Engineering Fracture Mechanics Vol. 21, ~0. 5, pp. 547-558 1987 0013-7944/87 $3.00 + 0.

00
Printed in Gnat Britain. $ 1987 Pergamon Journals Ltd.

AN ANISOTROPIC THEORY OF CONTINUUM

DAMAGE MECHANICS FOR DUCTILE FRACTURE
C. L. CHOW and JUNE WANG*
Department of Mechanical Engineering, University of Hong Kong

Abstract-The development of an anisotropic theory of continuum damage mechanics for ductile

fracture is described. A new anisotropic damage evolution equation and a constitutive equation
of plasticity are formulated using a damage effect tensor g(D) proposed recently by the authors.
This is followed by the development of a generalized damage characteristic tensor z to characterize
anisotropic damage evolution which is shown to be compatible with the damage effect tensor.
The tensorial equation of damage evolution for the proposed tensor z can be reduced to a scalar
equation if the damage state is isotropic. Verification of the proposed model has been performed
by conducting simple tension tests and the results are shown to be satisfactory.

1. INTRODUCTION
DUCTILE fracture has been the subject of considerable interest by fracture mechanics researchers
because of its practical engineering importance[l]. This has prompted the development of a number
of ductile fracture criteria including notably COD[2,3], Jintegral[4] and strain-energy density[5].
Considerable doubts have recently been cast on the use of COD and J-integral as valid single-
parameter fracture criteria for ductile fracture similar to the stress intensity factor for brittle
fracture[b8]. The energy density criterion has, on the other hand, shown to be viable for a variety
of structural applications[9-111. More recently, the theory of continuum damage mechanics
originally proposed by Kachanov[12] has been extended to solve a series of important practical
engineering problems which are however restricted to isotropic material damage[ 13-l 51.
Recent experimental observations confirm the development of anisotropic plasticity and
damage at probable failure sites even if the original material property is isotropic[16-18). Any
failure analysis failing to take into account the anisotropic effects of the material under investigation
into consideration may therefore yield questionable conclusions.
Researchers have in the past focused their attention mainly on the development of damage
theory for brittle and creep failures[19-211, although more involved analysis is needed to formulate
damage models capable of providing meaningful solutions to failure problems more commonly
observed in practice[19,22]. For the case of ductile damage, a simple yet effective damage model
was proposed by Lee, Peng and Wang[ 181 based on the work of Sidoroff and Cordebois[23,24]
and was successfully adopted to predict the forming limits of sheet metals. The damage effect
tensor M(D) for the effective stress equation and the damage characteristic tensor J developed for
the damage evolution model were however limited to a few special cases with a priori knowledge
of principal stress/strain directions such as the forming limit analysis, pressurized thick cylinders,
etc. For solutions of most practical engineering problems, the condition imposed is however too
restrictive. The authors have recently proposed a generalized damage effect tensor M(D) capable
of being applied to solve common structural engineering problems and then successfully derived
the constitutive equations of elasticity based on the proposed tensor[17].
This paper presents an investigation which further extends the application of the proposed
damage effect tensor for the development of the constitution equations of plasticity and for the
development of a new damage evolution equation. A generalized damage characteristic tensor J
in terms of general coordinate system is also developed to successfully characterize anisotropic
damage evolution.

2. ANISOTROPIC DAMAGE VARIABLES

Based on the concepts of the continuum damage mechanics, structural materials deteriorate
at different stages of loading. The deterioration is postulated to be attributed to the formation

*Currently on leave from Huazhong University of Science & Technology, China.

EFM
2):5-D
547
548 C. L. CHOW and JUNE WANG

and realignment of dislocations, micro-cracks, voids and other types of material defects and
flaws[19]. A damage variable which provides a measure of the change of an element surface area
S to s” after loading is defined as
S-s"
DC- (2.1)
s .

The corresponding effective Cauchy stress tensor 6 related to the effective resisting area s” is
postulated to be
s 6
(2.2)
==V=l-D
which may be expressed in a generalized form of anisotropic damage as

5 = M(D) : G (2.3)

where the symbol (:) means the tensorial product contracted on two indices. The damage effect
tensor M(D) is a symmetric 4th order tensor of 21 independent elements. Chow and Wang[17]
recently-proposed a formulation of the tensor taking into account the anisotropic material damage
in the principal coordinate system as

1
0 0 0 0 0
1 -D,
1
0 - 0 0 0 0
1 - D,
1
0 0 - 0 0 0
1 - D, (2.4)
N,@)l = 1
0 0 0 0 0
J(l - D3(1 - 03)
1
0 0 0 0 0
&l - D,)(l - D,)
1
0 0 0 0 0
,,/(l - D,)(l - D:

The above tensor was generalized using the law of coordinate transformation. Constitutive
equations of elasticity were derived using the concept of elastic energy equivalence. The equations
under tension may be expressed as

01 01
G=
E(1 - D,)*= E

"a, “I2
ES = - r a, (2.5)
E(1 - D,)(l - D,)= E

&e = - "01 “13

T (3,
3
E(1 - D,)(l - DJ = E

where

B = E(1 - D,)*
0 12 = VU - WU - 02) (2.6)

F 13 = “(1 - QM1 - 03)

are the effective Young’s modulus and Poisson’s ratios respectively. The damage variables D,,D,
and D, may be accordingly evaluated as
 Anisotropic theory of continuum damage mechanics 549

D, = 1 - (E/E) I’*; D, = 1 - k (1 - Dl)

(2.7)
0,=1-k (1 - 0,)

which may be used for the experimental determination of the damage variable.

3. CONSTITUTIVE EQUATIONS FOR PLASTICITY

In the damage characterization of materials undergoing large plastic deformations, constitutive

equations of anisotropic plasticity have been developed in the principal directions [ 181. For general
engineering problems, it is necessary to generalize the constitutive equations in any arbitrary
coordinate system incorporating the damage effect tensor of eq. (2.4). By invoking von Mises
yielding criterion, the yielding of material with damage is defined as

F,(a, D, R) = F,(& R) = eP - {R, + R(p)} = 0 (3.1)

where & = initial strain hardening threshold, R(p) = increment of strain hardening threshold
and 4 = effective plastic equivalent stress. The form of effective plastic equivalent stress chosen

(3.2)

where

fi = MT(D):H:M(D). (3.3)

If the vector notation for stress and strain is used, the positive semi-definite tensor H for orthotropic
materials is represented by 6 x 6 matrix as in principal coordinate system

G+H -H -G 0 0 0
H-IF -F 0 0 0
F+G 0 0 0
WI = s 2L 0 0 (3.4)
2M 0
2N

where F, G, H, L, M and N are parameters characterizing the current state of plastic anisotropy[25].
It can be readily verified that for isotropic materials, F = G = H = 1 and L = M = N = 3. The
increment of strain hardening threshold may be represented by a power law

R(P) = KP” (3.5)

where p = accumulative plastic strain, and K and n are material constants.

Following von Mises Theorem in the development of the plastic constitutive equations via
plastic potential[26], the constitutive plastic equations incorporating material damage may be
derived by taking the yielding criterion of (3.1) as the potential function
550 C. L. CHOW and JUNE WANG

where the Lagrange multiplier kp is evaluated as

; if F, = 0 and 2 : ir + g : D > 0.
a/)= (y(g) PO
(3.8)

i 0 ; if F, < 0,or F, = 0 and 2 :6 + g : D < 0.

From eqs (3.6) and (3.7) with eq. (3.2) the accumulative plastic strain rate i, is derived as

@=2 +-I :iP u2

= 2 f (d’iT[fi]-l{q2 (3.9)
( 1 [

where [a-’ is the generalized inverse (or g-inverse) of the singular matrix [a [18].

4. DAMAGE EVOLUTION
Consititution equations for damage evolution is required in the theory of continuum damage
mechanics due to the fact that a new variable (i.e. a damage tensor), is introduced in the theory.
A damage criterion in a quadratic homogeneous function of the damage strain energy release rate
Y was proposed as [23, 241

Fd = Y;i'
- (II0
+ B(P))= 0 (4.1)

where p = overall damage, B0 = initial damage threshold and B(P) = increment of damage
threshold depending on p and also other parameters to be discussed later. The quadratic function
Y,, was proposed to assume the form as
Y,, = ; YT : J : Y (4.2)

where J was a 4th order symmetric tensor, and Y known as the damage strain energy release rate
was formulated as [24]

y= -E’:(~4g:QS:Ee
(4.3)
c-1 : M-1. aM s
= -_B: : 0.
( . aD 1

The second invariant Y,, of damage strain energy release rate defined in eq. (4.2) suffers certain
anomalies which become evident when the three principal components of Y with vector notations
instead of second order tensor are expressed as

a, c,; '

- Y, = 1 _ oi(l _ D,)(l _ o,) Oj = &i Ci'q = 6j4 (4.4)

(no summation for i, i = 1, 2. 3)
for which the damage evolution equation is

B
Di = 2 JI,(- r,). (4.5)

It can be observed from eqs (4.4) and (4.5) that the components of damage increment depend upon
the sign of stress/strain components in work done. For instance a negative component of work
 Anisotropic theory of continuum damage mechanics 551

done involving a negative increment in damage, i.e. &c; < 0 implies a decrease in material
degradation irrespective of whether 4 > 0 and E$ c 0 or 4 c 0 and 4 > 0, a phenomenon
which is neither acceptable from physical consideration nor commonly observed in practice. In
order to remove these anomalies, we propose to replace the damage strain energy release rate Y
with the effective damage stress while preserving the form of quadratic function as

Fd = 8, - [B,, + B(p)] = 0 (4.6)

where the effective damage equivalent stress ed, similar to the effective plastic equivalent stress
4 defined earlier, is expressed as

(4.7)

where

5 = M’ : J :ljg. (4.8)

The damage characteristic tensor J has somewhat different physical meaning from that in eq.
(4.2), but characterizes the damage response of the material to the effective stresses instead.
If we again take the damage criterion as the potential function, the damage evolution equations
can be similarly deduced as

(4.9)

(4.10)

where

ifF,=Oand$:t+g:b>O

; if Fd < 0, or Fd = 0 and -
aFd: 15 + -
ac

Combining (4.9), (4.10) and (4.7), the overall damage rate /I is derived as

(4.12)

The damage criterion as defined in eq. (4.6) is based on the assumption that the magnitude
of damage threshold is the same in tension as in compression. In addition, the damage evolution
eq. of (4.9) implies that the effect of either tensile or compressive stresses to damage variables
are the same in magnitude but opposite in sign. Some experimental verifications to support the
above assumptions have been reported in literature[27]. Under uniaxial compression, initially low
values of Young’s modulus, E, and Poisson’s ratio, v, of rocks were found to increase with
compressive loading which implies that D, < 0 from eq. (2.7). The assumption that the stress-
damage relationship of (a - D) is the same in compression as in tension is made as a mathematical
simplicity and may not precisely reflect the material behaviour. But this assumption is considered
reasonable in view of the fact that similar assumption is often made for conventional stress analysis
552 C. L. CHOW and JUNE WANG

il
E F

B
t
(d)

I
E
F

t I3
(el (f)

Fig. 1. Plastic damage evolution process.

of most engineering materials undergoing elastic and/or plastic deformation. If necessary,

appropriate modifications to the assumption may be made to represent more realistic material
behaviour as in the conventional stress analysis.
For some brittle materials such as concretes and glass fibre reinforced composites, the
development of plastic strains is limited and the damage is largely confined to the elastic strain
development. The necessary and sufficient condition for damage evolution under these conditions
is Fd = 0,eq. (4.6), and the incremental damage threshold B is dependent only on B.
For ductile materials exhibiting large development of plastic strains, damage due to elastic
deformation may be considered negligible. In such cases, Fd = 0 is only a necessary condition,
while damage is assumed to have been developed when both conditions of Fd = 0 and F, = 0 are
satisfied, or the incremental damage threshold B is a function of not only B but also j.
The inter-relationship between material damage and yielding discussed above may be
graphically illustrated as shown in Fig. l.The damaging process is assumed to initiate at point B
and continue to develop until point C as both yielding and damaging criteria are satisfied during
this process. Although 8d as well as B both increase from C to D, the material at that stage
undergoes elastic unloading and loading without inducing additional damage, i.e. dB/dB = 0 when
b = 0.

5. DAMAGE EVOLUTION MODEL FOR ISOTROPIC MATERIALS

One remaining parameter yet to be defined in eq. (4.7) is the damage characteristic tensor J.
Even if the tensor is assumed to be symmetric, the operator in its general form has 21 independent
components. Such a tensor will render the formulation of the constitutive equations too
complicated for general engineering applications for which simplifications become necessary.
From the basic assumptions of continuum mechanics, physical properties such as E and v of
a structural component made of homogeneous and isotropic material should be independent of
 Anisotropic theory of ~o~t~~uurn damage mechanics 5.53

the material element and the reference system chosen. However, material damage induces some
degree of nonhomogenity and anisotropy. This has been incorporated into the constitutive
equations of elasticity and plasticity by replacing the conventional stress tensor emwith the effective
stress tensor 8. The resulting effective stiffness matrices [q for elasticity and [a for plasticity
in the modified constitutive equations, unlike the conventional stress analysis, then vary with
element coordinate dependent upon the degree of material degradation, and also vary with the
element orientation due to the anisotropic nature of damage. In other words, the constitutive
equations involving the use of the matrix [c] or [I?j may be considered as a distributed set of
constitutive equations.
The damage evolution equations have also been derived based on the effective stress tensor
ii representing an effective damage state of the material element. Similarly, we may assume that
the effective damage characteristic tensor [J”] dependent upon the damage characteristic tensor
[JJ is, like [C] and [H], independent of the element position and coordinate system chosen, This
implies that [J] is everywhere the same and [J’] = [.I] for which [J’] is the damage characteristic
matrix under any coordinate system xi.
A J-formulation which satisfies the above requirements is proposed as

iPP 0 0 0
P lc1 0 0 0
M =2 (5.1)
0 0 0 2(1 - /It) 0 0
000 0 2U - Pu) 0
0 0 2(1 0- PI!
-000
PPl 0 0

for which the damage tensor D is expressed as

(5.2)

where ,U is a material constant. Noting that the damage operator [J] should be positive semi-
definite because

which requires that

I221r 22Qo
I
and

2 2P 34
[4(1 - P11” 2P 2 2~ >Oforn=O, 1,2,3
2P 2P 2

the value of JJ is confined within the range of

1
--Q&l.
2
554 C. L. CHOW and JUNE WANG

Equation (4.9) now becomes

(5.3)

for which [J”] is derived as

By combining eqs (2.4) and (5.1), [J”] is expressed in the principal reference system as
-
-1 -P -P 0 0 0
2
Wl WlW2 WlW3

P 1 P
---
W2Wl W2 2 0 0 0
w2w3
P P 1
-
- - 2 0 0 0
W3Wl W3W2 W3
VI = WITIJIIM (5.4)
0 0 0 4-P) o 0
u’232
0 0 0 0 2-P) o
WI2
0 0 0 0 0 w - P)
WI2 2
-

where W , = 1 - D,, W2 1 - D,,

= w3= I -D,, W
23 = & - D,)U - D,),
W3, = (1 - D3)(1 - 0,)and ~1,~= ,/(l - DJ(l - D2).
It is evident from the above that [j] = [J(D)] is element position dependent due to damage
induced nonhomogenity. Moreover, it can be readily verified that in an arbitrary coordinate system
of x;,

[PI = [M’]=[JI[M’] # [3] (5.5)

due to the damage induced material anisotropy as [M’] is different from [M] except for isotropic
damage.
The proposed damage evolution model [J] is consistent with the assumption made for the
development of the damage effect tensor M(D) that the principal axes of damage coincide with
that of stress[17] as 6,, b)5 and b, in eq.q.3) become zero in the principal system of stress.
Furthermore, it can be observed from eqs (5.1) (5.3) and (5.4) that the damage evolution
equation (5.3) can be reduced to a scalar equation for the proposed [J] model when the damage
response is isotropic represented by p = 1. This reduction is realized not only in the principal
coordinate system but also in any arbitrary system xi due to the fact that [J”‘] = [5”] for isotropic
damage.

6. DAMAGE MODEL VERIFICATION

As only two independent components appeared in the proposed [J], verification of the model
is rather straight forward.
Under uniaxial tension, the damage evolution equation of (5.3) becomes

(6.1)
 Anisotropic theory of continuum damage mechanics 555

-EACH LINE
DIMENSION=lmm

Fig. 2. Tensile specimen used for damage measurement.

or

dD, 1 - D,
dD, - P(1 - 01)’
(6.2)

Integrating the above yields

W - &I = D,U - W2)

(6.3)
’ = D,(2 - 0,) Ddl-042)

or from eq. (2.7)

(6.4)

The measurement of damage variables was performed using tensile specimens of aluminium
alloy 2024-T3 depicted in Fig. 2 and some of the results had been reported in lierature[17]. The
measured damage evolutions, D, and D, against true strain are reproduced and shown in Fig. 3.
With the knowledge of damage evolution, p can be readily determined using eq (6.3). Figure 4
depicts the variation of D,(l - DJ2) with D,(l - D,/2) revealing a linear relationship over 70%
of DJD,, and 85% of D2/Dzr. This implies that the proposed [Jj model is a quite good
approximation of the actual damage evolution. p determined from the linear portion is 0.58 for
aluminium alloy 2024 T3.
556 C. L. CHOW and JUNE WANG

12.00 -

X x
DI

+ ‘J2
IO.50 - x
x x
x

xxx x
9.00 -
xx
xx
xx
x
< 7.50-
0

x
t-i! 6.00-
x x
s
xx
2
x
1.50- +
x + +++ +
x * ++ +
++ +4+ + ++ +
x x + + +
rx +
3.00 - +++ +
+ +
+
x + +
Xx ++
1.50- x”+++

0; 1 I I I I I
0 5.67 II.33 17.00 22.67 28.33 31.00
STRAIN &, %

Fig. 3. Damage variables of D, and D, vs strain.

The increment of damage threshold B(P) can also be readily determined experimentally. From
eq. (4.7) we deduce, under simple tension

CY’d= = 6, = o,/(l - 0,).

Substituting il, = jI into (6.1), we have

b, = j?/(l - 0,)

from which p is finally derived after integration as

j?= D, -fD;.
On the other hand, B in eq. (4.6) may be deduced under simple tension

B = ed - B, = o, - o,,,,

The development of B(P)is illustrated in Fig. 5 displaying a linear dependence of B against p.

7. CONCLUSIONS

A generalized anisotropic theory of continuum damage mechanics for ductile fracture is

presented. This is achieved for the first instance by developing the constitution equations of plasticity
 Anisotropic theory of continuum damage mechanics 557

12.00-

IO.50-
x
x x
x
9.00 - x x
xx

XxXx
x x
;;-”7.50 x
.
N .&x y x
x

q 6.00- x

x
P x
OL
1.50-
x
x x
x x
x x
3.00-

x xx a=
D2(I-D2121
D] cl-D1 121
1so x
X
X

/
OX / 1 I 8 I i
0 0.75 I.50 2.25 3.00 3.75 A.50
D*(l-D2/2) I %

Fig. 4. ~te~inafion of ft for AI 2024-T3.

based on the damage effect tensor M(S) proposed recently by the authors for general coordinate
system. New damage evolution equ%ions are then derived in terms of the effective stress tensor
& providing a more realistic physical representation of material degradation. A general anisotropic
damage evolution model is proposed and satjsfacto~ly verified by ex~~mental determination of
damage variables. The proposed model which involves the constitution equations of elasticity,
plasticity and damage provides a comprehensive theory of anisotropic continuum damage
mechanics capable of solving practical en~neering problems.

REFERENCES
D. G. H. Latzko, Post-Yield Fracwe ~ec~~jcs. Applied Science ~bi~she~ London (1979).
A. A. We&, Application of fracture mechanics at and beyond general yielding. Br. Weid J. 563-570 (1963).
F. M. Burdekin and D. E. W. Stone, The crack opening displacement approach to fracture mechanics in yielding
mechanics. J. Struin Anal. 2, 145-153 (1966).
J. R. Rice, A path independent integral and approximate analysis of strain concentration by notches and cracks. J.
appt. Mech. 35, 379-389 (1968).
G. C. Sib and E. Made&, Fracture initiation under gross yielding: strain energy density criterion. Engtrg Fracture
Mech. 18, 667-677 (1983).
J. H. Giovanola and I. Finnie, A review of the use of the J-integral as a fracture parameter, SM. Arch 9, 187-225
(1984).
J. H. Giovenola and I. Finnie, The crack opening dis~la~ment (COD) as a fracture parameter and a comparative
assessment of COD and J-integral concepts. S.M. Archs 9, 227-257 (1984).
H. W. Liu and Tao Zhuang, A dual-parameter elastio-plastic Fracture criterion. In& J. Fracture 27, R87-91 (1985).
C. L. Chow and Xu Jilin, Ductile crack propagation with the strain energy density criterion. Engng Fracrwe Mech.
21, 537-545 (l985)_
C. I... Cfrow and Xu J&n, Application of the strain energy density criterion to ductile fracture. J. them. a&. Frmrure
Me& 3, 185-191 (1985).
C. L. Chow and Xu Jilin, Mixed mode ductile fracture using the strain energy density criterion. Int. J. Fracture 28,
17--28 (1985).
L. M. Kachanov, On the creep fracture time. Izv Akud. Nauk USSR Otd. T&h. 8, 2631 (1958).
F. A. Leckie and D. R. Hayhurst, Creep rupture of structures. Prctc. R. Sot. AJ40, 323-347 (1974).
558 C. L. CHOW and JUNE WANG

Fig. 5. Increment of damage threshold.

[14] J. Lemaitre, How to use damage mechanics. Nucl. Engng Design 80, 2333245 (1984).
[15] J. Lemaitre, A Continuous damage mechanics model for ductile fracture. J. Engng Muter. Tech&. 107, 83-89 (1985).
[16] D. R. Hayhurst, Creep rupture under multiaxial states of stress. J. Mech. Phys. Solids 20, 381-390 (1972).
[17] C. L. Chow and J. Wang, An Anisotropic Theory qf Elasricity>,for Continuum Damage Mechanics (to be published).
[18] 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).
[19] D. Krajcinovic and G. U. Fonseka, The continuous damage theory of brittle materials. J. appl. Mech. 48, 809824
(1981).
[20] F. A. Leckie and E. T. Onat, Tensorial nature of damage measuring internal variables. IUTAM Collequium on Physical
Nonlinearities in Structural Analysis, pp. 14G155 (1981).
1211 S. Murakami and N. Ohno, A continuum theory of creep and creep damage. IUTAM Symp. on Creep in Strucrures,
pp. 422444 (1981).
[22] E. Krempl, On the identification problem in materials deformation modelling. Euromech 147 on Damuge Mechanics,
Cachan. France (1981).
[23] J. P. Cordebois and F. Sidoroff, Anisotropic damage in elasticity and plasticity. J. M&c. Theor. Appl., Numero Special
4560 (1982) (in French).
[24] J. P. Cordeois, Crittres d’Instabilite Plastique et Endommagement ductile en Grandes Deformations, These de
Doctorat. Presentee a l’universitt Pierre et Marie Curie (1983) (in French).
[25] R. Hill, The Mathematical Theory of Plusticify. Clarendon Press, Oxford (1950).
[26] R. von Mises, Z.A.M.M. 8, 161 (1928).
[27] E. D. Case, The effect of microcracking upon the Poisson’s ratio for brittle materials. J. Mater. Sci. 19, 3702-3712
(1984).

(Received 30 June 1986)