A Damage Mechanics for Composites Materials

Pierre LADEVEZE

Laboratoire de Mecanique et Technologie

E.N.S. de Cachan/C.N.R.S./Universite Paris 6

61, avenue du President Wilson - 94235 CACHAN CEDEX

1. INTRODUCTION
The fIrst concept is damage. The idea is due to Kachanov and Rabomov [8], [19] :

the deterioration of a material can be described by its effects on the elastic characteristics.

The classical theory of isotropic damage, more particularly developed by Lemaitre and al.

[14] for metallic materials and in [9] [16] for concrete, is not sufficient to study

composite materials. There are not a single one but several damage mechanisms. The

mechanisms are high! y anisotropic and present a strong unilateral feature depending on

whether the micro-defects are closed or open. These different aspects are included in the

general approach we proposed in [10] : Applications to composites are given in [5] [6]

[11] [12] [13].

In addition to the damage theory, the homogenization plays a prominent part as a basic

tool. For example in the case of laminates, on the single-layer level it allows to transfer

information on the nature of the damage, obtained from a more local level. On the

stIucturallevel it allows to construct the laminate behavior, knowing the layer behavior.

Before deriving a damage modelling, it is essential to specify the chosen scale. For

composite materials, berween the macro and micro scales, there exists an intermediate and

preferential scale where damage phenomena can be described in a simple way.

Moreover, one has obviously to take into account the specificities of the studied material.
What is common, is the method with its experimental methodology and its general tools.

This approach is detailed for rwo different materials, the three-dimensional composites

(ex. : 3D Carbon-Carbon) and the laminate composites (ex. : T300-914). A macro-

13
1. F. Dijksman and F. T. M. Nieuwstadt (eds.), Integration a/Theory and Applications in Applied Mechanics. 13--24.
© 1990 Kluwer Academic Publishers.
14

modelling is given for 3D materials and a meso-modelling for laminate composites. The

last part deals with the numerical simulation of such a model, i.e. the computation of

damages. The challenge is to build up a true rupture theory which includes, as a

simplified approach, the classical linear fracture mechanics.

2. Some general tools for the damage modelling

2.1. Damage concept
The idea we have followed is due to Kachanov and Rabomov [8] [19]. First, one

adds a new internal variable in the thermodynamics sense to describe the damage state.

The chosen damage indicator is the Young's modulus of the material.

Undamaged states: Eo modulus Damaged states: E modulus

Figure 1 : E as a damage indicator

It is clear that the relative variation of the modulus (Eo - E)lEo is characteristic of the

damage level. For anisotropic materials, and then for composites, we have developed

this idea in [10]. The main features of the theory are presented hereafter. Let us consider

a damaged material whose initial characteristics are denoted by '0' (No damage). The
effective volume which is studied, is submitted to a stress perturbation 0', in the space

direction rt which does not lead to additional anelastic strains. The strain perturbations

associated with 0' and rt are :

-e;.(rt) : longitudinal strain in the rt direction,
-e;.(rt) + e;(rt) + e;,(rt): volumic strain in the rt direction.
T, T' denote two transversal and orthogonal directions with respect to rt, and onhogonal

each other. In order to simplify what follows, we assume identical traction and

compression behaviors; we defme :

~ 0'
E(n)=--
*~
~(n )
 15

where ECrt) is Young's modulus in the rt direction and y(it) is a less classical
modulus, still in the rt direction. For an isotropic undamaged material, we have:

E(it) =Eo y(r{) =~

1-2'\)0

The fundamental result is the following theorem; the functions:

are independent and define entirely the Hooke's tensor of the material. It is to notice that

they are not arbitrary functions: E depends on fifteen scalar coefficients and yon six.

The damage functions are defined by :

Ii(r{) YO(rt) - y( rt) d(r{) =Eo(rt) - E(rt)
)t)(rt) Eo(rt)

The d and Ii functions may be represented in the three-dimensional space by two surfaces

Sd, So called damage surfaces:

<.N: = rt. d(rt)

For undamaged media these surfaces are reduced to a single point. These surfaces admit

the origin as a center of symmetry; their interest is to allow the visualization of the

damage state of the material. The central problem is, of course, to describe the damage

state of the material with the minimum number of variables. If, for example, the damage

surfaces may be approximated by means of two spheres, one obtains, for the damage, a

two variables description which constitutes the real isotropic damage theory. More
generally, the d and Ii functions may be expanded in Fourier series. Thus, one obtains,

depending on the choice of the truncation, relatively different damage kinematics.

2.2. MODELliNG of the Damage Evolution. Coupling Damage-Plasticity (or

viscoplasticity)

Let us consider a damage kinematics defined by the set of scalar damage variables:
d ; 0
The free energy P'I'is a function of :
P'I'(E e , d, 0, X)
16

where X denotes the hardening variables or any other. The conjugate quantities to dl, 0,

i.e the quantities which govern the damage evolution and then the rupture are :
i1\j1
Yd=·Pod (f,X: cst
i1\j1
Yo=-p (1) (f,X: cst

Where i:t denotes the chosen effective stress. For many cases, one has also:
OED
Yd = ad (J: cst (ED : strain energy)

Yo= as
oED
(J: cst

The micro-defects lead to sliding with friction and then to anelastic strains. A way to

model these phenomena is to use plasticity or viscopiasticity mechanical modelling. The

idea which seems to work quite well is to build the modelling upon quantities which are

called "effective" :
- effective stress a
- effective anelastic strain rate Ip
which verify: Tr [(J £p] = Tr [a lp]
A particular choice is
a =KQK-l (J Ip = K·l Ko I (K: Hooke tensor)

Remark 1 : In order to measure the damages, it is strictly necessary to use loading and

unloading. Figure 2 illustrates the type of test which must be performed.

cr

Figure 2 : Tensile test with unloadings

 17

Remark 2 : The micro-voids and micro-cracks may open or shut depending on the case.

Various possible modelling are given in [11] [12] [13]. The idea is to express this

unilateral character in te=s of energy.

23. Homogenization. Local Stress and Strain distributions.

It is clear that in a composite, many levels may be easily distinguished. Between the

electronic microscopic levels, and the structure macroscopic level, at least there exists an

intermediate level that we call the meso-level. For 2-D' s, each layer is schematized as

homogeneous. The modelling of the single-layer itself results from a homogenization at a

smaller scale, the scale of the fiber diameter.

Homogenization techniques have become classical, that is why we shall not emphasize

them.

3. A tirst example : Damage macro-modelling for 3D Composites

These materials (figure 3) can be defined at the meso-scale by the following constituents :

- fibre-yarns

- matrix-blocks

- interfaces

figure 3 : Geometry scheme for a 3D material

3.1. Damage kinematics - Macro-modelling

One supposes that only the shear moduli vary, the other elastic characteristics

remaining constant Moreover, one introduces only one scalar damage variable such that:
012 = (1 - d) G~2

023 = (1 - g) G~3

G31 = (1 - g) 0 3°1

The elastic energy is then:

[8

This expression agrees with the experimental observations. It can be also derived by

homogenization technics. For that model, the damage functions are :

oCit) =0 =
(X2 ~2jG
12
+ ~2 y2jG
23
+ y2 (X2jG
31
+ 1°Cit)
1 - dCit) (X2 ~2jGP2 + ~2 y2jG~3 + y2 (X2jG~1 + 1°Cit)

with :

a, ~, y components of the unitary normal rt

V° V° V ° V° V° V°
° EO
4 (.14
1 (rt) =~ + 1:... +
EO pO
4
L_ (X2~2 (.:..u +..:.2.1) _
EO pO
~2 Y. (.:.ll
pO
+.:Jl) _ y2 (X2 (..:.ll +.:.ll)
EO EO EO
12, 1-:1 -:1 3 31

The Young moduli are much larger than the shear moduli. So :
_--'-__ # (X2 ~2jG 12 + ~2 y2jG 23 + y2 (X2jG 31 1
=__
I - d(rt) (X2 ~2jGI02 + ~2 y2jG~3 + y2 (X2jG~1 1 - d.

if rt is not too near of an orthotropic direction

32. Damage evolution - Macro-rrwdelling

The conjugate quantity to!i is

y=aED I =_1_
ad a : cst (1 _ Ii) Et

where Et is the shear energy. The quantity Y is similar to the energy release rate in

Fracture Mechanics; it governs the damage evolution and then the rupture. For 3D

Carbon-Carbon we propose the following model:

d = h IT + gO::, Z)] if d < I otherwise d = I

Z = < all + a22+a33 >+
X t =sup,; $ t Y~

h, g are functions depending on the material. Figure 4 presents experimental results for

three tests: compression test at 45° (plane 1-2), tension test at 45° (plane 1-2), torsion test

(axis 3).
 19

Torsion
d

M Po

Figure 4 : Damage evolution for several tests

3.3. Damage and anelastic strains

The effective stress and strain are :
cr"
crij =~ for i l' j
I - cJ
lijp = Eijp (1- cJ) for i l' j

One uses a plasticity model with isotropic hardening. The elasticity domain is :
( cr 121012
2I 0
+ cf 23
2/ 0
023 +
cf 2/ 0) 1{l
31 031 - R (p):S; 0

where p --t R (p) is the hardening function which is identified from a tension test at 45°

(orthotropic plane).

To be complete. one has to add rupture criteria in the fibers directions: - Ec:S; Eii :s; lOt

More details can be founded in [5]

4. A second example : Damage meso-modelling of laminate

At the level we called meso-level the composite may be schematized by :

- a homogeneous single layer in the thickness

- an interface which is a surface entity connecting two adjacent layers and which

depends on the relative directions of their fibers

~ Figure 5 : Laminate modelling

20

These entities being modelled and identified, the mechanical behavior reconstituting

of any laminate is then a relatively easy task. It is to be noticed that we limit ourselves to

single layers with only one reinforced direction. The single layer is also analyzed at a

smaller level, the level of its constituents: fibers, matrix, interfaces. Some more or less

qualitative informations is thus transferred at the single-layer level by means of a

homogenization process.
The transverse rigidity in compression being supposed equal to ~, one obtains the

following energy for the damaged material :

p~
1 a" 2 ° + .........°)a
Y,o
__ (-'..L -(--'-*
Yo,
a +
<-a 22>+
2
+
<a 22 >+
2
+
2
a l2
)
-0-2 EO EO EO 11 22 EO (1 _ d')E O (1 _ d)G O
I I 2 Z Z IZ

where d and d' are two scalar damage variables which are constant within the thickness.

The conjugate variables associated with dissipation are :

Y -
a [EOJm 1 [a~zJm
- - ---,---""------:-
d - ad I a:cst - 2 G IOZ(! - d)2

Y -
a [EOJm 1 [< a Z2 >!Jm
- - --;,---==-'---;;-
d' - ad I a:cst - 2 E 10Z(1 - d,)2

Where [Jm denotes the mean value through the thickness. From experimental results

it follows that the governing quantities of damage evolution are :

'[ = ~~ [Y d + bY d,]l!2 '[' = ~~ = [Y'd]1/2

where b is a material constant. Experimentally, one obtains:

Y- YO
d= < ~ >+ ifd< 1 ; d= 1 otherwise

d' = bd if d' < 1 and '[' < Y'c ; d' = 1 otherwise

 21

Figure 6 : Material curve for the evolution of d

Models with delay effects are also used [13]. They differ from the previous one if the

damage rates are very high only.

To describe the anelastic phenomena due to damage one uses a plasticity model. Details

can be founded in [11] [12] [13]. The identification of the material parameters has been

done for severallarninates. Results for T300 - 914 and IM6 - 914 are given in [13] [6].

This modelling has been checked out on numerous experimental tests.

~ : Near the edges it is necessary to take out-plane stresses into account In order
to simplify Young's modulus E3 and the shear modulus G13 and G 23 are taken constant

and thus damage effects of out-plane stresses are assumed to affect interface behavior

only.

For the interface which is a mechanical surface entity similar modelling is used [13] [1]

[12].

S. RUPTURE COMPUTATION
The rupture phenomenon happens after two phases. In a first step, the micro-voids and

micro-cracks growth is nearly uniform: it is the initiation stage. From the critical point

(or from a point just beside) the strain and also the damages become more and more

localized; a macro-crack appears and growths until becoming unstable. If the ftrSt stage
22

is well described thanks to this damage approach, the full simulation of the rupture leads

to severe difficulties (22) (2) [3J.

For laminate composites, and more generally for meso-modelling these difficulties

partially vanish. To avoid them completely, one can use delay damage modelling. For

delamination, such computations have be done in [1].

6. CONCLUSION
This Damage mechanics Approach seems to be a powerful tool for the prediction of

complex structure deterioration [24J. It can be applied to most of composite materials.

Other further researches are of course necessary to solve completely the computational

difficulties in order to achieve a true Rupture Theory.

 23

REFERENCES

[1] ALLIX 0., "Delaminage par la mecanique de l'Endommagement", Calcul des

structures et Intelligence Anificielle, vol. I, Pluralis, Ed. I.M. Fouet, P. Ladeveze, R.
Ohayon, 1989.
[2] BAZANT Z.P. et PUAUDIER-CABOT G., "Non local damage: Continuum Model
and Localisation Instability", Repon nO 87-U Nonhwestem University, Evanston, also in
J. of Appl. Mech., AS ME, 1987, in press.
[3] BELYTSCHKO T. and LASRY D., "Localisation limiters and numerical strategies
for strain-softening materials", in Cracking and Damage, Edited by J. Mazars and Z.P.
Bazant, Elsevier. 1988, p. 349-362.
[4] CORDEBOIS J.P., SIDOROFF F., "Endommagement anisotrope en elasticite et
plasticite", Journal de MecaniQ.Ue Theorique et AppliQllee. numero special, 1982, pp. 45-
60.
[5] DUMONT J.P., LADEVEZE P., POSS M. and REMOND Y., ''Damage mechanics
for 3D composites", Composite Structure 8. 1987, pp. 119-141.
[6] GILLETTA D., GIRARD H., LADEVEZE P., "Composites 2D it fibres it haute
resistance: modelisation mecanique de la couche elementaire"JNC 5. Pluralis, Paris,
1986, p. 685-697.
[7] GILORMINI P., LICHT C., SUQUET P., "Growth of voids in a ductile matrix: a
review", Arch. Mech., 40, 1988, pp. 43-80.
[8] KACHANOV L.M., ''Time of the rupture process under creep conditions", Izv Akad
Nauk S.S.R. Otd Tech Nauk, 8,1958, pp. 26-31
[9] KRAJCINOVIC D., "Continuum damage mechanics", Appl. Mech. Rev., 37,1984,
p. 1-6.
[10] LADEVEZE P., "Sur une theorie de l'endommagement anisotrope", Rawort Interne
nO 34. Laboratoire de MecaniQ.Ue et Technologie. Cach an , 1983.
[11] LADEVEZE P., "Sur la Mecanique de I'Endommagement des Composites", lli,l,
Pluralis, Paris, 1986, p. 667-683.
[12] ALLIX 0., LADEVEZE P., LE DANTEC E., VITTECOQ E., Damage Mechanics
for Composites Laminates under/Complex Loading", IUTAM/ICM Symposium on
Yielding, Damage and Failure of Anisotropic Solids, GRENOBLE, August (France),
1987.
[13] LADEVEZE P., About a Damage Mechanics Approach, Int. Symp. on Mechanics
and Mechanisms of Damage in Composites and Multimaterials, MECAMAT, November,
St Etienne (France) (to appear in Int. J. of Fatigue and Fracture of Engineering Materials
and Structures 1990).
24

[14] LEMAITRE 1., "How to use Damage Mechanics", Nuclear En~neerin~ and Design,
80, 1984, p. 233-245.
[15] LECKIE F.A., ONAT E.T., "Tensoria1 nature of damage measuring internal
variable", Proceedin~s LU.T.A.M. Symposium Physical Non-Linearities in Structural
Analysis. 1980, Springer-Verlag.
[16] MAZARS 1., "Application de la mecanique de l'endommagement au comportement
non-lineaire et 11 la rupture du beton de structure", These d'Etat. Universite Paris 6,
1984.
[17] MAZARS I., PUAUDIER-CABOT G." "Continuum damage theory: Application
to concrete", Internal Rep. nO 71 LMT Cachan, and also I of en~n~ Mechanics ASCE.
1986~
[18] MURAKAMI S., "Notion of continuum damage mechanics and its application to
anisotrope creep damage theory", I. Engng Mat. Tech., 105,1983, p. 99-105.
[19] RABOTNOV Y.N., "Creep rupture", Proc XII, Int. Con. Appll., Mech., Stanford-
Springer.
[20 REIFFNIDER K., "Stiffness reduction mechanism in composite materials", ASTM-
STP 775-novembre, 1980.
[21] SANCHEZ-PALENCIA E., "Non homogeneous media and vibration theory", Lect.
Not. Phys., 127,~, Heidelber~, 1980.
[22] SAOURIDIS c., "Identification et numerisation objectives des comportements
adoucissants : une approche mu1tiechelle de l'endommagement du beton", These de
l'Universit6 Paris 6, 1988.
[23] TALREIA R., "Transverse cracking and stiffness reduction in composite laminates"
Iournal of composites materials, 1985, vol. 19, Iuly.
[24] ALLIX 0., LADEVEZE P., GILLETTA D., OHA YON R."A Damage Prediction
Method for Composite Structures"Int. Iournal for Num. Meth. in Engineering, Vol. 27,
p 271-283., 1989