Nes = Dee 23
E, i oa lky E, Lae lk Iknn ky

Goes es)
33 E

Note that compliance matrix isA symmetricG with

C i. <
cS, =C,,, C3, =C,; and CzyCae =ee C>,.
a, NSS e (ens ee é
In
e

this case, it is assumed that C35

2(1+v,) , Same as compliance component of intact rock. Note
:
that the equivalent compliance is dependent on the joint stiffness and inclination with respect
to global coordinate axis. If more than one joints are intercepted the element, equivalent
compliance matrix can be formed repeatedly considering parameters of one joint at a time. The
equivalent constitutive matrix is then evaluated as the inverse of the equivalent compliance
matrix and is used for determining the stiffness matrix of the element.

Deformation Equivalent Compliance with Dilation

As mentioned in Chapter 8, dilation resulted in normal deformation due to the effect of shear
deformation of a joint. Considering constant dilation angle, y;, we can then estimate the normal
deformation as (Figure 9.7) given in Eq. (9.41):
T
Oi realy (9.42)
tt

Figure 9.7. Equivalent compliance with dilation.

Hence, Eq. (9.39) can be rewritten as

Tom ey (9.43a)
r nn tt Ky

oO T cTtany;
6 = —[-V,0 5 ee -V,6,, | + Mo —~—5—~—__+ (9.43b)
E ky
 Finite Element Method: Concepts and Applications in Geomechanics

By manipulating the last two terms of Eqs. (9.43a) and (9.43b), we can rearrange the
equation as
1 oO G
Ox ==Ee [Ox —V,Oy- -~V,Fz]d+ +—-c
5—* (9.44a)
“mn tt
1 oO. 16
OF, {VO Oy, —V,O7- UE
& end Sie Aire (9.44b)
; Ran it
where
ck
c—stany

PEERS Skyy
ff Sat
e CL ainty,

Equations (9.39) and (9.44) are of similar forms except that in the latter shear stiffness
is changed. Based on the Eq. (9.44), the component of equivalent compliance can be established
similar to Eq. (9.41). Note that compliance matrix is symmetrical and so does the equivalent
stiffness matrix.

9.4.2 Equivalent Material Model: Method II

In this method, equivalent stress-strain law is established based on the concept outlined in
Section 8.5. The major assumption of this method is that in a joint set, individual joints are
parallel to each other but spacing between two consecutive joints may not be equal as shown
in Figure 9.8 (Pande et al., 1988). We assume that the average spacing between two consecutive
joints is d; for ith joint set. As before we can assume that the local coordinate axes of the joint
are x’ and y’ and intact rock material can be assumed as isotropic or anisotropic depending on
the problem in hand. Then global constitutive matrix for transversely isotropic rock is evaluated
from Eq. (9.36) and that for isotropic material is from Eq. (9.38). For ith joint set, the joint
compliance matrix is estimated using shear and normal stiffness, average joint spacing and
inclination of joint with global axis as elaborated in sub-section 8.5.4 and given in Eq. (8.45). The
cumulative compliance matrix of n number of joint sets is then combined with compliance
of
intact rock [Eq. (8.48)] and equivalent constitutive matrix is obtained using Eq. (8.50).

ith joint set

Figure 9.8 Orientation of joint in the rock mass.

 | Chapter 9 Finite Element Procedures of Rock Joints 239
It can be shown that for isotropic rock materi
al having a joint plane inclined @ degree
with positive x axis, components of equivalent
compliance matrix are:

aes : + + Seago
crs" C5Se : + ct + c's (9.45a)
E, d;k nn
49a
d IKw Er djKat d;ky

Dee Ded
imari
ai orCp eum
3
de at
Vv 2 2
0.458)
— =

Ti d; inn d; ky d;on d; ky

anne 28DCs 3 ters

2 in 2,
es Deed
4 es 2, a 2\2
i : (ej (9.45c)
> C33 =
d;ky d;Kan d;ky

9.4.3 Comparison between the Two Methods

The two methods described above yield identical equivalent
compliance matrix if the average
joint spacing d; is related with the sides of the rectangular element
, d and J, as shown below
(Figure 9.9):

Joints
Element

Figure 9.9 Comparison between two methods.

From Figure 9.9, we find d; = d sin @ = 1 cos@. Replacing d and / with d; in Eg. (9.41)
we find Eq. (9.45). For any other element configuration, equivalent compliance matrix
obtained by method I will be different from that estimated by method II. However, since
method II estimates all components of equivalent compliance matrix, this method is widely
applied in geotechnical field as compared to method I.

9.5 Tunnel Example

The tunnel problem shown in Chapters 3, 4 and 7 is considered again in this section. However,
the problem will now be solved considering rock mass around the tunnel boundary instead of
intact rock as shown in earlier chapters. For simplicity, we assume that the rock mass contains
two joint sets with the following properties:
 in Geomechanics |
| 240 | be Finite Element Method: Concepts and Applications

Table 9.3 Joint properties in the rock mass

how Ky Spacing Inclination

Joint
(GPa/m) (GPa/m) (m) (deg)
set no.
10.0 3.0 2.0 45
10.0 3.0 1.0 120
2

finite element mesh,

Figure 9.10 shows the schematic of rock mass with 2 joint sets. For
similar boundary and
quarter portion of the tunnel boundary is considered as before with
is develop ed with 230 quadratic quarilateral
loading conditions. Thus, the finite element mesh
elements having 757 nodes. As before, near the boundary of the tunnel, small size elements
size elements are
are developed for more accurate stress and displacement distributions. Large
analysis is
developed away from the boundary of the tunnel. In this case, linear elastic
rock model
performed. The results of finite element analysis are compared with those of intact
given in Chapter 4.

Joint set #1

Joint set #2

Excavation

Figure 9.10 Rock mass used for tunnel problem.

Using the finite element formulation described in method II, the author has developed the
finite element program to analyse rock mass using equivalent material. This program is used
to solve the above problem. Figure 9.11 compares the radial displacements around the tunnel

|
—— With joints —— Without joints|
0.0035
|
|
—u(m)
|
1 ES Z 2:5 3 BS 4
rla

Figure 9.11 Distribution of radial displacements.

 | Chapter 9 Finite Element Procedures
of Rock Joints | if241 |
made in intact rock and rock mass. It can
be seen that radial displacement at the tunnel
boundary made in rock mass has increased by
36% compared to similar tunnel made in intact
rock. Higher radial displacement is found until
r/a ratio is 2.0. As expected, tangential stress
is lower at the tunnel boundary as shown in
Figure 9.12. It is estimated that tangential stress
is about 18% lower for the rock mass assumed
in this analysis. The peak tangential stress of
11.59 MPa has shifted inside the rock mass. Overall, tangential stress is lower for the rock
mass model.

1.75
_ Nn

0,/Po
O,/Ro
or

I— Without joints With joints

SoNS
n

1 1.5 2 Ie) 3 35) 4

rla
Figure 9.12 Distribution of tangential and radial stresses.

9.6 Non-Linear Analysis with Equivalent Material

Non-linear or elastic plastic finite element procedure of equivalent material
is almost similar
to that of intact rock (Chapter 7). However, there are few exceptions which
are explained
below:
1. In this case, stresses evaluated at Gauss points can be applied for plastic corrections
for both yielded rock and joints. The general procedure is to first apply plastic
corrections for intact rock based on the procedure outlined in Chapter 7. The choice
of yield and plastic potential functions is lie on the user. After the correction is made
for intact rock the new updated stress tensors (o%)" = oe oe ae G) is
obtained based on Eq. (7.46), and is applied for the evaluation of joint slip or opening.
2. The updated stress tensor is transformed in the local x’y’ axes, where x’ represents the
shear direction as shown in Figure 9.8. Equation (9.46) is applied for the stress
transformation, where c = cos @ and s = sin @ and @ being the angle of inclination
of joint set with the x axis measured counterclockwise.

O7r'x’ SE es. 9 a :
Co Ey amir 268° 10 :
ty Ses cs Ce 5" 0 i, (9.46)
OF, Oe 0 0 1} |GN
| 242 | Finite Element Method: Concepts and Applications in Geomechanics

along the joint plane as

From Eg. (9.38), we can find shear stress and normal stress
ar: = Ty!

O, =Oyy (9.47)

to check the slip

Mohr-Coulomb or Barton’s shear yielding criteria are then applied
obtained above. In other
along joint plane based on the normal and shear stresses
applicable for this
words, general shear yield criterion expressed in Eq. (9.19) is also
case.
med into
3. Nodal displacement increment vector in global coordinate system is transfor
using the followin g interpol ation
the Gauss point displacement increment vector
functions:
n n
Aug => N)Au Avg = SIN AY; (9.48)
i=] i=l

where N; is the shape function at ith node and n denotes the number of nodes. Then
the displacement increments are dissolved in the directions of x’y’ axes using
transformation matrix expressed in Eq. (1.17). Hence displacement increments
dw’ = Au’, and dw” = Av; at a Gauss point are obtained.
4. Plastic multiplier d7 is calculated using Eq. (9.25). Shear and normal stiffness of the
joint set are specified by the user. After that, increments of shear and normal stresses
are simply calculated using Eq. (9.24). Hence, stress increments in x’y’ axes are
evaluated using the following equations:
doy, =0 (9.49a)
aT.) =dt= ke [aw -
s 99
2 an) (9.49b)

n 92
doy = Kan Gs aE ag in] (9.49c)

do,, =0 (9.494)
5. The increments of shear and normal stresses are then transformed back into the global
xy coordinate system using the following transformation procedure:

dO xx ciaustos PQenanOnidayg
dO yy + a te 2cs 0 ||4 yy’
aT xy és -es ci—s% 0|\2te% (9.50)

do, Oc fe 0 1|( 4022"

 Chapter 9 Finite Element Procedures of Rock Joints
| 243 |
The stress increment tensors thus obtained are added to the
stress tensor vector 0’. If
more than one joint set are present in the model, the procedures
outlined from steps | to 5 are
applied for the other jointsone after another. In each case, the
last updated stress vector is used
for verifying the condition of slip along a joint plane.

SUMMARY
Joints in rock mass are modelled in two ways—discrete joint element and
equivalent material
model. This chapter presented details about both the methods with suitable examples
. The finite
element formulation of rock joint or interface element was proposed by Goodman
and is widely
used in solving geomechanics problems. The step-by-step procedure is outlined for
non-linear
or elastic-plastic analysis of joint element. The concept of equivalent material
model was
presented in Chapter 8 and also elaborated in this chapter for specific implementation
in finite
elements. Two methods are discussed to formulate finite element stiffness matrix using
equivalent material model. It is author’s view that Method II will be the robust and more
accurate technique for implementation in finite element procedure. In this case also step-by-
step procedure is given for implementation of elastic-plastic formulation of equivalent material
in finite element procedure. Suitable examples are also solved to explain the procedures
outlined in this chapter.

EXERCISES
9.1 Show that Eq. (9.17) produces the similar matrix as given in Eq. (9.11).
9.2 Using the data given in Example 9.1, determine the normal and shear stresses at
Gauss points of the joint element. Compare the results obtained in Example 9.1.
9.3 Derive an expression of plastic multiplier dn [Eq. (9.25)] using Mohr-Coulomb
yield criterion with non-associative flow rule.

9.4 Displacement increment vector at the end of iteration i of a joint element inclined
at 45° counterclockwise with the horizontal is found to be q’ = [0 -0.3906 0
-0.7813 0 -2.5490 0 -2.1584] x 10% m.
(a) If k,, = kj, = 40 GPa/m, determine whether the joint has yielded considering
c; = 2 MPa and @; = 20°. Assume [Tt o,,];,; = [-3.1 -3.1] MPa.
(b) If joint has yielded, estimate the updated stresses at Gauss points after plastic
correction. Assume associative flow condition.
(c) If joint has yielded, estimate the updated stresses at Gauss points after plastic
correction. Assume non-associative flow condition as given in Table 9.2.
Consider & = 30°, k = 4.0 and o” = 30 MPa.

9.5 Recalculate Problem 9.4 based on Barton’s shear yield function. Assume
JRC = 12.0, JCS = 56°MPa and ¢, = 30° for joint length of 10 cm. For part (c),
apply the dilation angle given in Eq. (9.30).
 Finite Element Method: Concepts and Applications in Geomechanics

9.6 Figure P9.1 shows a rectangular element bisected by a joint. Assuming k,,, = ky, =
40 GPa/m, joint spacing of 0.5 m, E, = 5 GPa and n = 0.25.
(a) Determine the nodal displacement vector considering the equivalent material
model approach. (Any spreadsheet software like MS Excel can be used.)
(b) Determine the stresses at Gauss point | with r = 5 = qn.
(c) Compare the results obtained in parts (a) and (b) with no-joint element.
(d) Determine whether the joint fails at Gauss point 1 for no-cohesion joint moving
¢; = 30°.
20 MPa
3 20MPa sg

0.25m

Figure P9.1

9.7 If the part (d) of Problem 9.6 is satisfied, apply plastic corrections to adjust the
stresses at the Gauss point.
 APPENDIX

A.1 Introduction
In this text, problems in structural mechanics are addressed by variational principles or
Rayleigh-Ritz method. The principle of stationary energy is one such variational principle and
is explained in detail in this book. However, in other areas of physical science such as fluid
mechanics and heat transfer problems, finite element formulations may not be obtainable with
variational principles. In such cases, finite element equations are formulated with the help of
weighted residual method. Like the Rayleigh-Ritz method, a weighted residual method uses
integral expressions that contain the differential equations of a physical problem (Cook et al.,
1989). Both Rayleigh-Ritz method and weighted residual method enforce conditions of
differential equations in average or integral sense and hence they are called weak form. The
differential equations are strong form since they enforces the conditions at every point.

A.2 Concept of Weighted Residual Method

Let us consider Figure A.1 in which a straight bar of length L is fixed at one end B and a
concentrated load P is applied at the other end C. The parameter f,” denotes the body force
per unit volume in positive x axis direction. The goveruing differential equation of this physical
system can be written from the figure as

(A.1)

where A is the cross-sectional area of the bar. Considering u as the displacement field in
x direction, we find that

O;, = E— (A.2)
 Appendix A Galerkin Finite Element Method

where E is the modulus of elasticity of the material. Hence Eq. (A.1) can be rewritten as

-AE— - f? ~= (A.3)

Cross-section

Figure A.1 Example of one-dimensional bar.

The above equation can be generalized with a differential operator D and a force function
r as follows:
Diu] -— r=0 (A.4)
Sete oe b
where, in this example, D=—AE rE and r=f,.
x
The differential operator D must satisfy symmetric and positive definite conditions for the
entire domain of the physical system Q as given below:

|Plan = | Piva symmetric condition (A.5)

Q Q

[Diwan >0 positive definite condition (A.6)

Q
Readers can easily verify that both the conditions for the differential operator are given
in this example. As for example, positive definite condition can be verified with boundary

condition u = 0 at x = O and ou _ Q at x = L as given below:

Ox
ib 92

Vege a 2) +[ae(e
"a= [ae(%)dx (A.7)

The right hand side of the above equation is positive for all values of u.
The general solution wu = u(x) of any arbitrary physical system obeying Eq. (A.4) must
also satisfy the essential and natural boundary conditions, i.e.
 | Appendix A_ Galerkin Finite Element Method
| [247 |

B;[u]—1; = 0), boundary T;

(A.8)
where B; and 7; are the differential operator
and function respectively. In the above probl
the essential and natural boundary conditions em,
are
ut) Bt v=) (A.9a)
Wet pag Atgy eel,
ox (A.9b)
The exact solution of u = u(x) is difficult to find
for any arbitrary physical system and
instead we seek for approximate solution u=wu(
x) such that average residual vanishes. The
basic step of weighted residual method is to assum
e the approximate solution of the form
n

u =) a9; (A.10)
i=l
where @; linearly independent trial functions (shape
functions in Galerkin method) and a; are
the multipliers (displacements) to be determined from the
solution. The assumed function must
satisfy the essential or displacement boundary conditions.
The residuals of Equations (A.4) and
(A.8) are then obtained by replacing Eq. (A.10) in
place of u as

Rp = Diii]—r (A.11)
Rp = B;{u] -t,

For the exact solution, residuals are zero at every point. The
major idea of weighted
residual methods is then to find a good approximate solution
for which residuals are small at
all points of the solution domain. Residual methods such as
collocation, subdomain, least
squares, least squares collocation and Galerkin methods differ in
criteria for the calculation of
a; such that residuals are small in average or integral sense. In
general, the solutions a; are
obtained from the equations given below:

|W,(x)Rp(a;,x)dQ =0 (A.12a)
Q

[WcoRp(a;,,xNaT =0 (A.12b)
ir
where W; denotes weight of the generalized coordinates a;. The index i ranges
over all weight
functions.

A.3 Galerkin Finite Element Method

In Bubnov-Galerkin method or simply the Galerkin method, the weight function W; is
expressed as
Oa
Woe u
i ae (A.13)
 Appendix A Galerkin Finite Element Method

on by parts
In this method, the boundary residual Rg is used in combination with integrati
us revisit the bar example problem. Assume
for introducing natural boundary conditions. Let
as shown in Figure A.2. The shape
that the bar is discretized with a 2 noded bar element
functions are

Ny ek
1 z
iG
an ee eK 2
x
L (A.14)

Figure A.2_ Bar element.

The displacement field 7 can be expressed using the nodal point displacements a; as
A,

u= Sy aN; (A.15)
i=

The weight functions W; used in Galerkin method are equal to the shape functions as
given below:
Ou
We Tas=N; (A.16)

Thus, the Galerkin residual equation becomes

fi 2a
Uu b is ,
[y, ae 5 Jaro (A.17)

Note that the index i ranges over all shape functions. Considering the constant AE,
integration by parts of Eq. (A.17) yields

L 2. ae ees § % be
“J setts? ix=-|vae] (ee AEM ax [N,fPdx
:
=0 (A.18)
: ox ox |p | ox ox

The non-essential boundary condition at the end can be obtained from Eq. (A.9b) as
ou
AE =P. We can adopt the notation given in Chapter 3 as
x

On ON, ON r
ASR
a Bg h eB eh, a ey eZ
eo and q i ={a, a} (A.19)
| Appendix A Galerkin Finite Element Method 249

Rearranging Eq. (A.18), we find that

Jas L

: [wes \dx q= [INT#?ax. [N’P]

1,

0 0 =
te ey F (A.20)

Equation (A.20) depicts the standard finite element formula. It can be seen that the last
component in the right hand side of Eq. (A.20) comprises of concentrated load vector. In the
above example, No = [1 0] and N,; = [0 1]. Hence the concentrated force vector will be

rofeebhel vo
If the traction pressure is applied at boundaries, then the nodal traction force vector will
also be added in the right hand side of Eq. (A.20). For two and three-dimensional problems,
similar approach is adopted.
 APPENDIX

_Skyline Storage of
Stiffness

B.1 Introduction
The global stiffness matrix K is generally banded and symmetric. Since the matrix is symmetric
we need not have to store the lower diagonal terms. This reduces the storage requirement by
almost half. Again, since the matrix is banded, the leading zeros need not be stored either.
However, zeros inside the column have to be stored. There are mainly two methods of
computer storage of stiffness matrix—banded storage method and skyline storage method.
Here, we will discuss the skyline storage method which is more efficient and compact than the
banded storage method.

B.2 Concept of Connectivity

In practice, finite element codes should never store stiffness matrix as square array. Instead we
use the connectivity array to assemble the elemental stiffness matrices and force vectors directly
into the global stiffness matrix and global force vector. The connectivity array is part of the
input data for any finite element analysis whether it is provided by the user as a part of a text
data file or created by a graphical pre-processor. Figure B.1 shows one linear quadrilateral and
two triangular elements connected together. Then the connectivity array for each element can
be found as given in Table B.1.

Table B.1 Connectivity array in terms of node

Element Node 1 Node 2 Node 3 Node 4

(1) (J) (K) (L)
1 1 2 > 4
2 Z 3 5
3 3 6 5

250
 Appendix B_ Skyline Storage of Stiffness Matrix
251

Figure B.1__ Finite element mesh.

B.3 Skyline Storage Method

Recall that for two-dimensional problems, the dimens
ion of stiffness matrix is 2n x 2n, n being
the number of nodes. In the above problem, the global
stiffness matrix K is then of dimension
12 x 12 as shown in Figure B.2. In the Skyline method,
stiffness matrix is stored in a column
vector or in a single-dimensional array. Let us denote
it by ali]. First we have to define an
array storing the diagonal positions of stiffness matrix, for
example index position of kj in ai]
vector. Let us denote this array as Jdiag [j] where j refers
to the position of dof in the stiffness
matrix between 1 to 2n. Since there are 2 degrees of freedo
m at each node, the connectivity
in terms of degree of freedom can be expressed as given in
Table B.2.

kiss ks sz sg sq sq 5,19
K= kes ker keg keg kot. kon ko 12
kay yg yg yg ky ay
Sym. Keg gg gig gry gi
kog koiq ko} Ko
Koro Kou ‘1012

Figure B.2 Square stiffness matrix.

[ 252 | Appendix B_ Skyline Storage of Stiffness Matrix

Table B.2 Connectivity array in terms of degree of freedom

ee
ee Se —_S
Node I Node 2 Node 3 Node 4
Element (1) (J) (K)
SE Sas
(L) ES SS
aptionitaiidacht) a i ee) EE

ig ty if Bi Ve K,. 2
1 1 2 3 4 9 10 7 8
2 3 4 5 6 9 10
3 5 6 11 12 9 10

where d= 21 —"1, his = 2] and so on.

The height of each column in the skyline, say column J, is found using the following
formula:
Jdiag| J} = jdiag{j —1]+ j— min_dof +1 (B.1)

The min_dof for jth column is estimated based on the minimum dof found in connectivity
vectors (Table B.2) where dof j is present. As for example, min_dof for column 3 will be 1
since dof 3 is present in the elements 1 and 2 but the minimum dof in connectivity vector of
element | is 1. The initial value of jdiag array is taken as jdiag[0] = 0. The jdiag{j] array for
the above example is given in Table B.3. Hence the dimension of afi] array is 64, where i
ranges from 1 to jdiag[2n]. The computer storage of each component is shown in Table B.4.

Table B.3 jdiag|j] array

j
jdiag{j] I 3 6 Lieesl2 17 24 pestered 51 58 64

Table B.4 One-dimensional array of the stiffness matrix

1 ] 2 3 a4 15 16 17 Be 62 63 64

ali] ky kyo ky vee ky ks6 keg vee kono Kin Ky212

Alonoso, E., L.R. Alejano, F. Varas, G. Fdez-Manin and C. Carranza-Torres (2003), “Ground
Response Curves for Rock Masses Exhibiting Strain-Softening Behaviour”, Int. J. Num.
and Anal. Meth. in Geomech., Vol. 27, pp. 1153-1185.
Bathe, K.J. (1982), Finite Element Procedures in Engineering Analysis, Prentice-Hall, Inc.
Bandis, S., A.C. Lumsden and N. Barton (1981), “Experimental Studies of Scale Effect
on Shear Behaviour of Rock Joints, Int. J. Rock Mech. Min. Sci. and Geomech. Abstr.,
Vol. 18, pp. 1-21.

(1983), “Fundamentals of Rock Joint Deformation”, Int. J. Rock Mech. Min. Sci.
and Geomech. Abstr., Vol. 20, pp. 249-268.
Barton, N. (1976), “The Shear Strength of Rock and Rock Joints”, Int. J. Rock Mech. Min.
Sci. and Geomech. Abstr., Vol. 13, pp. 255-279.
Barton, N. and S. Bandis (1980), “Some effect of Scale on Shear Strength of Joints”, Int. J.
Rock Mech. Min. Sci. and Geomech. Abstr., Vol. 176, pp. 69-73.
Barton, N., S. Bandis and K. Bakhtar (1985), “Strength, Deformation and Conductivity
Coupling of Rock Joints”, Int. J. Rock Mech. Min. Sci. and Geomech. Abstr., Vol. 22,
pp. 121-140.
Barton, N. and V. Choubey (1977), “The Shear Strength of Rock Joints in Theory and
Practice”, Rock Mech., Vol. 10, pp. 1-54.

Beer, G. (1985), “An Isoparametric Joint/Interface Element for Finite Element Analysis”, Int.
J. Num. Meth. in Engg., Vol. 21, pp. 585-600.
Bieniawski, Z.T. (1974), “Geomechanics Classification of Rock Masses and Its Application in
Tunneling,” In: Proceeding of the 3rd International Congress on Rock Mechanics,
Denver, pp. 27-32.
Biron, C. and E. Arioglu 1983, Design of Support in Mines, John Wiley & Sons, p. 248.

253
 References ra
| 254 |

Bodonyi, J. (1970), “Laboratory Tests of Certain Rocks under Axially-Symmetrical Loading

Conditions”, In: Proceedings of the 2nd International Congress on Rock Mechanics,
Belgrade, 1, pp. 389-397.
Boresi, A.P., O.M. Sidebottom, F.B. Seely and J.O. Smith (1978), Advanced Mechanics of
Material, John Wiley & Sons.

Brady, B.H.G. (1987), “Boundary Element and Linked Methods for Underground Excavation”,
E.T. Brown (Eds.), Analytical and Computational Methods in Engineering Rock
Mechanics, Allen & Unwin Publisher.

Brady, B.H.G. and E.T. Brown (1985), Rock Mechanics for Underground Mining, George
Allen & Unwin Publisher, London.

Brown, E.T. (1987), Analytical and Computational Methods in Engineering Rock Mechanics,
Allen & Unwin Publisher.

Carranza-Torres, C. (2004), “Elasto-Plastic Solution Tunnel Problems Using Generalized Form

of The Hoek-Brown Failure Criterion”, SINOROCK2004 2B 28, Beijing, Int. J. Rock
Mech. & Min. Sci., Vol. 41, Supplement 1, pp. 629-639.
Carranza-Torres, C. and C. Fairhurst (1999), “The Elasto-Plastic Response of Underground
Excavations in Rock Masses That Satisfy The Hoek-Brown Failure Criterion”, 1999, Int.
J. Rock Mech. and Min. Sci., Vol. 36, pp. 777-809.

Chakrabarty, J. (1987), Theory of Plasticity, Vol. 1, McGraw-Hill International Editions.

Owen, D.R.J. and E. Hinton (1980), Finite Elements in Plasticity —Theory and Practice,
Pineridge Press.
Chandrupatla, T.R. and A.K. Belegundu (2002) Introduction to Finite Elements in
Engineering, 3rd ed., Prentice-Hall of India.

Cook R.D., D.S. Malkus and M.E. Phesha (1989), Concepts and Applications of Finite Element
Analysis, John Wiley Publication.

Deb, D. and S.O. Choi (2005a), “Singularity Adjustment on Generalized Hoek and Brown
Yield Surface for Elastic-Plastic Analysis”, In Proceedings of the 40th U.S. Rock
Mechanics Symposium, June 25-29, 2005, Anchorage, Alaska.

(2005b), “Comparison Between Direct and Indirect Implementation of Generalized

Hoek and Brown Failure Criterion in Numerical Analysis Procedure”, Tunnel and
Underground Space, Journal of Korean Society of Rock Mechanics, Vol. 15, No. 3,
pp. 142-149.

Deb, D., J. Bhattacharya, A.K. Verma and R.S. Reddy (2004a), “Stability Analysis of Shaft
Pillars Using Three Dimensional Finite Element Modeling—A Case Study”, Ohnishi and
Aoki (Eds.), In: Proceedings of 3rd Asian Rock Mechanics Symposium (ARMS), Kyoto,
Japan, pp. 1089-1094.
ees) as
Deb, D., S.O. Choi and H.S. Shin (2004b), “Analysis of Sinkhole Formation over Abandoned
Mine Using Active-Passive-Active Finite Elements”, Tunnel and Underground Space,
Journal of Korean Society of Rock Mechanics, Vol. 14, No. 6, pp. 411-422.
Edelbro, C. (2004), “Evaluation of Rock Mass Strength Criterion”, Ph.D. Thesis, Lulea
University of Technology, Department of Civil and Environmental Engineering, Sweden,
p. 152.
Fairhurst, C. (1964), On The Validity of The “Brazilian” “Test For Brittle Materials”, Int. J.
Rock Mech. Min. Sci., Vol. 1, pp. 515-546.

Franklin, J.A. (1971), “Triaxial Strength of Rock Material”, Rock Mechanics, Vol. 3 >

pp. 86-98.

Fung, Y.C. (1977), A First Course in Continuum Mechanics, Prentice-Hall Inc.

Gerrard, C.M. (1982a), “Elastic Models of Rock Masses Having One, Two and Three Sets of
Joints”, Int. J. Rock Mech. Min. Sci. & Geomech. Abstr., Vol. 19, pp. 15-23.

(1982b) “Joint Compliances as a Basis for Rock Mass Properties and the Design of
Supports”, Int. J. Rock Mech. Min. Sci. and Geomech. Abstr., Vol. 19, pp. 285-305.
Ghaboussi, J., E.L. Wilson and J. Isenberg (1973), “Finite Elements for Rock Joints and
Interfaces”, J. Soil. Mech. Dn. ASCE.

Goodman, R.E., (1989), Introduction to Rock Mechanics, 2nd ed, John Wiley & Sons.

(1976), Methods of Geological Engineering in Discontinuous Rocks, West

Publishing, St. Paul, MN.

Goodman, R.E., R., Taylor, and T.L. Brekke, (1968), “A Model for the Mechanics of Jointed
Rocks”, Journal of ASCE Geotechnical Division, 94SM(3), 637-59.

Gramberg, J. (1989), “A Non-Conventional View on Rock Mechanics and Fracture Mechanics”,

A.A. Balkema, p. 250.

Hartman, Howard L. (1992), SME Mining Engineering Handbook, Vol. 1, SME Inc.,
Littleton, Colorado, p. 1269.

Herget, G. (1988), Stresses in Rock, A.A. Balkema, Rotterdam, Brookfield, The Neatherland.

Hobbs, D.W. (1964) “The Strength and Stress-Strain Characteristics of Coal in Triaxial
Compression”, J. Geol., Vol. 72(2), pp. 214-23

Hoek, E. (2000), Practical Rock Engineering, http://www.rocscience.com/hoek/Practical Rock

Engineering.asp

Hoek, E., P.K. Kaiser and W.E. Bawden (1998), “Support of Underground Excavations in
Hard Rock”, A.A. Balkema.

Hoek, E. and E.T. Brown (1997), “Practical Estimate of Rock Mass Strength”, Int. J. Rock
Mech. Min. Sci. and Geomesh. Abstr., Vol. 34, No. 8, pp. 1165-1186.
 | 256 References ee

Hoek, E. and E.T. Brown (1988), “The Hoek and Brown Failure Criterion—A 1988 Update”,
15th Canadian Symposium of Rock Mechanics, Toronto, pp. 31-38, Dept., of Civil
Engineering, Univ. of Toronto.

(1980), “Underground Excavation in Rock”, Instn. Mine Metallurgy, London.

Hoek, E., C.C. Torres and Corkum B. (2002), “Hoek and Brown Failure Criterion—2002
Edition”, 5th North American Rock Mechanics Symposium and 17th Tunneling
Association of Canada Conference (NARMS-TAC), pp. 267-271.
Hudson, J.A. and J.P, Hudson, Engineering Rock Mechanics—An Introduction to the
Principles, Pergamon, Elsevier Ltd., UK.
ISRM-Commission on Standardization of Lab and Field Tests (1978), “Suggested Methods for
the Quantitative Description of Discontinuities in Rock Masses” (N. Barton et al.), Int.
J. Rock Mech. Min. Sci. & Geomech. Abstr., Vol. 15, pp. 319-368.
Jaeger, J.C. and N.G.W. Cook (1979), Fundamentals of Rock Mechanics, 3rd ed., pt (593,
Chapman & Hall, London.
Jaeger, J.C., (1960), “Shear Fracture of Anisotropic Rocks”, Geol. Mag., Vol. 97, pp. 65-72.
Jeremic, M.L., (1985), Strata Mechanics in Coal Mining, A.A. Balkema, Rotterdam, Boston,
The Netherlands.

Johnston, I.W. (1985), “Strength of Intact Geomechanical Materials”, Journal of Geotechnical

Engineering, ASCE, 111, p. 730-749.
Kwasniewski, M. (1984), “Anisotropy of Elasticity of Rocks and Its Effect on the Magnitude
and Distribution of Stresses around Underground Openings”, Application of Rock
Mechanics to Planning and Design Prior to Mining, A. Kidybinski and M. Kwasniews
ki
(Eds.), A.A. Balkema, pp. 5-38.
Krishnamorthy, C.S. (1987), Finite Element Analysis—Theory and Programming, Tata-
McGraw Hill Publication.
Kreyszig, E. (2002), Advanced Engineering Mathematics, 8th ed., John Wiley Publication.
Maier, G. and T. Hueckel (1979), “Nonassociative and Coupled Flow
Rules of Elastoplasticity
for Rock-like Materials”, Int. J. Rock Mech. Min. Sci. and Geomech. Abstr., Vol. 16,
pp. 77-92.
Obert, L. and W.I, Duval (1967), Rock Mechanics and the
Design of Structures in Rock, John
Wiley & Sons.
Press. “WH 5.4. Teukolsky, W.T. Vetterling and B.P. Flannery, (2002), Numerical Recipes
in C, Cambridge University Press.
Pande, G.N., G. Beer, and J.R. Williams (1990), Numeri
cal Methods in Rock Mechanics, John
Wiley & Sons.
 | References | | 257 |

Pande, G.N. and W. Xiong (1982), “An Improved Multilaminate Model

of Jointed Rock
Masses”, Numerical Models in Geomechanics, R. Dungar, G.N. Pande and J.A.
Studer
(Eds.), A.A. Balkema, Rotterdam, pp. 218—26.
Ramamurthy, T. (1995), “Bearing Capacity of Rock Foundations”, Yoshinaka and Kikuchi
(Eds.), Rock Foundation, Balkema, Rotterdam.
Rock Science Inc. (1999), “Phase 2: Finite Element Analysis and Support Design”, Toronto.
Rolshoven, S. and M. Jirasek (2002), “Nonlocal Formulations of Softening Plasticity”, Fifth
World Congress on Computational Mechanics, Vienna, Austria. H.A. Mang, F.G.
Rammerstorfer and J. Eberhardsteiner (Eds.).

Sheorey, P.R., A.K. Biswas and V.D. Choubey (1989), “An Empirical Failure Criterion for
Rocks and Jointed Rock Masses”, Engineering Geology, Vol. 26, pp. 141-159.
Sterpi, D. (1999), “An Analysis of Geotechnical Problems Involving Strain Softening Effect’,
Int. J. Numer. Anal. Meth. in Geomech., Vol. 23, pp. 1423-1454.
Sun, Z., C.M. Gerrard and O. Stephansson (1985), “Rock Joint Compliance Tests for
Compression and Shear Loads”, Int. J. Rock Mech. Min. Sci. and Geomech. Abstr..
Vol. 22, pp. 197-213.

Saeb, S. and B. Amadei (1992), “Modelling Rock Joints Under Shear and Normal Loading”,
Int. J. Rock Mech. Min. Sci. and Geomech. Abstr., Vol. 29, pp. 267-278.
Sharma, K.G. and G.N. Pande (1988), “Stability of Rock Masses Reinforced by Passive,
Fully Grouted Rock Bolts”, Int. J. Rock Mech. Min. Sci. and Geomech. Abstr., Vol. 25,
pp. 273-285.
Yoshida, N., N.R. Morgenstern, and D.H. Chan (1990), “Failure Criterion for Stiff Soils and
Rocks Exhibiting Softening”, Canadian Geotechnical Journal, 27(2), pp. 195-202.
Yoshinaka, R. and T. Yamabe (1986), “Joint Stiffness and the Deformation Behaviour
of Discontinuous Rock”, Int. J. Rock Mech. Min. Sci. and Geomech. Abstr., Vol. 23,
pp. 19-28.
Yu, M.H., L.N. He (1991), “A New Model and Theory on Yield and Failure of Material Under
Complex Stress State”, In: Mechanics Behaviour of Materials, Vol. 1, M. Jono and
T. Inoue (Eds.), ICM-6. Oxford: Pergamon, pp. 841-846.
Yu, M.H., Y.W. Zan, J. Zao and M. Yoshimine (2002), “A Unified Strength Criterion for
Rock Material”, Int. J. Rock Mech. Min. Sci. and Geomech. Abstr., Vol. 39, pp. 975-
989.
Yudhbir, W.L. and F. Pronzl (1983), “An Empirical Failure Criterion for Rock Masses” In:
Proceedings of the Sth International Congress on Rock Mechanics, Melbourne, Balkema,
Rotterdam, Vol. 1, B1I-B8.

Zeinkiewicz, O.C. (1979), The Finite Element Method, 3rd ed., Tata-McGraw Hill.

Zhu, W. and P. Wang (1993), “Finite Element Analysis of Jointed Rock Masses and
Engineering Application”, Int. J. Rock Mech. Min. Sci. and Geomech. Abstr., Vol. 30,
pp. 537-544.
 _w S84
-

Roel, ‘EB. Tarrks wo Coteek Ly abe

tind
ae Wa
i aaa cs ot
— soy ohgreeg’ CE uns 2 i
Hoban goiter t6anctaluene’? lela
Ae COON Df) dsent
yarsyy eT sired Sseoi9" PESOS ANA ShcA iY WO Mek ara
Nacriies, Peryamcn PimexiesTad, Cita avon ee 2 erent
Let oh witty inet robb edie sae noisy |
| , ee, ELIT Dh Soh RN TG za
ik eget
Rock, Mech Sch. & Gewmech. Abate. RAD.j
. it turtle nine griviovel enrsidord tested
inpgur. 1%, and Neth
rlzeit Uy ee nes2 diinditedihe mae )
Chap mb & Mall, Lot
wrt wes soneilqino ar aeA200” (28@1) neeenartyae Obes beret? 4. a?
POT Ce ar ie ee tet eee ate ee | 7%,
Jeremie, ML, GUSSS Serara Me chonics i Coat Afinins . AAS soba tank? hauth stfsh
oY
gotdellueytiic rsd? seb) anol tool aniflsbowt"* (S007) ishumA - bau
24
Sohovon, WEVFSRS, eee. o¥ pated esshamos hiatal Aaa. Aaa | i
Viedb ied: bench est sHekNP AIGR-Td Qilidare” (88Cl) sbasT VLD baa Dt am ida
ioe LOY, ayes Apgar pup, Jha,ls JooM. Ar ry, Anh ad dog ee s
tnd “Deerriiation of Sirtsses around. Us cigroomd Openmge. Ap (iu é :
bas Hod AeWA ons nila (OCR nally ae Bas _nbgtaneigrchdld RAK: sakes fe é
LOB-28P.aq AEKD vormmo’ lotindososy) nutbeas).."gainefie? gniidiixd aed 5
Nnetvmte® oudnial Do bre wotTeteaniky 4OeOU adem -P deadiningeriegy
ES 349 Gaede ietostken nt, 4 Ans doe 1 an. an suouniiaoseid To :
Kicysals I RTS) Avene! Leoherrgig Studies, Ae ed. Jone 5 WileyPitts.
¥ebalLF tesa |
tn slic’ bas bisi¥ a. wh stl tm fisboM wslt Le (f0ON) sH AT...
‘bars ore Wil lov)a sintiowt oats MS. ea, it ‘Sao Be Ae
J eit ita sy
7
. - - 14
OhE—T2 Aq nothagisT ‘bat “O-MOP<(2
t eh
paly Shoal §
19} norrsici J trons? henintl A: ASC snimtidagyY _M bas: oa A ak .W, y M }
aie & ee v ath wOsimisd Mth S52 snle “ae pest X am” > bald aS a
Sat ;
it “esazeM aoe Wt norman sult "eHie hae Paeiy \schayos Naeecow Rearen
ot itt Aight srramadlsté fell so th, US ix = meg 7 Ih MOO rine Ae? ih \o Nae 4
h. Be Sh aleray | ttn, Ne bioree a AER, Koha nse
itt 2 TO0M-st .Bs hf “wonstA Wiesel oil ant ACTRTy 30
m 29228M dood bomsél Fo stegiaas ghee: sunt” (€0CL) gmeW Fb
0 .AledA Asano fey S52 RE 2Hie as A ae Rms: sa meat
Chapter 1

1.2 -—3.4 MPa, -1.8 MPa

1.3 (a) 0; =-9.471 MPa, 0) = -5.198 MPa, 0; = —1.33 MPa

1, = 0.2866 (73.34°), m, = —0.4484 (63.35°), n; = —0.8466 (147.84°)
lL, = 0.0135, m) = —-0.8885, n. = -0.4644
1, = 0.9579, m3 = -0.1216, n; = 0.25989
(b) -5.33 MPa, 1.97 MPa
1.4 -1.4 MPa, -0.163 MPa, —1.226 MPa
1.5 o, = -9.21 MPa, o> = -3.96 MPa, 03 = 4.17 MPa
1, = 0.91204 (24.21°), m, = 0.02653 (88°), n, = —0.40922 (114.156°)
1, = 0.35719, m, = 0.4388, n, = 0.82453
1; = 0.20145, m, = -0.89819, n3; = 0.39073
Tmax = 6.695 MPa
1.7 0.8228
1.8 0, = 2 MPa, 0) = 0 MPa, o; = -4 MPa
1.9 €&,, = -0.00267, €, = -0.002, ¥, = 0
Evy = -0.0021675, eyy = -0.0025025, gy = 0.00029011
1.10 <’ = [-0.0025 - 0.009 0.0055]
1.11 o,, = 6lagy
Ty = 3ayy’

259
 260 Answers to Exercises

Chapter 2

1,193 0877 <-102-4m

22O, == 10900Ge00,, = 14227, +, = 0, = 0 MPa
Z.3 Ou = 24, Oyy = eon ete nt TEs
2.4 kel, m=2. faa ee

25 k= 3.114, m 25 eee — 2 olla i?

2.6 0, = -6.928 MPa, o; = -3.07 MPa; | = -0.1816, m = 0.9832
2.7 & = 27.46 x 10 Peso 10 55a l0cy we = 16 Ole 1

Chapter 3

1 1499
3.1 (a) = —hA, (b) ] = —-—— hA
: 60 (0) 1260

Bxe 2x 2 2
ee ay [eee i [boo ee
3.4 e” = {0.0034 0.0007 -0.0059}
3.5 x(0.2, 0.3) = 2.9, y(0.2, 0.3) = 2.8
3.60 Rita 95.93.375- 93.256 15ers onl Pose cat
3.7 F3, = Fy = -3.6667, Fe, = Foy = -6.0000, Fo, = Foy = -2.3333
3.8 Fy, = Fay = -1.333, Foy = Foy = -2.000, Fo, = Foy = -0.667, Fi24= Fy = -4.667,
Fis; = Fisy = 3.333
SLO Si 2.92 ys ao

3.12 395

Chapter 4

Al (a) Fan
43 758 3(b)) (3:25:98
7)

4.2 (a) At point P: oe tS , At point Q: aes 0.3375 men

Vly \edeS08 vj} |-4.2563
(b) e’ = {0.2114 -0.4114 -1.9329}x107
4.3 -0.5468N
 Antesiertlio’ Hxerciece a
34.33
1.7778
F3, = Fyy =-3.6667, Fo, = Foy =—6.0000, Fo, = Foy = ~2.3333
F3y = Fyy = —2.9830, Fe, = Fey = -9.3557, Foy = Foy = 5.2432, Fir, = Fizy = -0.8179
Fisx = Fisy = 0.8676
4.9 SCF = 1.469

Chapter 5

4.6111 1.0000 -1.3889 -0.5000 —-—2.8889 —-0.5000

1.0000 3.0000 1.0000 0.0000 —-2.0000 -3.0000
-1.3889 1.0000 1.6111 -0.5000 0.1111 -0.5000
5.1 K=27
~~|-9.5000 0.0000 -0.5000 1.5000 1.0000 -1.5000
2.55075 —2.0000 OTT 4:0000) 321111" 120000
—0.5000 -3.0000 -0.5000 -—1.5000 1.0000 4.5000

(27) l_3 27) rzly_ 27)h lo

5 ele +P), eee
me (Ponca): ae Or,swage

pi,=O
27) r4l
(A,+2)
1.439

1/1344

0.02381

Fi = -4.1667, Fo, = 4.5833, F3, = 4.5833

Fy = 0.2623, Foy = -0.2469, Fyy = -0.0154, Fs, = -3.7346, Fo, = 4.0123,
Fry = -3.5031
5.8 Fy, = Fe, = Fi, = Fs, = Fe, = Fy, = Fg, = 0.884 Vey

5.9 b,, = -0.1172, by = 0.0314, by; = 0.1485

Chapter 6

6.1 (ay 1 @=.30 4 c=-21.65 MPa

(b) Not safe
 Answers to Exercises

6.2 Safe
6.3 Outer apices is safe, Inner apices fails.
6.6 (a) @=51.25° and c = 4.39 MPa
(b) Jy +1.2176,, — 4.292=0
(C)e aj
Jot Oa 15g) 252 =—0
6.8 Yielding will occur
6.9 For Alternative-I Yield criterion: no yielding will occur.
For Alternative-II yield criterion: yielding will occur.
6.10 Yielding will occur.

Chapter 7

7.1 4.131, 4" iteration

7.2. 4.131, 8" iteration
7.3 4.103, 8" iteration
7.4 q, = 0.096, gq) = 0.149 and q3 = 0.186
7.5 (a) Satisfied

(b) Ae? =[-0.0134 -1.8096 5.6555 1.1085 -1.9877 -0.5734]'

x10~4
Ae* =[-0.8666 -4.3504 0.8245 -0.1485 3.9077 0.5734]' x10~
(c) o' =[-5.761 -16.043 -1.028 0.778 -1.602 0.115]’ MPa
7.6 0! =[-5.463 -15.172 -0.730 0.758 -1.552 0.091]" MPa

7.7 Ae? =[1.4956 -3.9031 4.9312 0.9816 -1.9632 oO] x10~4

Ae® =[-2.3756 -2.2569 1.5488 -0.0215 3.8832 o]’ x10“
o' =[-6.103 14.943 -0.477 0.804 -1.607 0] MPa
7.8 m=0.3862, of =[-5.596 -16.351 0.160 0.837 1.607 0.086]' MPa

7.9 (#) =1p Hr G42 “(Glaize 0) 0

eat pec i :
| Answers to Exercises |

oe sees 0° 0 «(0
—pestay iy 0 = 0-410
= es
LO eg Fe ee re
0 0 0 if2 60 0
0 0 0 On Ls2 0
0 0 0 0 Os

2(2q +1)? 2a aA 4)
where pI, a and Se ee
34(q° + 2) 21q@ +2) 54(q° +2)
7.11 Not satisfied

(a) Satisfied

(b) m=0.448

(c) Ae?’ =[0.241 -0.118 0.056 0.205]' x10~

(4) o' =[-1.052 -14.463 1.056 -4.777]' MPa
ii

7.12 (= -|a422 =f) w£soQ 0 0

00 q q q

B
ae B yi Y l= y O70 0
P 0 0 Dols Was etl) 0
0 0 0 ie Cee Ty
0 0 0 Onn 0 21/2

where
 Answers to Exercises

Chapter 8

8.1 f, = 30°, c; = 0.433 MPa.

8.35 — 5.815 MPa.
8.4 (b) JRC = 10.8

—0.500 -1.000 -1.500 —-2.000 -2.5000 -3.000 -3.500

(c)
96020 7.9703 7:0256 (6.3509 5.8276 ~~ S. A000" 5.0385

8.5

|-05
-1.0
| 08 | 38657
A
|ae)
04 |
-1.5 | 1.4 45°

teal a
—3.5
25 [as
6.906E9 3.0827E9 T7S57E7 2.9907E9
3.0827E9 7.3448F9 2.576E8 3.122E9
8.6 (a) (Pies
TITEL, 2576E8 ~2.1498E9" .998E7
ZINTIED “31A22E9* ~ 9O.98E7. 5 7 81829

0.599
hots ae
(b) e= x 107
0.140
—0.001

8.7 Displacement increases by 10 times.

8.8 Joint does not slip.

7/4k W4k 1/240 —2

Ary fae V4) 74k aera i/o
h oy ‘3 25, Ce
allele =i.” «3/ke 0 1/J2
0 0 0 1 0
[_ Answers to Exercises | [ 265 |
Chapter 9

9.2 u’=—39.24 & v = -20.87 and t = -7.847 MPa & o,, = -10.433 MPa

Teer sen k,dw” + tang; k,,,dw”

ky + tan? jk,
9.4 (a) Joint slips.
(b) 1; = -3.071 MPa, o,' = -3.181 MPa
(c) 1; =-3.123 MPa, o/ = -3.166 MPa
9.5 (a) Joint slips.
(b) 7; = -3.145 MPa, o,/ = -3.154 MPa
(c) 1 = —3.141 MPa, o' = -3.151 MPa
9.6 (a) qzy = -0.00177 m, day = —0.00188 m, all others are 0.
(b) of = [-5.68 -20.38 -0.53 -6.51] MPa
(c) 43, = —0.00167 m, gy, = -0.00167 m, all others are 0.
(d) F, = 0.969 > 0, joint slips.
Dea -8618en 29:2185 7.970- —6.515] MPa
 a. _- 2m
( 268 | ee
om a

an
1 5+eae 9 8 ee aaa
83 ~ S38 Pe.
ee wae = as ;
o, {030 -~lom --ya <8
Ww 7975 Take

~*
I . toe:
[es
| 3 | OA Ass| 04
a ae a a = 5M BL @.
“US20 je =a a Tin I21é- = eM ieee
LF mo :
21 eae = aarti
\i5 af si Tile eeo- se0c- pat}=
(3. te pleat Ca100.0-=ap T0100 = ue a=
— nat

1 o9068d TORE. - 2Ohgnily j9i0{ .0-<"202.0 = A o

TsET
ossgeo ttauseo aM setta,, 28 é1S.es- a .
3.6 (a) D, =
ms ©
ISTRT 25% picks SOnE7 |
> ae

-
Eea PAVED7 AIZES GMREP » FRe2E9}
.

9.3

: a | a F (
b) , 4 }
‘ e= tae
z “¢

VARI]

&? Deuces er by 16 tite

a

>
i : i> 4 .
<li
 m-plane, 140 Dilatancy, 198, 210
Dilation angle, 173, 201
Direct shear test, 197
Angle of Discretization with finite element
internal friction, 142 in one dimension, 53
instantaneous friction, 205, 206 in three dimensions, 117
Anisotropy, 45 in two dimensions, 54
Axisymmetric, 41, 110

Effective plastic strain, 168

B matrix Elastic-brittle plastic, 33
axisymmetric, 111 Elasticity, 33
quadrilateral, 98 Elastic-perfectly plastic, 33, 47
three-dimensional, 120 Elastic-plastic non-linear analysis, 166
triangular, 62 equivalent material, 241
Boundary condition implementation of Mohr-Coulomb criterion, 231
essential, 78 implementation of Bartons criterion, 231
natural, 78 plastic correction, 229
Brittle behaviour, 167 rock joint, 216, 228
Bulk modulus, 36 Elastic plastic constitutive matrix, 170
Element
cubic triangle, 54
Cohesion, 142 linear bar, 53
instantaneous, 205, 206 linear hexahedral, 119
Compliance matrix, 36 linear quadrilateral, 92
Concept of connectivity, 244 linear tetrahedron, 119
Constant strain triangle (CST), 64 linear triangle, 54
Consistency equation, 168 quadratic bar, 53
Continuum mechanics, 52 quadratic hexahedral, 120
Constitutive relation, 33, 37 quadratic quadrilateral, 92
Convergence criterion quadratic tetrahedron, 119
displacement based, 165 quadratic triangle, 54
force based, 166 Equilibrium equation, 67
Equivalent material approach, 233
comparison, 239
Deviatoric plane (see 7 plane) method I, 234
Deviatoric stress (see stress) method II, 238
267
[268
||_ Index

Euler’s method, 161 Joint frequency, 210

JRC (Joint roughness coefficient), 199, 203

Failure criteria
intact rock, 136 Lame constant, 36
rock mass, 138 Linear strain triangle (LST), 65
Finite element method, 50 Load (see force)
Flow rule Lode angle, 141
associative, 169, 229
non-associative, 170, 229
normality condition, 169 Material nonlinearity, 158
Force Material state parameters, 167
body, 28, 71, 100, 115, 126 Modulus of elasticity
concentrated (point) 67, 76 average, 34
reaction, 82, 127 secant, 34
traction 67. 725 74. LOT swlZ5 tangent, 34, 47
Functional, 65 Modulus of rigidity, 36
Mohr’s circle of
hydrostatic, 22
Galerkin residual method (see Weighted residual pure shear, 22
method) stress, 20
Gauss point, 70, 99 strain, 25
Global stiffness matrix triaxial compression, 22
reduced, 79 uniaxial compression, 22
unreduced, 76, 78 uniaxial tension, 22
Goodman’s joint element
direct formulation, 223
numerically integrated, 226 Natural coordinates, 55, 56
GSI (Geological strength index), 147 area, 56, 92
definition, 55
volume, 117
Hooke’s law, 35 Nonlinear isotropy, 45
general, 37 Nonlinear solution method
Euler, 161
initial stress, 160
Incremental method (see Euler’s method) modified Newton—Rapshon, 160
Isoparametric finite element, 55 Newton—Rapshon, 159, 163
Isotropic, 34 Normal stiffness, 192
Numerical integration, 69, 99

Jacobian, 61
JCS (Joint wall compressive strength), 199, 203 Pascal triangle, 54, 94
Jointed rock mass Penalty method, 78
deformability, 206 m plane, 140
fractured strength (multiple joint), 196 Physical system, 51
fractured strength (single joint), 193 Plane strain, 40
fractured strength (two joint), 195 Plane stress, 39
normal deformation, 206 Plasticity, 45
shear deformation, 208 Plastic potential function, 169
Joint compliance, 210, 212 Plastic potential multiplier, 169
multiple joints, 214 Poisson’s ratio, 35
Joint element
discrete, 221, 223
equivalent material, 221 Rayleigh-Ritz method, 65
— Index
| [269|
Residual friction angle, 199 Shear, 3
RMR (Rock Mass Rating), 148 volumetric, 27
Rock mass strength Stiffness matrix
compressive, 150 quadrilateral, 70, 99
tensile, 151 three dimensions, 124
Row and column adjustment, 79 two dimensions, 69
Strength
compressive, 144
Shaft pillar design, 127 shear, 3, 191
Shape function, 53, 55, 57 tensile, 144
direct method, 57, 95 yield, 33
Serendipity method, 58, 95 Strain hardening, 33, 47, 167
Lagrange’s interpolation method, 59, 96 Strain softening, 33, 167
Shear fractures, 135 Surface traction, 51
Shear stiffness, 192
Shear strength
Barton’s envelop, 199, 217 Tangent stiffness method, 178
Ladanyi and Archamcault’s envelop, 204 Transformation
Mohr-Coulomb envelop, 204, 216 displacement, 8
peak, 198 strain, 8
residual, 198 stress, 9
Skyline storage method, 250 three dimensions, 11
Stationary principle of potential energy, 52, 66 joint compliance, 208
Stress Tensor
at nodes, 83, 104 Cauchy’s infinitesimal strain, 6
average, 2 strain, 6
definition, 1 stress, 4
deviatoric, 138 Transverse isotropic, 43
equilibrium equations, 28 Tunnel
induced, | close-form solution, 85
in situ, 1 elastic plastic analysis (Hoek-Brown), 184
invariants, 14 elastic plastic analysis (Mohr-Coulomb), 180
maximum shear, 18 finite element solution, 85, 106, 115
normal, 2 with joints, 239
octahedral plane, 18
plane, 39
Variational methods, 65
principal in 2D, 13
principal in 3D, 15
shear, 2 Weighted residual method, 245
Strain
compatibility conditions, 30
definition, 3 Yield criteria
energy, 7 alternative, 153
maximum shear, 25 Drucker-Prager, 146
normal, 3 Hoek-Brown, 147, 151
plane, 40 implementation, 171, 176
principal strain in three dimensions, 24 Mohr-Coulomb, 142, 144
principal strain in two dimensions, 24 tensile, 155
 8 ae a [pteroenas ——
vo S
_ Pal +s <—h bem wes —

eS ae & oie =
Heke aah sl ic
s al ~ te,
>) 1a ae
e = Pay pe tnagtae™
lee Sieh | om, = BO =
aa + Bye eek? iad Et aqoeiants oH Lome: conten
—. ied, 3 a ay Jnoizeay Owl =<intot siratn
NE yite alae NEPERT o g |y ARURS eed ase pens
thee = - REE ove NQONOS inter grey fF
= sociated (a ret .E-wiede ;
ie
. be
sorpen- kine
i
iat zh Et
f stters!
«fionnt . vw
4
ot hissed condi, pas ; Ef bie wend
: ‘4 Ch .gimnobitl En? yp dere gti de
: by Mp ot ; oe £0 smastiog nieve are a, asf
An = ‘ Z " rn ~~ 3
T) Bphenatneet ¢petat) ¢ 1* HOUR SORMMS:, . gveragh, oe
ni a
277 5 sepa Bm a ra
. 4 “=o
<. 1? 6: =: tn Zant
avi Abate aoguifiite mantel e ° ses
8g “aotemetensyT pri eeecare a ui
me
: «
J 6 derraadiat. * bydremetatrg ane “b08,soi dmolne
a is pega oe the cr Weighed re ” pure whee! '
* y fp ves ~ 7 - ‘gh < Sosieoi ‘ Sasa, af ——-
SG age pate a 6:
oie
Te mn ‘ ay ¥ a iqreos amo :
orm
trax
‘ wy
ft Ses >
: cznet- .
7 * s cn ¥ . a anitaial comp eeyion, 23 I oT) :
es ‘ Agni me - Mawes nt fongian, 22 : ho 8 |
j a ae . on : :
firemites is , 2 } B 22302
a.”
: 7digees naw aay LE Stgotied sxsvenn? . sl Giesnelte “eee
"o> GST (Genlogical treagih indegh, 147 boanaT , 3S 2noitaaps mundi
28 noititoz «mii-szol> r
a. KS! (crwosit-isoH) ezyions pes oe 1 bee? . reci
J He eh, Jdaolvo-adol) sezeiens oiaslq sritsle ial seltuasin

% * seacen!
eit 01 as sees ol Peoem ohoy- ; PMMA.
Sa SOI ae
Petts
ee Eoler, 163 ;
ror ; “ ~witted sorete TAO tye is ee

age,
Bt f
naam
_—~ ‘ “im
Cert “erst Ldadihines Hetitied
ig the ers
—_

atin
. ttaxt if Le
Newtow
fF aesinny
apshon lap,»108 aiTagen
a

-1 69 - (G3
pearw
istsroy . otras stithees, {97 el st a:ok
2Ke te tt a erga eee 2
f ae eee “oe OF enoifibaae vis
Ske
SSa “d .
en) . Tt . 2 os
oll

pHa Et ap i kermtisbegesri Shs

‘
haze A bid). SRI dreolue adele =: ' «an *S atimenaaib Sos! at mia tanning, i
; 221 -shiens? Piece oo & gaptenamib' owt ni niewa laquan .
war «rt =f ; . Pmeic ootcetal fim Andes, 1é> at
diat maealWptiery 18's 3al a
 oy
x
Baa 88S
ys For the last three decades, numerical methods have found
ok a wide range of applications in geomechanics. Of all the n it e
8 numerical methods, the finite element method (FEM) has the
x maximum applications in geomechanics. This book offers an
insight into applications of FEM in designing, analyzing and |
optimizing structures or excavations made in the rock mass. +
the blend of finite element methodology and principles of id e mM e N
_ rock mechanics, the text emphasizes the basics of stress-
strain analysis, isoparametric finite element method, rock
mass yielding/failure behaviour and its formulation in FEM
procedure, and joint behaviour (as equivalent material and © O
discrete system).

KEY FEATURES | Concepts and

e In-depth analysis of strength and deformability of jointed
rock mass is provided. A\pplications in
¢ Many worked-out examples, including the tunnel
examples, are interspersed throughout the text to help ( i °
the readers grasp the concepts easily. eomecnanics
e The use of finite element method in elastic and elastic-
plastic rock joints is discussed.
Besides senior undergraduate and postgraduate students in }
civil (including geotechnical) and mining engineering, the book | Debasis BAS
will also benefit the practising engineers and researchers
who wish to acquaint themselves with state-of-the-art
techniques of FEM.

THE AUTHO
DEBASIS DEB, Ph.D. (University of Alabama), is an Assistant Professor in Department of Mining
Engineering at Indian Institute of Technology Kharagpur. Earlier, he worked as a Research Associate in
the University of Alabama and Korea Institute of Geoscience and Mineral Resources (KIGAM), South
Korea, and as Visiting Assistant Professor in the Southern Illinois University. He has undertaken seve
research projects related to applications of finite element method in geomechanics and has to his
credit several research papers published in various national and international journals.

Rs. 250.00

www.phindia.com
Vi nN
llHI

T e Wel ee