Mechanics of Materials 4 (1985) 121-136

North-Holland

121

ANALYSIS OF CRACKED LAMINATES: A VARIATIONAL APPROACH

Z. H A S H I N
Department of Solid Mechanics, Materials and Structures, Tel Aviv University, Tel Aviv, Israel
Received 7 June 1985

Cross-ply ([0 , 90L) laminates which contain distributions of intralaminar cracks within the 90 ply are analyzed by
variational methods for tensile and for shear membrane loading. Admissible stress systems which satisfy equilibrium and all
boundary and interface conditions are constructed and the principle of minimum complementary energy is employed to find an
optimal approximation. This yields approximate stress fields and rigorous lower bounds for stiffnesses. The analysis allows for
crack interaction and statistical distribution of cracks. Results for Young's modulus are in perfect agreement with experimental
data. Young's modulus and shear modulus results approach definite limits for large crack density. Typical stress variations are
presented for glass/epoxy and for graphite/epoxy laminates and their implications for the progressive damage and failure
process of laminates are discussed.

1. Introduction

It is well known that during loading, whether

static or cyclic, certain laminae in a fiber composite laminate will develop cracks along the fibers.
Such cracks are known in the literature as intralaminar cracks in contrast to interlaminar cracks
which develop on the planes in between the plies.
The intralaminar cracks are essentially normal to
the laminae planes and traverse the available distance throughout the laminate. Together with interlaminar cracks they are the major damage
suffered by a laminate during loading. The present
work is concerned with evaluation of laminate
stiffness reduction due to intralaminar crack distribution and with simplified analysis of the internal
stresses in such a cracked laminate.
The subject of analysis of stiffness reduction of
a cracked laminate has received repeated attention
in the literature. Reifsnider (1977) has made the
major discovery that intralaminar cracks tend to
form periodic saturation patterns which he has
called the Characteristic Damage State (CDS).
Such patterns are observed in static and in fatigue
loading. This has been elaborated by Reifsnider
and coauthors (see Reifsnider and Talug (1980),
and Reifsnider and Jamison (1982)) et al. who

have also devised a simple shear lag method to

evaluate stiffness reduction due to cracks. The
shear lag analysis is based on the following assumptions:
(a) The normal stress in external load direction
is constant over ply thickness.
(b) Shear stresses develop only within a
boundary layer of unknown thickness in between
plies.
(c) Cracks remain sufficiently far apart so that
their mutual interaction can be neglected.
This simple analysis has yielded reasonably good
predictions of stiffness reduction (Highsmith and
Reifsnider, 1982) and is of considerable conceptual value but it is not sufficiently accurate. In
particular, the thickness of the boundary layer
must be assumed in somewhat arbitrary fashion
and the transverse normal stresses cannot be
estimated. As will be seen these stresses may play
an important role in laminate failure. Furthermore, the assumption of no interaction of cracks is
inconsistent with experimental data as will be seen
in the course of present analysis.
Another method of analysis is due to Laws,
Dvorak and Hejazi (1983) and Laws and Dvorak
(1985). They have employed the 'self consistent
scheme' approximation (SCS) to estimate the stiff-

0167-6636/85/$3.30 1985, Elsevier Science Publishers B.V. (North-Holland)

122

Z. Hashin / Cracked laminatex

ness reduction of a cracked ply and have then used

conventional laminate analysis with equivalent homogeneous plies to obtain the stiffness reduction.
They have also obtained good agreement with
experimental data for [0/90~]~ glass/epoxy
laminate. One problem with this analysis is of
conceptual nature for their model of a cracked
material is infinite in all directions while a cracked
ply may be assumed to extend to infinity only in
one direction and crack opening is significantly
constrained by adjacentplies. Such constraint can
not be incorporated into the SCS. Another problem is that the method is incapable of evaluating
the stress concentrations produced by interlaminar
cracks; but such information is very important for
description of the failure process.
The problem of stiffness reduction has also
been considered by Talreja (1985) in terms of
continuum damage theory assuming vector field
description of damage. This results in a mathematical description of stiffness reduction involving
a set of unknown ply damage constants which
much be determined experimentally.
In the following we present a variational approach to the problems of stiffness reduction and
stress evaluation which incorporates all of the
important aspects of the problem and involves
only one assumption: normal ply stresses in load
direction are constant over ply thickness. We thus
construct admissible stress fields which satisfy
equilibrium and all boundary and interface conditions and we determine stiffness reduction and
stresses on the basis of the principle of minimum
complementary energy. The stiffness reduction of
the [0/90~], ply is in outstanding agreement with
experimental data and it is believed that the internal stresses obtained are useful simple approximations.

2. Analysis of cross-ply in tension

We consider a symmetric cross-ply laminate
which is subjected to uniform in-plane membrane
loads. It is assumed that the 90 ply has developed
continuous intralaminar cracks in fiber direction
which extend from edge to edge in y direction and
are perpendicular to the plane of the laminate, Fig.
1.

Xy

Fig. 1. Cracked cross-ply loaded in tension and shear.

When there are no cracks the layers are in states

oo(,,,) , o,,.
()1"'~, where
of constant plane stress o,,o ( . , ) , _,,.
rn is the ply index, 1, 2, which are found by
conventional laminate analysis. The cracks introduce stress perturbations which are denoted o~,;''~
where i, j range over x, y, z, but some components may vanish according to the case. Thus the
stresses in the cracked laminate are

o,';'""-- o,,'"'"'+

o,i'""

(2.1)

where the i, j in the first stress on the right, but

not in the second, are restricted to x, y.
Now let it be assumed that the laminate is
subjected to uniform in-plane membrane force N,,
in x direction. In this case the surviving components of virgin laminate stress are ~J~"'~ and o,~,~'''~
and the perturbation stresses are plane stresses in
the xz plane with components O~',"~(x, z),
o~"~(x, z) and %('"~(x, z). Otherwise, we shall

Z. Hashin / Cracked laminates

make only one single assumption with respect to
the perturbation stresses, namely that o,~,.-~1)and -,.,-12)
are functions of x only and are thus constant
throughout ply thickness. This is obviously incorrect for it is to be expected that there will be
considerable stress concentration, even a singularity in 6,,, at the crack tips. However, an analysis
to evaluate such concentration or singularity would
have to be based on the concept of homogeneous
anisotropic laminae which retain their properties
for elements which are infinitesimal relative to ply
thickness. But it must be remembered that a typical ply will contain about 20-50 fibers through the
thickness and that a representative volume element
which replaces the differential element of continuum mechanics must in itself contain many
fibers. It follows that variation of local average
stress throughout the thickness will be much more
subdued than that predicted by classical continuum mechanics and this reasoning implies that
the assumption is less severe than it might appear.
According to this assumption the o,., stresses in
the laminae may be expressed in the form
()~1) -_- 19"1,
O,:,~

{I) =

v,

--O1q')1

_ o(2) ~
O.,o:

(X),

_{2)__

Ox.x- - -

123

A relation between '~1 and q52 is easily obtained. Equilibrium in x direction of the undamaged laminate requires
f_J' o,, dz = U,, = 2(olt, + 2t2)

(2.6)

where N,~ is the membrane force and t 1, t 2 are

the ply thicknesses. If the same membrane force is
applied to the cracked laminate it follows from
(2.2) that the equilibrium condition in x direction
is
2(o1'i + o2t2) -2[Ol,lqh ( x ) + o2,2+2(x)] = N,,.
Therefore,

oltlq~ , (x) + o2t2q~2 ( x ) = 0.

(2.7)

We now place the origin of the system of coordinates at the center of the distance 2a between
any two typical cracks. The stresses in the rectangular region Ix ] ~ a, I z I ~ h, Fig. 2, must satisfy
the following interface and boundary conditions

o;I'(),, o) = o,

(2.8a)

0"2,

(2.2a)

tt)= o 2)(x, t,),

(2.8b)

--0"2~2(X),

(2.2b)

h) = o,

(2.8c)

%')'(x, t,)= o.'2:)(x, t,),

(2.8d)

where ~1 and q'2 are unknown functions.

The equilibrium equations for the plies are
(m)/~
36,,I r a ) / 3 x + ",oo~..
/ , , , = 0,

(2.3a)

I n l/),/.~, + goo._.
(m)/Ta_
o,.:
/~,, = 0.

(2.3b)

O( m )

The stresses o~i

are absent since they are constant in the plies and thus their derivatives vanish.
Inserting (2.2a) into (2.3a), 6,: is obtained by
integration. Inserting the resulting 6~_- into (2.3b),
o:: is obtained by another integration. The results

I:.: ~'

!-

are
(1) __

o~: - o 1[~'1 ( x ) z + f , ( x ) ] ,

o))'= -Ol[eP'l'(Z)z2 + f , ( x ) z + gl(x)],

(2.4)

0(~ )= 02 [~2 (X)Z + f 2 ( X ) ] ,

oj2) = _ol[dp~,(x)z2 q _ f 2 ( x ) z q _ g 2 ( x ) ] ,

(2.5)

where f,,(x) and g,,(x) are unknown functions

and primes on q~,, denote derivatives with respect
to x.

!:::

i/I,i]

_+__ h
Fig. 2. Laminate region between two cracks.

124

Z. Hashin / Cracked laminates

%{2:'(x, h) = 0,

(2.8e)

o~l,'(+a, z ) = - o , ,

Izl ~<t 1,

(2.9a)

a~'(++_a, z ) = 0 ,

Izl ~<t I.

(2.9b)

In (2.8), (a) follows from symmetry, (b, d) from

traction continuity and (c, e) since there is no
transverse load in z direction. Equations (2.9)
state that the crack surfaces are traction free since
in the absence of cracks o_o(~)_
o~ and u_o(~)_
O.
~: 2
-x. 2
-Next we denote

(2.10)

051( X ) = 05( X )

and thickness 2b are taken as the stress system

(2.11-2.12) with condition (2.15) for a = a,,. These
must be added to the virgin laminate stresses o,/"'
according to (2.1). Obviously the resulting stresses
satisfy equilibrium and all interface and boundary
conditions. They thus form a so-called admissible
stress system in the context of the principle of
minimum complementary energy. We shall therefore use this principle to obtain the function 05.
Since the laminate is entirely loaded by tractions the complementary energy functional assumes the form

l)(. = f s, a,5,16k, dV

and express 052 in terms of 05 from (2.7). Then all

stresses (2.2b) and (2.4)-(2.5) are expressed in
terms of 05. Inserting these stresses into (2.8) the
unknown functions gin(X) and g,,(x) are easily
expressed in terms of 05. The resulting perturbation
stresses are then

where S, jkt are the local compliances and 6,/ is the

admissible stress system. It may be shown that for
any elastic body containing cracks (2.16) can be
expressed in the form

o~' ) =

(2.1 la)

(j(. = U{9 + U[.

(2.11b)

where

(2.11c)

u{O
~ r~
o o dV,
= ~j~,jklo,/o~l

(2.18a)

U~. = fs,,k,o,l,o'~ ~ dV,

(2.18b)

- = Oijo nu Otl.
,
Or!

(2.18c)

o,05(x),

o { " - o,05'( x )z,

ozz(l} = 0 1 0 5 " ( X )

(ht l --Z2) '

o~2}= ol( tl/tz)05( x ),

(2.12a)

o~ )= 01(tl/tz)05'(x)(h - z),

(2.12b)

o(_.2)= 01( tl/t2)05"( x )( h - z) 2.

(2.12c)

In view of (2.11b,c) the crack surface boundary

conditions (2.9) assume the form
05(+__a)=1,

05'(+a)=0,

0~<lzl~<tl.(2.13)

This implies from (2.12a,b) that

O"(2)(xx,--+a, z) = o l ( t l / t 2 ) ,

(2.14)

o,T(+a, z)=0.
These conditions also follow directly from equilibrium of a transverse section through the crack,
(2.6). It follows that (2.13) applies for the entire
thickness of the laminate. Thus
05(+a) = 1,

05'(+a) = 0,

0~< Izl ~<h. (2.15)

Suppose now that the 90 ply, labelled 1, contains many cracks and let the distance between
any two adjacent cracks be 2a,,. Then the perturbation stresses in any rectangle of length 2a,

(2.16)

(2.17)

Here oi is the actual stress field in the elastic

body without cracks and thus (2.18a) is its actual
stress energy. The stress oij temporarily denotes
the stress o,(7'). This theorem is proved in Appendix 1. It facilitates the computation of f/c since
only the perturbation part of the admissible stress
system is needed to evaluate (2.18b).
Next, t h e effective elastic compliances of the
undamaged and the damaged laminate are defined
in terms of the stress energy stored. It is recalled,
(see e.g. (Hashin, 1983)) that when any composite
material body is loaded by homogeneous traction
boundary conditions
t

T~(S)=o-,,n i

(2.19)

where 8o are constant and nj are the components

of the normal, then the average stresses are 6 , and
the stress energy is rigorously expressed by

Uc = S,~kfl-so~y

(2.20)

Z. Hashin / Cracked laminates

where S,*~ is the effective elastic compliance

tensor.
When a laminate is loaded by membrane forces
N on its edges and the plane surfaces are traction
free

6,j= N i j 2 h ,

where 1 is in fiber direction and 2, 3 are transverse

directions and the elastic properties are
E A - axial Young's modulus in fiber direction,
vA - associated axial Poisson's ratio.

E r - transverse Young's modulus.

i, j = x, y,

V = 2hA,

125

v~ - associated transverse Poisson's ratio,

(2.21)

lit ~ lix" liv

GA axial shear modulus,

G r - transverse shear modulus.

where 2h is the laminate thickness and A is its

plane area. In the special case of loading by N,~
only we have for the undamaged and the damaged
laminate, respectively

U~9=(o(~/2E{))V,

Uc=(%/2E,)V

(2.22)

The plane perturbation stresses ..,,,("'~,, o,<'"): and

o(.j''~ are now identified in each ply with stresses in
(2.25). Thus
o (xIx) = O'22,

.
O~!)

0.,2)

O~ 2 ) -

0"23,

o ~::1 ) = O"33,

Oi2,

O.(~ )

O~.

where
Accordingly the energy densities are

% = N , , / 2 h = o l t l / h + o212/h

(2.23)

2 W"> = o~ ',)'-/E r - o~'~)oJ))2v T / E v

and E~~, E, are Young's moduli of undamaged
and damaged laminate, respectively.
It follows from the principle of minimum complementary energy that if the stress system
(2.11)-(2.12) with the constraint (2.15) is introduced into (2.18b) then

(2.23)

/--9,. >/ Uc.

In view of (2.17) and (2.22) this inequality assumes

the form
1 / E , <~ l/E{) + 2U~./Voff

(2.24)

which provides a lower bound on E,. The best

lower bound in the context of the present formulation will be obtained by specifying the function
q>(x) which minimizes U~.. This is a standard problem in the calculus of variations which we now
proceed to resolve.
First the stress energy densities of the different
laminae are needed. The general stress energy density W of a transversely isotropic unidirectional
fiber composite is given by

-- 0.:2%s 2 v

--

/ ' ~

(2.26a)

1" '

+o::(2)'-/ E t + oi2)'-/G
,: / A "

(2.26b)

We consider again the laminate region which is

bounded by two transverse planes through adjacent cracks, Fig. 2. For reasons of symmetry it is
sufficient to take the region - a ~<x ~ a, 0 ~ z ~< h
with unit thickness in y direction. Further symmetry consideration reveals that
q,(x) = q ~ ( - x ) .

(2.27)

The integral U~, (2.18b), for this region is given

by

U~-,, = 2

f)~ltlfo" W~dzdx + 2 j

t 4~1

j.[ ' h W2dzdx. (2.28)

Introducing (2.11)-(2.12) into (2.26), the results

into (2.28) and carrying out the ,: integrations we
have
o( fa

r,c()o4-+ t~c,)244"

+ t,~C220''2 + tC,,ep,2]dx

0.3s)2VA/E a

T/ET

--0.ds/GT + ( 0.(2 + O~s ) / G A

\c

2W,2) = 0.,,(~-':'r/r'~,- o~-,)o)).-'2v x / E x

G'.,, =-2 J

2 W = o , / , , = o2,/EA + ( o 2 + o d , ) / E v
- - O , l (0"22 q'-

+0.-(1-I:/E T + o ~l>~-/~

(2.29)

where
(2.25)

G)o = 1 / E T + 1 / X E A ,

(2.30a)

126

'7 Itashin

/ ('racked

(,,.--(p,/E,)(X+~)-(vAX/3Ex).

(2.30b)

('2~ = (;k + 1 )(3X2 + 12X + 8 ) / 6 0 E ~ .

(2.30c)

('~, = ' ~ ( I / G , + X / G x).

(2.30d)

X = I2/I I.

(2.30e)

lanlinates

Thus
A}=

2( a cosh ,~p sin tip + fi sinh aO cos fi0 )

c~ sin 2tip + fi sinh 2a0

A1

2(fl cosh ~O s i n / / p - c~ sinh e~O cos tip }

sin 2tip + B sinh 2~ 0

It is convenient to introduce the nondimensional

variable
~ = .\/t~

{2.31 )

(2.38)
This determines the stresses (2.11)-(2.12) and for
this purpose the required derivatives are listed

and thus (2.29) assumes the form

0'(x)=t
C'[,,= : O l U f
--

1 dq,

(ocAI+,SA 2 s i n h a ~ c o s f i , ~

1 d~

[C(),)02 + C()20(deO/dg; ~ )

+ ( a A 2 - flA, ) cosh a,~ sin fi,~.

+ C22(deq,/dU~ )2 + C l , ( d 0 / d ~ ) 2 ] d~

1 d%
q,"(

x ) =

--

- -

(2.32)
= [( c~-'- B e ) A , + 2aBAe]cosh cg: cos fl~

where p = a/t~,
The Euler-Lagrange equation for q5 is
da,;b/d,~ a + p d 2 ( ) / d ~ : + qO = O,
/) = ( ( ' 0 2 - - ( ' 1 1 } / C 2 2 ,

q = (;}t,/C:2

+ [(c~ 2 - fi2)A 2 - 2cq3A,]sinh a~ sin fl~.

(2.33)

The solution of (2.33) is of the form e ' where r

are the roots of the characteristic equation
r 4 + pr 2 + q = 0.

(2.34)

The four roots are

r= +(a+ifl),

i=~-l,

~ = ql 4 c o s 10"
tan 0 = ; 4 q / p : 2 I

Typical stress distributions will be presented

and discussed further below. At the present time
we return to the problem of stiffness reduction. To
evaluate U~.,, it would seem at first sight that
(2.37)-(2.38) must be substituted into (2.32). requiring very much work. Fortunately this is not
necessary. It is shown in Appendix 2 that by
proper utilization of the differential equation
(2.33). (2.32) can be expressed in the simple form

<i,, =

fi = q 4si n 20.
1

C.x(0).

X(O) = - ( d 3 0 / d ~ 3) 1~ ,,
(2.35)

cosh 2aO - cos 2tip

= 2 a f t ( a 2 + / 3 2 ) a sin 2/30 + fl sinh 2 a 0 "

provided that 4q > pc. Consequently the solutions

are of the form
0 = e ~ '' cos ,8,~,

e + '< sin ,8,~,

(2.36)

but it is more convenient to use hyperbolic functions instead of exponentials, because of the symmetry condition (2.27). It follows that only even
product functions are admissible solutions and
thus
q)= Ai cosh Mg cos fl~ + Ae sinh cd~ sin flid

(2.39)

(2.40)
We now consider the case of a large rectangular
laminate in which the 90 ply has an arbitrary
distribution of cracks and the distance between
any crack pair is 2a,,. Then. as has been pointed
out before, the stresses (2.11)-(2.12) with 0,
evaluated for regions - a,, ~< x ~< a,, are an admissible stress system. Consequently,
N

{2.37)

where A~ and A 2 are constants which are easily

determined from the boundary conditions (2.15).

E x(p,,),

el= I

p. = a . / & .

(2.41)

Z. Hashm / Cracked laminates

Let the length of the laminate in x direction by

L >> h. Then
N

L = 2 Y~ a,, = 2t, Y'~ p,,.

n

(2.42)

k - 1

If a,, is a r a n d o m variable then 0,, is also r a n d o m

and let its probability density function be P(P).
Then for a large number of cracks N

127

where c is the crack density, i.e. the number of

cracks per unit length defined by
c = 1 / 2 a = 1/2pt~.

(2.50)

When the distance between cracks is very small,

p ~ 0, we have
--

<

(2.51)

~_, p,, = N ( p } ,
n-

(2.43)

3. Analysis of cross-ply in shear

E X(P,,) = N(X(O)),
n

We now consider the case .when the cross-ply

laminate is loaded by a shear membrane force N,,
only or. equivalently, by an average shear stress

( 0 ) = fo~P(o)O do,
{ X ( P ) ) = f o ~ P ( P ) X ( P ) do.

(2.44)

Consider inequality (2.24). The volume V is given

by

V = Lh = L(t, + t2)

(2.45)

since U~.,, has been evaluated for unit thickness in

y direction and for 0 ~< z ~< h. Introducing
(2.41)-(2.45) with (2.30c) into (2.24) we have after
some rearrangement

(X(O))

(2.46)

where
r/(X)=(3)~ 2+12~,+8)/60,

? t = t 2 / t ,,

k j = oh/%,

(2.47)

and the value of k I and E~~ must be found from

conventional analysis of the undamaged laminate.
This defines a lower b o u n d on E, of the damaged
laminate. If the cracks are equidistant, a,, = a,
p,, = p = a / t I and

<X(p)>=X(O),

<p>=O.

(2.48)

When the cracks are far apart and do not

interact thus a >> t~ and 0 is large, (2.46) reduces
to

E--~<~-E~ + ~

k~(~)a(a'-+fl2)t,c

(2.49)

6 , , = N,,./2h = %.

(3.1)

If the laminate contains no cracks, and the laminae

are of the same unidirectional fiber composite, it
behaves as a homogeneous material with axial
shear modulus GA and the state of stress inside is
everywhere
o , , = o~i, = 6, =

~,

all others vanish.

(3.2)

It is not clear if such a state of pure shear will

produce cracks in the plies. However, if it does
there will be cracks along the fibers of both the 0
and the 90 plies. This is a complex situation
which will not be considered here. We shall assume again an intralaminar crack distribution ill
the 90 ply only similar to the one considered in
Section 2. Our problem is thus to determine shear
modulus reduction for such crack distribution
which is presumably due to a tensile load applied
previously or simultaneously with shear.
Such cracks will obviously perturb the homogeneous state of stress (3.2) and it may be shown
that the elastic fields in the presence of the cracks
are anti-plane with respect to the v direction.
Therefore, the perturbation stresses in the laminate
are o~',!'t(x. =). o,!2")(x, y), m = l . 2 ,
all others
vanish. It follows that the only surviving equilibrium equation is

a o~7'/ax + a o,!='"' / a = = 0.

(3.3)

12S

Z. ttashin / Cracked hmlinates

As in Section 2 we assttme that the primary,

load carrying, stress, o,, in this case, is a function
of x only. Thus, the perturbation stresses o~', * are

Therefore,

o~1,,= - Tol].:l(-~.').

2W~ = o~z':/O~ + o ~(: 2 ' /: U/ cl

O~21,= --ro~,:(x).

(3.4)

(_(1t: +o):II): )/G,~,

=~o,,

2Wj

(3.ll)
.

Equilibrium of any section in the vz plane

requires that

The stress o~l) must be a symmetric function of x

and therefore

t,<',) + < o 5 ' = 0.

4(x) = ~(-x).

(3.5)

Denoting ~,~(x) = + ( x ) , (3.4) then assume the form

<"=

-.~,),~ (.~- ).

<~'= -,,,[1 + { t , / , e ) + ( x ) ] .

(3.6)

Inserting (3.6) into (3.3) and carrying out the z

integrations we have

o!~'= ,,) [ ~ ' ( x ) : + f,{ x)],

o,C'=

Therefore U(,,, is again given by (2.28) with the

energy densities (3.11). Introducing the stresses
(3.6) (3.9) and carrying out the z integrations we
have
u/,,

Ga,

f"

(3.7)

o{~)(x, 0) = 0

from symmetry,

(3.8a)

(1 + 1/~,)4 ~

+3

~,,[{,,/te)~'(x):+f2{~)].

The stresses (3.6)-(3.7) must satisfy the following

boundary and interface conditions, Fig. 2

(3.12)

I+X~

~-

d,~

(3.13)

where again
0 = a/ll,

~ = g2/[l,

~=

x/t,.

(3.14)

o,,)'(x, t,) = o,(~>(x, t,).

(3.8b)

The Euler-Lagrange equation which minimizes

(3.13) is

o,,:'>(.v, h) = 0.

(3.8c)

d2~/d~j 2 + f.t2~b= 0.

o~1,)= 0

(3.8d)

/_

on crack surfaces.

3(1 + l/X)

(3.15)

1 + XGA/G l "

Insertion of (3.7) into (3.8a-c) easily determines

the functions f l ( x ) and r e ( x ) . Thus

The solution of (3.15) which satisfies (3.10) is

o (')

~b= cosh p,~/cosh/~p.

o()~b'(x)z

(3.9)

o,(!' = o,)( t,/ie )+'( x )( h - z ).

Considering again a typical region between two
cracks, Fig. 2, defined by - a ~ x ~ < a , 0 ~ < z 4 h
the boundary condition (3.8d) assumes the form

+(+.)=0.

(3.10)

(3.16)

Therefore the total, admissible, stresses in the

laminate are
6~1' = %(1 - cosh/.t,~/cosh/.tO),
sinh p,~ z
8())=,, %/Xcosh ~P tt
(3.17)

We now proceed precisely as in Section 2 and

evaluate the complementary energy in the above
region for 0 ~ y ~ 1. Recalling (2.25) the perturbation stresses are defined with respect to the material
system of axes x~, x 2, x 3 as follows:

o,: - %/-tcosh ~P

< 9 = o,:.

for any region between two cracks.

To evaluate (3.13) for the function (3.16) we
again make use of the Euler equation, (3.15). It is

~5,= ~,~.

(
6~.'=%

-(2)_

t, cosh/.t)
1-t ~ cosht~p '
sinh ~,~ h - z
t2

h=t)+&,

129

Z. Hashin / Cracked laminates

easily shown that

U~'-,, -

Table 1

'rt2 ( 1 + ~,-G~
3GA
GA)
T /~ tanh/~p.

Recalling the d e v e l o p m e n t in
(2.22)-(2.24), in the present case

(3.18)
Section

2,

Property

Glass/Epoxy

Graphite/Epoxy

E,,,
ET

41.7
13.0
3.40
4.58
0.30
0.42

208.3
6.5
1.65
2.30
0.255
0.413

GA

G1
PA
vT

- ~ A V,

Uc = 2-~, , V,

(3.19)

since now G~?, = G~,. Then (2.24) is replaced by

1 / G , , <~ 1 / G A + 2U~,/Vr,?.

(3.20)

Considering again the case of a distribution of

many cracks with mutual distance 2a,, we apply
the same reasoning as in (2.41)-(2.48) to arrive at
the result

G,,

>!

GA/(I + (tanh)/Y~/~(p)),

(3.21)

and when the distances between cracks are equal

G,,, >/GA/(1 + tanh i~O/Xl~p).

(3.22)

When the distances between cracks are large

O = a/t~ is a large number, tanh #p - 1 and (3.22)
assumes the form
G,, >/ GA/(1 + 1/2~t~p).

(3.23)

Conversely, if the distance between cracks is small

tanh t~pr#p and (3.22) becomes
G,,. >/GA/(1 + 1/2,).

(3.24)

This expression has a very simple interpretation

for the right side can be rewritten as G A t j h . This
implies that the shear stiffness of the 90 ply has
reduced to zero because of the dense crack distribution and the entire remaining shear stiffness is
provided by the undamaged 0 plies.

idirectional properties are chosen as those reported

by Highsmith and Reifsnider (1982) for scotch-ply
specimens. For graphite/epoxy it has been assumed that the fibers are T300 graphite at 0.60
volume concentration and the unidirectional properties are as computed by Hashin (1979) on the
basis of fiber and matrix properties. The properties are summarized in Table 1 with all moduli in
GPa.
In all cases the thickness of a single lamina is
taken after Highsmith and Reifsnider (1982) as
t = 0.203 mm. We consider the cases of [0/90],
(t I = t2) and [0/90~]~ (t I = 3t2) laminates. In
Table 2 we list the stiffnesses and the stress ratio
k~ = Ol/8,, in the 90 layer, for uniaxial tension
8,, = N , , / 2 h , obtained by conventional analysis
of the undamaged laminate.
Stiffness reduction measurements as a function
of crack density in the 90 ply has been given by
Highsmith and Reifsnider (1982) for glass/epoxy
cross plies. Figure 3 shows the comparison between present analysis and experiment for Young's
modulus E, of the [0/90~']~ laminate and it is
seen that there is outstanding agreement. Verification of the [0/90]~ experimental data is not as
good, the experimental results being between 5%
to 10% lower than the analytical results. This is
unreasonable since the analytical results are,
strictly speaking, lower bounds. However, the reTable 2

4. Results for stiffness reduction and stress fields

Glass/Epoxy

We now present numerical examples for stiffness reduction and stress fields based on the analysis given above and discuss their significance.
Results will be given for glass/epoxy and graphite/epoxy laminates. For glass/epoxy the un-

E
Ev
G
kI

Graphite/Epoxy

[0/901~

[0/90~]~

[0/90]~

[0/90~]~

27.56
27.56
3.40
0.464

20.30
34.75
3.40
0.636

107.59
107.59
1.65
0.060

57.05
157.75
1.65
0.114

Z. flas'hm / ('racked h*mim*te.~'

1 ~{I

ported measured value of E l ' of the undamaged

[()o 90oL is 22.7 G P a while the analysis result
given above is 27.57 GPa. Such a large discrepancy
is unusual and no attempt will be made here to
speculate as to its origin. (For the [0/90~']~ the
measured /--{) was 21.0 G P a versus the analytical
result of 20.3 GPa).
Figure 4 shows a comparison of reduction of
Y o u n g ' s m o d u l u s and shear m o d u l u s for
G l a s s / E p o x y [ 0 / 9 0 L, It is seen that the cracks
have a stronger influence on shear modulus reduction. The limiting values of stiffness are given in
the two cases by 0.522E{' and 0.25GII, = 0.25GA.
respectively. It is seen at once that the last result
implies that the effective shear stiffness of the 90
ply has reduced to zero. It is also easily shown by
conventional laminate analysis that the limiting
results for the Young's modulus implies that E w
and G~ of the 90 ply have practically reduced to
zero.
It will be recalled that the limiting values of E,
and G,, are given by the simple expression (2.51)
and (3.24). Denoting these by '( and G(, we list

Ex

YOUNG'S

MODULUS

OF

UNDAMAGED

Ex

YOUNG'S

MODULUS

OF

CRACKED

EXPERIMENTAL

in Table 3 the ratios E ( / E I > and G~,/G~, ', for the

various cases under consideration. The limiting
values are of considerable practical importance
since they give a simple estimate of order of stiffness reduction to be expected. It is seen that for
Young's modulus of G r a p h i t e / E p o x y the reduction is small. The reason for this is the large ratio
E x / ~ in this case which implies that most of tile
stiffness is due to the 0 plies.
We now consider stress fields as defined by
(2.11)-(2.12), (2.37)-(2.39) and (3.17). When the
cracks are far apart thus O = a / t ~ - , o~ it is more
convenient to put the origin of the x , z coordinate
system at the center of the crack and then the
function ,0 should be taken in the form (2.36).
Evaluating the stresses (2,11)-(2.12), the total (admissible) stresses are
' -1-e

'< c o s f l s e + = s i n

O]

O\_
-,,>

'8e +

_ _

o,

,8

LAMINATE

LAMINATE

IExlEx
1.00

"501p= ll2tlc
I0.0
I

5[0,,4.01 3.0
I
.25

2l'O

1.5]

.50

CRACK
o , Glass/Epoxy
Fig. 3. Stiffness r e d u c t i o n of [0 o / 9 031

i.Oi

[
.75

DENSITY,
laminate.

PER rnm

P
[ c

.90

(4.1a)

0 ,2

e "~ sin ,8~

(4.1b)

131

Z. Hashin / Cracked laminates

Z

,I

I L,I- 0-I

t.O0

t = 0.20ram
c : 1/2a

0.522
.50

G,~'G;,y
CRACK

I
1.0

I
2.0

0.250

DENSITY,

I
3.0

PER mm

I
4.0

o , G l a s s / E p o x y laminate.
Fig. 4. Young's and shear modulus reduclion of [0 / 9 03]

a [11
=

ol

B2

-}- ~ 2

Similarly, for the sheared laminate, the stresses

(3.17) for p --, oc assume the form

ea~

2fl

x ( f l c o s ,8~ - a s i n / 9 ~ ) ( 1

+ ?, -

z2/t~),

,,/r o - 1 - e ~

O,!)'/r,, = ff e "~,

(4.1c)
- { 2 ) ,'

o
cosB~+~sin

- +~-e
01

a,,/ro-I

)
B( ,

6,!~/ro = ~ e

~(2)

~2

o,

- - e

"/~ sin fl,~(1 + ~. - ~ ) ,

(4.1e)

B 2 q- a 2
-

o1

e ~,

(4.2)

~*~(1 + ~. - z / t ~ ).

02
X/~

--

(4.1d)

01

a(I)

--

"g;

2aft

X(flcos,8~-asinfl~)

(1 + ) ~ -

,
(4.1f)

~. = t 2 / t l ,

~ = x/t

1.

Examination of the stresses (2.11)-(2.12) reveals that on any section x = const, the maximum
of o,: is at the interface z = t~, where it assumes
the role of interlaminar shear while o_. has a
m a x i m u m absolute value at the laminate midplane
z = 0. In the case of the sheared laminate the shear
stress o . also has a maximum value at the interface z = t~ and is thus interlaminar.
The stress (4.1a) increases monotonically from
zero at the crack tip to the initial value a~. The
stress (4.1b) has a maximum value
r

a,(~}~*-t ~{}, t) = ( a 2 + f12 e-";{'

Table 3
Glass/Epoxy
/

E,./E,
G,! , . / G ~o,

Graphite/Epoxy

[0/90 k

[0/90~]~ [0/90]2

[0/90~k

0.770
0.500

0.522
0.250

0.914
0.250

0.970
0.500

(4.3)

~,,=~

tan

'(fl/a).

The stress (4.1c) has a maximum compressive value

at ,~ = 0 and m a x i m u m tensile value at ~o, given

Z. H(I.~,IIlllI

132

/ ('racked lammate.~

1.00
/fit
2.0

.50
1.0
o'zz (~,0)/o-,

i~ = X/t
6
6.0

2.0

~=x/t

o"xz ({,tl/o-,

~o

c . ~

it

1~

o-

3.0

-.51

{I/{yl~

~---.--- Cr 2

/y

{(,1 I/T O

-I.0*

"

-.i

__~
%

-2.0

Fig. 5. Stresses at isolated crack. [0/90], Graphite/Epox~

laminate under uniaxial tensic, n.

Fig. 6. Stresses at isolated crack. [0/90], laminate unde~

shear.

.50-

-I.00

Fig. 7. Stresses at interacting cracks. [0/90}, Graphite/Epoxy laminate under uniaxial tension.

Z. Hashin / Cracked laminates

by
I+X
-?--(a2+2),

,~g'(O,O 1 -

1 + x ,~2

o_(J)(~0,0)=~(
(0=~tan

+ / 3 2 ) e -"*,

-~ c ~ / B - B / ~

(4.4)

"

Figures 5 and 6 show plots of the stresses in the

90 ply near an isolated crack for the
graphite/epoxy [0/901~ laminate under tension
and under shear respectively. Figures 7 and 8 show
corresponding plots for interacting cracks which
are a distance 2a = 4t apart. These examples and
others show that for the laminate under tension,
when the cracks are far apart, a maximum interlaminar shear stress develops at some distance
from the crack while the maximum transverse
tensile stress o.: develops in the laminate mid-plane
at some other distance from the crack. At the
crack tip there is a large transverse compressive o..
in the midplane. The stresses o,.: and o:: decay to

2.0

't~c)/-to

~/CRACK
-

2.0

~_____________~ ~,/t ~ _ cK
1.0

- I . ~

=1.O-

2.0

~-2Q~I . . ~
a=2t

TO

-2.0

Fig, 8. Stresses at interacting cracks. [0/90 L laminate under

shear.

133

zero while at the same time (~,, builds up to its

initial value o~. As the crack density increases and
the cracks become closer the maximum tensile o_,
shifts to the center point between any two cracks
and a large compressive %: remains at the crack
tips. The stress (~,., assumes a maximum at midpoint between the cracks, is smaller than o] and
decreases as the cracks become close. This implies
that with increasing crack density the load N,, is
increasingly carried by the 0 layers and by interlaminar shear. The stress cr,(~~ in the 0 layer is
easily found from (2.6). It follows that at any
crack tip

I= o,t /t2 +

(4.5)

and as the crack density increases o,-{2)xapproaches

this value everywhere in the 0 layer.
In the case of the laminate under shear, Figs. 6
and 8 the effect of the cracks is to produce an
interlaminar shear o~.. which has a maximum at
~(2)
the crack tip and to increase the shear stress o,~.
in the layers. It follows from equilibrium that for
applied stress T0, o,(2) assumes the value 2% at the
crack tip and this is also its maximum value.
We recall the approximate nature of the stress
analysis in that it has been assumed that o,, and
(~,, stresses are constant through ply thickness. It
follows from Saint Venanrs principle that stresses
evaluated away from crack tips, such as maximum
of (~,.., and o.: for tension, should be more accurate
than stresses evaluated at crack tips.
Finally, we consider some interesting experimental data reported by Highsmith and Reifsnider
(1982) for a [0/90~L G l a s s / E p o x y Laminate.
For the case of uniaxial tension in 0 direction
they have measured the crack density in the 90
ply as a function of applied stress o0. We show in
Fig. 9 a plot of o] = 0.636o 0 versus c, the crack
density per ram. (It is recalled that o] is the stress
in the 90 ply when there are no cracks). Also
shown in the same plot are the maximum values of
O(1)
(1) and O(1)
.,,, c[,:
:. in the 90 ply versus crack
density, assuming that the distance between cracks
is equal, thus p = 1/2t]c. The maximum value of
(~) is always at midpoint between two cracks
while the maxima of stresses o .<1). and
. . o (~1 shift
with crack density, as has been explained above.

134

Z. ttashin / C r a c k e d laminate~ >

that it is reliable for this purpose. It also yields

approximate values for stress perturbations produced by the cracks. These stresses appear physically plausible but it is not known at the present
time to what extent they are quantitatively accurate. In this respect it would be useful to perform
numerical analysis of some of the problems which
have been treated variationally and to compare the
results with present stresses.
It is recalled that the one and only approximation employed in the analysis is that the load
carrying stresses, o,, for load in x direction and
o , , and .~v shear are constant through the respective plies. A more refined variational analysis can
be performed by taking these stresses in more
complicated forms. For example, the perturbation
stresses o,, can be taken as

We observe the interesting phenomenon that

max o,, which is responsible for new crack generation varies erratically between the values 40-54
MPa, while max o,: and max o:: consistently increase with load and crack density. This implies
that at some value of load (or crack density) the
90 ply will fail by delamination from the 0 ply
a n d / o r by splitting at its center plane z = 0.
Finally, we point out that a significant part of
the damage process takes place in the regime of
crack interaction. This can be seen from Figs. 4
and 10 which show that p goes down to a value of
about 1.2 while crack interaction need be considered from a value of O about 4.0 and less.

5. Conclusion

~',=

The variational method which has been employed to analyze cracked cross-ply laminates under tension and shear has yielded excellent results
for stiffness reduction and it is therefore believed

15C

- o, [ ~ , ( x ) + +, (x)( =/t,

o~-',)= -02[o~(.)+ +._(x)((h - :)/,,)"].

This allows for stress concentration at the crack

PERIMENTAL

/~o-,

= o 6 3 6 o-o

O0

///

50

i 'xz(mQx)

O'xxlmOx)

2,0 ip

~ ,4

.20

Fig. 9. Variation of ma:~imum o,,,

.40
o,:

and

o::

)"],

~
I

.60

,.,s p
1

.80

C per

mm

with crack density. [0/90~]~ Glass/Epoxy laminate.

Z. Hashin / Cracked laminates

tips at the price of introducing additional unknown functions ~k and considerably complicating
the variational treatment.
It should be borne in mind that present treatment has been confined to a very simple laminate.
The problem of extension of the method to practically more meaningful laminates such as [ 0 / +
45/90].~ is far from trivial and it is hoped to
address this problem in future work.

135

Then (3) assume the form

r,'(s~) = 0,

~'(s~) = - ~ .

(5)

The complementary energy functional for the

cracked body is given by

O~----fvSiikAi/Ok,dg-- f s,, Tiui d S .

(6)

Introducing (4) into (6) and recalling (2) this may

be expressed in the form

Acknowledgment

0~. = u,!'+ fvs,i. ok,o,'jd V

Support of the Air Force Office of Scientific
Research under Contract 84-0251, monitored by
Major D.A. Glasgow, is gratefully acknowledged.

Appendix 1. Complementary energy of cracked body

+ ~I /V Sijkto, tjoklt d V - ft Tit~ui dS.

s,,

(7)

Consider the first integral in (7). The stress o,'/

satisfies equilibrium since o,!I and 6,j do. Also

s,,,.,o~, = ,oq
We consider two elastic bodies with identically
shaped external surfaces S, bounding the same
volume V. Both bodies have space variable elastic
compliances S,j~.~. The second body contains a
distribution of cracks. Both bodies are subjected to
the same external mixed boundary conditions
~(s~)

= ~'~.

.,(s.)

= ~,,.

s~+ s. = s.

(1)

In the second body the tractions must vanish on

the surface S of the cracks.
Let the stress field in the first body be o ( x ) .
Then the complementary energy of this body is
defined by

v,o= f s,,,,o,,o,,O dV- f,

dS

(2)

Now define an admissible stress field 0i.i within

the volume V of the second body with associated
tractions ~ which satisfies only equilibrium and
the traction boundary conditions. Thus

~ ( s ~ ) = ~,

~(s~) = 0.

f;-- r ? + r/.

s,.

where the tractions in the second surface integral

are defined with respect to the inward normal to
S~.. Introducing (4.5) into (8) we have

--/o

dS - f_

dS.

(9)

The second integral is taken over the two adjacent surfaces of each crack. Since the normal on
one surface is the negative of the normal on the
other surface and since T, and u are continuous
across the crack surface, the integrals on the two
crack surfaces cancel one another and therefore
the total integral vanishes. Introducing the remainder of (9) into (7) we have

,iok, dV

(10)

(3)

0,. = v~ + ~ j / . k ,

(4)

This proves (2.17). Note that the proof given here

is for the more general case of mixed traction-displacement boundary conditions.

Write

ai,-- C + o,',,

0
where cij
is the actual strain field in the firstbody.
Therefore, by virtual work

136

Z. Hashin / Cracked laminates

Appendix 2. Evaluation of U~o

I n t r o d u c i n g (15) into (14), but leaving the term

m u l t i p l y i n g C02 as it stands, we find

To evaluate (2.32) we consider the integral

I =

fP

d p

I = -C22 ' ' ' ( p ) .

~
t~ a,,,2
,~
(C,,0" + C02'~" + ~22"~ + C,1 - ) d ~

(16)

This proves (2,40).

(11)
References
where we have denoted ~ derivative by primes.
The function satisfies the E u l e r - L a g r a n g e equation (2.33) which is written in the form
C22'v + (C~,2 - C,~ ) " + C~o = 0.

(12)

The associated b o u n d a r y conditions (2.15) are

' ( + p ) = 0.

(+0)=1,

(13)

Now multiply (12) by and integrate from - p

to p. Thus
2

-p

[ C221

+ ( Co 2 - C, 1 ) t , +

c00]d

=0
(14)

Integrating by parts and utilizing (13) we have

fo--p

de = 2'"(0) +fo--p

dC
(15)

--p

--p

Hashin, Z. (1979), "Analysis of properties of fiber composites

with anisotropic constituents", J. Appl. Mech. 46, 543.
Hashin, Z. (1983), "Analysis of composite materials", J. AppL
Mech. 50, 481.
Highsmith, A.L. and K.L. Reifsnider (1982), "Stiffnessreduction mechanisms in composite laminates", in: K.L.
Reifsnider, ed. Damage in Composite Materials, ASTM STP
115, ASTM, 103.
Laws, N., G.J. Dvorak and M. Hejazi (1983), "Stiffness changes
in unidirectional composites caused by crack systems",
Mechanics of Materials 2, 127.
Laws, N. and G.J. Dvorak (1985), "The loss of stiffness of
cracked laminates", Proc. IUTAM Eshelby Memorial Symposium, Cambridge University Press, London, 119.
Reifsnider, K.L. (1977), "Some fundamental aspects of the
fatigue and fracture response of composite materials", Proceedings 14th Annual Meeting of Society of Engineering
Science, Bethlehem, PA, 343.

Reifsnider, K.L. and A. Talug (1980), "Analysis of fatigue

damage in composite laminates", lnternat. J. Fatigue, 3.
Reifsnider, K.L. and R. Jamison (1982), "Fracture of fatigueloaded composite laminates", lnternat. J. Fatigue, 187.
Talreja, R. (1985), "Transverse cracking and stiffness reduction
in composite laminates", J. Composite Materials, to appear.