Professional Documents
Culture Documents
NORTH-HOLLAND
IN APPLIED
MECHANICS
AND ENGINEERING
51(1985) 31-49
COUPLED ELASTO-PLASTICI~
AND DAMAGE
CONSTITUTIVE EQUATIONS
Jean LEMAITRE
~~or~toir~ & ~~~an~~ et Techn~i5g~e, ENSET- Univ. Paris 6-CNRS, 94230 Cachan, France
Received 26 November
1984
The basic equations to mode1 the coupling between strain and damage behaviors are written within
the framework of the thermodynamics of irreversible processes. The damage is represented by a scalar
internal variable which expresses the loss of strength of materials during such processes as fatigue,
ductile or creep strains. Applications are given in elasticity coupled with damage for brittle failure of
concrete or high-cycle fatigue of metals, in plasticity coupled with damage for ductile fracture or
low-cycle fatigue, and in visco-plasticity coupled with creep and fatigue damages.
WMAGE
CRACK
~UTI~
LAWS
REAGAN
LAWS
STRESS AND
STRAIN FIELDS
equations.
32
(2) In a second step, these results are used to calculate the condition of a macro-crack
initiation with the help of a damage constitutive equation [2], for example, the number of
cycles to fatigue failure due to a loading variable with time.
(3) Finally, the growth and the propagation of the crack through the structure is analysed
by means of fracture mechanics, using concepts of stress intensity factors or strain energy
release rate [3].
This procedure implicitly assumes that the state of damage of the structure does not
influence the state of stress or strain, that is there is no coupling between constitutive
equations of strain and damage. This assumption was not too bad as far as the structure
calculations were not too accurate. With the progress of constitutive equations, taking into
account anisotropic strain hardening in plasticity or visco-plasticity, and the progress of
numerical calculations in relation with faster computers, this non-coupling hypothesis must be
released in order to have an accuracy of the same order of magnitude in the representation of
physics and in the numerical procedure.
From the physical point of view, this coupling is due to the nature of damage. The brittle or
fatigue or ductile damages are mainly the development of micro-cracks and micro-cavities;
that is, creation of surfaces of separation in the materials which reduce the rigidity of the
solids. We observe the following:
(1) The elasticity modulus decreases as the damage progresses.
(2) The plastic strain-hardening decreases also and may become negative just before failure.
(3) It is to be recognized now that the increase of plastic strain rate in tertiary creep is
mainly due to damage by intercristalline micro-cracking.
Roughly speaking, using uncoupled constitutive equations gives rise to small errors on
stresses or strains if the loading is far from the rupture conditions, but the errors may be of the
order of 10 to 50% close to the failure. This means that to predict the rupture of the structures
with a good accuracy, it is necessary to use coupled constitutive equations. Then, the
procedure of prediction is reduced to only one step in which stress, strain and damage fields
histories are obtained in the same time as shown by the scheme in Fig. 2. A macro-crack is
then considered as the set of points for which the local damage has reached its critical value at
failure.
Fig. 2. Prediction
equations.
33
2.1. Thermodynamic
variables
In order to model elasticity, plasticity and damage in view of structure calculations, the
following variables are introduced:
-the total strain tensor E,
-the elastic strain tensor Ed,
-the plastic strain tensor ~~ = F - Ed,
- anisotropic
strain hardening is modelled by the kinematic tensorial variable X which
represents the translation of the yield surface in the stress space, associated with the isotropic
hardening scalar variable R which represents the increase of the radius of that yield surface.
Their corresponding state variables are respectively (Y and p. Associated with X and R, they
define
the power dissipated by micro-internal stresses in the body (X : ci) and the power
dissipated to increase the yield stress (RJ~). (The symbol : denotes the contracted tensorial
double product.)
If the phenomenon of damage is regarded as isotropic, it can be defined by a scalar variable
D which represents the surface density of intersection of micro-cracks and micro-cavities with
any plane in the body. If SS is the area of the section of a finite volume element, (1 - D)SS is
the area which effectively resists to the stress. This proves that
(virgin element + ) 0 d D S 1 (+-broken
element) .
(2.1)
In fact, the volume element breaks by atomic decohesion before that value of 1; the criterion
for macro-crack initiation is D = DC (DC being of the order of 0.2 to 0.8 depending upon
materials). Let Y be the variable associated with D as Yfi is the power dissipated in the
creation of surfaces of separation.
If thermal effects are considered, the variable temperature T is introduced together with its
associated variable entropy s.
All these variables are summarised in Table 1.
Table 1
~e~odyn~ic
variables
State variables
Observable
Internal
Associated
variables
(T
Ee
EP
P
D
o--Ei
X
R
Y
34
_I. Lemaitre,
(2.2)
A very useful concept is that of an effective stress, that is, the stress relative to the effective
area [4],
cr
(2.3)
a=l-D
Associated to the hypothesis of strain equivalence [5], it gives the key to write an analytical
expression for 4. That hypothesis states that any strain constitutive equation of a damaged
body derives from the same function as for a virgin material, except that the stress is replaced
by the effective stress.
For example, the one-dimensional elastic strain energy of a damaged body is we = i&s, =
~E.E~(I - D), from which follows law of elasticity, namely,
(2.4)
where E is Youngs modulus.
Now, the thermodynamic
potential for isothermal
plasticity, not coupled with elasticity is
1
ICi=-a:r:ee(l-D)+tClp(tY,P),
2P
(2.5)
where a is the fourth-order tensor of elastic coefficients, functions of temperature and p the
density.
The coupling between elasticity and damage is introduced through the factor (1 - D) in the
first term of (2.5).
The law of elasticity is derived from (2.5) by
cr=p$=a:FC(l-D).
The thermodynamic
P-6)
35
is calculated
within the
tTeq
[5(1+ v) + 3(1- 2v) CZJ
2E(l - 0)
-y=
= 2E(;aD)*
R ,
(2.8)
where
v is Poissons ratio,
uH is the hydrostatic stress, uH = 4tr(a),
a,, is the Von Mises equivalent stress, ueq = (8~ : u*)~,
CT* being the stress deviator, CF*= CT- ~~1, and
R, = $(l + v)+ 3(1- 2~)(a~/cr,,)~ for short.
This quantity may be used as a damage criterion by the definition of an equivalent onedimensional stress ET*,giving the same value of Y as the equivalent one-dimensional case for
which R, = 1 (in (2.8)),
*2
-y=
(2.9)
(T
2E(l-
0)
or
u = u,,Rjf2.
(2.10)
This stress damage criterion may be used for damage theories as the Von Mises equivalent
stress is used for plasticity theories.
The dissipation, which must be positive in order for the second principle of thermodynamics
to be satisfied, is written through the Clausius-Duhem inequality as follows:
PdT,O,
a:iP-X:CC-R@-Yl%q.F.
(2.11)
.Gp, h, e, D are the mechanical dissipative flux variables; it is assumed that their laws of
evolution derive from a potential of dissipation, a convex function of all dual variables a, X,
R, Y, 0, grad T/T (the state variables acting as parameters),
(b(u,
X, R,
Y, y;
w*
au
a+*
P:_-g
(2.12)
NJ*
ff=-dX*
a=____a+*
8Y *
(2.13)
36
For phenomena which do not depend explicitly upon time, such as plasticity, $* becomes the
indicatrice function of the convex and the plasticity constitutive equations are written by
means of a plasticity multiplicator A calculated through a condition of irreversibility or
consistency,
(2.14)
The coupling between plastic strain and damage appears to be D and Y in the potential +*.
These constitutive equations have to be identified from experiments for each material and for
each kind of damage.
2.4. Identificationof damage constitutiveequation
The dependence of #* on a, X, R is found through one-dimensional strain hardening cyclic
tests for plasticity and also from creep or relaxation tests for visco-plasticity. The dependence
of #* on Y or D needs measurement of the damage variable. A way to do it is to evaluate D
through its influence on the elasticity modulus.
From the one-dimensional law of elasticity (2.4), calling (1 - D)E = E the damaged elasticity modulus, it is easy to obtain
D=l-g-.
(2.15)
Then, careful measurements of the Young modulus E on the virgin material and the elasticity
modulus I? on the damaged material yield D [7]. An example of the evolution of the damage
during a low cycle creep fatigue test is given in Fig. 3.
37
Knowing the evolution of damage in particular cases of brittle damage, fatigue damage,
ductile damage, creep damage, many empirical models have been proposed. A way to have a
synthetic view on these models is to use the potential of dissipation written as a sum of two
functions as follows:
f$* = qfqq x, R; T, D)+ 4;(Y;
T, D).
Most of these models may be obtained, at least for their main properties,
function of Y,
4; = so
so+
(qso+l(I
So
(2.16)
taking 4; as a power
_ 0)) )
(2.17)
(-yso>s(l_D)_l
(2.18)
fi=_K=_a@:=
aY
aY
or
a=-
(2.19)
for phenomena which depend explicitly, or not, upon time. So and so are the only two material
constants, temperature dependent, which must be identified for each type of damage and each
material. If D = D, failure of the volume element occurs.
38
tr(a2)
--
M41
.~
1-D
(3.1)
>
1-D
a**
l-Iv
Ee=P-=----dCr
1-D
y=p-=
av
aD
v tr(a) l
El-D
2
oeq
2E(l-
D)
r=p=.
R,
(3.2a)
w*
(3.2b)
The constitutive equation for elastic strain is coupled with damage through the factor (1 - D)
with the constitutive equation (2.19) as high-cycle fatigue does not depend explicitly, upon
time.
The different damage behaviors are taken into account by different expressions for h
related to i.
3.2. Application to high-cycle fatigue of metals
The high-cycle fatigue of metals is difficult to model by means of continuum mechanics
because at an early stage, a single crack develops which is not exactly in agreement with the
hypothesis of damage being equally distributed in the volume element!
Nevertheless, many continuous damage models have been proposed [8]. One of the
simplest, valid for three-dimensional
fatigue, is the following:
A possible constitutive equation for micro-plasticity is
(3.3)
- ,;)]2
(3.4)
where 6: are the components of an internal stress, which may be identified with the mean
values of the components of the stress deviator over a time period of one cycle. a: Max and
u: min being the maximum and the minimum values of the stress deviator during the cycle,
6y = $(u? Max - a: min) ;
(3.5)
+c)(+.$
g$.
(3.6)
39
This constitutive equation for high-cycle fatigue has the following properties:
(i) it is valid for three-dimensional
loading;
(ii) nonlinear evolution of damage;
(iii) nonlinear relation between number of cycles to failure and stress;
(iv) effect of hydrostatic stress given by (T*;
(v) linear cumulation if /? = const., but nonlinear cumulation if p is taken as a function of
(T* : p(cr*);
(vi) effect of mean stress given by a:, but equal damage for positive (tension) or negative stress
(compression). To take into account a different behavior in these two cases a concept of effective
stress in quasi unilateral conditions must be introduced [9].
The three coefficients to be identified for each material are B, so and p but most often
so = 1. This can be done from the classical Woehler curve which expresses the number of
cycles to failure as a function of the maximum stress, or its amplitude, from the results of
tension-compression
fatigue tests. Taking the case of mean stress equal to zero, so = 1, (3.6) for
a one-dimensional loading reduces to
(3.7)
Integrating over one cycle N defined by the maximum value (gM) and the minimum value
(a,,, = -uM) of the stress, and neglecting the variation of (1 - D)p+l during the cycle gives
SD
-=
SN
uM
4BaK
I o (/I + l)(l - D)p+ dD(a) .
(3.8)
+ D=O,
N=N,
D=D,-1.
(3.9)
We then obtain
N
-___ 1 Cr$+l)
/3+2
4B
= p +
(3.10)
Equation (3.10) is the equation of the Woehler curve given by the model when the mean stress
ti is zero. Then, (p + 1) may be calculated from the slope of the corresponding experimental
curve in a log-log plot. The nonlinearity of the phenomenon corresponds to values of p of the
order of 2 to 10.
The final constitutive equations for elasticity coupled with high-cycle fatigue damage are
(3.2a) and (3.6), with u* and Au,, given by (2.10) and (3.4), resp., and geeqand R, as defined in
Section 2.2.
As an example, Fig. 4 shows the evolution of the damage zone of a tulip specimen using
these constitutive equations implemented in a finite element code [lo] using a step-by-step
40
II
G
._
2
._
.z
.E
41
procedure. The difference with classical FEM calculations is that the rigidity matrix varies as it
is a function of the damage elasticity modulus E = E(1 - 0). The external loading is a periodic
0, + force and the state of stress is axisymmetrical.
3.3. Application to the behavior of concrete
From a phenomenological
point of view, it is better to write the damage as a function of
strain rather than stress. Then, Y and ti may be expressed as a function of strain through the
stress and the elasticity law.
Furthermore,
to take into account the large difference of strength in tension and compression, an equivalent strain E * is introduced. It contains the unilateral conditions of opening
and closure of micro-cracks:
E * =
where Ed, Ed, &3are the principal strains and the symbol ( ) denotes the positive part,
X
w= ( 0
ifX>O,
(3.11)
ifXa0.
3E FISSURE
LOADING
Q (kN)
NOTCHED
SPECIMEN
in a concrete
notch
specimen
loaded
in mode
1.
42
I
l_
D=
Q&-4+
&*
exp(G
* - EDI>
if e*>.zD,
(3.12)
if .5*<ED,
where EP (a damage threshold), a and b are material coefhcients which may be identified from
tension~ompression
tests. (D = DC means rupture of the volume element.)
This model is restricted to monotonic loading, i* >O, but it is valid for any threedimensional case.
In order to predict the failure of concrete structures it must be used together with the
elasticity law (3.2a). Using the finite element method a step-by-step procedure on displacement
is used.
An example is given in Fig. 5: it is a notch plate specimen for which the calculated damage
zone (D = DC) is compared with the path of the experimental crack [ll]. For that example, the
material coefficients are E = 3~~ MPa, Y = 0.2, ED= 10V4,a = 0.8, b = 2 - 104.
43
or
39f0,
f<O3
(4.1)
ip=O;
~*=f(o;R;D)+~;E(Y,d;D),
(4.2)
(4.3)
o;q-R
----ay=O,
1-D
(4.4)
Lj_pP3=
aY
(-y >soi
so
1-D'
44
.I. Lemaitre,
Coupled eiasto-plasticity
and damage
The multiplicator h is derived from the consistency condition f = 0 which ensures that the
actual yield state of stress follows the loading stress. Furthermore, we must have
h=O
now,
iff<Oorf<O;
af
af + af .
f=a,:ti+z~+aD~
or
R(p) . ez,-R
.
p+(1_D)2D=O?
(4.6)
or, with
-p-l-D
/i =
/i
-R
and
T_D)=~Y,
k&
H(f)
- D>
(4.7)
R(p) - CTY(-Y/S,,)o
iff<O,
H(f)={l
iff
(4.8)
=O;
R(p) characterises
strain hardening;
the nonlinear
strain hardening curve,
R=$$
h(p) = KptJM,
of the
(4.9)
where K and M are material coefficients which may be identified from tension strain
hardening test results. In the range of eP where the damage is zero or negligible, the equation
of the strain hardening curve is
(4.10)
a=ayfK&y,
which allows to determine
(TVand M from the slope of a plot of log(a - fly) against log sB.
fi=
(so>
-
1_D=
_yso
(3
&
ri;
(4.11)
45
(2.81,
so
~k~Rv
.
>
D)z P.
(2E&(l-
Ij=
(4.12)
Usually ductile damage occurs for strain large enough to consider the material as perfectly
plastic ~strain-h~dening saturated);, then we may consider
a-l
-=
(4.13)
const. = G,
1-D
&s&ED+
&=ER
ED and a rupture
D=Dc;
-+
then,
2
I)=?-
2ESo
cp + const.
or
D=D,c,
ER-CD
2ESo
DC
ER
and
ED
DC
fi= -----R,b.
ER-
equations
(4.16)
ED
for isotropic
elasto-plasticity
coupled
with ductile
(4.17)
46
&e_l+v
E
6
---___
1-D
ip = $H(f)
tr(a)
El-&
Y
uD
kLl>
K
(l-W/M
D;R,
(4.18)
*;
uy
EP
&R:
ED
DC
Lj,
-R@,
&R- ED
with
p =
(S&P
PC&D
: &P)1/2
--)
D = D, + crack initiation .
D=O,
This set of constitutive equations may be used to predict the limits in metal forming (D + DC)
[13] or to predict the influence of a forming process on the behavior of materials by the final
values of cP or p or R and D [14].
The identification of the material constants is easy to perform.
Example of steel 35NCD16 at room temperature:
E = 210,000 MPa ,
Y=
0.3 )
oy =
1200 MPa ,
K =
3340 MPa ,
M = 3.1,
&,, =
0.02,
&R =
0.37,
D, = 0.24.
d&
d = C (1 _
D>2
(a,, -
uy(l
- D))-lb,,
change of
(4.20)
Identification of coefficients C and M come from Woehler curve as explained in Section 3.2,
and a calculation of cyclic plasticity, coupled with low-cycle fatigue damage, uses the
elasto-plastic constitutive equations of Section 4.2 associated with the last one.
41
Visco-plasticity in metals occurs when the temperature is above approximately one third of
the absolute melting temperature. Two kinds of damage may occur: the creep damage and a
low-cycle fatigue damage if the loading is cyclic.
The coupled constitutive equations may be derived from the same formalism as previously.
The potential of dissipation is written as the sum of two terms, one for visco-plastic strain,
one for creep fatigue damage, both phenomena being time-dependent,
then
N p-NJM
1-D
((l_u;)K)N~,
(5.2)
-
D)- ,
(5.3)
This damage model generalises the one-dimensional Kachanovs law proposed in 1958 [4].
The final constitutive equations for isotropic visco-plasticity coupled with creep and fatigue
damages are (5.3) and
(5.4)
with u* and e as in (2.10) and (4.5), resp., and (TH,uD, ueq and R, as defined in Section 2. K, A4, N
are visco-plastic temperature-dependent
material coefficients which can be identified from
one-dimensional creep tests.
A and r, creep-fatigue damage coefficients in (5.3), are identified from one-dimensional
creep fatigue tests by means of Woehler (or Manson-Coffin curves) corresponding to different
frequencies.
48
1x1
DAMAGE
VERSUs
TIME
TIME : WOURS
DAMAGE: 1x10-
ROTbTlON
Fig. 6. Calculation
T = 550 C.
Equations (5.3) and (5.4) have been used for the calculation of the time for a crack initiation
in a model of a turbine disc [15]. Fig. 6 shows the result for the evolution of the creep damage
in the critical zone due to the centrifugal force. [16] and 1171 contains some other examples.
6. Conclusion
The sets of constitutive equations for elasticity, plasticity and visco-plasticity coupled with
brittle, fatigue, ductile or creep damage are given ready for use in step-by-step incremental
methods of structure calculations. Some problems of convergence may arise, but not more
difficult as those for classical plasticity.
The difficulty which remains is to obtain the material-dependent
coefficients for each
application. A handbook of such coefficients does not yet exist, and this situation may last
another ten years or more! If you want to make numerical calculations with the models
described, you have to identify the coefficients, by yourself, from tests results which may be
found in the literature for most of them. It is not difficult, the methods are given in this paper,
it is a good exercise, it gives a better understanding of the domain of applicability of the
constitutive equations, and it helps to make a good criticism of the validity of the numerical
results obtained!
49
References
[l] D.R.H. Owen and E. Hinton, Finite Elements in Plasticity (Pineridge, Swansea, 1980).
[2] J. Lemaitre
and J.L. Chaboche,
Aspects phenomenologiques
de la rupture par endommagement,
J. Met.
Appl. (3) (1978).
[3] G. Sih, Mechanics of Fracture, Vols. I-V (Noordhoff,
Leiden, 1973-1977).
[4] L.M. Kachanov, Time to the rupture process under creep conditions, Izv. Akad. Nauk SSR Otdel. Tehn. Nauk
8 (1958).
[5] J. Lemaitre, Evaluation of dissipation and damage in metals submitted to dynamic loading, Proc. ICMl, Kyoto,
Japan, 1971.
[6] J. Lemaitre
and D. Baptiste, On damage criteria, Workshop
NSF on Mechanics
of Damage and Fracture,
Atlanta, 1982.
[7] J. Dufailly, Modelisation
mecanique
et identification
de Iendommagement
plastique des materiaux,
These,
Universite
Paris VI, 1980.
[8] J.L. Chaboche,
Une loi differentielle
dendommagement
de fatigue avec cumulation
non lineaire,
Rev.
Francaise M&Z. 50-51 (1974).
[9] P. Ladeveze
and J. Lemaitre,
Damage effective stress in quasi unilateral
conditions,
IUTAM
Congress,
Lyngby, Denmark,
1984.
[lo] A. Benallal, R. Billardon and J. Lemaitre,
Failure analysis of structures
by continuum
damage mechanics,
Proc. ICF6, New Delhi, 1984.
[ll] J. Mazars,
Application
of continuous
damage
mechanics
to strain and fracture
behavior
of concrete;
Application
of fracture mechanics to cementitious
composites,
NATO ARW, Northwestern
Univ., 1984.
[12] J. Lemaitre,
A continuous
damage mechanics model for ductile fracture, J. Engrg. Materials Techn., Trans.
ASME 107 (1985).
[13] J.P. Cordebois,
Criteres dinstabilite
plastique et endommagement
ductile en grandes deformations,
These,
Universitt
Paris VI, 1983.
[14] B. Gattoufi, Effets de la prtdeformation
due au filage sur le comportement
des metaux, These, Universite
Paris VI, 1984.
[15] A. Benallal,
Finite element
analysis of the behavior
and failure of elasto-visco-plastic
structures,
Proc.
NUMETA,
Swansea, England, 1985.
[16] D.R. Hayhurst, P.R. Dimmer and C.H. Morrison, Development
of continuum damage in the creep rupture of
notched bars, Rept., Univ. Leicester, U.K., 1983.
[17] H. Riedel, A continuum damage approach to creep crack growth fundamentals
of deformation
and fracture,
Proc. Eshelby Memorial Symposium,
Sheffield, 1984.