Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Com~ffters& Stturrwes Vol.#. No. l/2. pp. 79-87, 1992 + 0.

00
~5-7~9/92 55.00
PrintedinGreatBritain. 0 1992PcrgamonPmssLtd

A DAMAGE COMPUTATIONAL METHOD FOR


COMPOSITE STRUCTURES

P. LADEV&E
Laboratoire de Mecanique et Technologie, ENS Cachan, CNRS, Universitb Paris 6,
61 avenue du President Wilson, 94235 Cachan Cedex, France

Abstract-The ~~ibiliti~ and computational difficusties of a damage ~mpu~tional method are


described in several examples. The composite material and especially the damage mechanisms are
described on an intermediate and preferential scale named the mesoscale. The challenge is to build up a
true rupture theory for composite structures.

1. I~RODU~ON and which depends on the relative direction of their


fibres (see Fig. 1).
Damage refers to the more and less gradual develop-
The characteristic length is the thickness of the
ment of microvoids and microcracks which lead to
elementary layer; to achieve mesomodelling one has
macrocracks and then to rupture. Brittle and
to add one more property [l]: the ‘meso’ damage state
progressive damage mechanisms are both present. In
is locally uniform within each elementary constituent.
composite structures it may be the main m~hanicai
Then, for laminates, the damage state is taken uni-
phenomenon. For such materials damage is generally
form in the thickness of the elementary ply. So a
of a highly complex nature; there is not one, but
mesomodel is not invariant by scale changing as in
several damage mechanisms. The mechanisms are
the classical models in continuum mechanics. A
highly anisotropic and present a strong unilateral
particular length has been introduced: the thickness
feature depending on whether the microdefects are
of the single layer. This point plays an important role
closed or open.
for the rupture computation, especially when one
Mesomodelling of these phenomena is proposed; it
tries to simulate cracks with a damage theory. Let us
is a damage mechanics approach where the composite
note that the property imposed to the damage state
is described on the intermediate and preferential scale
can be also prescribed to the other internal state
named the mesoscale. An important characteristic
variables but it is not essential.
is that such a modelling is not invariant by scale-
changing as in the classical models in continuum 2.2. General tools for damage rnadel~i~g
mechanics; damage constitutive relations are built up
The question is how to describe the damage state?
and when included in a structural analysis code, they
The idea we are following is due to Kachanov and
allow one to simulate the damage state at any point
Rabotnov [2,3]. In fact there are two ideas. First, one
of the structure until rupture.
adds a new variable to the usual state variables such
The purpose of the present study is to describe
as, plastic strain, and hardening variables, to quantify
the possibilities and the computational difficulties of
the damage state. The second, is to use the damage
this damage ~mputational approach on severai
indicator which is the relative variation of Young’s
examples concerning essentially laminate structures.
modulus. This scheme, i.e. the classical theory of
The challenge is to build up a true rupture theory for
isotropic damage, more particularly developed in
composite structures.
[4-g] for metallic materials and in [9, lo] for civil
engineering materials, is far from being sufficient in
2. MESOMODELLING OF A COMPOSITE MATERIAL
order to study composite materials. There is not one,
2.1. ~esos~ale -mgsomod~l~~ng but several damage mechanisms. The mechanisms are
highly anisotropic and present a strong unilateral
The first question is what scale are the damage
feature depending on whether the microdefects are
mechanisms described? For composite materials,
closed or open. Several approaches have been given
between the macro- and microscales, there exists an
for composites [ 11, 13,301. In the general theory [ 131
intermediate and preferential scale where damage
phenomena can be described in a rather simple
way. Let us consider the example of laminate com-
posite structures such as T300-914 composites. The
composite may be defined by two elementary con-
stituents: a single layer and an interface which is a
mechanical surface connecting two adjacent layers Fig. I+ Laminate modelling.

(IA.7 44,w.-o 79
80 P. LADEV~ZE

we proposed, these different aspects are included. We


d, d’ are two scalar internal variables which are
only present here the proposed models for laminate
constant within the thickness, they define the damage
composites.
of the single layer. The forces associated with the
mechanical dissipation are
3. MODELLING OF THE SINGLE LAYER

For simplicity, we consider a single layer with only


one reinforced direction. In what follows subscripts,
1,2 and 3 designate the fibre direction, the transverse
direction inside the layer and the normal direction,
respectively. y =’ <cfJ:,>>
1 yd’=
d 2 Ge(l - d)2 &St
3.1. Tension and compression behaviour in the jibre
direction
, a 1 (((a,,>: >>
Tension and compression behaviours are different; yd=dd’((E,))
to get the compression behaviour, we use a four-point o:cl,= i Et2 (1 - d’)2 ’
bending test described in [12]. The main result is that
the behaviour is purely elastic but non-linear. The $ is the free energy and ((. )) denotes the mean
compression modulus can be written value within the thickness. a’ is the effective stress
tensor; we will see its expression later.

Ei=Ey 1 -s(-cl,)+
I 1 , 3.3. Damage evolution law
For static loadings, we have
where a is a material constant and (.)+ the positive
part.
a’ I, = Ad (yd~,~yd’~, 7 Q t);

3.2. Damage kinematics


d’l,= A,.(Ydl,,Yd’l, z st),
Aside from brittle fractures in the fibre direction,
the matrix and the fibre-matrix interfaces inside the
layer deteriorate in a very particular manner; the where the operators Ad and A, are material
microcracks are parallel to the fibre direction. A characteristics.
homogenization calculation shows that the only Let us define:
moduli which are modified are the transverse modu-
lus E2 and the shear modulus G,2. The other indepen-
Yd=SUPIrdr Y, and YdP=supIrS,Yd’.
dent elastic characteristics E, and vu remain constant.
These properties are confirmed by experimental
observations. A simple law satisfied for carbon-epoxy materials is
The undamaged material strain energy, splitting defined by
up the energy into ‘tension’ energy and ‘compres-
sion’ energy is written in the case of plane stress
assumption d = &(Y,+ bYs),‘2 - Y;‘2)
e

if d<l;d=l otherwise

1
xo,,o*2+(022):+(-622):+o:2 d’=b’d if d’<l
--
E! E: GO12
and Y,. < Y:; d’ = 1 otherwise.
where cp is a material function defined such that
a2EDla2a,, = I/E:. YO. Y,, YL, b are damage characteristic constants of
The transverse rigidity in compression is supposed the material. Hence Yd and Y,, appear as the quan-
to be equal to Efj, one obtains the following energy tities which govern the damage increase and then the
for the damaged material transverse fracture of the single layer. Two different
damage mechanisms are introduced. The first one is
a progressive damage mode related to the micro-
cracking of the matrix and the deterioration of the
fibre-matrix interface. The second mode is a brittle
one; it concerns the deterioration of the fibre-matrix
. interfaces submitted to a transverse tension stress c22.
Damage computational method for composite structures 81

An important variant of the previous damage 140


evolution law is

k (Y, + b Y#)"2
- Y1'2(d)

( cijc
d =

yw
C >+
40 I* o TestC
if d<l; d=l otherwise
2or - Identified curve

with Y”2(d) = Y;‘2+ Yj”d 0’


0.2 0.4 0.6 0.6 1 1.2 1.4

P (%I
Fig. 2. Hardening curve for T300-914.

1 . 512
if d’ < 1; d’ = 1 otherwise, if2 = - -. t; 20.
2’R+R,,’

where n, n’, k, k’ are material parameters. This An example of such a hardening curve is given, for
modelling, which introduces delay effects, differs T300-914, in Fig. 2.
from the previous one only if the damage rates are
very high. It includes the following physical property: 4. INTERFACE MODELLINC

damage variations due to forces variations are not


4.1. Interface definition
instantaneous.
Remark. In order to simplify the general case The interface is a mechanical surface (limit of a
where out-of-plane stresses are present, Young’s small layer) which ensures stress and displacement
modulus E,, and shear modulus G,, and G,, remain transfers from one ply to another (see Fig. 3). It
constant and thus damage effects of out-of-plane depends on the relative orientations of the upper and
stresses are included in the interface behaviour only. lower plies. We suppose it is orthotropic. The N, , N2
axes are the bisectors of the angle formed by the fibre
3.4. Damage -plasticity (or viscoplasticity) coupling directions of the adjacent layers.
The kinematic variable is the displacement
The microcracking of the matrix and the deterio-
discontinuity denoted by
ration of fibre-matrix interfaces leads to sliding with
friction and then to inelastic strains. A way to model [VI = U+ - U- = [U],N, + [U],N, + [UN,
these phenomena is to use plasticity or viscoplasticity
mechanical models. The idea which seems to work which verifies the contact condition
quite well is to build the modelling upon quantities
which are called ‘effective’: effective stress B and [Ufl>O.
effective inelastic strain i, which is defined by
Tr[G 51 = Tr[u i,]. The energy of an undamaged interface is
The effective stress is a choice; it defines the
coupling between classical stress and damage state & = f [k”[W12+ k#A12 + k;[UJ2]
which is involved in inelastic strains. For laminates,
we introduce

k,, ky, k! are initial elastic characteristics.

4.2. Kinematics and evolution of the damage


9, = ep,(l - d); 9, = L$,(l -d’); S{, = c{, . The deterioration of the interface can be described
by three damage variables
From a homogenization calculation, and assuming *-
+<-%X
that only the matrix has an inelastic behaviour, we
kO(l - d)
have cf, = 0. The elastic domain is defined by

f = (ai, + a26:2)‘/2 - R - &.


42

+k;(l-d2)+k;(l-d,)
4,

1 *

The hardening is assumed to be isotropic, which


means that the threshold R is a function of the
cumulated strain p; p -, R@) is a characteristic func- inte
tion of the material. The yield conditions are written
as follows: Fig. 3. Interface modelling.
82

The forces associated with the dissipation are

y
d

yd2=k;(
=’ @33X
2k0(1 -C+

42

1 -d2)2’
Yd’=/cp(l -dj
4,
P. LADEV~ZE

=.,
--__
500
uL (MPa)

A simple modelling is to consider that the damage prediction

evolution is governed by
I I I I I
c
-0.04 -0.02 0 0.02 0.04
Fig. 5. Analysis of [- 12,781, laminate (T300-914).
where y, and y2 are coupling constants. In term of
delamination modes, the first term is associated with Then, we have
the first opening mode, the other two with the second
and third modes. The damage evolution law is
defined through a material function w, such that

d=u(Y) ifd<l; d = 1 otherwise

d, = y, o(Y) if dl c 1; d, = 1 otherwise

d2 = y2w(Y) if d2 < 1; d2 = 1 otherwise.


5. RUPTURE COMPUTATION
A variant with delay effect can also be introduced. An
Let us consider a composite structure with given
inelastic response can be taken into account by
displacements and loads on its boundary. All the
introducing an inelastic part of the displacement
data depend on the time t E [OZJ. The material is
discontinuities: [U, y; [U,r. Let us consider the
described with mesomodelling which is a kind of
following effective quantities
constitutive relation which depends on: the stress a;
[8,lp = [ir,J(l -d,); [02i,r = [ir,yl(l - d2) the strain L, the temperature 8; the inelastic strain L,,
the hardening variables and the damage variables
413 a23 *33 (meso). The problem is to compute the structure until
5, = -. cf2 = -. d,=-.
l-d,’ 1 -d,’ l-d rupture with such a modelling. A finite element
approach is used. Before presenting the possibilities
We build a plasticity-like model upon effective and the difficulties of this approach, several examples
quantities will be detailed.

f(a,,a,,a,,R)=J~-d,-RR(p),<O, 6. ANALYSIS OF LAMINATES SUBMITTED TO


TENSION LOADINGS
where a,, a2 are two material constants and R(p) the
hardening function Damage computations are done using plane stress
assumption for several cases [ * 4512,, [O],, [45], ,
P=
S[ t

0
'",I"'

a1
I KJ2YZ
2.
a2 1 dr
P, W,, , [67,5; 22, Aa, [ f 22,%,, 1f 1%. The cm-
parison with experimental tests is good except for the
last two tests (see, for example, Fig. 4).
uL (MPa) For the two last stacking sequences, the compari-
250 ----Test
son is satisfying for the curves, but not for the rupture
-Model
200 + Failure prediction P
t

-0.04 0.02 0 0.02 O.oI


Fig. 4. Analysis of [f45], laminate (T3W914), submitted
to tension. Fig. 6. DCB specimen.
Damage computational method for composite structures 83

a uIx104
Fig. 7. Initiation delamination problem and propagation problem.

values (see Fig. 5). The rupture experimental values grow does not correspond to the maximum load.
are lower than those given by our modelling. This fact Such a simplified delamination approach has been
is not surprising because the plane stress assumption applied to more complex situations [15]. Let us con-
is no longer correct. Such a damage delamination sider the structure defined by Fig. 8 where a tension
mode is taken into account due to the complete loading is applied. The edge effects are the only
damage modelling including the interface modelling. ones to be computed taking the interface damage
modelhng into account. The results are given in Figs
7. A SIMPLIFIED DELAMINATION ANALYSIS
9 and 10.
First there appears a crack in the central
The computation is done with the following interface and after a certain tension load, two new
hypotheses: the single layer is elastic and the interface cracks in the interface (+ 30, - 30). All these cracks
is also elastic but can be damaged. The material grow together until rupture. These first results have
function o(Y) is considered linear been obtained with a certain range of parameters.
Comparisons with experimental [16] results are in
agreement.
o(Y)=;,
c

where Y, is a material constant. Then, the interface 8. RUPTURE COMPUTATION: POSSIBILITIES AND
model depends on three material constants: the coup- COMPUTATIONAL DIFFICULTIES

ling coefficients I,, y2 and Y,. Let us consider a DCB The rupture phenomenon happens after two
specimen constituted with two layers as shown in phases. In the first step, the microvoid and micro-
Fig. 6. A pure mode I loading is imposed. crack growth is nearly uniform: it is the initiation
The delamination length is a and the thickness of stage. From the critical point (or from a point
the single layer is h. Under the plane strain assump- nearby) the strain and also the damages become more
tion, 2D calculations are carried out using the finite localized; a macrocrack appears and grows until
element displacement method. For a P h, Fig. 7 gives becoming unstable. If the first stage is well described
the computed critical value of the energy release rate due to this damage approach, the full simulation of
as a function of the delamination length u. The the rupture leads to severe difficulties. The goal is
critical value corresponds to d = 1 at the tip of the to build up a true rupture theory for composite
delamination front. It remains constant and the value structures.
which is G, = YJ2 can be also determined by a
theoretical analysis. Then, fracture mechanics
appears to be an approximation of damage mech-
anics for large delamination area. Starting with a = 0,
one obtains the load displacement curve given in
Fig. 7.
It is interesting to notice that the initiation point,
i.e. the point where the delamination front begins to

F F

F t
IL
Fig. 8. Delamination analysis of a tension test. Fig. 9. Delamination pattern.
84 P. LADE&E

nterface 30,-30: $
1.0
rltelface -30.90: *
ntetface 90.90: C F 0.8

0.6
3 : (30,-30) crack t
3’: (90,90) crack

I I 1 1 , ‘IO-* a,a’
0.2 0.4 0.6 0.8 1.0

Laminate : (30,-30,90) aym under tension


Force F versus a and a’
1
Fig. IO. Delamination analysis.

Figure 11 shows these difficulties considering a been obtained in [29]. The mesh dependence is very
damage macromodelling, but not a damage meso- important during the second stage of the rupture
modelling. It concerns a concrete structure and has process (after the critical point). The load-

i mesha 12

“;” I
,
mesh b
meshc 10

I a

I
i 4

0 1 2 3 4 6

mesh s meshb mesh c


Fig. 11. Rupture computation using a damage macromodelling.
Damage ~mpu~tional method for Composite structures 85

Fig. 12. Tension test for a [90], specimen.

-~~~~n
Fig. 13. Delamination analysis.

80 1.0 -
70
0.6 -
60
50
d 0.6 -
40
30 0.4 -
20
10
0.2-

0.05 0.10 0.15 0.20


Time Time
Fig. 14. Results for the loaded Gauss point of the interface.

displacement curves are very different for the three


meshes. It is also the case for the three totally damage
102J
areas which simulate the crack. This surprising result
exists also for other instabilities such as localized
necking.
Thus the problem is how to follow a post-critical
20 d--l
solution independently of the element size. An answer
is given by the localization limiter concept introduced
10 in [17,18]. It is a regtdarization procedure; the
additional terms are built up from a second gradient
approach or from a non-local approach. Delay terms
0.06 0.10 0.15 0.20 can also be. introduced. To go further in the under-
Time standing of these diBiculties,let us consider an exact
Fig. 15. Strain energy divided by 1 with respect to 2. solution.
86 P. LADE&ZE

Pure tension analysis of a composite plate using TAM Symposium an Inelastic Deformation of
Composites Materials, Troy (1990).
damage macromodelling
2. L. M. Kachanov, Time of the rupture process under
The fibres are supposed to be perpendicular to the creep conditions. Izv. Akad. Nauk. S.S:R. otd. Tech.
loading direction (see Fig. 12). From the critical Nauk. 8, 26-31 (1958).
3, Y. N. Rabotnov, Creep rupture. Proc. HI, Int. Gong.
point, numerous different solutions can occur. It is Appt., Meek. Stanford. Springer (1968).
a non-denumerable family parameterized by the 4. J.~~Lemaitre, How to use damage mechanics. Nuclear
thickness of the localization zone. Emma Des. 80. 233-245 f1984).
This example shows that the theoretical post-criti- 5. P. Giormini, b. Licht and P. Suquet, Growth of voids
in a ductile matrix: a review. Arch. Mech. 40, 43-80
cal solutions are much more numerous than in
(1988).
physical reality. Something homogeneous to a length 6, F. A. Leckie and E. T. Onate, Tensorial nature of
is missing; then classical models from continuum damage measuring internal variable. Proceedings
mechanics are not sufficient. IUTA M Symposium Physicai Non -linearities in
Damage mesomodelling introduces particular Structural Analysts. Springer (1980).
7. S. Murakamis, Notion of continuum damage mechanics
lengths and consequently these difficulties partially and its application to anisotropic creep damage theory.
disappear. For laminate composites, let us recall that J. Engng mat. Tech. 105, 99-iOS (19‘83). _ _
the interface is the limit of a small layer and then the 8. J. P. Cordebois and F. Sidoroff, Endommanement
model contains a characteristic length. For the single anisotrope en elasticit& et plasticite, Jo&at de
M~can~que Th~or~q~ et Applique 45-60 (1982).
layer modelling we have prescribed a constant
9. J. Mazars. Anolication de la mecaniaue de
damage state through the thickness and so the thick- l’endommagementau comportement non-liniaik et a la
ness of the elementary ply has been introduced as a rupture du beton de structure. These d’Etat, Universitt
characteristic length. To make sure, we now system- Paris 6 (1984).
atically use damage mesomodelling with delay effects 10. D. Krajcinovic, Continuum damage mechanics. Appl.
Mech. 37, l-6 (1984).
even to describe the brittle fracture of the tibre. 11, D. H. Allen, C. E. Harris and S. E. Groves, A thermo-
Several computations have been done. Let us mechanical constitutive theory for elastic composites
consider here the first one achieved in [19,20]. It is with distributed damage. Int. J. Solids Structures 23,
the delamination analysis of a laminate structure 1301-1338 (1987).
with an initially circular hole taking into account 12, E. Vittecoq, Comparison between compression and
tension behaviours of composite laminates. Thesis
all the damage mechanisms described previously (see (1990).
Fig. 13). The problem which is both non-linear and 13. P. Ladevize, Sur la mecanique de ~endommagement
three-dimensional is solved by a semi-analytical des composites. JNC 5, pp. 667-683. Pluralis, Paris
method. (1986).
14. S, S. Wang and I. Choi, Boundary layer effects in
Several results are given in Figs 14 and 15 for a composite laminates, I. Free edge stress singularities,
[0,90], laminate loaded in mode I. Normal displace- and II. Free edge stress solutions and basic character-
ments are prescribed on the edge of the hole (radius istics. J. Appl. Mech. 48, I, 541-548 and II, 5499560
to) under the following form: U(t) = i(t)U,, where (1982).
1s. L. Daudeville, Thesis (1991).
A(t) = t/T. Figure 14 concerns the loaded Gauss
16, R. Y. Kim and S. R. Soni, Experimental and analytical
point of the interface; it gives the evolution of the studies on the onset of delamination in laminated
peeling stress and the evolution of the damage vari- composites. J. Compos. Mat. 18, 70-80. (1984).
able d. Figure 15 shows the global energy divided by 17. Z. P. Bazant and G. Pijaudier-Cabot, Non-local
L, i.e an equivalent force with respect to A. damage: continuum model and localization instabiiity.
Report 87-2. Northwestern University, Evanston (1987).
18. ‘I. Belytschko and D. Lasry, Localization limiters
and numeri~l strategies for swain-softening materials.
9. CONCLUSION
In Cracking and Damage, pp. 349-362 (Edited by
J. Mazars and Z. P. Bazant): Elsevier (1988).
This damage computational method can be applied 19. 0. Allix. P. Ladeveze. D. Gilletta and R. Ohavon. A
damage prediction method for composite structuies. bt.
to other composite structures. The use of a damage J. Numer. Meth. Engng 27, 271-283 (1989).
mesomodelling avoids the main computational 20. 0. A&x, Delaminage par la m&canique de
difficulties and involves an finer modelling of physical ~Endommagement. Calcul des Structures et intelligence
phenomena. Nevertheless, this kind of modefling Arttjicieite, (Edited by J. M. Fouet, and R. Ladeveze and
R. Ohayon), Vol. 1, p. 39-56. Pluralis, Paris (1989).
needs a multiscale computational approach. To be
21. J. P. Dumont, P. Ladeveze, M. Poss and Y. Remond,
complete, the sensitivity to imperfections, to large Damage mechanics for 3D composites. Compos.
defects has to be studied. Further research is of Structure 8, 119-141 (1987).
course necessary to completely solve the compu- 22. Z. Hashin, Failure criteria for unidirectional fiber com-
tational and theoretical problems in order to achieve posites. J. Appl. Mech. 47, 329-334 (1980).
23. D. Gilletta, H. Girard and P. Ladeveze, Composites 2D
a true rupture theory for composite structures. a fibres a haute resistance: modelisation m&anique de
la couche blementaire. JNC 5, DD. 685497. Pluralis.
REFERENCES Paris (1986).
24. P. Ladeveze, Sur une theorie de l’endommagement
1. P. Ladeveze, 0. Allix and L. Daudeville, Mesomodelling anisotrope. Rapport Interne 34, Laboratoire de
of damage for laminate composites. Proceedings IU- Mecanique et Technologie, Cachan (1983).
Damage computational method for composite structures 87

25. E. Le Dantec, Contribution a la modelisation du com- 29. C. Saouridis, Identification et numerisation objectifves
portement mecanique des composites stratifies. These de des comportements adoucissants: une approche multi-
l’UniversitC Paris 5, Cachan (1989). tchelle de l’endommagement du b&on. These de
26. J. Mazars, and G. Pijaudier-Cabot, Continuum damage 1’Universite Paris 6, Cachan (1988).
theory: Application to concrete, Internal Report 71 30. R. Talreja, Transverse cracking and stiffness reduction
LMT Cachan (1986). in composite laminates. J. Compos. Muter. 19, 355-375
27. A. Needleman, Material rate dependence and mesh (1985).
sensitivity in localization problems. Camp. Merh. Appl. 31. S. S. Wang, Fracture mechanics for delamination prob-
Mech. Engng 67, 68-85 (1987). lems in composite laminates. J. Compos. Mufer. 17,
28. K. Reiffnider, Stiffness reduction mechanism in com- 210-213 (1983).
posite materials. ASTM-STP 775 (1980).

You might also like