Download as pdf or txt
Download as pdf or txt
You are on page 1of 23

Engineering Fracture Mechanics Vol. 34, No. 3, pp. 679-701, 1989 0013-7944/89 $3.00 + 0.

00
Printed in Great Britain. 0 1989 Perpmon Press pk.

ON EVOLUTION LAWS OF ANISOTROPIC DAMAGE


C. L. CHOW and T. J. LUt
Department of Mechanical Engineering, University of Hong Kong, Pokfulam Road, Hong Kong

Abstract-This paper is focused on the development of evolution laws of anisotropic damage.


Material damage in the form of microcracks and microcavities leading to macro-mechanical
property variations is modelled phenomenologically with some distortion and blurring of fine
details, arguing that the extent of material property changes may serve as an appropriate, indirect
measure of damage. The experimentally observed phenomenon of damage hardening is also
described. The well-established thermodynamic theories of irreversible processes are then appealed
for deriving constitutive equations of the materials suffering from progressive deterioration,
together with a new hypothesis of energy equivalence instead of the restrictive hypothesis of strain
or deformation equivalence hitherto proposed. Thermodynamic conjugate force of the damage
variable, the “damage energy release rate”, is deduced in analytical form by assuming that the
energy involved in the plastic flow and damaging processes can be uncoupled. Evolution laws of
anisotropic damage are established following the concept of a damage surface in the space of
affinities as well as the principle of maximum damage dissipation. An important outcome of the
investigation is the identification of a new damage characteristic tensor J which can accurately
describe the anisotropic nature of damage growth and yet is simple with only one unknown
parameter to be determined from a standard tension test.

1. INTRODUCTION
ONEBASIC hypothesis in most isotropic and anisotropic models of continuum damage mechanics
(CDM) is that, ignoring the mode shape of microscopic damage growth and micro-damage-ge-
ometries, damage can be viewed as a macroscopic state variable which affects the macro-mechanical
properties of the material in the sense of an “effective stress” or “net stress” introduced first by
Kachanov-Rabotnov[ l-21 and developed thereafter by Leckie-Hayhurst[3-6], Lemaitre-
Chaboche[7-121, Murakami-Ohno [13-141 and many others. This basic hypothesis of effective
stress can be stated this way: There exists an “effective damage tensor” M(D) applied to stress
tensor u which defines the effective stress tensor 6(D)
G(D)=M(D):a’ (1)
and which is to be substituted into constitutive equations of materials suffering progressive
deterioration instead of current stress tensor u, to describe elastic, plastic, creep, fatigue and other
deformation behavi0urs.S Note that three fundamental concepts of CDM have been introduced
in the foregoing hypothesis, i.e. the damage variable D, the damage effect tensor M(D), and the
effective stress tensor 6.
Because the material deterioration in the process of creep is one of the most important
phenomena in the design of high-temperature equipment of longer service life, the problem of
developing comprehensive theory to describe creep damage mechanisms and damage accumulation
has been an objective of many investigations. Kachanov in 1958[1] first proposed a phenomeno-
logical theory of creep damage by introducing a scalar damage variable D to characterize the
material degradation which is responsible for tertiary creep and rupture. The underlying physical
implication behind the scalar characterization is that creep damage in a continuum is assumed to
be everywhere the same and independent of the specific orientation chosen. Thereafter, Rabot-
nov[2] interpreted that the damage variable D represents the fraction of damaged material, thus
reducing the effective resisting area of the local material element from A to A = (1 - D)A. In the
one-dimensional case, the effective stress was thus defined as
d = F/ii = F/(1 - D)A = o/(1 - D). (2)

tcurrently on leave from Xian Jiaotong University, Xian, P.R.C.


$In eq. (l), a boldface symbol implies that this quantity is of tensorial nature (e.g. u, M, D) and (:) denotes the tensorial
product reduced to two indices.

EFM34/3--J 619
680 C. L. CHOW and T. J. LU

Rabotnov makes the hypothesis of strain equivalence (the term “strain equivalence” is, in fact,
first proposed by Lemaitre[7] and will be discussed in some detail elsewhere, e.g. in [15], which
assumes that, the conventional steady stage creep law, say, the Norton-Hoff power law
i =A&’ (3)
still retains its validity in tertiary stage of creep, provided that the “real” stress o is replaced by
the effective stress of (2), i.e.
( = Ad” = A[a/(l - D)] (4)
where A and n are material constants to be identified from steady stage creep data. The following
evolution law of damage is then assumed to complete the model
B = BP = B[a/(l - D)]” (5)
which can be integrated to give the rupture time (corresponds to critical value of D, D, = 1) as
t, = a-m/B(m + 1) (6)
and the two material parameters B and m are to be determined from creep rupture tests.
The definition of damage due to Kachanov-Rabotnov is of brittle type in which the nucleation
and propagation of microcavities and microcracks are assumed to be the only cause leading to
material degradation. From the geoemetrical point of view, it is thus assumed that the ratio of the
cavity area to the total virgin area of a cross section can serve as a measure of damage. Although
this idea is intuitively appealing and has been universally accepted, it needs careful re-examination.
Consider, for instance, a solid weakened by a set of multitude (but finite) random planar
microcracks (Fig. 1) and subjected to external load D in the direction n. Since the number of
microcracks while large is still finite and the number of cross sectional area A normal to the loading
direction n infinite, the probability of a cross section containing a microcrack is nil[16]. Thus,
according to the definition of damage due to Kachanov-Rabotnov, the cross section A is not
affected by the microcrack system at all, and hence its load bearing capability is the same as that
when no microcrack exists in the material element which is not in accord with intuitive imagination
and physical reasoning. This drawback of Kachanov-Rabotnov’s idea may be illustrated in an
alternative way. Since the volume fraction of cavities in a specimen is equal to the area fraction
of cavities[l7], very flat, i.e. cracks, with average volume fraction zero, should not influence
deformation behaviours of the material. The problem may be resolved if one notes the fact that
the actual amount of load bearing area in a cross section like A is strongly affected by the
microcracks in its proximity. A characteristic length 1 may thus be introduced[ 161by assuming that
only those microcracks within the influence band ( - 1 < 5 < 1) will impair the load bearing capacity
of A. Nevertheless, the specification of the magnitude of I raises an additional problem to be solved.
Furthermore, in Kachanov-Rabotnov’s model, the cavity shape and orientation do not enter the
damage characterization. Even though the shape of a cavity may be of negligible effect, it is evident
that macro-mechanical properties of the material must be subjected to different influences by
cavities at different orientations [18]. The final comment is referred to one peculiar notation

Fig. 1. A material element weakened by a cloud of random planar microcracks and the definition of the
influence band (-I < r < I).
Evolution laws of anisotropic damage 681

Fig. 2. Typical stress-strain curve for an elastic-plastic material subjected to simple tension.

concerning the Kachanov-Rabotnov model and the hypothesis of strain equivalence in that, for
a cylindrical bar subjected to simple tension (Fig. 2), the model predicts a constant value of the
Poisson’s ratio v not affected by material damage while the elastic Young’s modulus E is reduced
to E (1 - D) after damage has occurred. This result is obviously in variant from experimental
observations[ 19-221.
On the other hand, despite the complexity of their mathematical structure, tensors of first,
second and higher ranks can describe the spatial distribution state of microcavities and their
macro-mechanical effects on material responses more accurately, and hence, considerable efforts
have already been made to develop damage models based on these tensors. Recently, we introduced
a damage model incorporating a symmetric second order damage tensor D,, used to define the
extent of material degradation induced by microstructural rearrangements as an indirect measure
of damage. The model was then applied to investigate the crack-tip shielding effect due to
anisotropic microcracking in brittle solids[ 151.The effect of anisotropy of microcracking was found
to be significant with an isotropic damage case but the least effective in mitigating the remote
applied load.
This paper is intended to describe the establishment of evolution law of the damage tensor
D,under the following sections of sequence. The parametrization of material damage through D,
is first introduced briefly. The thermodynamic work-conjugate force of D,or damage energy release
rate Y, is next determined following the phenomenological nature of the conventional definition
of damage. The concept of damage surface in the space of the generalized force YVas well as the
principle of maximum damage dissipation are then introduced such that a suitable framework for
the formulation of an analytical model may be presented. Attention is finally focused on the
development of a new damage characteristic tensor J, which is demonstrated to provide better
physical representation of damage evolution in a material element.

2. CHARACTERIZATION OF DAMAGE
An appropriate selection of measures to define and quantify the damage variables plays an
essential role in the theory of CDM since it not only affects the accuracy of the developed model,
but also the evolution of the magnitude of the damage variables. Most of the microcrack
representation retained the intuitively appealing idea due to Kachanov-Rabotnov adopting the
void area density or net area reduction as the measure of damage in the observed plane. The present
representation of damage is no exception. Nevertheless, unlike those methods which model a solid
weakened by a cloud of microdefects analytically or micromechanically, the present approach will
model the solid experimentally or phenomenologically. The effect of all microstructural rearrange-
ments on macro-mechanical properties of the material such as the gradual degradation of elastic
stiffness is smeared into a locally homogeneous field, with more or less distortion and blurring of
the fine details, and the field quantities of the field will then be utilized to characterize the damage
state.
682 C. L. CHOW and T. J. LU

Damage characterization is realized on the “basic cell” of CDM, the so-called “local material
element” V,, the dimensions of which being small compared to the external dimensions of the solid
under consideration and yet large enough to guarantee statistical homogeneity of the material
heterogeneities and microdefects in V, . The definition and selection of VM can be found in
[15,23-241. Having established a proper averaging process to transmit the information from
mesoscale (of aggregate, grain or fiber size) to macroscale, it can then be proved, with the help
of homogenization method[25-261, that the classical laws of continuum mechanics for a “fictitious”
continuous, homogeneous, crack-free medium retain their validities in the macroscopic description
for a “real” discontinuous, fragmented, irregular medium except that all the variables involved are
replaced by their corresponding average measures[l5,27-281.
Based on the analytical result that the effective compliance tensor S”(D) of the element VM
embedded with an arbitrary number of elliptic planar or penny shaped microcracks possess
orthotropic symmetry[l5,29-301, we assume that those irreversible microstructural occurrences
leading to macroscopic, measurable material mechanical property variations can be characterized
by a homogeneous field D(x), the “damage field”, which is a second order symmetric tensor, i.e.
D, = D,.Instead of using isolated microscopic entrances, it is noted that D, are indirect, average
measures of microdefects presenting in the local element V,. The following conditions are imposed
to further specify the damage field D:
(i) the principal directions of D coincide with the principal directions of stress field at the
specified material element with its three principal values denoted by (D,, D2,D3);
(ii) for a local element V with surfaces (A,, AZ,A3) perpendicular to a set of Cartesian
coordinate axes (x, , x2, x,) with unit base vectors (e, , e,, e3) which coincide with the principal
directions of local stress field, the function of D is to reduce the virgin undamaged surfaces
(A,, AZ,A,) to (b,, &,, 6,) through the relations (Fig. 3)
&=(l-D))A,e,=(l-D,)A,
&=(l-D,)AZe,=(l-D,)A,
A,=(1 -D3)A,e,=(l -D,)A,; (7)

93 =A3e3
t

(a) 42=A29
x3

T %I

(b)

Fig. 3. Local material element: (a) before damaging; (b) after damaging; and (c) after transformation
through damage tensor D.
Evolution laws of anisotropic damage 683

(iii) the effect of material damage on the compliance tensor S of the local element V,,, is
reflected through a “damage operator” M*(D)
S(D) = M*(D) : S(0) (8)
and the operation of M*(D) should be such that S(D) possess the usual symmetric properties and
(iv) the mapping between D and S(D) is unique and D is completely determined when S(D)
is obtained through measurements.
A significant feature of the present characterization of damage which differs from most other
damage models is to describe the experimentally observed “damage hardening” behaviour through
a scalar parameter k’ under the assumption that damage hardening is isotropic. Consider, for
example, a cylinder loaded in simple tension and the measured true stress-strain curve shown in
Fig. 2. At point A beyond initial yielding, upon unloading, we generally get a reduced elastic
modulus E < E due to damage. With further loading, no new damage is induced (i.e. E does not
change its current value) until the new yield point A is surpassed. This phenomenon is believed
to be closely related to the heterogeneous microstructure of the material. k’ may be taken as the
total damage energy, viz.
k’= Wd=f- Y,,dD,. (9)
An alternative measure of damage hardening is the accumulated overall damage w which is defined
incrementally as
dw = [A ‘JiikrdD, dDk,]“* (10)
where J is the damage characteristic tensor to be defined later and A’ is a material constant. The
damage hardening parameter, k’, is then assumed to be
k’=w (11)
where w is the result of integrating dw over the damage path. Similar to the two measures of plastic
hardening[31-321, the two measures of damage hardening are equivalent for any quadratic or linear
damage criteria.
Based on the properties of D,, we are able to construct a new damage effect tensor M(D)
expressed in the principal coordinates as

1
0 0 0 0 0
1 -D,
1
0 0 0 0
l-D2
1
0 0 0
Pf,(WI = l-D,
2-D2-D3 (12)
0 0
2(1-D&l-D,)
2-0,-D,
0
20 -D,)(l-D,)
2-D,-Dz
S
2(1 -D,)(l-Dz)
where the Voigt notation for stresses and strains has been adopted.
It is noted that, since the new set of damage variables (Dii, w) describe mechanically the
internal microstructural state of materials suffering progressive deterioration, they are precisely the
internal state variables in continuum thermodynamic theories of irreversible processes[33-351.

3. DETERMINATION OF DAMAGE ENERGY RELEASE RATE Y


Experimental observations reveal that the gradual degradation of macro-mechanical material
properties occurs as a result of a complex interaction of two dissipative mechanisms: propagation
of dislocations or line defects vs the nucleation and evolution of microcavities. The microcaviting
nucleates in the planes of local maximum shear stresses where plastic distortion is significant. In
684 C. L. CHOW and T. J. LU

other words, “dilatational” damage in the form of microcaviting commences only when the local
yielding condition has been satisfied or the positions of dislocation serve as the potential sites of
microcaviting. The following properties of microcaviting have been recognized[3638]: (a) the
microcavities are initially formed when the shear stress applied along a preferentially oriented slip
plane with relatively weak grain boundary exceeds a certain threshold value; and (b) the
microcavity nucleated initially in the slip plane “eventually switch to the plane normal to the
tension axis”.
Analytically, the problem is complicated significantly by the presence of two different
dissipative mechanisms and their corresponding sets of different variables. One of the most
important steps consists of the establishment of appropriate expressions for the Helmholtz free
energy + and the dissipation potential cp which will enable us to derive constitutive equations and
damage evolution laws conveniently from the well-established thermodynamic theories of irre-
versible processes[33-351.
Identifying the kinematic and isotropic hardening of plastic flow by clijand p and the brittle
effects (microcrack density, distribution and orientation) by a vector damage variable wi, the
following form of Ic/is assumed by Krajcinovic[36]
$ = $(b)(Ce,w, T) + *yz,p, w, T) (13)
where ee is the elastic strain tensor, T is the temperature while the superscript “b” and “d” signify
brittle and ductile components, respectively. The thermodynamic conjugate force of w is given by

Ri= a@b’
-paw- (14)
I

where p represents the material density. The brittle component Rjb) is shown[39] to have maxima
only in the directions of the principal elastic strains. Thus the ductile component of the free energy
@ must be selected in such a manner that Ri acquires initially maximum value along the normal
to the slip plane. Furthermore, in accordance with the properties (a) and (b), the dominance must
gradually shift from Rid) to Rib) as the deformation progresses ahead.
The foregoing argument regarding the determination of themodynamic conjugate force R
stems from the definition of damage variable w. If the shape of the microcavities is considered to
be of little significance, then damage is assumed by Krajcinovic[39] to be characterized by a vector
given by
Wi= W,ni (15)

where the scalar w0 denotes the void density in a cross section defined by the normal ni. The
evolution of wi follows directly from (15)
. .
wi = w,n, + w,n, (16)
in which the second term signifies the “rotation” of the cavity. We can therefore see that, in the
model proposed by Krajcinovic, “damage” is gradually shifted from the initial slip plane to the
final plane perpendicular to maximum tension which necessitates the introduction of ductile term
Rid) in the thermodynamic conjugate force and hence the ductile term eCd)of free energy must
depend explicitly on wi.
In the present model, damage is defined phenomenologically instead of microscopically.
Indeed, to associate each microdefect with a damage vector such as (15) would unnecessarily
complicate the problem to such an extent that it can hardly be resolved by available analytic tools.
The alternative strategy taken here is to assume that the extent of material deterioration (property
and strength detriments measured on a macroscale) induced by structural rearrangements occurring
on a microscale may act as a suitable, indirect measure of damage with some distortion and blurring
of the fine details. The shifting process of a microcavity from the initial slip plane to the final plane
normal to maximum tension is considered to be of negligible effect. It is true that damaging and
yielding are two interactive mechanisms, viz. damage occurrence necessitates the yield condition
to be satisfied a priori and the position of slip plane acts as the potential site of microcaviting, but
the energy dissipated in the yielding and damaging processes, on the other hand, may be assumed
reasonably to be independent.
Evolution laws of anisotropic damage 685

Fig. 4. Schematic illustration of the various energy dissipating mechanisms during plastic flow and
damage growth.

Figure 4 illustrates schematically the various form of energy stored or dissipated in a simple
tension test. The evolution of hardening during plastic flow OA is denoted by the curve OA’ while
AB and AC parts correspond to the hardening and increment of elastic strain during damaging.
The total energy is divided into four parts: (1) the energy stored in the material due to plastic
hardening; (2) the energy dissipated by heat; (3) the energy stored in the material due to damage
hardening; and (4) the energy released by the system during damaging, -Y: 6 D, which is eventually
converted into heat.
Neglecting the kinetic part of plastic hardening, the following hypothesis is proposed:
Hypothesis. The elastic properties of the material depend only on the accumulated damage D
and not on the dislocation density p; the energies involved in the plastic flow and damaging
processes, dissipated by heat or stored in the material due to hardening, are independent.
Consequently, in the present model, the Helmholtz free energy takes the following form

pll/(ce, P, D, w, T) = wehe, Q T) + $J.P, T) + h,(w 7’) (17)


where we (ce, D, T) is the elastic strain energy, $r(p, T) the free energy due to plastic hardening
and I,+~(w,T) the free energy due to damage hardening. The complementary energy is obtained from
Legendre transformation of the free energy with respect to strain, i.e.

= WY@, D, T) - &(P, T) - h(w T). (18)

For the derivation of constitutive equations in CDM, instead of the restrictive hypothesis of strain
or deformation equivalence, we have proposed to use the following hypothesis based on energy
considerations[ 151: the damaged material state can be replaced by a fictitious undamaged material
state which is characterized by effective stress tensor and effective strain tensor in the sense that
the total energy involved in the two processes, reversible or irreversible, should be equal. The
effective elastic strain tensor is defined by
;‘= MT.-l : Ce = N : L~.
(19)
686 C. L. CHOW and T. J. LU

When the Voigt notation for stresses and strains is used, eq. (19) is reduced to
r: = N,(D)cT, i, j = 1,2, . . . ,6

where

l-D, 0 0 0 0 0
l-D* 0 0 0 0
l-D, 0 0 0
2(1 -&)(1-Q)
W,(D)1= 0 0 . (21)
1 -D,-D,
20 -D3(1-D,)
0
1 -D3-0,
2(1 -D&l -DJ
s
l-D,-D,
When the inelastic strain in neglected, the above hypothesis is equivalent to the elastic strain energy
equivalence hypothesis due to Sidoroff_Cordebois[21].
According to the energy equivalence hypothesis, the elastic strain energy we (Lo,D, r) and the
complementary elastic energy we (a, D, T) can be evaluated as
we(ce, D, 7’) = we(P, 0, T) = se : C : P
=’ & : C : Le
(22a)
we@, D, T) = we@, 0, T) = $ : S : d
+ :S:a (22b)
where C is the elastic stiffness tensor and it has been assumed that the dependence on temperature
T is reflected only through material parameters for isothermal conditions.
Following the rules of thermodynamic theories of irreversible processes, the thermodynamic
conjugate force Y of damage tensor D is given by
a$ awe(ce,D, T)
y=qjjj= aD W)
or

Wb)
in which the superscript “s” means that the symmetric part should be taken. In the derivation of
(23), we have noted the fact that ce (or a) and D do not appear explicitly in tip@, T) even if p is
dependent on fe (or a) and D through the dissipative potential (Lemaitre[40]).
The density of elastic strain energy we is defined by
dwe=d :dce (24)
and it is possible to show that

-Y = i g at constant u and temperature T. (25)

In fact, for isothermal conditions, we can write that


awebe,D, T) = pce=~-I:~:~T.-I.Ce
tS=p . .
a@
At constant u, one follows from (26) to see that

da =C:dc’--2 M-‘:g:C ‘:Ec:dD=O (27)


>
Evolution laws of anisotropic damage 687

and from which

(28)

so that

dw’ = u : dc’ = C : cc : de’ = 2~’ *.(hP:g:C)‘:re:dD. (29)

With (29) and (23a), the validity of (25) is obvious. -Y, as such, can then be viewed as the elastic
strain energy release rate associated with a unit damage growth analogous to G in linear elastic
fracture mechanics and thus is termed “the damage energy release rate”. This energy property of
-Y makes it an extremely important quantity in the theory of CDM.

4. DAMAGE EVOLUTION LAW AND DAMAGE SURFACE


The damage evolution laws may take a very complex form which should reveal partly, the
anisotropy of the damage state and, partly, the high degree of nonlinearity of the physical
phenomena associated with the nucleation and propagation of microdefects. It is therefore
generally accepted that, the formulation of these laws has been, perhaps, if not the weakest then
at least the most arbitrary aspect in the theory of CDM[23].
The complex nature of damage growth may be revealed by the following examples. From the
damage-strain curve shown in Fig. 5 which was obtained by Chow and Wang[l9,41] for an
aluminum alloy through simple tension test, it is readily seen that D, in the direction of maximum
principal tensile stress grows much more rapidly than D2 does, i.e. d, > & and at the moment D,

12.00
t

x
g 6.00 -
a x x

5 .”
n x
L.50 - x + + ++. +
x * ++ + + +
x x ++ +++ + + +
+ + +
4% +
3.00 - + + P +
+
x + +
Xx +
+
1.50 - xx+++

0; I I I I I I
0 5.67 Il.33 17.00 22.67 26.33 34.00
STRAIN E , %

Fig. 5. The evolution curve of damage variables D, and D, for aluminum alloy 2024-T3 under simple
tension.
688 C. L. CHOW and T. J. LU

reaches its critical value which also signifies the fracture of the specimen, the corresponding value
of D, is relatively small. During the final stage of fracturing process, 6, g 0, implying that D,
approaches an asymptotic value and it is D, which dominates the specimen rupture. Nonlinear
evolution nature of microcracking is also observed in fatigue tests of short-fiber reinforced
thermoplastics. Quantitative measurements of the degradation in in-plane stiffness properties due
to microcracking have been collected by Wang et a1.[42]. From their results and the present
definition of damage through the extent of macro-mechanical material property changes, we are
able to obtain the damage evolution curves at three fatigue stress levels as shown in Fig. 6.
Obviously, D, in the direction of axial-loading grows much faster than D2 does. (SEM examination
of these fatigue-induced microcracks shows that they possess strong anisotropy relative to fatigue
loading. Generally, in a matrix-dominated area with dispersed fibers, almost all microcracks were
formed nearly normal to the loading direction. In a fiber-dominant region with fibers oriented
parallel to the loading, microcracks were developed mainly in the matrix, also normal to the
loading, but had rather small crack lengths limited by the interfiber spacing. In a fiber strand
oriented with an angle to the loading direction, microcracks usually grow along the interface
between the fiber and matrix.) It thus becomes evident that it is important to take into account
the above observations regarding the non-linear dependence of the rate of damage growth under
applied loads in the formulation of damage evolution laws. In what follows we intend to obtain
a formulation which is sufficiently simple to be applicable but indentifiable by only a few specially
designed tests while furnishing an adequately refined representation of the anisotropic phenomena.
Acoustic emission experiments on short-fiber reinforced thermoplastics clearly show that there
exists a finite region in strain space within which only elastic change occurs (Fig. 7)[43]. The
additional microcrackings in the form of fiber-end cracking, interface debonding and matrix
cracking, etc. occur only when loaded beyond the elastic region. Similar observations have been
obtained by Holcomb in geomaterials[44]. On the other hand, for ductile materials, it is commonly

0.50

O-GOt

“.__

0.50

0.10
-
---
-_-
NUMBER

u’nax=
Gmx=
um
OF CYCLES

0.60 uuts
0.60 Guts
= 0.10 Duts

NUMBER OF CYCLES
Fig. 6. The evolution of damage in short-fibre reinforced thermoplastics (SMC R-SO) during fatigue
loading (R = 0.05, f = 2 Hz).
Evolution laws of anisotropic damage 689

Fig. 7. Typical tensile behaviour of a short-fibre reinforced thermoplastics.

observed that the reduction of the elastic Young’s modulus E occurs only when the initial yield
condition is satisfied. It is therefore quite reasonable to assume the existence of a flow potential
@in the space of affinities which envelops all states which can be met without any energy dissipation
(i.e. without any irreversible changes in the microstructure of the material). Together with the
convexity and normality properties, this intuitively appealing idea presents a suitable framework
for the fomulation of analytic evolution laws of anisotropic damage.
In the present model, when the heat conduction is neglected, the total dissipative energy power
is
a+'--Rfi-Y:b-B+O (30)
where LP is the inelastic strain tensor, R and B are the thermodynamic conjugate forces of the
internal variables p and w known as the “increment of strain hardening threshold” and the
“increment of damage strengthening threshold”, respectively. According to the assumption
introduced in the last section that the energy dissipated in plastic flow and damaging processes is
uncoupled, (30) can be separated into two parts.
d :P-R@>O (3W
-Y:lb-BG>O (31b)
from which we can assume that there exist a plastic dissipative potential and a damage dissipative
potential, separately. For plastic dissipative potential, it is generally suggested[21,41,45] that it is
the von Mises yield criterion in stress space expressed in the following form
F&9, R)= d,- (&+ R(p)}=0 (32)
where the effective equivalent plastic stress d, is
CD= (9 : H : 6p2 = ($I : R : up2 (33)
690 C. L. CHOW and T. J. LU

while the fourth order tensor H specifies the plastic anisotropy. It can be easily shown that
d, = u : P&(D) (34)
where the Lagrange multiplier i,(D) is given by

(35)

where

1, ifF,=Oand~:d+~:~zO
C= (36)
0, if F,<O, or Fp=O and z:t++g:fi”O

Thus we see from (33) that the effective equivalent plastic strain 6,, which acts as the potential
function in the derivation of plastic constitutive equations, is proportional to the plastic dissipative
power and a non-negative quantity since, by the second law of thermodynamics, C-J: 8’ 2 0, in the
absence of damage and heat conduction. The plastic constitutive equations, on the other hand, can
be derived directly from (32) following the principle of maximum plastic dissipation[31].
In a similar way to the arguments leading to plastic dissipative potential, we assume that there
exists a surface Fd = 0, in the strain space, which separates the damaging domain from undamaging
domain. It is further assumed that Fddepends on the strain c only through the thermodynamic force
Y conjugate to D, i.e.
Fd= Y,-[B,,+B(w)]=O (37)
where the equivalent damage energy release rate Y, is defined by
Yd= (;Y : J : Y)liz (38)
and B,, is the initial damage hardening threshold. Following the principle of maximum plastic
dissipation[31], one may similarly assume that, for a given local history of strains, the actual
damage variables D and w are those variables rendering a maximum damage dissipation, and they
are usually called “the principle of maximum damage dissipation” which becomes the basis for the
development of damage evolution laws[46].
The principle of maximum damage dissipation may be expressed mathematically as follows.
Let
D* = {D, w}~
Y* = {-Y, - BIT
(39)
i
and introduce the convex cone
R* = {Y* E [L2(R)]‘IFd(Y*) GO}. (40)
The damage variables set D* is then characterized as the argument of the following principle

D* = argmax (pDd = -Y : b - Bti = Y*.b*}. (41)


Y*ER’

The Lagrangian functional is next introduced


Ld = pD, - &(D*)F,(Y*) (42)
where the Lagrange multiplier MD*) belongs to the positive cone
K* = {1, (D*) E L*(R) I& (D*) > 01.Therefore, we can easily obtain the following damage evolution
laws as a consequence of optimizing the functional L,,

b* = &(D*) y$ (43)
Evolution laws of anisotropic damage 691

or equivalently

b=&(D*)$+ -$J:Y
d

86
* = Ad CD*) a(- = nd(D*) (4)
where

(45)

while

d =
r 1, if Fd=O and g:u>O

I 0, if

From (43) and (37) one concludes that


Fd< 0, or if Fd= 0 and 2: Y < 0. (46)

i+ = 2($b : J : lb)“2 (47)


where J-’ is the generalized inverse of J.
When vector notation instead of a second order symmetric tensor is adopted, the three
principal components of Y can be expressed as
s..
- y’=&(l-& -Dj)
ai = ~$~crj = c,f$ (no summation for i, i = 1,2,3) (48)

for which the evolution law of principal components of D becomes

bi=&Ju(-k;).
d

5. DEVELOPMENT OF DAMAGE CHARACTERISTIC TENSOR J


The determination of a suitable damage characteristic tensor J, which is simple enough to be
applied and yet describes accurately the nonlinear nature of damage growth, may well be the most
important aspect in the present formulation of anisotropic damage evolution laws. Since the several
existing formulations of J are of more or less simplified nature, a new form of J is proposed and
its physical implications expounded.
According to the creep test results of Johnson el a1.[471 for commercial pure copper, of
Hayhurst and Storakers[48] for aluminum alloys, and of Dyson et a1.[49] for Nimonic 80A, the
grain-boundary voids and fissures were observed to form predominantly on the planes which are
perpendicular to the applied maximum tension stress rr,,. On the other hand, the creep damage
may in some cases develop isotropically as observed by Johnson[SO] for R.R. 59 aluminum alloy.
In view of these experimental results, Murakami and Ohno[l3] introduced the following general
evolution law of creep damage

where r and F’), Qc” are a scalar function and fourth rank tensor functions of d and M(D) while
fIti) and fi$ denote the principal directions corresponding to the positive principal values of the
effective stress tensor d and its deviator S, = d - $(tr 8)1, respectively. Following the suggestion
by Hayhurst[6] that the creep rupture times of polycrystalline metals can be related to a linear
combination of the maximum tensile principal stress, the octahedral shear stress and the hydrostatic
692 C. L. CHOW and T. J. LU

stress, a simplified form of (49) was then taken as [13]

~=B,[a,(+trS2,)“2+/?0 d(“+(l -a0-f10)(ftrB)]k[tr@#12[~01+(1 -~o)fi(‘)@fic”] (51)

where d(” is the maximum positive principal stress of 8, QI, = (1 - D)-‘, and B, k, 1, ao, POand ‘lo
represent the material coefficients. We can see from (40) that, in the principal system of coordinates,
the damage characteristic tensor J in Murakami-Ohno’s model may be expressed as

1
1 0 0
[Jl= 0 & 0 , 0 < ?o< 1 (52)
[0 0 ‘lo

which has also been adopted by Sidoroff and Cordebois[21-221 to study elastic-plastic damage
evolution, A more simpler model of (49) was used by Murakami and Sanomura[Sl] to study creep
damage of commercial copper under complex states of stress
b= B,[a,d(‘)+(l -ao)(~trS~)“2]kii(‘)~~ii(‘) (53)
which is the special case of (50) when I]~= 0. The limitation of the above model due to
Murakami-Ohno lies in the fact that it is “too simplistic” to take into account the coupling effects
between the principal stresses which is obviously insufficient to describe those material damages
caused by complex stresses, such as in the damage process-zone ahead of a macrocrack tip. Even
in the case of simple tension of a smoothed cylinder, using (51) will result in a nonzero value of
damage, D, , in the direction of applied load only, with zero damage values in other directions, i.e.
D2 = D, = 0. This is of course contrary to experimental results (Fig. 5).
An alternative formulation of J was proposed by Chaboche[l l] for creep damage based on
the assumption that the degree of anisotropy of the deteriorated material could be expressed by
means of only two coefficients, a and /I, with /I = 1 corresponds to the perfectly isotropic case and
/I = 0 stands for the ideal anisotropic case. The resulted form of J is

1 0 0 0 0 0
u-mv 1 o o o
0
l-v
-
[Jl = -010
(1 P)v 0 0 (54)
l-v
0 0 0 0 0 0
0 0 0 0 l+a-a/3 0
0 0 0 0 0 l+a-a/j I_

Nevertheless, the physical ground for generalization of this model, which is based on some
homogenization results on elastic media weakened by completely parallel microcracks in order that
the unknown coefficients involved in the model may be reduced, to more complicated cases such
as plasticity, creep and fatigue problems is questionable.
Recently, Lee er a/.[521 proposed to use the following form of J

1 CL P
[Jl= p 1 P (55)
[P P 11

to describe anisotropic damage evolution. However, the assumption that ~1is a material constant
will be proved not realistic in later discussions.
On the other hand, Chow and Wang[41] noticed from eqs (47)-(48) that “the components of
damage increment depend upon the sign of stress/strain components in work done”. As a
consequence, a negative component of work done involving a negative increment in damage, i.e.
6,~; < 0 may imply a “decrease” in material degradation irrespective of whether dl > 0 and ~7 < 0
or a”,< 0 and c;I > 0, a phenomenon which is neither acceptable from physical consideration nor
Evolution laws of anisotropic damage 693

commonly observed in practice. One way to remove such anomalies is to replace the equivalent
damage energy release rate Y, in (36) with an “effective damage stress” cd defined as
zd = ($6 : J : $)“* = (4~ : j: o)‘/* (56)
where 3 = MT : J : M, such that the new damage criterion becomes
Fd=6d-[[B,,+B(w)]=0. (57)
Taking again the damage criterion of (56) as the damage dissipative potential, the damage evolution
laws are then derived as

(58)
However, in the above derivation, the thermodynamic conjugate force of D is assumed to be the
stress tensor u instead of -Y which is arbitrary, since, for example, u : 1) does not have the unit
of energy power as that of the damage energy dissipative rate -Y : b. Another uncertainty in this
approach comes from the damage evolution law of (57) which implies that the effects of either
tensile or compressive stress to damage rate are the same in magnitude but opposite in sign which
is generally in contrast with available experimental results[20,53] which reveal that, under
compressive loading, serious damage with significant anisotropy occurs and increases monotoni-
cally with load increment.?
Another way to surmount the anomalies raised in [41] is to set certain restraints on the damage
characteristic tensor J and retain the physically reasonable damage criterion of (36). The physical
reasoning that supports the selection of (36) is as follows. The most general form of damage
criterion in the space of affinities is
Fd( YV)= 0. (59)
For initial isotropic material, the form of Fd should not be changed under the transformation of
coordinate systems. Consequently, (58) may be rewritten as

Fd(Y,, Yu, Yin) = 0 (60)


where Y, , Y,,,Y,,,
are the first, second, and third invariants of the damage release rate YU.Assuming
that the damage criterion has the same form (no change in magnitude and direction) for both
tension and compression cases, Fdshould then also not be changed after the stress reversal. In other
words, F must be an even function of Y, and Y,,,, respectively. In the present model, we have defined
a damage criterion of (36) which is a function of Y,, only because Y,, is intimately related to the
dissipative damage energy power ti,, = -Y : b. Indeed, from (43) we have

Y,,=jY:J:Y=-$Y:b=+,. (61)

On the other hand, since Y,, = Yi, (60) is reduced to


Ydti = d (62)
or
y = -Y:dD
d (63)
dw *
Thus we see that the equivalent damage energy release rate Yd is the energy dissipated through
damaging required to cause an unit overall damage increment. In other words, the damage criterion
of (36) is in fact a dissipative damage energy criterion.

tThe elastic modulus E and the Poisson’s ratio v of a typical brittle solid such as concrete or rock under uniaxial
compression increase somewhat during the initial stage of loading; this is due to the fact concrete or rock is a specific kind
of material with numerous initial microcracks, which when subjected to compressive loading, those defects oriented normal
to the loading tends to “close” or become passive damage and thus “enhance” the material moduli; however, with further
loading, those cracks parallel to compression tend to open because of lateral extention and the moduli decrease
consequently. The so-called “active damage” and “passive damage” have been discussed in [16].
694 C. L. CHOW and T. J. LU

In the principal coordinate system, the most general form of J is

Although it is not an absolute necessity to require J to be symmetric, J is taken here to be symmetric


for the following reasons. A typical form of (37) when expanded in the principal coordinate system
is (iJuYiYj) v2. When i #j, there are thus two terms in the bracket with the same variables, i.e.
fJ, Yi Yj and $Jji Yi Yi and which, when added, give f(Ju + Jji)Yi Yj. If then J is not symmetric we
can substitute for it a symmetric tensor J’ such that J;i = Jii and Jk = Jii = i(J, + Jji) without
changing the value of Yd.
Upon substitution of (63) into (48) and making use of (47), the following expression for D,
is obtained

1
=I
(1 -D,)2 (1 -D,T;; -D,) (1 -D,;; -D3)
1
X a2 (65)
(1 - D2)’ (1 -D,,; -D,)

S =3

=-& {a)TIZ]{O), i = 1,2,3

where

(66)
and

Jil - vJi, - vJ,,


(1 -D,)3 (1 - D,)2(1 - D2) (1 - D,)2(1 -D,)

[.q] =
vJiz
- J,2 - vJn . i=l,2,3 (67)
(1 - Q)U - D212 (1 - D213 (1 - D212(1 - D3)

-vJi3 - vJi3 Ji3


_ (1 - D,)(l - D3)’ (1 - D2)(l - D3)2 (1 - D313

Within the context of the present model of CDM, it is reasonable t oa ssume that a “final”
negative damage growth rate should not exist, or in an alternative way, damage per se cannot
“strengthen” the material and we should always require that b 2 0. The word “final” is used here
to intentionally exclude the momentary “strengthening” behaviours of rocks or concretes under
compression due to “closure” or “healing” of initial microdefects[20] whose damage becomes
“finally” positive. According to eq. (64), [A] of (65) must be a positive semidefinite tensor so that
D, (i = 1,2,3) is non-negative for all loading spectrums. After trivial derivations, it is found that
the necessary and sufficient condition to fulfill the above requirement is
Jii 2 0 (68)
which enables us to retain (36) as the damage criterion without violating the thermodynamic
condition that b 2 0.
Evolution laws of anisotropic damage 695

Next we will show the components of J are not material constants. Consider, for example,
the problem of bi-axial loading which, according to (43), has the following components of damage
growth rate

Jllyl+Jl*y2+Jl,y3
J2, Y, + J22Y2 + J2s Y, (6%
JH Y, + J32Y2 + J,,Y,

where, from (47)

1
VP
---
I-02

1
-- -
P (70)
+1-D2
0

while

1
VP
---
l-D2

1
1 --
VP P2
g2=(l -D2)2 l-D,+- ’ (71)

From (69) one obtains


dD, _ J,,g, + Jngz
(72)
dD2- Jag, + Jza *
Assume that J,, =J 22= 1 (the reason of doing this becomes clear later) and note the symmetric
property of J, one can solve from (72) for J,2 such that
g, - &?2 1 - aLr2/g,)
J --=
(73)
‘2-&?, -g2 r7. - L?2/g,

where

g2
-=- l-4 - VP + P2(1 - D,)/(l - D2)
(74)
g, l-D, (1-D,)/(l-D,)-VP
when P # 0, it is readily seen from (73) and (74) that J,2 depends upon the damage state (and hence
varies with the position and orientation of local material element V, due to the anisotropic nature
of damage), the material type, the external loads as well as the load history. When P = 0 in the
case of uniaxial loading, (73) reduces to
1 dD2
J,2=;i =dD (75)
I
Integrating one may obtain

J ,2=- ; 6,. (76)


I
696 C. L. CHOW and T. J. LU

However, even in this simple case, J,2 is not a material constant since from the experimental results
of Fig. 5, one calculates that

J,* & 1.0, when t % < 0.5%

Jlz & 0.73, when C% = 1.37% (77)

J,* & 0.48, when 6% = 20.70%.

Generally, J,2 changes continuously from 1.0 when no damage occurs to 0.0 at final rupture. One
concludes therefore that any formulation of J which fails to recognize the above feature would be
inadequate to describe damage growth behaviours.
The purpose of introducing a damage characteristic tensor J is, above all, like the introduction
of plastic characteristic tensor H in the theory of plasticity, to account for the anisotropic nature
of damage growth. An appropriate formulation of J should therefore be intimately correlated with
the real damage growth behaviours, avoiding as far as possible any arbitrary aspect in the
approach, usually of inductive nature. As to the physical implications of J, consider, for instance,
a cylinder under simple tension (or compression) along the principal axis x, (Fig. 8a). The elastic
strain-stress relation for a damaged material is

(78)

which, when expressed in the principal coordinate system, takes the following form for the

DI) D2

do;+Q
t i

Ib)

-Y*
RELOADING
,Dl

Fig. 8. (a) A cylindrical bar subjected to simple tension. (b) Anisotropic nature of damage propagation.
(c) Damage hardening behaviour.
Evolution laws of anisotropic damage 69-l

present example
01 61
“=E(l -D,)2=g
_
E2- - ---d
‘- E(l-D,;;l-D,)= yd;’ ’

Vbl VI3

“= - E(1 -D,)(l -D,)=


---CT
E, ’
(784
where

from which we can measure the magnitudes of damage variables


D, 1u, = 1 - (E, /E)2

D,Io,= 1 -&(1 -0)

D,Io,= 1 +l - 0,). (80)

Consequently we can generally obtain the damage growth curves shown in Fig. 8(b). From (48)
we have

y2 = Y3 = 0 (81)

which, supplemented by Fig. 8(b), displays the - Y,-D, curve of Fig. 8(c) where the damage
hardening behaviour is also depicted. Upon substitution of (81) into (69) we get

0:
(82)
W-D,)’

from which
dDz dD, da =5
dD do,=?, (83)
I II dD, ’ J,,’
Take J,, = 1, then we have

J=%jg
2’ dD,
19 J3, =a dD3
da I. (84)

When there is a load increment, all the three damage components, D,, D2 and D3, will have
corresponding increments, dD, , dDz and dD,. Usually, dD, # dD2 # dDj due to the anisotropic
nature of damage growth (Fig. 8b). Thus, from (84), we see the physical implication of J2, and J3,
is that they measure the extent to which dD2, dD, deviate from dD, due to a load increment da,
in the x, direction. Recognization of this underlying physical meaning of J is very important in
the understanding and description of anisotropic damage evolution, which also promotes us to
define J as:
DeJinition. The component of the damage characteristic tensor J is defined by

Jv=$ daj, i,j=l,2,3


J
698 C. L. CHOW and T. J. LU

which is the damage increment dD, in the ith direction when an unit damage increment is induced
in the jth direction by a load increment daj in the same direction.
Consequently, the assumption J,, = 1 in the foregoing derivations is appropriate since, from
(83), we have
Jii= 1, i= 1,2,3. (86)
On the other hand, for initial isotropic media, it is obvious that

J =dDIda =dD2da = J
” dDz I ’ dD, I ’ 2’

J,2 = 2 da, = 2 da, = J,3 = J3, (87)


2 I 3 I

J,2 = 2 da, = 2 da, = J,2 = J23.


2 I 2 I

1
Thus we conclude the damage characteristic tensor J takes the following form in principal
coordinate system

1
J,2 J12
[Jl = 1 J,2 W3)
[ S 1
and the identification of J is reduced to the determination of only one parameter, i.e. JIz, which
is very simple!
Using the uniaxial tension test, we can obtain at least two damage characteristics. One is D,, 1a,
which denotes the critical damage value that can be induced by the load applied along x, direction
while the other one is the largest damage value Dzc1a, in x2 direction that can be generated by this
same load. For aluminum alloy 2024-73, according to Fig. 5, Dlcla, = 0.114 and D&la, = 0.044.
Usually, when D2 I a, has reached its critical value DZEI a, further load increment in x, direction can
no longer induce additional damage in x2 direction, i.e. d, Ida, = 0 or J,2 decreases to zero. Further
increment in D2 can only be generated by load applying in x2 direction. The J12-0, curve for
aluminum alloy 2024T3 is shown in Fig. 9. With this curve and eq. (80), we can describe the
anisotropic nature of damage growth (Fig. 5, for example), satisfactorily.
Under the complex loading case {da} = (da,, da,, da, IT, the total damage increment dD,l da
in ith direction is decomposed into three parts
dDiI da = dDiI da, + dDiI da, + dD,I da, (89)
where by dD,I daj (j = 1,2,3) we mean the increment of damage in ith direction caused by a load
increment da, in the jth direction. In the initial stage of loading, damage is dilutely distributed all
over the material body and can be considered as isotropic, hence dail daj = dDjl doi and Jti = 1 for
. .
= 1,2,3. Nevertheless, with further loading, damage component dDiI dai grows increasingly
baiter than dD,I da. (i #j) does and the former eventually dominates the growth of Di.
j)
Accordingly, Ju (i # changes gradually from 1 to smaller values. When @ dDiI da, = Dil aj (i #j),

Fig. 9. The J,,-E curve for aluminum alloy 2034-T3, where t, is the failure strain of the material.
Evolution laws of anisotropic damage 699

which is the total amount of damage in ith direction induced by load applying in jth direction up
to aj, reaches its critical value D, 1oj, a material characteristic, further increase in rsj, i.e. uj + daj,
can no longer generate any damage in ith direction corresponding to the zero value of Jo (i #j).
However, it is emphasized that the total damage in ith direction, Dilda, may be subjected to further
increase if external loads other than aj (i #j) exist. In the present model, we have from (69) for
multiaxial loading

Since the direction of {Y,, Y,, Y,} coincide with those of (6,) IT*, 03}, and with the experimental
curve for J,2 such as Fig. 9 for aluminum alloy 2024-T3, the anisotropic behaviour of damage
growth under complex loading can thus be adequately described by eq. (88).
Now we turn to the discussion of the damage criterion of (37) again. Under the assumption
of plane stress, the equivalent damage energy release rate is
Y, = ($Y: J : Y)“’ = (Y; + 2J,2 Y, Y, + Y:y2

1
(91)
in which we have used the following generalized expression for J

1
J12 J,2
[Jl=2 1 J12 . (92)
[ s 1

The addition of a multiplier “2” is to ensure that, for uniaxial case, Yd is reduced exactly to Y, .
The damage criterion of (37) can then be changed to the following form

” ” (&>‘= 1 (924
m+~w$+B)+
and Y, , Y, are calculated from eq. (70). Apparently, the locus of this damage criterion is dependent
upon the magnitude of anisotropic ratio J,2. In the initial stage of loading, J,2 = 1, the damage locus
is the curve a in Fig. 10. As the load increases, J,2 decreases gradually until it reaches zero when
rupture of the local material element occurs and damage locus is accordingly shifted from curve
a to curve b and finally to curve c.
In summary, the evolution of anisotropic damage growth under both simple and complex
loading spectrums can be adequately described by the present definition of a damage characteristic
tensor J which involves only one unknown variable Ji2 to be determined. Identification of J,2 is
easy since a simple tension test is all that is required. The J12curve can be considered as an intrinsic

-Y,/IB.+B)
t

Fig. 10. Damage loci in the space of thermodynamic conjugate forces.


700 C. L. CHOW and T. J. LU

property of the specified material, similar to the basic stress-strain curve. It is recommended
therefore that the simple tension test for J,z determination be standardized.

6. CONCLUSIONS
In the present communication, damage is defined phenomenologically or experimentally
instead of analytically or microscopically, arguing that a phenomenological theory like CDM itself
does not need to describe those microstructural occurrences leading to material degradation in any
direct sense. An appropriate local material element VMembedded with a cloud of microdefects is
selected such that the defined damage variable represents the average measure of material
degradation within VM. The damage hardening behaviour is also described, different from most
other damage models. The constitutive equations of the damaged material follow directly from the
thermodynamic considerations with internal variables, together with the hypothesis of energy
equivalence. The thermodynamic conjugate force of the damage variable, the “damage energy
release rate”, is derived according to the established form of the Helmholtz free energy which is,
in turn, based on the phenomenological nature of the defined damage variable. The evolution laws
of anisotropic damage are developed by adopting the damage surface concept as well as the
principle of maximum damage dissipation. The model is completed by introducing a new damage
characteristic tensor J which involves only one unknown variable to be determined. The developed
model of anisotropic damage growth can be conveniently implemented into a finite element
program to solve practical engineering problems from which initial results have shown promise.

REFERENCES
L. M. Kachanov, On the creep fracture time. Izu. A/cud. Nuuk U.S.S.R. Otd. Tekh. 8, 2631 (1958).
Y. N. Rabotnov, Creep Problems in S~ructurul Members. North Holland, Amsterdam (1969).
F. A. Leckie and D. R. Hayhurst, Creep rupture of structures. Proc. R. Sot. Lond. A340, 323-347 (1974).
F. A. Leckie, The constitutive equations of continuum creep damage mechanics. Phil. Truns. R. Sot. Lond. A288, 27
(1978).
F. A. Leckie and D. R. Hayhurst, Constitutive equations of creep rupture. Acta Mefull. 25, 1059-1070 (1977).
D. R. Hayhurst, Creep rupture under multiaxial state of stress. J. Mech. Phys. Solids 20, 381-390 (1972).
J. Lemaitre, Evaluation of dissipation and damage in metals, submitted to dynamic loading. Proc. ICM 1,Kyoto, Japan
(1971).
PI J. Lemaitre and J. L. Chaboche, A nonlinear model of creep-fatigue damage evaluation and interaction. Proc. IUTAM
Symp. of Mechanics of Visco-Elastic Media and Bodies. Springer, Gothenburg (1974).
[91 J. L. Chaboche, Description Thermodynamique et Phenomenologique de la Visco-plasticite Cychque avec Endom-
magement. Univ. Thesis of Paris VI and published in ONERA, 3 (1978).
J. Lemaitre, A continuous damage mechanics model for ductile fracture. J. Engng Muter. Technol. 107,83-89 (1985).
;t:; J. L. Chaboche, The concept of effective stress applied to elasticity in the presence of anisotropic damage.
EUROMECH Colloque 115, Grenoble Edition du CNRS (1979).
J. L. Chaboche, Continuum damage mechanics, part I and part II. J. uppl. Mech. 55, 59-72 (1988).
;t:; S. Murakami and N. Ohno, A continuum theory of creep and creep damage. 3rd IUTAM Symp. on Creep in
Structures, Leicester (1980).
P41 S. Murakami, Notion of continuum damage mechanics and its application to anisotropic creep damage theory. J.
Engng Muter. Technol. 105, 99-105 (1983).
[I51 C. L. Chow and T. J. Lu, A continuum damage mechanics approach to crack tip shielding in brittle solids (submitted
for publication).
[I61 D. Krajcinovic, Continuous damage mechanics revisited: basic concepts and definitions. J. app. Mech. 52, 829 (1985).
E. E. Underwood, Quantitative Srereology, pp. 25-27. Addison-Wesley, Reading, Massachusetts (1970).
tt:; S. Jansson and U. Stigh, Influence of cavity shape on damage parameter. J. uppl. Mech. 52, 6099614 (1985).
v91 C. L. Chow and J. Wang, An anisotropic theory of elasticity for continuum damage mechanics. Inr. J. Fracture 33,
3-16 (1987).
PO1 E. D. Case; The effect of microcracking upon the Poisson’s ratio for brittle solids. J. Mater. Sci. 19, 3702-3712 (1984).
plasticity. J. Mech. Theor. Auul.,
PII J. P. Cordebois and F. Sidoroff. Anisotronic damage in elasticity_ and _ __ Numero Special,
45-60 (1980).
P21 J. P. Cordebois and F. Sidoroff, Damage induced elastic anisotropy. Euromech 115, Villard de lans (1979).
1231D. Krajcinovic, Constitutive theories for solids with defective microstructure, in Damage Mechanics and Continuum
Modeling (Edited by N. Stubbs and D. Krajcinovic), pp. 39-57. ASCE, New York (1985).
J. Lemaitre and J. Bataille, Damage measurements. Engng Fracture Mech. 28, 643661 (1987).
;z; P. Suquet, Plasticite et homogeneization. These, University Pierre et Marie Curie, Paris (1982).
WI E. Sanchez-Palercia, Nonhomogeneous Media and Vibration Theory, in Lecture Nores in Physics. Springer,
Gothenburg ( 1980).
[271 A. C. F. Cocks and F. A. Leckie, Creep constitutive equations for damaged materials. T. & A. M. Report No. 480,
University of Illinois at Urbana-Champaign (1986).
Evolution laws of anisotropic damage 701

PI D. A. Allen, C. E. Harris and S. E. Groves, A thermomechanical constitutive theory for elastic composites with
distributed damage--I. Theoretical development. Int. J. Solids Struck 23, 1301-1318 (1987).
[291E. S.-M. Chim, Tensile fatigue damage and degradation of random short-fiber SMC composite. Ph.D. Thesis,
University of Illinois at Urbana-Champaign (1985).
[301D. Krajcinovic and D. Fanella, A micromechanical damage model for concrete. Engng Fracture Mech. 25, 585-596
(1986).
R. Hill, The Mathematical Theory of Plasticity. Oxford University Press, Oxford (1950).
t:;;D. R. Bland, The two measures of work-hardening. Proc. 9th Int. Cong. Appl. Mech., Bruxelles (1956).
B. D. Coleman and M. E. Gurtin, Thermodynamics with internal state variables. J. Chem. Phys. 47, 597-613 (1967).
[ii;J. R. Rice, Inelastic constitutive relations for solids: an internal variable theory and its application to plasticity. J. Mech.
Phys. Solids 19, 433455 (1971).
(351J. Kestin and J. Bataille, Irreversible thermodynamics of continua and internal variables, in Continuum Models of
Discrete Systems, pp. 39-67. Univ. of Waterloo Press, Ontario (1977).
D. Krajcinovic, Creep of structures-A continuous damage mechanics approach. J. Struct. Mech. 11, l-l 1 (1983).
;:;;H. Conrad, The role of grain boundaries in creep and stress rupture, in Mechanical Behauiour ofMaterials at Eieuated
Temperatures (Edited by J. E. Dom). McGraw-Hill, New York (1961).
[381J. R. Low, Jr, Microstructural aspects of fracture, in Fracture of Solids (Edited by D. C. Drucker et al.), Met. Sot.
Conf., Vol. 20. Interscience, New York (1963).
t391D. Krajcinovic and G. U. Fonseka, The continuous damage theory of brittle materials--Parts I & II. J. uppl. Mech.
48, 809-824 (1981).
J. Lemaitre, Private communication (1988).
::; C. L. Chow and J. Wang, An anisotropic theory of continuum damage mechanics for ductile fracture. Engng Fracture
Mech. 27, 547-558 (1987).
1421 S. S. Wang, E. S.-M, Chim and H. Suemasu, Mechanics of fatigue damage in random short-fiber composites--Part
I & II. J. uppl. Mech. 53, 339-353 (1986).
1431 P. T. Curtis, M. G. Bader and J. E. Bailey, The stiffness and strength of a polyamide thermoplastic reinforced with
glass and carbon fibers. J. Muter. Sci. 13, 377-390 (1978).
]441 D. J. Holcomb, Discrete memory in rock: A review. J. Rheology 28, 725 (1984).
]451 C. L. Chow and J. Wang, An anisotropic continuum damage theory and its application to ductile crack initiation,
in Damage Mechanics in Composites (Edited by A. S. D. Wang and G. K. Haritos), pp. l-9. The Winter Annual
Meeting of ASME (1987).
]461 J. J. Marigo, Modelling of brittle and fatigue damage for elastic material by growth of microvoids. Engng Fracture
Mech. 21, 861-874 (1985).
1471 A. E. Johnson, J. Henderson and V. D. Mathur, Combined stress creep fracture of a commercial copper at 250°C.
Engineer, London 202, 261-265 and 299301 (1956).
1481 D. R. Hayhurst and B. Storakers, Creep rupture of the Andrade shear disk. Proc. R. Sot. Lond. A349,369-382 (1976).
[49] B. F. Dyson, M. S. Loveday and M. J. Rodgers, Grain boundary cavitation under various states of applied stresses.
Proc. R. Sot. Lond. A349, 245-259 (1976).
[50] A. E. Johnson, Metall. Rev. 5, 57 (1960).
[51] S. Murakami and Y. Sanomura, Creep and creep damage of copper under multiaxial states of stress, in Plasticity Today
(Edited by A. Sawczuk and G. Bianchi), pp. 535-551. Elsevier, London (1985).
[52] H. Lee. K. Peng and J. Wang, An anisotropic damage criterion for deformation instability and its application to
forming limit analysis of metal plates. Engng Fracture Mech. 21, 1031-1054 (1985).
[53] A. Dragon and Z. Mroz, A continuum model for plastic-brittle behavior of rock and concrete. Int. J. Engng Sci. 17,
121-137 (1978).

(Received 18 November 1988)

You might also like