A Course On Damage Mechanics-Springer-Verlag Berlin Heidelberg (1996) Professor Jean Lemaitre (Auth.)

A Course on Damage Mechanics
Springer-Verlag Berlin Heidelberg GmbH

Jean Lemaitre
A Course on
Damage Mechanics
With a Foreword by
Professor H. Lippmann
Second Revised and Enlarged Edition
With 118 Figures
Springer
Professor Jean Lemaitre
Laboratoire de Mecanique et Technologie
E.N.S. de Cachan, C.N.R.S.
U niversite Paris 6
61, avenue du President Wilson
F-94235 Cachan Cedex, France
ISBN 978-3-540-60980-3
Library of Congress Cataloging-in-Publication Data
Lemaitre, J. (Jean), 1934-

A course on damage mechanics 1Jean Lemaitre ; with a foreword by
H. Lippmann. - 2nd and enl. ed.
p. om.
Includes bibliographical references and index.
ISBN 978-3-540-60980-3 ISBN 978-3-642-18255-6 (eBook)
DOI 10.1007/978-3-642-18255-6
1. Fracture mechanics, 2. Continuum damage mechanics, I. Title.
TA409.L46 1996
620.1'126-dc20
This work is subject to copyright. All rights are reserved, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data
banks. Duplication of this publication or parts thereof is permitted only under the provisions
of the German Copyright Law of September 9, 1965, in this current version, and permission
for use must always be obtained from Springer-Verlag. Violations are liable to prosecution
under the German Copyright Law.
Springer-Verlag Berlin Heidelberg 1996
Originally published by Springer-Verlag Berlin Heidelberg in 1996
The use of general descriptive names, registered names, trademarks, etc. in this publication
does not imply. even in the absence of a specific <tatement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
Typesetting: Fotosatz-Service Kohler OHG, Wiirzburg
Cover design: de'blik, Berlin
SPIN: 10710372 55/3012 - 5 4 3 2 1 - Printed on acid-free paper
Foreword
A new branch of science usually develops thus. Somebody publishes the basic
ideas. Hesitatingly at first, then little by little, other original contributions appear,
until a certain threshold is reached. Then, overview articles are printed, conferences
are held, and a first mention is made in textbooks, until specialized monographs
are written. Continuum damage mechanics has reached that status now.
To analyze or, if possible, to predict the failure of machine parts or other structures
is one of the main goals of engineering science. Consequently fracture mechanics
became one of its leading branches. It was based on the analysis of existing cracks.
However, especially under conditions of cyclic loading, this might be too late to
prevent a disaster. Therefore, the question regarding the precursory state, that is,
the evolution of internal damage before macrocracks become visible, was then
posed. One of the successful approaches to the problem was Weibull's theory which
examined, in a statistical manner, the " weakest link " in the material volume under
consideration. Unfortunately it proved too difficult mathematically to be applied to
complicated parts or structures. Therefore it was highly appreciated by the scientific
community when L.M. Kachanov published in 1958 a simple model of material
damage which subsequently could be extended to brittle elastic, plastic or viscous
materials under all conditions of uniaxial or multiaxial, simple or cyclic loadings,
so that it may be considered nearly universal. As usual in a living science many
questions are still open, and the limitations of the model need further research.
Jean Lemaitre, the author of the present work, has become one of the most dis-
tinguished representatives of damage mechanics. His monograph highlights all the
aspects of the subject, from the basic ideas, to theoretical and thermodynamical
perspectives, up to numerical or experimental applications. It also devotes suf-
ficient space to the problems associated with parameter identification. I like the
style of presentation, which is always vivid or humorous and never dry, so that
the understanding even of the more difficult passages is facilitated . An excellent
book has finally appeared, most valuable to any engineer or scientist dealing with
modem problems of strength of materials.
I wish it every success.
Munich, 6 August 1990. Professor Horst Lippmann

Lehrstuhl A fUr Mechanik
Technische Universitiit Miinchen,
Germany
Introduction
Why an apple on the cover page?

And why not? Do you know any other fruit species with so many different flavors?
Besides, I like apples! It is also a joke from the early 1970s when it was so
fashionable in France to teach "modern" mathematics. At that time you could
draw a potato on the blackboard as the set of points having the neighbouring
Bernoulli ' s property of representing a beam! So, to tease my colleagues, I began
to draw an apple as an elementary solid, and the students just loved it!
Why damage?
Since people began to design and make components most of their efforts have been
spent trying to avoid failure! Stress and strain concepts were hardly developed
before the concept of damage appeared; this was probably because they are related
to displacement, which can be seen and touched. Damage is the deterioration which
occurs in materials prior to failure, and generally you see and you touch nothing!
For centuries fracture has been studied as a "yes or no " process related to a critical
value of load, stress, strain , and time or number of cycles of loading. A. Palmgreen
(1924), M.A. Miner (1945), and E.L. Robinson (1952) pioneered the concept of a
variable related to the progressive deterioration prior to failure, but 1958 is the
year to be considered the starting point of continuum damage mechanics, when
L.M. Kachanov published the first paper on a field variable 7jJ called " continuity".
About fifteen years later D = (1 - '1jJ ) received the status of an internal state
variable in the thermodynamical sense: 0 ::; D ::; I (0 for the undamaged state
and I for failure). During these fifteen years, this concept was practically ignored
and only one important result appeared, in 1968, with the concept of effective
stress introduced by Y.N. Rabotnov. The basic developments of damage mechanics
occurred during the 1970s, at least ten years after the tremendous development of
fracture mechanics. In the 1980s, the theory was set up on a more rigorous basis
using thermodynamics and micromechanics, and applications to engineering began
as many more people were involved in this discipline.
Why a course?
The idea of assembling, into a coherent instructional unit, the main results of
damage mechanics came from Prof. Lee Hao when he invited me to teach a course
in 1983 at Huazhong University in Wuhan (China). Afterwards, it became the
"Damage Circus" which travelled to Rio de Janeiro (Brazil) in 1985 invited by
Prof. R. Sampaio, to Nagoya (Japan) in 1986 invited by Prof. S. Murakami, to
Miinchen (Germany) in 1987 invited by Prof. H. Lippmann and based in Paris
for a graduate course "Mecanique de I'endommagement " of a joint " Universite-
viii Introduction
Grandes Ecoles" 3rd cycle "Mecanique et materiaux" initiated by D. Fran~ois,

A. Pineau, and A. Zaoui. The opportunity to write "Damage Mechanics" in book
form was offered to me by the Humboldt Foundation through an award which
gave me the possibility to spend some "free time" in 1988 and 1989 with
Prof. H. Lippmann and Prof. 1. Najar at the Technische Universitiit Munchen.
The second opportunity to put pen to paper was provided by an invitation of
Prof. F. Leckie to spend the whole academic year, 1989-90 at the University of
California, Santa Barbara (USA) where I benefited from the stimulating company of
F. Leckie, A. Evans, 1. Hutchinson and R. McMeeking. By the way, the "Damage
Circus" also played as the "Damage Brothers" in 1986 for a CISM course in
Udine (Italy) together with my friends 1. Hult, D. Krajcinovic, F. Leckie, and
S. Murakami and as the "Modem Damage Quartet" in 1988 and 1990 with
my young colleagues and friends A. Benallal, R. Billardon and D. Marquis. For
this synthesis, which is certainly neither complete nor accurate, I have benefited
from several conferences devoted to damage mechanics listed in the literature
references. The French group of labs "GRECO-CNRS" "Grandes deformations et
endommagement", which included more than 25 groups working together between
1980 and 1989, was the melting pot of the French solid mechanics school initiated
by P. Germain, in which many ideas and applications were discussed. But I must
say that it was first in ONERA with my friend 1.L. Chaboche and later, after 1975,
at the "Laboratoire de Mecanique et Technologie" at Cachan, that I found an
exceptional atmosphere of research and friendship in which to discuss, develop or
reject ideas. I thank you all for your help and your friendship.
Why four chapters?
The main purpose of this book is to develop mechanics, which is a macroscopic
modelling, so that it can be used in the design, processing, testing and control of
mechanical components in engineering. Then it becomes physics for mechanicans,
mechanics for mathematicians, mathematics for computer specialists, computations
for engineers, and engineering for physicists!
- The first chapter deals with the phenomenological aspects of damage with a
description of the different kinds of damage and the main ideas of modelling
applied to the one-dimensional case. An important section is devoted to damage
measurements, the field which probably needs most further research.
- The second chapter concerns the main part of damage mechanics in which the
couplings between damage and strains are written for the three-dimensional case.
Thermodynamics is the main framework of the theory, although micromechanics
is widely used to derive the analytical form of the general functions introduced.
- The whole third chapter is devoted to the kinetic laws of damage evolution. It
is an attempt to unify many models that have been proposed for brittle damage,
ductile and creep damage, low and high cycle fatigue of metals, polymers,
concrete, ceramics and composites.
- In the fourth chapter, several methods are described for predicting the conditions
for crack initiation in structural components: either simple methods with a few
calculations or more accurate methods involving computerized calculations, and
Introduction ix
the finite element method. A post-processor of damage growth, up to crack

initiation, is described in detail.
These chapters are written as a course for the senior-undergraduate or graduate
level with detailed calculations and exercises at the end of each chapter section.
This is a course for universities or "Grandes Ecoles" but also, I hope, a new course
for the permanent education of engineers in research centers and industry.
Many people have read the manuscript to improve shortcomings in my English and
my explanations, so let me then thank them for their advice: Prof. H. Lippmann
who wrote the Foreword, Prof. F. Leckie, Drs. A. Benallal, R. Billardon, I. Doghri
and D. Marquis; some of my students at UC-Santa Barbara: S. Gunawardena,
K. Herath, J. Manderscheid, D. Sherman; and June Finney and Sabine Lemaitre
who typed the manuscript with much care. The cover picture was taken by
Annie Lemaitre. Merci.
Spring 1988 - Spring 1991 . Jean Lemaitre

Contents
Notation xv
Chapter I Phenomenological Aspects of Damage

1.1 Physical Nature of the Solid State and Damage
1.1.1 Atoms, Elasticity and Damage I
1.1.2 Slips, Plasticity and Irreversible Strains . 3
1.1.3 Scale of the Phenomena of Strain and Damage 3
1.1.4 Different Manifestations of Damage 4
1.1.5 Exercise on Micrographic Observations . . to
1.2 Mechanical Representation of Damage ... . II
1.2.1 One-Dimensional Surface Damage Variable II
1.2.2 Effective Stress Concept . . . . . . 12
1.2.3 Strain Equivalence Principle . ... 13
1.2.4 Coupling Between Strains and Damage;
Rupture Criterion; Damage Threshold . . . . . . . . 14
1.2.5 Exercise on the Micromechanics of the
Effective Damage Area 17
1.3 Measurement of Damage 19
1.3.1 Direct Measurements 19
1.3.2 Variation of the Elasticity Modulus 21
1.3.3 Variation of the Microhardness 25
1.3.4 Other Methods . . . . . . . . 28
1.3.5 Exercise on Measurement of Damage by the
Stress Amplitude Drop ...... . . 36
Chapter 2 Thermodynamics and Micromechanics of Damage 39

2.1 Three-Dimensional Analysis of Isotropic Damage 39
2.1.1 Thermodynamical Variables, State Potential 39
2. 1.2 Damage Equivalent Stress Criterion 44
2.1 .3 Potential of Dissipation ..... 46
2.1.4 Strain-Damage Coupled Constitutive
Equations . . . . . . . . . . . . . . . . . . 50
2.1.5 Exercise on the Identification of Material Parameters 59
xii Contents
2.2 Analysis of Anisotropic Damage . . . . . . 59

2.2.1 Geometrical Definition of a Second-Order
Damage Tensor . . . . . . . 60
2.2.2 Thermodynamical Definition of a
Fourth-Order Damage Tensor . . . . . . . . 63
2.2.3 Energetic Definition of a Double Scalar Variable 64
2.2.4 Exercise on Anisotropic Damage in
Proportional Loading 66
2.3 Micromechanics of Damage . 68
2.3.1 Brittle Isotropic Damage 68
2.3.2 Ductile Isotropic Damage 72
2.3.3 Anisotropic Damage 75
2.3.4 Microcrack Closure Effect, Unilateral Conditions 80
2.3.5 Damage Localization and Instability 86
2.3.6 Exercise on the Fiber Bundle System 90
Chapter 3 Kinetic Laws of Damage Evolution . . . 95

3.1 Unified Formulation of Damage Laws . . 95
3.1.1 General Properties and Formulation 96
3.1 .2 Stored Energy Damage Threshold 98
3.1.3 Three-Dimensional Rupture Criterion 100
3.1.4 Case of Elastic-Perfectly Plastic and
Damageable Materials . . . . . . . . . . 105
3.1.5 Identification of the Material Parameters 109
3.1.6 Exercise on Identification by a Low Cycle Test 110
3.2 Brittle Damage of Metals, Ceramics,
Composites and Concrete 112
3.2.1 Pure Brittle Damage 113
3.2.2 Quasi-brittle Damage 113
3.2.3 Exercise on the Influence of the Triaxiality
on Rupture .... . .... . ... 116
3.3 Ductile and Creep Damage of Metals and Polymers 117
3.3.1 Ductile Damage ............ 117
3.3.2 Exercises on the Fracture Limits in Metal Forming 119
3.3.3 Creep Damage . . . . . . . . . . . . . 123
3.3.4 Exercise on Isochronous Creep Damage Curves 126
3.4 Fatigue Damage . . . . . . . . . . . 127
3.4.1 Low Cycle Fatigue . . . . . . . 127
3.4.2 Exercise on Creep Fatigue Interaction 132
3.4.3 High Cycle Fatigue . . . . . . 135
3.4.4 Exercise on Damage Accumulation 141
3.5 Damage of Interfaces . . . . . . . . 143
3.5.1 Continuity of the Stress and Strain Vectors 143
3.5.2 Strain Surface Energy Release Rate . . . 145
Contents xiii
3.5.3 Kinetic Law of Debonding Damage Evolution. . 146

3.5.4 Simplified Model . . . . . . . . . . . . . . 147
3.5.5 Exercise on a Debonding Criterion for Interfaces. 150
3.6 Table of Material Parameters. . . . . 152
Chapter 4 Analysis of Crack Initiation in Structures 154

4.1 Stress-Strain Analysis 154
4.1.1 Stress Concentrations 156
4.1.2 Neuber' s Method . . 163
4.1.3 Finite Element Method 165
4.1.4 Exercise on the Stress Concentration Near a Hole 167
4.2 Uncoupled Analysis of Crack Initiation 169
4.2.1 Determination of the Critical Point(s) . . . . . 169
4.2.2 Integration of the Kinetic Damage Law . . . . 169
4.2.3 Exercise on Fatigue Crack Initiation Near a Hole 172
4.3 Locally Coupled Analysis . . . . . . 176
4.3.1 Localization of Damage . . . . 176
4.3.2 Postprocessing of Damage Growth 177
4.3.3 Description and Listing of the Postprocessor
DAMA~~ .... . . . . . .. . 179
4.3.4 Exercises Using the DAMAGE 90 Postprocessor 196
4.4 Fully Coupled Analysis ...... . . . 201
4.4.1 Initial Strain Hardening and Damage . . 202
4.4.2 Example of a Calculation Using the Finite
Element Method . . . . . . . . . . 203
4.4.3 Growth of Damaged Zones and Macrocracks 206
4.4.4 Exercise on Damage Zone at a Crack Tip 208
4.5 Statistical Analysis with Microdefects 209
4.5.1 Initial Defects . . . . . . . . 210
4.5.2 Case of Brittle Materials . . . 210
4.5.3 Case of Quasi Brittle Materials 212
4.5.4 Case of Ductile Materials . . 214
4.5.5 Volume Effect . . . . . . . . 215
4.5.6 Effect of Stress Heterogeneity . 217
4.5.7 Exercise on Bending Fatigue of a Beam. 219
Bibliography 222
History of International Damage Mechanics Conferences 223
Authors and Subject Index . . . . . . . . . . . . 225

Notation
The choice of notation is always a difficult problem in a new field because all letters
have already been assigned a meaning. I have tried to be as classical as possible,
but I could not completely avoid using the same letter for different meanings. Be
careful and please forgive me.
Operators
X a scalar
xi components of a vector x
x ij components of a second-order tensor x
[xl matrix
X time derivative of x(x = dx/dt)
Xi ,j gradient of x
divergence of x

Xij,j
Dij Kronecker delta: Dij I if i = j, Dij = if i -I- j

x kk trace of x
XijX ij trace of x2
D
x ij components of the deviatoric tensor x8 = x ij -
I
3XkkDij
Ixl absolute value of x

sign(x) + or - as the sign of x
[xll

discontinuity of x
(x) Macauley bracket (x) = if x < 0, (x) = x if x >
x mean value of x
~x range of x (peak to peak amplitude)

d,o,D differential operators
H Heaviside function: H(x) = if x < 0, H(x) I if x >
Ln Napierian logarithm
Symbols
a crack length
aijkl elastic moduli
A Gurson-Tvergaard material parameter
A crack area
b isotropic hardening material exponent

B Gurson-Tvergaard material parameter
XVI Notation
c Paris' crack growth material parameter
d length of microcells
D damage variable
Ds,D" damage variables in shear and tension
Dij , D components and second-order anisotropic damage tensor
D ijkl components of anisotropic fourth-order damage tensor
Dc critical damage at crack initiation
DIe critical damage in pure tension
specific internal energy rate

Young's modulus of elasticity
elastoplastic tangent modulus
effective damage modulus of elasticity, i;+ in tension, i;- in
compression
f yield function
l body force
F force
F function potential of dissipation
FD damage potential function
strain-stress function
vector of strain rate discontinuity
strain energy release rate
energy toughness of materials
h crack closure parameter

H microhardness
H Heaviside function
H ijkl tangent elasto-plastic stiffness matrix components
Z intensity of electricity
Ikl unit second-order tensor components
Ikl
'J .
unit fourth-order tensor components
factor
specific kinetic energy rate
stress intensity factor
cyclic plasticity material parameter
creep material parameter
viscosity material parameter
stress concentration coefficient
Notation XVII
length of the mesovolume element
m mass
M material point
M strain hardening material exponent
n number of defects
n viscosity material exponent
ii unit normal vector
N number of cycles
N Norton's creep material exponent
NR number of cycles to rupture
P accumulated plastic strain

PD damage threshold plastic strain
Px specific external force power
P porosity
if heat flux vector

Q input heat rate
T resistivity
T strain of isotropic hardening
R isotropic hardening stress variable
RXJ isotropic hardening material parameter
Ro,Ra stiffness
Rv triaxiality function
s specific entropy
s crack area
S sectional area
SD damaged sectional area
S damage strength material parameter
t time
T temperature
i1 displacement vector
v wave speed
V electrical potential difference
V volume
we elastic strain energy density

w*e elastic complementary energy density
xviii Notation
elastic strain energy of a body
coordinate
unidimensional and components of tensorial kinematic hardening
stress variable
kinematic hardening material parameter
y strain energy density release rate

y:" Y" strain energy density release rate in shear and tension
Yijkl strain energy release rate tensor components
Y" critical strain energy density release rate at rupture
unidimensional and components of tensorial variable of kinematic

hardening
nonlinear kinematic hardening material parameter
amplitude
C, Ci j ' uniaxial and tensorial total strains

e e
Ce, Cij ' uniaxial and tensorial elastic strains
p p
cp , Cij' uniaxial and tensorial plastic strains
c"j, hydrostatic elastic strain
Ceq equivalent total strain
CpD damage plastic strain threshold in pure tension
cR strain to rupture
Paris' crack growth material exponent
dissipation
Lame's coefficient of elasticity

plastic multiplier
Lame's coefficient of elasticity in shear
v Poisson's ratio of elastic contraction

v* elastoplastic contraction coefficient
scalar and tensorial volume damage variables
p mass density
uniaxial and tensorial stresses

Notation xix
stresses at microscale
(5H hydrostatic stress
D D deviatoric stress tensor
(5 ij , ( j
(5"fj von Mises equivalent stress

(5* damage equivalent stress
0- , o-i j ' iF unidimensional and tensorial effective stresses
fatigue limit stress
yield stress
plastic threshold stress

ultimate stress
stress to rupture
Helmholtz specific free energy

Chapter 1
Phenomenological Aspects of Damage
The damage of materials is the progressive physical process by which they break.
The mechanics of damage is the study, through mechanical variables, of the mecha-
nisms involved in this deterioration when the materials are subjected to loading. At
the microscale level this is the accumulation of microstresses in the neighborhood
of defects or interfaces and the breaking of bonds, which both damage the material.
At the mesoscale level of the representative volume element this is the growth and
the coalescence of microcracks or microvoids which together initiate one crack.
At the macroscale level this is the growth of that crack. The two first stages may
be studied by means of damage variables of the mechanics of continuous media
defined at the mesoscale level. The third stage is usually studied using fracture
mechanics with variables defined at the macroscale level.
When studying engineering materials such as metals and alloys, pofymers and com-
posites, ceramics, rocks, concrete, and wood, it is very surprising to see how such
materials, which have different physical structures, are similar in their qualitative
mechanical behavior. All show elastic behavior, yielding, some form of plastic or
irreversible strain, anisotropy induced by strain, cyclic hysteresis loops, damage
by monotonic loading or by fatigue, and crack growth under static or dynamic
loads. This means that the common mesoscopic properties can be explained by a
few energy mechanisms that are similar for all these materials. This is the main
reason it is possible to explain material behavior successfully with the mechan-
ics of continuous media and the thermodynamics of irreversible processes, which
model the materials without detailed reference to the complexity of their physical
microstructures.
1.1 Physical Nature of the Solid State and Damage

1.1.1 Atoms, Elasticity and Damage
All materials are composed of atoms, which are held together by bonds resulting
from the interaction of electromagnetic fields. Elasticity is directly related to the
relative movement of atoms. The physical study of the properties of an atomic
lattice leads to the theory of elasticity, but a much easier way is to write the math-
ematical constitutive equations directly at the mesoscale level using the property
of reversibility of strain, which implies a one-to-one relationship, and eventually
incorporate the properties of linearity and isotropy.
When debonding occurs, this is the beginning of the damage process. For example,
metals are organized in crystals or grains: a regular array of atoms except on many
2 Phenomenological Aspects of Damage
lines of dislocations where atoms are missing. If a shear stress is applied, the
dislocations may move by the displacement of bonds, thus creating a plastic strain
by slip without any debonding as shown in Figure 1.1.
If the dislocation is stopped by a microdefect or a microstress concentration, it
creates a constrained zone in which another dislocation may be stopped. This
second process cannot occur without a debonding damage as shown in Figure 1.2.
Several arrests of dislocations nucleate a microcrack. Other damage mechanisms
in metals are intergranular debonding and decohesion between inclusions and the
matrix.
All these mechanisms create plastic microstrains.
- In polymers, damage occurs by the breakage of bonds that exist between the
long chains of molecules.
- In composites damage is the debonding between the fibers and the polymeric
matrix.
- In ceramics it is mainly microdecohesions between the inclusions and the matrix.
- In concrete, the early damage mechanism is also a decohesion between aggre-
gates and the cement with the complex influence of water.
- In wood, the weak point where damage occurs is the bonding of the celulosic
cells.
In all cases elasticity is directly influenced by the damage, since the number
of atomic bonds responsible for elasticity decreases with damage. This coupling,
I I
.L
I I
Fig. 1.1. Elementary plastic strain by slip due to dislocation movement
Fig. 1.2. Elementary damage by nucleation of a microcrack

due to an accumulation of dislocations (after D. Krajcinovic)
1.1 Physical Nature of the Solid State and Damage 3
which occurs at the level of the state of the material defined here by the elastic
strain and the damage, is called a "state coupling" .
1.1.2 Slips, Plasticity and Irreversible Strains

Plasticity is directly related to slips. In metals, slips occur by movement of dislo-
cations, as shown schematically in Figure 1.1 or by the climbing of dislocations
and twinning. In no case does it induce any appreciable volume change.
In other materials, irreversible strains may occur by different mechanisms:
- rearrangement of molecules in polymers;
- microcracks in ceramics where the large lattice resistance does not allow move-
ments of dislocations;
- slips along surfaces of decohesions in concrete;
- rearrangement of cells in wood.
They may induce a volume change.
In all cases, damage influences plastic or irreversible strains only because the
elementary area of resistance decreases as the number of bonds decreases. The
damage does not directly influence the mechanism of slip itself; that is, there is no
state coupling. The indirect coupling owing to an increase in the effective stress
arises only in the kinetic constitutive equation, it is called "kinetic coupling".
1.1.3 Scale of the Phenomena of Strain and Damage

- Elasticity takes place at the level of atoms.
- Plasticity is governed by slips at the level of crystals or molecules.
- Damage is debonding from the level of atoms to the mesoscale level for crack
initiation.
Continuum mechanics deals with quantities defined at a mathematical point. From
the physical point of view, these quantities represent averages on a certain volume.
The "Representative V<;>lume Element", must be small enough to avoid smoothing
of high gradients but large enough to represent an average of the microprocesses.
For experimental purposes and numerical analysis it is useful to consider the fol-
lowing orders of magnitude of the Representative Volume Element which is the
scale of mesomechanics:
- metals and ceramics: (0.1 mm)3

-polymers and most composites: (I mm)3
-wood: (10 mm)3
-concrete: (100 mm)3
Another important property to consider is that the damage is always much more
localized than the strain. Remember that the damage, or debonding of atoms, is
restricted to a surface, although the strains, being movements of atoms by variation

of their distance or by many slips, occur throughout the volume. If damage exists in
a single plane at the mesoscale, there is no way to study it by classical continuum
mechanics. Fortunately, most often it exists on many planes at the microscale,
but always with a high space gradient. Micromechanics helps a great deal in for
modeling the damage behavior, as we shaH see in Chapter 2.
In conclusion:
- the microscale is the scale of the mechanisms used to consider strains and
damage;
- the mesoscale is the scale at which the constitutive equations for mechanics
analysis are written;
- the macroscale is the scale of engineering structures.
1.1.4 Different Manifestations of Damage

Even if the damage at the microscale is governed by one general mechanism of
debonding, at the mesoscale it can manifest itself in various ways depending upon
the nature of the materials, the type of loading, and the temperature.
Brittle damage
The damage is called brittle when a crack is initiated at the mesoscale without a
large amount of plastic strain. Just to give an order of magnitude, let us say that
the ratio of plastic strain to elastic strain is below unity:
cp
-<
ce
This means that the cleavage forces are below the forces that could produce slips
but are higher than the debonding forces. The degree of localization is high. An
example is given in Figure 1.3.
Ductile damage
On the other hand, the damage is called ductile when it occurs simultaneously with
plastic deformations larger than a certain threshold PD' It results from the nucleation
of cavities due to decohesions between inclusions and the matrix foHowed by their
growth and their coalescence through the phenomenon of plastic instability. As a
consequence, the degree of localization of ductile damage is comparable to that of
plastic strain. An example is given in Figure 1.4.
Creep damage
When a metal is loaded at elevated temperature, for instance a temperature above
1/3 of the melting temperature, the plastic strain involves viscosity; that is, the
material may be deformed at constant stress. When the strain is large enough, there
MPa cr
600
400
200
100llm o 0.1 0.210-2

a b
Fig. 1.3. a Fractograph showing microcracks damage around an inclusion in a silicon nitride ceramic;
b Tensile stress-strain curve up to rupture (after F. Hild)
are intergranular decohesions which produce damage and an increase of the strain
rate through the period of tertiary creep. As for ductile damage, the gradients of
creep damage are similar to the visco-plastic strain gradients. An example is given
in Figure 1.5.
Low cycle fatigue damage

When a material is subjected to a cyclic loading at high values for stress or strain,
damage develops together with cyclic plastic strain after a period of incubation
preceding the phases of nucleation and propagation of microcracks. The degree of
a
o'(MPa) Damage
500
400
300
200
100
e
b 0 10 20 30 40 50 60 70 80 >10- 2
Fig. 1.4. a Fractograph showing microcavities damage at the grain boundaries of a steel (after L. Engel
and H. Klingele); b Tension stress-strain curve of A!S! 1010 annealed steel (after J. Dufailly)
damage localization is higher than for ductile or creep damage. Because of the high
values for the stress, the low cycle fatigue is characterized by low values of the
number of cycles to rupture, N R:
N R < 10000 cycles.
If the material is strain loaded, the damage induces a drop of the stress amplitude
as shown in Figure 1.6 for two stress-strain loops corresponding to the stabilized
cycle and a cycle close to the rupture.
1. 1 Physical Nature of the Solid State and Damage 7
a
-2 ep
.10
20
10
t (h)
b o 10000 15000
Fig. 1.5. a Fractograph showing intergranular cavitation damage formed at 800 C in a nickel-based
alloy (after L. Engel and H. Klingele); b Creep curve under constant tension stress of A 542 stainless
steel at 550C (after A. Benallal)
For metals, the damage can be either intergranular or trans granular microcracking
following slip-band arrests.
High cycle fatigue

When a material is loaded with lower values for stress, the plastic strain at the
mesolevel remains small and is often negligible. It may be high at some points at
t- - - - -- - - - - - - - - - - --
- 0.2
----- ----~-------------------------
b
Fig. 1.7. a Fractograph of transcrystalline fatigue fracture in a nickel alloy (after L. Engel and
H. Klingele); b Cyclic tension compression curves for high cycle fatigue of A 316 stainless steel
Note that for brittle damage and high cycle fatigue damage, a stress-strain curve ob-
tained from a classical tension-compression test at the meso-scale usually does not
represent the "true" behavior for strain and damage because the space localization
induces microplastic and damage zones much smaller than those of the specimens.
Nevertheless, it is used because mechanical experiments at the microscale are diffi-
cult to perform; but please bear in mind that the results are averages of nonuniform
quantities over a mesovolume. The microhardness test may help to characterize a mi-
crovolume as it involves a size of the order of microns but its state of stress is complex.
Fig. 1.6. a Fractograph showing ductile cycle striations of fatigue in an 18:8 austenitic steel (after
L. Engel and H. Klingele); b Cyclic tension compression curves for low cycle fatigue of A 316 L
stainless steel (after J. Dufailly)
the micro level where transgranular microcracking occurs only on some planes and
most often at the surface of the specimen by the mechanism of intrusion-extrusion.
The number of cycles to failure may be very large:
NIl> 100000 cycles.
As a consequence, the localization of damage is high and the drop of stress at the
mesolevel for an experiment similar to that of Figure 1.6 occurs later on Figure 1.7.
1.1.5 Exercise on Micrographic Observations
Look carefully at the pictures of materials in Figure 1.8 taken with a microscope
and find the mechanism of strain damage or rupture for each case. The answers
are, in random order:
A - Transcristalline fatigue crack propagation in a nickel alloy at 650C

C - Brittle transcristalline cleavage fracture in pure tantalum at room temperature
A - Creep voids at the grain boundaries of a nickel-based alloy at 800
Fig. 1.8. Micrograph pictures taken from An Atlas of Metal Damage by L. Engel and H. Klingele and
also from D. Fran,>ois and Y.Berthaud for the two last pictures
1.2 Mechanical Representation of Damage II
N- Fracture of a glass fiber epoxy composite

P- Fatigue crack in a steel specimen at room temperature
E- Tensile crack in a concrete
H- Ductile dimple fracture around manganese sulphide stringers in a steel at
room temperature
M - Thermal cycling cracks in the nozzle of a metal squirting device
G - Stress corrosion transcristalline cracking in an austenitic stainless steel.
Hint: if the answers are correctly matched to the pictures, the sequence of capital
reference letters should spell the name of a most famous wine.
1.2 Mechanical Representation of Damage
1.2.1 One-Dimensional Surface Damage Variable (L.M. Kachanov , 1958)
It follows from the preceding section that damage may be interpreted at the mi-
croscale as the creation of microsurfaces of discontinuities: breaking of atomic
bonds and plastic enlargement of microcavities. At the mesoscale, the number of
broken bonds or the pattern of microcavities may be approximated in any plane by
the area of the intersections of all the flaws with that plane. In order to manipu-
late a dimensionless quantity, this area is scaled by the size of the representative
volume element. This size is of primary importance in the definition of a variable
continuous .in the sense of continuum mechanics. At one point, it must be the rep-
resentative effect on failure of microdefects over the mesoscale volume element.
It is similar to plasticity where the plastic strain Ep represents, at one point, the
average of many slips.
Let us consider a damaged body and a Representative Volume Element (RVE) at
a point M oriented by a plane defined by its normal ii and its abscissa x along the
direction ii Figure 1.9.
- Let 85 be the area of the intersection of the plane with the RVE;
- Let 85 D", be the effective area of the intersections of all microcracks or micro-
cavities which lie in 85;
Fig. 1.9. Micro-meso definition of damage (Apple after 1. Lemaitre, 1975)

- The value of the damage D(M, ii, x) attached to the point M in the direction
ii and at the abscissa x is:
-)
D(M ,n,x 8S D x
=~.
In order to define a continuous variable over the RYE for its deterioration to failure
in two parts, one must look at all the planes varying with x and consider that which
is most damaged:
D(M ,ii) = Max[D(M
(x )
ii x )] ,
, ,
The coordinate x disappears, and:
Ir-D-(-M-'ii-)-=-~-:S-D--'I
It follows from this definition that the value of the scalar variable D (which depends
upon the point and the direction considered) is bounded by 0 and 1:
0::::; D::::; 1;
D = 0 undamaged RYE material;
-+
D =1 fully broken RYE material in two parts.
-+
In fact, the failure occurs for D < I through a process of instability (see Sec-
tions 1.2.4 and 3.1.3.).
Consideration of the simple one-dimensional case of a homogeneous damage pic-
tured in Figure 1.10 leads to the simple definition of damage as the effective surface
density of microdefects:
D_ SD
- S
1.2.2 Effective Stress Concept (Y.N. Rabotnov, 1968)

If the RYE of Figure 1.10 is loaded by a force f = iiF, the usual uniaxial stress
is:
F
(J = S.
t:F
.. +! .
50
...
-, I ~ 5
Fig. 1.10. One-dimensional damaged element

1.2 Mechanical Representation of Damage 13
If all the defects are open in such a way that no microforces are acting on the
surfaces of microcracks or microcavities represented by 5 D , it is convenient to
introduce an effective stress 0- related to the surface that effectively resists the
load, namely (5 - 5 D ):
F
Introducing the damage variable D
F
or L. -a_
I_1=D - ___a_--,I
This definition is the effective stress on the material in tension. In compression, if

some defects close, the damage remaining unchanged, the surface that effectively
resists the load is larger than 5 - 5 D. In particular, if all the defects close, the
effective stress in compression 0-+ is equal to the usual stress a . This effect is
studied in details in Section 2.3.4.
Coming back to the definition of the "effective" area of microcracks 85 D ' "ef-
fective" has to be understood as "strength", taking into account the microstress
concentrations and mutual interactions of defects loaded in tension or shear. Only
micromechanics may give a precise meaning of this concept which will be taken
into account globally at the mesoscale through identification of the damage variable
by means of its coupling with elasticity or plasticity (see exercise in Section 1.2.5.)
1.2.3 Strain Equivalence Principle (J. Lemaitre, 1971)
A way to avoid a micromechanical analysis for each type of defect and each type
of mechanism of damage is to postulate a principle at the mesoscale.
In thermodynamics, the method of local state assumes that the thermomechanical
state at a point is completely defined by the time values of a set of continuous
state variables depending upon the point considered. This postulate applied at the
microscale imposes that the constitutive equations for the strain of a microvolume
element are not modified by a neighboring microvolume element containing a
microcrack. Extrapolating to the mesoscale, this means that the strain constitutive
equations written for the surface 85 - 85 D are not modified by the damage or that
the true stress loading on the material is the effective stress 0- and no longer a.
The following principle results:
"Any strain constitutive equation for a damaged material may be derived in the
same way as for a virgin material except that the usual stress is replaced by the
effecti ve stress"
Undamaged material Damaged material

D = 0 0< D < I
C = ;:(a,) c=:F(l~D'''')
Same derivation
t t
This statement is a principle because it has been demonstrat d only in some partic-
ular cases of damage through homogenization techniques. It will be applied either
to elasticity or plasticity.
1.2.4 Coupling Between Strains and Damage; Rupture Criterion;

Damage Threshold
In accordance with the description of the damage mechanics of Sections 1.1.1 and
1.1.2 and as a direct application of the strain equivalence principle, one may write
the uniaxial laws of elasticity and plasticity of a damaged material.
Elasticity law
This is direct state coupling through the concept of effective stress:
Undamaged material I Damaged material
D=O O < D<l
a a
ce = E ce = E( I - D)
Contraction for isotropic damage:
c22 = c33 = -vce;
E is Young's modulus of the undamaged material;
v is Poisson's ratio.
The elasticity modulus of the damaged material defined by the ratio E=
E= E(I -D).
Figure 1.11 shows an example of the variation of the elasticity modulus as ductile
damage progresses. It can also be evaluated from the microcrack pattern by means
of micromechanics (8. Budiansky, R. O'Connell, 1976).
Plasticity
This is a kinetic coupling on the evolution of plastic strain which has to be written
in the plasticity criterion used to derive the kinetic constitutive equations.
In order to model plasticity two kinds of strain hardening are usually considered:
- the isotropic hardening related to the density of dislocations or flow arrests;
- the kinematic hardening related to the state of internal microstress concentra-
tions. The corresponding back stress defines the center of the elastic domain in
tension compression (or in three dimensions).
a' (MPa) E/E

1.o+---_.
200 \
I
E \
I 0.5
100 99000 \
I
e
oOL.--2~0--'-~4~0--6-r0-~8~0-.Ll0~0---l'1~0- 2 0 0'---~20-~4~0--60
~-~8~
0 --I~OO----C'1~0 -2
Fig. 1.11. Variation of the elasticity modulus with damage for copper (99.9%) (after J. Dufailly. 1975)
If a y is the yield stress, R the stress due to isotropic hardening and X the back
stress, both functions of the plastic strain, the one-dimensional plasticity criterion
defining the current threshold of yield limit, and represented in Figure 1.12 is
a=ay+ R + X
or f = la-XI-R-a y = O.
f is the yield function from which the kinetic constitutive equation for plastic strain
is derived (see Section 2.1.3):
f = 0 f < 0
:lip -::j:. 0 if { ~nd { or
f = 0 j < 0
To achieve this, let us write the total strain as
-------------
-- - ~
-
~...,..::;;;:,;=
Oy r - - - - i - -__ OR
p
O~--_+---~------_4-----+
pD I
D I
t ~I[k
o
~
L..-_ _ _"-~_ _ _ _ ___I._ _ -+.
p Fig. 1.12. Schematic evolution of the plas-
ticity yield criterion with damage
When damage occurs, according to the principle of equivalence, the yield function
f must be written as:
f = II~D-XI-R-O-y = 0
Experiments and the equation
0- = (o-y+R+X)(I-D)
show that the damage equally decreases the yield stress, the isotropic strain hard-
ening stress and the back stress as shown in Figure 1.12.
Rupture criterion
The rupture at the mesoscale is a crack initiation which occupies the whole surface
of the RYE; that is, D = I. In many cases this is caused by a process of instability
which suddenly induces the decohesion of atoms in the remaining resisting area.
It corresponds to a critical value of damage Dc' which depends upon the material
and the conditions of loading.
The final decohesion of atoms is characterized by a critical value of the effective
stress acting on the resisting area. Let us call it 0- 00 because it is the maximum
stress that could be applied to the material (see Figure 1.12):
_ 0-
0-=---=0-
I - Dc 00
Practically 0- 00 may be approximated by the ultimate stress 0-u' which is easier to

identify but is always smaller. Then
0-
De '::: 1 - -
o-u
gives the critical value of the damage at a mesocrack initiation occurring for the
unidimensional stress 0-. The ultimate stress o-u being identified as a material char-
acteristic, Dc may vary between Dc '::: 0 for pure brittle fracture to Dc '::: I for
pure ductile fracture but usually Dc remains of the order of 0.2 to 0.5.
This relation, applied to the pure monotonic tension test, which is taken as a
reference, defines the corresponding critical damage DI e considered as a material
characteristic:
where 0- R is the stress to rupture.
Damage threshold
Before the microcracks are initiated, creating the damage modelled by D, they must
nucleate by the accumulation of microstresses accompanying incompatibilities of
microstrains or by the accumulation of dislocations in metals. This corresponds in
the pure tension case to a certain value EpD of the plastic strain below which no
damage by microcracking occurs:
Ep < EpD ----+ D = O.
Finally, the four main relations which comprise the basis of damage mechanics
are:
(J ..
Ep = for elasticity;
E ( l-D )
Il-D
-(J- - X 1- R - (Jy
.
= 0 as the plastic yield criterion;
Ep < cpD ----+ D = 0 as the damage threshold;

D = Dc ----+ crack initiation.
1.2.5 Exercise on the Micromechanics of the Effective Damage Area

In Section 1.2.1., the one-dimensional variable of damage has been defined as the
"effective" surface density of intersections of microcracks and cavities in the most
damaged section. Let us illustrate the meaning of this, using a particular example.
Consider a three-dimensional mesovolume element containing a single circular
crack loaded in mode I by a load F perpendicular to the plane of the crack
(Figure 1.13.).
The overall damage defined by the continuum damage theory is the damage in the
plane of the crack,
D = 85 D .
85 '
namely,
where k is the correction factor due to the stress concentration in the vicinity of
the crack tip line to define the effective damaged area.
-u
Fig. 1.13. Single cracked meso volume element
Calculation of k as a function of a/I

Applying the strain equivalence principle together with the effective stress con-
cept, the damage may also be defined as the variation of the stiffness Ra of the
mesoelement
P = (I - D)Rou,
where u is the elastic displacement corresponding to the load P, and Ro the initial
stiffness of the uncracked element:
Ra = Ro(l - D) ,
with R = EI2 = EI
~"O I '
E being the Young's modulus of the material.
Calculation of Ra from the two concepts of classical fracture mechanics

the strain energy release rate,
W being the strain energy and A the crack area:
laWI 1 F2 dR
G = -2" aA F=const = -2" R~ 2Jrada;
and the stress intensity factor for a circular crack in an infinite body loaded in
mode I:
2 2F
K = -a00,j1i(i = - -12 ,j1i(i.
Jr Jr
For a case of generalized plane strain (1/ being the Poisson 's ratio),
K2
G = e(l-l}).
Writing the equality of the two expressions of G gives a differential equation,
1- 1/ 2 2
- e a da ,
which can be integrated from a = 0 --+ Ra = Ro, to the current state (a, R,J.
The result is
R = [~ 16(1-1/2)a3 ]-'
a Ro + 3EI4
Assuming a small crack (a/ I I) giving a small variation of the stiffness,
R = R ( _ 16( I - 1/ 2 )a3 R )
a H{) I 3EI4 o'
This gives an expression for D:
Jra 2 16(1 - 1/2 ) a

r 3Jr I'
from which it follows that
k -_ 16(137r
- V2) ~ '" ~
I - 1.55 I .
This means that for this particular case and for a/I I, the value of the damage
is always smaller than the crack density.
Determining the effective stress

In this simple example where the damage is D, the effective stress can obviously
be calculated as in tension when the crack is open;
a F+ .. .
a= [2 (1 _ D) In tensIOn when the crack IS open;
I-D
a= a = rF- in compression when the crack is closed.
This crack closure effect will be studied in Section 2.3.4.
1.3 Measurement of Damage

1.3.1 Direct Measurements
Returning to the definition of the damage as D = oSD / 0S, the direct measurement
consists in the evaluation of the total crack areas oS D lying on a surface oS at
mesoscale. Assuming the correction factor k to be equal to I, this can be done
by observing micrograph pictures. To observe a picture of an approximately (100-
mm)2 section of a RYE, a maximum magnification of 1000 is sufficient for metals,
and I to lOis enough for concrete. This is of course a destructive method and
tedious to practice. An example is given in Figure 1.14 showing ductile microcracks
and dimple formations in an high-alloy steel:
At the scale of real observation:
OSD = 1163 mm 2
oS = 11317 mm 2
In the plane of observation:
OSD
D = oS '::::' 0.1
If the damage consists of microcracks which are not perfectly homogeneous in the
RYE, there is almost no chance of obtaining even one crack in any plane of inter-
section! Therefore the damage can only be evaluated by means of the lines which
indicate the intersection of the microcracks with the plane of observation. An equiv-
alent isotropic damage may be obtained as follows : Consider a micromechanical
element of n cells of size d x d x d, each containing a crack of different appar-
ent size a i (Figure 1.15). Assuming square cracks for simplicity, and a correction
factor equal to I, the damage in the plane of the crack of each cell i is
aT
)
D = OSD
1 as d2 '
Fig. 1.14. Direct evaluation of damage (after Engel, KlingeJe and C. Lienard)
The equivalent isotropic damage in the plane of observation of the n cells is taken as
the mean value of the damage in each cell. This is the simplest way to homogenize
discrete quantities in order to define a continuous variable
n a2
~-t
D = i=1 d
n
If a micrograph of size 12 is considered with n cells corresponding to n cracks, the
surface of the micrographic may be expressed as
12 = n. d2 .
The value of the damage is simply:
~a2
t
D
y'
An example of damage is shown in Figure 1.15., where grain boundary precipitates
of titanium carbide caused intercrystalline failure in a martensitic NiCoMo steel
subjected to triaxial stresses.
/' / / / /" /'

, ,.-' ,
"""'"'
"'" /'
Id
/
/7
,
-- (
~
V
I /~ \ - ""' r
~ /
, ~ )
~ ......
/'
~
'- , .... ('
Fig. 1.15. Micromechanic model and direct evaluation of damage (after L. Engel, H. Klingele and
C. Lienard)
At the scale of real observation:

~a2, = 9000 mm 2
[2 = 10200 mm 2
In the plane of observation
D -'r ~a2
0.88
1.3.2 Variation of the Elasticity Modulus

This is a nondirect measurement based on the influence of damage on elasticity
through the state coupling as explained in Section 1.2.4:
(J
cF =
E(I - D)'
This destructive method requires the machining of specimens in order to run me-
chanical tests. It assumes uniform homogeneous damage in the specimen gauge
section.
If E = E( I - D) is considered as the effective elasticity modulus of the damaged
material, the values of the damage may be derived from measurements of E,
provided that Young's modulus E is known:
This very useful method requires accurate strain measurements. Strain gauges are
commonly used and E is most accurately measured during unloadings. An example
() (MPal
soo
400
300
a
Ol
Ol a
200 co
Ol
co
C1)
a
"
0"
Ol 'I
co 0I C
Ol
100 Ol I I
~
"
UJ UJ iii iii
00 10 20 30 SO 60 70 80 90 100 110
o
0.8 0.85 - - - - - - -- - - - - - - - - - -
I
0.6 I
I
I
0.4 I
I
I
I
0.2 I
I 2
I ep .10
OL-_.---r--~_,--_.--._--.__.--_r--r_~~
o 20 40 60 1 80 100
Fig. 1.16. Measurement of ductile damage on 99.9% copper at room temperature (after J. Dufailly)
is given in Figure 1.16 for ductile damage of copper induced by large deformations,
where a v is the true stress a v = a( 1 + E)
This technique may be used for any kind of damage as long as the damage is
uniformly distributed in the volume on which the strain is measured which is the
main limitation of the method. If the damage is too greatly localized, as for high
cycle fatigue of metals, for example, another method must be used.
Some other precautions related to non-linearities have to be taken. At the beginning
and at the end of the unloading paths in the plane (a, E) there are small nonlin-
earities, owing to viscous or hardening effects and also due to the experimental
devices as well. It is best to ignore them and to identify E in the range
O.ISF:s,"x < ~ < O.8S F s ax
It is most important to always use the same procedure to evaluate E and the
evolution of E.
For ductile or low cycle fatigue damage in metals, the procedure could be somewhat
disturbed by an early decrease of E at low strain levels or during the first cycles.
This is due to microplasticity related to reversible movements of dislocations, and
to texture development, but not to damage. As this phenomenon is rapidly saturated,

it is easy to consider:
D = 0 for E < ED (the damage strain threshold)
or
D = 0 for N < N * (the number of cycles to stabilization).
For polymers or composites, to avoid the viscous effect of viscoelasticity, the
strain rate during unloadings must be the same for the measurement of E and E.
An example of a damage measurement on a carbon-carbon composite is given in
Figure 1.17.
For concrete it is important to check the "uniform" distribution of cracks, as the

method is no longer valid if a single big crack develops. An example is given in
Figure 1.18.
Ultrasonic waves propagation

Another technique for evaluating damage based on the variation of the elasticity
modulus consists in measuring the speed of ultrasonic waves.
/
0.5
/
II
0.4
II
/
r<)
co
-I'
0.3 X3
/
o
0.2
!!
III~
II
!II
I
0.1
6~
o 2 4
- a'n (MPa)
Fig. 1.17. Evolution of brittle damage in a shear direction of a three-dimensional carbon-carbon com-
posite (after M. Poss)
Concrete
Ultimate stress 40 MPa _X.~. ~. (~X~~ om

Cylindrical specimen: J
. Xz
25c
Equivalent strain 'i = .r;r:;:;r
1 2 = f:1 X f2
t
28~
.
,
'"Q 26
.,
)( 24 \.
0
0.. 22
::iE
20
w'"
18
16
O ~--~----~----~--~--~---~
0.5 1.0 1.5 2.0 2 .5
xl0 3
0.5
Fig. 1.18. Evolution of brittle damage in compression for a concrete (au =40 MPa) (after J. Mazars)
For frequencies higher than 200 kHz the longitudinal wave speed v L and the
transversal wave speed vT in a linear isotropic elastic cylindrical medium are:
2 E 1- v 2 E I
vL = P(I + v)( I - 2v) VT = P2(1 + v) '
where E is the Young's modulus, p the density and v Poisson's ratio. A measure
of the longitudinal wave speed of a damaged material gives
E I-v
p (I + v) (I - 2v)'
where E and p are the actual damaged elastic modulus and density. Poisson's ratio
does not vary with damage if elasticity is isotropic , and damage a scalar.
The damage is calculated by:
If the damage consists mainly of microcracks or if small cavitation is considered,

(PI p) ~ I and:
-2
D ~ 1_ VL
v2 .
L
The method considered here is destructive: indeed, to measure the speed v L or the
time a wave takes to propagate through a certain thickness, one needs to limit that
thickness by two surfaces. If the space distribution of the damage is not uniform,
the thickness must be of an order of magnitude coherent with that of the RYE,
therefore that requires the body to being analyzed to be cut into parts.
The present limitations of this method are related to the size, to be analyzed, which,
for metals, is too small in comparison with ultrasonic transducers size and accuracy
of time measurements. Nevertheless the method gives good results for concrete.
The ranges of frequency to be used are:
1.0 to 50 MHz for metals,

0.1 to 5 MHz for polymers,
0.1 to 5 MHz for woods,
0.1 to I MHz for concrete.
An example is given in Figure 1.19.

This method can be improved so as to be used in situ without causing any destruc-
tion by working with Rayleigh's surface waves and attenuating of the ultrasonic
signal.
1.3.3 Variation of the Microhardness
This is another nondirect measurement based on the second fundamental equation

of the conclusion of Section 1.2.4: the influence of damage on the plasticity-yield
criterion through the kinetic coupling. Assume an equivalent unidimensional state
of stress, it is written
I_ a
I- D
-xl-R-a II
=0
As microhardness is a process of very small indentation, this method may be

considered as practically nondestructive. The test consists in inserting a diamond
indenter in the material, the hardness H being defined by the mean stress in space:
F
H=a=-
S
01
tot / toto
0.6
0
0.4
,
I
0.2
o 0.5
(1' / (1'r
Fig. 1.19. Evolution of brittle damage from the propagation lime ~t of ultrasonic waves in concrele
(after Y. Berthaud).
The load F on the indenter is chosen so as to obtain a projected indented area S

of the same order of magnitude as that of the RVE.
Theoretical analyses and many experimental results prove a linear relationship be-
tween H and the plasticity threshold (1." k' being the coefficient of proportionality:
H = k'a,.
This threshold corresponds to the actual yield stress,
as = (all + R + X) (I - D).
then
H = k'(all+R+X)(I-D).
In fact, the microhardness test itself increases the strain hardening by an amount
which corresponds. to a plastic strain e~ of the order of 5 to 8%. H is then always
related to (el' + e~), eli being the current plastic strain.
If H* = k'(a y + R + X) is the microhardness of the material which would exist

without'any damage for (Ell+E~), and H = k'(a,,+R+X)(I-D) the actual
micro-hardness for (E I) + E~), then
H is measured, and H* has to be evaluated as proposed below:

- For high cycle fatigue, damage occurs while the stress remains below the yield
stress so that Ep '::::' 0 and
H* may be measured on a nondamaged part of the material.

- For the case of low cycle fatigue, we may consider that the strain hardening is
saturated so that R = Rx and X = X x:
H(*o P +oH)
Jl
= k'(a'I+Rx

+X x ) = const.
H* is obtained by a measurement on the fully strain-hardened but not yet

damaged material.
- For the case of ductile damage, damage and strain hardening occur simultane-
ously and H* has to be obtained by some extrapolation procedure.
Figure 1.20. gives a comparison of results obtained by measuring the evolution
of ductile damage and those obtained by measuring the variation of the elastic
modulus.
Apart from the quasi-non destructive character of this method, it is also a con-
vernient way to evaluate damage fields if the microhardness test is automatized
(Figure 1.21.). Automatization makes it possible to measure the state of damage
of a structure submitted to some accident which, for safety purposes can be in-
troduced as an initial condition in the calculation of the residual strength (see
Section 4.4.1.).
Another possibility of this method of measurement is to deduce the value of the
strain hardening that may exist in some part of the structure after loading.
Let us call Hy the microhardness measured at a point where the material was
certainly loaded below the yield stress a y :
and let H be the microhardness at the point where the strain hardening has to be
evaluated:
H = k'(a y + R+ X)(I - D)
H
or H = --...1t.(a y +R+X)(I - D) ;
ay
H
/
3000
1000
900
800
700
600
SOO
400
300 o'y=320 D=I-.t!.
0.5 H*
200
lOa Fig. 1.20. Evolution of ductile damage by
0.1 microhardness on AIS! 1010 at room tem-
00 10 20 30 40 perature (after R. Billardonj
then
or
H* having the same meaning as previously. (R + X) is the sum of the isotropic

strain-hardening stress and the kinematic back stress of an uniaxial stress equivalent
to the three-dimensional state of stress produced by the indentation. If no kinematic
hardening is assumed, R may be introduced easily as an initial value, for example,
in a three-dimensional analysis as explained in Section 4.4.1.
1.3.4 Other Methods
Several other methods, which are based upon the influence of damage on some
physical or mechanically measurable properties, may also be used.
0.3
0.2
0.1
Fig. 1.21. Damage field on a plastic deformed built notched specimen as obtained from microhardness
measurements on A)S) 1010 (after J. Dufailly, R. Billardon)
Variation of density
In the case of pure ductile damage, the defects are cavities which can be assumed
to be roughly spherical; this means that the volume increases with damage. The
corresponding decrease of density is measurable with apparatuses based on the
Archimedean principle.
If (p - p) / p is the relative variation of density between the damaged state p and
the initial nondamaged state p, it easy, by means of micromechanics, considering a
spherical cavity of radius l' in a spherical R.V.E.of initial radius R and mass m, to
derive the following relationships between the surface damage D and the variation
of density or porosity, assuming no residual microstress:
m
p = 4 '
-7rRJ
3
m
p-p RJ _rJ
- I
p RJ + rJ R' + 1'3 '
bS D 7r1'2 ( rJ ) 2/3
D
bS 7r(RJ + rJ)2fJ RJ + rJ
D ( 1--p)2/3
p
An example is given in Figure 1.22.
Variation of electrical resistance (potential drop)
The effective intensity of the electrical current can be defined in the same way as
the effective stress was defined, using the surface definition of damage:
--: 1
1 = I-D'
I being the intensity which actually exists in the cohesive parts of a damaged
volume element.
Given the potential difference V, Ohm's law for a non damaged element of length
I, area s and resistivity r is written as
V = r-l;
s
0.15
~ 0.10
a.
-.
a.
"'? 0.05
" " - - - ' - - - - - ' - - - - - - ' - - - - - _ e eq

o 0.5 1.0
Q
14
12
.
M
52 10 c
0 8
6
4
2
eeq
0 0.5 1.0
Fig. 1.22. Evauation of ductile damage on steel for different strain hardenings (after F. Moussy)
whereas, for a damaged element of the same size,
- l -
V = f-i,
8
where f is the resistivity affected by the damage by means of the change of volume
only (p - p)1 p (p being the density). Bridgman's law gives f as:
f = r (I +KP~P) = r(1 +KD3 /2) .
K is a coefficient which is approximately 2 for metals.

If the same intensity i is considered for the nondamaged and the damaged elements,
the damage D may be derived from the two expressions
-~
V r 8
- -
f
V
81-D
D = I- ~~ with ~ = 1+ K D3 /2.
Vr r
For small values of D, the correction term (fir) due to the volume change is close
to I (for instance, D c::: 0.1 ~ f i r = 1.064); then
V
D"'I--
- V
This method is known as the "potential drop method".
Some comparisons between measurements of damage obtained with this method
and from elasticity modulus variations D = I - EI E or methods described
hereafter are shown in Figure 1.23.
As the damage measurements are made during tension tests, a correction of these
examples is required to account for the length change l( I + f), the area change
8(1 -2VCe -lOp) and the volume change due to elastic strain: 8l(1 -2v)c e . The
agreement between the results obtained by the different methods is sufficient for
the practical accuracy involved.
Variation of the cyclic plasticity response (stress amplitude drop)
The influence of damage on plasticity may be used to measure the low cycle fatigue
damage.
The one-dimensional law of cyclic plasticity at stabilization may be written as a
power relationship between the amplitude of stress 6..(7 and the amplitude of plastic
strain 6..c p (see Section 3.4.1.):
!I
0.2 0.2
!I
0 308 MPa I 0 0' = 240 MPa !I ~
<.>
E
I <.>
!I
I ~it
0.1 ,
~ 0.1 !I I E.2 G>
>
~ / G>
i_ -~
~
0.5 0.5
a N/NR c tltR
0 .2 :t 1% !I 0.2 0' = :t 205 MPa

o o
0.1 0.1
b d
Fig. 1.23. Comparison between damage measurements by variation of the electrical potential and by
a the elasticity modulus change on stress-controlled fatigue of A 316 stainless steel, b the elasticity
modulus change on strain-controlled fatigue of A 316 stainless steel, c the tertiary creep method on
creep damage of IN 100 superalloy at IOOOoC. d the cyclic plasticity response on stress controlled
fatigue of IN 100 superalloy at IOOOC (after G. Cailletaud)
for a non damaged material, and
.1.', ~ [K,(~~ D)1 M
for a damaged material, in application of the principle of strain equivalence, where

Kp and M are material parameters.
Considering a test at constant plastic strain amplitude, if ~a* is the stress amplitude
at stabilization at the end of the cyclic softening or hardening period and before
the beginning of the damage process:
[K,(~~D) 1 M

~O'
D = 1--.
~O' .
Another way to derive this expression is proposed as an exercise in Section 1.3.5.

This method successfully identifies the evolution of damage during low cycle fa-
tigue of metals except when stabilization of the cyclic softening or hardening does
not occur. An example is given in Figure 1.24.
Tertiary creep response

Creep damage occurs in metals loaded at temperatures above approximately one-
third of the melting temperature. To identify creep damage during creep tests
(constant 0') it is convenient to use once more the principle of strain equivalence
applied here to Norton's law of secondary creep (see Section 2.1.3.).
N
. ( 0' )
cp = K '/I
'
K" and N being temperature-dependent material parameters.
12 -
10.23- 10 r-- ilo'*
0
0...
:E 8 cp =It 0.3 'I.
N
S2 6 -
0
<J
4
2
I I I I
0 5 10 15 20
1.0 N'10 2
0.8 -
0.6 I-
0
0.4
'I
0.2 :1_ one mesocrack
I
I I develops
0 5 10 15 20
N .10 2
Fig. 1.24. Evolution of low cycle fatigue damage on AISI 1010 deep drawn steel
Assuming that the damage process begins at the end of secondary creep, during
tertiary creep one may write:
(!:r,
from which one derives
D = 1-
where i; is the minimum creep rate.

4
~ 2
a.
w
e -*
o
t (h)
1.0 ...... ... . . .. .. ............'j

~
. t
0.8
I:
/:
o
0.6
r
(
0.4
I :
I :
0.2 +- one meso crack
~ develops
:1
o 20 30
t (h)
Fig. 1.25. Evaluation of creep damage on IN 100 superalloy at lOOOee (after H. Policella)
This method yields good results which are in accordance with those obtained
by measuring the variation of the elasticity modulus. An example is given in
Figure 1.25.
Acoustic emission
It is a good method for detecting the location of the damaged zone, but the results
remain qualitative as far as the values of the variable D are concerned.
By way of conclusion, Figure 1.26 offers, in the form of a wine selection chart (!),
some advice for choosing the proper method of damage measurement, depending
on the kind of damage involved.
Low High
cycle cycle
Damage Brittle Ductile Creep fatigue fatigue
8S D
Micrography D= * ** ** * *
8S
Density D= ( pr/
1- -
1
** * *
P
Elasticity E ** *** ***

D= I- - ***
modulus E
Ultrasonic {i2
D= I - --.b.
V2
*** ** ** * *
waves L
Cyclic stress /:;.(1

D= I - - * * ** *
amplitude Ll.a*
Tertiary
D = 1-
("*yIN
:: * *** *
creep
Micro-hardness D= 1- -
H ** *** ** *** *
H*
Electrical V
D= 1 - -:::- * ** ** * *
resistance V
three stars *** means "very good" like "Bourgogne Chambertin 1985"
two stars ** means "good" like " Bourgogne Pommard 1983 "
one stars * means "try to see" like " Beaujolais 1989"
no stars means "do not try" like ???
A votre sante I
Fig. 1.26. Quality chart of methods of damage measurement

1.3.5 Exercise on Measurement of Damage by the Stress Amplitude Drop
Consider a cyclic strain-controlled test in tension compression where the drop of

the stress range caused by low cycle fatigue damage after saturation of the strain
hardening is recorded as a function of the number of cycles (Figure 1.27). The
amplitude of strain is such that the maximum stress reaches a value close to the
ultimate stress au at saturation.
From the one-dimensional coupled strain-damage constitutive equations of Sec-
tion 1.2.4, demonstrate that the elastic strain amplitude remains constant:
a
E(I - D)'
a
(I-D) -X -R-ay = O.
When the strain hardening is saturated and the damage still zero,
then aM
I-D
_----ttrr --- ------
o 0.2 0.4 0.6 0.8 1.0
Fig. 1.27. Cyclic strain controlled test on MAR M 509 stainless steel.
0
1.0
0.8 -
0.6
0.4
0.2
Fig. 1.28. Evolution of the damage
0 measured by the stress amplitude drop
0 0.2 1.0 and by the elasticity modulus change
From the previous equation,

D = 1_ O"M ,
0"11
From Figure 1.27b, draw the evolution 0/ D as a/unction o/the numher o/cycles N
and compare in Figure 1.28 with the evolution of D calculated Fom the elasticity
modulus change measured on Figure 1.27a:
Chapter 2
Thermodynamics and Micromechanics of Damage
A classical way to fonnulate in three dimensions the phenomena identified in one

dimension as described for damage in Chapter I is to postulate the existence of
energy potentials from which one can derive the state laws and the kinetic consti-
tutive equations. In the thennodynamics of irreversible processes, two potentials
are introduced and identified within the framework of the "State Kinetic Coupling
theory" (J. Lemaitre, D. Marquis, \988) .
The state potential, written as a function of the state variables, defines the state laws
and the variables associated with the state variables to define the power involved
in each physical process. For the damage variable, an energy damage criterion is
derived from the damaged elasticity potential.
The potential of dissipation written with the associated variables accounts for the
kinetic laws of evolution of the flux dissipative variables. A constitutive equation
for the damage gives the damage rate as a function of its associated variable.
This method is straightforward if the phenomena can be considered as isotropic. If
not, the formulation at the mesoscale has to be guided by micromechanical analysis
of basic mechanisms responsible for the anisotropy. The study of the growth of
microcracks or microcavities restricts the degree of anisotropy to be introduced; the
study of their behavior makes, it possible to take crack closure effects into account
and to analyze the degree of localization of the damage in many cases. It also
justifies the choice of the continuous damage variable and accounts for a particular
analytical expression of the kinetic law of damage evolution.
2.1 Three-Dimensional Analysis of Isotropic Damage
2.1.1 Thermodynamical Variables, State Potential
The method of local state in the thennodynamics of irreversible processes identifies

one state variable per phenomenon to be analyzed. As the damage always occurs
simultaneously with elastic and plastic strains, their classical variables must be
taken into consideration.
Within the hypothesis of small strains and small displacements, the state variables
at the mesoscale are divided as follows:
- Observable variables:
* the total strain tensor of components Eij associated with the Cauchy stress
tensor (j of components (Jij;
* T the temperature associated with the entropy density s .
40 Thermodynamics and Micromechanics of Damage
- Internal variables:
* " the elastic strain tensor of component cij associated with the stress tensor
0';
* p the plastic strain tensor of component cfj. It turns out that its associated
variable is -0' due to the partitioning of strains:
*r the damage accumulated plastic strain, whose associated variable is the

isotropic strain hardening variable R. The von Mises criterion used in the
plastic potential requires that its rate r be related to the damage and be pro-
portional to the accumulated plastic strain rate defined by
._(2. )I/2
P -
p .p
"jCijCi j ,
with the notation p= t and the summation convention of Einstein:

if/fj = L: (ifj)2
In Section 2.1.3 we shall also demonstrate that R measures the increase in
radius of the yield cylinder in the stress space, this is an extension of the
one-dimensional definition of R introduced in Section 1.2.4;
*a or O'ij the back strain tensor, whose associated variable is the back stress
X D or XB. X D represents the kinematic hardening, translation of the center
of the yield surface in the deviatoric space, which makes that tensor a deviator:
tr(X D) = X Be = O. Figure 2.1 gives the three-dimensional extension of
Figure 1.12;
* D the damage variable. As defined in Section 1.2.1, D = 8S D / 8S depends
a priori upon the plane (ii) of the section 8S, but if the damage is considered
Yield surface after

/" isotropic and kinematic
hardenings
Fig. 2.1. Plastic yield locus in the principal stress space (The trisectrice is normal to the page!)
2. 1 Three-Dimensional Analysis of Isotropic Damage 41
to be isotropic, it has the same value in all directions, and the scalar
D = 8S D ~
8S \in
characterizes completely the three-dimensional state of damage in the RYE at
the point considered.
Let us call Y the associated variable of D which will be derived from the
state potential. As D is dimensionless and since the product - Y b is the power
involved in the process of damage, we already see that - Y is a volume energy
density.
Figure 2.2 is the chart of variables involved in the continuous mechanics of thermo-
elasticity, plasticity and damage.
Let us postulate that the state laws are derived from a state potential: a continuous
scalar function, concave with the temperature, convex with the other state variables
and containing the origin. Taking the Helmholtz free energy,
For elastoplasticity or elastoviscoplasticity, the strains act only through their dif-
ference e - eP = ee:
'lj; = 'lj; ([e - el'], T , r, 0 , D) = 'lj; (e' , T , r , 0 , D).

Assuming that the density p is constant, which is an approximation for ductile
damage, the second principle of thermodynamics written as the Clausius-Duhem
inequality gives the state laws (see Section 2.1 .3):
a'lj; a'lj;
(j = P-;:;- and also (j = p~
uee ue
a'lj;
s
State variables Associated

observable I internal variables
e . .. . ..... .. ... {j
T ...... . .. . . .......... . .... .. .. . ....... s

eC 0'
eP ..... . . .. . . . . - {j
r ..... .. . . .. R
9 0 ..xD
D ..... ... . .. .. . .... Y
Fig. 2.2. Chart of thermodynamic variables (1 . Lemaitre 1976)

The associated variables are defined by

a'lj;
R=Par '
XD _ a'lj;
- P00. '
- a'lj;
Y=p-
aD
The analytical expression for 'lj; is chosen in keeping with the known experimental
observations and the micromechanical results through the so called " State Kinetic
Coupling theory":
- the linear isotropic elasticity;
- the state coupling of damage with elastic strain as shown experimentally in
Section 1.2.4, and analytically by B. Budiansky and R. O'Connell (1976);
- the principle of strain equivalence stated in Section 1.2.3 of Chapter 1 together
with the concept of effective stress written for the three-dimensional case as
a
iF = - - or a . = _ '_J_.
0'
1- D 'J 1- D '
since the damage is isotropic, it has the same influence on all the components
of the Cauchy stress tensor. For anisotropic damage, the variable is no longer a
scalar and the effective stress needs the introduction of a more complex operator
(see Section 2.2). The difference of the behavior in tension and compression is
not taken into account here either, it will be, however, in Section 2.3.4.
- no state coupling neither between plasticity and elasticity nor between damage
and plasticity,
- classical expressions for isotropic hardening and kinematic hardening giving
saturation of hardening for large plastic strain (see Section 2.1.4),
- isothermal processes:
'lj; = ~ [~aijkl<jEkl(1 - D) + Rx:, [r+ ~ exp( -br)] + x~')' (};i/ J: i j ] .
The law of elasticity coupled with damage is
a ij
a'lj; =
= P~ e (
a 'i j klEkl 1-
D) ,
UE ij
where a is the fourth order elasticity stiffness tensor or, by inversion for the
isotropic case,
e _ I +v ~ _ ~ akk
Eij - E 1 - D E I _ D bij ,
E being Young's modulus, v Poisson's ratio and bij the Kronecker delta. Values
for E and v are given for some materials in Section 3.5.
The isotropic strain hardening scalar stress is expressed as:
R = pa'lj; = Rx [i -exp(-br )].

or
R oo and b are two parameters which characterize the isotropic strain hardening for
each material (see Section 5).
The kinematic strain hardening tensorial stress is expressed as
D a'lj; 2
X ij = p~ = -3 X -x/yo:,j.
uO:' J
Xoo and I are two parameters of nonlinear kinematic hardening written with the
factor 2/3 to ensure a simple expression in one dimension. They must be identified
for each material (see also Section 3.5).
The associated variable for D is defined as
Y a'lj; I ~~
= p aD = -iaijklCij Ckl
In order to work with a positive quantity, let us write
Y = -Yo
Let us now find the relationship between Y and the elastic strain energy density we.
By definition,
dW e = (Jijd<j.
Integrating with the law of elasticity, and assuming no variation of damage, that
is, D = const, yields.
we J aijklckl(l- D) dC:j = ~aijklc7jCA:l( I - D).

This shows that
IY ~I
Y is also equal to one half the variation of the strain energy density corresponding
to a variation of damage at constant stress: d(J ij = 0 (J.L. Chaboche 1976).
Starting with the law of elasticity,
d(Jij = aij hl [( 1 - D)dck:l - ck:ldDj = 0,
d ee dD
or ci.:l = ch:[ 1_ D '
together with
dweliT=const
or dW e I
dD iT=const
and from the definition of Y = - y
y = ~ dW e I
2 dD a=const
This last equation allows Y to be called the strain energy density release rate.
This is the energy released by loss of stiffness of the RYE in which the damage
occurred.
44 Thennodynamics and Micromechanics of Damage
2.1.2 Damage Equivalent Stress Criterion
The strain energy density release rate, by definition of an associated variable in

thennodynamics, is the principal variable which governs the phenomenon of dam-
age. It is interesting to give it the meaning of an equivalent stress like the von
Mises equivalent stress for plasticity.
We start with
Y=~
I-D
Let us split we into two parts: the shear energy and the hydrostatic energy cor-
responding to the deviatoric part and the spherical part of the stress and strain
tensors, respectively:
lJij = IJE + IJ H 8ij '

C~j = cW + CI-Aj
IJ Hand c'H are the hydrostatic stress and strain: IJ H
We = JlJijdc~j J = IJEdcg + 8i Aj J IJHdc'H.
From the law of linear isotropic elasticity coupled with damage,
1-211 ~
E I-D
It is interesting to introduce the von Mises equivalent stress defined as
which leads Y to be equal to
2
Y =
We
I-D = 2
lJ e q
2E(I-D)2 ["3(1+11)+3(1-211) ( lJIJHeq ) 2] .
H
() e q
is the triaxiality ratio, which plays a very important role in the rupture of
materials, the measured ductility at fracture decreases as it increases. Remember

what is known from practice: "High triaxiality makes materials brittle!"
Let us introduce
Rv = ~(I
3
+11)+3(1-211) H)2 (IJ
lJ eq
as the triaxiality function; then:

(j2
Y = ~R
2E //
The damage equivalent stress is now defined, similarly to the plastic equivalent
stress, as the one-dimensional stress a* which, that for the same value of the
damage, yields the same value of the elastic strain energy density as that of a
three-dimensional state.
For the one-dimensional case:
2
~l '
-a* 0 0
0 3
[ a' I * I *
[a] = ~ 0 aH 3 a , [aD] 0 --a 0
0 3
0 0 I *
--a
3
::' :
ap'l
~' }
3
W
e 2E(1 - D) '
Rv = I
Writing the equality wf'(a*) = wp (O') = Y(I - D) ,
a;",R,/
2E(1 - D) 2E(1 - D)
defines the damage equivalent stress as
Ira-*-=-a-p-q-R-~/-2-'1
This differs from the von Mises equivalent stress by the triaxiality function, which
is in accordance with the physics described in Section 1.1: plasticity is mainly due
to slips which do not depend upon the hydrostatic stress; damage is debonding
influenced by the hydrostatic stress or the triaxiality ratio. This phenomenon is
influenced by Poisson's ratio, which governs the elastic volume change; the factor
RI/ increases in most cases when the Poisson 's ratio decreases.
As an example, let us consider the damage equivalent stress for a case of plane
stress:
[~
Figure 2.3 represents this expression in the plane (~, ( 2 ) for different values of
a* a *
Poisson's ratio.
2.1.3 Potential of Dissipation
Having defined all the state and associated variables, a second potential will give
the kinetic constitutive relations to describe the evolution of the phenomena.
First of all, we must write the second principle of thermodynamics to ensure the
validity of the model chosen. Let us start from the Clausius-Duhem inequality,
. . Ti
a i/: ij - p( 'Ij! + sT) - qi T : : 0,
where if is the heat flux vector associated with the temperature gradient
grad T (= 8x =8T
i
T .)
"
for non isothermal processes.

The free energy is a function of all the state variables. Let us write its rate as
. 8'1j! .e 8'1j! . 8'1j! . 8'1j! . 8'1j! .
'Ij! = 8E; e E ij+ 8T T + 8r r + 8a aij + 8D D .
'J 'J
Together with iij = i ij + ifj it becomes
8'1j! ) .e
( a ij - P8E;'fj Ci j -
( 8'1j! )
P s + 8T
.p -
t + a ijCij 8'1j! . 8'1j! .
P 8r r - P8a ij a ij
8'1j! . Ti
- p - D - q """:'" > 0
8D ' T -
11=0
0.25
0.28
0.32
Fig. 2.3. Representation of the damage criterion
0.5 in the case of plane stress (after D.Baptiste)
or, with the definition of the associated variables,
aii~j - Ri' - X ij
D
aij -
-'
YD - q.i;T 2: 0.
To always satisfy this inequality of a positive dissipation and particularly for an

isothermal process where the plastic dissipation is negligible, we must require that
-yiJ 2: 0.
As - Y is a positive quadratic function, - Y 2: 0, the damage rate iJ must be

a non-negative function. The variable damage introduced can account only for
deterioration of materials and not for recovery of strength.
Returning to the dissipation inequality, we see that it is the sum of the products of
rate or flux variables (with the negative sign) multiplied by their dual variables as
shown by the table in Figure 2.4.
Let us postulate that the kinetic laws are derived from a potential of dissipation, a
scalar continuous and convex function F, of the dual variables, the state variables
having the possibility to act as parameters.
For the isothermal case:
F(u , R ,X D , Y; (ge, r , a , D)).

'-.,-' ~
variables parameters
The laws of plasticity or viscoplasticity coupled with damage are derived from this
potential by means of a scalar multiplier which is always positive. This ensures the
normality condition of yielding for plasticity, which does not depend upon time
explicitly, and the viscosity property for the viscoplasticity. Within the so-called
"normality rule of generalized standard materials,"
of
-A
ou
of
-oRA (Visco) Plasticity constitutive equations;
of .
- oX DA
Flux variables Dual variables

ell. . .. . ... . . .. . ..... . ... 0'
- r .......... . .... . . ... . . .. R

- 0 . . . . . . . . . . .. . . . . . . .. ... XD
- b .. ..... .. . ... ........ - y

fj. .. . .. . . . ........... --r
gradT
Fig. 2.4. Table of dissipative variables

of /\\
D = - -= ---7
.. I
Kmetlc f
aw 0 damage '
evolutIOn.
oY
In order to define the yield condition, we need a loading function which generalizes
in three-dimensions the one-dimensional plasticity criterion in tension introduced
in Section 1.2.4, namely,
I-1-I J
D
-XI-R-IJ
y
=0
Let us work with the von Mises criterion which states that the (visco) plastic strains
are governed by the elastic shear (or deviatoric) energy density, the so-called "J2
theory". The corresponding equivalent stress is
lJ eq = (~1J31J3) 1/2 ,
1
(TD being the stress deviator:1J3 = lJ ij - 3lJkkDij' Together with kinematic hard-
ening, the von Mises criterion is applied to define the size of the yield locus
regardless of the translation X D defined in Section 2.1.1 and shown in Figure 2.1.
Therefore, it acts upon the difference between (TD and XD.
Furthennore, in the presence of damage, the coupling between the damage and the
plastic strain is written in accordance with the principle of strain equivalence (see
Section 1.2.3). The yield criterion is written in the same way as for a non dam aged
material except that the stress is replaced by the effective stress, which, for isotropic
damage is:
_ (T
(T = - - .
I-D
Then, the loading function f is written as
If = (a-D-XD)eq-R-lJ y I
with (a- D _ XD)
eq
= [~( 1J3 _XI)) ( 11J3
2 1- D - D
_XI)) ]1/2
Z] Z]
Remember that XD is a deviator and IJ y is the yield stress of each material which
depends upon the temperature (see Section 3.5).
Plasticity (see Figure 2.5)

Plastic strain occurs only when the state of stress reaches the actual yield stress;
this corresponds to the satisfaction of the yield criterion f = O. The plastic strain
continues to grow if the yield criterion is continuously satisfied, that is, if j = O.
Then plasticity deals with these two conditions, which define loading, or unloading
with f :::; 0:
f = 0
gP =I 0 if { . and
f = 0
2. 1 Three-Dimensional Analysis of Isotropic Damage 49
__ . _ ___ . _ _. ____..--f:
.. 0 ...........
. /Ioading
/1<0 f: 0
unloading "-- f < 0 /
/
/
o Fig. 2.5. Loading-unloading condition in one di-
mension for plasticity and damage
The expression of the multiplier ,\ is deduced from these two conditions f 0

and j = O.
Viscoplasticity
The plastic strain becomes a time-dependent phenomenon as the temperature in-
creases. For metals, this plastic viscosity is important if the loading occurs at a
temperature above about a third of the melting temper~ture T m:
I 0
T 2 3Tm =K.
For the uniaxial state of stress without any damage, Norton's law expresses the
relation between the steady state plastic strain rate and the viscous stress:
(1v
-- K vC pI IN ,
where K" and N are two material parameters. This relation is usually applied for
low plastic strain rates and does not account for the saturation of viscosity where
the strain rate is very high. Then for damage mechanics dealing with conditions
close to failure , it is more appropriate to work with an expression like:
(1" = K oo [I - exp ( - ~ )] or i p = Ln (I _; : )- n ,
where K ex:, and n are two material parameters (see Section 3.5)
(i P --+ 00) => ((1" --+ K ex:, = const).
The stress a is now the sum of the yield stress, the strain hardening stresses Rand
X as for plasticity, and a viscous stress (1,, :
a = (1y + R + X + (1v'
which means that the loading function f may be positive:
f = la-XI-R-(1y = a" > O.
Referring back to the multiplier ~ of the three-dimensional case and using the von
Mises function ,
ifj of ~ , f
oa = [(u D -X D )eq-R-(1y] > o for visco plasticity;
ij
by extension of the uniaxial case, ). is taken such that (see also Section 2.1.4)
.
p = 1 - D = Ln
~ (f )-n
1 - K 00
Had using Norton's law, we would have obtained
2.1.4 Strain-Damage Coupled Constitutive Equations
An important problem in the mechanics of materials is the choice of an analytical

expression for the two potentials and particularly for the potential of dissipation.
Thermodynamics provides the general framework and some restrictions on the
functions that can be used, but only experiments and micromechanics can give the
details. As the constitutive equations must be general and valid for all materials
(or almost!), we have to consider the general trends of basic experiments and
micromechanics and let only the value of a few parameters to be characteristic of
each material.
Let us look carefully at the results of a schematic experimental test in tension with
some unloadings and compressive loadings, as shown in Figure 2.6.
I) As already mentioned in Section 1.2.4, below a certain value of the plastic strain,
a threshold cpD' no damage occurs. This allows us to identify the (visco) plastic
constitutive equations in their (large) domain of validity without considering any
coupling with damage.
2) The kinematic back stress X , defined as the locus of the center of the elastic
domain,
- increases with the plastic strain;
cr
-;h;--
tR ,
cry /
cry i E<E
E ,_-
I
X
-~--
/
I
I
\
\
I
0
I 0
1 1 I
1
I , I
-~
I
..J
~ ,-"
0 Fig. 2.6. General trends of kinematic and

cPD isotropic hardenings
- is nonlinear with the plastic strain;

- tends to saturate to some value Xoo '
This allows us to choose for the evolution of X an expression like
X = Xoo [1-exp(-rE)],
which is more or less justified by physical and micromechanical considerations.
Xoo and r are parameters to be identified for each material and each temperature
(see table Section 3.5). \
3) The isotropic hardening stress R, identified as
R=a-a y -X
(where the yield stress a y is a constant),
- increases with the plastic strain;
- is nonlinear with the plastic strain;
- tends to also saturate to some value Roo '
Let us take for the same (good!) reasons as for X,
R = Roo [l - exp( -bEp)],
as already written in Section 2.1 .1 Roo and b are also parameters to be identified
for each material and temperature.
Those equations, valid for one-dimensional tension, serve as a guide to writing the
three-dimensional potential of dissipation where the kinetic coupling acts only by
the effective stress deviator aD /(1 - D). In the framework of the "State Kinetic
Coupling theory", it corresponds to:
3
F = ( a- D -X D) eq-R-ay + 4X
D D
X ij X ij +FD (Y:(r.D))
00
The first three terms correspond to the loading function f. Working with a potential
equal to the loading function F = f gives rise to the so-called "associate plasticity
theories." The fourth term _3_
4Xoo
XB XB is responsible for the nonlinear kinematic
hardening, including the factor 3/4 to ensure a simple expression in one dimension.
The last term F D is the part of the potential from which the kinetic law of damage
evolution is derived. Experiments show that it does not depend explicitly upon a,
X or R:
iJ = _ BF ). = BFD ).
BY BY
The analytical expression for F D will be discussed in detail in Chapter 3.
Now let us derive the general constitutive equations valid for plasticity and vis-
coplasticity:
.p
Eij
I
3 (a&
2.. 2 I=D - X i j I=D -
D) I oX& .
a;;:;
2. (~_XD)
I-D
A,
eq
.p
3 ((jlJ-XlJ)
'J IJ
~
Cij = 2. (iYD - XD) eq 1- D

An interesting variable in plasticity is the accumulated plastic strain p, defined by
its rate p:
p = (~ififj )'/2 ,
which is, for the one-dimensional case, simply the absolute value of the strain rate
ip:
p= lipl
Let us calculate p from the expression of ifj:
(
2 ) 1/2 fi 3 ~
P= :3ififj = V:3 2. ( (jD )
I_D-XD (I-D)
eq
a& - XiD)
. [( I=D j
(I=D
a& - XiD)]1
j
/2,
which, together with the definition of
( ~_XD)
I-D eq
= [ ~2 ( I afJ- D -XlJ) ( I afJ- D -XlJ)] 1/2
'J IJ
yields
~
p= I-D
The constitutive equation for the isotropic strain hardening is
of .
r= - ORA = A.
This shows that
Ir = ~ = p( I - D) I
When no damage occurs, the state variable for isotropic hardening is the accumu-
lated plastic strain,
D = 0 -+ r = p.
For incremental calculations it is more convenient to work with a constitutive
equation for the associated variable R. From the state potential,
R = Roo [l-exp(-br)]
or R = R oobexp( -br)r
or R= b(Roo - R)i
The last equation accounts for kinematic hardening:
. 8F . (8((j-D - XD) eq 3 D).

(Xij = - 8XIJ A = - 8X D + 2X X ij A.
D D 00
The same calculation as for : : gives

ZJ
(1E _ XD)
( I-D 8( -Xm
3 ZJ 8XIJ 3
XDij A,
__ -XD
2-'--(:--cr""""D==-----'-:--)--'-- + 2X
(X . ZJ
ZJ
00
I-D eq
_
(xij - Cij I -
.p ( D) 3
- 2Xoo ij A ,
xD ;
or, with the definition of X D from the state potential in order to have a constitutive
equation for this associated variable,
D 2 D 2 .
X ZJ = 3X oo l'(Xij or X ij = 3Xoo l'(Xij'
XE = I' [~Xooifj(1 - D) - Xi~~] .

The last calculation to be performed is to find the multiplier A. (ouff!!)
Plasticity (f = 0)
The plastic multiplier is derived from the consistancy condition

f = o.
Let us start from
since (1 y is a material constant.

8f and 8 f D have already been calculated to determine ifj and (xij:
8(1ZJ 8XZJ
8f
8(~-XD)
I-D eq
3 (~I-D
-XIJ)
ZJ
- X D)
2 ( -cr-D
8(1ZJ I - D'
\-D eq
aD)
aj
a(~_XD)
I-D 3
(t=D- ij D
X ij
eq
aX8 -2(~_XD) ,
I-D eq
aj
aD
a(5-
aD
x t = ~ (~[ a8
aD 2 1- D
-XIJ ] [a8
'J 1- D
-XIJ]) 'J
1/2
3 a ijD ) a ij D
-2
12
( ---x
1- D
D
ij (I - D)2
- 2 ( ___
uD XD )
I-D eq
aj = -I
aR .
Then:
. 3
( 1~jD ) ( .
D - X8 a a ijDD )
XD)
ij . D .
j = 2 (~ _ 1- D - X ij + (I _ D)2 - R = o.
I-D eq
Replacing X8, iJ and R by
D
X ij =
( ~_XD)
I-D
(I-D) .
~q
D)
( I=DIJij D
- Xij
. XOO-Y -2(~_XD)
3
8Y
8FD
( D D
IJ ij )
-yX ij +(I_D)2 +b(Roo -R)
I-D eq
or, in a more compact form with iT

I-D
D
The damage influences the plasticity by the effective deviatoric stress afl ~
J-D
and by the partial derivative of the damage potential 8:-:.
Viscoplasticity (f > 0) (A. Benallal 1985)
This case is much simpler, as the loading function gives the viscosity effect by the
viscosity law:
p= I ~D = Ln ( I - /xJ -n
A = Ln [ 1- (iT -X )e q. - R - I J y
D D ]-n
J-D Kx
Let us summarize the complete set of constitutive equations of the elasto-(visco)-

plasticity coupled with damage: (The case of perfectly plastic materials is treated
in Section 3.1.4)
eij = eij +efj'

e I+v_ v_
eij = T(Tij - E(Tkkbij ,
-D XD .
.p 3 (Tij - ij A
eij - 2 (iT D - XD)eq 1 - D'
R = b(Roo - R)~,
. D _ "I
Xij -
[23" X .p (
ooeij 1 - D) - XDJ
ij A ,
. BFD
D = BY A,
(T ..
- tJ
with (Tij = 1 - D'
[23 ( 1 - D
D -
) (D
(Tij
1_ D -
. D
Xij
(Tij D
X ij
)]' /2 ,
D 1
(Tij = (Tij - 3"(Tkk 8ij'
f = (iT D - XD) eq - R- (Ty .
Plasticity (f :S 0)
if f = 0 and j = 0
~ ~ ~ (,;fj - Xg) &ij / (I - D) [(<T D - X D)e,[Xoo 0+ b(Roo - R)J-
3 -D D (Tij BFD D
-2((Tij - X ij ) (
-D
1 _ D BY + "IXij ) ]
if f < 0 or j = 0
~ =0
Viscoplasticity (f > 0)
.
A [D
(iT -X D)e -R-(T ]-n
- - = L n \- q y
I-D Koo
Identification of the material parameters

All the characteristic coefficients of the materials may be identified from one-
dimensional experiments in tension compression. Let us write this interesting case
in which damage does not exist, D = 0:
-~J '
0
~ [~ ~l ~ " [~ ~ 1+ [~
0 0 1
--
[a] a 0 lEI -/.I
'p 2
0 0 -/.I 0 0
2 2
- 0 0 - 0 0
3 3
I I
[aD] = a 0 -- 0 [xD] = X 0 - - 0
3 3
I 1
0 0 -- 0 0 --
3 3
From the constitutive equations obtained previously with D = 0 and iJ = 0,
[~(~(a-X)2+~(a-X)2 + ~(a-x?)r/2 = la-XI,
f = la-XI-R-a y ,
C = c e + cP '
a
c e = E'
a-X .
Cp = la _ XI A = Asign(a - X),
R= b(Roo - R)~ ,
2 2 2
-X
3 = -X
3 ool'VE p - -'VXA
3 ' ,
~ = r = p = IEpl.
32
X)a 23(a -
For plasticity, A = --------'=-=-----=-3-=2----'
la - XI [Xx; ,+ b(Roo - R)]- 23(a - X),X
or
For viscoplasticity,
The final equations are:
C = Ce +c P '
a
Ce = E'
R= b(Roo - R) ICpl '
x = /' (X(x/ p- X lipl) ;

10 = if for plasticity
P
Xoo /' + b(Roo - R) - /,X sign(o - X)
i p = Ln [1 -
10 - XI- R -
Koo
0
Y
]-n for viscoplasticity
For monotonic loading (Iipl = ip) the two strain-hardening equations may be
integrated with the initial condition lOp = 0 -+ R = 0 and X = 0:
The complete set of parameters to be identified are:

- E and v for elasticity by classical methods;
- 0 Y the yield stress taken as the stress which produces the first offset of plastic
strain defined on a tensile test as;
- Xoo and /, for the kinematic hardening. The partition of stresses is made as
shown in Figure 2.6, the plot of X as a function of lOp allows us to identify X x
and /, either on a semi-Log diagram or by numerical identification through the
least square method, for example;
- Rand b for the isotropic hardening by the same method.
A schematic example is given in Figure 2.7.
For viscoplasticity two more parameters need to be identified. The most convenient
method is to derive them from a relaxation test at constant strain for a large value
of the plastic strain when the strain hardening is almost saturated (X ~ X x' R ~
R x ) and when the damage has not yet occurred:
o
10 = const -+ i = 0 -+ ip +i p = 0 -+ ip -_.
E'
if = Ln (0 -
1_ Xx - Rx - 0)-n Y
E Kx
If the stress is recorded as a function of the time and if (X x + R x + 0 Y)is known
from the above method for a very low strain rate tension compression test, one may
calculate i p and plot it as a function of 0 ; then n and K x are the values which
2.2 Analysis of Anisotropic Damage 59
Fig. 2.7. Identification of elastoplas-

p p tic parameters
fit best with the experimental points. A systematic method consists in a numerical
identification of both coefficients nand K :xl using the least square method.
2.1.5 Exercise on the Identification of Material Parameters

A result of an experimental test in tension compression is given in Figure 2.8. It
gives the stabilized stress-strain loops after 5 cycles for different amplitudes of
strain, and also the radial strain to determine Poisson ' s ratio.
Please follow carefully the method described in Section 2.1.4 under "Identifica-
tion of the material parameters" in order to derive the values of the elastoplastic
parameters, you will be glad to find the following answers!
Elasticity E :::: 200 000 MPa
v :::: 0.3
Plasticity o-y :::: 260 MPa
X x :::: 200 MPa
"( :::: 2
RCXJ :::: 300 MPa
b :::: I
Check that the relative error on predictions remains within 10% on stresses
2.2 Analysis of Anisotropic Damage

Assuming the damage to be isotropic is realistic in many cases, especially under
conditions of proportional loading when the principal directions of the stresses
remain constant. This assumption allows us to predict the conditions for crack
initiation with an accuracy often admissible in engineering applications where a
prediction of the lifetime with an uncertainty of 10 to 50% is already very helpful.
o'(MPa)
+0'
e r (%)
~
0.05 0.1 e(%)
O~----~----,-.
-3 3 e(%)
0.05
Fig. 2.8. Cyclic test result on AISI 316 L stainless steel at room temperature (after J.L. Chaboche)
Nevertheless all materials under special loading conditions and most brittle mate-
rials develop anisotropic damage for which the damage variable can no longer be
a scalar.
Maximum principal stress damage concept (F. Leckie, D. Hayhurst 1973)

A simple way to model anisotropic damage is to assume that it occurs only in the
plane normal to the maximum principal stress. The damage is then characterized by
its intensity D and by the unit vector of the orientation of the maximum principal
stress rip. It is a vector: fj = Drip'
The effective stress is written in the principal stress coordinates:
[a] = [
0'1
0
0
0'2
0
0
1 if 0'1 < 0'2 < 0'3'
o 0 ~
I-D
2.2.1 Geometrical Definition of a Second-Order Damage Tensor

(S. Murakami 1981)
For a more complex type of anisotropy and for use in nonproportional loading,
it is possible to extend the surface definition of the damage introduced for one
dimension in Section 1.2.1
Recall that if oS represents the surface area of the current configuration with
damage of surface oSD' then oS - oS D = oS (1 - D) represents the nondamaged
area of the strain-equivalent undamaged configuration. Let us develop this idea for
anisotropic damage.
Consider a RYE of a damaged material unloaded in its current configuration and a
sectional area defined by its orientation ii, its surface 8S and its shape (Figure 2.9).
Consider also the strain-equivalent undamaged configuration in the same coordinate
system. The R.Y.E. has changed and particularly the sectional area has been trans-
formed into a section of different orientation ii due to the anisotropy of damage
and Of the smaller area 85 (Figure 2.9). Let us assume that the shape remains the
same; later on we shall see that this hypothesis corresponds to orthotropic damage.
ii8S is transformed into fi8S
By analogy with the one-dimensional case, the second-order tensor which trans-
forms the vector ii8S into the vector fi8S characterizes the damage and is denoted
by (I - D) where 1 is the unit second order tensor
(I - D) . ii8S = fi8S or (Iij - Dij )n j 8S = ni8S
D is the second-order tensor damage variable. It is symmetric and has the three
principal values D l , D 2 , D3 in the principal coordinates (Xl ' X2 , X 3) which cor-
responds to the reduction of the areas of a tetrahedron drawn in the Xl' X 2 , X3
0' 1- D i1
s -
s
one dimension
I
_['1<
e
L:2 x1 L::: x1
3 dimensions
Current damaged Strain equivalent

configuration undamaged configuration
Fig. 2.9. Isotropic damage in one dimension and anisotropic damage in three dimensions.
coordinates. Let PQR be a triangle of area 8S in a tetrahedron OPQR in the

current configuration which transfonns into a triangle ?QR of area 8S in the
strain-equivalent undamaged configuration (Figure 2.10) then we have:
PQR -+ ?Q-R -+ (I . -D.)n8S = n8S'

'J 'J J "
ii = X, OQR -+ OQR -+ (1- D,)8A, = 8A"

ii = X2 ORP -+ OR? -+ (I - D2 )8A2 = 8A 2 ,
ii = X3 OPQ -+ O?Q -+ (I - D3)8A3 = 8A 3.
This property of symmetry with respect to three planes makes the damage or-
thotropic. Nevertheless it covers a wide range of applications.
To define the effective stress, let us first write the stress vector T in the two
configurations. For the one-dimensional case;
TiDS = Ti DS.
Introducing the effective stress tensor ij as
we can write
or with the definition of (I - D),
n j 8S = (Iij - Dij )ni 8S,

a ij n j = ijij (Ikj - Dkj)nk 'Vii.
Then, the effective stress tensor could be defined by

jj = O'(I-D)-' ,
R
.....
n
'"'-'
OA, = (1-0, lOA,
x2
Q
'"'-'
OA) = ( 1- 0) I iiA)
'"'-'
0A 2 = (1-02 10A 2
Fig. 2.10. Principal values of the damage tensor

but this leads to a nonsymmetric tensor. As only the symmetric part accounts for
the constitutive equations of elasticity and plasticity, let us consider
iT = 2I [u(I-D) - 1+(I-D)-I u ].
When the principal directions of stress and damage coincide (see Exercise 2.2.4.),
0"1
I-DI
~
[aJ 0
l-D2
0
0"3
0 0
l-D3
The definition of D as a symmetric second order tensor and of its associated
effective stress are the basis for developing a complete theory of the orthotropic
damage mechanics in the same maner as for the isotropic case. It is only "a bit"
more complicated! In particular, the associate variable also becomes a second-order
tensor Y.
The Helmholtz free energy state potential for linear elasticity coupled with damage
is written as a scalar in a form similar to that in Section 2.1.1:
1/J = ~ U(Iij - Dij)aimklcjmc!.l] + 1/Jp ,

where 1/J p is the plastic part which does not depend upon D.
Then
or
2.2.2 Thermodynamical Definition of a Fourth-Order Damage Tensor

(lL. Chaboche 1978)
Another way to define the damage variable is to use the concept of effective stress
associated with the principle of strain equivalence for linear elasticity.
Remember that for the one-dimensional case the law of elasticity coupled to damage
is
a Ec e ,
or 0" Ec e ,
E
for which (I - D)
E"
Following the same ideas, the three-dimensional law of elasticity of a damaged
RYE may be written, using the principle of strain equivalence, as
It may also be derived from the state potential:

I _ e e
'lj; = 2p a ij k:lcij ckl + 'lj;p,
a 'lj; _ e
a ij = P ace . = aijklCkl'
1)
where a is the elasticity tensor modified by damage. Assuming it has an inverse,

= akli
--I
jaij" ckl
e
Substituting in the previous equation for ai j with a change of dummy indices,

- --I
a ij = a i jklaklrsars'
By definition of the effective stress, (ai jkl . iikl~s) is the operator of damage. It is
a fourth-order tensor, by analogy with the one-dimensional case let us call it:
(/ D ) -,
ijkl -
--I
i jkl = a i jrsarsk l'
where D i jkl are the components of the fourth order tensor damage variable D, and
I ijkl are the components of the fourth-order unit tensor I for symmetrical fourth
order tensors:
D ij kl = I ijkl -
- aijr s ar skl '
-I
or D=I-a : a- 1 ,
where the symbol (:) means the contracted product upon two indices.
The associated variable Y is also a fourth order tensor:
I -
.1.
'f/ = 2p a i jklCij Ckl
e e
+ .'f/p
1.
'
with iiij kl = (li jrs - D ij rs )ars kl = aijkl - D i jrs ar skl ;
'lj; = 21p (lij r s - Dijrs )arsklC~jCkl + 'lj;p'

Yijkl = a 'lj; I ee
or P-;;--D = --2aklrs Cij Crs '
U i jkl
Y
i j kl = "2aklr
l ee
sCtJCr s'
This theory corresponds to the general case of anisotropy. We shall see in Sec-
tion 2.3.3 that micromechanics of a general pattern of microcracks leads to damage
representation as a combination of a scalar, a second-order tensor and a fourth-order
tensor.
2.2.3 Energetic Definition of a Double Scalar Variable (P. Ladeveze, 1983)
For materials such as composites, in which the mechanisms of damage may be

different in tension and shear it is interesting to develop a theory defining two
independent scalar variables which characterize the damage that influences the
elastic shear energy and the elastic hydrostatic energy:
- D s acts on the deviatoric stress components;
- D n acts on the hydrostatic stress.
The damage may be isotropic or anisotropic if the material is not subjected to
damage in certain material directions.
For the general case of anisotropic elasticity, but without any damage, the comple-
mentary energy density is
1S
we = 2" ij klai jakl ,
where S is the compliance fourth-order tensor S = a-I. It may be split into two
parts as in Section 2.1.2 by means of the stress deviator (j D and the hydrostatic
stress a H :
aH = 3akk '
a[)
tJ
= a t J - aHDtJ'
1 D D 1 2
2" S ijk/aij a kl + 2"SijklaHDijDk l'
~~
shear energy hydrostatic energy
When damage occurs in all directions,

1 StJkl D D 1 Sijkl 2
We = 2" (1 - D s ) a 'Jaij + 2" (1 - Dn) aHDijDkl'
This allows us to determine
eD ow;
{
Cij = oa[)
- the law of elasticity, '" t1 ;
e
CH a
uWe
H _OW'
__e OW
a {Y s or Ys = oDe
oDs s
- the two variables associated to D s and Dn' _ OW' ,
_ OW'
__e or Yn = __
Yn e
oDn oDn
the damage dissipation being
YsD s + YnDn '
With this theory it is also possible to consider anisotropic damage for which
D n i= 0 in certain material directions, and D n = 0 in the non damageable
directions. To illustrate this point, let us develop the whole theory for an elementary
ply of a fiber composite made of (0. Allix 1989)
- parallel fibers considered as linearly elastic, anisotropic and nondamageable up
to brittle failure;
- a matrix whose behavior is governed by anisotropic linear elasticity coupled
with damage by progressive cracking in the direction of the fibers;
- an interface which is linearly elastic and damageable in shear and elastic and
brittle in transverse tension perpendicular to the fibers.
Assume plane stress conditions in the ( X l' X 2 ) plane. Take EI as Young's modulus
in the X I fiber direction, E2 as Young's modulus in the transverse x 2 direction, v l2
and v21 as the corresponding Poisson's ratios, and G I2 as the shear modulus. As
the damage does not affect the behavior in the X I direction, must be equal to Dn
in the x 2 direction and equal to Ds in (1,2) shear, then the complementary energy
density must be written as
The plane strain components of the law of elasticity coupled with this particular
damage are derived from this elastic stress potential (0: and (3 being indices taking
the values I and 2):
The result is also obtained by homogenization techniques from the constitutive

equations of the three components of the ply, namely the fibers, the matrix and
the interface; it allows us to identify the damages D" and D , by experiments in
tension (T22 in the x 2 direction and in shear (T12 '
The associated variables needed to model the kinetic laws of evolution of the
damages are:
ow'~
oD,
ow'
1'" --
~
oD"
2.2.4 Exercise on Anisotropic Damage in Proportional Loading
The case of proportional loading occurs quite often in practice. Considered as a

special case, it is of great interest because it allows three-dimensional problems to
be solved by working with scalar quantities only.
Definition: A loading on a particular structure is considered to be proportional
when the stress field may be written as the product of a tensor which depends only
upon the point considered, M, multiplied by a scalar which depends only upon the
time:
Demonstrate that the principal directions oj the stresses remain constant with time
(but may possibly be different at each point):
[a] =
SI
[ 0
0
S2
0
0
1 T(t).
o 0 S3
Demonstrate that the triaxiality ratio remains constant with time:
I
SH = :3 SH :'
sg = Sij-SHbij S~q = (~S8SB)' /2 ,

aH S H function of M only
a~q Seq
Then the triaxiality function Rv also remains constant with time.
- Considering isotropic, linear elasticity coupled with anisotropic damage repre-

sented by the second order tensor Dij of Section 2.2.1, write the law oj elasticity
coupled with damage by means oj the effective stress associated with the prin-
ciple oj strain equivalence,
2I [Sik ( 1- D) kj+
- I (
1-
D) -I S ]
ik kj T(t ) =
E ,.
l+v Cij+( I+v)(I_2v)cu,u;j,
vE e i:
and show that the elastic strain tensor and the strain energy release rate tensor
are also "proportional":
c
C'i j(M.t) = eij(M) T(t ),
y l ee y 2
ij = 2a i lllUCj'ITIEU = ;j(M)' T(t ) .
- Assuming that the damage rate depends linearly on the strain energy density
release rate (see Chapter 3), show that the damaf?e tensor is also' 'proportional"
Jor the initial condition Di j (t = 0) = 0:
D ij rv Y;j -> D ij(M.t) = 6. ;j(f\f) D(t ) ,
where D(t) is a scalar function of time.

- Finally, demonstrate that the damaf?e has the same principal directions as the
stress, allowing us to write the effective stress matrix as (see Section 2.2.1)
al
0 0
I-DI
[0-] 0 ~ 0
I-D2
a]
0 0
I-D]
These results are very important for the calculation of the damage evolution
in all applications satisfying the conditions of proportional loading. The tensor
a has to be calculated only once from the reduced stress S, and the further
evolution depends only upon the scalar D through a straightforward integration

of the kinetic law of damage described in Chapter 3.
2.3 Micromechanics of Damage
Micromechanics consists of deriving the behavior of materials at the mesoscale

from the study of specific mechanisms at the microscale. The mechanisms must
be well defined from physical observations by precise geometries and kinematics.
Their mechanical modeling is performed with elementary, common constitutive
equations for strain, crack growth and fracture known at the meso or macroscale.
The interest comes from the interaction between mechanisms and from the homog-
enization brindging the gap between micro- and mesoscales.
2.3.1 Brittle Isotropic Damage
Microcracks and the damage variable
The main mechanism of brittle damage is the nucleation and growth of microcracks
and their coalescence to initiate a crack at the mesoscale. Let us derive from the
microcrack pattern the relation between the isotropic damage variable D defined
at the mesoscale, and the surface density of microcracks.
Consider for simplicity a R.Y.E. at the mesoscale as a cube of dimension l l l
and define the microscale of cells as the cube d d . d in which a microcrack of
any area Si and any orientation may lie (Figure 2.11). The number of cells is [3 / d 3
although the number of cracks, is n(n ::; l3 / d3 ).
The geometry having been defined, modeling consists here in writing the balance of
the dissipated energy calculated by classical fracture mechanics and the dissipated
energy calculated by continuous damage mechanics.
/ /
/ /
L
MESO SCALE RVE
~
V
/ ~ / Micro scale cell
- .
-~ Id
\ ~
/ - ~ Microcrack area Sj
/
- /
Fig. 2.11. Micro-meso element for brittle damage

2.3 Micromechanics of Damage 69
If G is the strain energy release rate corresponding to a crack of area s, D the

equivalent damage of the cell and Y the strain energy density release rate for one
cracked cell loaded by a given state of stress;
Gs = yDd 3 .
For the n cracked cells of the mesocube :
11 11
~ GiSi
" .="~ Y;Di
'3 d .
I I
Assuming that brittle growth of microcracks occurs at G = G e = cons!., and that
brittle damage occurs at Y = 1";: = cons!.,
". 3'"
n n
Ge ~Si = Ycd ~Di '
Furthermore, when s; =
~
I I
D; = 0, the obvious integration yields

11
"
Ge LSi = Yed3 L D i'
I I
The simplest homogenization consists in defining the damage D at the mesoscale
by the mean value of the damages of all the cells, whether cracked or no!. Their
number is [3/ d 3 :
d3 n
D rLDi
I
G
Y;3 LSi'
11
or D =
e I
The term Gjy)3 may be calculated from the rupture criterion for mesocrack
initiation in one plane or here in a set of cells occupying the flat volume [ .[ . d.
Due to the localization of damage, other microcracks may be neglected.
If one assumes that the mesocrack is initiated when a part k of this flat volume
has microcracks of size d2 , corresponding to D = Dc at the mesoscale, then
n
LSi = kP
I
This allows us to write D as
n
D =
2:>
I 1 Dc
-[2-T '
which shows that D is the surface density of microcracks multiplied by a correction
factor (here, De/ k) as was mentioned for the definition of the damage variable in
Section 1.2.1.
If the simplest fracture criterion is considered,
"
then, the damage is simply
D=
This result was also found by the purely geometrical considerations described in
Section 1.3.1.
By the way, this calculation gives the order of magnitude of a characteristic length
which permits correlation between fracture mechanics and damage mechanics,
namely l, the size of the R.Y.E.
Taking the simple fracture criterion k = I, Dc = I,
Gc
YP
c
[2 ;
then It ~ G,
Y c
I
for most metallic materials from the table in Section 3.5:
light alloys steels and alloys
0.005 <
- Gr: <- 0.05 MPa.m
2 -< Yc <- 10 MPa
2.5 10- 3 :::;l:::; 5 10- 3 m
and for concrete in tension: G c :::::' 3 10- 5 MPa.m, ~: :::::' 1.5 10- 4 MPa: l
2 10- 1 m.
This shows that the size of the physical RYE must be of the order of millimeters for
metals and of the order of decimeters for concrete as pointed out in Section 1.1.3,
where slightly smaller values were involved.
Micromechanics and damage growth
As a specific example, let us derive the kinetic damage evolution law at the
mesoscale which corresponds to fatigue microcrack growth at the microscale of
Figure 2.11.
With Dc = k, the damage was found to be
or, for its rate,

."
I: si
yI .
The surface growth rate of one crack Si may be expressed as a function of the
strain energy release rate of one cell G i by means of the Paris law of fatigue crack
growth.
If, for a two-dimensional problem,
- 2a is the crack length, :~ is the crack increment per cycle,

- N the number of cycles of loading in mode I ,
- and K M the amplitude of the stress intensity factor (with K min = 0), then
6a 1/
6N = CKM ,
where C and 1/ are two material parameters, 1/ ~ 4 for many metaIlic materials.
Paris' relation may be considered as the integration over one cycle of the rate
equation:
a = 1/CK1/-I1<.
Or, for the cell problem with the relations
EI /2
K = (EC)I /2, k = -2-C-1 /2C, si = 2ei ai '
where ei is the width of the crack (i),
si = 1/Ce i E1/ /2C7/ 2- IC i
A relation between C i and Y; may be found through their definitions by the elastic
energy. If We i is the elastic strain energy density and W e; the elastic strain energy
of the elementary cell,
C oWe; y _ oWei y _ y oWei
;= as;' ; - aD;' ; - - ; = - aD; ,
with
o(we j d3 ) dD;
aD; ds;'
s
D ,.......i..
with j - d2 '
C i = Y,d, Gi
= Y;d ,
and . = C .E,I /2d,I /2y,, /2-l y.
S; TJ e l . 1 '1 '
The damage rate is
D = ES j 1/CE,I/2d,, /2 ~ y 'I/2-l y'

[2 = [2 L.. e;; ;.
1
Assuming that all the n cracked cells have the same strain energy density release
rate Y; = Y", the homogenized strain energy density release rate for the R.V .E. is
Y = nY" and also nY" , Y=
Also assuming the same width for all the cracks e; = e,
"?1 1 ?1
"" 2- .
L.. ei Y, Y, = neY"
2-1 .
Y" = en
1_ ?1 ?1_ I
2Y 2 y ,
1
This example shows that the damage rate is an increasing function of the strain
energy density release rate and is quasi-proportiona~ to it for most materials for
which 'T/ '::::: 4. It is also proportional to its rate Y. This fact will be used as
a guideline for the derivation of a general kinetic law of damage evolution in
Chapter 3.
2.3.2 Ductile Isotropic Damage

Microcavities and the damage variable
The main mechanisms of ductile damage are the nucleation, growth and coalescence
of microcavities by large local plastic deformations. As in the previous section, let
us derive a relation between the isotropic damage variable D at the mesoscale and
the density of microvoids.
Let us again consider a RYE at the mesoscale as a cube ll l with n cavity cells
of dimension d3 (Figure 2.12).
On this very simple geometry let us again write the balance of energy calculated
from the growth of cavities and from the damage concept.
According to the Gurson model, the porosity at the mesoscale P is equal to the
1
hydrostatic part of the plastic strain E~ = 3E~k due to the growth of voids. For
the geometrical model under consideration,
nd3
P=P'
where n is the number of cavities.
Writing the equality for the rates at the mesoscale,
p = i~
The total power density dissipated at the mesoscale for the homogenized stress aij
and plastic strain rate ifj is
/ /
/ /
/ /
RVE
~ /
Micro cavity
~/
I ~
'/~ ~ I~ I~ Id
~ /
~, /
/
Fig. 2.12. Micro-meso element for ductile

I
~ -----------> damage
2,3 Micromechanics of Damage 73
which can be split into two parts by means of the deviatoric and the hydrostatic
quantities:
(aB + aHbiJ (i;jD + ijA j )

3 0
ij + aH EH + '
D , pD 'p
or aijE
The first term is the power dissipated in pure plasticity by slips. The second term,
which corresponds to the irreversible change of volume, may be interpreted as the
power dissipated for the increase of material discontinuities in the RYE by the
growth of cavities. It is this part which has to be equal to the damage dissipation,
3a H i~Z3 = Y DZ 3
, 3aH'
or D = yP,
For simplicity, let us assume proportional loading, perfect plasticity,
aH / Y = const and the initial condition P = 0 -+ D = 0; then, integration
yields
3a d 3
D = -1Ln-
Y p'
As for the brittle damage, it is assumed that the mesocrack initiation criterion is
reached when a set of cavities fulfill a flat volume of dimension Z' Z d, the other
cavities in the RYE being neglected because of the the localization phenomenon.
The critical value of the porosity corresponding to D = I is
Z2d d
Pc = r Z'
which allows for the calculation of the term 3a H / Y in the damage equation
3aH d
l=yZ'
Finally, we obtain the equation
d2
D = nr'
which was introduced in Section 1,2.1.
Microcavities and damage growth

The kinetic law of damage evolution may be derived from the above expression
for D:
, d2 dd
D = rn+2n"'j2'
The first term accounts for the increase in the number of cavities (n signifies the
number of cavities per unit time), and the second term accounts for the cavity
growth.
a) Growth by nucleation of the number of cavities:
, d2
D = [in ,
In the Gurson model, the porosity rate is also the sum of two terms accounting
for nucleation and growth. For nucleation the Tvergaard kinetic law is used:
P = Aaeq + BaH'
where A and B are material parameters.
Assuming for simplicity a sudden nucleation of cavities of a fixed size d,
. d3
P = nr'
. .1 1
D = Pd = d (Aaeq+Ba H)
or b ~ ~ff" (A+B!:)
It is convenient to express the damage as a function of the accumulated plastic
strain rate p = (~ififj) 1/2, which is easy to introduce by means of the

plastic tangent modulus E r .
Assuming proportional loading,
p= aeq .
Er'
(JH (JH
and also
a eq (J eq ,
~ET (A+B::) P
b) Growth by enlargement of cavities of a fixed number n:
. dd
D = 2n r
d2
or with D = n(2'
The problem of crack growth has received much attention in the past 20 years. An
essential result for this problem is offered by the McClintock and Rice and Tracey
analysis, which derives the rate of growth of a cylindrical or a spherical cavity
of volume V in a perfectly plastic infinite body as a function of the accumulated
plastic strain rate p and the triaxiality ratio (J H/ (J eq:
V = 0.85Vpexp (~(JH)
2 (Jeq
or, with V = d3 ,
3d2 d = 0.85d 3 pexp (~(JH)

2 (Jeq
,
b = O.57DjJexp (~(JH)
2 (Jeq
.
These two examples of cavity nucleation and growth show that the damage rate is
- proportional to the accumulated plastic strain rate ;
- an increasing function of the triaxiality ratio (J H/ (J eq;
- a function of the state through ET or D.
These properties will also serve as a guideline for the derivation a general kinetic
law of damage evolution in Chapter 3.
2.3.3 Anisotropic Damage (D. Krajcinovic 1981)
A way to modelling anisotropic damage is to consider its influence on the stiffness

or on the compliance of the material at the mesoscale, that is, to generalize the
isotropic elastic strain energy density of a damaged material to the anisotropic case:
I _ e e
we = 2a 'Jkl E'J Ekl'
in which the damage D, here isotropic, is defined as
aijkl = aijkl(l- D).
Analysis of a single crack cell
Consider a cell at the microscale containing a circular crack defined by Euler's an-
e,
gles if; of the orientation of its plane, by its radius a and its area A (Figure 2.13).
The matrix is isotropic linear elastic and the response is perfectly brittle.
X2
e)
"'-\
[rrl'j
Micro Ell x,
radius a
[
- Crack area A
~j
X3
MESO[ Volume V
Fig. 2.13. Single crack cell

Let aM and EM be the stress and elastic strain tensor fields in the cell, decomposed
in
aM = aO +a c ,
EM = EO + EC,
where aO and EO are the stress and elastic strain "far fields", which are constant
in the cell when no crack is involved, and a C and E C are the disturbances in stress
and strain due to the crack. These fields verify the classical equations of continuum
mechanics, assuming a static case and no body forces:
aO .. = 0,
tJ,J

aijnj -- Fi'

Eij
= Soijklakl,
where So is the elastic compliance tensor of the matrix.
The total stress and strain also verify that
E~j + E~j = Sfj kd a k1 + akl),
and the stress disturbance verify that
C . . = 0,
a tJ,}
C
a ij n j -- 0 ,
since aO is considered to be a "far field" for the crack.

The conditions of free surfaces are written on the two surfaces of a crack of area
A as
(afj +afJ nj = 0 on A.
The important role in the loss of stiffness of the cell due to the crack is played
by the crack opening displacement vector b defined by the discontinuity of the
displacement across the opened crack,
bi = [[u~ + udl.
The relation existing between b and the "far field" stress (70 is somewhat difficult
to obtain. Using Eshelby's result for an ellipsoidal inclusion of vanishing thickness
and Mura's computations, the result with respect to the crack coordinates, where
e~ is normal to the crack in its center, is
b: ~ [1- (:) T2 B:k";;a,

where xj < a is measured along the coordinate axis ej and B:
k is a symmetric
matrix of influence coefficients.
This equation represents the ellipsoidal void created by the opening of a circular
plane crack. It allows for the determination of all the stress and strain fields in
the cell. "This involves arduous computations using Eshelby tensor and elliptic
integrals" (D. Krajcinovic). In fact, to characterize the mesoresponse of the cell it
is not necessary to perform these calculations because it is governed by the averages
of the microstress and -strain fields. The local fluctuations are of a second order
nature.
1 1
The total stress averaged over the volume of the unit cell is
, IJ
'J
=~
V
(lJ o
'J
+ IJ'Jc. ) dV = 1J'J0 + ~
V
IJc
'J
dV.
V v
To perform the integration by parts let us write the equation together with
xj, k = Djk as
IJ
'J = IJ'oJ + V1- 1 ,. x J,, k
IJk
C dV
v
or
IJ 'J = 1J0
'J +V ~ [I IJcknk
'J
x dA - 1 IJ"" X
'.' ,.k J. dV]
A v
The equilibrium equations for the stress disturbance require the two integrals to
vanish and, as expected; we obtain
o
lJij = lJ ij
The average of the strain may be deduced from the stress and from the contribution
of the crack-opening displacement:
Eij = V1 I( 0
Eij + Eij C )
dV = 0
Sijkl VI I( 0
IJkl + IJkl
C ) *,
dV + Eij
V V
Eij o llJkl + Eij
= sijk *
The determination of cij first requires the calculation of the average of the crack-
opening displacement over the crack area:
1
A
b; dA ;
then, the transformation from local (e;, e2'eJ ) to global (X I ,X2, x3 ) coordinates by
means of the rotation matrix gij:

b; = gijb j and lJ;j = g i kgjllJkl'
In concise form, the result for tension is
Eij
* = a
3bij kllJkl '
where the fourth order tensor b has the form ,
1T
bijkl = 3Bmn (njgm i + n i gmj)(gnl - f8 2n ndnk'
in which the nonvanishing components of Bare
[Bill
B22 = 8 (1 - v 2 ) [ 22 . 1
B33 1TE(2-v) (2-v)
E and v are Young's modulus and Poisson's ratio, and f is Coulomb's coefficient
of sliding friCtion between the two surfaces of the crack.
The final expression for the homogenized stress strain relation in a single crack
cell loaded in tension is
iknjnl(Jkl
+Oilnjnk(Jkl
+Ojknlni(Jkl
+Ojlnink(Jkl
-2vn i n j n k n l(Jkl
Analysis of an anisotropic distribution of cracks
Consider now a set of cells each containing a crack of different orientation. Let us
calculate the strain energy of the whole.
According to Taylor's hypothesis we consider a reasonably dilute concentration of
microcracks which allows us to neglect the interactions between the cracks. The
above analysis may be used considering a random distribution of orientations and
sizes of the crack (Figure 2.14). The strain energy is simply the superposition of
the contribution of each cell.
For one active crack cell, the strain energy density is calculated from
with Eij taken from the previous expression.

In order to obtain We as a function of the strain, a Fenchel transform is performed:
32 (,\ + 2J-l)J-l a3
-- -
3 3'\ +4J-l V
/' - - , \
a'11
- - / /
-
[
-" ~ Micro e It
I / \ /
'- \ I / -
, / - - /'
Fig. 2.14. Multiple crack cell element

where A. and f.1. are Lame's constants:
A. _ vE E
f.1.=
- (I + v) (I - 2v) , 2(1 + v)"
3
The measure of the damage is here ~, its relation with the D variable used all
along the course is obvious.
Considering a cubic cell of length [ : V = [3 ,
and
V
If we now consider a set of cells with a random distribution of cracks, the calcu-
lation of the mean value of the energy requires us to consider the following terms
as random variables:
a3
vn i nj '
a3
V n i njnkn/.
Calling p ( ~ , ii) the probability density associated with the relative void area of
a3
the cracks V and their orientations ii, let us introduce the Leckie and Onat tensors:
Then, the elastic strain energy density of an anisotropic mesoelement contain-

: ( ~~:) :Iined by the prob,bility density 01 their relative ,i,. and orientation
v
-Wijkl"2cijCkl
By analogy with the isotropic case, this expression shows that the description of
the anisotropic damage is achieved by three variables:
- a scalar W as in the isotropic case ;
- a second-order tensor w ij as may be obtained by a purely geometric analysis
(see Section 2.2.1);
- a fourth-order tensor wijkl as it is obtained by a pure thermodynamical analysis
(see Section 2.2.2).
2.3.4 Microcrack Closure Effect, Unilateral Conditions
Derivation of a crack closure parameter
In Section 1.2.1 the effective stress in tension was defined as:

_ a
a---
- I-D
This is also the effective stress in compression if the microcracks and microcavities
remain open. For certain materials and certain conditions of loading, the defects
may close in compression. This is often the case for very brittle materials. If the
defects close completely, the area which effectively carries the load in compression
is no longer (S - S D)' but S. This means that the effective area is defined by a
unilateral condition :
a+
+ S-SD in tension a+ ----7 + = S(I-D) ----70-+ = - -
I-D'
- S in compression a- ----7 0-- = a- .
In fact, the real defects of complicated shapes do not close completely. The effective
area in compression is such that
S-SD < < S.
Let us write this expression as S - hS D S(1 - D h), where h,
(0 ::; h ::; 1) is a crack closure parameter which depends a priori upon the material
and the loading. Let us determine h from a very simple micromechanical model.
Consider again a cubic R.Y.E. of size lll containing a plane crack of area s
loaded in mode I by a force F with a damaged zone created by the growth of the
crack (Figure 2.15).
The material is assumed to be linearly elastic with a Young's modulus E. It is also
elastic in the damage zone with an elasticity modulus jj; = E(I - Dc)' where Dc
is the critical value of the damage during the initiation and growth of the crack.
/ /
/ /
Crack area s
Damaged zone
Fig. 2.15. Microcrack in quasi-

unilateral condition
Let us write the overall equilibrium equations in the plane of the crack.
- In tension, P = P+ , the area of resistance is ([2 - s):
p+ = J a(M) dS = J E c:tM) dS.
For a rough approximation we may consider the Lin-Taylor hypothesis, which

states that the strain is uniform in the plane of the defect; then
p+ = Ec:+(l2 - s).
The homogenization at the mesoscale consists in writing the mesostress as a+

p+
[2'
s
or with the simple definition of damage D [2 and the principle of strain
equivalence,
I-D
- In compression, P = P - , the damage as a state variable is still D but its effect
is different. The resisting area [2 may be decomposed in two parts: ([2 - s),
for which the elasticity modulus is E; and s, for which the elasticity modulus
is one of the damaged zone with the crack closed E = E(l - DJ. Then the
equilibrium equation is written as
P- = J
l' -s
arM) dS+ J arM) dS
or, with the same Lin-Taylor hypothesis as above,
P- = Ec:-(l2-s)+E(1 - Dc)C: -s,

which gives for the homogenized mesostress:

F-
a- = [2 = Ec-(I - D) + E(I- DJc- D,
a- = Ec-[l- D + (I - DJD]'
a- = Ec-(I - DDc)
The principle of strain equivalence requires that ij = Ec- and
_ a
a - -:-----..,..
-(I-D.Dc)"
This simple example shows that the crack closure parameter h is of the order of
magnitude of Dc' a quantity which depends upon the material and the loading:
Ih ~ Dc I
In practice, h is considered to be a constant, in order to avoid a tremendous
amount of algebra! A value of h = 0.2 often gives results close to the experi-
ments. It is also possible to identify its value from the measurements of elasticity
modulus in tension and in compression on a damaged material as explained in
Section 3.1.6.
Effective stress for the three-dimensional case

To define tension or compression by the sign of a scalar quantity in the one-
dimensional case is easy. In three dimensions, speaking of the sign of a tensor
or a matrix is like speaking of one color of a flag! A norm must be defined in
accordance with the basic features of damage mechanics.
D is a state variable, it has the same value for tension and compression. Then the
only problem remaining is to define the effective stress, which will be done under
the following scheme.
- Decomposition of the stress tensor in a positive part and a negative part, from
the signs of the principal stresses a i :
(x ) = x if x 2 0
or with the Macauley brackets { (x) = 0 if x < 0
[a] = [ (~I ) (~2) ~]_ [ (-~I )

o 0 (a3 ) 0
denoted as a i j = (a i ) - ( -a i j )
or u = (u ) - (-u ).
Using the property (x) . (-x) = 0 it is easy to demonstrate that
((jkk ) - (-(jkk )'

((jkk)2 + (-(jkk )2,
((jij) ((jij) + (-(jij)( -(ji) '
- Partition of the specific enthalpy in two tenns which are functions of ((7) and
(-(7). Instead of deriving the law of elasticity from the state potential specific
free energy 7f; e (Ee) as in Section 2.1.1, it is more convenient here to work with
its dual transfonn: the Gibbs specific enthalpy 7f;; ((7).
For linear elasticity without any damage,
7f;; may be written as a function of ((jij) and (-(ji) :
7f;; = 2;E [( 1+ V) ((jij) ((jij) - V((jkk )2 ] +
+ 2;E [(I + v) ( -(jij)( -(jij) - V(-(jkk?]

When isotropic damage occurs, in accordance with the concept of effective stress,
the unilateral conditions are written by means of (I - D) for the first tenn, and
(I - Dh) for the second tenn:
7f;; = 2pE(: _ D) [(1 + V)((jij) ((jij) - V((jkk)2] +
+ 2PE(/- Dh) [(I + v)(-(jij)(-(jij) - v(-aHl]

From this, the law of elasticity is derived as
07f;*
E7j = P '"
u(j 1.Je,.
with the properties demonstrated by P. Ladeveze (1984):
00 ..
a'J
(-21((ji) ((jij ) ) = ",0 . .
u(j.'J
(~ ((ji~,)(akj )8ij) = (aij ),
o~ (~ ((jkk)2)
'.J
= o~ . (~ ((jk:k)((jH,))
1.J
= ((jkk )8ij ,
I+ v v I +v
E(l - D) (a ij ) - E(l - D) (akk)8 ij - E(I _ Dh) (-(jij)
v
+ E(l - Dh) (-(jkk )8ij
- Application of the principle of strain equivalence. The effective stress a must

verify the general form of the law of elasticity,
e 1+v_ v_
Cij = -yaij - EakkDij
The identification of the two expressions for the elastic strain cfj gives the effective
stress as a function of (a), (-a), D, h and v if the law of elasticity is written in
the form of the following equivalent expression:
1 +v [(aij) + ~((akl)Dkl- (akk))Dij

E 1-D
_ (-ai) + ~((-akl)Dkl- (-akk))Dij ]

1-Dh
_~ ([(a k1 ) + ~((ars)Drs - (a rr ) )Dkl

E 1-D
_ (-akl)+~((-arS)Drs-(-arr))Dkl] )
1 - Dh Dkl Dij
Care should be taken with respect to the order of the operations; for example,
([(arr)DkllDkl)Dij = (a rr )3Dij ,
and also tr(a) :j: (tr(a))
or (akl)b kl :j: (a kk ),
v v
(a ij ) + ~ ((akl)Dkr-<akk) )Dij (-a ij ) + ~(-akl)Dkl - (-akk)D ij
aij 1- D 1- Dh
Let us introduce more convenient notation:
1 .
a H = :3 a kk the hydrostatic tress;
+ _ 1 1
a H - :3 (akl)D kl , aH = :3(-akl)Dk1 ;
a + -_ (
a) + -3v- (a + - (a H)) D .
'J 'J 1 - 2v H 'J'
3v
aij = (-aij ) + 1_2)aH-(-aH))Dij ;
+ -
~-~
1- D 1- Dh
The two terms with the factor ~

1 - 2v
are coupling terms which account for shear
effects on the stiffness of the material. They disappear in the one-dimensional case;
which we consider here as an exercise:
[a[ ~ [~ ooo 0]00

tension 0'+ compression a
I + I .
aH 30' aH = 30'- (negatIve)
I
-a
3
+
at = 0
I
a H = -30' - (positive)
ali - (-a H ) =0
a
I-Dh
The damage mechanics in unilateral conditions may be developed as it has been
done all through this chapter by simply using the new expression for the effective
stress in the potentials.
Damage equivalent stress in unilateral conditions

From the state specific enthalpy potential one can also derive the strain energy
release rate density Y and the damage equivalent stress 0'* .
Starting from 'Ij;; ,
'Ij;; = 2PE(: _ D) [(I + v) (aij)(aij ) - v(akk) 2]
+ 2pE(I I - Dh) (I
[
+ v) (-aij )( -ajj ) -
2
v( -akk ) ] ,
-Y = -P aD
eN; = - 2E(1 1_ D)2 [
(I + v) (aij )(aij ) - v(akk )
2]
- 2(1 ~ Dh)2 [(I + v) (-aij)( -aij ) - v(-akk)2]

The minus sign comes from the fact that, with damage, the specific enthalpy 'Ij;* is
only a partial Legendre-Fenchel transformation of the specific free energy 'Ij;.
Y _ (I + v) [ (aij)(aij ) h(- aij)( -aij ) ] _ ~ [ (akk)2 h (-akk?]
- 2E (I-D)2 + (I-Dh)2 2E (I-D)2 + (I-Dh)2 .
The damage equivalent stress was defined in Section 2.1.2 as the one-dimensional
stress (we add here : in tension), which, for the same value of the damage, gives
the same value of the strain energy density or, here, of the complementary energy
density w; = p'Ij;;.
For a tensile stress 0'*,
2E(1 - D) '
Writing the equality with the expression of w; for the three-dimensional case,
eJ* = {(l + lI)(eJij )(eJ ij ) - lI ( eJkk ) 2 +
I D } 1/2
+ I ~Dh [(I + 1I)(-eJij ) (-a;) -1I(-akk)2]
It is easy to check that

for tension (a+),
_ ( 1- D ) 1/2 _
for compression (CJ -), eJ* - I _ Dh leJ I
2.3.5 Damage Localization and Instability (R. Billardon, I. Doghri,

G. Geymonat, 1989)
Damage, as a softening process, may induce localization of strain, as it is the case

for plastic shear bands in plasticity. This is the beginning of the process of failure
by mesocrack initiation which transforms a volume phenomenon into a surface of
separation.
Consider a RYE loaded by a uniform stress (F inducing a strain e either elastic
or plastic, and a damage D. Let i1 be the displacement. Now let us search for the
conditions of existence of a surface of discontinuity of the strain or of the strain
rate e,as the velocity problem is easier to solve (Figure 2.16).
The constitutive equation of the material is assumed to have been worked out to
have the rate form
aij = H ij kl kl '
where H is the tangent stiffness tensor, a function of the state of stress or elastic
and plastic strains and of damage as a softening process.
Since this is a problem of continuum mechanics, the displacement and the stress
vector are continuous even across the surface (n). With the notation of the discon-
tinuity of a quantity X across a surface,
Fig. 2.16. Damage locali zation in a R.V.E

[llill = 0,
[aijnJ = O.
Using Maxwell's compatibility equation for the general shape of the discontinuity
of a second order tensor gradient, = ~ [grad ii + (grad ii)T], which must be
written as
. I
[cijll = 2 (ginj +nigj ) ,
where is a vector to be determined:
- = 0, no discontinuity;
- i= 0, sufficient condition for the discontinuity to exist.
Let us see for which property this corresponds to H, starting with
o = [ai jnjll = [ai jlln j = [Hi jkl Eklllnj '

H being the stiffness tensor of a continuous material, we may assume that it has
no discontinuity across the surface (n): H+ = H-, then:
[Hijk/ Eklllnj = H ij kl[Ekllln j = H i jkl [~(gknl + n kgl )] nj = O.
Furthermore, H has the minor symmetries,
Hkl
'J = H '--lk
J -' H 'J kl = H JI-
- kl ,
then, H i jklgknl = H i jklnk gl
and H i jkl [~(gknl + nkgl) ] n j = n j H ij klnlgk = o.

n H n. The vector n H n has to be zero
nj H ijklnl is a second order tensor
with a nonzero vector to ensure the existence of the discontinuity of the strain
e
rate together with the continuity of the stress vector rate 0- . n:
(n j H ijklnl )gk = 0, with i= o.
This requires the matrix [n H . n] to be singular; that is,
1det( n . H . n) = 0 I
This is a sufficient condition for the existence of a surface of discontinuity of strain,
which can be used for damage localization if H depends upon D.
If an orientation n can be found in an R.Y.E. such that det(n H n) = 0, then a
mesocrack will be initiated in the plane perpendicular to n.
As an exercise, let us work out the simple tension of the plane sheet problem
depicted in Figure 2.17:
a
[a] = [ 0 0 0 ,
0 0] o
-v *c; o ].
000 o -v* c;
Fig. 2.17. Crack initiation in tension
The constitutive equations of the material do not need to be introduced. Let us only
consider a constant contraction coefficient v* and the general law
(Jij = Hijkl(e , D)i kl

For the one-dimensional problem:
a= H""i-v*i(H1l22+H'133)'
0= H2211i-v*i(H2222+H2233)'
0= H3311i-v*i(H3322+H3333).
The components of the matrix [H] may be calculated analytically from the consti-
tutive equations of each material.
Nevertheless, this particular case may be solved by considering the discontinuities
of the strain rates. The shear strain '2 being zero in the tension problem con-
sidered, the discontinuity of its rate is assumed to be zero also. This corresponds
to a particular mode, the principal one, among several others also possible. From
Maxwell's compatibility equation,
. I
[[E'2 ]] = 2" (g,n2 +n,92) = o.
For the problem in planes (x l ,X2) or (Xllx3)' the components of the unit normal
vector ii to a possible discontinuity line defined by its angle () with the x I axis are
n [:~~l.
g is a vector in the (x" x 2 ) plane; let us call it
9 [~:~~l
Then [e 12] = ~ (-y sin X cos () + ')' sin () cos X) = 0
or sin(x + ()) = 0 --> X + () = k7r with k = 0, I, 2, ... ,

which gives the components of g:
9 [~~i~: ~1( -I )'

We may now write the discontinuities of the normal strain rates:
[rill] = gIn, - ')' sin 2 B . ( _I)k ,
[i 22 ] = g2 n 2 +')' cos 2 B . (-I )~:,
and we know from the strain matrix in tension (Figure 2.17) that E22
Assuming this relation is still valid locally,
This leads to
cos 2 B = v* sm
. 2 B or tan B = ~- .
v*
This important result shows that the orientation of the mesocrack that will be
initiated does not depend upon the entire constitutive equation, but only upon the
contraction coefficient.
Two cases are of practical interest:
* v* = 0.5 corresponds to rigid, perfectly plastic damageable materials like most

metals close to rupture;
B = Arc tan v-rs = 54~ is the classic angle of the Luder's bands which may
be easily observed by experiments;
* v * = v, Poisson's ratio, corresponding to elastic damageable materials:
v = 0.2 is the value for concrete: B = +65.9~ ;
v = 0.3 is the value for cast iron B = +61.3~;
v = 0.4 is the value for plexiglass B = + 57 . 7~.
The stress strain condition for damage localization is deduced from the equilib-
rium condition, namely, no discontinuity upon the stress vector, together with the
constitutive equations:
aij = Hijklikl '

[aijn j ]] = [aij ]nj = 0;
expanded for the first component,
[[a,,] = H"" [i ,,]+ H,I22[i zz] + H" 33 [i 33 ] = O.

The directions x2 and x3 playing identical roles, we have
H"22 = H,m '

Also
then H"" [i ,,]- 2v* HII22 [i Ill] = 0;
when localization occurs: [[i I,ll :f. 0,
which implies = v*.

This localization condition corresponds for this tension problem to the peak of the
stress strain curve:
a- II = HIIIIEII +Hll2zE22+ H I1 33E33'

as HII22 = HII 33 and E33 = 10 22 = -V*E II '
0"11 = (HIIII-2v *HIl22) EII =0,
dO"II = 0.
d C II
This result is a necessary and sufficient condition for damage localization; it does
not give the condition for crack initiation, which may occur later by the rupture
criterion D = Dc'
To summarize, the mesocrack:
- is initiated for D = Dc;
- growths in the direction normal to n defined by det( n. H . n) = 0;
- has a size of the order of magnitude of the RYE, also determined by the corre-
lation between damage mechanics and fracture mechanics, l = ~e .
e
2.3.6 Exercise on the Fiber Bundle System (1 . HuIt 1982, Chrzanowski 1986)
A nice way to understand the mechanical development of damage is to play with

_ a set of parallel tensile fibers as represented in Figure 2.18(a).
There is no interaction between the fibers but they are all subjected to the same
elongation. In fact, the analysis is simplified by considering the continuum model
of Figure 2.18(b).
,!//////// / ////P///,ij-
Cross section
A
~ ~
i~ I ~~.
a
~
h b
Fig. 2.18. Fiber bundle model

Brittle and ductile failures

The fibers (dx in the continuum model) are linearly elastic with a Young's modulus
E(x) and brittle with an ultimate strength aR( x ). They are arranged in such a way
that E ( x )and a R (x )are linear monotonically increasing functions of x:
E(x) = E[l +a(2x-I)]'
aR(x) = O'R[l +b(2x -I) ],
where E and 0' R are the values of E and a R in the middle of the fiber bundle
system (x = 1/ 2) . a and b are characteristic coefficients.
If 0 < b < I and a < b, show that for increasing value of the load F. the first
fiber rupture will occur at x = 0 for F = Fo:
I
Fo = jE( X)EA dx = O'RA 1- b .

I-a
o
As the load increases, consecutive fibers fail, let us call x D the rupture front
defined by
a(D) = EE(D) = aR(D )
D is the relative surface of decohesion ; it is a measure of the damage of the fiber
bundle system.
Determine the following relation between the damage D and the strain E:
from EE =a -+D=EE(I-a)-O'R(I-b)
(x= D) R(x=D) 2(0' Rb - Eca) .
Determine the relation between the load F and the damage D in correspondence
with the three particular cases given in Figure 2.19:
F = 7J A(I-D+aD-aD 2)(I-b+2bD).
R l -a +2aD
The first case is interesting because it shows the possibility of gradual rupture from
F = Fo until of/aD = 0 in a way somewhat similar to ductile damage failure
(but here without any plastic strain).
From of/aD = 0, deduce the following equation from which the smallest root
gives Dc (This is a good exercise for practicing algebra!):
8a 2bD3 + (2a 2 + IOab - 12a2b)D2 + (2a + 4b - 2a 2 - IOab + 6a 2b)D +
+ I - 3b + a2 + 2ab - a2b = O.
Note that Dc is much smaller than I.
Creep rupture
The fibers are now considered to be linearly viscoplastic, with the constitutive
equation:
a( x, t)
i (t) = K N(x ) where K N = K N [I + c(2x -I)] .
2
I+a <b<1
3- 2a+ a 2
F F F
o o
o Dc o o
Progressive Instantaneous brittle rupture
ductile rupture
Fig. 2.19. Motion of the rupture damage front D in the fiber bundle model
The elastic strain is neglected but the ultimate tensile stress is still
aR = aR[1 +b(2x-I)]'
with o< b < I, - 1 ::; c ::; 1 and b > c.
As the creep strain progresses, the volume remains constant.
F F
Determine the true stress A as a function of the engineering stress - where Ao
Ao
is the initial cross-section area,
dL F F
dE = _. LA = L A ---+ - = -expE.
L ' 0 0 A Ao
The load F is constant ; it induces creep and failure of the fibers. Determine
the creep rate E: as a function of time before failure, with the initial condition
t = 0 ---+ E = O.
From the equilibrium equation
J
I
F
a(x ,t)dx ---+E:
A
o
and by integration,
it follows that
Determine the time at which the first fiber breaks at x = O.

The rupture condition is
* if
there is immediate rupture of the fiber upon load application.
F I-b . f' .
* if -A ::; aR-- -> to IS mIte,
o l- c
there is delayed rupture of an incubation time to > O. These two cases are
represented in Figure 2.20.
The failure propagation is governed by two equations:
* failure at x = D: a ( t ,x=D) = aR(x =D )
or i( t)K N [1 +c(2D-l)] = aR[1 +b(2D-I)] ;
J
I
* equilibrium a (x, t ) dx = :
D
Jc\
I
F
or t) K N [I + c(2x - 1) J dx = Aexp C:(t)
()
Deduce the following relation between the damage D( t) and the creep strain i :
2(b- c)(I+ cD)(I-D)-[1+ c(2D-I)j2[I+b(2D-l)]b = F .

[I + c(2D -1)]2 AoaR c:exp c:.
(This is also a good training in algebra!)
After the first fiber is broken, either an instantaneous break of the whole fiber
bundle or gradual failure propagation will ocur, depending upon the relative values
of band c.
Express the damage rate when the first fiber breaks.
D=O { i= ~exp c: }
AoK N
.
-aR I - b
c: = = - - - -
b = (l- c)(I-b) aR .
2(b- c) - (l- c)2((I-b) K N
KN I- c
Note that the damage rate does not depend upon the load.
D D
.E..~o: l-b ...~o: l-b
AD Rl-e AD R l-c
1+c 2 1+e 2
b~ 3-2c+c 2 b> 3-2c+c 2
Immediate Immediate
instantaneous gradual
failure failure
D D
...<0: l-b .E.. < 0: .!.:.P..

AD R l-c AD Rl-c
1+c 2 1+c 2
b~ 3-2c+c2 b>---
3-2c+c 2
Delayed Delayed
instantaneous gradual Fig. 2.20. Creep rupture of the
failure failure fiber bundle model
Determine the conditions of the development of damage:

* instantaneous rupture of the bundle:
iJ = 00;
if b::;
3 - 2c + c2 '
* gradual development of the damage:
iJ finite,
1+ c2
if
b > 3 _ 2c+ c2'
These two cases appear in Figure 2.20.
Courageous readers can also calculate the time to rupture and the critical value
of the damage x = Dc which produces the final instantaneous failure (some
approximations must be introduced)!
Chapter 3
Kinetic Laws of Damage Evolution
The physical nature of damage has been defined in the first chapter. The defini-
tion of the damage variable as the effective surface density of microcracks in a
Representative Volume Element associated with the effective stress concept and
the principle of equivalence has given rise to methods of damage measurement
through changes in elasticity or plasticity. These concepts have been generalized
to the three-dimensional case in the second chapter by means of two potentials.
Accordingly, state coupling occurs between elastic strain and damage, and kinetic
coupling takes place between plastic strain and damage, which allowing us to cal-
culate strains and damage up to failure if the constitutive equations for the damage
are known.
Many models have been proposed in the past; their basic properties are contained
in a unified formulation of the potential of dissipation, which will be studied in
detail in this chapter. After the kinetic law of isotropic damage is derived, it is
applied to several classical cases of loading giving rise to different kinds of dam-
age such as brittle, quasi-brittle, ductile, and low cycle fatigue or high cycle fa-
tigue, with special emphasis on the particular properties of metals, ceramics, poly-
mers, composites and concrete. The common main feature is the proportionality
of the damage rate to the strain energy density release rate and to the accumu-
lated plastic strain rate beyond a plastic strain threshold and up to a critical value
of the damage variable. The accumulated pla~ic strain which governs the dam-
age is defined on the meso R.V.E. or at the microscale when the damage is very
localized. The case of perfect plasticity is of great interest for further develop-
ments.
3.1 Unified Formulation of Damage Laws, (1. Lemaitre 1987)
In Section 2. 1.4, the potential of dissipation was written as
D ijD+FD (-
F = ((7-D -X D) f'q-R-a y + 4X3 XiJX Y; (r,D )) ,
ex:
with no analytical expression for the damage term F D from which the damage rate
b is derived
b = _BF~ = BFD~
BY BY'
with ~ = r= zi(\ - D),
96 Kinetic Laws of Damage Evolution
and
The choice of this function F D is of course the key to representing the damage
evolution. It is a matter of logic through the "State Kinetic Coupling theory", of
physical observations, of experimental data and of micromechanical mechanism
modeling to quantify the effects. Like Hooke 's law for elasticity or the Prandtl-
Reuss law for plasticity, a unique kinetic law of damage evolution may be de-
veloped to model the general trends of all kinds of damage in many kinds of
materials.
3.1.1 General Properties and Formulation

Let us list the main trends and properties of damage as presented in Chapter I.
- Damage is always related to some irreversible strain either at the microlevel
or the mesole~el. This property is taken into account in the damage law by
the multiplier A, which is proportional to the accumulated plastic strain. When
damage occurs,
. BFD .
D = BY p( I - D).
The variable p, which governs the damage evolution, also gives the irreversible
nature of the damage, as p is always positive or zero.
- As the accumulated plastic strain increases from zero, the damage remains equal
to zero during the nucleation of microcracks. This corresponds to the accumu-
lation of micro-stresses, or dislocations in metals, generating microcracks. A
one-dimensional damage threshold related to the plastic strain EpD has already
been introduced in Sections 1.2.4 and 2.1.4. As the equation of damage is gov-
erned by the accumulated plastic strain, and as P = Ep in one-dimensional
monotonic loading, it is logical to introduce a threshold on the variable p, or on
r, as r = p when D = 0:
D = BFDp'(I_D) ifp >

BY - PD;
D= 0 if p < PD'
This allows us to introduce a step function in the potential F D which has to be
a function of the state variable r taken as a parameter:
I if r 2:: PD
Oifr < PD
In monotonic uniaxial loading, PD may be identified with the uniaxial damage
threshold EpD but in fatigue or creep processes, PD is a function of the applied
stress as will be established in Section 3.1.2.
- On the basis of a thermodynamical analysis, the main causal variable for the
damage is the strain energy density release rate Y, as it is the dual variable of
3.1 Unified Fonnulation of Damage Laws, O. Lemaitre 1987) 97
the flux D. Then F D must be a function of Y:

FD = FD(Y .).
- Another important feature of fracture is the influence of the triaxiality ratio
a H , (aH is the hydrostatic stress, a e q is the von Mises equivalent stress). The
a eq
modeling of this effect is contained in the expression of Y by the triaxiality
factor R" (see Section 2.1.2).
y = 2E~;'~;)2 ' R, = ~(I+V)+3(1-2V) ( : : ) '

- In order to choose the proper and simplest expression for FD let us recall the
kinetic damage relations obtained by micromechanics for particular mechanisms
in Section 2.3.
* Brittle damage by fatigue growth of microcracks:
D = 1]cE1) / 2d1) / 2e 1) / 2- 1
[2n1) / 2- 1 Y Y ,
1] being of the order of 4 for most materials, D is proportional to Y . Y:
D"" YY.
Here, no plasticity has been introduced, but it always exists at the microscale at
the crack tips of the microcracks and it is possible at least formally to relate Y
to the accumulated microplastic strain rate ii' through a plasticity constitutive
equation:
a~qRv .
Y = 2E(1 _ D)2 ---+ y,(;'eq
v ), aeq (.,'p ) f- a eq (pr,,).
* Ductile damage by nucleation of microcavities:

D = ~ET (A+B::) p.
* Ductile damage by enlargement of microcavities:
D= 057D [exp (t:) 1p

A general qualitative statement of these three results can be obtained by con-
. aH
sidering D to be proportional to Y (function of - and p):
aeq
or
- Here, as in every constitutive equation, a scale factor like

1]CE1) / 2d1) / 2e
2
l n1) -
/2 1
l
,-
d
ET
or 0.57 seen above must be introduced. Let us take
S(I - D),
where 5 is a material constant. The term (1 - D) is considered here to be

cancelled, with (1 - D) coming from ~ = p(1 - D), because experiments
show a non decreasing damage rate when Y and p are constant:
y2
FD '" 5(1 - D)
- The rupture criterion D = Dc introduced in one dimension in Section 1.2.4,
which reduces to D = D l c in the pure tension reference case, also defines
conditions for mesocrack initiation in three dimensions. Dc will be established
as a function of D l c ' au and the loading 0-* in Section 3.1.3.
- Finally, according to the quantitative properties listed above, the damage poten-
tial is logically written as:
y2
FD(y; (r, D)) 25(1 - D) H (r-PD)'
The factor 1/2 is used here to avoid 2 in the derivation:
. uFD Y.
D = uY'\ = 5(1 _ D)p(1 - D)H(r_PD)
or, with r = p when D = 0,
I'D-'-=-~-p- I,
. H-(-P--P-D-")
with the rupture condition for crack initiation,
D = Dc'
- Three material parameters are introduced to characterize the damage evolution:
* S --t the energy strength of damage;
* p D --t the damage threshold function of the material and the loading (see
Section 3.1.2.)
* Dc --t the critical damage, a function of the material and the loading (see
Section 3.1.3).
The effects of the temperature T are taken into account by the variation of these
coefficients with T and by the accumulated plastic strain rate p, also a function of
the temperature.
Several properties not directly introduced in the formulation will be observed as
consequences:
- the nonlinear accumulation of damage;
- the effect of mean stress in fatigue;
- the nonlinear interaction of different kinds of damage;
- the brittleness of materials as the applied stress or the triaxiality increase.
3.1.2 Stored Energy Damage Threshold

The damage threshold PD ' or cpD in one dimension, corresponds to a nucleation
of microcracks which does not produce any change in the mechanical properties.
3.1 Unified Formulation of Damage Laws, (1. Lemaitre 1987) 99
It is related to the amount of energy which is stored in the material. Experiments

in fatigue have shown that the total plastic strain energy dissipated may reach
tremendous values before failure but the stored energy remains constant at micro-
crack initiation, no matter what the stress is,
This stored energy is the result of microstress concentrations which develop in the
neighborhood of dislocation networks in metals and of inhomogeneities in other ma-
terials, For a unit volume, it is equal to the difference between the total plastic strain
t
energy J(Ji/fjdt and the energy dissipated in heat given by the Clausius-Duhem
o
inequality of the second principle of thermodynamics (see Section 2.1.3). For an
isothermal transformation (if = 0). of a material having kinematic and isotropic

- 7 r = p), the power dissipated is.

strain hardenings and no damage (D =
-+. --
'I' (JijCij P - X ijD a' ij >
.p - R' _ ,
which may be calculated from :
- the potential of dissipation (Section 2.1.4),
F = (uD-XD)eq-R-(Jy+_3_xBxi~;
4Xoo
- its associated normality flow rule,
.p of . of
Cij = o(J . A, P = -oRA A,
'J
- and the yield criterion (Section 2.1.3)
f = (u D - XD) eq - R - (Jy = 0,
oF of D of ) .
= ( (Jij 0(J . + R oR + Xij oXD P 2 0,
'J 'J
3 (J3 - XB D( 3 0"3 - XB 3 D)] .

= [(Jij2(uD_XD) eQ-R+Xij -2(u D-XD)eq +2Xoo Xij p,
= [(U D -X D)e q - R +2;oc XBXB]p ,
= ((Jy + 2;00 XBXB) p.

Then, the stored energy w s as a function of time is
J J
t t
WS ( t ) = (Ji/fj dt - ((Jy + 2~x XB XB ) pdt.

o 0
Some approximations make this formula much easier to deal with:
Neglecting the effect of the kinematic hardening,
3 (J[)
- -' p
J .
2 (J eq ,
t DDt
j ~2 CY CY pdt - j CY ypdt
CYijij
eq
o 0
or with, (~cygcyg) 1/2

p
Ws = j(CYeq -CYy)dp.
o
This fonnula may be used; however, it is still easier to assume a perfectly plastic
material of plastic threshold CY eq 2: CY y' whose stored energy is a function of
the difference between CYeq and the fatigue limit CY j' that is the stress, still in the
conventional elastic range below which no damage may occur. It is identified as
the stress amplitude which corresponds to a very large number of cycles to failure
in a tension-compression fatigue test: NRc::,- 106 to 107 cycles. Some values are
given in the chart in Section 3.5.
Then, as CYeq = const.
Ws = (CY eq - CYj)p.
The characteristic value of that energy corresponding to microcrack initiation
(D = 0, b > 0) is taken as that of the pure tension reference case having a
damage threshold C PD and a plastic threshold CY u' the ultimate stress.
(CY u -CYj)CPD'
Writing the equality of these two energies gives the value of the damage threshold
PD for any kind of loading:
(CY eq -CYj)PD = (CY u -CYj)CPD'
Remember that in this fonnula the material is considered to be perfectly plastic.

Then for a varying loading cyeq(t), a more accurate calculation consists in perfonn-
ing the integration for W s!
3.1.3 Three-Dimensional Rupture Criterion
A one-dimensional rupture criterion was introduced in Section l.2.4 as a 'relation

between the applied stress at rupture and a critical value of the damage depending
upon the material and the loading:
CY
I-Dc
Applied to the pure tension test, which is taken as a reference, D c = Di e'
CY = CYR' the stress to rupture, and CY oo (J"u' the ultimate stress (whose val-
ues for several materials may be found in the table in Section 3.5). We then
3.1 Unified Formulation of Damage Laws, (J. Lemaitre 1987) 101
obtain
In three dimensions, the damage criterion introduced in Section 2, I ,2 allows us to

generalize this one-dimensional relation of rupture,
Let us study the condition of stability of the damage process related to a non
positive evolution of the density of kinetic energy k:
k ::; O.
The first principle of thermodynamics, which gives the balance of energy, is written
as:
e+k = p,,+Q-Yj).
The additional term - Yj) is the power disspated to create new micro free surfaces
acting as variable boundaries; it is characterized for each material by a critical
energy Ye,
e is the internal energy density rate, Assuming the stage of instability to be brittle,
e is the sum of the elastic strain power density we
and the heat rate Q received by
the RYE:
e = +Q, we
p" is the external force power density equal to th~ internal force power density if
the inertia forces are neglected for the condition k approaching zero:
Px = a i / ij = a i / 7j
since only elastic energy is involved in brittle processes.
The equation for the first principle of thermodynamics becomes:
k= a i /7 j -we- Y)J < O.
From the elasticity potential (see Section 2,1.1),
We = "2I aijklc ei /=:kl~ ( 1 - D

)
.
We aijklci/:kl
e 'e (I -
D) -
1
"2aijklCijCk/
e e D'
or, with the law of elasticity:

akl = aijklC ~j(l - D) ,
and, remembering the definition of the strain energy density release rate (Sec-
tion 2.1.1),
Y 1 e e
= "2aijklCijC kl'
k = ai/rj - ak1Eki + Y D- Y"D ::; o.

D always being positive, this yields the stability condition
Y-Yc ::;O
and the instability criterion
Y > Y,.
Physically, this means that an instability may occur if the energy released by loss
of stiffness becomes equal to or greater than the energy needed for the creation of
new damage surfaces of decohesions.
In fact, during this stage of instability prior to failure, the strain energy density
release rate is almost constant. For example, if a perfectly plastic material of plastic
threshold as is submitted to a proportional loading for which Rv = const. (see
Section 2.2.4),
a eq
- - - as = O,
1- D
a;Rv
2E = const.
The instability may occur at any time during the damage growth when some per-
turbation slightly increases the quantity Y. Then this instability condition cannot
simultaneously be the rupture criterion. From both the practical and theoretical
points of view, a physical condition of rupture must be added.
To this end, let us assume that the final stage of damage rupture by atomic decohe-
sion, (that is mesocrack initiation) is governed by the amount of energy dissipated
in damage growth:
jY
Dc
dD = const. at failure .
o
This quantity may be identified from the uniaxial case in tension, still taken as a
reference:
a eq = a R }
Rv = 1 Y
D = Die
and
In three dimensions, the case of perfect plasticity in proportional loading is, for
simplicity, considered as an approximation:
Dc Dc 2 Dc
a;Rv D
j YdD = j 2E(I-D)2
a eqRv dD = ja;R vdD
2E 2E e
0 0 0
a*
as a-* a* a eq R vI / 2 an d a eq = a .. (1 - D).
I-D'
Then the rupture criterion gives the critical value of the damage as a function of its
value for the one-dimensional reference case Di e' the effective damage equivalent
stress ij* and the ultimate stress a" with the limiting value Dc = I:
ij*2 a2
2EfD" = 2f;;D lc '
Dc = Die -:2
a
a
2
: :; 1
3. 1 Unified Fonnulation of Damage Laws, (J. Lemaitre 1987) 103
This formula shows that the critical value of the damage for mesocrack initiation
decreases as the effective damage equivalent stress a* increases either by the
stress a eq' the damage D or by the triaxiality ratio a H / a eq contained in the Rv
expression.
DI e is the critical value of the damage of the case of reference under tension. Any
other one-dimensional case (like fatigue, for example) may have a different value
of De' depending upon the applied effective stress a.
It is interesting to study the different possible cases in the graph ( Dc, a*) of
2
Figure 3.1, where the rupture relation Dc Die ~
'a
:2 corresponds to a line which
defines:
- a safe domain below which a stress-damage path does not produce crack initi-
ation,
- a rupture domain upon which the material is certainly broken.
One-dimensional examples (Figure 3.2)
- A fatigue case at constant amplitude of stress is represented by path I: the

critical value of the damage decreases as the stress amplitude increases.
- A fatigue. case at constant amplitude of strain corresponds to a quasi-constant
amplitude of effective stress as shown by path 2.
- A ductile case in tension is the case of reference represented by path 3 for which
the critical value of the damage is of course Die'
- A creep path at constant stress is represented by path 4. The critical value of
the damage also decreases as the applied stress increases.
Rupture line
Safe
domaine
01e +--- - - - - --4,.,(,/
Fig. 3.1. Rupture criterion at the meso-

o scale
o 0
2
o o
3 4
Fig. 3.2. Damage rupture paths. I) Fatigue !::!.U const; 2) Fatigue !::!.C: const; 3) Ductile case;
4) creep u = cons!.
This mesocrack initiation is completed by the localization criterion giving the ori-
entation of the crack developed in Section 2.3.5.
The entire set of equations governing damage evolution can be summarized as
follows:
b
o if P < PD
Crack initiation if D = De with De
The material parameters are:
- S, cPD and Di e' which must be determined from damage me!lsurements;

- au and a f' which are classical characteristics "easy" to find in handbooks or
to identify by tensile and fatigue tests.
In most cases Y is the strain energy density release rate defined in Section 2.1.1. In
order to take into account the micro-crack closure effect responsible for different
damage behaviors in tension and compression, the only change to be made is in
the definition of Y derived in Section 2.3.4 as a function of the crack closure
parameter h.
3.1 Unified Fonnulation of Damage Laws, (1. Lemaitre 1987) 105
3.1.4 Case of ElasticPerfectly Plastic and Damageable Materials

In Section 3.1.1 we have observed that damage occurs only if the accumulated
plastic strain p (or the plastic strain in tension) has reached a certain threshold PD:
P < PD --+ iJ = 0 or Ep < E P D --+ iJ = O.
For many materials, this threshold is large enough to assume that the strain harden-
ing is close to saturation making the behavior of the materials perfectly plastic for
further straining. This hypothesis corresponds to the saturation of kinematic and
the isotropic hardenings.
P > PD
The plasticity criterion is simply

f=o- eq-0'8=0,
where 0'8 is the threshold into consideration for each application. The choice of its
value for a given material is somewhat subjective because real materials are never
perfectly plastic. But it has to be bounded by 0'Y and O'u:
O'y :::; 0'8 :::; O'u'
Taking 0'8 = O'u ' the maximum value of the stress on a stress strain curve in
tension, where dO' = 0, is often a good choice (Figure 3.3 for example).
dE
Let us write the corresponding kinetic laws for plastic strain and damage taken
from sections 2.1.4 and I. I of the present chapter.
of .
-A if {f = I ~~ - 0'8 =0
OO'ij j = 0 with ~ = p(1 - D),
O';qRv .
ifp > PD with Y = 2E(I_D)21
from the plasticity criterion:
from the consistancy condition:
aeq +0'8D = 0;
. 0'2
then: D = 2IlSRvp,
or, replacing p by its expression in tenns of D for P > PD
.p _ 3ES a iJ
Cij - 71-DR'
s v
This shows that for perfectly plastic behavior when P > PD' the plastic strain is
directly related to the damage.
Another interesting result coming from the consistency condition is
iJ = _ a eq .
as
Notice that the equivalent stress rate is negative, owing to the softening damage
process.
The last interesting result comes from the plasticity criterion written in tenns of
the elastic strain: the law of elasticity coupled to the damage written in tenns of
the deviatoric and the hydrostatic stress and strain is
e eD + cHe fj ij' c eH
I e
Cij = Cij 3' c kk '
I
a i j = aB+aHfjij' aH 3' a kk '
c eD I +v D
I
1-2v
c eH
tJ E(I_D)a ij , E(I_D)a H
It follows that
and
or, with f = 0,
eD e D 1/ 2 _
(Cij Ci j ) - V~3' (I +E)
v \
a 8
_
- const .,
which makes the second invariant of the elastic strain deviator constant for any
state of stress verifying the plasticity criterion.
In tension, where .
a eq
I-D
EC e - a s = 0,
as
Ce = E = const.
This property is shown schematically in Figure 3.3 for a classical tension test and
for a cyclic tension-compression test, strain controlled at constant strain amplitude.
It proves that the measure of the damage by the variation of the elasticity modulus
and by the stress amplitude drop are equivalent.
3.1 Unified Formulation of Damage Laws. O. Lemaitre 1987) 107
The complete set of constitutive equations is
I+ v_ v_
E aij - Eakkti;j '
-D
3 a ij . . _
.p
Eij --P If a eq -a~ = 0,
2 as
2
D= 2isRj) if P 2: PD'
where Rv may be expressed as a function of the strain:

E(I - D) e
I _ 2v EH I
---=---:-=-=---=-:-- with ElI = -3 t T (e - e P ) EH since e P is a
a eq a,(l - D)
deviator;
2
2
Rv = - (I + v) + -3- ( -
EEH )
- .
3 I - 2v as
In order to take into account the micro crack closure effect explained in Sec-
tion 2.3.4, the assumption of perfect unilateral conditions may be made; it corre-
sponds to a zero value of the crack closure parameter:
h = O.
The strain energy density release rate from Section 2.3.4 reduces to
I +v < aij >< aij > v < au: >2
y = 2E (I-D)2 -2E (I-D)2
and, with this expression, the kinetic law of damage remains unchanged:
. y
D = -p if P 2: PD
S
a b
Fig. 3.3. Constant elastic strain. a tension test; b cyclic tension compression test
Viscoplasticity
The case of nonhardening behavior reduces the set of elastic perfect viscoplasticity
constitutive equations from those of Section 2. 104 to
Cij = e
Cij
+ Cij'
p
.p
c1,)
.
2
2 3 ( ECH )
with -(I+v)+-- -_-
3 1- 2v a eq
Damage threshold
au -af
PD = cPD
a eq - af
For perfectly plastic materials this becomes
and
Rupture criterion
For a perfectly plastic material,
0-* = ~RI/2 = a RI/2

I-D v s v
Then
and
D .
== --1. If a == au.
Rv S
3.1 Unified Formulation of Damage Laws, (1. Lemaitre 1987) 109
3.1.5 Identification of the Material Parameters
The detennination of the coefficients 5, [PD' and Di e that characterize the damage,
together with a f and au related to plasticity, must be worked out for each material
and temperature, from experiments perfonned on those cases in which the damage
is easiest to measure. This involves:
- damage as unifonn as possible;

- tensile experiment;
- measurement by means of elasticity change as explained in Section 1.3.2.
This restricts us to monotonic tensile experiment strain imposed at a constant strain

rate, or very low cycle fatigue at a constant amplitude of strain.
Let us assume that a good tensile test has been perfonned with measurement of
the damage during unloading by elasticity change (Figure 3.4):
For the one-dimensional case:
a-2eq R lJ a2
y=-- 2E(I-D)2 as R,/ = I,
2E
p = (~i~i~j) 1/2 = lipl and in monotonic loading p = [p'
. a2
D = 2E5(1 - D)2 Ii p IH(" - PD) .
- [p D is the plastic strain below which there is no appreciable damage (see
Figure 3.4).
- au is the classical ultimate stress as shown in Figure 3.4. It is also interesting
to know the yield stress a y as it may help to make a better choice of the plastic
threshold as when needed.
- a f is the fatigue limit taken, as already mentioned in Section 3.1 .2, as the
stress amplitude corresponding to a number of cycles to failure in a tension
O'c - -- - - - - - - -
dO
Ep
o~----~----------~
Fig. 3.4. Identification of damage coefficients

lID Kinetic Laws of Damage Evolution
compression fatigne test of about 106 to 107 . Usually:

2
30"Y < O"f < O"y'
- The last parameter S is determined from the slope of the curve: damage D
versus the plastic strain cp :
0"2
D- .
- 2ES(1 - D)2c p
or
dD 0"2
dc p 2ES(1 - D)2'
At each point of the curve, D is known, 0" is known from the stress strain curve,
dD / dc p is estimated and E is known from a previous identification:
0"2
S-
- 2E( I - D)2 dD .
dc p
Several points may be considered in order to obtain S as the best average.
The main difficulty involved in this identification lies in obtaining a good stress
strain curve in the softening range where necking occurs. To avoid any instability, a
"soft" machine is required with an excellent feedback system, which can accurately
impose the strain in the damaged region of the specimen. This strain must be
measured locally by a small strain gauge (see Section 1.3).
3.1.6 Exercise on Identification by a Low Cycle Test

A way to avoid, at least partly, the problem of localization by necking is to perform
a tension compression test at constant amplitude of strain imposed at such a value
that the strain-hardening is saturated at the first cycle or after a few cycles.
The general kinetic damage law in one dimension is written as:
-2
iJ = 2~S lipl H(c p -cPD ),
with -
0" = -0"- un d er tension
.
I-D
and I _0"Dh under compression (see Section 2.3.4),
with the condition for crack initiation being,
2
D Ie ~
0-*2
=D Ie'
O"eq
as I and 1- D = O"u
Determination of s, Po' D,., and h from the experimental result shown in

Figure 3.5 for a particular material.
I) Plot the evolution of the elasticity modulus (divided by Young's modulus) in
tension and compression as a function of the number of cycles (Figure 3.6).
3.1 Unified Formulation of Damage Laws, (1. Lemaitre 1987) III
N cycles
0-10
200
iif!~f 2000
2300
2400
2500
2550
Fig. 3.5_ Low cycle fatigue stress strain pattern, for AISI 316 L stainless steel (after J. Dufailly)
From these ('urves. deduce the crack closure coefficient h.
h = 1 - ~- / E = 0.2
\- E+/E '
.tE
1.0 t---~=======::::::~~~=~
_______ E- -
0.8 '\
E'
0.6
0.4
0.2
N
o 2000 2100 2200 2300 2400 2550 (cyc les)
Fig. 3.6. Elasticity modulus evolutions

taken as the mean value for several values of N. Also deduce the damage as a
function of N by
t+ E-]
D = 1-- or D = h
I [ 1- .
E'
2) Calculate the accumulated plastic strain as a function of the number of cycles.

For one cycle i,
P = 2 LAcPi'
i=1
Plot the damage as a function of P (Figure 3.7.).

This graph allows for the determination of PD and Dc:
PD = Max(p(D=O)) :::: 6,
Dc = Max(D(p)) :::: 0.15.
3) Determine the parameter S as the average value of several computations such
as:
S -- 0"7w
dD ' S:::: 7M Pa
2E(I-D)2-
dp
3.2 Brittle Damage of Metals, Ceramics, Composites and Concrete
Damage is generally termed brittle when it occurs by decohesion without any

sensible plastic strain at the mesoscale. Damage mechanics, as developed, always
o
1.0 1 - - - - - - - - - - -_ _- ,
0.8
0.6
0.4
I
0.2 D1c =0.15 ,I
r -- ---~
p
o 2 4 6 8 10
Fig. 3.7. Damage evolution for a low cycle fatigue, on AISI 316 stainless steel
3.2 Brittle Damage of Metals, Ceramics, Composites and Concrete 113
associates damage growth with plastic dissipation: D is proportional to p. For cer-

tain kinds of materials such as ceramics, concrete or high strength quenched steels,
there is no measurable plastic strain at the mesoscale up to failure. Nevertheless,
although p may be considered to be zero at this scale, at the microscale, where
the damage develops as a localized phenomenon, some plastic strain always oc-
cur in these very small damaged volumes: slips at grain boundaries or interfaces,
movements of dislocations in neighborhood of defects in metals.
In cases in which these permanent micro strains may be neglected, the damage is
refered to as "pure brittle"; otherwise, it is called "quasi brittle".
3.2.1 Pure Brittle Damage
As there is no plastic strain (p = 0) when brittle damage is involved, from

the general law of kinetic damage, the damage is also zero. Coming back to the
instability criterion of Section 3.1.3. the pure brittle rupture occurs at the same time
as the instability, that is, for:
(>2
u
2E
or, with D = 0,
y
2E
or
I (> * -- (> eq RIv /2 -- (> u I

This equation shows the strong influence of the triaxiality ratio contained in Rv:
Rv = ~3 (I + /I) + 3(I _ 2/1) ((>

(>eq
H.) 2
The characteristics of the material are expressed by (>..,; this is the stress to rupture
in a one-dimensional tension test.
3.2.2 Quasi-Brittle Damage

When the behavior is brittle at the mesoscale but localized damage growth occurs
at the microscale, we are dealing with quasi-brittle damage.
Consider once again a mesovolume element exhibiting elastic behavior everywhere
except in a small microvolume JL representing a weak defect subjected to elasticity,
plasticity and damage (Figure 3.8).
The matrix is elastic with a yield stress (> y and a fatigue limit (>f' The inclusion has
the same properties as the matrix except that it is perfectly plastic with a plastic
threshold (>~ and a fatigue limit (>/. Its weakness comes from the value of the
plastic threshold, which may be taken equal to the fatigue limit of the material, as
it is the lowest stress giving rise to possible damage:
(>~ = (> f'
Elastic (plastic)
E(o'y' o'u.o'f)
r:t---+ Elastoplastic and damageable

E o'!',o't
S.po.Dc
Fig. 3.8. Two-scale volume element
Furthennore, the weakness also comes from the fatigue limit aj assumed to be
reduced in the same proportion as the plastic threshold:
a~
J1
at = at-
ay
The complete inclusion problem will be solved numerically in Section 4.3. by "lo-
cally coupled analysis". Here, however, some approximations allow us to derive
the rupture conditions without resolution of the complete set of constitutive equa-
tions. Starting with the kinetic damage law for the inclusion,
. yJ1
D - - _ s p' J1 ,
we wish to express yJ1 and pJ1 as functions of macroscopic quantities such as the
elastic strain and the stress (T.
According to the Lin-Taylor hypothesis, we may assume that the inclusion is sub-
jected to the state of strain (or strain rate) of the matrix, which is taken to be
unifonn:
Neglecting the elastic strain e J1 in comparison to the plastic strain PJ1 in the
inclusion allows us to write:
P
'J1 _
- (
32 Cij
'PJ1 'PJ1)
Cij
1/2 _
-
(23 .J1D .J1D
Cij Cij
)
1/ 2
if the yield criterion is satisfied.

Next, what about yll?
The inclusion being perfectly plastic, then, from the yield criterion,
all
eq _ II
l-D -as'
The triaxiality factor may also be expressed as a function of the stress at the
mesoscale:
R~ 2
3(I+v)+3(1-2v) ( ;~J1)2
a~q (I - D)a~;
3.2 Brittle Damage of Metals, Ceramics, Composites and Concrete 115
E( 1 - D) cj/ from the damaged elasticity at the microscale,

1-2v
but cj/ = cHat the mesoscale since g F- Jl + gPJl = g and tT (gPJl) = 0;
cH -ea
1-2v
H from pure elasticity at the mesoscale;
aJl a
then a~ = (I - D)a H and ff
a eq
~.
as
au-a! au-a!
The damage threshold PD = cP D becomes PD = cP D Jl Jl because
a eq - a! as - a!
for the inclusion, a eq = a~ and a! = aj, the fatigue limit remaining a! for the
case of reference.
It is not possible to write the yield criterion exactly in terms of macroscopic
quantities. The following approximation is made:
D= 0 if (J eq < (J J .
Finally, 2aE~S2 [-32 (I + v) + 3(1 - 2v) (:~s )2] E:eq v

if Ceq PD'~
and if (J eq > (J J
with
caution is advised here: E:eq =

2 )
(3E:;/;j 1/ 2
and not dt
d (23 C;jC;j
) 1/ 2
!
The integration of this damage rate equation may be perfonned for the following
initial and final conditions:
Ceq < PD --+ D = 0,
c - co R +- D - D - D __ a u_
2
Ceq - Ceq - " - Ie Jl2 R .
as v
This last condition is, strictly speaking, a crack initiation at the microscale but can
be shown to also correspond to the brittle fracture of the matrix ! The strain energy
release rate at the mesoscale corresponding to a microcrack of surface d 2 is
G = _~ 8WI
2 8A <1=consl.
I
But -28W1<1=consl is also equal to the energy dissipated in the inclusion by the
damaging process:
8A
Assuming a constant strain energy density release rate Y Yr and a critical value
of the damage Dr = I yields
G(DA) = Y,d.
Using the result of Section 2.3.1 in which a simple relation was found between the
critical value of the strain energy density release rate Y" and the dimension of the
crack (l in Chapter 2, d here), and the toughness G e ,

Yed = Ge
shows that G(D = De at the microscale) = Gc (at the mesoscale).
We can conclude that condition of crack initiation at the microscale satisfies also
the brittle crack initiation criterion at the mesoscale.
In the case of constant hydrostatic stress loading, the integration of the damage
rate equation becomes obvious:
Calling R"(~) = ~(I + v) + 3(1 - 2v) (:i!)' = cons!.,

a} (R )
2ESRv(;f) Ceq - PD
or, with the assumption, a~ = a I'

R au-a l 2ESD
Ceq = CP D 2 + a2 R e
al_a l I V(;f)
ay
This result accounts for any kind of constant hydrostatic stress loading: monotonic,
fatigue, etc., with the proper values of cPD and De; remember, however, that Ceq
is the accumulated total strain.
3.2.3 Exercise on the Influence of the Triaxiality on Rupture
In practical problems of engineering the triaxiality ratio varies from 0 (pure shear)
to 5 or 6 at the tips of very sharp notches, a H = ~ being the value for pure
a eq 3
tension.
For pure brittle damage , plot the von Mises equivalent stress at failure divided
by the one-dimensional damage equivalent stress as a function of the triaxiality
ratio. Take Poisson's ratio as a parameter: v = 0.2, v = 0.3, v = 0.4
(Figure 3.9).
- three-dimensional failure: a* = aeqRl,p = au;

- one-dimensional failure: a* = a R = au'
This exercise should convince you that using the von Mises criterion a eq = const.
for failure is like using a hammer to open a bottle of wine!
all
For quasi-brittle damage, plot versus the triaxiality at microscale
a eq
the equiv- ff
alent strain to rupture (minus PD) divided by the one-dimensional strain to rup-
ture (minus C PD) for the three values of Poisson's ratio v = 0.2,0.3,0.4. For
simplicity take De = Die' and a = 3a~ for the one-dimensional case at
the mesoscale which induces a three-dimensional state of stress at the microscale
(Figure 3.10).
0.8
0.6
0.4
0.2
o 1/3 2 3 4 5 6
Fig. 3.9. Stress criterion of brittle failure
3.3 Ductile and Creep Damage of Metals and Polymers
Recalling Section 1.1.4, the damage process is called ductile or creep when it
occurs simultaneously with large permanent strains. At the microscale it mainly
involves the nucleation and growth of cavities causing ductile damage in metals
and polymers; in metals submitted to elevated temperature, it causes creep damage
on the form of grain boundary sliding and decohesion.
3.3.1 Ductile Damage
The general damage law is straightforward in its application to ductile damage:

. Y
D = SPH(p - PD)'
The damage evolution may be calculated for any process of loading defined by
the histories of the accumulated strain rate p( t), the von Mises equivalent stress
(a eq (t)) and the hydrostatic stress a H (t).
Case of proportional loading
This is a particular case of loading which often occurs in structures for which
integration is simple.
R
Ceg- PD =RV(I)
CR-CPD Rv(~
V=O.2
o 2 3 4 5 6
Fig. 3.10. Strain criterion for quasi-brittle failure
The following results were obtained in Section 2.2.4. Starting with the definition
of a proportional loading in a structure:
it follows that
O"H
const .,
O"eq
Rv = const.
Assuming as in Section 3.1.4, that the ductile damage occurs when the strain-
hardening is saturated or almost saturated, then from the plasticity criterion:
This approximation makes the strain energy density release rate a constant:
(J~qRv (J2 R
Y = 2E (1 _ D)2 ~Ev = const.
and the integration of the damage constitutive equation becomes obvious:
J~~~
t
~'E~ Jdp;
2 P
D = p(t)H(p - PD)
o Po
This makes it possible to determine the value of the damage at each point of a
structure as a function of the accumulated plastic strain, provided that the loading
is proportional.
In one dimension and monotonic loading:
Rv = I,
p = Ep,
(J2
Dc
.
= Dl c ~
(Js
;
the three dimensional equation becomes

(J2
D = 2IiS ( Ep - Epo )'
Writing this for the critical condition of crack initiation,

D D c -+ Ep = E pR '
(J ;
2ES(cPR - cpo )'
This allows us to write the one-dimensional evolution of damage as
E -E
D = D P PD,
cE - E
PR Po
which is the equation of a straight line lying between the two points (D 0,
Ep = Ep o ) and (D = D r.,E p = E p R ) as shown in Figure 3.11.
3.3.2 Exercises on the Fracture Limits in Metal Forming

Metal forming is a process by which a component of a given shape is obtained
from a cast product by means of large plastic deformations: rolling, forging, deep
drawing, extrusion, etc. The state of strain to be obtained is limited by two phe-
nomena: the plastic instability giving rise to unexpected local changes of shape,
and the damage, which may produce unexpected cracks. Let us analyze the later
phenomenon.
Dc - - - - - - - - - - - - -
o Fig. 3.11. Monotonic tensile damage evolution
Master curves of ductile fracture
Calculation of the maximum accumulated plastic strain PR corresponding to the

critical value of damage as a function of the triaxiality ratio
- using the general damage law:
. y
D = SPH(P-PD)'
- in the case of proportional loading:
<J H = constant during the entire the process,

<J eq
- for the critical value of the damage: De = DIe = 1.

Integrating the kinetic law once in the three-dimensional case and once in the
one-dimensional case gives the result:
which may be plotted on the graph (p R=P

cPR
D )
CPD
,(<J
<J eq
H) as shown In
Figure 3.12, which is similar to Figure 3.10. The one-dimensional case is of course
represented by the point
These limit curves depend upon the type of material by means of PD and slightly
by means of Poisson's ratio if 0.25 :S v :S 0.33. They strongly depend upon
the triaxiality ratio. The ductility measured by the maximum accumulated plastic
strain at crack initiation decreases substantially with the triaxiality ratio; it is well
known that a "sharp notch makes the material more brittle". (In fact it is not the
material itself but the effect of the hydrostatic loading on it).
Fracture limits in the deep drawing process (J.P. Cordebois 1983)
Deep drawing consists of indenting a sheet of metal to obtain a three-dimensional

form (Figure 3.13). In practice it is interesting to know a priori the limits in the
strain space between the states of possible strains and the states of strain that do
not occur without a crack.
Calculation of the relation between the principal strain components corresponding
to the critical value of damage may be performed:
- in the case of plane stresses;
- neglecting elastic strain;
- using the general damage law with c P D = 0 and Dc = 1;
- assuming proportional loading;
rr
- neglecting the kinematic hardening in the plastic constitutive equation.
From the master curve of ductile fracture with PD = 0, as c P D = 0,
PR ~ [~(I + ") + 3( I - 2") ( ; :

'R
1.2
0.8
0.6
0.4
0.2
o 1/3 2 3 4 5 6
Fig. 3.12. Master curves of ductile fracture limits

Fig. 3.13. Elementary deep-drawing process
PR may be calculated as a function of the strain by its definition together with

gP = g:
The plane stress condition imposes for the principal stress components that:
o
a2
0]
0 ,
o 0
from which the plastic incompressible condition is written for tr.([c]) = 0:
101 ~ [~ ; -(o,~+J
then
2 (2 2 )\/2
PR = J3 c\ +c2 +c\c2 .
The triaxiality ratio a H / a eq may also be calculated as a function of the strain from
the plastic constitutive equations.
From Section 2.1.4,
-D D) A.
. p _ 3 (aij - X ij
Cij - 2 (u D - XD) eq 1- D'
Assuming no kinematic hardening (X = 0) and proportional loading, it is straight-
forward to demonstrate the proportionality between strains and components of the
deviatoric stress:

Then replacing PRand (J HI (J eq in the first equation gives the result:
A plot of e21 eRas a function of elle R is the classical failure limit in deep drawing
of sheets (Figure 3.14). As for the master curve of ductile fracture, the curves
depend slightly upon Poisson's ratio but the result for each particular case depends
strongly upon the type of material by means of its characteristic e R'
3.3.3 Creep Damage
The application of the kinetic damage law to creep damage is also straightforward.
It differs from the case of ductile damage only in that the accumulated plastic strain
rate comes from a visco plastic constitutive equation:
in which Y may possibly take into account the microcrack closure effect (see
Section 3.1.4) Remember that the material parameter 5 depends upon temperature.
From Section 2.1.4,
. - L 1-
(iT D - XD )e q - R - (J
y
]-n
P- n[ K '
x
with p= I-D
~0.8eR
"" \ '
\ 0.6 v = 0.33
" 0.54 v = 0.27
0.45 \ 0.4 /
0.64
'" \ /
"" 0.2
,,\ '\ /
/
/1
./
//
///
/./
Fig. 3.14. Limit curves of deep
- 0.1 o 0.1 drawing
For simplicity, let us consider a simple perfect visco plastic Norton's law with
neither strain hardening nor yield stress (see Section 2.1.3). Then the visco plastic
multiplier reduces to:
~ . [ (Je q ] N
I- D = P= K" (I - D)
where K" and N are material parameters which can be found for some materials
from the table in Section 3.5. Then
. (J;qRV [ (Je q ] N
D = 2ES(1 _ D)2 K,,(I _ D) H(p - PD)
(J~+2Rv
or D = 2ESKt'(l-D)N+2H(P-PD).
The damage becomes a time dependent phenomenon like the visco plastic strain.
In the one-dimensional case this constitutive equation reduces to the Kachanov's
earlier model:
b =[ (J ] N +2 H( )
A(l- D) P - PD ,
I
with A = (2ESK::) N+2.
Let us calculate the evolution of the damage as a function of the time in a simple
creep process where (J = const.
Let t* be the time needed to reach cp = PD by creep without any damage,
dD = (~) N+2 (I _ D)-(N+2) dt

or
which leads to
D ~ 1- [1- (N +3) (~(+2 (t -t'r~3

The critical value of the damage defining the rupture of the volume element under
tension is obtained by (see Section 3.1.3)
(J2
De = Die (J_";' :::; 1,
but as Norton's creep law is used here, (J 00 -+ 00 at rupture when ip -+ 00.
Then De = 1 is the proper rupture criterion for this analysis. The time to rupture
tr(D = 1) is
_ * _ _1_ (::.) -(N+2)
tR t - N +3 A .
The time t* must be calculated from a visco plastic model. For simplicity let us
again take the Norton's perfect visco plastic law:
or
7 (;y j
o
dE,
0
dt , t' = PD (;J- N
tR = PD (;J -N+ N ~3 (~) -(N+2) with PD = EP D :u~::

Then it is possible to draw the graph of the evolution of the damage as a function
of time for constant stress creep processes (Figure 3.15).
Viscoplastic damage behavior must be accurately calculated using the viscoplastic
constitutive equations coupled with damage as shown in Section 2.1.4; this is only
possible by means of numerical calculations. In order to give a general idea based
on analytical results, let us again assume the simple Norton's perfect visco-plastic
law and calculate the evolution of strain with time for a simple creep process at
(J = const.:
(J
E =
e E'
Ep [Kv(I(J_D)] N
D [A(I~D)] N+2 H(Ep-PD) ;
For Ep < PD and with the initial condition t = 0 ~ Ep = 0:
D = 0 ~ E
p
= (~)
Kv N t.'
o
Fig. 3.15. Evolution of the damage

in creep processes 0'3 < 0'2 < 0'1
o t*1 (schematic)
Taking (1 - D) from the preceding calculation gives for t > t*
i = UY [1- (N +3) (~t+2 (t-t')r~3

(;.J j [I-(N +3) (~t+2 (t-t')r~3 dt
t'
or
The shape of the graph of the total strain C = c e + cp as a function of time for
different values of (J' is given in Figure 3.16.
3.3.4 Exercise on Isochronous Creep Damage Curves
In engineering practice it is of interest to know the stress which, when applied

constantly to a given material at a given temperature over a certain period of time,
causes a certain amount of plastic strain or damage. Those isochronous curves
may be obtained by interpolation of many experimental results; they can also be
calculated from the constitutive equations of plastic strain and damage.
Fig. 3.16. Evolution of the strain up to rupture in creep processes, 0"3 < 0"2 < 0"1
3.4 Fatigue Damage 127
Calculation of the time to rupture in pure creep
This calculation may be perfonned for:

- tension
- in the case of Norton's perfect viscoplasticity for the material AISI 316 at 600C
and constant damage threshold PD = E pD :
I
K v = 765 MPa hi / N A = 900 MPa h N +2
PD = 2.10- .
N = 8.2 E 2
For Ep < PD the relation between the stress, the plastic strain and the time
calculated in Section 3.3.3 is
Ep = (; )N tor t = Ep (.!!..-)-N
v Kv
Then, it is easy to plot t as a function of (J taking Ep as a parameter (Figure 3.17).
For Ep 2: PD' the relation between the stress, the damage and the time is:
D 1-[I-(N+3)mN+2(t-t'r~3
with t* = PD (;,J- N
PD I - (I _D)N+3
(;J +
or t = --=---=N N +2 .
(N+3)(~)
Taking D as a parameter one may plot t as a function of (J (Figure 3.17).
At rupture, the critical value of the damage Dc is
Dc = I,
which gives the time to rupture to be plotted on the same graph (Figure 3.17).
3.4 Fatigue Damage

The damage process is calIed fatigue when the loading is repeated a large number of
times with the same amplitude or with varying amplitude. It can be one dimensional,
in tension or tension-compression, or three dimensional.
3.4.1 Low Cycle Fatigue

The case of low cycle fatigue occurs when the loading is large enough to produce
plastic strain of the order of magnitude of the elastic strain. This means that the
number of cycles to failure N R is relatively low. An order of magnitude is
N R < 10000 cycles.
STRESS (MPa)
700
~~ - D=O.Ol
""'" ~"
.~ - D=0.02
- D=0.05
~~ ~ ::::; ... ~ "!!o.. - D=O.l
......... ~ "I
~ r....~ . - D=0.2
:::::r-. ~ ........
~
~ .;:<~ - D=l
r'" .........;;
::::
.
~ ~~ . t:--. ~ ~~
~ f:::r-.r- ~~, ~
.....
- Ep=O.ool ~ ::::~ tS:; r:::
r-...
)D
- Ep=0.002
~
i""
r...r--
~ .....-...,
- Ep=O.005
~~
100 -
-
-
Ep=O.Ol
Ep=O.02 ~ ............. ) Ep
80 I I I II Ill!
10 100 1000 10000 100000
TIME (hours)
Fig. 3.17. Isochronous creep damage curves (after F. Hild)
One-dimensional periodic fatigue
Let us start with the general kinetic damage law written for the one-dimensional
case defined by the stress a(t) and the strain c(t) as functions of time t:
. y
D = Sp H(p-pv),
a~qRv a2
y = 2E(I-D)2 - 2E(I-D)2'
P= Ic'pl,
2
D = 2ES(~ _ D)2 lipl if cp> PD'

The number of cycles to failure is the sum of two parts: the number of cycles No
to reach p D for microcrack nucleation, and the number of cycles N D during which
the damage occurs, initiating a crack at the mesoscale:
NR = No +Nv
No must be calculated by the classical elastoplasticity constitutive equations without
any coupling with damage, as during this period D = O.
Let us consider a periodic strain-controlled loading of constant amplitude ~ c

(Figure 3.18).
The elastoplastic constitutive equations of Section 2.1.4 show that after some cycles
the elasto plastic process is stabilized to an amplitude of stress which remains
constant until p = PD:
~c = const . ~ ~a = const. ,
as ~c = ~ce +~cp and ~c e = ~a /E
~cp = const.
The accumulated plastic strain after one cycle is
8p
8N
J
I cycle
lipl dt = 2~cp'
and for N cycles,
Then
au - at
No = -PD- wIth
'
PD = cpD----"-
2~cp' a M -at
The calculation of N D requires the integration of the kinetic damage law over one
cycle and then over N cycles:
0'2
D 2ES(1 _ D)2 lipl
Fig. 3.18. Strain controlled low cycle fatigue

Normally this calculation is performed by computer, as _a_ is a function of

I -D
cP ' but the general trends may be obtained analytically from some simplifica-
tions:
- The material is first assumed to be perfectly plastic during one cycle with a
threshold as = ~a, and the variation of the damage is neglected for the
integration over one cycle (Figure 3.19).
- The strain and damage behaviors are assumed to be identical both under tension
and compression, this corresponds to the crack closure parameter h being equal
to I (Section 2.3.4). This allows us to calculate the damage per cycle as:
A cyclic relationship between flc p and fla is assumed. It can be derived numer-
ically from the equations of Section 2.1.4 and has strong experimental support:
Kp and M are material parameters which can be found for some materials in
the table in Section 3.5.
It follows that:
oD K~ flc(M+2)/M
oN 4ES p ,
Fig. 3.19. Simplified stress strain cycle

D.Ep being constant throughout the damage process, the damage is linear with N:
D = J
N
8N 4ES p
2
8D 8N = Kp D.E(M+2) /M(N - N ).
0
N"
The number of cycles to failure is reached when D = Dc' with No
D =
c
K~
4ES
D.E(M+2)/M (N
p R
_~)
2D.E '
P
N - ~ 4ESD c D. - (M+2 )/M

R - 2D.Ep + K~ Ep
The power relationship between N Rand D.Ep is known as the Manson-Coffin

law of low cycle fatigue. An example is given in Figure 3.20. For many metallic
materials (M + 2) / M is of the order of 2.
Coming back to the evolution of D we have
K2
D = - p D.E(M+2) /M(N - N.)
4ES p 0
4ESDc A - (M+2) /M
and N R - No = K2 UEp .
P
This shows that

D = Dc N -No.
NR-No
In a small range of D.Ep the critical value of the damage does not vary too much:
Dc c:,: canst .
Fig. 3.20. Manson-Coffin curve of low cycle fatigue of a strainhardened steel: AISI 10 10 at temperature
T = 20C
o 0
~-------------- ~
I
I
1
I
I N -NO N
1---
=---------______ ~-~
- L_ _
~
o 0
Fig. 3.21. Evolution of the damage as modeled by the simplified theory
Then in practice a single curve exists in the diagram (D, N - No ), but the
NR-No
damage evolution curves differ by No in the (D, ~) diagram (Figure 3.21).
An example of real low cycle damage evolution measured by the variation of the
elasticity modulus is given in Figure 3.22.
Three-dimensional complex loading in time

The kinetic law of damage evolution has to be integrated with the coupled elasto-
plastic constitutive equations for the history of the stress or strain under conderation.
This can only be done numerically by one of the methods described in Chapter 4.
Here again there is a strong influence of the triaxiality.
3.4.2 Exercise on Creep Fatigue Interaction

An important feature of low cycle fatigue of metals at high temperature is the accu-
mulation of fatigue damage and creep damage giving rise to nonlinear effects. This
can be modeled by a complete numerical computation of the coupled constitutive
equations but the main trend of linear or nonlinear interactions may be pointed out
analytically with some approximations.
Consider a periodic one-dimensional stress loading and assume a stabilized cycle
in the (stress versus plastic strain) diagram as shown in Figure 3.23.
Calculation of the expression
where:
- N R is the number of cycles to failure of the creep fatigue process corresponding
to some value of Dc;
1.0 D
0.8
0.6
A -2
) tiE. = 1.8,10
-2
0.4 ,. M:=!,IO
" -2
I, 6.E. = 0.3,10
0.2
N
NR
o 0.2 0.4 0.6 0_8 1.0
Fig. 3.22. Damage evolution in low cycle fatigue of strain-hardened AISI 1010 steel
102 < N R < l<f cycles
(J' (J'
B C B C
~eFp .le~
A 0 ep
F E
~ep
Fig. 3.23. Creep fatigue loading
- N'R is the number of cycles to failure of a pure cyclic creep process at constant
stress amplitude a = aM for which the critical value of damage is D~ ;
- Nt; is the number of cycles to failure of a pure fatigue process without any
hold time,
!:l.t = 0, for which Dc = D{.

The same damage threshold PD allows for the above three cases since the same
stress amplitude is assumed.
Using the general kinetic damage law,

, a2
D = 2ES(1 _ D)2li p l if c p > PD'
1) Calculate the damage per cycle in the creep fatigue process, neglecting the
variation of damage during that cycle:
8D a2
8N = ES(1 ~ D)2 ~cP'
from which the number of cycles to rupture is
NR - ~cp
_ -I [PD
2 +
ES [1 - (1 - Dc?]
2
1
'
3a M
2) The cycle in Figure 3.23 is partitioned in
- (AB + DE) considered as a pure fatigue process of plastic strain amplitude
~cF ,
p'
-(Be + EF) considered as a pure cyclic creep process of plastic strain ampli-
tude ~c~ ,
The damage per cycle of the pure fatigue process is
8D a~ F
8N = ES(1 _ D)2 ~cp ,
from which the number of cycles to failure is
NF = ~ F-I [PD ES[1 - (1 - D[)3 1]

R cp 2 + 3a2
M
'
3) The damage per cycle of the pure cyclic creep process is
8D a~ ~ c
8N = ES(1 _ D)2 c p
and
It follows that
N N
The expression ~ + ---.!i may be smaller or greater than I depending upon
NR N'R
the relative values of Dc' D; and D~ . This property is known as the nonlinear
creep fatigue interaction. As 6.c p 6.c: + 6. c~, the sum of the two ratios is
equal to I if Dc = D; = D~ :
NR NR
N F + NC = I,
R R
which is the Taira rule of linear creep fatigue interaction.
At very high temperature, the superalloys show a strong nonlinearity in their creep
fatigue interaction in the sense that a small amount of creep may considerably re-
duce the number of cycles to failure as shown in the bottom curve of the interaction
diagram in Figure 3.24.
3.4.3 High Cycle Fatigue
If the amplitude of the loading is low, the amplitude of the plastic strain may be
very small, even negligible at the mesoscale in comparison with the amplitude of
the elastic strain. This corresponds to high values of the number of cycles to failure .
For instance,
N R > 100000 cycles.
Another feature which makes the damage analysis of high cycle fatigue difficult is
its high degree of localization. Very often only a very small microelement damaged
at the free surface of the body gives rise to one microcrack by slips (stage I) which
later propagates perpendicularly to the loading (stage 2). Then, for high cycle fa-
tigue, materials may be considered as quasi-brittle and modeled by a damageable
microinclusion embedded in an elastic mesoelement. Its complete numerical anal-
ysis, including high cycle fatigue, is performed in Section 4.3. Nevertheless, the
linear interaction
o Fig. 3.24. Creep-fatigue interaction diagram

main features of high cycle fatigue may be derived from the simplified analysis of
the quasi-brittle damage presented in Section 3.2.2.
The following kinetic law of damage evolution
2
. af .
D = 2ESRv (~reqH(ceq - PD) WIth PD =
has been derived from the three hypotheses:
- perfectly plastic behavior of the inclusion with a plastic threshold a~ = a f'

the fatigue limit
- equality of the total strains in the inclusion and in the matrix (Lin-Taylor's
hypothesis).
- elastic strain neglected in comparison to the plastic strain in the inclusion,
This law accounts for any kind of three-dimensional high cycle fatigue, Ceq being
the accumulated total strain.
One-dimensional case of fatigue

Consider a periodic process of fatigue !n tension compression at the mesoscale,
stress controlled ~a = const. as shown in Figure 3.25. Note that a one-dimensional
state of stress at the mesoscale accounts for a three-dimensional state of stress at
the microscale due to the mismatch of Poisson's ratio at the mesoscale and the
plastic contraction coefficient at the microscale.
In one-dimension at the mesoscale:
a
3at'
r1 r1
o~ _ _ _ .'
Fig. 3.25. Stress controlled high cycle fatigue

and
Then
A first integration over one cycle gives the damage per cycle.
da,
JD =2a7
-
IN
-
E2 S
[ (l+v) (!la
--af) +--
2
1 - 2 V [(!la)3
9a7
- -af
2
3]]
A second integration as a function of the number of cycles gives the evolution of
the damage for the initial condition,
D = E2a7 [ (!la ) 1 - 2 v [(!la)

2 S (l+v) ""2- af + 9a7 ""2 -a f 3 3]] (N-N O) '
where No is the number of cycles for Ceq to reach the threshold PD:
N = PD
o 2.:lc
The number of cycles to failure is reached when D = Dc = DI e _

a
:2'
a2
Taking
D = D Ie as a "good" approximation,
DIe = E2aj [ (!la
2 S (1 + v) ""2 - af ) + 1-
9a72v [(!la)3
""2 - af 3]] (NR - No)
or
The graph of the amplitude of stress .:la as a function of the number of cycles to
failure is known as the Woehler curve. An example is given in Figure 3.26.
Using the expression for (N R - No) ' it is possible to express the evolution of the
damage in a simpler way:.
M(MPa)
1200
+
800
+
400
( cycles)
Fig. 3.26. Woehler curve of high cycle fatigue of AISI 316 stainless steel
This shows that the damage is a linear function of the number of cycles. With
the approximation De = Di e' a unit straight line represents the evolution of the
damage in the graph (D, ~--~o)' but as No decreases with the amplitude of
the stress, different lines represent the evolution of the damage as a function of
the number of cycles. These types of damage evolution are shown schematically
in Figure 3.27.
An example of real evolution measured by means of elasticity changes is shown
in Figure 3.28.
Influence of the mean stress

The previous study has been worked out with tension and compression stresses
equal in their absolute values, that is, a mean stress equal to zero. The influence
of the mean stress, an important feature of fatigue, can be analyzed numerically
as described in Chapter 4, using the concept of unilateral conditions developed in
Section 2.3.4, which takes the microcrak closure effect into account. Nevertheless,
using a very simple example, let us illustrate how the stress amplitude which
induces a certain number of cycles to failure decreases as the mean stress increases.
o 0
: N -No N
: NR-N o NR
o ~------------------~I--~ 0
Fig. 3.27. Evolution of high cycle fatigue damage as modeled by the simplified theory
1M =600MPa
0.2
/ ,1'.0'=500
/ ,
0.1 / 1/1'.0'=480
/ /I N
Fig. 3.28. Fatigue damage evolu-
~ tion for AISI 316 stainless steel
o 0.2 0.4 0.6 0.8 1.0 104 < N R < 106 cycles
Consider, as above, a periodic process of fatigue loaded in such a way at the

mesoscale that we have a one-dimensional loading at the microscale, R" = I.
The maximum stress is aM' the minimum stress is am and the mean stress is
a = aM ~ am, (Figure 3.29). Then, the stress amplitude is
~a = aM -am = 2(a M -a).
A pure unilateral condition of microcrack closure is assumed for simplicity. From

Section 3.1.4.
I + v (aij)(a ij ) v akk
( )2
2E (I-D)2 -2E(I-D)2'
-2 -1 /0 / 2 0
./
/
/
/ / -1
/ /
I
I
/
-2
/
-3
-4
I
/
Fig. 3.29. Goodman 's diagram (schematic)

which reduces to
Y in tension a > 0;
2E(1 - D)2
Y = 0 in compression a < O.
The kinetic law of quasi-brittle damage developed in Section 3.2.2 was
D = YII..~.q 1f' {a eq ~ af
5 Ceq ~ CD
.
With the results obtained previously and the new conditions, it becomes
. af2
D = 2E 2 5 iF in tension
b =0 in compression.
The integration of the damage rate concerns only the positive stress taking the
micro plasticity threshold equal to zero for simplicity
bD
bN =
J
aM
a2 a2
2E;5 da = 2E;5 (aM-af)
l1 f
and the number of cycles to failure for D = Die is

Di e
NR = NO + 2 )
af
2E25 (aM - af )
where No remains unchanged, No = ~~~ if a non unilateral condition is assumed

for the plastic strain. Replacing D.a by its value as a function of aM and a gives
the relation which defines all possible combinations of the maximum stress and
the mean stress to obtain a given number of cycles to failure N R:
EPD
-.,.---=-=-'7 + Di e
= NR
4(aM - a) a}
2E25 (aM-af)
If a MO is the value of the maximum stress aM which gives the number of cycles to
failure for a zero value of the mean stress, the graph of aM as a function of a
aMO a MO
is known as Goodman's diagram. Figure 3.29 shows an example corresponding to
.
an hypothetIcal case where No = T
N
:
aM _ _--;;-_
a MO a MO 2 _ a MO .
aM
Experiments show the same trends but the graph is closer to a straight line starting
from the point aM = (j = au'
Other parameters may act on the phenomenon of fatigue. The influence of temper-
ature is taken into account by the material parameters which are functions of the
temperature. In high cycle fatigue the influence of frequency is negligible; in the
model it disappears when using pure plasticity at the microscale and pure elastic-
ity at the mesoscale. The influence of environment can be evaluated by chemical
means used to define appropriate internal variables; it is, however beyond, the
scope of the damage mechanics.
3.4.4 Exercise on Damage Accumulation
The question of linear or nonlinear accumulation of fatigue damage due to different

levels of stress is very important for the applications.
If the damage due to sequences of loading of different amplitudes may be added
independently of their order in time up to failure, this property is known as
Palmgreen-Miner's rule: If n i are the numbers of cycles of sequences of peri-
odic loadings defined by there amplitudes /}"a i , N R being the number of cycles
to failure due to those sequences; and if N Ri is the number of cycles to failure
that would exist for the same material submitted to a periodic loading of constant
amplitude /}"a i up to failure, then
NR = ~ni'
~~ = 1.
NRi
This "rule" is in fact no longer a rule if a kinetic damage law is used; it is only
a consequence of the mathematical properties of the model used. Let us see under
which conditions the model developed in Section 3.4.3 has the property of linear
or nonlinear accumulation.
Consider two levels of a one-dimensional loading defined in Figure 3.30.
Using the high cycle damage evolution model written as
N - No(f':.a)
D = Die ,
N R - No(f':.a)
(j
J~M~ I I t
mn
I
0 l'1
\ I
;
n1 n2
NR
Fig. 3.30. Two-level fatigue case
determine the number of cycles to failure when n l > N ol :
Find for which condition the Palmgreen-Miner's rule of linear accumulation would
be valid
...!2:L + ~ = I only if NNOI NN02 that is, if a damage threshold proportional

NRI NR2 RI R2
to the number of cycles to rupture is assumed.
This corresponds to a unique damage evolution curve independent of the amplitude
of stress if D is plotted against N / N R (see Figure 3.27).
In fact this property is a general property of any differential equation with separated
variables. All damage equations of the following general shape show the property
of linear accumulation:
iJ = II (D)I2(x) with x = a or E or p
if associated with constant initial and final conditions.

In the model studied, the nonlinear accumulation comes from the different values
of the damage thresholds.
The experiments show that the Palmgreen-Miner property may be applied if the se-
quences do not vary too much in amplitude (the variations of PD may be neglected)
;2
but it may give wrong results if the amplitudes differ by more than 10 or 20%.
To quantify these differences please draw the graphs as a function of ...!2:L
R2 NRI
corresponding to the two-level fatigue sequence for the conditions in Figure 3.31 :
3.5 Damage of Interfaces 143
\
,..,
.
" .... ...... ,
\"
~
\ '- ...........
0.8 ''''
\
\
...
\
\
0.6
0.4
Fig. 3.31. Diagram of nonlin-

0.2 ear accumulation of two level
111 fatigue processes:
low level first No I / N HI
NRI
0.4, No2lNR2 = 0.2
high-level first No 1/ N HI
o 0.2 0.4 0.6 0.8 0.2, No2 / NR2 = 0.4
3.5 Damage of Interfaces
Composites or multimaterials are made of different layers of possibly different

materials bonded together by glue, polymerization or diffusion. One of the main
failure mechanisms is the debonding of interfaces. As an interface is "nothing
between two somethings!", it is interesting to consider it as a two-dimensional
surface medium of zero thickness and to develop a specific damage model in which
the inputs are the components of the stresses and the strains which load the
interface. These are the components continuous through the interface which may
result from a structure calculation of the multimaterial in which the interface is
ignored except that the stress vector and the displacement vector are continuous
through the interface surface. Then, the calculation of the damage and the debon-
ding acts as a post processor of the structure calculation.
3.5.1 Continuity of the Stress and Strain Vectors

Let us study the basic problem of a mesovolume element made of two different
layers A and B bonded together by an interface I of negligible thickness and loaded
by a three-dimensional state of stress (Figure 3.32).
On the surface of normal n3 the equilibrium imposes that
i = 1, 2, 3.
Then, three components of the stress in the interface are known from the stresses
in layers A or B:
, (1- .
I,j
J Fig, 3.32. Interface at the mesoscale
On the interface, there is also continuity of the displacement vector:
- From the first equation
ou A ou A
~dx}+~dx2
UX I UX2
in which dX2 = 0 -+ S11 = sfl'

ou~ ou~
and dX 1 = 0 -+ -
oX 2 OX2
- From the second equation
dx} =0 --+ E~ = E~,
dX 2 = 0 --+ out = auf

ox} ox}
Then, three components of the strain in the interface, complementary to the known
stresses, are known from the strains on layers A or B (plus a rotation condition
which does not play any role in classical continuum mechanics):
The missing terms in [0'/] and [1: / ] are of course those which cannot be defined due
to the surface character of the interface: no area to support loads in the x 1 and x 2
directions, no length to define strains in the X3 direction.
3.5.2 Strain Surface Energy Release Rate

The plane strains I:!/l (with (X and f3 = 1 or 2) and the antiplane stresses O'f3 are
considered to be independent state and associated variables of the interface. Fur-
thermore, qualitative experiments show that many interfaces exhibit elastic, vis-
cous, plastic and damageable behavior. To follow the State Kinetic Coupling
theory used in Chapter 2 it is necessary to split the associated variable s!/l to I:!/l
into an elastic part s:/l and an irreversible part s:ii
and the same for the state variable ei3 to which 0'i3 is associated
By analogy with plasticity, a scalar strain hardening state variable n is introduced,

its associated variable is named r. The present study is restricted to isothermal
behavior which avoid introducing the temperature and the entropy variables. D is
the damage variable, the density of microdebondings, ji is its associated variable
as usual.
The state potential is chosen by analogy with the potential used in Chapter 2 except
that it is an energy per surface unit;
1 k/1 1+1'
If; = - [k,: 1:1 : el + k,,: ee: ee](l - D) + -"- n- I'-
2 1+/1
(the symbol: means the double contracted tensorial product)
where k" k", k" and /1 are tensors and scalar parameters characteristic of the
nature of the interface.
It follows that:
e _ olf;
S - 01: 1 ' s:/l have the dimensions of force per unit length,
olf;
0'1 = oe = k,,: ee(1- D), ef3 have the dimensions of length,
olf;
r = -on = k n n 1/1' , r has the dimensions of stress,
ji = ~~ = - (~' : 1: 1 : 1: 1 +~ : e:e).
Y =:= - ji is the strain surface energy release rate which can be written as a function
of the variables known by a classical structure calculation el and (JI
3.5.3 Kinetic Law of Debonding Damage Evolution

The dissipation satisfies the second principle of thermodynamics if
san: e + (JI : ean - ni: + yD ~ O.
The kinetic law of damage evolution derives from potentials of dissipation which,
in the case of stress and strain controlled phenomena, needs to be:
- A viscous-type potential Q , a nonlinear function of the accumulated equivalent
strain e!q defined by
2 I~'I
Q = ~ eeqi>eq
2 St
Where 1] and St are material constants and eD' also a material constant, is the
damage threshold.
- A plastic-type potential F governed by a criterion/written as a function of an
equivalent stress (J!q taking into account the unilateral character of interfaces
which do not show any damage in compression
The ratio of the ultimate normal stress (JR to the ultimate shear stress r R models
the large difference of strength which often exists between tension and shear in
interfaces. <>are the Macauley brackets, II is the absolute value.
(J!q = 1133 in pure tension,
l1!q = 0 in pure compression,
- a tension decreases the strength in shear,

- a compression increases the strength in shear.
The coupling with damage is neglected in the plastic criterion and potential:
f = (J;q - r - (Jy = 0,
ji2
F=f+ 28 ,
"
(Jy is a yield stress and S" another parameter characteristic of the interface.
The kinetic laws of evolution derive from these potentials by the normality rule
an _
oQ
SaIl - ;;-:- ,
ueafJ
of. .
n=--A=A
or '
D = _ oQ _ of i = y (e;: i;q + ~)
oji oji Se S,,'
i ~ 0 is the plastic multiplier determined by the consistency condition J = 0,
.I .
(Jeq-r= 0,
or, together with r = kn7rYI' and n= i

'f {f=
f' 0
1
=0 .
D = y [e;: i ;q(if eeq ~ eD) + ((J;q~ (Jy Y-1:S" (if {j: ~)1
Finally D = 0 if not,
D = Dc -+ debonding of a Representative Surface Element.
The rheological parameters of the interface introduced are k", ke' Se, 1] , (JR, !R' eD'
(Jy, k", 11, Dc: too many! In fact some of them may be identified as group parame-
ters and a simplified model restricts them to four only.
3.5.4 Simplified Model

Different classes of models may be derived from the above expression. Let us
restrict ourselves to the two most common mechanisms of debonding:
- debonding in mode I due to the normal stress (J 33 ,
- debonding in mode II due to the shear stress (J 13 or (J 3
This corresponds to a loading such as e!/l ~ 0 or a material for which ke approaches
zero ans Se is large enough to neglect the first term in the damage law.
Furthermore, consider an isotropic interface which reduces to two the number of

independent coefficients in the matrix [k a ], for the same reasons as for isotropic
elasticity. Take ka for mode I and k~ for mode II, then
Last assumption to make the model as simple as possible! The yield stress, always
difficult to determine, is considered to be zero. Then
if {f=
1=0
0
or
Qualitative identification
It is possible to relate the parameters ka' k~, k" and Sa to engineering parameters
such as (1 R or 'R and others easier to measure.
- Consider the case of mode I debonding in pure tension
The damage law reduces to
A mesodebonding occurs for D = Dc when the tensile stress reaches the ultimate
tensile stress (1R . These conditions are obtaned by integration of the damage
differential equation with the initial condition of an undamaged state:
0"33=0 -+ D=O
from which
J1 1 - (1- Dd 3 J1 +2
2kak~Sa (1~+2 3
- Consider now the case of mode II debonding in pure shear:
. O"i3 (O"R (113)/l - 1 J1 O"R 0"13

D=k~(1_D)2 'RIC; k"'RS:
The conditions of a mesodebonding initiation D = Dc when 0" 13 = 'R are obtained
by the same integration as previously and for the same initial conditions:
from' which
A more phenomenological writing of the Damage law is obtained by replacing the

two calculated group parameters by their values as functions of aR , 'R and Dc :
Quantitative identification
For each nature of interface, four coefficients need to be determined:
aR: the ultimate stress in pure tension,
, R: the ultimate stress in pure shear,
Dc: the critical damage at debonding,
fJ.: the exponent characteristic of the nonlinearity
Even in the case of glue the identification tests must be performed on a multimate-
rial that is the interface and its two adjacent layers because an isolated interface,
glue for example, has different properties than a built-in interface.
- A pure tensile test allows the determination of a R .
- A pure shear test with unloading allows the determination of 'R and Dc by
extrapolation of the Damage evolution measured by
E G
D=l - :E=l- G,
Gbeing the damaged shear modulus and G its value in the undamaged state. This
formula using G = /2(1 + v) assumes that the Poisson's ratio is not affected by
the damage which is exact in the case of isotropic damage.
- The best way to obtain the exponent fJ. is to perform fatigue tests in tension
(aM = 0, aM) or in shear and to deduce fJ. from the slope of the Woehler curve
giving the logarithm of the stress amplitude as a function of the logarithm of
the number of cycles to debonding.
The equation of the Woehler curve is obtained by integration of the above damage
law over one cycle first neglecting the variation of D:
oD
oN
(a
aD = ~ 1- (1- Dd 3 M)Il+ 2
oN 3 (1-D)2 aR '
and then as a function of the number of cycles with the conditions
N=O ~ D=O,
N=NR +- D=D c ,
Dfe 3(1- D)2 _

---------=-3bD-
NfR (a- M)Il+ 2bN,
l-(1-Dd aR
o 0
Having obtained the Woehler curve from experiments we calculate the exponent
Jl from
3.5.5 Exercise on a Debonding Criterion for Interfaces

A simplified criterion for debonding of interfaces loaded in proportional loading
consists in writing
or
- Show that this criterion is the exact solution given by the damage law when
integrated in the case of proportional loading in shear and tension:
The integration reduces to
- In the case of shear and compression, the integration is more complex due to
the sign of a 33 < 0 but the numerical values are not far from those given by
I _
a eq - aRo
For example show that for a constant value of 0" 33/0" R = -1,0"23 = 0 and Il = 4 the
integration of the damage law gives 0" d, R = 1.371 and the simple criterion
O"13/'R = 1.414.
Damage law:
_~1-(1-Dd3(
dD - 3 ( 2 1+
O"i3)(_
2
O"i3)3/2(_
1+ 2
O"i3)- 1/20"13
1+ 2 2 dO"13'
i-D) 'R 'R 'R 'R
Integration:
Criterion:
- Graph of the debonding criterion in the tension-shear plane (Figure 3.33):
0"3310"331 + 0"13 = 1
O"~ ,~ ,
0"33> 0 --+ half a circle,
0" 33 < 0 --+ two branches of an hyperbola.
4 'to
cr
ll
r----r----r----+----+----+----+----4--0-+---- 1----~---
cr,
-8 -7 -6 -5 -4 -3 -2 -1 2
-2
-3
-4
Fig. 3.33. Debonding criterion (from J.P. Sermage)
3.6 Table of material parameters
Users should be aware that all numbers of the following table are to be considered as
orders of magnitude only, because each set corresponds to a specific material which
may show slightly different properties from another one identically denominated.
When performing a structural calculation, use available characteristics chosen with
the name of the material and be prepared to perform calculation of corrections with
the numbers identified, if possible in situ, on the real material of the real structure.
Furthermore, keep in mind that each parameter is related to a specific model.
For example, the value of the yield stress of a given material may be different
for a perfectly plastic model, a linear kinematic hardening model or an isotropic
hardening model.
Dear "identify-or" , each time you perform an identification please write the results
in the following chart and send it to me! Thank you.
Material Temps Elasticity Plasticity Strain hardening Viscoplasticity DAMAGE Crack
= C closure
E v af ay au X oo -y R oo b K oo n S pD Die h
MPa MPa MPa MPa MPa MPa MPa S-1 MPa
Steel
AISI 20 190 000 0.28 200 320 700 - - 2.4 0.44 0.2 0.2
1010
Stainless 20 200 000 0.30 180 260 700 200 2 300 I - - 7 0.10 0.15 0.2
steel 600 140000 0.32 3 6 - 150 12 0.2 0 0.5 0.5
AISI316
Copper 20 100 000 0.33 100 190 300 - - 0.4 0.35 0.85 0.6
Alumi-
nium 20 72 000 0.32 250 300 500 - - 1.7 0.03 0.23 0.2
alloy
2024
Ceramic 20 400 000 0.2 300 306 500 - - - - - - 6 0.10 0.99 0.2
Alumina
at micro
scale
Concrete
in 20 30000 0.2 1.5 2.5 3 - - - - - - 2.510- 7 0 0.2 0.2
tension ;.>
V>
at micro
~e ____ L - _____ ,
o
-
3
~
o
...,
5"
&
;;;>
~
V>
\.>
-
Chapter 4
Analysis of Crack Initiation in Structures
The fundamental reason for studing damage mechanics is to understand why and
how materials break. Together with physics, metallurgy and chemistry this knowl-
edge allows us to improve the mechanical properties of materials and to design
new multimaterials. The practical reason for studing damage mechanics is to pre-
dict when materials, as they are currently made, will break upon submission to
mechanical and thermal loading. This involves the analysis of real components of
structures in real or presumed situations.
This analysis requires two main steps: modeling behavior of the material and solv-
ing the boundary value problem of stress, strain and damage fields in the structure.
The first step has been studied in detail in the previous three chapters (the con-
stitutive equations for plasticity and viscoplasticity are given in Section 2.1.4; the
constitutive equations for damage are given in Chapter 3). In the present chapter
we shall deal with the second step, by considering how these constitutive equations
may be integrated in the classical framework of continuum mechanics, the basic
variables of which are stress, strain, temperature and time. In the first section, we
review the classical equations and methods. Then, several levels of computation
are developed, whereby the coupling between damage and strain is: neglected for
a rough approximation; taken into account only for the most damaged points in
small-scale damage zones; or, taken into account everywhere for a fully coupled
analysis of structure in large-scale damaged zones.
4.1 Stress-Strain Analysis
Consider the general problem of a structure S defined by its geometrical contour

as on which a load density F(t) given on aS F and displacements u(t) given on
asu (f)SF + asu = as) are applied as a function of'time t. A body force density
~t) may be applied in S and the temperature distribution in S is given by T(M ,t)
(Figure 4.1).
The mechanical properties of the material are represented by the strain and damage
constitutive equations possible with initial values of stress (residual stresses), of
strain hardening (components obtained by metal forming, for example) and of
damage (analysis of the residual life of an accidentally damaged component, for
example).
Point M
Structure S - j - - -
Boundary 6S
Fig. 4.1. Boundary value problem
The general equations of the problem are:
- The equilibrium equations derived from the principle of virtual work,

a ij .j + Ii = pili in S
d271 . .
p is the density and -IT = dt 2 IS the acceleratIOn
- The strain-displacement equations, given here for the small deformation theory:
I
E
ZJ
= -2 (u '-J. + U t,J ) in S
- The set of constitutive equations, given in Sections 2.1.4 and 3.1:

Eij = Eij + Efj ,
Eij = Eij(aij' D , T),
ifj = ifj(aij , aij' X ij ' R , DT),
,
R = R(aij,aij, X iJ, R , D,T),
Xij = X ij (a ij , aij' R, D ,T),
D = D(a ij , D, Efj , T) .
- The boundary conditions
aijn j = Pi given on 85 F'
Ui = u i given on 85u '
T = T given on 5 +85.
- The initial conditions
aij(t = 0) = a?j
R(t = 0) = !In,
X(t
'.J
= 0) = XO,
'.J
D(t = 0) = Do.
156 Analysis of Crack Initiation in Structures
Due to the complexity of the partial differential equations and the nonlinearities
introduced by the constitutive equations, there is no analytical method for solving
this type of problem, in the general case. It can only be solved by means of
approximations in some cases and by numerical timelike incremental analysis,
usually with the aid of computers.
4.1.1 Stress Concentrations
Regarding damage analysis in structures, the critical points at which a mesocrack

may be initiated are almost always located in regions of stress concentration caused
by sharp variation of geometry: holes, notches, etc. This means that a high degree
of accuracy of stress strain analysis is needed only in those regions.
There is no general method for easily calculating stress concentrations, not even
for cases of elasticity without any plastic strain or damage. A useful concept is the
stress concentration factor K T , defined, at one point loaded in one dimension, as
the ratio of the stress (1'A at the maximum loaded point A to the nominal stress
that would exist in the absence of any effect of the geometrical perturbation on the
stress field, (1'~ :
Figure 4.2 gives an example of a small circular hole in a large elastic plate. On
the boundary of the hole where x 2 = 0 the stress is one-dimensional.
Many particular. geometries have been analyzed either by approximate analytical
methods, photoelasticity or the numerical finite element method. Some classical
results taken from handbooks listed in the literature references are given below.
All concern isotropic elasticity.
Fig. 4.2. Stress concentration factor in a plate with a hole

- Elliptic holes in infinite plates under tension (Figure 4.3)
h I
20
--t1.
I.
Fig. 4.3.
{
fla > lO
"Infinite" plate ---+
hlb > lO
Stress concentration factor at A and in the direction of O"~
o a
K T -- 0"22
0 1+2-
0"2 b
Examples: -circular hole ~ = I ---+ KT = 3
- horizontal sharp ellipse ~ = lO ---+ K T = 21
- vertical sharp ellipse ~ 1~ ---+ KT = 1.2
-Circular holes in infinite plates under biaxial tension (Figure 4.4)
valid for -I < ~ < 1

- 0"2 -
Examples: - equibiaxial tension

I ---+ KT =2
I
o
- -~~- 0'1
Fig. 4.4.
- equal tension and compression

a( = -1 -+ KT = 4
a2
- Edge etTect in circular holes in plates under tension (Figure 4.5)
aA or B or C
22
KT aO
2
~
KT
0'0
2
4
L1
3
C
2
B
C A B
- 0
3 4 5 I/r
-1
-2
-3
-4
Fig. 4.5.
{ Kf
K~ = 3.5
1
Example: - = 2 ....... =0
r
Kfj = 3.15
- Multihole plates under biaxial tension (Figure 4.6)
\
J
\
i\ .........
r-- K2
I
I
I- 0.8 0.6 0.4

K1
0.2 o
d/b
Fig. 4.6.
Example: dlb = 0.4 a~ = 3.3a~ - la?
- Two V notches in members of rectangular section (Figure 4.7)

rib 0.05 0.10 0.15 0.20 0.25 0.30 0.40 0.50
A
0'22 {TenSion 4.1 3.0 2.5 2.2 2.0 1.9 I.7 1.55
Kr = -0-
a 22A Bending 2.9 2.2 1.9 I.7 1.6 1.5 1.4 1.3
o
a 22A
_- a 0 l
2b
Fig. 4.7.
- One V notch in member of rectangular cross section (Figure 4.8)
Approximation for hlb

A
(122
Tension -- KT = -0- 1 +2..jhF,
(122 A
Bending -- KT
Example: hlr = 4 -- KT = 5
Fig. 4.8.
- Square shoulder with fillet in rectangular bar (Figure 4.9)

A T
K - a 22 - 0.05 0.10 0.20 0.27 0.50 1.0
T - 0
a 22A
~\
Tension 0.5 1.70 1.60 1.53 1.47 1.39 1.21
1.0 1.93 1.78 1.67 1.59 1.42 1.22
1.5 1.89 1.72 1.65 1.43 1.23
2.0 1.95 1.80 1.70 1.44 1.23
3.5 2.10 1.93 1.78 1.47 1.24
A
K - a 22 T
- 0.05 0.10 0.20 0.27 0.50 1.0
T - 0
a 22A
~\
0.5 1.61 1.49 1.39 1.34 1.22 1.07
1.0 1.91 1.70 1.48 1.38 1.22 1.08
Bending 1.5 2.00 1.73 1.50 1.39 1.23 1.08
2.0 1.74 1.52 1.39 1.23 1.09
3.5 1.76 1.54 1.40 1.23 1.10
Fig. 4.9.
-Square shoulder with fillet in circular shaft subjected to torsion (Figure 4.10)
C--;c
Fig. 4.10.
K -
T -
A
a R8
a 0R8A
r 0.005 0.01 0.02 0.03 0.04 1.06 0.08 0.10 0.12
~\
Bending 2.00 3.0 2.25 2.00 1.82 1.65 1.51 1.44 1.39
1.33 2.7 1.16 1.91 1.76 1.60 1.48 1.40 1.35
1.20 3.00 2.5 2.00 1.75 1.62 1.50 1.40 1.34 1.30
1.09 2.20 1.88 1.53 1.40 1.30 1.20 1.16 1.15 1.15
- V notch in circular shaft subjected to torsion (Figure 4.11)
A
K - a R8 h
a\r
T - 0 0.5 3 5 9
a R8
0 1.85 2.01 2.66 3.23 4.54
60 1.84 2.00 2.54 3.06 3.90
90 1.81 1.95 2.40 2.64 3.12
120 1.66 1.75 1.95 2.06 2.13

0:
Fig_ 4_11.
4.1.2 Neuber's Method (Neuber 1947)
When plasticity occurs, even locally, it is much more complicated. Neuber's method
allows us to calculate the elastoplastic stress concentration from a purely elastic
calculation. It applies for all notch problems at the root of which there is a stress
concentration.
The basic heuristic hypothesis, checked for thousands of applications, postulates
that for a one-dimentional elastic problem, the product of the stress by the strain
a . at the most loaded point is equal to the same product for the same problem
analyzed in pure elasticity a E . E
I", ~ aea E ~~ I
Assuming the elastic problem to be solved, a E and E are known. The second
equation with which to calculate the two variable a and is the elastoplastic
constitutive equation. For the simple case of monotonic loading, let us take some
stress function representing the stress strain curve:
Then
a2
agc(a) = : = const.
is the equation with which to find the stress at the root of an elasto plastic notch.
A simple geometrical construction follows from the above two equations.
In the stress strain coordinate system;
a2
.-K
a
E
is an hyperbola containing the point (a = a E' aE )
E .
The other equation,
is the classical elastic strain hardening stress strain curve. The solution is obviously
given by the intersection of these two curves (Figure 4.12).
This method may be generalized straighforwardly to cyclic loading by using the
cyclic elasto plastic stress strain curve 9 c to relate the amplitude of stress and strain:
(~O"E?
~O" . ~c = ~O" E . ~c E = E
~c = 9c (~0").
The generalization to a three-dimensional state of stress is also easy if we restrict
the application to proportional loadings.
The basic heuristic hypothesis is
With the same notations as previously,
O"ijc'fj = 2w e , the elastic strain energy which has been calculated in Section 2.1.2
as:
2(I+v) 2 3(1-2v) 2
2we = 3E O"eq+ E O"H
with D = 0,
\
\
\
\
\
dE ------- Elastic solution
E Fig. 4.12. Neubers method

.
since p .IS a d
Cij ' '8i jCij
eVlator. P-O-
For monotonic proportional loading the plastic strain may be written as (see Sec-
tion 2.1.4)
3 aIJ
1.)
-2 g ( a eq ) - ,
a eq
where 9 is the hardening function, then
The Neuber hypothesis written in three dimensions is:
But a5c5 may also be written as

2(1+v)( E )2 3(1-2v)( E)2.
3E a eq + E aH ,
then
2(1+v) 2 _ 2(1+v)( E)2 3(1-2v) [( E)2_ 2].

3E aeq + g(aeq)aeq - 3E aeq + E aH aH ,
a;; being known from the elastic calculation, a eq may be determined if a~ (which
cannot be calculated without the complete analysis) is considered close to (a~)2
to make the last term negligible.
4.1.3 Finite Element Method (A. Benallal, R. Billardon, I. Doghri, 1988)
This method allows us to numerically solve the set of equations from problems
in mechanics as described in Section 4.1. It consists in replacing the problem of
partial differential equations by a problem of linear algebraic equations in which
the unknowns are the displacements of the nodes defining the finite elements into
which the structure is devided. In each element the continuum mechanical equa-
tions are solved for a class of simple displacement fields: linear, quadratic or of
higher order with space coordinates. By consequence of the piecewise approxi-
mation of the displacement, and of the strain and stress fields, we are obliged
to considerer a mesh size small enough to discretize the gradients of stress or
strain.
The algebric system to be solved is linear for elasticty, the only difficulty being
the size of the banded matrix for 1000, 10000, 100000 or even more degrees of
freedom. When plasticity and damage occur, the problem becomes nonlinear and
the only way to solve it is by time like or load incremental linearization.
It is not the purpose of this course to describe the finite element method in detail;
many good books already do. But as we are concerned with damage, two dificulties
arise: the choice of the meshes and the choice of the time or load increments.
Convergence regarding these two parameters is difficult to ensure and only heuristic
procedures may be advised:
- The algorithms used to integrate the highly nonlinear constitutive equations must
be "strong" relative to stability.
- Implicit integration schemes associated with Newton's iterative method are
preferred
- The local integration of the constitutive equations is performed by using an
"elastic predictor" to calculate the first increment as elastic and a "plastic
corrector" to ensure, by iterations, that the plasticity criterion and kinetic laws
are satisfied up to the required accuracy.
- In order to compute the tangent stiffness matrix of the structure at each it-
eration, a tangent modulus J consistent with the discretization, calculated as
baij = Jijklbckl is used instead of aij = LijklEkl. This ensures a much better
convergence.
- As to the space discretization is concern, the convergence regarding the mesh
size is of the same order of difficulty as for classical elastoplastic calculations
up to crack initiation, that is just before the localization of damage. Afterwards
the problem is no longer elliptic; its solution involves strain rate discontinuities
which cannot be modeled by classical finite elements.
- When a periodic loading is considered, as for fatigue, it is not possible to
perform the tremendous number of increments needed for all of the thousands,
or millions of cycles. A method for saving computer time involves the use of
a "jump-in-cycles" procedure which avoids calCulating the stress strain cycles
when the process is a quasi-steady state process.
Assume a piecewise periodic loading and consider a number of cycles N i for
which the damage is known to be Di. The structural calculation of this cycle (i)
is performed several times until the stabilization is reached at each point. The
coupled constitutive equations are such that for constant damage, a periodic
input implies an asymptotic periodic output. Only two or three calculations are
sufficient.
This gives (~~) i at each Gauss point of the elements of the structure, this
rate will be assumed to be constant during a certain number of cycles !:1N i,
which may be very large.
The criterion for choosing !:1N i, is a certain amount of damage !:1D at the most
damaged point M*. !:1D must not be so large as to violate the coupling, but large
enough to obtain a reasonable computer time. For applications !:1D = D lc /50
is a good compromise. Then

tlD
tlNi
( 8D) ,
8N i(M*)
where
( 8D) = Max (8D)

8N i(M') (M) 8N i(M)'
The damage at each point for the next step of the number of cycles is
Ni+ 1 = Ni + tlNi ~ Di+I(M) = Di(M) + (~~) i(M) tlNi ,

and the process may be continued in damage increments.
When needed for variable amplitude loadings, this procedure also applies for
the plastic strain, even without any damage, where the game is played with the
accumulated plastic strain p. After stabilization is reached with a cyclic plastic
increment 8p/8N, a jump of tlN cycles is performed like:
tlN = tlp
8p ,
8N
where tlp is the best compromise between accuracy and rapidity. CR being the
order of magnitude of the strain to rupture in pure tension, tlp = ~~ gives
good results. Nevertheless, be careful when using those heuristic procedures!
4.1.4 Exercise on the Stress Concentration Near a Hole
Consider a plate with a central hole of small radius in comparison to the dimensions
of the plate, which is loaded in tension by a "far" stress 0'22 = O'~ (Figure 4.13).
Assuming the material to be linear elastic, find the stress concentration factor at
the boundary of the hole (XI = r,x2 = 0) from Section 4.1.1;
K T -- 0'22
0
-
-
3.
0'2
Considering an increase of the external load O'~, plasticity may occur at the point
of stress concentration. Assuming the power strain function,
C = 9,,(0') = (:0) M for 0'22 2:: O'y'
where O'y is the yield stress Ko and M are other parameters, determine the stress
concentration factor KTfor 0'22 2:: O'y by Neuber's method;
0'22 c 22 = O'E c E ,
Fig. 4.13. Plate with a central hole
0"22)M _ (30"g)2
0"22 ( K - E '
E
KT is now a function of O"g which has to be equal to 3 for O"g

= O"y
3
Then
= 3 (SL)
30"g
~:;::
0"0
Draw the graph of KT as a function of -1. for M = 3 and for the limiting case
O"y
of perfect plasticity M = 00 (Figure 4.14).
3 + - - - -....
Fig. 4.14. Evolution of the stress con-

o 0.2 0.33 0.6 0.8 1.0 centration factor with plasticity
4.2 Uncoupled Analysis of Crack Initiation
Damage analysis of a structure or component subjected to a given history of loading

consists in the calculation of the evolution of the damage as a function of the time
at the most loaded point(s) and the critical time at which the damage reaches its
critical value corresponding to mesocrack initiation.
An assumption which simplifies the analysis consists in neglecting the coupling
between the damage and the strains. In a first step the stress and strain field histories
are calculated by a classical method. Let us suppose that the results are
c(M, t) ,
a(M, t)
everywhere in the structure loaded by external forces given as a function of
time.
4.2.1 Determination of the Critical Point(s)

The second step consists in determining the point(s) where the damage has its
maximum value. If the loading is proportional, M* is such a point where the
damage equivalent stress is maximum at any time:
a*(M*) = Max(a*(M))
M
--; M* ,
,..* -,..
v -
RI /2 , Rv =
V eq v ~(I+V)+3(1_2V)(aH)2
3 a eq
(see Section 2.1.2)

If the loading is nonproportional, the damage equivalent stresss may vary differently
at different points as a fonction of time. A small a* for a long time may be more
damaging than a large a* for a short time but less damaging than a very large
a* as an overload! There is no rigorous way to select the critical point(s) except
to calculate the evolution of the damage field in each point. This is possible, but
usually, an "intelligent" look at the evolution of a* as a function of space and time
will restrict the number of dangerous areas where the damage must be calculated.
4.2.2 Integration of the Kinetic Damage Law

Assume that at the critical point the structural calculation gives
cij(M* , t), < j(M*, t) , cfj(M*, t) ,
aij(M*, t).
It is easy to deduce:
- the accumulated plastic strain rate,
- the strain energy density release rale,

(1;q R V
l(t) = 2E(1 - D)2'
from
3 ) 1/2
(1(eq) = ( '2(18(18 '
R" ~ ~ (I + ") +3( I - 2") ( ; :

I
f
(1H = 3(1kk;
and to write the general kinetic damage law as
. Y(t,D) .
D = 5 p(t)H(p - PD) '
or
. (1;q(t)Rv(t).
D = 2ES(1 _ D)2 P(t)H(p - PD)'
The initial condition for the damage evolution is the end of the period of incubation
of micro defects corresponding to the value PD of the accumulated plastic strain
(see Section 3.1.2). Let us start the process with the initial conditions
t =0 ---> P = 0,
(1u-(1j
t = to --->
{P = PD = CP D Max ((1eq) - (1 j .
D
=0
The damage evolution is given by the integral
D t
j(1-D)2 dD = 2~S j (1;q(t)Rv(t)p(t)dt

o ~
or
t
~ [-(I-D?+l] 2~S j (1;q(t)Rv(t)p(t) dt

to
or
The critical time t R at which a crack is initiated is reached when the damage
itself reaches its critical value given by the rupture criterion (see Section 3.1.3)
(12
Dc = Dlc (1-:2 : :;
1.
Let us take here for simplicity
Dc = 1.
I
t R is given by the expression
1= 1- [I - 2~S ";q)R")p) dt]

or
J
tR
a;q(t) Rl.I (t)p(t) dt

to
where to is given by the elastoplastic analysis: to = t(p = PD)
Lower bound on the life time

In the case of strain-controlled conditions it is possible to demonstrate that the last
expression is in fact a lower bound on the critical time to rupture in comparison
with the strain-damage coupled calculation.
Assume at first the material to be elastic-perfectly plastic and the loading to be
proportional:
a eq = as = const .Vt,
Rl.I = const ."It.
The last expression becomes
J
tR
p(t) dt
to
The strain history is imposed on the mesovolume element considered; this allows
us to take p( t) as a given function, which is particularly simple to calculate if the
elastic strain is neglected:
2 ) 1/ 2
p(t) = ( "3 Ei j (tAj (t )
Then t R may be determined.

Let us now consider the coupled case of the same material loaded under the same
conditions:
- for p < p D no damage occurs; the same calculation as previously gives the
same result,
to = t(p = PD);
- the material is perfectly plastic with the same threshold as' which allows us to
write the coupled plasticity criterion as:
~ as = const. ;
l-D
- the loading is proportional:
Rl.I = const.;
- the same strain history is imposed and the elastic strain is again neglected: the
function p( t} is the same.
The critical time for crack initiation in the coupled case t'n is deduced from the
same kinetic damage law:
. a;qR v .
D = 2ES(1 _ D)2P(t}H(p-PDl
or
iJ .( }H
= a;Rv
2ES P t (p-PDl'
where (I - D)2 disappears due to the coupling in the plasticity criterion. The
integration is obvious:
t
2ES J(}d
D = a;Rv P t t.
to
Taking again D = 1 as the critical condition:
D = 1~ t = t'n,
J
t fR
2ES
p(t) dt ~R
as v
to
Comparison with the uncoupled case shows that
J J
t~ tR
p(t} dt = 3 p(t} dt,

to to
which implies that
tR < t'n.
In the particular case of monotonic loading where p = const.,
Uncoupled case Coupled case
2 ES ES
tR = to+-3~R
as vp as vP
t'n = to+2~R.
The uncoupled calculation is always conservative. It is a lower bound, but far from
the coupled solution. In other terms, component design may benefit from coupled
calculations, whkh can prove the enhanced safety of components or indicate how
light-weight economical components could be built.
The scheme of the uncoupled method is sketched in Figure 4.15.
4.2.3 Exercise on Fatigue Crack Initiation Near a Hole

Consider the same plate with a central hole as in Exercise 4.14, but now loaded
with a stress at infinity varying with the number of cycles to give rise to a low
cycle fatigue process:
os.
a2(t) "2N smwt ,
where N is the number of cycles (Figure 4.16).

Determine the number of cycles to crack initiation.
The whole process may be divided into three steps:
1. No plasticity, no damage until the number of cycles reaches the value Ny

for which the maximum stress concentration (J M = ~(J /2 equals the yield
stress (J y'
2. Plasticity occurs but no damage until the number of cycles reaches the value No
for which the accumulated l>lastic strain is equal to the damage threshold PD'
3. Plasticity and damage occur to develop a damage process up to the critical value
corresponding to the number of cycles to crack initiation N R'
Step J: 0 ~ N < Ny.

The material is elastic everywhere; from exercise 4.1.4, the stress concentration
factor is
KT = 3.
Elasto-(plastic) Kinetic
constitutive equations damage law
Structural analysis Damage mechanics
Fig. 4.15. Uncoupled analysis of crack initiation
0:0
2 0'0
2
Ad,ot,
vL
-r
I
. x1 --
rfy
o'M
Po
P D
.".
,-
,
./ _:~---
I
....,I
I I
0 Ny NO NR
Fig. 4.16. Low cycle fatigue of a plate with a central hole

The loading history is such that
Lla~ = 8N.
From the Neuber method, Section 4.1.2,
Lla= KTLla~ = 38N.

The number of cycles Ny corresponds to Lla = 2a y:
2a y
N =-
y 38
Step 2: Ny ::; N < No.
The material is elastoplastic at the free boundary of the hole (Xl = r, x 2 = 0).
Then the stress concentration factor is now a function of the stress at infinity and by
consequence, a function of the number of cycles. For a cyclic loading (Section 4.1.2
and Exercise 4.1.4),
KT = 3(3~;1) :::;: ,
M-I
2a ) M+I 2
Lla = KTLla~ = 38 ( 3: N M+I .
Using the cyclic strain-hardening power function already mentioned in Section 3.4.1
(here without any damage),
The damage threshold PD is determined by the accumulation of 2Llcp at each

cycle. Take PD = cPD for simplicity:
J
No
2 LlCp(N) dN = PD = cPD
Ny
from which
If you work with numbers for a specific material this formula will not look so bad!
Step 3: No ~ N ~ N R"
A low cycle fatigue damage occurs with a varying amplitude of stress. Take the
damage per cycle derived in Section 3.4.1:
8D K2 M2
= P LlE M
8N 4ES P
The uncoupled hypothesis assumes the plastic strain amplitude to be related to the
stress amplitude without any consideration of damage:
or, with the relation of Lla derived in step 2,
38(~)~ M
NMI.
2M
Kp
The number of cycles to crack initiation is obtained by integration of the kinetic

law of damage:
M2
2a )
38 ( - y
~:;::
8D K2 2(M2)
38
-p- N MI ,
8N 4ES
from the initial condition N = No ...... D = 0;

to the final condition D = Dc ...... N = N R'
Taking again for simplicity D, = I,
M2
2a ) ~:;::
38 ( - y
38
J
NR
N
2(M2)
MI 8N,
o No

.M.!..
- (M2) 3M5
2a )
38 ( - y
~:;::
3M5 4ES
N MI + __ 38 3M+5
o K2 M+I
p
4.3 Locally Coupled Analysis (J. Lemaitre, 1988)
Quite often, the damage is so localized that the volume of the damaged material
is small in comparison to the macroscale of the structural component and even
to the RYE. This allows us to perform an uncoupled analysis at the macroscale
and to consider the coupling between strain and damage only on the RYE of the
critical point as shown schematically in Figure 4.17. This is the case of small-scale
damage.
The uncoupled analysis differs from the coupled calculation at the level of the
constitutive equations of the critical point. The latter gives a more accurate result,
but remember that the uncoupled analysis is a lower bound on the lifetime.
The method of locally coupled analysis may often be used with good accuracy for
brittle and fatigue types of damage.
4.3.1 Localization of Damage
Damage localization results from stress concentration, of course, but also occurs,
because some weakness always exists at the microscale. Let us generalize, for any
kind of damage, what has been said for quasi-brittle materials and for high cycle
fatigue in Section 3.2.2 and 3.4.3. The mechanical model was, a two-scale volume
element, elastic or elastoplastic at the mesoscale and elastoplastic and damageable
at the microscale (Figure 3.8).
The only material characteristics which differ in the matrix and in the inclusion is
the yield stress of the inclusion a~ , which is assumed to be equal to the fatique
limit a f of the material, and its fatigue limit, assumed to be reduced in the same
proportion:
Elasto-(plastic) Coupled
constitutive equations elasto-plastic and damage
constitutive equations
Fig. 4.17. Locally coupled analysis of crack initiation

4.3 Locally Coupled Analysis 177
This takes into account the microinternal stresses and the weak defects always
existing everywhere in all materials. As the fatigue limit is smaller than the yield
stress, it allows for plasticity which induces damage at the microscale, whereas the
matrix remains elastic or elastoplastic and undamaged for the same loading.
The second assumption which simplifies calculations is the Lin-Taylor strain com-
patibility hypothesis, which states that the state of strain at the microscale is equal
to the state of strain at the mesoscale as derived from the classical structural cal-
culation,
g'" = g
Then, there is no boundary value problem to be considered. Only the set of coupled
constitutive equations must be solved for the given history of the strain.
As in Section 3.2.2, the crack initiation condition at the microscale D = Dc is
also the brittle crack initiation at the mesoscale G = G c.
4.3.2 Postprocessing of Damage Growth (c. Lienard, 1989; I. Doghri, 1990)
As the input of the locally coupled method is the strain at some point, it can be
set up as a postprocessor to any finite element code of structural calculations. The
most sophisticated constitutive equations to be used are:
- the set of equations developed in Section 2.1.4 for linear elasticity and for
isotropic and kinematic plastic hardening,
- associated with the kinetic damage law developed in Section 3.1;
- in the case of unilateral damage microcrack closure conditions as explained in
Section 2.3.4.
Such a postprocessor exists in the French Framasoft + CSI SYSTUS Finite

Element Code.
It is also possible to develop a simpler code by assuming the material to be perfectly
plastic at the micro-scale and by neglecting the microcrack closure effect. The
equations are those of Section 3.1.4 with the conditions of quasi-brittle damage
developed in Section 3.2.2.
Let us recall that
(J eq
1 _ D - (J~ = 0 and f = 0; if not, ifj = 0,
IL2
D=~R . if not, D = 0;
2ES vP
2
Crack initiation if De = Dl e f -
(J~ Rv
These equations may be used for piecewise perfect plasticity by considering several
values of the plastic threshold (J~ as the loading or the timelike parameter vary.
This allows us to take some cases of high values of strainhardening and the cyclic
stress strain curve for multilevel fatigue processes into consideration.
Then, the material parameters must be considered as follows:
* E and v for elasticity;

* (J f' (J y and (Ju for plasticity:
* (J~ = (J f for pure plasticity;
* (J~ = (J si given as input for piecewise plasticity;
* S , c PD' DI e for damage.
The input of the calculation is the time history of the strain components Cij (t)
which may come from the result of a finite element structural calculation or be
given directly. Note that the set of equations does not depend explicity upon time.
(This would not be the case for viscoplasticity). The reference for history is a
timelike parameter which accounts for the succession of increments of the loading
only.
The outputs are:
- the timelike parameter t,

- the damage evolution D(t) , the last point corresponding to D De' that is,
crack initiation,
- the accumulated plastic strain evolution p(t) ;
- the evolution of the von Mises equivalent stress (Jeq(t) and the damage equiv-
alent stress (J' (t);
- the evolution of the stress components (Jij (t) and the input strains Cij (t).
The numerical procedure is a strain-driven incremental timelike algorithm using an

elastic predictor and a plastic corrector. The hypothesis of perfect plasticity, even
coupled with the softening effect of damage, allows us to explicitly formulate this
plastic correction. The constitutive equations are discretized in an incremental form
corresponding to a fully implicit integration scheme having unconditional stability;
but since the unknowns are explicitly updated, it has the advantage of an explicit
scheme in which no linear system needs to be solved.
After an elastic increment given by the law of elasticity, the plasticity criterion f :::;
o is checked. If f > 0, the plasticity corrections are obtained by Newton's iterative
procedure applied to a system of two equations deduced from the constitutive
equations:
f = aeq - (J~ = 0,
with >. = Ev E
(I - 2v)(1 + v) J.l = 2(1 + v)'
Efjn is the plastic strain at the beginning of the increment, t:::.p is accumulated
plastic strain increment, and the other quantities are the values at the end of the
increment.
The iteration operates on
J+ ::'CfJ
'J
= 0,
&h &h
hij 'J 'J C = 0 ,
C kl+-&
lI
+-&- p
akl P
where CfJ and C p may be explicitly calculated. The resulting expressions are
C =
J -Nh
'J 'J 'th N .
p 3J.l ' WI 'J
4.3.3 Desciption and Listing of the Postprocessor DAMAGE 90

(I. Doghri, 1990)
DAMAGE 90 is a friendly computer code with fewer than 600 FORTRAN instruc-
tions; it calculates the evolution of damage up to crack initiation with the method
described in Section 4.3.2. It may be used in an interactive way either as a postpro-
cessor of the strain output of a finite element code or as an autonomous program.
It is written in FORTRAN 77 as available on a Digital V AX computer. The CPU
time for one complete execution is a few seconds. It distinguishes between two
loading cases:
- general loading history where the history of loading is defined by the values
of the strain components at given timelike parameter values. DAMAGE 90
interpolates linearly between these values;
- piecewise periodic loading for which the loading is a certain number of blocks
of cycles defined by the two consecutive maximum and minimum set of strain
components and the number of cycles in each block. The strain interpolations are
also linear. For a large number of cycles, the jump-in-cycles procedure (which
alows "jumps" of cycles to save computer time as explaned in Section 4.1.3)
may be used. It is capable of accounting for initial values of damage Do and
plastic strain Po'
t 80 Analysis of Crack Initiation in Structures
The questions asked by DAMAGE 90 to introduce the input data and the charts of
the output data as they appear on the screen of the terminal are given below. (The
subscript f..l is omitted a!' == as).
The complete listing is also given hereafter for those of you who wish to play with
DAMAGE 90.
DAMAGE 90 Input Data

- Example of the first exercise in Section 4.3.4 .
* * * **a.* *I*III II.*I.* ** *

Give material constants and the strains history
UI DAMAGE91) u,
wI II give you the damage gl'owth up to crack I nit I at Ion
***111111111111111111111111111.11111111111111111111111
u ELASTICITY . Give
YOUNG's modulus
200 . E3
POISSOW s I'at. 10
0.3
I I PERFECT PLASTICITY: plastic threshold SIGs given with loading. Give
Fatigue limit. SIGf
21)1).
Yield st.ress SIGy
31)1).
Ultlmat.e str'ess SIGu
51)0.
DAMRGE EVOLUTION dD (Y/ S ) dp . Give : ~;
13.El6
u DAI1AGE THRESHOLD dD =13 If p(pD. Do you know I'D? 'Y' or 'Ii'
Y
Give t.he value of pD :
10.E-2
CRACK INITIATION : D'Dc . Do you know Dc ? 'Y' or ' N'
Y
Give the va I ue 0 f Dc (remember : 13 ( Dc ( 1) :
0.99
U : INITIAL CONDITIONS Give :
The value of po
0.
The value of Do
D.
H LOADIIiG
Is t.he stress state uniaxial? ' Y' or ' N'
Y
Is t.he st.raln history cyclic? 'Y' or 'N'
Ii
tnn YOUR LOADIIiG IS liOT CYCLIC.
Give lhe numbe,' 0 f po i nls wh i ch de fine lhe h lslo,'~! (21313 ma>: I

5
Give lhe values of time al lhese poinls :
13. l 1 ... 2 . I 3 . , 4.
Give lhe values of 11 -strain al these limes
0. 0.25E-2 . 1.5E-2 5.E-2 , 25.E-2
Give lhe values of the plastic lhreshold sl,'ess SIGs crl these limes :
21313 . 31313 . 4130 . 51313 . 5130.
Suggest. an init.ial lime increment. < to inler. \... cd Ie bet.ween fir' sl 2 points
0. 1
III CRACK INITIATION .
HI THE JOE: IS EliDED . YOUR RESULTS FILES ARE
di re cl.out shew.... out

FORTRAli STOP
CLR
- Example of a low cycle fatigue tension case at the macroscale strain-imposed
Ell = 0.425%, with material data of the aluminum alloy considered in the
exercises in Section 4.3.4.
Itl I I II.II.IIIII.II.I.III 1
Give male,' i cd conslanls and lhe sl,'a ins h i slo,'~!
", i 11 1 i lie ~ou lhe damage wol~lh up lo c,'ack in i II ell i on

111111.111111 11111111111111111,1111111'11111, ,1
II ELASTICITY. Give
YOUNG ' e modu 1us
72.E ... :::
POISSOW s "'al io
0.32
I I PERFECT PLASTICITY: plastic lhr eshold SIGs given wilh loading. Gille
Fatigue limil SIGf
303 .
Yield stress SIGy
3136.
Ultimate slress SIGu
500.
:1:1 DAI'1AGE EVOLUTIOli dD (Y/ S) dp . Give : S
6.
u Dl'li'1AGE THRESHOLD dDO if p<pD. Do you know pO? ''I'' or' "1'1 '
Ii
Give EpD in lens ion: pDEpDI(SIGu -S IGf) / [SIGs- IS IGfl.2/ SIGyIJ
10. [-2
II CRACK INITIATION : DDc . Do you know Dc ? 'Y' or ' N'
Ii
Give D1e in lension Dc D1e I [(SIGu/ SIGs).12J / Rv
13 . 99
INITIAL CONDITIONS Give:
The value of po
0.
The ....J'due of Do
0.
U : LOADIIiG
Is lhe sl,'ee.s st.ale uniCl xia l ? ""1" or' ' 1'1 '
N
Is lhe sl,'a in h i 5lo,'~! eye 1 ic ? ",// or- ' Ii'
v
nu* YOUR LOADIIiG IS CYCLIC.

Do lJou wish the jump in clJcles procedul'e fOI' Iw'ge Ii ? 'Y' or 'Ii'
Y
Give the number of b locks of constant amp I i tude (50 ma):)
Give the numbel' of c~lcles fOI' block:

120000
Give for each st.'ain component 1'1AX and min ,'a lues :
1st peak 2nd peak of 11-strain
0.425E-2 , -0.425E-2
Give for each stra in component 11AX cmd mi n va lues
-0.136E-2 , 0.136E-2
Give for each 5t.'ain component ~1AX w,d min vcdues
1st peak ,2nd peak of 33-strain
-0.136E-2 , 0.136E-2
Give fOI' each stra in component 1'1AX and min va lues
0. , 0.
Give fOI' each strain component 1'1AX w1d min vcdues
e. , e.
Give fOI' each strain component ~1AX and min value5
1st peak 2nd peak of 23-5train
e. 0.
Give the value of the plastic threshold stress SIGs for thi5 block
3133.
Sugges t a numbel' 0 f i nCI'ements pel' clJc Ie (m i n i mum : 4) :
4
*** CRACK INITIATION.

n! THE JOB IS EliDED. YOUR RESULTS FILES ARE :
direct.oklt shear'. out fatigue.out

FORTRAN STOP
CLR
DAMAGE 90 Output Data

- Partial results of example of the first exercise in Section 4.3.4.
TIME DAMAG P MISS Q. STRESS DAMAG Q. STRESS

0.0000+00 0 . 0000+00 0 . 0000+00 0.0000+00 0.0000+00
0 . 1000+00 0.0000+00 0.0000+00 0.5000+02 0.5000+02
0.2100E+00 O. OOOOE+OO 0.0000+00 0.1050+03 0.1050E+03
0.3310E+00 O.OOOOE+OO O. OOOOE+OO 0.1655E+03 0 . 1655+03
0.4641E+00 O. OOOOE+OO 0 . 1603-03 0.2000E+03 0.2000E+03
0.6105E+00 O.OOOOE+OO 0.5263E-03 0.2000E+03 0 . 2000E+03
0.7716E+00 O.OOOOE+OO 0.9289-03 0 . 2000+03 0.2000E+03
0.9487+00 O.OOOOE+OO 0 . 1372-02 0.2000E+03 0 . 2000E+03
0.1000+01 0.0000+00 0.1500E-02 0 . 2000E+03 0.2000+03
0.1056+01 O.OOOOE+OO o. 1705E-02 0 . 3000E+03 0.3000E+03
0.1118+01 O.OOOOE+OO 0.2481E-02 0.3000+03 0.3000E+03
0.1187E+01 O. OOOOE+OO 0.3334E-02 0 . 3000+03 0.3000+03
0 . 1262+01 0.0000+00 0.4273E-02 0.3000E+03 0 .3 000E+03
0.1344+01 0 . 0000+00 0 . 5305-02 0.3000+03 0.3000+03
0.1435+01 0.0000+00 0.6441-02 0.3000+03 0.3000+03
0.1535+01 O.OOOOE+OO 0 . 7690-02 0 . 3000E+03 0 . 3000+03
0 . 1645+01 O.OOOOE+OO 0.9064E-02 0.3000E+03 0 . 3000E+03
0.1766E+Ol O.OOOOE+OO 0.1058-01 0.3000+03 0 . 3000+0"3
0.1899+01 0.0000+00 0.1224-01 0.3000+03 0 . 3000+03
0. 2000+01 0.0000+00 0.1350-01 0 . 3000+03 0.3000+03
0.2111+01 O. OOOOE+OO 0.1689-01 0.4000+03 0 . 4000+03
0.2233+01 0.0000+00 0.2116-01 0 . 4000E+03 0 .4 000E+03
0.2368+01 O.OOOOE+OO 0.2586-01 0.4000E+03 0.4000E+03
0 . 2515+01 O.OOOOE+OO 0 . 3104E-Ol 0 . 4000E+03. 0 . 4000E+03
0.2605E+01 O.OOOOE+OO 0 . 3417E-Ol 0.4000E+03 0 .4 000E+03
0.2703+01 O.OOOOE+OO 0.3761E-Ol 0.4000E+03 0 . 4000E+03
0 . 2811E+01 O.OOOOE+OO 0.4140E-Ol O. '1000E+03 0 . 4000+03
0. 2930E+01 O. OOOOE+OO 0 . 4556E-01 0.4000E+03 0.4000E+03
0 . 3000E+01 O. OOOOE+OO 0 . 4800E-01 0 . 4000E+03 0.4000E+03
0.3023E+01 0.0000+00 0.5213-01 0.5000+03 0 . 5000+03
0 . 3049+01 0.0000+00 0 . 5723-01 0 . 5000+03 0.5000E+03
0 . 3077E+01 O. OOOOE+OO 0.6284E-01 0.5000E+03 0 .5000E+03
0 . 3108+01 0.0000+00 0.6901E-01 0.5000E+03 0.5000E+03
0.3126+01 O.OOOOE+OO 0.7274E-01 0.5000E+03 0.5000E+03
0.3147E+01 0.0000+00 0.7684E-01 0.5000E+03 0 . 5000+03
0 . 3169+01 0.0000+00 0 . 8136E-01 0 . 5000+03 0 . 5000+03
0.3194E+01 0.0000+00 0 .86 33E-01 0.5000E+03 0.5000E+03
0 .32 21E+01 0.0000+00 0.9179-01 0 .5 000E+03 0.5000+03
0.3252+01 0.0000+00 0 . 9780-01 0 . 5000+03 0 .5000+03
0.3270+01 0.3788-01 0 . 1014+00 0.4811+03 0.4811+03
0.3290E+01 0 .7 954-01 0.1054+00 0 . 460 2E+03 0.4 60 2+03
0 . 3312+01 0.1254+00 0 . 1098+00 O. 'l373E+03 0 . 4373E+03
0 . 3336+01 0.1758+00 0 . 1147+00 0.1121+03 0.4121E+03
0.3362E+01 0.2312+00 0.1200+00 0 .38 44+03 0 . 3844+03
0.)392E+01 0 . 2922+00 0.1259+00 0.3539+03 0.3539E+03
0 . 3424E+01 0.3593+00 o. 1323E+OO 0 . 3203E+03 0 . 3203+03
0 . 3443+01 0 . 3999+00 0 . 1362E+00 0 . 3000E+03 0.3000+03
0.3465E+01 0.4446+00 0.1405E+00 0 . 2777E+03 0.2777E+03
0 . 3488E+01 0 . 4937+00 0 . 1452+00 0 . 2531+03 0 . 2531+03
0.3514E+01 0 . 5477+00 0.1504+00 0.2261+03 0.2261E+03
0.3543E+01 0 .6 072+00 0 . 1561+00 0 . 1964+03 0.1964+03
0.3574+01 0.6725+00 0.1624+00 0.1637+03 0 . 1637E+03
0.3593E+Ol 0 . 7121E+00 0.1662E+00 o. 1439E+03 0 . 1439E+03
0.3614E+Ol 0.7556E+00 0.1703E+00 o.1222E+03 0 . 1222E+03
0.3637E+01 0 . 8035E+00 o. 1749E+00 0.9826E+02 0 .9826+02
0.3662E+Ol 0 . 8561+00 0 . 1800E+00 0.7194+02 0.7194+02
0.3690E+Ol 0.9140E+00 0 . 1855E+00 0.4299+02 0.4299E+02
0 .3 721E+01 0.9777E+00 0.1917E+00 o. 1113E+02 0.1113+02
0 . 3739E+Ol 0 . 1016+01 0 . 1954E+00 -.8136E+01 -. 313 GC +Ol
- Partial results of the low cycle fatigue in tension II = O.425%

CYCLE DAMAGE P MIStS EQ . STRESS DAMAGE EQ. STIIESS
O.OOOOE+OO O.OOOOE+OO 0.0000&+00 0.0000&+00 0.0000&+00
0 . 2500&+00 0.0000&+00 0.3667&-04 0.3030E+03 0.3034E+03
0.5000&+00 0.0000&+00 0.3667E-04 0.3000E+01 0.2814E+01
0 . 7500E+00 0.0000&+00 0.1100E-03 0.3030E+03 0 . 3034E+03
0.1000E+01 O.OOOOE+OO 0 . 1100E-03 0 . 3000E+01 0.2814E+01
o.4547E+05 O.OOOOE+OO 0.6669E+01 0.3030E+03 0.3031E+03

0.4547E+05 O.OOOOE+OO 0.6669E+01 0 . 3000E+01 0.2314&+01
0 . 4547E+05 0 . 7811E-05 0 . 6670E+01 0.3030&+03 0.3034E+03
0.4547E+05 0.7811&-05 0 . 6670E+01 0.3000E+01 0.2814E:+01
0.4547E+05 0.1562&-04 0 . 6670E+01 0 . 3030E+03 0.3034&+03
0.4548E+05 0 . 1562E-04 0.6670E+01 0.3000&+01 0.2814E+01
0.4640E+05 0.1450&-01 0 . 6806E+01 0.2986E+03 0.2990&+03
0.4640E+05 0.1450E-01 0.6806&+01 0.2957E+01 0.2773E+01
0.4640E+05 o. 1450E-01 0.6806E+01 0.2986&+03 0.2990E+03
0.4640E+05 o. 1450E-01 0.6806E+01 0.2957&+01 0 . 2773&+01
0.4640E+05 o. 1451E-01 0 . 6806E+01 0.2986E+03 0.2990E+03
0.4640E+05 o. 1451E-01 0.6806E+01 0.2956E+01 0.2773E+01
0.4640E+05 0.1451E-01 0.6806E+01 0.2986E+03 0.2990&+03
0.4640E+05 0 . 1451E-01 0 .6806&+01 0.2956E+01 0 . 2773E+01
0.4733E+05 0.2899E-01 0.6942&+01 0.2942E+03 0 . 2946E+03
0.4733&+05 0.2899E-01 0.6942E+01 0.2913E+01 0.2733&+01
0.4733E+05 0.2899E-01 0 . 6942E+01 0.2942E+03 0.2946E+03
0.4733E+05 0 . 2899&-01 0.6942E+01 0 . 2913&+01 o. 2733E+01
0.4733E+05 0.2900E-01 0.6942E+01 0.2942E+03 0.2946E+03
0.4733E+05 0.2900E-01 0.6942E+01 0.2913&+01 0.2733E+01
0.4733E+05 0.2901E-01 0 . 6942E+01 0.2942E+03 0.2946E+03
0.4733E+05 0.2901E-01 0.6942E+01 0.2913E+01 0 . 2733E+01
0.4826E+05 0.4349E-01 0.7078E+01 0.2898E+03 0 . 2902E+03
0.5569E+05 0.1595E+00 0 . 8168E+01 0.2547&+03 0.2550E+03

0.5569E+05 0.1595E+00 0.8168E+01 0.2522E+01 0.2365E+01
0.5662E+05 0.1740E+00 0.8304E+01 0.2503E+03 0.2506E+03
0.5662E+05 0.1740E+00 0.8304E+01 0 . 2478E+01 0.2325E+01
0.5662E+05 0 . 1740E+00 0.8305E+01 0.2503E+03 0 . 2506E+03
0 . 5662E+05 0.1740E+00 0.8305E+01 0.2476E+01 0.2325E+01
0.5662E+05 0.1740E+00 0 . 6305E+01 0.2503E+03 0 . 2506E+03
0.5662E+05 0.1740E+00 0.8305E+01 0.2478E+01 0.2325E+01
0.5662E+05 0.1740E+00 0.8305&+01 0.2503&+03 0.2506E+03
0.5662E+05 0 . 1740E+00 0 . 8305E+01 0.2478E+01 0 . 2325&+01
0.5755&+05 0.1885E+00 0 . 8441E+01 0.2459E+03 0.2462E+03
0.5755E+05 0.1885E+00 0.8441&+01 0.2435E+01 0.2284E+01
0.5755E+05 0.1885E+00 0 . 8441&+01 0.2459E+03 0.2462E+03
0.5755E+05 0.1885E+00 0 . 8441E+01 0.2435E+01 0.2284E+01
0.5755E+05 o.1885E+00 0.8441E+01 0.2459E+03 0.2462E+03
0.5755&+05 o.1885E+00 0.8441&+01 0 . 2'135&+01 o. 22~ L.!. E+Ol
0.5755E+05 0.1885E+00 0.8441&+01 0.2459&+03 O.24G 2E+03
0.5755&+05 0.1885&+00 0.8441&+01 0 . 2435&+01 O.2234E+Ol
0 . 5848&+05 0.2030&+00 0.8577&+01 0.24151::+03 0.2 4181::+03
0 . 5848&+05 0.2030&+00 0.8577&+01 0.2391E+01 0.2243E+01
0.5848E+05 0.2030E+00 0.8577&+01 0 . 2415E+03 0.2 118E+03
0 . 5848E+05 0 . 2030E+00 0.8577E+01 0 . 2391E+01 0.22'131':+01
0.5848&+05 0 . 2030E+00 0 . 8577&+01 0.2415E+03 0.2 418E+03
0.1077&+06 0.9713E+00 0.15801':+02 0.8686E+01 0 . 8696E+01

0 . 1077E+06 0 . 9713E+00 0.1580&+02 0.8600E-01 0 .8067&-01
0.1086&+06 0 . 9858E+00 0.1593E+02 0 . 4298&+01 0.4303E+01
0.1086E+06 0.9858E+00 0 . 1593E+02 0 .4 255&-01 0.3992E-01
0.1086E+06 0.9858&+00 0 . 1593&+02 0.4298&+01 0 . 4303E+01
0.1086E+06 0.9858E+00 0.1593E+02 0 .4 255&-01 0.3992E-01
0.1086E+06 0.9858&+00 0.1593E+02 0.4296&+01 0.4301E+01
0.1086E+06 0.9858E+00 0.1593&+02 0.4253&-01 0.3990&-01
0.1086&+06 0.9858&+00 0.1593E+02 0 . 4293E+01 0.4298E+01
0.1086&+06 0 . 9858E+00 0.1593E+02 0.4251&-01 0.3987E-01
0.1096E+06 0.1000E+01 0.1607&+02 - . 9480E-01 - . 9492E-01
Listing of DAMAGE 90
c**********************"'*"*""',.
C'" DAMAGE 90
C Elastic-perfectly plastic law coupled to a ductile damage model.
C'l' Fully implicit integration scheme.
C'2' Jump in cycles procedure.
C'" Written by Issam Doghri
C U. C. SANTA BARBARA, DEPT . Of MECHANICAL ENGINEERING
C'" Version : May 1990
C"*'It*'ltlt""'***"*"'*""*"'*"*'**"""'*""** **'*""'*'*'"
IMPLICIT REAL'S (A-H,o-Z)
CHARACTER'l UNIAXI,ANsPD,ANsDC,CYCLIC,JUMP,CONVER,sTAB,COUPL
CHARACTER'20 flLE1,flLE2
CHARACTER' 6 COMMENT
PARAMETER(NTENs=6,NSTATV=ll)
REAL'8 sTRS(NTENS) ,sTATEV(NsTATV) ,STATEVI (NsTATV),
sTRAN(NTENs) ,DsTRAN(NTENS) ,sIGHI (6) ,SIGHf(6),
INCUSE(50),TIM(200),HIsT(6,200),EPsB(6,50),EPss(6,50),
PERDIV(6),sYIEL(50)
INTEGER ISLOPE(6),IPER(6),INDEX(6),NBCYCL(50)
COMMON/ETIQ1jEO,XNU,sIGS,SMPD,SO
COMMON/ETIQ2/0NE,TWO,THREE
DATA INDEX 111,22,33,12,13,231
C'" READ DATA
MS=6
MB=5
NAXI=6
NCYCLE=O
sMPD=l.D+IO
DCRIT=O . 99DO
WRITE (MS , ')'
WRITE (MS , ')'
WRITE(MS,"')' ft. It". **10 It **, ***,. * ,** **11 *It.,. *"" **'" ,," ft** "It, #tit.
WRITE(MS,'), Give material constants and the strains history'

WRITE (MS , ')'
WRITE(MS,*)' It'"~ DAMAGE90 It.ftl
WRITE (MS , ')'

WRITE(MS,')' will give you the damage growth up to crack initiatio
.n'
WRITE(MS,') ,
WRITE(MS,')
WRITE(MS, *)' It* ELASTICITY . Give
WRITE(MS,') , YOUNG"s modulus
READ(MB,' )EO
WRITE(MS,') , POISSON"s ratio
READ(MB,')XNU
WRITE(MS,')'" PERfECT PLASTICITY plastic threshold sIGs
. given with loading . Give :'
WRITE(MS,'), fatigue limit SIGf
READ(MB,' )SIGf
WRITE(MS,'), Yield stress sIGy
READ(MB,' )SIGY
WRITE(MS,'), Ultimate stress SIGu .,
READ(MB,')sIGU
WRITE(MS,')'" DAMAGE EVOLUTION dD = (Y/S) dp . Give S'
READ(MB,')SO
FORMAT (A)
WRITE(MS,')'" DAMAGE THRESHOLD dD=0 if p<pD . Do you know pD?
, ,y" or ' 'N'"
READ(MD,l)ANsPD
If( (ANSPD.EQ. 'y') .OR. (ANSPD.EQ. 'Y') )THEN
WRITE(MS,')' Give the value of pD :'
READ(MD,')sHPD
If(SMPD.EQ.0.)SMPD=1.D-6
ELSE If( (ANsPD.EQ. 'n') .OR. (ANSPD.EQ. 'N') ITHEN
Give EpD 1n tens10n :

pD=EpD'(SIGU-SIGf)/(SIGs-(SIGf"2/SIGy))'
READ(MB.)EPD
IF(EPD.EO.0.)EPD=1.D-6
ELSE
STOp WRONG DATA . GOOD BYE . '
END IF
WRITE(MS.) CRACK INITIATION: D=Dc . Do you know Dc ? "Y"
or ' 'N'"
READ(MB.l)ANSDC
IF( (ANSDC.EO . 'y') .OR. (ANSDC.EO. 'Y') )THEN
WRITE(MS.) Give the value of Dc (remember o < Dc <1) .
READ(MB.)DCRIT
ELSEII'( (ANSDC . EO. 'n') .OR. (ANSDC.EO. 'N') )THEN
WRITE (MS ). Give Dlc in tension
Dc = Dlc (SIGu/SIGs)**2) / Rv
READ(MB.)DlC
ELSE
STOP' WRONG DATA . GOOD BYE . '
END IF
WRITE(MS.) INITIAL CONDITIONS. Give
WRITE (MS ). The value of po
READ(MB. )SMPO
WRITE (MS ). The value of Do
READ(MB.)DO
WRITE(MS.) LOADING'
WRITE(MS.) Is the stress state uniaxial , 'V" or ' 'N'"
READ(MB.l)UNIAXI
IF (UNIAXI .NE. 'y' .AND. UNIAXI .NE. 'n' .AND.
UNIAXI.NE.Y.AND.UNIAXI.NE.N)STOp WRONG DATA.
GOOD BYE .
IF( (UNIAXI .EO. 'y') .OR. (UNIAXI .EO. 'Y')) NAXI=l
WRITE(MS.) Is the strain histol~ cyclic? "Y" or "N'"
READ(MB.l)CYCLIC
IF(CYCLIC.NE . y.AND.CYCLIC.NE.n .AND.
CYCLIC . NE . Y.AND.CYCLIC.NE.N)STOp WRONG DATA .
. GOOD BYE.***'
IF( (CYCLIC.EQ. 'y') .OR. (CYCLIC . EO. 'Y') )THEN
WRITE(MS.)
WRITE (MS , .. )' ***** YOUR LOADING IS CYCLIC . '
WRITE (MS )'
WRITE(MS,), Do you wish the jump in cycles procedure for
1aLge N? ,ty" or "N'"
READ (MB, 1 )JUMP
IF (JUMP .NE. 'y' .AND.JUMP .NE. 'n' .AND.
JUMP . NE . 'Y' .AND.JUMP . NE . W )STOP' WRONG DATA.
GOOD BYE.***'
WRITE(MS,) Give the number of blocks of constant amplitude
(SO max)'
READ(MB,')NBLOCK
DO IBL=l,NBLOCK
WRITE (MS, ) , Gi ve the number of cycles for block :', IBL
READ(MB , ' )NBCYCL( IBL)
NcyCLE=NCYCLE+NBCYCL(IBL)
DO IST=l, NAXI
WRITE(MS,*)' Give for each strain component MAX and min values
WRITE(MS,3)INDEX(IST)
FORMAT(' 1st peak 2nd peak of ',12,'-strain')
READ(MB.)EPSB(IST,IBL),EPSS(IST,IBL)
END DO
WRITE(MS.) Give the value of the plastic threshold stress
SIGs for this block :'
READ(MB,)SYIEL(IBL)
WRITE(MS.)' SU9gest a number of increments per cycle
(minimum : 4) :
READ(MB,)INCUSE(IBL)
END DO
ELSE IF( (CYCLIC.EO . 'n') .OR. (CYCLIC. EO . 'N') )then
WRITE (MS , .)'
WRITE(MS.) YOUR LOADING IS NOT CYCLIC . '

WRITE(MS.)
WRITE(MS.) Give the number of points which define the histor
.y (200 max)
R&AD(MB.)NGIV
WRITE(MS.) Give the values of time at these points .
R&AD(MB.) (TIM(IG).IG=l.NGIV)
DO IST=l. NAXI
WRITE(MS.2)INDEX(IST)
FORMAT(' Give the values of '.12.' -strain at these'
. ' times !'}
R&AD(MB,') (HIST(IST.IG) ,IG=l.NGIV)
&NO DO
WRITE(MS.) Give the values of the plastic threshold stress S
~IGs at these times : '
R&AD(MB . ) (SYIEL(IG).IG=l.NGIV)
WRITE(MS , '), Suggest an initial time increment to
,intervalle between first 2 points'
R&AD(MB,' )DTIME
&NO IF
C''INITIALIZE . STATEV= r,p,D,dD,(pstrn),dp
DO ISTzl, NSTATV
STATEVI(IST)=O .
END DO
STATEVI(2)=SMPO
STATEVI (3 )=00
STATEVI(4)=0 .
DO IST=1,6
STRAN(IST)=O .
DSTRAN(IST)=O.
END DO
IF( (lINSPD . EQ. 'y') .OR. (lINSPD.EQ. 'Y') ) EPD=SMPD
ENERG1=(SIGU-SIGF)'EPD
ENERGO=O.
ENERG3=O.
TOLENE=2. S~2
ILOOKO=O
COUPL='y'
IPASS=O
ONE=l . Df.OO
'1'WO=2.Df.OO
THREE=3. Df.OO
RAC2=DSQRT('1'WO)
NUl=l1
NU2=12
FlLE1='direct.out'
FILE2='shear.out'
OPEN(UNIT=NU1,name=FILE1,status='unknown' )
OPEN(UNIT=NU2 , name=FlLE2,status='unknown')
COMMENT=' TIME'
IF( (CYCLIC . EQ. 'y') .OR. (CYCLIC . EQ. 'Y') )COMMENT= ' CYCLE'
WRITE(NU1,3l2)COMMENT
WRITE(NU2,3l2)COMMENT
WRITE(90,31S)COMMENT
312 FORMAT ( 2X,A6,21X, 'STRAINS ' ,27X, 'STRESSES' ,16X, 'DAMAGE',
. 9X, 'p' ,SX, 'MISES' ,2X, 'DAM.EQ .STRESS ')
WRITE (NUl ,313)
WRITE (NU2, 314)
313 FORMAT(l9X,' 11' ,9X,' 22' ,9X,' 33' ,lOX,' 11' ,9X,' 22' ,9X,' 33' ,
11X, '0' ,23X, 'SIG eq' ,SX, 'SIG*')
314 FORMAT ( 19x,' 12' ,9X,' 13' ,9X,' 23' ,lOX,' 12' ,9X,' 13' ,9X,' 23' ,
11X, '0' ,23X , 'SIG eq' ,SX , 'SIG")
315 FORMAT(SX,A6,lOX ,' DAMAGE ' ,SX,'p' ,SX,'MISES EQ . STRESS'
,2X,'DAMAGE EQ.STRESS')
C'" IF TilE LOADING IS NOT CYCLIC
IF( (CYCLIC.EQ. 'n') .or . (CYCLIC . EQ . 'N') )THEN
JUMP='n'
TIME=TIM(l)
DTMIN=DTIME/10000.
TFIN=TIM(NGIV)
IGIV=l
CALL OtrrPtrr (NUl. NU2 TIME. STnAN. STRS STATEV NTENS NSTA'1'V
STRSB.STAR)
~IILE(TIME+DTIME.LE . TFIN)
DO IST=l. NAXI
DSTnAN( IST)=DTIME' (IIIST( 1ST. IGIV+l )-IIIST( 1ST. IGIV) )
/(TIM(IGIV+l)-TIM(IGIV))
END DO
CONVER='y'
DO IST=l.NSTATV
STATEV(IST)=STATEVI(IST)
END DO
SIGS=SYIEL(IGIV)
IF(SIGS . LT.SIGF.OR.SIGS . GT.SIGU)STOp .. PLEASE REVIE:W TilE
VALUES OF SIGf SIGu and SIGs .'
CALL INTEGR(STnAN.DSTnAN.NTENS.NSTATV.UNIAXI.COUPL .
& CONVER.STRS.STATEV.STRSB.STAR.RNU)
C'" IF NO CONVERGENCE. DIVIDE TIME INC. BY 2
IF(CONVER.EQ.y . AND . STATEV(ll).GT.O.)THEN
ENERG2=ENERG3+(SIGS-SIGf'(SIGF/SIGY))'STATEV(11)
IF(ANSDC . EQ.n)DCRIT=DIC((SIGU/SIGS)2)!RNU
IF(ILOOKO. EQ. 0 .AND. (ANSPD. EQ. 'y' .OR.ANSPD. EQ . 'Y' )
.AND.STATEV(2).GT.l . OSSMPD.OR.
ILOOKO . EQ.O .AND. (ANSPD. EQ . 'n' . OR .ANSPO. EQ . N' ) .111'10.
ENERG2 .GT . 1. OSENERG1 . OH .
STATEV(3).GT . l . OSDCHIT)TIIEN
CONVER='n'
END IF
END IF
IF(CONVER.EQ.n)THEN
IPASS=IPASS+l
DTIME=DTIME/2.
C'" IF CONVERGENCE
ELSE
IF( (ANSDC.EQ . 'n' . OR . ANSDC . EQ . 'N')
.AND . DCRIT.GT.O.99DO)DCRIT=O.99DO
IPASS=O
DO IST=l.NSTATV
STATEVI(IST)=STATEV(IST)
END DO
DO IST=1.6
STRJIN ( 1ST) =STHIIN ( 1ST) +DSTRJIN ( 1ST)
END DO
TIME=TIME+DTIME
CALL OtrrPtrr (NUl. NU2 TIME. STRAN. STRS. STATEV NTENS NS1ATV
STRSB.STAR)
IF(STATEV(3).GE . DCRIT)THEN
WH1TE(MS.)
WHITE(MS.) CRACK INITIATION.'
GO TO 308
END IF
IF(STATEV(3).GE.l.l)STOp DAMAGE CANNOT ~XCEED 1 . '
C'" FIND pD
IF(ILOOKO.EQ.O .AND . STATEV(ll) . GT.O. )THEN
ENERG3=ENERG2
IF( (ANSPD.EQ. 'n' .OR.ANSPD.EQ.W)
.AND. ENERG3 .GE.ENERGl )TIlEN
SMPD=STATEV(2)
ILOOKO=l
ELSE IF( (ANSPD.EQ. 'y') .OR. (IINSPD.EQ. 'Y') .IIND.
STATEV(2).GE.SMPD)THEN
ILOOKO=l
END IF
END IF
IF(DAllS(TIME-TIM(IGIV+l)) .LE . DTMIN)TIIEN
IGIV=IGIV+l
END If
END IF
DTIME=DTIME1.1
IF(TIME+DTIME.GT.TIM(IGIV+l)+DTMIN.AND . IGIV+l.LE.NGIV)TIlEN
DTIME=TIM(IGIV+l)-TIME
END IF
IF(IPASS.EQ.13.0R . DTIME.LT.DTMIN)TIIEN
WRITE(MS,')"" NO CONVERGENCE'
WRITE(MS,')'DTlME=' ,DTlME,'IPASS=' ,IPASS
GO TO 308
END IF
END DO
END IF
C'" IF TIlE LOADING IS CYCLIC
IF( (CYCLIC.EQ. 'y') .OR. (CYCLIC . EQ. 'Y') )TIIEN
TPER=1.
DTMIN=TPER/40000 .
TlME=O.
ICYCLE=1
NRCYCL=ICYCLE
DIITlME=O.
DO 1=1,6
SIGIII(I)=O.
END DO
SMPI=SMPO
0111=00
IF( (JUMP.EQ. 'y') .OR . (JUMP.EQ. 'Y') )THEN
COUPL='n'
OPEN (UNIT=80,name='fatigue .out' ,status='unknown')
WRITE (80, ' ) 'REAL TIME TIME D p'
WRITE (80, #I ) , #I " " #I" #I #I" 11"" #I "'. #I"" #I *" .. *""" .. I< #I ,
#I
WRITE(80,') 'BEGINNING OF TilE CYCLE'

WRITE(80,')'CYCLE (REAL) =' ,NRCYCL
WRITE(80,')'CYCLE (MACHINE) =' ,ICYCLE
WRITE(80,400)RTlME,TlME,DHI,SMPI
END IF
CALL OUTPUT (NUl, NU2, TIME, STRAN, STRS, STATEV, NTENS, NSTATV,
STRSB,STAR)
IBL=1
DTIME=TPER/INCUSE(IBL)
DO IST=l,6
IF(EPSD( 1ST, IOL) 'EPSS( 1ST, IOL) . LT.O. )TIIEN
PERDIV(IST)=TPER/4 .
ELSE
PERDIV(IST)=TPER/2.
END IF
IPER(IST)=1
ISLOPE(IST)=l
END DO
NCYSUM=O
DOWlIILE (NRCICL. LE . NCYCLE)
DO IST=l , NAXI
IF (TIME+DTMIN . LE . PERDIV ( 1ST) )THEN
DSTRAN(IST)=EPSB(IST,IBL)'DTlME/PERDIV(IST)
ELSE
xxnu= epsb(ist,ib1)-epss(ist,ibl)
DSTRAN( IST)=ISLOPE( 1ST) '""nu'2 . 'DTlMEjTPER
END IF
END DO
CONVER='y'
DO IST=l,NSTATV
STATEV(IST)=STATEVI(IST)
END DO
SIGS=SYIEL(IBL)
EH1=SIGS/lOOO .
EPSMIN=SIGS/EO/10000.
IF(SIGS . LT.SIGF.OR .SIGS .GT.SIGU)STOP'''' PLEASE HEVI~W TilE
VALUES OF SIGf SIGu and SIGs .'
CALL INTEGR(STRAN,DSTRAN,NTENS,NSTATV,UNIAXI,COUPL,
& CONVER,STRS,STATEV,STRSB,STAR,RNU)
C'" If NO CONVERGENCE, DIVIDE TIME INC. BY 2
IF( (ANSDC .EQ. 'n' .OR.ANSDC.EQ. 'N') .lIND .
STATEV(ll).GT.O.)
DCRIT=DlC'SIGU/SIGS)"2)jRNU
IF(CONVER.EQ . 'n' )TIIEN
I Pl\SS;:::I P.1\5S+1
DTlME=DTIME/2 .
C IF CONVERGENCE
ELSE
11'( (ANSDC.EQ. 'n' .OR.ANSDC.EQ. 'N')
.AND.DCRIT . GT.O. 99DO )OCRIT=O. 99DO
IPJlSS=O
DO IST=1. NSTATV
STATEVI(IST)=STATEV(IST)
END DO
DO IST=1.6
STRAN(IST)=STRAN(IST)+DSTRAN(IST)
END DO
TIME=TIME+DTHIE
RTIME=TIME+DHTIME
CALL OUTPUT(NU1.NU2.RTIME.STRAN.STRS.STIITEV.NTENS.NSTATV.
STRSIl.STIIR)
IF (STIITEV( 3) .GE .DCRIT)TIlEN
WRITE(MS. *).
WRITE(MS.) CRACK INITIATION .'
GO TO 300
END IF
IF (STATEV( 3) .GE . 1.1 )STOI' .. DAMAGE CANNOT EXCEED 1.'
DO IST=1.6
11'( DABS (TIME-IPER( 1ST) *PERDIV( 1ST) ) .LE. DTMIN)TIIEN
IPER(IST)=IPER(IST)+1
DTIME=1./INCUSE(IBL)
END IF
11'( ISLOPE( 1ST) .GT. 0 .AND. DABS (STRAN( IST)-EPSB( IS1'. IIlL) ) . LE.
EPSMIN.OR.ISLOPE(IST).LT.O.AND.DABS(STRAN(IST)-EPSS(IST.IBL))
. .LE.EPSMIN)THEN
ISLOPE(IST)=-ISLOPE(IST)
END IF
END DO
C'" END OF A CYCLE
IF(DABS(TIME-TPERICYCLE) . LE . DTMIN)THEN
IDELTN=O
DO 1=1,6
SIGHF ( I )=S'rRS ( I )
END DO
SMPF=STATEV(2)
DHF=STATEV(3)
DELTP=SMPF-SMPI
DELTD=DHF-DIII
C'" FIND pD
IF(ILOOKO.EQ.O)THEN
ENERG3=ENERGO+(SIGS-SIGF*(SIGF/SIGY))'
DELTP'(NRCYCL-NCYSUM)
11'( (IINSPD.EQ. 'n' .OR.ANSPD.EQ. 'N')
.AND.ENERG3 . GE . ENERG1)THEN
SMPD=SMPF
NCYCLO=NRCYCL-NCYSUM
ILOOKO=l
ELSE 11'( (ANSPD.EQ. 'y' .OR.ANSPD.EQ. "';') .AND.
ST~TEV(2).GE . SMPD)THEN
NCYCLO=NRCYCL-NCYSUM
ILOOKO=l
END IF
END IF
C'" JUMP IN CYCLES PROCEDURE
11'( (JUMP . EQ . 'y') .OR. (JUMP . EQ. 'Y') )TIIEN
WRITE (00, .) END OF TIlE CYCLE'
WRITE ( 00.400 )RTIME. TIME. Dill' .S~IPF
DDMIIXO=DCRIT/50.
11'( (IINSDC.EQ. 'n') .OR. (IINSDC . EQ.N'))
DDMllXO=DlC/50 .
11'( (UNIAXI .EQ. 'y' ) .OR . (UNIAXLEQ . 'Y') )TIIEN
YYY=SIGSSIGS/ 2./EO
ELSE
TRDEPS=O .
DO IST=l.3
TRDEPS=TRDEPS+EPSIl(IST.IBL)-EI'SS(IST.IIlL)
END DO
YYl=(l.+XNU)*SIGS'SIGS/3./E0
YY2=EO*TRDEPS*TRDEPS/(l.-2.*XNU)/6.
YYY=YYl+YY2
END If
DPMAXO=SO'DDMAXOjYYY
DPMAX=DPMAXO
PSLOPE=NBCYCL(IDL)*DELTP
If(DPMAX.GT.PSLOPE)DPMAX=PSLOPE
WRITE(60,')'ILOOKO=' ,ILOOKO,'NCYCLO=' ,NCYCLO,'pD=' ,SMPD
If(COUPL . EQ . 'n' .OR . STATEV( 2) . LT. SMPD) Til EN
STAB='y'
DO KK=l,6
If(DABS(SIGllf(KK)-SIGUI (KK) ) .GT. Eill ) STNI= , n'
END DO
WRITE(80,*)'ENERGl=' ,ENEHGl,'ENERG3=' , ENERG3
If(STAB.EQ. 'y') TIIEN
WRITE(60,') 'STABILIZED CYCLE'
If (STATEV( 2) .LT .SMPD)T1IEN
C** * JUMP Of CYCLES BEfORE DAMAGE GRowrll
IDELTN=IDINT(DPMAX/DELTP)
If(NRCYCL+IDELTN.GT.NCYSUM+NBCYCL(IBL))
IDELTN=NCYSUM+NBCYCL(IBL)-NRCYCL
ENERG2=ENERGO+(SIGS-SIGY) *DELTP' (NRCYCL+IDELTN-NCYSUM)
If(ENERG2 . GT.ENERGl'l . 05)
IDELTN=(ENERGl-ENEllG3)/(SIGS-SIGY)/DELTP
WRITE(60,*)'*** JUMP Of ',IDELTN,' CYCLES'
WRITE(60,')'DPMAX=' ,IDELTN'DELTP, 'YYY=' ,YYY
ELSE
WRITE(60,') 'COUPLED CO~WUTATION fOil NEXT CYCLE
COUPL='y'
END If
END If
I::LSI::
C'" JUMP Of CYCLES ,\FTI::H DNlAGI:: GHOWfIi
DOMJ\X=DDNA.,(O
DSLOPE=(NBCYCL(IBL)-NCYCLO)'DELTD
If(DDMAX . GT.DSLOPE)DDMAX=DSLOPE
NCYCLO=O
IDELTN=IDINT(DDMAX/DELTD)
If(NRCYCL+IDELTN.GT . NCYSUM+NBCYCL(IDL))
IDELTN=NCYSUM+NBCYCL(IBL)-NRCYCL
If(Dllf+DELTD'IDELTN.GT.DCRIT'l.05)
IDELTN=IDINT((DCRIT-DHf)/DELTD)
If(IDELTN.GT.IDINT(DPMAX/DELTP))IDELTN=IDINT(DPMAX/DELTP)
WRITE (80, * ) , ** * JUMP Of ',IDELTN,' CYCLES'
WRITE(80,')'DDMAX=',IDELTN*DELTD, 'DPMAX=' ,IDELTN'DELTP
COUPL='n'
END If
If(DELTI'.GT . DPMAX)WRITE(80,')' .. PHOIJL"M
DPMAXO IS TOO SMALL . '
If(DELTD.GT . DDMAX)WHITE(80 , ')' ... PHODLEM
DDMAXO IS TOO SMALL . '
If(DELTD . GT . DDNN( .OH . DELTP .GT . DPMAX) IDEL1'N=0
If(IDELTN . LT.O.)STOP"'* PROBLEM: NEGATIVE JUMP Of CYCLES . '
END If
DO J=l,6
SIGIII(J)=SIGllf(J)
END DO
SMPI=SMPf+DELTP'IDELTN
DUI=Dllf+DELTD' I DELTN
STATEVI(2)=SMPI
STATEVI(3)=DIII
D1ITlME=DIITIME+TPJ::H'IDELTN
NRCYCL=NRCYCL+IDELTN
R'fIME=TlME+DIITIME
If( (JUM.I'.EQ . 'y') . OIL (JUll1' . EQ . ', ' ) )THEN
W1\ITE (80, ' ) ,-------- -----------------,
WRITE(60,') 'BEGINNING Of TilE CYCLE'
WRITE(80,')'CYCLE (REAL) =' ,NHCYCL+l
WRITE(80,')'CYCLE (MACHINE) =' , ICYCLE+l
WRITE(80,400)RTINE,TIME,DHI , SMl'I
END If
IF (NRCYCL.EQ.NCYSUI1+NBCYCL( IBL) )TIIEN

ENERGO=ENERGO+(SIGS-SIGF'(SIGF/SIGY))'DELTP'(NRCYCL-NCYSUI1)
NCYSUI1=NCYSUI1+NBCYCL(IBL)
IBL=IBL+l
END IF
ICYCLE=ICYCLE+l
NRCYCL=NRCYCL+ 1
END IF
400 FOR/1J\T(lX.4(E12.6..))
END IF
00 IST=1.6
IF(TII1E+DTII1E.GT.IPER(IST)PERDIV(IST)+DTI1IN)TJIEN
DTII1E=( IPER( 1ST) 'PERDIV( 1ST) )-'rI11E
END IF
END 00
IF(IPASS.EQ . 13 . OR . DTII1E . LT. DTI1IN)TIIEN
WRITE(I1S.) .. NO CONVERGENCE'
WRITE (115 ) DTII1E=' DTII1E IPlISS=' I PASS
GO TO 306
END IF
END 00
END IF
306 WRITE (115 ) .... TilE JOB IS ENDED. YOUR HESULTS FILES All!::
WRITE (115 )
IF( (JUMP .EQ. 'n' ) .OR. (JUMP .EQ. 'N' ) )~IRITE(I1S )FILEI. FILE2
IF( (JUMP.EQ. 'y') .OR. (JUMP.EQ. 'Y'))
. WRITE(I1S.)FlLEl.FlLE2.fatigue.out
STOP
END
SUBROUTINE INTEGR(STRAN.DSTRAN.NTENS.NSTATV.UNIAXI.COUPL.
CONVER.STRS.STATEV.STRSB.STAR.RNU)
II1l'LICIT HEIIL'S (A-II.0-Z)

REIIL'S S1'ATEV(NSTATV) .STll1lN(NTENS) DSTfu\N(N'fENS)
STRN(6). Ili(G. 6) .AS( 6).
STRS(6).ESTRS(6).ESTRSO(G).CESTRS(6).KESTRS(6).
& N(6).PSTRN(~).DPSTRN(6)
REAL'S LAMDA. NN
LOGICAL CRITERE
CHARACTER'l CONVER.COUPL.UNIAXI
C0I1l10N/ETIQl/EO.XNU.SIGS.SI1PD.SO
COMMON/ETIQ2/0NE.TWO.TIIREE
C
NN=TIIREE/TWO
RAC2=DSQRT('l'WO)
TOLRES=SIGSl.D-6
NISS=NTENS
C'" IDENTITY MATRIX
00 I=1.NISS-1
00 J=I+1. NISS
ID(I.J)=O .
ID(J. I )=0.
END 00
ID(I.I)=ONE
END 00
ID(NISS.NISS)=ONE
C'" IDENTITY VECTOR
00 I=1.NISS
IF(I.LE.3)AS(I)=ONE
IF(I.GE.4)AS(I)=0.
END 00
C'" LAME COEFf.
XI1U=EO/(ONE+XNU)/TWO
LAMDA=EO'XNU/(ONE-TWO'XNU)/(ONE+XNU)
C
SI1R=S1'A'l'EV(1)
SI1P=STATEV(2)
D=STATEV(3)
00 I=1.NISS
PSTRN(I)=S'l'A'l'EV(I+4)
END 00
C*" TOTAL STRAIN

00 I=l,NISS
STRN ( I ) =STRAN ( I \ +DSTRAN ( I )
IF(I.GE.1)STRN(I)=STRN(I)*RAC2
END 00
C*" ELASTIC PREDICTOH
DSMP=O.
IF( (UNIJlXI.EQ. 'y') . OR. (UNIJlXI.EQ. 'Y') )THEN
STRN(2)=-XNU*STRN(1)J(0.5-XNU)*PSTPR(1)
STRN(3)=STRN(2)
END IF
C .... Trace of strain tensor
TREPS=STRN(1)+STHN(2)+STHNI3)
C.... Effective stress
00 I=l,NISS
ESTRS( I )=LllMDA*THEPS*lIS I I )+2. *XMU* (S'fltNl I )-PSTltNl I) )
END 00
CALL INVMI ESTRS, ESTRSII, ESTRSll, NISS)
SIGN=l .
IFIESTRS(l).LT . O. )SIGN=-l.
C*** TEST THE YIELD CONDITION
F=ESTRSB-SIGS
IFIF.LE.O.)THEN
GO TO 500
ELSE
C*" IF THE ELASTIC PREDICTOH OOES NOT VERIFY THE YIELD CONDn'ION
00 I=l,NISS
N( I )=NN* tESTRS( I )-ESTHSII*lIS( I) )/ESTRSn
END 00
e"" COL'rections over the elastic predictor
C.... Correction over p
DENOM=3. *XMU
CSMP=F/DENOM
C . . . .Correction over effective stress
00 I=I,NISS
CESTRS(I)=-2.*F*N(I)/3.
END 00
C*** LOOP ON THE PLASTIC CORRECTIONS
NFOIS=O
CRITERE=.TRUE.
00 WHILE(CRITERE)
NFOIS=NFOIS+l
DIVl=ESTRSB
NTEST=2
IFINFOIS.EQ.l)NTEST=l
C*** UPDATE 'filE STATE
C. ... Update inc. of p
DSMP=DSI1P+CSMP
C.. . . Update effective stress
00 I=I,NISS
ESTRSII)=ESTRSII)+CESTRSII)
END 00
CALL INVMIESTRS,ESTHSH,ESTHSll,NISS)
00 I=l,NISS
NII)=NN*IESTRSII)-ESTRSH*lISII))/ESTRSB
END 00
C
IF(NTEST.EQ.2)THEN
DIV2=ESTRSB
REL=(DIV2-DIVl)/DIVI
END IF
C.... If we converge too slowly/or \oIe diverge,exit . The main
C . .. . program '-lill propose a smaller "time" increment.
IFINFOIS.EQ.50.0H.
(NTEST. EQ. 2 .AND.REL.GE . 15.0-02) )TIlEN
CONVER='n'
GO TO 500
END IF
C'" compute residual functions

F=ESTRSB-SIGS
IF( (UNIIIXI.EQ. 'y') .OR. (UNIIIXI.EQ. 'Y') )'rHEN
STRN( 2 )=-0. 5'STRN( 1 )+(0. 5-XNU) 'SIGN'SIGSjEO
STRN(3)=STRN(2)
'rREPS=STRN(1)+STRN(2)+STRN(3)
END IF
DO I=l,NISS
KESTRS(I)=ESTRS(I)-LAMDA'THEPS'IIS(I)-
2. 'lCMU' (STRN( I )-PSTRN( I )-DSMI"N( I) )
END DO
C Max .
residual5
HESTRS=O.
DO I=l,NISS
IF(DABS(KESTHS(I)) . GT.HESTRS)RESTRS=DABS(KESTRS(I))
END DO
C'" ITERATIVE TEST ON TilE YIELD CONDITION
IF(DADS(F) . LT.TOLRES . AND.HESTHS.LT . TOLH~ S )TII~N
CHITERE= . FALSE .
ELSE
C'" PLIISTIC CORRECTIONS
CALL VTRANVl(N,KESTHS,XNKES,NISS)
C. . . Correction over p
DENOM= 3 . 'XMU
CSMP=(F-XNKES)jDENOM
C.. .. Correction over effective stress
DENOM=DENOM'DSMP jESTHSIJ+ (1. )
DO I=l,NISS
CESTRS(I)=(2.j3.)'(XNKES-DENOM'F)'N(I)-KESTRS(I)
CESTRS(I)=CESTHS(I)/DENOM
END DO
C'" END OF THE ITERATIVE TEST ON THE YIELD CONDITION
END IF
C'" END OF THE LOOP ON THE PLIISTIC CORRECTIONS
END DO
C'" END OF TEST ON THE YIELD CONDITION
END IF
500 CONTINUE
IF(CONVER.EQ . 'y' )TIIEN
IF( (UNIIIXI .EQ. 'y') .OR. (UNIIIXI .EQ. 'Y' ) )THEN
DSTRAN(2)=STRN(2)-STRAN(2)
DSTRAN(3)=DSTRAN(2)
END IF
C.... Plastic strain increment
DO I=l,NISS
DPSTRN(I)=N(I)'DSMP
END DO
C. ... Damage increment
DO=O .
CALL DAMAGE(THEPS,ESTRSIJ,Y,RNU)
IF(DSMP.GT . O. . AND.SMP+DSMP.GE.SMPD.AND .
COUPL . EQ . 'y' )TIIEN
DO=Y'DSMPjSO
END IF
STAR=(ONE-D-DD)'DSQRT(2.'EO'Y)
C ... plastic multiplier increment
DSMR=(ONE-D-DD)'DSMP
C.... Compute stresse s
DO I=l,NISS
STRS(I)=ESTRS(I)'(ONE-D-DD)
END DO
STHSIJ=ESTHSIJ'(ONI::-D-DD)
C.... STORE STATEV AT THE END OF THE INCHEMENT
STATEV(l)=SMR+DSMR
STATEV(2)=SMP+DSMP
STATEV( 3 )=D+DD
STATEV(4)=DD
DO I=l,NISS
STATEV( 1+4 )=PSTIlN ( I ) +DPSTRN ( I)
END DO
STATEV(ll)=DSMP
END IF
RETURN
END
SUBROUTINE INVAR(V,~I,VB,NISS)
C*~*********************************
C . . . 1st and 2nd stress invariant s
IMPLICIT REAL'8(A-H,O-Z)
REAL'O V(6)
COMMON/ETIQ2/0NE ,TWO ,THREE
C
VH=(V( 1 )+V( 2 )+V( 3) )/TIIREE
CONST=O.
00 I=l,NISS
IF(I.LE . 3)CONST=CONST+(V(I)-VH)'(V(I)-VH)
IF(I.GE.4)CONST=CONST+V(I)'V(I)
END 00
VB=DSQI\T(TIIREE'CONSTjTWO)
RETURN
END
SUBROUTINE DAMAGE (TREPS ,ESTRSB,Y ,HNU)

C***********************************
IMPLICIT REAL'S(A-II,o-Z)
COMMON/ETIQ1/EO,XNU,SIGS,SMPD,SO
C
TERMI=ESTRSB'ESTRSB'(l.+XNU)/EO
TERM2=EO'TREPS'TREPS/(1 . -2.'XNU)/2.
Y=(TERMI+TERM2)/3.
RNU=2 . 'EO'Y/ESTRSB/ESTRSB
RETURN
END
SUBROUTINE VTRANVI(V,V1,VTV1,NISS)
c*******************************
C . . . Inner product of 2 symmetric 2nd order tensor s
IMPLICIT REAL'O (A-H,O-Z)
REAL'O V(6),V1(6)
C
VTV1=0.
00 I=l,NISS
VTV1=VTV1+V( I) 'VI (I)
END 00
RETURN
END
SUBROUTINE OUTPUT(NU1,NU2, TIl1E,STRAN, STRS ,STATEV,

_ NTENS ,NSTATV,STRSB ,STAR)
c****************************************************
IMPLICIT I\EAL'O(A-II,O-Z)
HEAL'O STRAN(NTENS),STRS(NTENS),STATEV(NSTATV)
C
WRITE (NUl ,130 )T1ME, (STRAN( 1),1=1,3), (STRS( 1),1=1,3),
, STA'fEV (3) ,STATEV (2) ,STRSIJ, STAR
WRITE(NU2 ,DO )T1ME, (STRAN(I) ,1=4,6) , (STRS (1) ,1=4, (.) ,
, STATEV( 3) ,STATEV( 2) ,STHSIJ,STAH
WIlITE(90 , 444 )TIME,STATEV( 3) ,STATEV( 2) ,STHSIJ, STAll
130 FORMAT ( 1X,E10 . 4,' , ',3 (E10.4,' ,'), 2X, 31E10 . 4,' " ),
3X,E10.'1,' , ' ,E10.4,' ,',2IE10. 4 ,',') )
444 FORMAT(3X,E10.4,5X,E10.4,3X,E10 . 4,3X,E10.4,4X,E10.4)
RETURN
END
4.3.4 Exercises Using the DAMAGE 90 Postprocessor

DAMAGE 90 may become your friend if you bring it to life in your computer.
The following examples which were already performed with DAMAGE 90 will
help you to set up the program without too many "bugs". Please try them.
Case of ductile damage
The material properties are those of a stainless steel:
E= 200 ()()() MPa 1/=0.3
a f= 200 MPa a y= 300 MPa au= 500 MPa
S= 0.06 MPa Ep D = 10% D l e=0.99
Pure tension at microscale
Take as the main input
PD = 10%, De = 0.99 and E ll ' DAMAGE 90 computes the other strains:
10 22 = 1033 = -2I 10 11 + ( 21 -1/ ) all

E(I-D) '
For a better representation of the large strain hardening, use the piecewise perfect
plasticity procedure with
E ll % 0 0.25 1.5 5
a s MPa 200 300 I 400 500.
The results are plotted in Figure 4.18.
Plane strain at meso-(or micro-)scale
Take the same material data as previously but use the perfect plasticity procedure
with
a s = 500 MPa.
(j (MPa)
SOO
400
300
200
0.6
100 0.4
0.2
0 0
0 5 10 15 20 Fig. 4.18. Numerical simulation of a pure
e ('!.) tension test at the microscale
Perform proportional loading in strain histories:
and find for each numerical test the set of strains (E 1\ , E22 ) which corresponds to
crack initiation. You will obtain the so-called "damage forming limit curve" in
plane strain, as shown in Figure 4.19.
-20 -15 -10 -5 o 5 10 15
Fig. 4.19. Crack initiation limit curve in plane strain
Case of brittle damage

The material data are those of a ceramic :
E= 400 000 MPa v= 0.2
af= 200 MPa a y= 250 MPa a u =300 MPa
S= 0.00012 MPa EPD =0 D l e = 0.05
Perform calculations for the case of an elastic tension case at the mesoscale: Ell
given with E22 = E33 = -VE I \' with a plastic threshold a s = 200 MPa, and plot
the result at the mesoscale (an elastic-brittle material) and at the microscale. as in
Figures 4.20 and 4.21.
Low and high cycle fatigue

The main hypothesis of the damage model is that there is no difference in the
equations for low or high cycle fatigue because they obey the same energetic
mechanism. Furthermore, written as a damage rate the constitutive equation is
valid even if a cycle cannot be defined.
The material data is that of an aluminum alloy:
E= 72 000 MPa V= 0.32

af= 303 MPa a y= 306 MPa a u = 500 MPa
S= 6 MPa E pD = 10% D l e= 0.99
300 *
200
100
Fig. 4.20. Stress strain and brittle

O~-.---.--.---.--.---r--r--.----- damage at the mesoscale in pure
o 2 3 4 5 6 7 8 E11 ( 10') tension
200
*
100 o
0.10
0.05
O--'r-~---r--~~O~~-r--~~ Fig. 4.21. Stress strain and damage at the
o 2 3 4 5 microscale when pure tension is applied
at the mesoscale
Consider a pure tension case at the microscale in which E II is the main input, the
other components E22 and E33 being calculated by DAMAGE 90 as
The cyclic strain hardening is taken into account by using piecewise perfect plas-
ticity according to the cyclic stress strain curve:
Ell % 0.425 0.43 0.47 I 3.5 4.5

as MPa 300 305 308 370 440 460.
Perfonn some numerical simulations to obtain the number of cycles to crack ini-
tiation as a function of the stress amplitude, and plot the results together with the
number of cycles No to crack nucleation corresponding to P = PD (Figure 4.22).
Consider the case Ell = 3.5% (~E = 7%). The number of cycles to failure is
40. Compare to an hypothetical case of pure tension at the mesoscale where the
input is
E" = 3.51O- 2 , E22 = E33 = -VEil'
Due to the effect of the triaxiality the number of cycles to failure would be only 8.
Plot the graphs of stresses, strains and damage for the pure tension case at the
microscale as in Figure 4.23.
t. E11 ("!o)
8.0 \
\
\
\
4.0 \
\
\ No NR
\
2.0 ,
"-
.......
' ,,-
1.0 . . . . . .......
-. -
- ....,---
~""'
0.5 t - - - - . , - - - - , - - - , - - - . - - - - , - - - -
10 10~
N Fig. 4.22. Fatigue rupture curve
Two-level fatigue case

The non linear accumulation of fatigue damage was also dealt with in Exer-
cice 3.44, in which the graph of Figure 3.30 was drawn with the same material
data as for the above example of low and high cycle fatigue and for the one-
dimensional case at the mesoscale:
E22 = E3} = -VE,l ' E I2 = E 13 = En = O.

Ell 0.47% ~ N R = 7720 cycles;
Ell = 0.425% ~ N R = 109570 cycles;
e" = 0.47% during 3088 cycles and E" = +0.425% after ~ N R 52400
cycles;
Ell = 0.425 % during 43828 cycles and Ell = +0.47% after ~ N R 50000
cycles;
Check that for
- the highest loading first, NNI + N2 = 0.85 < I;

RI NR2
. . N N
- the lowest loadIng fIrst, __
I + _2_ = 1.2 > I.
NRI Nm
Figure 4.24 gives the results ::2 R2

when ::1
RI
varies in both cases. Compare these
with Figure 3.31 .
Multiaxial fatigue
You may also play with DAMAGE 90 to obtain contours of the number of cycles to
failure in biaxial fatigue as it is shown for the same aluminum alloy as previously:
- for plane strain in Figure 4.25:
cil = x, c22 = y, c33 = OJ
- for tension and shear in Figure 4.26:
c II = x, c 12 = y, all other components = OJ
"
0.04 (J",,(MPa)
600
400
0.02
200
0
N 0
"
-0.02 -200
-400
-0.04
-600
-0.06 -800
0 10 20 30 40 -0.06 - -0.02 0 0.04
(J"eq (MPa)
SOO 5
p
400 4
300 3
200 2
100
D
0 0
0 10 20 30 40 50 0 10 20 30 40 so
N N
Fig. 4.23. Very low cycle fatigue simulation in tension at the microscale
4.4 Fully Coupled Analysis 201
~
NR
2
1.0
8"
0.8
N
0.6
0.4 Linear accumulation
0.2
O+----,r----.-----,----,---~~---
o 0.2 0.4 0.6 0.8 1.0
Fig. 4.24. Accumulation diagram for two-level simulation in tension at the mesoscale
dE22 %
2
.7
.6
.5
.4
NR=7800 cycles
.3
.2
.1
O~""'T"---'--r--r-r--+-~--_ Fig. 4.25. Biaxial fatigue envelope for

o .1 .2 .3 .4 .5 .6 .7 dEli % the case of tensile plane strains im-
2 posed
4.4 Fully Coupled Analysis

For ductile-and creep-type damage, the damage is often distributed over a large
volume of the structural component under consideration. Then, for a good accuracy
the coupling between damage and strain must be considered all over the structure
or at least over a substructure. This makes the calculation (corresponding to the
scheme in Figure 4.27) more complex, but there is no other way to proceed.
L\12 %
-2-
.4
.3
.2
.1
.1 .2 .3 .4 .5 .6 L\1I %
-2- Fig. 4.26. Biaxial fatigue envelope for the case of ten-
sile and shear strains imposed
The classical way to perform this calculation is to implement the coupled constitu-
tive equations from Section 2.1.4 and the kinetic damage law from Section 3.1 in
a finite element code as explained in Section 4.1.3. The difficulty for convergence
of iterations and for the computer time of calculations are of the same order as
for classical elastoplastic analysis. Such routine is available in the French code
SYSTUS and possible with the American code ABAQUS.
4.4.1 Initial Strain Hardening and Damage
What makes this method interesting is that it allows us to take into account pre-
existing strain hardening and damage as initial fields in some part of the structural
component to be analyzed. It is always of great interest to evaluate the remaining
strenght or lifetime of a component after an accident which has induced plastic
deformation and damage.
Three quantities may be introduced as material initial conditions at the Gauss points
of the finite element meshes:
- Ro, the isotropic strain hardening stress as an initial value for its differential
constitutive equatio.n,
Coupled
elasto-plastic and damage
constitutive equations
Fig. 4.27. Fully coupled analysis of strain and damage

or, in the absence of damage, Po' an initial value of the accumulated plastic
strain since in this case, there is a one-to-one relationship between p and R:
R = Roo[l- exp( -bp)] ;
- X{?, the kinematic strain hardening stress tensor as an initial value for its dif-
ferential constitutive equation,
XB = I' [~Xooifj(I-D)-XB~];
- Do an initial value of damage for its kinetic law,
b Y.
= SP.
One more quantity which often exist as a characteristic of the structure, but not
of the material, is a state of residual stresses, after welding, for example. A
self-equilibrated residual stress field can also be introduced at the Gauss points
as an initial condition:
o
aij'] = oIn S }
a?j such that { Figure 4.1
a?j,n j = 0 on 8S
The first three material quantities are not difficult to introduce in a finite element
code, but the main problem is to identify them. As far as isotropic strain hardening
and damage are concerned, if the damaged structure is available for testing, one
can evaluate Ro and Do, at least on the surface, by microhardeness measurements
as explained in Section 1.3.3. As far as kinematic strain hardening is concerned its
initial value may be evaluated by means of phenomenological considerations(!!)
or neglected; the latter is not an approximation for the stabilized cycle which does
not depend upon the back stress X{?:
Note that those initial conditions may be considered in the uncoupled analysis as
a constant reduction of the elasticity modulus and the strain hardening variables.
4.4.2 Example of a Calculation Using the Finite Element Method

(R. BILLARDON, 1988)
It is out of the scope of this course to describe in detail the finite element procedure
for elastoplasticity or viscoplasticity coupled with damage calculations. The set
of constitutive equations of Section 2.1.4 together with the kinetic damage law
identified in Section 3.1.4 may be implemented in any classical finite element code
having plasticity routines. Best results are obtained with the algorithm described
briefly in Section 1.3 of the present chapter.
The practical case considered here schematically is an underground gas pipeline
which has been damaged by the shock of a machine digging in the soil. The
pressure inside the pipe varies periodicallly twice a day due to gas consumption. It
induces fatigue of the material especially where the shock occured. The question
is: What is the remaining strength or lifetime of the damaged pipe before it breaks?
Obviously, it is of primary importance to decide when and how to make a repair.
The pipe and the loading being considered are defined in Figure 4.28.
A simple calculation of the undamaged cylinder unconstrained at the ends by
the expansion joints gives the state of stress everywhere. Assuming the material
to be elastic, one can deduce the displacements everywhere and particularly at
the boundaries A and B of the substructure containing the damage that will be
studied in more detail. Figure 4.29 shows the geometrical model submitted to a
one-dimensional displacement within the hypothesis of plane strain. Note that this
substructuring neglects the redistribution of stress due to plasticity that will occur
at the root of the notch as a result of the loading 71(t).
The material is a low-carbon steel. Its elastoplastic and damage parameters have
been determined from several tension tests according to procedures described in
Sects.2.1.5 and 3.1.5:
- E, v for elasticity;
- a y' Roo b, X oo ' 'Y for plasticity;
- S, cPD ' Die for damage together with af and au
More interesting is the strain hardening and the damage induced by the shock in
the idented zone; these must be considered as initial conditions, as explained in
Section 4.4.1. These damage and strain hardening fields have been identified from
microhardness experiments on a polished section of a plate being subjected to a
shock similar to that which damaged the pipe. The method used is described in
Section 1.3.3. The results are shown in Figure 4.30.
A I B p
o~----------------------------~
Fig. 4.28. Damaged pipeline under pressure
N
u
42mm
1----- - - - - - - - ------------
Fig. 4.29. Notched substructure

0.1 0.20.3 0.3 0.2 0.1
Fig. 4.30. Initial strain-hardening and damage fields
The values of Ro and Do are introduced at each Gauss point of the finite ele-
ment meshes represented in Figure 4.30. The rest of the meshing is indicated in
Figure 4.29. The results of the structural calculation corresponding to a mono-
tonic loading performed in order to determine the remaining strength, are shown
in Figures 4.31 and 4.32.
From the evolution of the damage in Figure 4.31, the displacement which will
cause crack initiation is u c:::' 0.12 mm. It turns out to be about 50% of the elastic
displacement corresponding to the load of the pipe when the von Mises equivalent
stress is equal to the yield stress ay. This particular damage has reduced the strenght
of the pipe by a factor of about 50%.
If the initial strain hardening and damage are not taken into account the damage
evolution shows a critical displacement of about 0.20 mm, that is, an error of 66%.
As far as crack initiation by fatigue due to the cyclic pressure in the pipe is con-
erned, the same calculation is performed with a periodic displacement as input.
To avoid a considerable amount of calculation, the procedure of integration by
damage increments (instead of time or cycle increments) described in Section 4.13
is used. For the case under consideration, the number of cycles to crack initiation
of the indented substructure (or the pipe) was about 100 times less than the fatigue
life of the initial pipe. As the nominal number of cycles to failure of the undam-
aged pipe line was of the order of 105 , this corresponds to 1000 cycles, that is,
more than a year to organize the repair without cutting off gas delivery. Thank
you.
o'eq(MPa) 0
600 1.0
0.8
400
0.6
0.4
200
0.2
u u
0
0 0.02 0.04 0.06 0.08 0.10
mm
0
0 0.02 0.04 0.06 0.08 0.10 mm
Fig. 4.31. Evolution of the von Mises equivalent stress and the damage at the critical point M *
4.4.3 Growth of Damaged Zones and Macrocracks

Damage mechanics may predict crack initiation at the mesoscale, generally with
a damaged zone surrounding the cracked mesovolume element. As the loading, or
the time, continue to increase, the calculation may continue with a fully damage
point until another mesovolume element reachs its critical state of damage, and so
on, to describe the evolution of a fully damaged zone. No stress intensity factors,
no fracture mechanics to model the phenomenon of crack growth. Nice isn't it!
But nice only in principle because this ideal process does not work so easily,
due to the localization of damage at its critical stage. The problem is no longer
ellptic (see Section 2.3.5). A localized bifurcation of the solution occurs and, if
classical finite elements are used, the convergence regarding the mesh size is not
ensured.
We may say in 1991 that this is still an open problem despite several attempts to
introduce concepts which avoid this difficulty:
- neglecting the coupling between damage and elasticity: D. Hayhurst (1978);

J.L. Chaboche (1982);
- partial regularization of the problem by viscosity: A. Needleman (1988);
- introduction of a characteristic length for each material taken as the minimum
size of the meshes in the finite element calculation: Z. Bazant (1976);
- definition of a "nonlocal" damage variable as a weighted mean value over
a characteristic volume: G. Pijaudier-Cabot and Z. Bazant (1986), K. Saanouni
(1988);
o'eq (MPa)
600
u =0.12mm
400
u=O.OSmm
200
0.1 0.20.3 0.30.20.1

z
0
0 2 3 (mm)
D
O.S
0.4
u=0.12mm
0.2
z
OL-------~~----r_------._--_.
o 2 3 (mm)
Fig. 4.32. Von Mises equivalent stress and damage along the z axis
- derivation of special finite elements able to model discontinuities of strains due

to damage localization: M. Ortiz (1987), T. Belyschko (1988), R. Billardon and
I. Doghri, (1989);
Then, to cover the whole process of fracture of a component, the following steps
may be worked out (Figure 4.33):
I) structural calculations either as an uncoupled, locally coupled or fully coupled

analysis,
2) conditions of mesocrack initiation:
* loading or time or number of cycles;
* direction n of the plane of the crack determined by det( n . H . n) = 0 in
elastoplasticity (see Section 2.3.5);
* crack area: its value may be derived from the concept of characteristic
Fig. 4.33. "The frightened apple": damage mechanics

and fracture mechanics as complementary sciences (after
J- P. Balloffet)
length matching damage mechanics and fracture mechanics developed in

Section 2.3.1.
for a two-dimensional problem, where [ is the crack length of the mesocrack

initiated by the damage process. For a three-dementional problem, take [2 as
the crack area.
* the crack shape is not a problem in two dimensions. In three dimensions, a
circular crack can be asumed, as it is the shape for which the stress intensity
factor in mode I is constant all along the circumference;
3) Fracture mechanics analysis to obtain the growth of the initial crack as defined
above.
4.4.4 Exercise on Damaged Zone at it Crack Tip

Considering a macrocrack in a two-dimensional structure made of a ductile mate-
rial and loaded in mode I, we may consider the stress field as elastic far away from
the crack tip then elastoplastic near the crack tip and much affected by damage in
the process zone very close to the crack tip. Let us try to analyze some features of
this process zone (Figure 4.34).
Assuming:
- small deformations,
- small-scale yielding,
- a strain-hardening power law
which induces a stress field with the H. R. R. singularity (Hutchinson-Rice and

Rosengreen) in polar coordination (r, ()):
1
(Jeq"""'" - 1 - '
r M +1
Fig. 4.34. Plastic zone and damaged zone

at a crack tip
- stress and strain redistribution due to damage negligible,

- the plastic zone ry(8) delimited from the elastic zone by an offset
py(py ~ 0.2 10- 2 for example),
- the damaged zone rD (() delimited from the plastic zone by the damage
threshold PD'
Show that the damaged zone may be deduced from the plastic zone by an homo-
tetic transformation of ratio
rD = ( py)M; 1 .
ry \PD
For most of materials this damaged zone is very small.
Example: py = 0.2 10 - 2 , PD = 10 10 - 2 , M = 5, rD ~ 10- 2 .

ry
As the energies involved are more or less proportional to r1 and r~, this corre-
sponds to a very small energy dissipated in damage and fracture in comparison to
the energy dissipated in plasticity. This is the reason why the critical value of the
strain energy release rate Gc at fracture is so small in comparison to the strain
energy involved.
4.5 Statistical Analysis with Microdefects (F. Hild 1994)

The fracture of any material is mainly governed by defects at a micro level which
always exist after the process of elaboration. In the deterministic analysis, they are
taken into account by their mean effects in the identification of the material
parameters. But for materials such as ceramics or cast iron. The large size of some
defects imposes the introduction of a statistical effect. Furthermore the only way
to explain the size effect on fracture or the high scatter in fatigue failure for
example is to take into account the probability of having defects of a given size in
the tested sample or structure.
4.5.1 Initial defects

Since the geometries of the microcracks and microcavities are always complex and
most of the time unknown, a good approach consists in modeling them by initial
equivalent damages.
Let us recall the results of Section 2.3 where it has been shown that:
- the equivalent damage D of a crack of area s lying in a cell of volume I x I x I is
- the equivalent damage D of a cavity of diameter d in a cell of volume I x I x I is
d2
D~[2 .
Then, the statistical characteristics of the surface of initial defects may be directly
translated into the statistics of initial damage Do provided the definition of the size
a representative volume element be possible. Practically, this size I may be the
mean distance between the centers of adjacent microcracks, that is the inverse of
their linear density. These defects may be randomly distributed in size, orientation,
shape ... but we restrict ourselves to the distribution in size only given by the
probability density f(Do} of the equivalent initial damage Do
Prob (Do :s;; Do :s;; Do + dD o) = f(Do} dD o .
The function f(Do} may be deduced from measurements of defects through obser-
vations with a microscope. We shall see that some statistical characteristics may be
obtained more easily from mechanical characteristics at measoscale.
The probability density characterizes the distribution of the defects or the initial
damage in a structure but it is also the probability of having an initial damage of
a given value in each representative volume element.
4.5.2 Case of Brittle Materials

This case applies to ceramics or cast iron. Consider a Representative Volume
Element damaged by one defect characterized by an equivalent initial damage Do
having a probability density f(Do} and loaded by a damage equivalent stress a*.
We look for statistics on the stress ak which induces a crack initiation, that is when
the pure brittle failure criterion is reached. From Section 3.2.1 this is when
au being the ultimate stress of the material. With

a*2
y = - - -----"
2E(1- D o)2 '
a1; = au (1- Do),
Do being a random variable 0::; Do < 1, the mean value of the failure stress, also
a random variable, is
1
a1; = Jau (1- Do) f(Do) dD o ,
o
f(Do) being a probability density
1
Jf(DotstDo = 1,
o
a1; = a{l -l Dof(Do)dDo] ,
or Ia1; = a u (1- Do) I,

1
150 = JDof(Do)dDo being the mean value of the initial damage distribution.
o
The standard deviation of the stress at fracture is obtained by
_ 1
J
(a1;)2 = (a1;(Do) - a1;)2 f(Do)dDo,
o
1
(a1;)2 = a; J(150 - Do)2 f(Do) dD o ,
o
Do = D (Do -15 0 )2 f(D o)dD o J/ 2 being the standard deviation of the initial
damage distribution.
As an example, consider an uniform distribution of initial damage
the mean value and the standard deviation of the failure stress are
4.5.3 Case of Quasi-Brittle Materials

Let us recall that materials considered as quasi-brittle are those for which plastic
strain axists only at microscale without any substantial effect at mesoscale.
The same representative volume element damaged by one defect of magnitude Do
as previously is considered but the damage may grow stable on reaching its critical
value Dc corresponding to crack initiation. Section 3.2.2 gives the damage evolu-
tion law for quasi-brittle materials taken here with a zero value of the two thres-
holds for simplicity
where (Jf is the fatigue limit of the material, S the damage strength, Rv the
triaxiality function.
Assuming a proportional loading for which Rv = const.
D(t) = Do + 2is(J2
Rvceq(t)'
where Ceq is the accumulated strain
Ceq = tGiBi8yl2 dt.
The time to crack initiation tc is reached when D = Dc, considered here as a

constant,
Do becoming a random variable with a probability density f(Do), the mean value
of the accumulated stain to crack initiation is
or with
Its standard deviation is given by
(e:q)2
Dc
= J(e:q(D o ) -
o
e:q) f(Do) dD o ,
2
-
where Do =
[DC! (Do _DO)2 f(Do)dDoJ1 /2 is the standard deviation of the initial
damage distribution.
- Application to uniaxial periodical fatigue of strain amplitude ~e (Rv = 1).
For this case eeq = 2~eN,

N being the number of cycles.
At crack initiation
the mean value of the number of cycles to failure NR and its standard deviation NR
immediately follow
Writing the standard deviations as
- NR -
NR = Do
Dc-Do
explains a well-known experimental result which shows that the standard devia-
tion of the number of cycles to failure NR , or the scatter in fatigue, increases with
the number of cycles to failure NR . An example is given in Figure 4.35.
Another interesting consequence is related to the identification of initial damage
or defects of a given material for which the Woehler curve is known by the mean
value of the number of cycles to failure NR and its standard deviation NR .
Solving the above equations for Do and Do gives the mean value and the standard
deviation of initial damage
Experiments
2 .--~...,..,-~,---~~-,.-j -- PF =50 %
-- - --PF = 90 %
\
'- ,
......... P = 10 %
.... F
1.5 ___ ~----------r---~
----------- -
-E
CI3
1
. .... ..
.. ,....

.
............... ..... ......... ......... _....
zo 0.5 L.4~~~.........L-
5 ~~~'--'--'----6~~~....... 7
10 10 10 10
Number of Cycles to Failure
Fig. 4.35. Woehler curve of a sa cast iron with its scatter (from A. S. Beranger)
2 A
- aff, -
Do = D ---NR
C ES '
4.5.4 Case of Ductile Materials

The same kind of analysis may be performed with the damage evolution low of
Section 3.3.1.
. a;Rv .
D = 2ES P ,
or in the case of proportional loading with an initial damage Do
a; Rv
D = Do + 2ES P ,
When the value of the initial damage is randomly distributed with a mean value
150 und u standard deviation Do, the mean value PR and the standard deviation
PR of the accumulated plastic strain at crack initiation defined by D = Dc are
4.5.5 Volume Effect

It has been observed many times that the stress to failure of a specimen or a
component decreases when the size of this component increases. This phenomenon
is important in brittle materials and it is mainly due to the probability of finding
a large defect which increases when the considered volume is larger.
Consider a structure of volume V loaded by a uniform stress field and having a
distribution of initial damage f(Do).
Consider the representative volume elements of the material having a volume Vo ,
their number in the structure is
V
n= - .
Vo
The hypothesis of the weakest link is adopted in the sense that if a crack initiation
occurs in an R VE, the whole structure breaks.
Let us call the failure of the structure probability PF and the corresponding
probability of survival Ps = 1 - PF .
Considering the case of brittle materials and no interaction between defects, a
crack initiation occurs in an RVE (i) if
at ~ au(l- Dc) = aR
The probability of survival of the structure considered as the set of the R VE links
is given by the product
Ps =l-PF = o PSi (at <aR )

i=!
Taking the Logarithm

n
Ln(l-PF} = I Ln(l-PF;),
i=!
but
!
where PFO = J f(Do} dD o ,
Dc
1
then Ln(1- PF) = - JLn(1- PFo)dV,
Vo v
or PF = 1 - exp (~ JLn (1- PFO ) dV).

Vo v
If the probability PFO does not depend upon the space coordinates,
PF = 1 - exp (~ Ln (1- PFO )

or PF = 1-(1-PFO)V/Vo I.
This formula explains the size effects: Vo being fixed by the nature of the material,
if the volume of the structure V increases, the probability of failure PF also
increases.
The well-known Wei bull formula may be derived for a three-dimensional state of
stress with two additional hypothesis.
- The probability PFO is small with respect to 1,
Ln(1- PFO ) ~ -PFO ,
PF = 1- exp ( - ~P FO) '
- The probability density of initial damage is a power-law function
where m is the Wei bull exponent

1
PFO = I f(Do)dDo = (1- Dc)m,
Dc
with
The Weibull formula is
Knowing the material parameters Vo , au, m, the probability of failure is an

increasing function of the volume V and of course an increasing function of the
damage equivalent stress a*.
As an application, let us calculate the mean value of the damage equivalent stress
to fracture Uk of a structure as a function of its volume V in the case of an uniform
distribution of the initial damage
1
f(Do) = - 0 < Do < DM ,
DM
with
then
The mean value of the stress is given by
Changing the variable (1: for Do with

(1: = (1.(1- Do),
Let us consider the normalized mean failure stress ratio
This result is plotted in Figure 4.36 with the asymptotic value
v
- --+ 00 =>
Vo
The main difficulty in a volume effect analysis is the identification of the elemen-
tary representative volume Vo. Practically, Vo maybe taken as the volume of the
uniform stressed part of the specimens used to identify the constitutive equations
of the material under consideration, but it is difficult to avoid a certain degree of
subjectivity!
A better identification would be to determine the probability density of defects
f(Do} which allows for the calculation of PFo if Dc (1*) is known. Then from tests
on a real structure PF may be derived and Vo calculated from
PF = 1- (1- PFo}v/vo .
But this procedure needs two large sets of experiments!!
4.5.6 Effect of Stress Heterogeneity

When a structure is subjected to a nonuniform stress field, the probability of
failure depends upon the probability of having large defects located in the stress
concentration zone. If a stress field has a high degree of heterogeneity, the proba-
cr~ (V) ~
00
cr~ (V,,)
1.2
>-
::l
'"
~o
DM =0,1 ...,
~ "-"-'-"-"-' .. _.. _.. _.. _..-.. _..-.. -.. _.. _. --------_.._.._._.._.._.. _..-.. _.._.._.._.._.._.._.._.'-"-"-"-"-"-"-"-'"
-~.--.-.-- - - . - - - - - - - - . - - - _ _ _ . _ _ _ _. __ 0' ._ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ ~
~
D" =0,2 e.g.

0.8 g.
::l
"-"-"-"---"-'--" -"----'-"-' -_. ..

__.._--------_ _.._ _._----_.._ _.. _.._.._--.. .. s
m
0.6 D .. =0,5 ~
z
a
0.4
0.2
V
o ~1--~4-~-4--~4--+~--~4-~-4--~4-~-4--~4-~-4--+-4--+~--+-~-+~~+-~
Vo
o 5 10 15 20 25 30
Fig. 4.36. Volume effect on the mean damaged equivalent stress at fracture
bility of failure for the same distribution of defects is larger than for a uniform
stress field involving the same strain energy.
To model this effect, it is convenient to introduce the concept of effective volume
V eff defined by an equivalence in elastic strain energy as it governs the phenomenon
of damage.
If We is the elastic strain energy density
W~AX its maximum value at the stress concentration point of the structure,
Jwe(M)dV = VeffW~AX,
V
or
Then, the probability of failure formulas are written with the effective volume
calculated by an elastic calculation instead of the real volume of the structure,
PF = 1- (1- PFoteffiVo ~ 1- exp (~~f P FO ) ,

or for the Weibull formula
PF = 1- exp ( - Va [O'~AXJm)
Veff
~ .
4.5.7 Exercise on Bending Fatigue of a Beam

The considered beam is of rectangular cross-section and subjected to a pure
periodical moment MMAX = M, M min = -M (Figure 4.37)
L
..
Fig. 4.37. Beam in bending fatigue
By a classical deterministic elastic analysis show that

- The one dimensional maximum stress is
- The strain amplitude is
12M
L1e = Ebh 2 .
- The number of cycles to failure from Section 4.5.3 is
By a statistical analysis considering a probability density of initial damage f(Do)

and with
show that:
- The probability of crack initiation is
DM
PFo = J f(Do)dDo ,
(Do)e
(Do), being the initial damage becoming critical after NR cycles
NR aJL1e
(Do), = D, - ES .
- For an uniform distribution

P _ 1_ ~ NR aJL1e
Fo- DM + ESD M
Taking into account the effects fo volume and stress heterogeneity linear with X2
show that:
- The effective volume is
Lbh
Veff = -3-
Notice that the effective volume is reduced to Lbh/6 if pure unilateral conditions
are considered that is no damage in compression.
- The volume ratio is
Veff 1
Vo 3'
if the constitutive equations of the material are identified by tests in tension on

specimens of the same size as the beam.
- The probability of crack initiation of the beam in bending is
D N (12.11:)1 /3
P =l-(l-P )VerriVo=l _( _c_ R f
F Fo DM ESD M
JE
Bibliography
Atlas of Metal Damage. L. Engel, H. Klingele,

Wolfe Science Books, C.H. Verlag, Munich, 1981.
Mechanics of Solid Materials. J. Lemaitre, J.L. Chaboche,

Cambridge University Press, 1990
Translated from "Mecanique des materiaux solides," 2nd edition, 1988,
Dunod 1985
Also translated in Chinese by T.C. Yu, 1991
Mechanil!s of Damage and Fatigue. S.R. Bodner and Z. Hashin,

Proceedings Symposium JUT AM, Haifa, Israel, 1985.
Pergamon Press, 1986.
Introduction to Continuum Damage Mechanics. L.M. Kachanov,

Martinus Nijhoff Dordrecht, The Netherlands, 1986.
Continuum Damage Mechanics-Theory and Applications. D. Krajcinovic,

J. Lemaitre, CISM Lectures, Springer Verlag, 1987.
Damage Mechanics. D. Krajcinovic,

in Mechanics of Materials Journal, 8, Elsevier, 1980.
Formul~ for Stress and Strain. R.I. Roark,

McGraw Hill Book Company, 1965.
Stress Concentration Factors. R.E. Peterson,

Wiley Interscience Publication, 1974.
Handbook of Formulas for Stress and Strain. W. Griffel,

Frederick Ungar Publishing Co., 1976.
Finite elements in plasticity. D.R.I. Owen, E. Hinton,

Pineridge Press Ltd, 1980.
History of International Damage Mechanics Conferences
Damage Mechanics, Euromech Colloquium, Cachan, France, 1981.
Mechanics of Damage and Fracture, NSF Workshop, Atlanta, USA, 1982.
Mechanics of Damage and Fatigue, JUTAM Symposium, Haifa, Israel, 1985.
Continuum Damage Mechanics, CISM Lectures, Udine, Italy, 1986.
Large Deformations and Damage, Mini Symposium, JUT AM Congress, Greno-

ble, France 1988.
Strain Localization and Size Effects Due to Cracking and Damage, CNRS-
NSF Workshop, Cachan, France, 1988.
Applications of Continuum Damage Mechanics, Euromech Colloquium, Crakow,

Poland, 1989.
Damage Mechanics in Composite Materials Design, Santa Barbara UCSB Work-

shop, USA 1990.
Continuum Damage Mechanics, Sandia National Lab Workshop, Pleasanton,

USA,1990.
Damage Mechanics in Engineering Materials, ASME Winter Annual Meeting,
Dallas (USA), 1990
Recent Advances in Damage Mechanics and Plasticity, ASME Summer Mechanics

and Materials Conference, Tempe (USA), 1992
Computer aided Assessment and Control of Localized Damage, CAACOLD'92,

Southampton (UK), 1992
Fracture and Damage of Concrete and Rock, FDCR 2, Vienna (Austria), 1992
Impact Damage in Composites, Euromech Colloquium, Oxford (UK), 1993
Computer Aided Assessment and Control of Localized Damage, CAACOLD'94,

Udine (Italy), 1994
Structural Damage Assessment using Advanced Signal Processing Procedures, lnt.

Workshop, Pescara (Italy), 1995
224 History of International Damage Mechanics Conferences
Inelastic Deformation, Damage and Life Analysis, Int. Symposium. Hawal (USA),
1995
Micromecbanics of Plasticity and Damage of Multipbase Materials, IUTAM Sym-
posium, Paris (France), 1995
Mecbanical Bebaviour of Damaged Solids, Mecamat Workshop, Fontainebleau
(France), 1995
Author and Subject Index
A 542 stainless steel, 7 crack closure, 19

ABAQUS,202 crack closure parameter, 80, 111
accumulated plastic strain, 52 crack initiation, 102, 115, 169, 172
acoustic emission, 35 creep damage, 4, 33, 123
AISI 1010 steel, 6, 28, 29, 33, 133 creep fatigue interaction, 132
AISI 316 stainless steel, 32, 111, 139 creep rupture, 10, 91
AlIix,65 critical damage, 98, 103
anisotropic damage, 59, 75, 80 critical point(s), 169
associated variable, 40, 42 critical time, 162
back stress, 14, 40, 50 DAMAGE, 90,179

Baptiste, 46 damage accumulation, 141
Bazant,206 damage equivalent stress, 44, 85
Beaujolais, 35 damage growth, 70, 73, 177
Be1yschko, 206 damage localization, 86
Benallal, viii, ix, 7, 55, 165 damage measurement, 35
Beranger,214 damage orthotropic, 62
Bernoulli, vii damage threshold, 16,98, 108
Berthaud, 10, 26 damage variable, II
Billardon, viii, ix, 28, 29, 86, 165, 203, 206 damaged zones, 205
boundary value problem, 155 debonding of interface, 208, 212
Bourgogne, 35 decohesion, 16
Bridgman's law, 31 deep drawing, 119
brittle damage, 4, 68, 91, 112, 113, 197 density, 29
Budiansky, 14,42 deviatoric stress, 65
discontinuity, 86
dissipation, 47
Cailletaud, 32 Doghri, ix, 86, 165, 177, 179,206
carbon-carbon composite, 23 double scalar variable, 64
Cauchy stress tensor, 39 ductile damage, 4, 27, 72, 117, 196
ceramics, 2, 3, 5, 112 ductile fracture, 11
Chaboche, viii , 43, 63, 206 Dufailly, 6, 8, 15, 22, 29, 111
characteristic length, 70
Chrzanowski, 90
Clausius-Duhem inequality, 41, 46, 99 effective damage area, 17
cleavage, 4 effective stress, 12, 19,42,62, 84
composites, 2, 3, II, 112 elastic predictor, 166
concrete, 2, 3, 24, 26, 112 elasticity modulus, 14, 21
consistancy condition, 53, 105, 155 electrical resistance, 30
constitutive equations, 55, 107, 155 energy strength of damage, 98
convergence, 166 Engel, 6, 7, 8, 9, 10,20, 21
copper, 22 equilibrium equations, 155
Cordebois, 121 Eshelby, 76
Coulomb's coefficient, 77 Euler's angles, 75
coupled constitutive equations, 50 Evans, viii
coupling, 14 extrusion, 119
226 Author and Subject Index
fatigue crack, 11 jump-in-cycles procedure, 166, 179

fatigue limit, 109 June Finney, ix
Fenchel transform, 78
fiber bundle system, 90
fiber composite, 65 Kachanov, v, vii, 11
finite element method, 165 Kachanov's model, 124
first principle of thermodynamics, 101 kinematic hardening, 14, 40, 43, 50
forging, 119 kinetic damage law, 169
fourth-order damage tensor, 63 kinetic energy, 101
fracture limits, 119 kinetic law of damage, 47, 48
Fran90is, viii, 10 Klingele, 6, 7, 8, 9, 10, 20, 21
free energy, 46 Krajcinovic, viii, 2, 75, 76
frightened apple, 208 Kronecker, xv
fully coupled analysis, 192
Ladeveze, 64, 83
Gauss point, 166 Lame's constants, 79
generalized standard materials, 47 Leckie, viii, ix, 60, 79
Germain, viii Lee Hao, vii
Geymonat, 86 Legendre-Fenchel, 85
Gibbs specific enthalpy, 83 Lemaitre, v, 11, 13, 39, 41,95,176
Goodman's diagram, 139 Lienard, 20, 21, 168
Gunawardena, ix Lin-Taylor hypothesis, 81 , 114, 136, 177
Gurson model, 72, 74 linear accumulation, 142
linear creep fatigue interaction, 135
Lippmann, v, vii, viii, ix
listing of DAMAGE, 90, 185
Hayhurst, 60, 206 loading function , 48 , 49
Heaviside, xv, xvi loading-unloading condition, 49
Helmholtz free energy, 41 localization, 86, 176
Herath, ix locally coupled analysis, 176
high cycle fatigue, 7, 27, 135, 197 low cycle fatigue, 5, 27, 127, 197
high-alloy steel, 19 lower bound, 171
Hild, 5, 128,209 Luder's bands, 89
homogenization, 69
H ult, viii, 90
Hutchinson, viii, 208 Macauley, xv
hydrostativ stress, 65 macrocracks, 205
macroscale, 4
Manderscheid, ix
identification, 57, 109, 110 Manson-Coffin law, 131
implicit integration, 166 MAR M 509, 36
implicit scheme, 178 Marquis, viii, ix, 39
IN 100 superalloy, 32 master curves, 120
inclusions, 4 material parameters, 57, 98, 104, 109, 152
incremental linearization, 165 Maxwell's compatibility equation, 87
initial conditions, 155 Mazars,24
initial damage, 179, 202 McClintock, 74
initial plastic strain, 179 McMeeking, viii
initial strain hardening, 202 mean stress, 138
input data, 180 measurements, 19
instability, 86 mechanism of damage, 10
interface, 66, 208 mesoscale, 4, 116
intergranular decohesions, 5 metal forming, 119
intergranular microcracking. 7 metals, 1, 3, 117
internal variables, 40 micro crack closure effect, 107
isochronous creep damage curves, 126 micro-meso element, 68, 72
isotropic damage, 39 microcavities, 72
isotropic hardening, 14, 40, 42, 51 microcrack, 2, 19, 68
Author and Subject Index 227
microcrack closure effect, 80 Rabotnov, vii, 12

micrograph pictures, 19 Rayleigh's surface waves, 25
micrographic observations, 10 representative volume element, 3, 11
microhardness, 25 residual stresses, 203
micromechanics, 68 Rice, 208
microscale, 4, 116 Rice and Tracey, 74
Miner, vii Robinson, vii
Moussy, 30 rolling, 119
multiaxial fatigue, 200 Rosengreen , 208
multi hole, 159 rupture criterion, 16, 100, 108
multiplier, 49 rupture domain , 103
Mura,76
Murakami, vii, viii, 60
Saanouni, 206
Sabine Lemaitre, ix
Najar, viii safe domain, 103
Needleman, 206 Sampaio, vii
Neuber's method, 163 scale, 3
Newton's procedure, 166, 178 second-order damage tensor, 60
nonlinear interaction, 132 Sermage, 151
nonlinear kinematic hardening, 51 Sherman, ix
normality condition, 47 slips, 2, 3
Norton, xvii state kinetic coupling theory, 42, 96, 210
Norton's law, 33, 49, 124 state potential, 39
notches, 156 strain energy density release rate, 43, 107
nucleation of microcracks, 5 strain energy release rate, 18, 69
numerical simulation, 196 strain eq ui valence, 13
strain hardening, 27
strain measurements, 21
O'Connell, 14, 42 strain-displacement equations, 155
observable variables, 39 strain-driven algorithm, 178
Ohm's law, 30 stress amplitude drop, 31, 36
Onat, 79 stress concentration, 17, 156, 167
Ortiz, 206 stress concentration factor, 156
output data , 183 Stress corrosion cracking, 11
stress intensity factor, 18
summation convention of Einstein, 40
Palmgreen, vii
SYSTUS, 177, 202
Palmgreen-Miner's rule, 141
Pafis, 'XVi, xviii
Paris law, 70 Taira rule, 135
perfe9tly plastic materials, 105 tangent modulus, 166
Pijau'dier-Cabot, 206 tangent stiffness tensor, 86
Pineau, viii Taylor, 78
plane strain, 66 tensile crack, 11
plastic corrector, 166 thermal cracks, 11
plasticity, 3, 14,48, 53 thennodynamics, 39
plasticity criterion, 105 threshold, 98
Poisson, xviii transcristalline cleavage, 10
Poisson's ratio, 18, 42 transcristalline fatigue crack, 10
Policella, 34 transgranular microcracking, 8
polymers, 2, 3, 117 triaxiality, 116
Poss,23 triaxiality function , 45, 67
postprocessing, 177 triaxiality ratio, 44, 67, 113
potential drop, 30 Tvergaard, 74
potential of dissipation, 46, 95 two-scale volume element, 114
proportional loading, 66, 117
ultimate stress, 16, 109

quasi-brittle damage, 113, 135 ultraconic waves, 23
228 Author and Subject Index
uncoupled analysis, 169 Woehler curve, 138

unified damage laws, 95 wood, 2-3
unilateral conditions, 80
Young, xvi
viscoplasticity, 49, 55, 108 Young's modulus, 19,42
von Mises, xix
von Mises equivalent stress, 40, 44 Zaoui, viii
Weibull, v, 216
wine selection, 35

A Course On Damage Mechanics-Springer-Verlag Berlin Heidelberg (1996) Professor Jean Lemaitre (Auth.)

Uploaded by

Copyright:

Available Formats

A Course On Damage Mechanics-Springer-Verlag Berlin Heidelberg (1996) Professor Jean Lemaitre (Auth.)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Course On Damage Mechanics-Springer-Verlag Berlin Heidelberg (1996) Professor Jean Lemaitre (Auth.)

Uploaded by

Copyright:

Available Formats

A Course on Damage Mechanics

Springer-Verlag Berlin Heidelberg GmbH

Second Revised and Enlarged Edition

With 118 Figures

Library of Congress Cataloging-in-Publication Data

Lemaitre, J. (Jean), 1934-

Munich, 6 August 1990. Professor Horst Lippmann

Why an apple on the cover page?

Grandes Ecoles" 3rd cycle "Mecanique et materiaux" initiated by D. Fran~ois,

the finite element method. A post-processor of damage growth, up to crack

Spring 1988 - Spring 1991 . Jean Lemaitre

Chapter I Phenomenological Aspects of Damage

Chapter 2 Thermodynamics and Micromechanics of Damage 39

2.2 Analysis of Anisotropic Damage . . . . . . 59

Chapter 3 Kinetic Laws of Damage Evolution . . . 95

3.5.3 Kinetic Law of Debonding Damage Evolution. . 146

Chapter 4 Analysis of Crack Initiation in Structures 154

History of International Damage Mechanics Conferences 223

Authors and Subject Index . . . . . . . . . . . . 225

Dij Kronecker delta: Dij I if i = j, Dij = if i -I- j

Ixl absolute value of x

b isotropic hardening material exponent

c Paris' crack growth material parameter

specific internal energy rate

h crack closure parameter

length of the mesovolume element

P accumulated plastic strain

if heat flux vector

we elastic strain energy density

elastic strain energy of a body

y strain energy density release rate

unidimensional and components of tensorial variable of kinematic

nonlinear kinematic hardening material parameter

C, Ci j ' uniaxial and tensorial total strains

Paris' crack growth material exponent

Lame's coefficient of elasticity

Lame's coefficient of elasticity in shear

v Poisson's ratio of elastic contraction

scalar and tensorial volume damage variables

uniaxial and tensorial stresses

(5"fj von Mises equivalent stress

plastic threshold stress

Helmholtz specific free energy

1.1 Physical Nature of the Solid State and Damage

Fig. 1.1. Elementary plastic strain by slip due to dislocation movement

Fig. 1.2. Elementary damage by nucleation of a microcrack

1.1.2 Slips, Plasticity and Irreversible Strains

1.1.3 Scale of the Phenomena of Strain and Damage

- metals and ceramics: (0.1 mm)3

restricted to a surface, although the strains, being movements of atoms by variation

1.1.4 Different Manifestations of Damage

100llm o 0.1 0.210-2

Low cycle fatigue damage

N R < 10000 cycles.

High cycle fatigue

1.1.5 Exercise on Micrographic Observations

A - Transcristalline fatigue crack propagation in a nickel alloy at 650C

N- Fracture of a glass fiber epoxy composite

1.2 Mechanical Representation of Damage

Elasticity E * ***