You are on page 1of 28

Engineering Frachue Meckanics Vol. 29, No. 5, pp. 557-584, 1988 0013-7944/88 $3.00 + .

a0
Printed in Great Britain. @ 1988 Pergamon Press pk.

DELAMINATION-A DAMAGE MODE IN


COMPOSITE STRUCTURES

AMAR C. GARG
Aeronautical Engineering Department, Indian Institute of Technology, Bombay-400 076,
India

Abstract-The composite structures have a tendency to delaminate which reduces the strength
and stiffness and thus limits the life of a structure. This behavior of composites has caused
concern amongst the designers to find the ways to delay or prevent delamination in order to
increase the life and the load bearing capability of the structure. Some of the solutions being
considered are: proper lay-up sequence to minimize the interlaminar stresses, improved
structural configurations such as discrete stiffness design, stitching, tough resin systems and the
hybrids. The present paper discusses the state of the art of delamination behavior. A few of the
aspects considered here are: causes of delamination and its effect on structural performance,
analytical and experimental techniques to predict its behavior and some of the preventive
measures to delay the delamination so as to make the structure more damage tolerant.

1. INTRODUCTION

THE COMPOSITES are finding an increased use in various structural applications as they offer a
significant weight savings and have attractive mechanical properties. Like metals, these
materials are notch sensitive and lose much of their structural integrity when damaged. Damage
can be caused during manufacture or during service. The manufacturing defects may be in the
form of delamination and voids due to improper lamination and curing or damage may be
introduced by machining the components for fastener holes and design cutouts etc. Also the
damages during manufacture or maintenance may be caused by dropping tools accidentally. The
service damage may result from the impact by runway debris, hailstones, bird strike, ground
service vehicles, ballistics etc. In many instances, the damage caused by such impacts may not be
visible or barely visible on the surface but may significantly reduce the strength of the
component. For example, Fig. l[l] shows the applied compressive strain, E, while being
impacted by an aluminium projectile (12.5 mm diameter ball) at velocities shown as abscissae for
a T300/5208 graphite epoxy panel of lay-up (*45/O,/rt45/0,/*45/0/90)2,. On this figure are
indicated the regions where the damage may be visible on contact and impact surfaces. It is seen
from this figure that the damage may be visible on contact surface only at the impact velocities
higher than 80-100 m/s, but the compressive strength of the panel is significantly reduced by the
invisible damage. Since normal inspection practice relies primarily on visual inspections, it is

0 Failed on impact
0 Did not fail on impact
n Residual strength
I
8
n

I r\ In ,I I I I
0 25 50- 75 -lo0 125 150 175
Projectile velocity, mlsec

Fig. 1. Effect of impact damage on compressive strain for T300/5208 graphite/epoxy (*45/O,/ f
45/0,/*45/O/90),, laminate[l].

557
558 AMAR C. CARG

l 52 1%

50- ;ze/$y

00
Weight
saving 40- I
I
vs
Aluminimum / /----e37’8%
baseline 30 - ;upper
surface
/ &%
!
PerIX!flt

i25.6,
I
20 - I
I
-Current design allowables

I I
0.5 0.6
Design allowable ultimate strain - i%l

Fig. 2. Effect of design strain levels on weight saving potential for wing structures using graphite/epox~
composites compared to aluminium[Z].

therefore possible that the internal damages (mainly delaminations) caused by low level impact
may not be detected for the life time of a structure. Therefore, damage tolerant design criteria
are established for composite structure to ensure that growth from non-visible damage will not
degrade the strength to less than design ultimate load during the life of the aircraft. This
requirement for the current graphite/epoxy composites is ensured by limiting the design strain to
0.4% [Z, 3] even though the fibres have a breaking strain of over 1.3%. Such a restriction leads
to an increase in structural weight and limits the exploitation of full potential of composites.
Figure 2[2] shows that the weight saving can be up to lo-12% if the design strain can be
increased from 0.4% to 0.6% for a wing structure.
When a load (static or cyclic) is applied to a laminated composite structure, the damage
initiation and propagation occurs in a series of events, whereby individual modes develop and
interact with each other[4]. Due to the anisotropic and heterogeneous nature of the composite
materials, the possibility of damage modes is large[4,5]. However, in a laminated composite,
there are three principal failure modes (intra-ply cracking, interlaminar matrix delamination and
fiber failure which play a role in affecting their mechanical properties. Other types of damages
may simply alter the load levels at which these three damage modes may occur.
Amongst the three principal damage modes mentioned above, the interlaminar matrix
delamination is of major importance. The delamination in a structure subjected to inplane loads
is a subcritical failure mode [4,6] whose effect may be stiffness loss, local stress concentration in
load bearing piles, and a locat instability causing its further growth leading to compressive
failure. In the latter two cases the delamination leads to a redistribution of structural load paths
which, in turn, precipitates structural failure. Hence, delamination indirectly affects the final
failure of the structure thus affecting its life. Therefore, delamination is known as a most
prevalent life limiting damage growth mode[7].
Since delamination may influence the structural performance, its behavior needs to be
thoroughly understood and the techniques for its delay or prevention need to be developed.
Some of the approaches for the prevention or delay in delamination being considered are:
improvement in the current resins by toughening them or developing new tougher resins,
adhesive interleaving, braiding, 3-D weaving, through the thickness stitching and novel designs
to reduce interlaminar stresses. The purpose of the present paper is to review the state of art in
delamination highlighting its importance, reasons for its growth, its influence on structural
performance and the remedial measures for its prevention. The environmental influence on
delamination will not be considered here and it forms the part of another paper.

2. CAUSES OF DELAMINATION
The delamination may result from the interlaminar stresses created by impacts, eccen-
tricities in structural load paths or from discontinuities in the structure. Some of the design
Delamination 559

Free Notch PlY Bonded Bolted Cracked lap


edge (hole) drop joint joint shear

Fig. 3. Sources of out-of-plane loads from load path discontinuities[3].

details which may induce the local out-of-plane loads leading to interlaminar stresses are shown
in Fig. 3[6]. These are (1) straight or (2) curved (near holes) free edges, (3) ply terminations or
ply drop for tapering the thickness, (4) bonded or co-cured joints, (5) a bolted joint and (6) a
cracked lap shear specimen. In all these cases, even if the remote loading is in-plane the local
loads near the discontinuities may be out-of-plane.
Besides mechanical loads, the moisture and temperature also may cause interlaminar
stresses in a laminate. These may be caused by (1) residual thermal stresses due to cool down of
the laminate from the elevated curing temperature or due to the difference between test and
stress free temperatures, (2) residual stresses created by the moisture absorption in the laminate
and (3) moisture gradient through thickness of the laminate.

2.1 Free edge stresses


The occurrence of interlaminar stresses near the free edges of a laminated composite is an
important phenomenon. These stresses (Fig. 4) occur due to the mismatch in engineering
properties, i.e. mismatch in Poisson’s ratio (v.+,) and coefficient of mutual influence (v,,,~)
between layers[8]. If there is no mismatch of v._ or v,,,~ between layers, interlaminar stresses do
not exist even if there is mismatch in elastic and shear moduli[8]. The statement does not mean
that the delamination between layer of same orientation will not occur. The delamination
between such layers may occur if there exists the interface moment which is caused by the
neighboring plies of different orientations or elastic properties. For example the delamination
at the free edge of a laminate (*45/O/90), may occur at the midsurface i.e. at the interface 90/90
plies where a large interface moment and tensile uz exist as shown in Fig. 5.
The mismatch in v,,(~v,,) give rise to interlaminar normal (oz) and shear (T,,=)stresses and
the mismatch in ~,y,X(S~.Y,X) causes interlaminar shear stress (T=~) near the free edge of a
laminate (Fig. 4). The magnitude of these stresses depend on the magnitude of ]SvXyland ]8~~,~(,
elastic and shear moduli, stacking sequence, mode of loading and environmental conditions.
Herakovich’s[8] analysis has shown that largest values of ~~~~~~~~ and ]SV,.,] occur for
(+ 11 So/-- 11.5”) and (22”/90”) layups, respectively. These are in agreement with his test data [8]
on (‘t @),laminates for 10” % 0 5 30” which were found to be delamination prone indicating high
interlaminar shear. Since high interlaminar shear is more likely to exist in the cases where the
angle plies (* 0) are adjacent to each other, one should try to avoid placing them together in a
laminate.
The layup sequence[8-121 has considerable influence on the magnitude and direction of
interlaminar stresses. Figure 5 shows the interlaminar stresses for (&45/O/90), and
(90/45/O/-45), laminates under the action of an uniaxial tensile strain lX= 0.5%. It is observed
that these laminates produce high tensile and compressive uz at midsurface, respectively. Thus
the latter laminate is less delamination prone compared to the other. Herakovich’s[8] studies
have shown that interspersing f 45 layers such as to create alternate positive and negative loads
560 AMAR C. GARG

‘,r +
I

Equilibrium free body diagram


4. Interlaminar stresses in a ply of a symmetric balanced laminate under axial
load.

Max shear
= 49.6 MPa

(7” = 93-
93-
-3.45+
-1X$2-7--+
(MPa)

cry = 182.7 ----3

93 -
--3-45---a
93 -
IMPaJ

Fig. 5. Int~rlami~ar stresses near boundary for (~45/0/9~)* and t90/45/0/-45)~ laminates: The
stresses are shown for the laminates under uniaxial tension by eX= 0.5%. At the interfaces, the interface
moments are also shown in Nm/m[S].
561

O.OlO- __-____--
/-
0172
01
0.008-
032
A2

0.008-

0404-

0.002-

I I 1 I I 1 1 I
O- 5 6 7 8 9 10 11 12
Number of plies

Fig. 6. Parametric study of variation in delamination onset strains with thickness[l3].

(o,,), reduce cZ and thus reduces the chances of delamination. In a study, Herakovich[8] has
shown that the strain for the onset of delamination in [*45/O/90], laminate is approximately half
the strain at first delamination in a [O/45/90/-45], laminate. The first laminate was seen to
delaminate along the midplane due to high o, and the latter delaminated in the 90” layer,
apparently due to oZ and 7,, effects.
The other parameter affecting the interlaminar stresses is ply thickness. Thick plies tend to
encourage higher interlaminar stresses thus causing early delamination. This is obvious from Fig.
6[13] where the critical strain (E,) for the onset of delamination for various laminates is shown. lc
is seen to decrease with increase in 90” ply thickness in the laminates.
So far the discussion have been based on the mechanical loads ignoring the effects of
residual stresses caused by temperature and moisture changes. Residual thermal stresses are
always present due to cool down of the laminate from the elevated curing temperature and these
may have considerable influence on interlaminar stresses[S, 14-201. However for the present
discussions, we shall not consider it.

2.2. Delamination caused by impact


When a laminate is hit by a projectile or an impactor[l, 3,211 (Fig. 7) the material directly
under the projectile is compressed and translates laterally in a time frame much less than that is
required for the overall response of the structure. The highly localized deformation gradient
causes large transverse shear and normal stresses which can cause the damages to propagate and
even failure of the laminate. Another effect of the impact is the creation of a compression stress

0
Projectile

Laminate

Fig. 7. Deformation by a projectile hitting a laminate[21].


562 AMAR C. GARG

Fig. 8. Geometry of impact test and primary failure modes[3]. Damages (b) and (c) may occur at low
velocity and (d) may occur at high velocity.

wave which travels from the impact surface through the thickness of the laminate. This wave is
reflected from the back surface as a tension wave which can cause failure at the first weak
interface, resulting in chipping or splintering of the parts of the rear ply[3].
Both internal stress waves and local out-of-plane deformations may initiate delamination at
interfaces where there is major change in the angle between plies. The amount and type of the
damage in the laminate depends upon the size, type and geometry of the laminate, impact energy
and the loading on the laminate at the time of impact. At relatively low impact velocities, the
laminate can respond[3] by bending and failing either by shear resulting in delamination (Fig.
8b) or flexural failure (Fig. 8c) depending on short or long beam, respectively.
At higher velocities somewhat different damage modes occur which may be caused by the
combination of stress waves and the out of plane deformations as described in previous
paragraphs. The damage may result in delamination, fiber failure, matrix cracking etc.

lmm SC&Z
7i
Impact L,
2mm

CFlPEEK

7i
Impact

CFkpoxy 1

-7

I /
- -
,
----a \ 77/ ,_._JF, , I !

Fig. 9. Scaled diagram comparing cracking due to a 75 impact in graphite/PEEK and graphite/epoxy
laminates [*45, O,, *45, OJ. lay-up, cross-sectioned parallel to 0” fibers[23].
Delamination 563

12 0’

10
1

“E
3 8- Graphite/epoxy
e
m /

$ 6-
%
*,/!m+
.5
E
m 4-

zi

2 8 0
0
t /
? I I I ! I
0 2 4 6 8 10 12
Impact energy J0Uk.S

Fig. 10. Variation of delaminated area due to dropweight impact in graphite/epoxy and
graphite/PEEK laminates with [ * 45, 03, f 45, O,], lay-up[23].

If the velocity of the projectile is fairly high, the laminate acts as relatively rigid resulting in
shear out and complete penetration of projectile (Fig. 8d).
Figure 9 shows [22-241 the damage development in graphite/epoxy and graphite/PEEK?
laminates by 7 J impactors. Delamination is seen in both laminates but more predominant in
epoxy specimen compared to PEEK. Figure 10 shows the delamination area as a function of
impact energy for graphite/epoxy and graphite/PEEK [*45/0,/+45/02], laminates. Delamina-
tion for PEEK is relatively smaller than epoxy specimens. It is due to higher fracture toughness
of PEEK compared to epoxy.

2.3. Delamination caused by matrix cracks


Another cause of the delamination development in a laminate is the matrix cracking in
off-axis plies. These off-axis ply cracks create interlaminar stresses as shown in Fig. 11. In a

X
Y

IL z

Fig. 11. Interlaminar stresses due to matrix crack in off-axis plies.

tPEEK: Poly-ether-ether-ketone.

EPH 29:5-D
564 AMAR C. GARG

1000
6

-6
-1000
0 0.25 0.50 0.75 1 .oo
z/t

Fig. 12. Distribution of interlaminar stresses uz and 7Xxnear a crack in the 90” plies of a [0, ~t45,90)~
graphite/e~xy laminate (a half-thickness edge view is shownf[lZ].

\
0
- I - - 90
+45
I ( 1 - 45
! -45
h 1 \ +45
I - \- \ 90
0
/

/
Pre-fracture patterns near
fracture site

Fig. 13. Pre-fracture crack patterns showing a breakdown of the characteristic damage state by crack
coupling and longitudinal cracking[25].

quasi-isotropic graphite/epoxy laminate (O/90/&45),, Reifsnider and coworkers[4,12,25] have


estimated these interlaminar stresses near the tip of the matrix crack in 90” ply. In Fig. 12 the
distribution of cr, and r,, are shown for such a laminate loaded to a strain of 1000 /_Min axial
direction. These interlaminar stresses frequently cause local delamination which grow along ply
interfaces near matrix cracks. Schematic representation of such delaminations and matrix cracks
in (0/9O/lt45), and (Ofzlz45/90), laminates are shown in Fig. 13 just before fracture of the
specimen.

3. EFFECT OF DELAMINATION
It has been discussed in preceding sections that strong interlaminar stresses arising out of
various reasons lead to delamination which may have influence in strength and stiffness of the
laminate and thus may affect the structural performance. It is observed in general that the tensile
Delamination 565

behavior of a laminate may not be significantly affected by the occurrence of delamination but
compressive behavior may be critical.
The tensile static and cyclic loading behavior of quasi-isotropic graphite/epoxy laminates by
Reifsnider, Stinchcomb and coworkers[4, 11,12,25] indicated that the damage development in
the form of matrix cracks and delamination does not appear to influence the residual strength of
these laminates. However, the stiffness is found to decrease. For example, the percentage change
in stiffness of (O/&45/90), (Type I) and (O/90/*45), (Type II) laminates after 500 cycles at
393 MPa was seen to reduce by 7.2% and 13.6%, respectively. On the other hand the residual
strengths were within or above the tensile static strength. Some of such results are shown in
Table 1. Such a behavior is somewhat surprising as considerable damage occurred in both
laminates during fatigue loading. Of course the damage was more extensive in type I laminate
than type II. For example, fatigue loading (393 MPa for 500 cycles) of types I and II laminates
caused edge delaminations extending almost 100% and 0% in 90” plies and 21% and 12.5% in
f 45 interfaces, respectively [ 111.
Though in the preceding example, the damage does not appear to affect the tensile strength
appreciably, the delamination of the laminate causes the redistribution of the stresses. For
example the longitudinal stress (a,) in 0” plies of a undamaged (O/90/*45), laminate is
2632 MPa, but its delamination into sublaminates (O/90) and *45 causes 0; to increase in 0”
plies to 3 13 1 MPa [4] for crX= 1000 MPa applied to the laminate. The redistribution of the stress
may influence the fatigue life and precipitate early failure of the structure.
Delamination plays a more critical role in affecting the compressive behavior as it may
cause localized buckling and the high interlaminar shear and normal stresses at the edges of the
buckled regions (Fig. 14) which often lead to cyclic delamination growth. Such a phenomenon is
designated as instability-related delamination growth by Whitcomb[27-291. If the localized
buckling propagation is not arrested or the load is not redistributed, the damage may propagate
until the panel fails by general instability. Whitcomb studied such a delamination growth in
compressively loaded composite laminates analytically and experimentally. As shown in Fig. 15,
for a laminate with one surface delaminated partially (Fig. 14), the critical load PC decreases with
the increase in delamination length (a). The decrease also depends upon the interlaminar
fracture toughness, G rC and GrIC (ignoring the contribution of GmC). For large delamination
length, GrC increase has no influence but improvement in G rrC causes substantial improvement
in critical load. The location of delamination along thickness also effects the critical load, PC.
The influence of depth for two types of laminates designated as type I [O,/O/45/9O/-45)7,/O,1 and
type II [0~/(0/45/90/-45),,/0/452/0] where the buckled regions are 3 or 4 zero degree plies deep,
is shown in Fig. 16[28]. The critical load is higher for short delamination for thicker buckled
region but the trend is reversed for long delaminations.
Yin, Simistses and coworkers[3&34] also studied extensively the delamination growth due
to compressive loads in various laminates under different edge conditions. The results of one of
such studies are presented in Fig. 17[32] where the nondimensional load P( = PC/PC, where Per
is the max buckling load for undamaged specimen and PC is the critical load for buckling
initiation of delaminated surface) is shown as a function of delamination length (a = a/L) for
various depths of delamination (h= h/t) of a wide column of length, L with simply supported

Table 1. Residual strength and stiffness data[ 1 l]

Max. cyclic Residual Percentage


Laminate stress Cycles strength+ change in
typet MPa MPa stiffness

I 234 10,000 434 -5.5


I 289 10,000 451 -8.3
I 393 500 470 -7.2
II 234 10,000 546 -7.0
II 289 10,000 555 -6.1
II 393 500 483 -13.6

? Laminate types-Type I: (O/+45/90),, Type II: (O/90/*45),.


$ Static strengths-Type I: 462 f 36 MPa, Type II: 486 f 23 MPa.
566 AMAR C. GARG

Fig. 14. A delaminated laminate under compression (8, is initial imperfection).

ends. It is seen that p decreases significantly as the delamination grows. As long as Z < h, the
buckling load is not appreciably affected by the presence-of delamication. Simitses et a/.[321
have also computed the (total) load carrying capability PD where PD is defined as the total
compressive load carried by two delaminated sections divided by maximum compressive load for
undamaged panel. &, is shown to decrease with the thickness of delaminated section and it is
minimum when the delamination occurs at mild thickness. This sets a lower bound on the load
carrying capability of the panels with delamination. The behavior is similar to one shown by
Whitcomb[28] that for long delaminations thicker buckled surface has lower critical load than
thinner one. Similar effects on the buckling behavior of composites with delamination have been
shown by Majumdar and Suryanarayana[35].
The experimental and analytical studies of Chatterjee et a1.[36] show the decrease in failure
load (under three point bend) of the laminates with implanted disbonds at various depths of a

I I I I I
0 10 20 30 40 50

Delamination length, mm

Fig. 15. Effect of G,, and Gn, on critical load predicted by mixed-mode criterion laminate:
[(0,/45/90/-45),,/0/45,/0]. Initial imperfection, 6” = 0 (Buckling curve shown for reference)[28].
Delamination

loo-

80 -

60 -

5
d?
40 -

20

i
01 I I I I I
0 10 20 30 40 50

Delamination length. mm

Fig. 16. Effect of laminate type on critical load predicted by mixed-mode criterion. (G,, = 200 Jm-*;
1/
G IIc = 1000 Jm-*; initial imperfection, S0 = 0)[28], Type I: [0,/O/(0/45/90/-45),,/0,], Type II:

[011~0,45/90/-45),,/0/45,/0], (Vertical arrow shows the location of delamination).

laminate [04/f452/~452/04]2s. This is shown in Fig. 18[36]. Konishi and Johnston[37] have also
shown the reduction in compression strength of laminate with delamination.
Rhodes, Williams and coworkers [ 1,2,2 1,34-401 extensively studied the compression
behavior of impact damaged composite panels. These studies have indicated that the residual
strength significantly deteriorates with the increase in impact velocity (or the impact damage) as
shown in Fig. 1.
The fatigue loading of the delaminated panel may accelerate the damage growth parti-
cularly under compressive cycling. This tends to reduce the residual strength and stiffness. Such
a behavior has been studied by several workers[37,41-421 where the delamination was either
caused by impact or by implanted flaw. Figure 19[41] shows the ratio of peak cyclic strain to
static failure strain vs cycles to failure (compression-compression cyclic loading with stress ratio
R = 10) for the impacted panels about 32 plies thick with the lay-ups 38% 0” plies, 50% *45”
plies and 12% 90” plies. It is seen from this figure that the failure strain decreases considerably
with the increase in number of cycles. Similar results were obtained by Ram Kumar[42] for

0.00 0.25 0.50 0.75 1

Fig. 17. Effect of symmetric delamination on buckling loads of wide column with simply supported
ends[32].
AMAR C. GARG

010

Disbond after Ply No

OO 2I 4I 6I 8I I
10 II
12 14 I
16 1
18 11 22
20 24 11 28
26 1
30 f
32 :

Fig. 18. Failure load for disbonds of 2.54 cm length in 2.54 cm wide beam[36].

impact damaged laminates where it has been shown that compression-compression fatigue
cycling considerably reduces the threshold strain for impacted panels.
The disbonds or delamination defects caused during fabrication of joints may grow during
service and thus may have deteriorating influence on joint strengths. Frame and Jackson’s[43]
studies at RAE England have shown that the joint strength can get reduced by about 22% as a
result of 8% reduction in bond area caused by disbond (Fig. 20) at room temperature dry
condition but no effect of disbond could be seen for hot/wet condition. Deo and Ratwani[44]
have also shown the degradation in adhesive joint strength due to defective bond for stepped
adhesive joint.

O.lO-

0.9 -

0.8-

Specimen Impact
0.5 -
0 125mmBY25Omm 23 J
Al25mmBY25Omm 4OJ
0 3 STR PANEL 136J
0.4 -

103 104 105 106

Cycles to failure

Fig. 19. Constant amplitude fatigue of impact damaged specimens[41],


Delamination 569

Bond len th = 20 mm
4 80- Bondw 3 th = 25 mm
:
C
r,
F 60-
?
z

+
1 40-

z
$
2 20-
s
2
$

0 7.0 Nil 7.0 Nil ’ 7.0 Disbond dia.mm


20 I 120 120 20 ’ 20 Temp.“c
AR 1 1% , 1% 1% 1 1%
Static Residual

Fig. 20. Effect of disbond on bonded joint tensile strength[43]. A. R. represents as-received panels.
The residual strength is the tensile strength after 10h cycles of fatigue loading.

The extensive studies of adhesively bonded joints by Hartsmith[45-461 demonstrated that


joints having thin adherends are more damage tolerant than thick adherends for which minor
flaws may create problems. The location and size of flaw in the overlap region is important in
effecting the strength. If small flaw lies in middle of overlap it may not have any influence on the
performance of the joint. But if the flaw is large or it occurs near the edge of the joint, it shifts the
critical location of stress peak from the edge of the joint to a location adjacent to the flaw. When
this happens, the load redistribution is high and substantial loss of joint strength occurs. In order
to make thicker bonded structures damage tolerant, the mechanical joint is required in
conjunction with adhesive bond.
The ply drop is quite common in aircraft structural components where tapering of the
thickness is required. The ply drop regions act as stress raisers and thus affect the strength of the
laminate. The studies of Grimes and Dusablon[47] on the ply terminated specimens show that
the fatigue degradation of such specimens is higher than plane laminates, and the worst affected
were the laminates with 0” ply drop. The results have indicated a significant endurance limit ( 10h
cycles) strain level degradation caused by ply drop offs in a laminate under compressive fatigue
loading typical of upper wing skins.

4. SUPPRESSION OF DELAMINATION
It has been discussed earlier that the damage or flaws may degrade the mechanical
properties of the laminated structures thus affecting the structural performance. The problem
the designer faces is to consider the invisible or a barely visible damages on the surface which
may be responsible for deterioration in the structural performance. The present usage of
composites is limited to secondary structural components and these are designed to operate at
design ultimate strains sufficiently low that impact damage does not degrade their structural
performance. However, the heavily loaded primary structures such as wing panels are designed
to efficiently carry loads at high strains and the structural performance may be degraded by the
impact damage. If the designer keeps the design strains lower, the structure is bound to become
heavy and the advantages of using a composite is lost. Thus in order to make these structures
570 AMAR C. GARG

more damage tolerant or to reduce the delamination growth, various methods are being
considered. Some of these are described below.

4.1. Improved resins


The delamination growth in a composite depends on its interlaminar fracture toughness
which in turn depends on the toughness of the matrix material. The current resins (epoxies) used
as matrix materials are brittle having mode I fracture toughness of about 80-200 J/m2 whereas
the desirable fracture toughness of the matrix material is considered to be about 1.9-3 kJ/m* [48]
so as to obtain interlaminar fracture toughness of composites to be about 700-1000 J/m’.
Williams and Rhode[21], Sarah and coworkers[23,24] demonstrated that the tough resins can
significantly reduce the damages caused by impact and substantially improve the residual
strength following impact. Figure 10 shows the delamination area caused by the impact damage
for composites using epoxy and PEEK as matrix materials. It is seen that the composite with
PEEK has relatively, smaller damage due to its higher toughness. Similarly, the drop in residual
strengths following impact as shown in Figs 21 and 22 are higher for EPOXY than PEEK
specimens. However for high impact energy (12 J), the degradation in tensile strength of
PEEK is more than epoxy. This was the result of broken fibers throughout the thickness for
PEEK specimens at high impact energy. The compressive strength, however, is significantly
improved (Fig. 22).

[+45,0x, +45,0,1, LAY-UP

0
0 G/peek
OWE

01 I 1 I I I I
2 4 6 8 10 12

Impact energy JOIdeS

Fig. 21. Residual tensile strength a, of graphite/PEEK and graphite/epoxy laminate following
impact[23].

[?45,03, f45.021. LAY-UP

I I I I I I
0'
2 4 6 8 10 12
Impact energy JOLIlt%

Fig. 22. Residual compressive strength o, of graphite/PEEK and graphite/epoxy laminates following
impact[23].
Delamination 571

Since the increase in resin toughness may improve significantly the performance of the
composite, four general approaches have been considered[48]. These are:

(a) Toughened thermo-sets. The present thermosets (epoxies) are brittle and may be
toughened by adding thermoplastics, interleafing ‘soft’ and ‘hard’ layers, increasing the length
between crosslinks, creating interpenetrating networks, using novel curing agents, and adding a
second phase. Second phase may consist of reactive and/or unreactive rubber[4&591, chopped
fiber, fibrils and inorganic particles[60]. Adding rubbers to epoxies has been observed to cause
10-30 fold increase in fracture energy of the epoxies[W-521, however the addition of rubber
causes some degradation in Youngs modulus (E), glass transition temperature (T,) and tensile
strength of the epoxies. Also it causes some increase in coefficient of thermal expansion.
(b) Ljghtly cross-linker ~hermoplas?ics. This includes the addition of cross linkers to
polysulfones.
(c) Crystalline thermoplastics such as polyesters, polyarylethers (PEEK, etc.). PEEK appears
to be a promising material as it has good mechanical properties and high fracture resistance. The
fracture toughness for mode I delamination of graphite/PEEK is about 2.5 kJ/m2[6 l-621.
(d) Linear fhermoplastics such as legible polyimi~es. These appear to be attractive candidate
materials for composite matrices, as these have high fracture toughness combined with the high
flexural modulus and solvent resistance.

4.2. Through thickness reinforce~nt


Various studies have shown that 3-dimensional integrated structures formed by 3-D
braiding process[63-641 or through thickness stitching [65] offer improved damage tolerance,
fracture toughness, and capability to withstand out-of-plane tension loads as the delamination is
suppressed by such means. Using polyester and Kevlar Yarns at various stitch spacings and pitch
for stitching through the thickness graphite/epoxy laminates, Dexter and Funk[65] demon-
strated a considerable improvement in damage tolerance. Figures 23-25 illustrate that the
impact damaged area is reduced, the residual compressive strain and GIc are increased as a
result of stitching the laminate by Kevlar Yarn. However, the stitching reduces the compressive
and tensile strength of the laminate compared to unstitched laminates for undamaged specimens
by about 20-25%.

4.3. Interleafing
Interleafing of the laminate, i.e. sandwiching thin films of high toughness and shear strain
the~oplastics between graphite/epoxy layers has been demons~ated by Evans et al. [66] to be an
effective way to improve the damage tolerance of composites. One of their studies has shown a gain
of about 100% in compressive residual strength after impact for interleafed graphite/epoxy
laminate compared to first generation graphite/epoxy composite.
All the above measures are useful only if the failure occurs through delamination. If it is by
transverse shear these approaches are ineffective and an improvement in shear modulus of the
resin is important.

Stitched T300/3501-6
2.5
125mm spaced Kevlar
3 stitches/cm
2.0 T300/5208 unstitched

Damage 1.5 6.25mm spaced Kevlar


are% 3 stitches/cm
cm
1.0

0 0.9 I.6 2.7 3.6 4.5 5.4


impact energy per unit thickness, kN.m/m
Fig. 23. Damage area as a function of impact energy per unit thickness for Kevlar stitched laminates
and unstitched laminates[65].
572 AMAR C. GARG

Stitched T300/3501+

12.5mm spaced Kevlar


3 stitches/cm

3 stitches/cm

T300/5208 unstiched

I I I I I 1
0.9 1.8 2.7 3.6 4.5 5.4

Impact energy per unit thickness, kNm/m

Fig. 24. Compressipn failure strain as a function of impact energy per unit thickness for Kevlar stitch
panels and unstitched panels [65].

T300/3501-6 Quasi-isotropic

6.25mm Kevlar
3 Stitches/cm

Standard _
deviation

k
0.6

T300/5208

Unidirect 0 Stitch
Tape direction
(unstitched)

Fig. 25. Interlaminar fracture toughness of Kevlar stitched laminates and unstitched laminates[65].

4.4. Design considerations


It has already been stated that the cause of delamination is the interlaminar stresses which
are dependent on the lay-up sequence. The extensive studies of Herakovich[8] have shown that
one should avoid using angle plies (* 0) together, it is preferred that these should be interspersed
between other plies in order to reduce the interlaminar shear. The other guideline which he has
suggested is to minimize interface moments so as to reduce the interlaminar normal stresses. The
attempt of the designer should be to minimize the free edges and the geometric discontinuities
which may not always be possible. Kim[67] suggested that the reinforcement of free edges is an
effective way to prevent and/or delay initiation of delamination. His studies have indicated that
the reinforcement of edges of various delamination prone laminates by fiber glass cloth
significantly delays or prevents the initiation of edge delamination as shown in Table 2.
Some of other design features[l] which may be used to arrest the propagating damage to
improve the damage tolerance of the structures are:

(1) discrete-stiffness design-here the panel has the regions of high (0” plies predominant) and
Delamination 573

Table 2. Delamination threshold strain for unreinforced and rein-


forced graphite/epoxy laminates[67]

Delamination threshold strain, percentage

Laminate Unreinforced Reinforced

[* 45/o/901, 0.52 No delamination


[O/f 45/901, 0.66 No delamination
0.54 No delamination
0.39 No delamination
0.26 0.50
0.22 0.48

low (only +45” plies) axial stiffness. It uses the concept that regions of low axial stiffness are
more tolerant to impact[l].
(2) Mechanically fastening-mechanically fastening the panels together is shown to be an
effective way of arresting the propagating damage thus improving the damage tolerance.

5. PREDICTION OF DELAMINATION BEHAVIOR

5.1. Prediction of interlaminar stresses


Prediction of delamination behavior in composites has been the subject of many years. Most
of such work has been to determine the stresses in the boundary layer near the free edges
whether curved or straight. The problem of determining the edge stresses involves the solution
of an interfacial crack between two highly anisotropic fiber composite laminae under general
loading conditions. The analytical approaches to such problems have been to determine stress
intensity factors at the tips of the crack at the interface of two or more isotropic layers or
traversely isotropic half planes[36,68-751, or for an edge delamination in an angle ply
laminate[76].
Because of the complexities, the analytical solutions to edge delamination has been rather
limited. Therefore more emphasis has been placed on the use of either finite
difference[l6,18,77] or finite element methods[7,29,78-971. Using these numerical tech-
niques it has been possible to incorporate the effect of moisture and temperature on the
interlaminar stresses.
However, for thick realistic structural laminates, solving edge stress problems tend to be
expensive using such methods. The only exception to this criticism is an approximate solution
provided by Pagan0 and Pipes[98] for the normal component of interlaminar stress state, gz.
Though, it is less accurate compared to other numerical methods, it may give an idea of
delamination prone and resistant stacking sequences on the basis of approximate cr, distribution.
This ability of differentiating the stacking sequences offers promise that the approximation when
employed with a suitable failure criterion can accurately predict edge delamination. Accord-
ingly, Rodini and Eisenman[99] used the approximate o, distribution combined with the
statistical failure criterion to predict the onset of delamination in various composites.
Pagano[lOO, 1011 proposed a theory to define a complete stress field within an arbitrary
composite laminate using the Reissner’s variational principle to laminated bodies. The theory is
simple and gives reasonable results for the free edge stresses which are adequate for the
structural design purposes. However, the method may not be very suitable for thick laminates.
Another useful approximate method has been proposed by Whitney[l02] to estimate the
free edge stresses. This method gives the interlaminar stresses which are close to the exact
estimates.

5.2. Onset of delamination


Basically there are two approaches: one based on fracture mechanics and the other on
strength. The fracture mechanics method used the assumption of the delamination as an edge
crack at the interface and thus determines either stress intensity factor, K or the strain energy
514 AMAR C. GARG

releases rate, G. On the basis of the K or G approach, the prediction of delamination onset or
propagation is made.
Rybicki, Schmueser and Fox[83], modelled the edge delamination as an interfacial crack at
the free edge and analysed it by finite element method. They determined that the strain energy
release rate remained nearly constant during the propagation of delamination, indicating that the
critical strain energy release rate (Ge) may be a useful criterion for delamination growth. In
addition an approach to predict the delamination onset or initiation was also suggested. In order
to do this a characteristic length of the flaw equivalent to one ply thickness was assumed and the
onset of delamination in (* 301 f 30/90/90), laminate was successfully predicted.
Wang[76] also determined delamination onset criterion in angle ply laminates on the basis
of fracture mechanics. His studies have indicated that there exists a critical delamination length
a* (for example, a* = 2 h for b/h = 16 where b is specimen width and h is ply thickness) for
various composite laminates. This indicates that any edge flaw a0 inherently in the composite,
which is less than a*, will experience rapid unstable growth as the load or G reaches a critical
value. Any initial delamination greater than a* will experience a stable growth under
monotonically rising load showing that there exists an inherently built in crack arrest mechanism
for edge delamination. This behavior of G remaining constant with delamination propagation
beyond a characteristic flaw length has been noted by other researchers also[7,77,78,82]. The
a* may be an important quantity in the life prediction for delaminated composite materials and
structures subjected to static and cyclic loading.
The importance of an existence of an initial flaw of a characteristic length a* has also been
discussed by Crossman, Mauri and Warren[l03]. a* can be determined using the equation

E*Gc
a* -
m::
where E* is the effective modulus and mc is the critical stress at which the pre-existing flaw in the
material begins to grow.
Assuming E* (= 14.5 GPa) to be equivalent to transverse modulus, uc (= 35 MPa)
equivalent to transverse tensile stress, and Gc = 154 J/m*, a* was estimated to be about
0.58 mm for a wet graphite/epoxy laminate. This is about 44 ply thickness. On this basis they
argued that unless a, exceeds uc at a depth a* from the edge, the delamination will not initiate.
However, the characteristic flaw appears to be large because of wet matrix having low ac and
higher Gc. Usually for the current dry epoxy composites such as T300/934, ac = 50-55 MPa,
Gc = 80-100 J/m* and E2 = 9.5-10.5 GPa, and thus a* = 0.09-0.11 mm, which is about one ply
thick as obtained by Rybicki er al.[83].
The concept of delamination prediction on the basis of strain energy release rate was further
explored by O’Brien[79] who has given a simple method using only the laminate theory to
predict the onset of delamination. His method makes use of the drop in stiffness of the laminate
during delamination which can be related to the strain energy release rate. Accordingly, the
initiation of delamination in a laminate occurs at the critical strain, lc given by

EC =
JG%m2Gc- E*)
where Gc is the critical strain energy release rate, E* is the stiffness of the laminate completely
delaminated along one or more interfaces given by

E*=‘=’
(3)
t

with Ei being the stiffness of the ith sublaminate of the thickness ti+ Elam is the extensional
stiffness of the laminate computed from laminate theory.
O’Brien[79] has used the relation (2) to successfully predict the onset of delamination in
Delamination 57.5

0.008
r

[+45J-45dOJ90.1,
0.006 - ‘b

Test data ‘\ 0

l, 0.004 -

EC =
0 Test data
0.002 - 0 Predicted value

n=l n=2 n=3


I I I
0 8 16 24

Number of plies

Fig. 26. Edge delamination onset in O/90 interfaces of T300/5208 laminates[6].

various laminates. The predicted behavior is seen to agree well with test data. A few of his results
are presented in Fig. 26[6] where the strain for delamination onset at the O/90 interface for
[+45./-45,/0,/90,1, for rt = 1,2,3, is shown.
Whitney and Browning[l04] modified eq. (2) in order to account for any discrepancies
between theoretically and experimentally determined modulus of the undamaged laminate i.e.

2Gc
” = J t( 1 - E*/E,,,)&,,

where I?,,, is experimentally determined laminate modulus. Additionally, E* should be


determined using the moduli of the sublaminates which are usually unsymmetric rather than
using analysis of symmetric sublaminate.
Whitcomb[29] also used the fracture mechanics approach in order to study the instability
related delamination growth i.e. growth of delamination in a panel under in-plane compression.
The other approach to predict the delamination onset and propagation involves the detailed
stress analysis near the free edge used in conjunction with a failure criterion. The failure
criterion considered are similar to Nuismer and Whitney’s[lOS] point and average stress failure
or Wu’s[lO6] fracture criterion for composites.
Kim and Soni[78] used the point stress and average stress criterion to predict delamination
initiation in laminates such as (&30,/90,),, (0,/*45,/90,), and (0,/90,/+ 45,), for several
values of n. Their studies have shown the point stress criterion to be conservative but average
stress failure criterion predicts the behavior which agrees well with the test data. For average
stress criterion, they have used a characteristic distance, a* as equivalent to a ply thickness and
a, equivalent to transverse strength of the composite material and the assumption is similar to
one by Rybicki et a1.[81] for characteristic dimension. Figure 27 shows a few of their results.
However, the point or average stress approaches are suitable only when the delamination
onset is governed by mz. If the delamination occurs under the influence of shear and normal
stresses, the approach can be modified and the criterion similar to Wu[106] defined as “the onset
of delamination, will occur when at a characteristic distance, a* from the edge, the stress
state satisfies the following failure criterion”[lOb]

where Fi and Fij are strength tensors whose components can be expressed in terms of the
material principal strengths, and gi are the stress components at the interface at a point a* away
from the edge.
AMAR C. GARG

I I I I I
0 1 2 3 4
n

Fig. 27. Comparison of prediction and experiment for onset of delamination[80]. -Average stress
criterion. - - - - maximum point stress criterion. 0 experiments.

Herakovich et a/.[841 used this approach for the prediction of onset of delamination in
various composite laminates. However, it is not obvious from their work whether they have
evaluated the failure criterion (eq. 5) at a characteristic distance. According to their paper, the
tensor polynomial (5) seems to have been evaluated near the edge for each finite element to
determine the location of initial failure in various laminate configurations. In addition the
individual terms of the polynomial were examined to identify modes of failure. It has been shown
by these studies that for angle ply (* 0) laminates, 8 = 15” is the most critical angle ply laminate
with the shear stress r,, dominating failure. It is interesting to note from these studies that as the
fiber angle is increased above 15”, the failure mode of (* 0), laminates shifts to mixed shear (T,,
and ryr), mixed shear and normal (T,,, 7Yzand az) and finally to transverse tension (Us). In fact
contribution to failure is dominated by transverse tension, v,, for 8 2 37.5”. The location of
initial failure also shifts from the f 8 interface to the midplane with increasing 8 i.e. for 132 45,
the failure is at mid plane.
At present it is difficult to make the comparison between two approaches, i.e. fracture
mechanics or strength based for the prediction of delamination onset. However, fracture
mechanics appears to be more accurate.

5.3. Growth of delamination


Delamination growth depends upon the stress state of the crack tip which is governed by the
mixed mode stress intensity factors Kr, Kn and Km or the strain energy release rates Gr, G,, and
Gm. It is not necessary that all three modes exist together. Only one or two modes may dominate
the fracture propagation. For example, the analysis by Wang[76] for edge delaminations at the
interface of angle ply (* 0) laminates show that Kn is considerably lower than Kr and Km (Table
3) and edge delamination is dominated by KIII i.e. the tearing mode. The relative magnitudes of
KI, KII and KIII depend on ply angle 8 as shown in Table 3. On the other hand, O’Brien’s[79]
analysis for the edge crack at the - 30/90 interface of [+ 30/ f 30/90/90], laminate show that
mode III contribution is negligible.

Table 3. Stress intensity factors K> for edge delamination in


(he), angle ply graphite/epoxy (T300/5208) laminate sub-
jected to uniform axial strain, eX[76].

15 0.095 0.01334 -5.009 -52.13


30 0.256 0.0366 -4.022 15.71
45 0.148 0.0152 -1.425 9.63
60 0.2225 0.00149 -0.1951 0.88
75 0.0069 -0.000324 0.0899 13.02

tKi (MPa &) are scaled by lo3 9.


Delamination 577

The growth of edge delamination is a stable fracture process in laminates subjected to


tensile loading[76,79,83], i.e. applied load has to be increased to force the delamination to
grow. Such a growth has been characterized by O’Brien[79] using the concept of crack growth
resistance (R-curve) curves whereby GR vs delamination growth curves were produced. The
initial value of GR represents the critical G, or the onset of delamination. O’Brien[79]
considered that G, is a material property and is independent of ply orientation. This is contrary
to the observations of Johannesson and Blikstad[l07] for delamination of angle ply laminates.
Their studies have shown that the G, is strongly dependent on the ratio KrIIfKl for angle ply
laminate and obtained a relationship in the form

G, 03Jl + &I/&)~. (6)

The mixed mode delamination growth is not observed to follow a single propagation law.
Various laws have been used by different investigators in order to fit their test data. The simplest
mixed mode delamination propagation law is defined as “the delamination growth occurs when the
total strain energy release rate GT (= Gi + Gn) reaches a critical value G,“, i.e.

GT= Gr+ Grr= G, (7)


where Gr and Grr are the mode I and II strain energy release rates, respectively.
However, eq. (7) may describe the mixed mode delamination growth only for the materials
when Grc = Grro. Rybicki et a!.[831 and O’Brien[79] have used eq. (7) to describe their mixed
mode delamination growth
For most of the graphite/epoxy composites, GrC < G rrc and thus eq. (7) may not adequately
describe the delamination growth. A more appropriate interaction relation to describe the
delamination growth is considered as

(~)~+(~)n= 1

where the exponents m and n have been found to have different values for different cases.
Johnson and Mangalgiri[108], and Jurf and Pipes[l09] showed that m = n = 1 agrees with
their test data for mixed mode delamination growth. Our recent studies[l35] have also shown
that the mixed mode delamination growth in T300/934 graphite/epoxy laminate follows eq. (8) if,
m = II = 1. On the other hand Law[88] has used eq. (8) with m = l/2 and n = 1 to fit their test
data for free edge delamination behavior in (&25/90), laminates. Whitcomb[27-291 considered
eq. (8) with m = n = 2 and eq. (6) as Gi = Gio to describe the delamination growth of a laminate
under uniaxial compression. Both these criterion lead to quite different results. However
Whitcomb has not verified his results with test data. Hahn[llO] has proposed another mixed
mode crack preparation law as

where g = GrolGrrc.
For composites with epoxy matrices g = 0.1-0.2. For negligible g, eq. (9) is similar to (8)
with m = l/2, n = 1, which fitted the test data law[88].
On the basis of morphology of fracture surfaces a geometrical fracture criterion was
suggested by Hahn and Johannesson[ 11 l] for mixed mode fracture as

Go = LYE
+ LYE
41-t (Kn/K,)i (IO)

where cyl may be considered to be the critical strain energy release rate due to fiber-matrix
debonding and a2 is related to the surface energy of the resin, -ym,by

a2 = 2YInU - 4 (11)

where Uf is the fiber volume fraction


578 AMAR C. GARG
The constants cq and o2 were determined to be 61 and 3.1 J/m2 for delamination of angle
ply laminates of T300/934 graphite/epoxy by Johannesson and Blikstad[l07].
Donaldson[l12,.113] compared various crack propagation laws for graphite/epoxy and
graphite/PEEK composites. He has shown that mixed mode delamination of graphite~epoxy can
be described by

Gc = -e-(CtM+cz) + C,
(12)

where

M = ‘f1+ ( Gil/ GI)&JE; (13)

and Ci = 0.25, C, = -6.31, C, = 503.3. EL and ET are, respectively, the longitudinal and
transverse moduli for the laminate. The delamination growth for graphite/PEEK was found to
follow the relation

G Gn 3’2
IC (GUC
) =-
1

(14)
G+--
Russell and Street[l14] also studied various crack propagation laws and demonstrated that
mixed mode delamination growth may be better represented by the Wu’s[ 1061 tensor polynomial
criterion. Russell and Street modified the original Wu’s criterion which was for the plane stress
condition to plane strain conditions which are more appropriate to delamination. The plane
strain model was found to fit the experimental data for mixed mode delamination fairly well.

6. EVALUATION OF ~NTERLA~INAR FRACTURE TOUGHNESS


In order to use the prediction methods for delamination growth and to compare the
performance of one material to another, it is necessary that a correct assessment of delamination
fracture toughness be made using the test methods. Until a few years back, the short beam shear
test method was the only interlaminar test commonly employed in materials characterization.
However, delamination is usually caused by interlaminar normal (peel) stresses. Therefore a test
method for mode I delamination was suggested by Wilkins et a1.[7] using a double cantilever
beam (DCB) specimen similar to DCB used for adhesive bond tests. Since then several
investigators[61,62,104,115-1261 have used such a specimen to characterize mode I
delamination behavior. The other types of specimens such as width tapered double cantilever
beam (WTDCB)]127,128], Arcan test fixture[l09,130] and free edge delamination (FED)[79]
have also been used. Amongst these, FED specimen proposed by O’Brien provides a viable
approach to determining delamination initiation as it is relatively simple test specimen.
However, FED tends to give higher G rc compared to DCB specimen[l04]. Because of the
complex stress state in the free edge zone, the FED specimen does not necessarily produce a
pure mode I delamination. Therefore DCB is considered as a most suitable to determine Grc.
To characterize the mode II delamination, several specimens such as cracked lap shear
(CLS)[7], Arcan test fixture[ 109,130] and end notched flexure (ENF)[114,131] have been
proposed. CLS specimen develops mixed mode with mode II as dominating but Arcan test
specimen and ENF are proposed as developing pure mode II behavior. However, Arcan test
fixture is more complicated and very thick specimen is required for this test. ENF appears to be
more suitable as specimen is same as DCB and test is simple three point bend test. A few of the
test specimens are described here.
In all these specimens, the aim is to measure critical strain energy release rate Gc, during
onset or extension of delamination. While in EDT specimen the increase in compliance at the
onset of delamination is determined by test, other specimens are calibrated by means of a
relation between G and compliance C[132].

,=ES (1%
2Waa
Delamination 579

where a is the crack length, W the width of the specimen, P the applied load and C is the
compliance defined by the slope of the load (P) vs displacement (6) curve for the specimen i.e.

c = 8/P. (16)

6.1. Double cantilever beam (DCB) specimen


A DCB (Fig. 28) is tested under displacement controlled conditions and P vs 6 curves are
obtained for various crack lengths (a) which are then used to obtain C and aC/aa. The value of
Gc may then be determined using the critical load PC for the initiation of delamination and the
values of X/&r in the eq. (15). Also, assuming DCB specimen to consist of two cantilever beams
joined at the end of crack tip, the compliance may be expressed as

&tC (17)
WEh3

where E is longitudinal modulus and h is the thickness of each beam. Substituting this into eq.
(15), one obtains

3P2C
G=-
2Wa’
(18)

The beam relation (17) can be used as long as the shear deformations can be ignored and the
deflections are small, otherwise the corrections for these should be applied[116, 1171. For details
of the data reduction for DCB specimen, one may refer to refs [115] and [116].

(a) Double cantilever beam (DCB) specimen

(Ehh

P P
t t
(b) Cracked lap shear (CLS) specimen

(c) End notched flexure (ENF) specimen

\______/
- -
______
I

(d) Free edge delamination (FED) specimen

Fig. 28. Schematic representation of various interlaminar fracture test specimens.

EFM 29:5-e
580 AMAR C. GARG

The crack growth for DCB is observed to be stable for graphite/epoxy (Dry) composites but
for graphite/PEEK material the propagation is usually unstable and follows the stick slip
mechanism[69]. Thus the crack growth in graphite/PEEK is highly rate dependent which
is attributed to plastic and viscoelastic effects in the process zone around the crack tip. Garg and
Ishai[ 1251 observed the similar stick slip mechanism even in a graphite/epoxy composite when
wet containing about 2% moisture by weight.

6.2. Cracked lap shear (CLS) ~~cimen


Using the specimen shown in Fig. 28, the P - 6 curves may be obtained for various crack
lengths and aC/aa determined. The substitution of PC and X/au in eq. (1.5) give the value of
Go. Gc may also be obtained using

where subscripts (1) and (2) refer to the sections indicated in Fig. 28(b). Since CLS gives the total
energy consisting of modes I and Ii, the individual components may be obtained by using finite
element analysis[7].

6.3. End notched flexure (ENF) specimen


The specimen is the same as the DCB but it is loaded in a three point flexure which results in
almost pure in-plane shear delamination growth by mode II. Russel and Street]1 141 give an
expression for the determination of Gut using the beam theory as

9P$a2C
Grc = Gw
2 W(2L3 + 3a3)

where PC is the critical load for delamination, W, L and a are defined in Fig. 28(c) and C is the
compliance given by

C = (2L” t- 3a3)18EhW (21)

where E is the flexural modulus and 2 h is the thickness of specimen.


Equations (20) and (21) were derived by Russell and Street[ll4] ignoring the influence of
interlaminar shear deformation and friction between the crack surfaces on Grr and C. Recently,
Carlsson et a1.[13 l] modified the beam theory solution to consider the influence of shear
deformation and frictional effects. Later Gillespie et a1.[133] refined the analysis of ENF
specimen using the finite element method including the effects of local shear deformation around
‘the crack tip (not accounted by beam theory). This resulted in higher Grr. These studies have
revealed that inclusion of shear deformation in the derivation of Gtr by beam theory improves
the results. Yet the discrepancy between finite element and beam theory solutions co&d be of
the order of 20-40% for typical graphite composites. Therefore they have suggested an im-
proved data reduction scheme which retains the simplicity of beam theory and accuracy of the finite
element method [ 1331.

6.4. Edge delamination tension (EDT) test


This specimen provides a viable approach to determine delamination initiation. The test
method has been well described by O’Brien[72]. It makes use of tensile specimens of the lay-ups
which are deIamination prone such as (*30~*30/~~)~. The point at which the linearity in stress
vs strain curve is deviated is considered to be the point of onset of delamination and thus Gc
may be obtained by eq. (2). Once again Go obtained by this test is the total G. The individual
components Gr and Grr may be obtained by a finite element analysis[79].
Though the EDT test is quite simple, one needs to know the delamination interface as (1
priori in order to analyse the test data. Also, the finite element analysis is required to obtain the
individual components of strain energy release rates i.e. Gr and Grr.
Johnson and Mangalgiri [ 1081 have compared the Gr and Gu obtained during delamination
Delamination 581

Laboratory environment

rv’ CLS

40-

ENF
4
0-, I
I I I I I ,,I I I I
0 40 80 120 160 2od’3orJ 600sw1200
G,,,Jlm’

Fig. 29. Interlaminar fracture toughness of T300/5208 composites. The data spread is the maximum
and minimum values [ 1081.

using various specimens for T300/5208 composite material as depicted in Fig. 29. According to
this figure, DCB and ENF appear to be most suitable for characterizing modes I and II
delamination behavior, respectively. The figure shows GIIc to be about 8-10 times Gro. On the
other hand, EDT and CLS can only give the total Ge for fracture. Obtaining Gre or Gut from
these specimens is rather difficult. The interpretation of the results for Go obtained by testing
CLS by Wilkins ec al. [7] and Garg and Ishai[l25] do not seem to be in order. These investigators
have considered Gnc = 0.75 Go, which is not appropriate. For example, Wilkins et a1.[7]
reported Gut = 154 J/m2 for T300/5208 whereas Fig. 29 shows it to be about 600-1200 J/m2 as
determined by Russell and Street [ 1141. In order to use CLS test data to evaluate Gut one should
use an appropriate mixed mode criterion such as eq. (8). If we use Wilkins’[7] test data for CLS
as Gi = 51 J/m2, Gu = 154 J/m2 (Gr = 0.75 GT), Gie = 88 J/m2 in to eq. 8 with m = l/2 and
n = 1, GIIc = 645 J/m2 which agrees with the value shown in Fig. 29 for ENF test. Similarly, Giro
for AS/3501-6 as obtained by various test methods also differs considerably. Wilkins[l26],
Russel and Street[ll4] and Jurf and Pipes[l09] have reported it to be 322, 420-500, 670 J/m2
using CLS, ENF and Arcan test respectively.
There is hardly any standard test method available to determine tearing mode fracture
toughness, i.e. Gmc in delamination. Wilkins et a1.[7] appear to have attempted to study mode
III behavior but their attempts have been unsuccessful. However, they have mentioned that for a
conservative estimate Gmc can be considered to be the same as Gno. Recently Ripling, Santner
and Crosley[l34] have used tapered double cantilever beam specimen with a scarf joint for the
study of modes I and III. By increasing the scarf angle they could achieve a higher percentage of
tearing. This appears to be a viable specimen to study mode III behavior. More recently,
Donaldson[l36] has successfully determined G mc by the use of DCB specimen loaded in
tearing mode for AS4/3502 graphite/epoxy laminates. He has designated the test specimen as
split cantilever beam (SCB) test.

CONCLUSIONS
In the preceding section, various aspects of delamination have been considered. Based upon
the discussions, the following conclusions are made.
(1) The delamination is an important damage mode which may not be visible on the surface
but is capable of affecting strength and stiffness.
(2) The delamination may not deteriorate in-plane tensile behavior appreciably but
compressive behavior is significantly affected.
(3) Root cause of delamination is poor interlaminar toughness. This may be improved by
development of tougher resins, through-thickness reinforcement of laminate or interleafing
graphite/epoxy layers by thermoplastic films.
(4) The prediction of onset of delamination requires extensive computations thus there is a
need to develop a simpler methods.
(5) various test methods considered show that modes I, II and III delaminations can be
studied by the use of DCB, ENF and SCB specimens, respectively.
582 AMAR C. GARG

REFERENCES

[l] M. D. Rhodes and J. G. Williams, NASA-TM 85748 (1984).


[2] A. M. James and E. Williams, Lockheed Horizon, pp. 31-43 (1986).
[3] G. Dorey, AGARD-LS 124, pp. 6.1-6.11 (1982).
[4] K. L. Reifsnider, E. G. Henneke, W. W. Stinchcomb and J. L. Duke, Proc. Znt. Union of Theor. Appl. Me&,
Blacksburg, Virginia, (1982). Pergamon Press, New York (1983) pp. 339-390.
[5] C. K. H. Dharan, J. engng Maw. Technol. 100, 233-247 (1978).
[6] T. K. O’Brien, NASA-Tm-85768 (1984).
[7] D. J. Wilkins, J. R. Eisenmann, R. A. Camin, W. S. Margolis and R. A. Benson, ASTMSTP 775, 168-183 (1982).
[8] C. T. Herakovich, J. compos. Mater. 15, 336-348 (1981).
[9] N. J. Pagan0 and R. B. Pipes, Znr. J. Mech. Sci. 15, 679-688 (1973).
[lo] I. M. Daniel, R. E. Rowlands and J. B. Whiteside, Expl. Mech. 14, l-10 (1974).
[I I] D. 0. Stalnaker and W. W. Stinchcomb, ASTM STP 674, 620-641 (1979).
[12] A. L. Highsmith, W. W. Stinchcomb and K. L. Reifsnider, ASTM STP 836, 194-216 (1984).
[13] T. K. O’Brien, NASA-TM-84592 (1983).
[14] C. T. Herakovich, Znt. J. Mech. Sci. 18, 129-134 (1976).
[15] F. W. Crossman and A. S. D. Wang, J. compos. Mater. 12, 2-18 (1978).
[16] G. L. Farley and C. T. Herakovich, ASTM STP 658, 143-159 (1978).
[17] P. W. Hsu and C. T. Herakovich, J. compos. Mater. 11, 422-428 (1977).
[18] G. L. Farley and C. T. Herakovich, Two dimensional hygrothermal diffusion into a finite width composite
laminate, VPI-E-77-20 (1977).
[19] M. W. Hyer, C. T. Herakovich, S. K. Milkovich and J. S. Short Jr, Composites 14, 276-280 (1983).
[20] C. T. Herakovich and D. M. Wang, Expl Mech. 17,409-414 (1977).
[21] J. G. Williams and M. D. Rhodes, ASTM STP 787,450-480 (1982).
[22] G. Dorey, Impact and crashworthiness of composite structures, in Sbuchcral Impact and Crashworthiness, Vol. I
(Edited by G. A. 0. Davies), pp. 155-192. Elsevier Applied Science, London (1984).
[23] S. M. Bishop, Compos. Strucl. 3, 295-318 (1985).
[24] G. Dorey, S. M. Bishop and P. T. Curtis, Compos. Sci. Technol. 23, 221-237 (1985).
[25] K. L. Reifsnider, E. G. Henneke, and W. W. Stinchcomb, Defect Property Relationship in Composite Materials,
AFML-TR-76-81(1979).
[26] K. L. Reifsnider and R. Jamison, Znr. J. Fatigue, 187-197 (1982).
[27] J. D. Whitcomb, ASTM STP 836, 175-193 (1984).
[28] J. D. Whitcomb, Compos. Sci. Technol. 25, 19-48 (1986).
[29] J. D. Whitcomb, J. compos. Mater. 15, 403-426 (1981).
[30] W. L. Yin, S. Sallam and G. J. Simitses, Proc. AZAAIASMEIAHSIASCE 25th Structures, Srrucrural Dynamics
and Mater. Conf., Palm Springs, CA, pp. 159-165 (May 1984).
[31] W. L. Yin and J. T. S. Wang, J. appl. Mech. 51, 939-941 (1984).
[32] G. J. Simitses, S. Sallam and W. L. Yin, AIAA J. 23, 1437-1444 (1985).
[33] W. L. Yin and Z. Fei, Mech. Res. Commun. 11, 337-344 (1984).
[34] W. L. Yin, Znt. J. Solids Struct. 21, 503-514 (1985).
[35] P. M. Mujumdar and S. Suryanaryayana, Proc. Znr. Conf. on Camp. Mater. Strucr. Madras, India (6-9 January,
1988) pp. 273-283.
[36] S. N. dhatterjee, R. B. Pipes and R. A. Blake Jr, ASTM STP 836, 161-174 (1984).
r371 D. Y. Konishi and W. R. Johnston. ASTM STP 674. 579-619 (1979).
[38j M. D. Rhodes, J. G. Williams and J. N. Starnes, Jr,’ Presented‘at thk 23rd SAMPE National Symposium and
Exhibition, Anaheim, CA (2-4 May 1978).
[39] J. G. Williams, M. S. Anderson, M. D. Rhodes, J. H. Starnes, Jr and W. J. Stroud, NASA, TM-80077 (1979).
[40] M. J. Shuart and J. G. Williams, AZAA J. 24, 115-122 (1986).
[41] J. M. Hopper, E. Demuts and G. Milizianto, Damage tolerant design demonstration. AZAA/ASME/ASCE/AHS,
25th Structures, Structural dynamics and Mater. Conf., Palm Springs, CA, pp. 15-20 (May 1984).
[42] R. L. Ram Kumar, ASTM STP 813, 116-135 (1983).
[43] C. S. Frame and G. Jackson, AGARD-CP. 355, pp. 21.1-21.17 (1986).
[44] R. B. Deo and M. M. Ratwani, Experimental Investigation of delamination initiation and propagation in
composite to metal stepped-lap joints. AZAAIASMEIASCEIAHS, 25th St. Structural Dynamics and Mater.
Conf., Palm Springs, CA, pp. 257-263 (May 1984).
[45] L. J. Hartsmith, Effects of flaws and porosity on strength of adhesive-bonded joints, McDonnel Douglas A/C Co.
Douglas Paper 7388 (1984).
[46] L. J. Hartsmith, Adhesively bonded joints for fibrous composite structures, McDonnel Douglas A/C Co., Douglas
Paper 7740 (1986).
[47] G. C. Grimes and E. G. Dusablon, ASTM STP 787, 513-538 (1982).
[48] N. J. Johnston, NASA-CP 2321, pp. 75-95 (1984).
[49] A. F. Yee and R. A. Pearson, NASA-CR. 3718 (1983).
[50] W. D. Bascom, R. L. Cottington, R. L. Jones and P. Peyser, J. appl. Polym. Sci. 19, 2545-2562 (1975).
[51] W. D. Bascom, R. Y. Ting, R. J. Moulton, C. K, Riew and A. R. Siebert, J. Mater. Sci. 16, 2657-2664 (1981).
[52] D. L. Hunston and W. D. Bascom, ACS Advances in Chemistry Series. No. 208, Rubber Modified Thermoset
Resins (Edited by C. Keith Riew and J. K. Gillham), pp. 83-99. Am. Chem. Sot. (1984).
[53] J. M. Scott and D. C. Phillips, J. Mater. Sci. 10, 551-562 (1975).
[54] C. B. Bucknall, Toughened Plastics. Applied Science Publishers, London (1977).
[55] A. J. Kinloch, S. J. Shaw, D. A. Tod and D. L. Hunston, Part I, Polymer 24, 1341-1354 (1983).
[56] A. J. Kinloch, S. J. Shaw, D. A. Tod and D. L. Hunston, Part II, Polymer 24, 1355-1353 (1983).
[57] D. N. Shah, G. Attalla, J. A. Manson, G. M. Conelly and R. W. Hertzberg, Am. Chem. Sot. Ser. 2m, 117-135

[58] gE$j, Douglas, P. W. R. Beaument and M. F. Ashby, J. mater. Sci. 15, 1109-1123 (1980).
Delamination 583

[59] A. C. C&g, Toughening Mechanisms in modified epoxy resins, Presented at ASME Winter Annual Meeting,
Anaheim, CA (December 1986).
[60] A. C. Moloney, H. H, Kausch and H. R. Stieger, J. mater. Sci. l&208-216 (1983).
1611 D. C. Leach and D. R. Moore, Compos. Sci. Technol. 23, 131-161 (1985).
[62] D. C. Leach, D. C. Curtis and D. R. Tamblin, ASTM STP 937 (1987).
[63] S. S. Yau, T. W. Chou and F. K. Ko, Composites 17,227-232 (1986).
[64] F. K. Ko and C. M. Pastare, ASTM STP 864,428-439 (1985).
[65] H. B. Dexter and J. G. Funk, Presented at AIAA/ASME/ASCE/AHS, 27th Structures, Structural Dynamics and
Mat. Conference, San Antonio, TX (19-21 May 1986).
[66] R. E. Evans, J. E. Masters and J. L. Courter, Advanced composites, Conf. Pmt., American Society for Metals,
Dearborn, Michigan, pp. 249-257 (2-4 December, 1985).
[67] R. Y. Kim, A technique for prevention of delamination. AFWAL-TR-82-4007, pp. 218-230 (1982).
[68] F. Erdogan and G. D. Gupta, Znt. J. Solids Struct. 7, 39-61 (1971).
[69] F. Erdogan and G. D. Gupta, Znr. J. Solids Struct. 7, 1089-1107 (1971).
[70] A. H. England, J. appl. Mech. 87, 631-636 (1985).
[71] M. Comninou, J. appl. Mech. 44,631-636 (1977).
[72] K. S. Parihar and A. C. Garg, Znt. J. Engng Sci. 14, 831-843 (1976).
[73] A. C. Garg, Znr. J. Engng Sci. 19, 1110-1124 (1984).
[74] K. S. Parihar and A. C. Garg, Engng Fracture Mech. 7, 751-759 (1975).
[75] J. R. Rice, J. appl. Mech. 87. 418-423 (1965).
[76] S. S. Wang, NASA-CR 165439 (1981). Also Published in J. compos. Mater. 17, 210-223 (1983).
[77] E. Altus, A. Rotem and M. Shmueli, J. compos. Mater. 14, 21-30 (1980).
[78] T. K. O’Brien, NASA-TM 85728 (1984).
[79] T. K. O’Brien, ASTM STP 775, 140-167 (1982).
[80] R. Y. Kim and S. R. Soni, Experimental and analytical Studies on the onset of delamination in laminated
composites, Univ. of Dayton Res. Inst. Dayton, Ohio, Rept. UDR-TR-83.40 (1983).
[81] T. Mohlin, A. F. Blom, L. A. Carlsson and A. Gustavsson, The Aeronautical Research Inst. of Sweden, FAA Rept.
FAA-TN-183.57 (1983).
J. D. Whitcomb, ~AS~-TM-86301 (1984).
E. F. Rybicki, D. W. Schmueser and J. Fox, J. compos. Mater. 11,470-487 (1977).
C. T. Herakovich, A. Nagarkar and D. W. O’Brien, Modem Developments in Composite Materials and S@ucrures,
(Edited by J. R. Vinson), pp. 53-66. ASME, New York (1981).
T. K. O’Brien and I. S. Raju, AIAA/ASME/ASCE/AHS, 25th Structures, St. Dynamics and Mat. Conf. Palm
Springs, pp. 526-536 (May 1984).
A. S. D. Wang and F. W. Crossman, J. compos. Mater. 11,92-106 (1977).
F. W. Crossman and A. S. D. Wang, J. compos. Mater. 12, 2-18 (1978).
G. E. Law, ASTM STP 836, 143-160 (1984).
S. S. Wang, ASTM STP 674,642-663 (1979).
A. S. D. Wang and F. W. Crossman, J. compos. Mater. 11, 300-312 (1977).
I. S. Raju and J. H. Crews, Jr, Comput. Struct. 14, 21-28 (1981).
A. D. Reddy, L. W. Rehfield and R. S. Haag, Influence of prescribed delaminations on stiffness controlled
behavior of composite laminates, Presented at ASTM Symposium on effects of defects in Comp. Mat. San
Francisco, CA (December 13-14 1982).
S. R. Soni and N. J. Pagano, Global local laminate variational model, AFWAL-TR-82-4028 (1982).
R. L. Spilker and S. C. Chou, J. compos. Mater. 14, 2-20 (1980).
K. S. Kim and C. S. Hong, J. compos. Ma&r. 20,423-438 (1986).
C. G. Shah, Proc. Works~p-cum-seminar on Deluminafions in Com~s~tes, Bangalore, pp. 271-294 (March
1987).
A. V. Krishnamurty and K. Vijaykumar, Proc. Workshop-cum-Seminar on delaminations in Composites,
Bangalore, pp. 271-294 (March 1987).
N. J. Pagan0 and R. B. Pipes, Znt. J. Mech. Sci. 15, 679-688 (1973).
B. T. Rodini and J. R. Eisenmann, Fibrous Composites in Structural Design (Edited by E. M. Lenoe) pp. 441-447,
Plenum Press, New York (1980).
N. J. Pagano, Znr. J. Solids Sfrucr. 14, 385-400 (1978).
N. J. Pagano, Znt. J. Solids Struct. 14, 401406 (1978).
J. M. Whitney, ASTM STP 521, 167-180 (1973).
F. W. Crossman, R. E. Mauri and W. John Warren, NASA-CR-3189 (1979).
J. M. Whitney and C. E. Browning, ASTM STP 836, 104-124 (1984).
R. J. Nuismer and J. M. Whitney, ASTM STP 593, 117-142 (1975).
E. M. Wu, Sirength and Fracture of C#m~sife~, Composite Materials, Vol. V (Edited by L. J. Broutman). pp.
191-247. Academic Press, NY (1974).
T. Johannesson and M. Blikstad, A Fractographic study of delamination process. Rep. from Linkopine. . _ Inst. of
Tech. Linkoping, Sweden (1984). - _ .
W. S. Johnson and P. D. Manealeiri. NASA-TM-87571 (19851.
R. A. Jurf and R. B. Pipes, J.ycokp: Mater. 16, 386-394 (198i).
H. T. Hahn, Camp. ‘Z’ec‘e.Reo. 5,26-29 (1983).
H. T. Hahn and T. Johannesson, Mechanical behavior of materials -IV, ZCM4, Vol. 1, Stockholm, pp. 431-488
(August 1983).
S. L. Donaldson, Composites 16,103-112 (1985).
S. L. Donaldson, Compos. Sci. Technol. 33-44 (1987).
A. J. Russell and K. N. Street, ASTM STP 876, 349-370 (198s).
Anon. Standard Tests for toughened resin composites, NASA RP 1092 (1982).
P. E. Keary, L. B. Ilcewicz, C. Shaor and T. Trostle, 1. Compos. Mater. 19, 1.54-177 (1985).
J. M. Whitney, C. E. Browning and W. Hoogsteden, J. Reinforced Plastics Compos. 1 (1982).
584 AMAR C. GARG

[118] W. S. Johnson and P. D. Mangalgiri, NASA-TM-87716 (1986).


[119] D. J. Nicols and J. P. Gallagher, J. Reinforced Plastics Compos. 2 (1983).
[120] D. F. Davitt, R. A. Schapery and W. L. Bradley, J. compos. Mater. 14, 270-285 (1980).
[121] K. S. Hahn and J. Koutsky, J. compos. Mater. 15, 371-388 (1981).
[122] K. S. Hahn and J. Koutsky, Composites 14, 67-70 (1983).
[123] F. X. Decharentany, M. Bethmont, M. Benzeggagh, and J. F. Chretien, ICM-4, Cambridge, England, pp.
241-248 (1979).
[124] F. X. Decharentany, J. M. Harry, Y. J. Prel and M. L. Benzeggagh, ASTM STP 836,84-103 (1984).
[125] A. Garg and 0. Ishai, Engng Fracture Med. 22, 413-427 (1985).
[126] D. J. Wilkins, A Comparison of the delamination and environmental resistance of a graphite/epoxy and graphite
bismallimide, Naval Air Systems Command, Washington NAV-GD-0031 (1981).
[127] S. M. Lee, J. compos. Mater. 20, 185-196 (1986).
[128] W. D. Bascon, J. L. Bitner, R. J. Moulton and A. R. Siebert, Composites II, 9-18 (1980).
[129] M. Arcan, Z. Hashin and A. Voloshin, Expl Med. 18, 141-146 (1978).
[130] L. A. Carlsson, J. W. Gillespie and A. J. Smiley, .I. compos. Ma&r. 20, 594-604 (1986).
[131] J. F. Knott, Fundamentals ofFracture Mechanics. Buttenvorth, London (1973).
11321 J. W. Gillespie, Jr, L. A. Carlsson and A. J. Smiley, Compos. Sci. Technol. 25, 1-15 (1987).
[133] J. W. Gillespie, Jr, L. A. Carlsson and R. B. Pipe< Comks. Sci. Technol. 24, 177-197 (1986).
r1341 E. T. Rinline. J. S. Santner and R. B. Croslev. J. mater. Sci. 18, 2274-2282 (1983).
i135] S. C. Joshi, Delamination in fiber reinforcedcomposite materials, M. Tech. Thesis; Aero. Engng Dept., I.T.T.
Bombay, India (December 1987).
[ 1361 S. L. Donaldson, Mode III interlaminar fracture characterization of composite materials, M.S. Thesis, University of
Dayton, Ohio, (April 1987).

You might also like