Low-velocity impact damage on dispersed stacking sequence laminates.

Part I: Experiments
C.S. Lopes
a,b
, O. Seresta
d
, Y. Coquet
c
, Z. Grdal
a
, P.P. Camanho
b,
*
, B. Thuis
e
a
Faculteit Luchtvaart-en Ruimtevaarttechniek, Technische Universiteit Delft, Kluyverweg 1, 2629 HS Delft, The Netherlands
b
DEMEGI, Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal
c
Institut Suprieur de lAronautique et de lEspace, 1 place Emile Blouin, 31500 Toulouse, France
d
ADOPTECH, Inc., Blacksburg, VA 24060, USA
e
Nationaal Lucht-en Ruimtevaartlaboratorium NLR, Voorsterweg 31, 8316 PR Marknesse, The Netherlands
a r t i c l e i n f o
Article history:
Received 10 October 2008
Received in revised form 5 February 2009
Accepted 5 February 2009
Available online 14 February 2009
Keywords:
A. Structural composites
B. Delamination
C. Finite element analysis (FEA)
C. Damage tolerance
Low-velocity impact
a b s t r a c t
The stacking sequence design of composite laminates is often limited to combinations of 0, 90, and 45
bre angle plies. Furthermore, in order to comply to certain stiffness requirements, clustering of plies
becomes unavoidable. Although such laminates might have the desired stiffness properties, they may
show poor impact and/or compression-after-impact behaviour.
A method to redesign the traditional stacking sequences such that the alternative laminates have
improved damage resistance whilst keeping similar in-plane and bending stiffness properties as their ori-
ginal traditional stacking sequences is proposed. This method makes use of optimisation tools based on
genetic algorithms. In the alternative laminates, the difference between bre angles of two consecutive
plies is maximised and allowed to vary in the 090 bre angle range at intervals of 5. Manufacturing of
such laminates is practical nowadays as the industry is changing its production techniques into accurate
automated bre-placement and tape-laying technologies. A two-step approach is proposed for the design
of laminates. In the rst step, the optimal laminate is designed in the traditional fashion to cope with the
expected quasi-static loads on the structure. The second step consists of redesigning this laminate to bet-
ter withstand impact loads by dispersing its stacking sequence while keeping similar stiffness properties
as in the rst step.
A traditional laminate and two dispersed stacking sequence alternative layups were tested under low-
velocity impact and compression-after-impact loads in order to compare their impact resistance and
damage tolerance characteristics. The evaluation of these laminates will also be carried out by the inno-
vative numerical tools proposed in the follow-up of the present paper.
2009 Elsevier Ltd. All rights reserved.
1. Introduction
In industrial practice the stacking sequence of laminates is often
limited to combinations of 0, 90 and 45 bre angle plies which
is inline with the limitations of traditional layup processes in
assuring a precise bre-placement. Furthermore, in order to com-
ply to certain directional stiffness requirements, clustering of plies
becomes unavoidable. This practice, in spite of being advantageous
due to its simplicity and readiness, can be inefcient in terms of
structural behaviour. Although a laminate might have good stiff-
ness properties, it may show a poor response to impact loads in
particular when plies with the same orientations are clustered to-
gether. The optimisation of composite laminate designs towards a
better impact damage resistance and tolerance is often overlooked
in favour of efcient in-plane, statically loaded designs. The
response to impact damage is, in general, not accounted for in
the early design phase but evaluated for those designs that meet
the static load requirements. More than often, there is margin to
improve the impact response of a laminate previously designed
to withstand in-plane loads in a optimal way, just by changing
its stacking sequence.
Experimental research on the damage response of composite
laminates has been carried by many authors (e.g. [5,14]). Cantwell
and Morton [6] made an extensive review of the research work on
impact damage up until 1991 and identied the fundamental
parameters determining the impact resistance of CFRPs. The effect
of varying bre orientations was also addressed. In particular, the
inuence of the stacking sequence on the impact response of lam-
inated composites has been studied by several authors. Dost et al.
[7] investigated the damage resistance and residual strength for
several quasi-isotropic laminates under low-velocity impact.
Post-impact compressive behaviour was found to be a strong func-
tion of the laminate stacking sequence. Strait et al. [8] carried
0266-3538/$ - see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compscitech.2009.02.009
* Corresponding author. Tel.: +351 225081753; fax: +351 225081315.
E-mail address: pcamanho@fe.up.pt (P.P. Camanho).
Composites Science and Technology 69 (2009) 926936
Contents lists available at ScienceDirect
Composites Science and Technology
j our nal homepage: www. el sevi er. com/ l ocat e/ compsci t ech
instrumented drop-weight impact tests on cross-ply, quasi-isotro-
pic and 0/45 bre angle based laminates. Stacking sequence was
found to have a signicant effect on the impact resistance, particu-
larly at higher impact energies. Fuoss et al. [9,10] studied the inu-
ence of three parameters on the impact damage resistance of
composite laminates: interface angle, ply orientation relative to a
xed axis and ply grouping. The guidelines given in their work
for a better damage resistance include the avoidance of ply cluster-
ing or small interface angles. It should be noted that the work of
Fuoss et al. [9,10] was based on transverse quasi-static loading
and not dynamic impact. In addition, their conclusions were based
on simplied FE models that were not able to accurately simulate
delaminations.
In the previous investigations, the stacking sequence of lami-
nates was changed with no regard for the changes in laminate stiff-
ness. Low-velocity impact events can often be approximated to
quasi-static loads. In such situations, the delaminated area is
highly dependant on the out-of-plane displacement of the plate
during impact [11]. This means that the bending stiffness plays
an important role on the way damage develops on an impacted
laminate. To avoid misinterpretations of the results, in this work
both the in-plane and the bending stiffness of the studied lami-
nates are maintained while redesigning the stacking sequence.
Using optimisation tools based on genetic algorithms (GA), alterna-
tives to the traditional 0, 90, and 45 bre angle based laminates
are designed where the plies are dispersed through the 090 bre
angle range at intervals of 5. These alternative laminates maintain
similar in-plane and bending stiffness properties to the baseline
from where they were derived. This procedure is possible since
in the design of composite laminates multiple optima exist i.e.
there is more than one stacking sequence that satises a given de-
sign criterion. It is also highly efcient due to use of lamination
parameters that represent the laminate stiffness properties instead
of ply orientation angles directly. Furthermore, manufacturing of
such laminates is practical nowadays as the industry switches from
hand laying processes to accurate automated bre-placement and
tape-laying technologies.
In laminates with dispersed stacking sequences, matrix crack
propagation through-the-thickness may be prevented because of
the increased bre bridging effect. Part of the energy dissipated
by matrix cracking in traditional, clustered laminates can be used
to load the much stronger bres which bridge matrix cracks in
neighbouring plies. Additionally, the extent of delaminations may
be reduced. Since delaminations do not usually propagate between
plies of the same orientation, the interlaminar shear stresses at
interfaces with clustered plies, which act as a single ply, are rela-
tively high. These have the potential to cause wide delaminations
that split the laminate into sublaminates which have a lower resis-
tance to buckling. This can greatly reduce the compression-after-
impact (CAI) strength of a panel which is the same as to say that
its tolerance to damage is compromised [12]. In laminates with
dispersed stacking sequence not only unidirectional layers are
thinner but also more interfaces with the potential for delamina-
tion become available. This means that the same impact energy
can be dissipated in more delaminations of controlled extension,
hence reducing the impact footprint and hopefully increasing the
CAI strength.
To improve the impact damage resistance and tolerance of com-
posite laminates, a two-step approach to the design of laminates is
proposed. In the rst step, the optimal laminate is designed in the
traditional fashion to cope with the expected quasi-static loads on
the structure. The second step consists of redesigning this laminate
to better withstand impact loads by dispersing its stacking se-
quence. This is done without compromising the initial stiffness
properties. This second stage of design will be cost efcient since
the candidate laminates will be those with known stiffness proper-
ties, thus minimising the number of designs for which impact test-
ing and/or progressive failure analysis is required.
In this work, the proposed stacking sequence dispersion meth-
od is rst applied on one traditionally designed laminate, resulting
in two alternative layups. The three congurations are build on
carbon/epoxy material by means of automated tow-placement.
The impact resistance of specimens laid according to each layup
is evaluated by analysing their response to drop-weight impact
loads of different energies. Then, their damage tolerance is ascer-
tained by conducting CAI tests. The results of such experiments
are presented and compared in this paper.
2. Stacking sequence dispersion
The basic idea of stacking sequence dispersion is to maximise
the difference between bre orientation angle of two successive
plies. For the purpose of numerical simulation, it is required to de-
ne dispersion in the stacking sequence of any given laminate. We
dene a numerical entity d
i
, between a pair of successive plies (i
and (i + 1)) having bre orientation angle h
i
and h
i+1
, to be given by
d
i

1
sin
2
h
i
h
i1
d

1
1 d
; 1
where d is a small number used only to make Eq. (1) stable even
when the two successive plies have the same orientation. In this
work, we assumed d = 0.0001. From the above equation, it is evident
that as the value of d
i
approaches zero, the bre orientation angle
difference or dispersion between the successive plies approaches
90. The qualitative dispersion of an n-ply laminate can be judged
from the numerical value of d which is given by
d
X
n1
i1
d
i
: 2
which is to be minimised.
2.1. Dispersed stacking sequence using genetic algorithms
The baseline laminate, i.e. the design with traditional 0, 90
and 45 layers is assumed to be known a priori. This baseline lam-
inate can come from any existing design optimised for some re-
sponse characteristic such as buckling or vibration. The idea is to
improve the damage resistance and tolerance of the base laminate
via stacking sequence dispersion. The stacking sequences gener-
ated are subjected to the following conditions:
The generated stacking sequences must have the same, or simi-
lar, in-plane and exure stiffness as the base laminate. All the
stiffness coefcients are linear combinations of eight lamination
parameters. Therefore, the objective is to minimise the differ-
ence between the base laminate and the dispersed stacking
sequence lamination parameters.
Both the dispersed stacking sequence and base laminate are
symmetric. In the proposed optimisation formulation, only half
of the laminate is considered and hence the symmetry require-
ment is automatically satised.
The dispersed stacking sequence must be balanced. This is
imposed via a penalty function approach.
The external two layers of the baseline and dispersed stacking
sequence laminates are xed to 45 following standard indus-
trial practice in counteracting eventual shear loads.
A standard GA is used to generate the dispersed stacking
sequence. The design space is discretised into 19 possible ply ori-
entations from 0 to 90 at intervals of 5. A "Fortran 90" GA
framework that was developed in an earlier research effort, specif-
C.S. Lopes et al. / Composites Science and Technology 69 (2009) 926936 927
ically for composite laminate design, is used in the current work
[13]. This framework consists of a GA module, encapsulating GA
datastructures, and a package of GA operators like crossover and
mutation. The module along with the package of operators consti-
tutes a standard GA. An integer alphabet is used to code the ply
genes. The implementation details of the GA modules and GA pack-
ages can be found in the work carried by McMahon et al. [13]. The
present objective function f is given by
f
X
i8
i1
W
i
W
b
i

2
Cd
!
P; 3
where the W
i
s refer to the lamination parameters of the candidate
designs and the W
b
i
s to the lamination parameters of the baseline
laminate. C is a bonus parameter associated with the dispersion d
of the candidate design. It is imperative to use a small value of C
so that the optimisation process is guided by the error function
and not by the dispersion measure of the candidate designs. P is
the penalty due to violation of the balanced laminate condition. A
negative sign is added because the GA used maximises the objective
function.
The optimisation algorithm developed attempts to keep the in-
plane and bending stiffness properties of the baseline laminate
selected. Additionally, the algorithm attempts to maximise the
global dispersion of the laminate by introducing a bonus factor
in the objective function rather than the maximisation of the bre
orientation difference between all adjacent plies. In practice, this
results in the modication of the bre orientation angles of clus-
tered plies.
Lamination parameters represent the laminate layup congura-
tion in a compact form [14]. The in-plane behaviour of composite
laminates in the classical lamination theory can be fully modelled
using only four lamination parameters regardless of the actual
number of layers:
W
1
; W
2
; W
3
; W
4

Z
1=2
1=2
cos2h

z; sin2h

z; cos4h

z; sin4h

zd

z
4
in which hz is the distribution of the bre orientation angle
through the normalised thickness z z=h (h is the total laminate
thickness). Another four parameter dene the bending behaviour:
W
5
; W
6
; W
7
; W
8
12

Z
1=2
1=2

z
2
cos2h

z; sin2h

z; cos4h

z; sin4h

zd

z
5
2.2. Baseline and alternative laminates
Specimens were cut out of laminated plates with three different
stacking sequences and tested. One constitutes the baseline cong-
uration and the other two have dispersed stacking sequences and
are the proposed alternative congurations. The plates were fabri-
cated with Hexply AS4/8552 carbon epoxy tows resulting in a
nominal ply thickness of 0.182 mm. The lamina properties of this
material are summarised in Table 1.
The baseline laminate constitutes an example of a laminate that
might be the result of the traditional industrial practice in design-
ing a laminate to comply with certain stiffness requirements. In
this work, it is assumed that such requirements have led to a sym-
metric and balanced 24-ply laminate having about 60% of plies in
the main loading direction (0), 10% of 90 plies to add transverse
stiffness and another 30% at 45 to counteract shear loads and im-
prove buckling resistance. Placing 45 plies on the surface of lam-
inates is also the best strategy to improve impact resistance [6].
Mimicking industrial practice in avoiding large interlaminar stres-
ses, the ply contiguity was limited to four plies. Thus, the described
stacking sequence dispersion technique is applied to the following
baseline laminate:
Baseline : 45=90=0=45=0
4
= 45=0
2

s
In spite of its good elastic performance, the damage tolerance of
the baseline laminate may be poor due to the relative high number
of 0 clustered layers allowing for an easy propagation of through-
the-thickness matrix cracks degenerating in wide delaminations at
interfaces with neighbouring plies due to the high interlaminar
shear stresses. The low number of interfaces can lead to wide del-
aminations necessary to dissipate the impact energy. The stacking
sequence dispersion procedure previously described results in the
following alternative layups:
Alternative 1: 45=0=70=70=0=15=10=10=15=15=15
s
Alternative 2: 45=80=5=20=20=10=80=10=5=15=15
s
For the sake of identication, these two laminates are termed
alternative 1 and alternative 2, respectively. These two congura-
tions were generated through two different runs of the dispersion
algorithm. To generate the alternative 1 laminate, a value of the
bonus parameter C equal to 10
17
was used in the objective func-
tion. The alternative 2 conguration was generated with C = 10
13
.
This means that there was more freedom to disperse the stacking
sequence when generating alternative 2.
The baseline laminate axial modulus, transverse modulus, shear
modulus and Poissons ratios are:
E
b
x
89:2 GPa; E
b
y
29:6 GPa; G
b
xy
15:2 GPa; m
b
xy
0:413
As for the laminates with dispersed stacking sequences, the
same engineering constants take the values:
E
a1
x
88:1 GPa; E
a1
y
30:4 GPa; G
a1
xy
15:3 GPa; m
a1
xy
0:406
and
E
a2
x
86:4 GPa; E
a2
y
34:0 GPa; G
a2
xy
14:8 GPa; m
a2
xy
0:352
respectively, for the alternative 1 and alternative 2 laminates. The
axial and shear modulii of the alternative laminates are within 3%
difference from the baseline. The transverse modulii are very simi-
lar between the baseline and alternative 1 but alternative 2 shows a
15% shift. A similar comparison can be made for the Poissons ratio.
This means that the alternative laminate 1 can be considered a
slightly better match to the baseline than the alternative 2. Overall,
the GA is successful in generating laminates having dispersed stack-
ing sequences but stiffness properties very similar to those of the
baseline.
3. Drop-weight impact tests
The experimental programme follows the standard test method
for measuring the damage resistance of a bre-reinforced polymer
matrix composite to a drop-weight impact event [15], as devised
by the American Society for Testing and Materials (ASTM). A sim-
ilar procedure is proposed by AIRBUS

[16]. The objective of the

test is the evaluation of the level of damage induced by an impact
event on a given laminated composite material plate congura-
tion. The specimen geometry and dimensions for this test are
sketched in Fig. 1.
Table 1
Hexply AS4/8552 ply properties.
Density 1590 kg/m
3
Elastic properties E
1
= 135 GPa; E
2
= 9.6 GPa; G
12
= 5.3 GPa; m
12
= 0.32; m
23
= 0.487
Strength (MPa) X
T
= 2207; X
C
= 1531; Y
T
= 80.7; Y
C
= 199.8; S
L
= 114.5
928 C.S. Lopes et al. / Composites Science and Technology 69 (2009) 926936
3.1. Specimen preparation and test procedure
Laminated composite panels, about 500 500 mm in size, were
fabricated with Hexply AS4/8552 carbon epoxy tows at the Dutch
National Aerospace Laboratory (NLR) by means of an Automated
Fibre-Placement Workcell from Automated Dynamics.
Curing was performed along a standard autoclave procedure. A
two-step 3 h cycle at 7 bar was followed. In the rst 1-h step a
temperature of 110 C was maintained and the second step took
2 h at 180 C. Non-destructive Inspection (NDI) by means of
through-transmission ultrasonics (C-scan) was carried on the
cured plates. C-scan results showed no delaminations or porosities
of signicative size. Overall, the achieved laminate quality was
higher than for traditional hand-laid laminates. Each panel was
cut in 15 test specimens with a water-cooled disk saw to within
a tolerance of 0.5% with respect to the standard dimensions. The
specimen thickness was measured at three specimen points and
determined to be of 4.37 mm on average with a standard deviation
of 0.7%.
The standard impact test consists of dropping a spherically
shaped steel dart on a free fall, guided by two rails, onto a
150 100 mm composite laminate specimen supported on all
sides in such a way that the specimen region free for an impact
is a rectangle of 125 75 mm in size, as shown in Fig. 1. The mass
of the impactor and the height at which it is dropped in a free fall
must be chosen to meet the impact energy required. According to
the standard procedure [15], four toggle clamps should be installed
along the longer specimen sides in order to prevent any in-plane
movement. In the present case, the clamps are replaced by a
10 mm thick steel plate with a 125 75 mm hole that completely
covers the supported side areas preventing movement in the out-
of-plane direction. In Fig. 1, the supported areas are drawn in grey.
Either the standard or the adopted procedures simulate the simply
supported conditions found in a typical structural composite panel.
The latest was found simpler to produce and mechanically more
reliable.
The drop height, H, is related with the impact energy, E
i
, with
H
E
i
m
i
g
where m
i
is the mass of the impactor and g is the accelera-
tion due to gravity (9.81 m/s). The impactor velocity at the mo-
ment of contact with the specimen, V
i
, is related with the impact
energy by E
i

1
2
m
i
V
2
i
. In this experimental programme, specimens
of the three different stacking sequence congurations were im-
pacted with energies varying between 5 J and 50 J, according to
the procedure proposed by AIRBUS

[16]. The test matrix is shown

in Table 2. The drop heights ranged from 77 cm to 168 cm and the
impactor mass was either 951 g, 1.331 kg, 2.441 kg or 4.186 kg, as
function of the impact energy required.
The specimens were impacted on the laminate tool side. A re-
bound impact on the specimens was prevented by means of a man-
ually controlled device. The impact velocity was calculated as a
function of the travelled time between a couple of sensors placed
close to the impact surface. The impact force as function of the im-
pact contact time, t, was recorded digitally by reading the output of
a load-cell installed inside the impactor. The indentation origi-
nated on the impact face of the specimens was measured once
immediately after each test and again after a relaxation time of
one week. The visible damage was characterised and the speci-
mens were again inspected to evaluate the extent of the delami-
nated projected area, commonly referred to as impact footprint.
The location of major delaminations through-the-thickness of the
specimens was identied by means of Fluorescent Penetrant
Inspection (FPI).
3.2. Results
With the goal of nding the differences between the impact re-
sponse of the three congurations tested, the data pertaining im-
pact force and dynamics as well as the resulting impact damage
is analysed in the next paragraphs.
3.2.1. Impact dynamics
In a typical low-velocity impact process, the initial impactor
velocity is gradually reduced as its movement is opposed by the
deforming composite specimen. This deceleration is associated to
a reaction force on the impactor. The kinetic energy is transferred
to the laminate and temporarily stored as elastic strain energy. If
the local strength of the material is reached, part of this energy
starts to be dissipated through irreversible damage. The impactor
velocity is eventually reduced to zero as the penetration reaches
a maximum. Gradually, the major part of the accumulated elastic
strain energy is transferred back to the impactor. The impactor
accelerates and is impelled away from the specimen, but the resti-
tuted energy is lower than the impact energy. Part of the accumu-
lated energy is kept in the form of panel vibrations and eventually
dissipated by damping. Another part corresponds to the energy
dissipated by material damage.
As an example, Fig. 2 shows the impactor reaction force histo-
ries during 30 J impacts on the baseline and alternative laminates.
For the remaining cases, the three designs compare in a similar
way. In the cases shown, the whole impact takes less than four mil-
liseconds, since contact is rst established until it is broken. The
impact on the alternative 2 laminate specimen takes slightly longer
than for the other congurations.
Harmonic oscillations are recorded from the beginning of the
impact event. These may be caused by dynamic coupling between
impactor, the specimen and the supports. Due to these oscillations,
differences between baseline and alternative laminates are difcult
to identify. In order to overcome this difculty, high order polyno-
mials where tted to the force vs. time history curves. These tted
curves show three different behaviours. All of them show a closely
linear part until about 0.3 ms of impact duration. Then, the curve
gradually acquires a lower slope and a second closely linear part.
Along this second part of the interpolating curve, the amplitude
of the oscillations in the actual force histories increases. The
change of slope is suspected to be caused by the start of delamin-
ations when the impact force reaches about 6 kN. The sudden
propagation of delaminations seems to amplify the dynamic re-
sponse, as experienced by Schoeppner and Abrate [2]. All curves
1
0
0
1
5
0
1
2
5
7
5
D=16
0
45
90
Fig. 1. Illustration of the devised impact test setup (dimensions in mm).
C.S. Lopes et al. / Composites Science and Technology 69 (2009) 926936 929
reach the maximum force value slightly before 1 ms of impact
time. After this point, the reaction force corresponding the baseline
laminate specimen drops to a plateau at 5 kN and, after 2.2 ms,
gradually decreases to zero as the impactor bounces back. For
the alternative laminate specimens, the behaviour is similar except
for the plateau level which is lower for the alternative 1 and higher
for the alternative 2. The plateaus indicate a sudden drop in stiff-
ness caused by a more severe damaged mechanism such as bre
breakage. The alternative 1 laminate behaviour denotes more dam-
age than the baseline as opposed to the alternative 2 laminate. Be-
tween 1 ms and 2.2 ms, corresponding to this third part of the load
history curve, the harmonic oscillations are slightly damped. How-
ever, the damping is more effective during the unloading phase.
By integrating the force history, the impactor acceleration (A),
velocity (V) and displacement (S) can be calculated, respectively
by [11]:
Vt V
i

Z
T
0
Atdt V
i
gT
1
m
i
Z
T
0
Ftdt 6
St S
i

Z
T
0
Vtdt 7
with V
i

2E
i
=m
i
p
. S
i
is the impactor position at the moment of
contact with the specimen and T is the total impact time.
By analysing the force vs. displacement behaviours on Fig. 3, the
actual stiffness of the specimens can be evaluated. Four distinct re-
gions can be identied by relying on the help of simple data inter-
polations: (i) a nonlinear loading, (ii) a plateau (or slight unloading,
in the case of the alternative 1 laminate), and then (iii) a nonlinear
unloading path. The nonlinearity in the loading path corresponds
mainly to nonlinear geometrical effects. Major delaminations seem
to start propagating when the impact force reaches about 6 kN. The
plateau on the force level does not reveal the start of bre break-
age, but the sudden decrease in the amplitude of the oscillations
is suspected to be related to the initiation of this damage mode.
Here, the alternative 1 specimen behaves differently from the other
two specimens by showing a negative stiffness. The maximum
impactor displacement on this specimen is also higher by almost
1 mm. Both things reveal that the alternative 1 specimen suffers
a higher damage level.
3.2.2. Non-destructive inspection (NDI)
The necessary data needed to support the arguments givenin the
previous section consists of information regarding the specimens
impact damaged areas. The extent of these was determined by NDI
based on ultrasonic C-Scan. The results of C-Scans on specimens im-
pacted at four energy levels are shown in Fig. 4. An attenuation of
6 dB is considered the threshold value for damage identication.
For a better comparison, the area exceeding the threshold attenua-
tion value for the baseline and both alternative congurations are
superimposed in Fig. 4a and b. The exceedance area represents a
rough estimation of the extent of damage, mostly delaminations,
in the impacted specimens, i.e. the impact footprint.
In general, the alternative 1 conguration shows a larger impact
footprint than the baseline conguration, for the impact energy
range used. On the other hand, if the alternative 2 is compared
to the baseline, the trend is opposite. The impact energy is plotted
on Fig. 5 as function of the measured exceedance area. An expo-
nential law, crossing the vertical axis at 5 J, is tted to the results.
As a rough estimation, 5 J was considered the threshold energy le-
vel for impact damage for all three congurations since impact
tests at this energy level did not produce any damage that could
be detected by ultrasonic C-Scan. In reality, this threshold value
is somewhere between 5 J and 10 J.
The trendlines on Fig. 5 indicate that, in general, the [45/80/5/
20/ 20/10/ 80/ 10/ 5/15/ 15]
s
(alternative 2) laminate
produces an 8% narrower impact footprint than the [45/90/0/45/
0
4
/ 45/0
2
]
s
(baseline) conguration but the [45/0/70/ 70/0/
15/10/ 10/ 15/15/ 15]
s
(alternative 1) laminate exceeds it
by 30%. From these results, it is clear that by dispersing the stack-
ing sequence of a laminate while keeping its elastic performance, it
is possible to improve its damage resistance. However, this result is
not achieved for every alternative conguration generated by the
current method.
Displacement [mm]
F
o
r
c
e

[
k
N
]
0 1 2 3 4 5 6 7
0
2
4
6
8
10
Baseline
B - Curve Fit
Alternative 1
Alternative 2
Delaminations
Matrix Cracking
Delaminations
Matrix Cracking
FibreFailure
Fig. 3. Impactor force vs. displacement for 30 J impacts on baseline and alternative
laminates.
Time [ms]
F
o
r
c
e

[
k
N
]
0 1 2 3
0
2
4
6
8
10
Baseline
B - Curve Fit
Alternative 1
A1 - Curve Fit
Alternative 2
A2 - Curve Fit
First Major Delaminations
Delaminations
Matrix Cracking
Delaminations
Matrix Cracking
Fibre Failure
Fig. 2. Impactor reaction force histories for 30 J impacts on baseline and alternative
laminates. The oscillatory behaviour is interpolated by high order polynomials.
Table 2
Test matrix for the drop-weight impact experimental programme. E
1mm
corresponds to the energy necessary to produce a 1 mm dent depth, considered to be the threshold for
Barely Visible Impact Damage (BVID).
Number of specimens 1 1 1 2 1 3 3 1 1
Target impact energy (J) 5 10 15 20 25 E
1mm
30 40 50
Drop-weight (kg) 0.95 1.33 1.33 2.44 2.44 2.44 2.44 2.44 4.19
930 C.S. Lopes et al. / Composites Science and Technology 69 (2009) 926936
3.2.3. Fluorescent penetrant inspection
The intraply damage (i.e. matrix cracking and bre failure) can
be evaluated by visual inspection of the specimens transversal sec-
tion at the impact location. However, there are obvious limitations
of inspecting a single section cut of an impacted specimen. Besides,
the lack of image contrast makes the identication and measure-
ment of delaminations difcult. To overcome these, a different
technique, the uorescent penetrant inspection (FPI), was adopted.
The FPI is a type of the known Liquid Penetrant Inspection tech-
nique and is based upon capillary action, where a low surface ten-
sion uid penetrates into the open cracks of a material. In the case
of FPI, the penetrant is applied by dipping the sectioned specimens
in a dye that uoresces when excited by ultraviolet radiation (also
known as black light). FPI is performed in a darkened environment,
and the excited dyes emit bright yellow-green light that contrasts
strongly against the dark background. In these conditions, a high
resolution picture of the specimen section is taken for later pro-
cessing. In order to obtain the required results, the correct type
of dye (on a sensitivity scale) and the adequate dipping time must
be chosen carefully. Also, the correct cleansing of the section sur-
face must be performed.
Within the present research, FPI was performed on three spec-
imens, one per layup conguration, previously impacted at an en-
ergy level of 30 J. In order to evaluate the actual extent of damage
not only layerwise but also in the plan of the laminate, 5 mm thick
strips of the specimens were cut parallel to the smaller specimen
axis and then inspected at both surfaces. As examples, Fig. 6 shows
pictures of FPI technique as applied at sections of the baseline and
alternative laminates close to the impact location. The location of
the widest delaminations is identied.
The extension of the larger delaminations was measured at each
inspected section. Then, the measurements were interpolated to
generate a contour plot of each major delamination. Fig. 7 shows
the superimposition of these delaminations for each of the
inspected specimens. For the baseline laminate specimen impacted
at 30 J, there are four major delaminations. The widest one has an
Fig. 4. Impact Footprint area exceeding an attenuation of 6 dB, as measured by non-destructive inspection by means of through-transmission ultrasonics. Comparison
between baseline (grey) and alternative (dark) laminate specimens for four different levels of the impact energy.
Exceedance Area [mm^2]
I
m
p
a
c
t

E
n
e
r
g
y

[
J
]
0 1000 2000 3000 4000
0
10
20
30
40
50
60
Baseline
Alternative 1
Alternative 2
Fig. 5. Impact footprint area as function of the impact energy. An exponential law is
tted to the discrete results with the 5 J energy level showing zero exceedance area.
C.S. Lopes et al. / Composites Science and Technology 69 (2009) 926936 931
ellipse-like shape with its major axis oriented at around 75 and is
located between plies 21(0) and 22(90). The second largest
delamination occurs at the interface between plies 19(0) and
20(45) and has a more elongated shape oriented at around 30.
21/22
19/20
15/16
14/15
9/10
5/6
21/22
3/4
9/10
20/21
19/20
20/21
19/20
21/22
16/17
7/8
7/8
16/17
21/22
Fig. 6. Section cut views of the three different laminate specimens impacted at 30 J. A uorescent penetrant liquid is applied for better identication of through-the-thickness
damage, namely the location and extent of delaminations. Major delaminations are identied by the neighbouring ply numbers.
Fig. 7. Superimposition of the major delaminations on baseline and alternative laminate specimens impacted at 30 J. Each delamination is identied by the neighbouring ply
numbers. Comparison is made with the C-Scan results (grey area).
932 C.S. Lopes et al. / Composites Science and Technology 69 (2009) 926936
There are still two other delaminations of relative large size, but
smaller than the previous. These are two elongated delaminations
along 0 and 40, one between plies 15(45) and 16(0) and the
other one between layers 14(0) and 15(45).
Regarding the alternative 1 laminate specimen impacted at 30 J,
only two major delaminations are observed: one between layers
19(0) and 20(70) oriented at 60, and the other one between
plies 20(70) and 21(70) oriented at around 60. The alternative
2 laminate specimen, loaded with the same impact energy, also
suffers from two major delaminations between layers, one square
shaped between layers 16(10) and 17(80) and the other one
between plies 21(5) and 22(80), elongated along the 60 axis.
There is also a smaller, but still relatively large, square shaped
delamination between layers 7(10) and 8(80).
From these observations, it is concluded that all major delamin-
ations occur at the interfaces close to the backface of each speci-
men. The plies most prone to delaminate are the ones with the
largest difference in bre angle. These are also the ones which de-
velop the highest interlaminar shear stresses. Also, in accordance
with previous investigations [11], delaminations develop, in gen-
eral, along an axis which orientation is close or equal to the bre
orientation of the underlaying ply.
The superimposition of major delaminations, as determined by
FPI, are compared in Fig. 7 with the results from the NDI presented
above. In all cases, the impact footprint area measured by the NDI
inspection is smaller than the one determined by the FPI. This
shows a known limitation of the C-Scan technique. Far from being
100% reliable, better results are achieved with the FPI. In spite of
the accuracy of each method, the results of the FPI are inline with
the ones of the NDI and presented in Figs. 4 and 5. Overall, these
reveal that the delaminations on the alternative 1 laminate speci-
men are wider than on the baseline whilst the ones on the alterna-
tive 2 laminate specimen are smaller. However, the post-impact
strength of each specimen certainly depends on the through-the-
thickness location of these delaminations, as well.
3.2.4. Indentation
The indentation on the impacted specimens, measured immedi-
ately after the tests, is plotted against the impact energy in Fig. 8.
For the same impact energy, the resultant dent is deeper on the
laminates with dispersed stacking sequences than on the baseline.
This indicates a higher concentration of matrix cracking and possi-
bly bre breakage around the impact zone in the alternative lam-
inates. The energy necessary to produce a 1 mm dent depth
(E
1mm
), considered to be the threshold for Barely Visible Impact
Damage (BVID), is interpolated at 26.4 J for the baseline laminate.
For the alternative laminates this value is approximately 25.5 J.
Therefore, for a given impact energy, damage is easier to identify
on the laminates with dispersed stacking sequence.
The increased depth of indentation of the redesigned laminates
due to impact is a bonus given the industry requirements for BVID
based on indentation depth. A wider range of impacts can be iden-
tied on a structure designed with dispersed stacking sequence
laminates leading to a higher readiness in structural maintenance.
4. Compression-after-impact tests
The standard test method for compressive residual strength prop-
erties of damaged polymer matrix composite plies [17], as devised
by the American Society for Testing and Materials (ASTM), was car-
ried on the previously impacted specimens. A similar procedure is
proposed by AIRBUS

[16]. This experimental procedure is com-

monly known as the compression-after-impact (CAI) test used by
many authors (e.g. [1,12,18,19]) to evaluate the residual post-im-
pact strength of composite laminates i.e. their tolerance to damage.
4.1. Test setup and procedure
The setup for the CAI test is illustrated in Fig. 9. A more com-
plete representation is available in the document dening the stan-
dard test procedure [17]. The previously impacted specimens are
loaded along their major axis by means of a compression loading
machine with parallel platens. In the present test programme, an
80 kN load range compression machine driven by four hydraulic
actuators was used. Being the only option, the edge displacement
was manually controlled. A rate lower than 0.5 mm/s was achieved
during all the tests.
A special purpose xture [17] was developed for the CAI tests. It
applies simply supported boundary conditions to the longer side
edges while clamping the horizontal edges (except for the horizon-
tal in-plane displacements). As shown in Fig. 9, two couples of
strain gauges are attached to each specimen in a back-to-back con-
guration. They are placed in opposite corners in order to avoid the
Impact Energy [J]
I
n
d
e
n
t
a
n
t
i
o
n

[
m
m
]
0 10 20 30 40 50
0
1
2
3
4
5
6
Baseline
Alternative 1
Alternative 2
Fig. 8. Indentation as measured immediately after specimen testing. Second order
polynomials are tted to the discrete results.
100
1
5
0

z
=

u
=
0
y=z=u=v=w=0
y z=u=v=w=0 y

z
=

u
=
0
y
x
z
25
3
5
SG3&4
SG1&2
Fig. 9. Compression-after-impact test setup (dimensions in mm). View from the
backface where bre splits are visible.
C.S. Lopes et al. / Composites Science and Technology 69 (2009) 926936 933
large bre splitting areas on the backface of the specimens as result
of the high energy impact events. The purpose of using the strain
gauges is to check for any out-of-plane bending, abnormal (non-
parallel) loading and to determine the maximum strains achieved
before specimen failure.
4.2. Results
Twelve out of the fourteen previously impacted specimens of
each conguration were compressed up to failure. One virgin
(non-impacted) specimen of each conguration was tested as well.
From the test matrix presented in Table 2, only two of the speci-
mens per conguration, corresponding respectively to 30 J and
E
1mm
, were not compressed after impact. The residual strength val-
ues and failure strains achieved are presented in the following
paragraphs.
4.2.1. Compressive residual strength
As observed previously, an impact event on a laminated plate
cause delaminations at several interfaces through its thickness.
These cause the subdivision of the laminate into several thinner
sublaminates which have lower bending stiffness than the original
plate. The compressive residual strength of the specimens tested is
generally determined by the resistance to buckling of the sublami-
nates rather than their in-plane strength. Even the compressive
strength values obtained for the virgin specimens do not corre-
spond to their failure under pure in-plane compression because
of plate instability before those values could be reached. While
the Hashin criteria [20] predict pure compression rst-ply failure
of the 0 layers at 1016 MPa, under test conditions none of the
specimens reached a value higher than 500 MPa.
The CAI strength as function of the impact energy, for all the
specimens tested, is represented in Fig. 10. Power-law curves are
tted to the results as a model of the residual strength degradation
of the laminates as function of the impact energy. In reality, an
asymptotic behaviour is expected as the compressive residual
strength of completely penetrated specimens does not vary with
impact energy. In the present cases, the asymptotes of minimum
CAI strength approach the 200 MPa level.
The specimens impacted at 5 J did not show any damage under
the scrutiny of the ultrasonic C-Scan. Their compression-after-im-
pact strength is, in spite of the expected results scatter, the same as
for the virgin specimens. Therefore, for low impact energies, there
is a plateau in the residual strength plot of each specimen corre-
sponding to its undamaged compressive strength, i.e. there is a
range of impact energies which cause no damage to the specimens.
In the absence of more data, the threshold impact energy to guar-
antee specimen damage is considered to be 5 J. If more rened data
was available, this value would be found to be somewhere be-
tween 5 and 10 J, possibly different for each of the three congura-
tions tested.
Further analysis of Fig. 10 reveals that the response behaviour
of the three laminate congurations is similar. Furthermore, there
is no consistent trend of superior performance of any of the alter-
native laminates as compared with the baseline. On the contrary,
the [45/80/5/20/ 20/10/ 80/ 10/ 5/15/ 15]
s
(alternative
2) laminate consistently underperforms the [45/90/0/45/0
4
/
45/0
2
]
s
(baseline) conguration by 10% to 20% even though alter-
native 2 had shown to have a smaller impact footprint compared to
the baseline design. Given the scatter in the results, the compres-
sive residual strength of the baseline and [45/0/70/ 70/0/15/
10/ 10/ 15/15/ 15]
s
(alternative 1) laminates cannot be
differentiated.
4.2.2. Failure strains
The designer of composite structures is often more interested in
knowing the failure strain of an impacted laminate rather than its
residual strength. With this simple information one can draw an
upper limit for the strain that a given structure should hold with-
out failure, supposing that it had suffered any single impact within
an expected range of impacts. Increasing the failure strains of a
structure is of utmost importance for the designers as this increase
can drive the reduction of structural weight.
The CAI failure strains for the tested specimens are represented
in the form of bars in Fig. 11. The strain results for the alternative 2
laminate slightly underperform the ones achieved for the baseline.
However, in general, the alternative 1 shows much better perfor-
mance than the baseline. While the minimum failure strain value
reached for the baseline conguration was about 2700 lstrain,
for the alternative 1 laminate this value is on the vicinity of
4600 lstrain, i.e. 70% higher in absolute value.
The previous observations are substantiated by the impact
damage results. As shown in Fig. 7, considerably sized delamina-
tions occur at several interfaces through-the-thickness of both
baseline and alternative 2 specimens as a result of impact events.
This divides each of these laminates into a few sublaminates. Im-
pacts on the alternative 1 laminate specimens on the other hand
seem to produce relatively large delaminations only on interfaces
19/20 and 20/21. This leaves one relatively thick sublaminate,
build up by plies 0 to 19, almost intact. This sublaminate carries
Impact Energy [J]
R
e
s
i
d
u
a
l

S
t
r
e
n
g
t
h

[
M
P
a
]
0 10 20 30 40 50 60
0
100
200
300
400
500
600
Baseline
Alternative 1
Alternative 2
Fig. 10. Residual compression-after-impact strength of the three specimen cong-
urations as function of the impact energy. Power-law curves were tted to the
results after the residual strength plateaus.
Impact Energy [J]
S
t
r
a
i
n

[
-
]
0 10 20 30 40 50
-0.008
-0.007
-0.006
-0.005
-0.004
-0.003
-0.002
-0.001
0
Baseline
Alternative 1
Alternative 2
Fig. 11. Strains reached at failure during the compression-after-impact tests. The
values shown are the average of the strains measured at the impact face.
934 C.S. Lopes et al. / Composites Science and Technology 69 (2009) 926936
most of the compressive load on the alternative 1 specimens.
While the sublaminates on the other specimens fail quite abruptly,
this one seems to buckle before failure, hence allowing higher
strains. This is corroborated by the load vs. strain plots on Fig. 12
for specimens impacted at 30 J. The buckling of the alternative
specimen is evidenced by the sudden divergence of measurements
made by back-to-back strain gauges SG3 and SG4 before failure.
5. Discussion
As opposed to what was expected in the design phase of this
study, the experiments carried out showed no clear improvement
in the impact performance of the dispersed stacking sequence lam-
inates [45/0/70/ 70/0/15/10/ 10/ 15/15/ 15]
s
(alternative
1) and [45/80/5/20/ 20/10/ 80/ 10/ 5/15/ 15]
s
(alterna-
tive 2) over the chosen traditional conguration [ 45/90/0/45/
0
4
/ 45/0
2
]
s
(baseline), even though there are discernable differ-
ences in various response quantities.
The alternative 2 laminate appears to offer an improved impact
resistance, as compared to the baseline, measured by the smaller
impact footprint on the specimens. For the alternative 1 laminate
on the other hand the opposite occurs. According to a general rule
of thumb in the impact damage eld [11], the wider the damaged
area, the lower the compressive residual strength. However, the
CAI tests results show a reversed trend for the congurations stud-
ied, i.e. the damage tolerance of the alternative 1 laminate speci-
mens is similar or higher than for the baseline, while for the
alternative 2 layup the opposite is observed.
A closer look at the through-the-thickness damage distribution
(Fig. 7) of the impacted specimens reveals the reasons for such
unexpected behaviour. The impact footprint shown by the alterna-
tive 1 laminate specimens is constituted mainly by the superimpo-
sition of only two wide delaminations between plies close to their
backface. This leaves an almost intact stiff sublaminate capable of
holding relatively high CAI loads before buckling. On the baseline,
but mostly on the alternative 2 laminate, the delaminations resul-
tant of the impact loads, some of which close to the laminate sym-
metry plane, subdivide the specimens in several sublaminates with
less resistance to buckling.
The evidence of more concentrated damaged regions around the
impact locations as well as the larger impact indentations on the
dispersed stacking sequence laminates (Fig. 8) are signs that the
mechanism of bre bridging is playing a somewhat more promi-
nent role in the prevention of spreading of intralaminar damage
in these laminates than in the traditional conguration. This might
be good news for residual tension strength. However, the residual
compressive strength of laminates after impact is a stronger func-
tion of a distinct damage mechanism: interlaminar damage or del-
aminations. As a rule of thumb, the narrower the delaminations, the
less affected the laminate buckling resistance is, therefore the high-
er its residual strength. However, the through-the-thickness loca-
tion of the delaminations is also very important, as already
discussed for the alternative 1 laminate which has wider delamin-
ations occuring more towards the backface of the specimens. How-
ever, testing reveals that this is not enough to improve substantially
its compressive residual strength over the baseline conguration,
although it allows higher failure strains.
The simple strategy of dispersing a traditional laminate stacking
sequence while maintaining its in-plane and bending stiffness does
not lead automatically to higher performance congurations under
impact events. The main reason for this seems to be the high bre
angle differences allowed between consecutive layers in the alter-
native laminates. These easily delaminate due to the high interlam-
inar shear stresses, specially if they are located close to the
backface of the impacted specimens. On the other hand, interface
angle differences of 5 are common on the alternative laminates.
Such a small bre angle difference between plies is probably not
enough to promote a signicant bre bridging effect and not en-
ough to trigger delaminations.
Delaminations are, in any case, unavoidable and actually nec-
essary to dissipate part of the impact energy. The strategy to keep
them under controlled size is to spread them among the most
laminate interfaces as much as possible and prevent high inter-
laminar shear stresses caused by large differences in bre angles
between neighbouring plies. Therefore, constraints regarding
maximum and interface angle differences should be imposed to
the method of designing stacking sequence dispersed laminates.
However, these requirements may lead to an empty solution
space if the lamination parameters are to be matched exactly. A
relaxation in the requirement to keep the elastic stiffness un-
changed might be necessary. Also, for an improved CAI response,
extra control on the through-the-thickness location of delamina-
tions is needed.
6. Conclusions
In the present study, a new stacking sequence design method is
proposed. It is shown that it is possible to disperse the stacking se-
quences commonly used on composite applications in such a way
that no neighbouring layers have the same bre orientation angle
while the resultant laminate still keeps similar axial and bending
stiffness properties.
A traditional laminate was compared with two alternative con-
gurations generated according to the method proposed in terms
of impact performance. This was done by means of drop-weight
impact and compression-after-impact experiments. In general,
the results show no clear improvement in terms of impact resis-
tance or impact damage when the dispersed laminates are com-
pared with the baseline conguration. However, there are
indications that the strategy might work if constraints in the bre
angle difference between neighbouring plies and in the through-
the-thickness location of delaminations are imposed.
The impact performance of new dispersed stacking sequence
designs is evaluated in a follow-up paper [21] by means of reliable
numerical simulations. The correct numerical simulations of the
impact and compression-after-impact tests will provide insight
on the impact damage phenomena, besides being a valuable tool
to replace part of the physical tests by reliable computational tools
in the evaluation of candidate laminates.
Strain [-]
L
o
a
d

[
k
N
]
-0.006 -0.005 -0.004 -0.003 -0.002 -0.001 0
0
50
100
150
Baseline
Alternative 1
Alternative 2
SG4 SG3
Fig. 12. Strains measured by SG3 and SG4 during the compression-after-impact
tests on specimens impacted at 30 J.
C.S. Lopes et al. / Composites Science and Technology 69 (2009) 926936 935
Acknowledgements
The funding of this research through the scholarship SFRH/BD/
16238/2004 from the Portuguese Foundation for Science and Tech-
nology (FCT), and the participation of the Dutch National Aero-
space Laboratory (NLR) in the experimental programme are
gratefully acknowledged. The nancial support of the FCT under
the project PDCT/EME-PME/64984/2006 is acknowledged by the
5th author.
The specimens tested in this research were fabricated at the
NLR facilities in Noordoostpolder using the Automated Fibre-Place-
ment Workcell from Automated Dynamics under the technical
supervision of Mr. Wilco Gerrits. The uorescent penetrant inspec-
tion was also conducted at NLR with the much appreciated help
of Mr. Jacco Platenkamp. The help of Mr. Hans Verheim in the
conduction of the CAI experiments is gratefully acknowledged as
well.
References
[1] Zhang X, Davies G, Hitchings D. Impact damage with compressive preload and
post-impact compression of carbon composite plates. Int J Impact Eng
1999;22:485509.
[2] Schoeppner GA, Abrate S. Delamination threshold loads for low velocity
impact on composite laminates. Compos Part A: Appl Sci Manuf 2000;31:
90315.
[3] Sutherland LS, Soares CG. The effects of test parameters on the impact response
of glass reinforced plastic using an experimental design approach. Compos Sci
Technol 2003;63:118.
[4] Sutherland LS, Soares CG. Impact behaviour of typical marine composite
laminates. Compos Part B: Eng 2006;37:89100.
[5] Vlot A. Low-velocity impact loading on bre reinforced aluminium laminates
(arall) and other aircraft sheet materials, Ph.D. thesis. Delft (The Netherlands):
Delft University of Technology; October 1991.
[6] Cantwell W, Morton J. The impact resistance of composite materials a review.
Composites 1991;22:34762.
[7] Dost E, Ilcewiz L, Avery W, Coxon B. Effects of stacking sequence on impact
damage resistance and residual strength for quasi-isotropic laminates. In:
OBrien T, editor. Composites materials: fatigue and fracture, vol. 3. American
Society for Testing and Materials; 1991. p. 476500 [aSTM STP 1110].
[8] Strait L, Karasek M, Amateau M. Effects of stacking sequence on the impact
resistance of carbon ber reinforced thermoplastic toughened epoxy
laminates. J Compos Mater 1992;26(12):172540.
[9] Fuoss E, Straznicky P, Poon C. Effects of stacking sequence on the impact
resistance in composite laminates part 1: parametric study. Compos Struct
1998;41:6777.
[10] Fuoss E, Straznicky P, Poon C. Effects of stacking sequence on the impact
resistance in composite laminates part 2: prediction method. Compos Struct
1998;41:17786.
[11] Abrate S. Impact on composite structures. Cambridge (England): Cambridge
University Press; 1998.
[12] Sierakowski RL, Newaz GM. Damage tolerance in advanced composites. Basel
(Switzerland): Technomic Publishing AG; 1995.
[13] McMahon MT, Watson LT, Soremekun G, Grdal Z, Haftka RT. A fortran 90
genetic algorithm module for composite laminate structure design. Eng Comp
1998;14:26073.
[14] Grdal Z, Haftka RT, Hajela P. Design and optimization of laminated composite
materials. New York (NY): John Wiley and Sons, Inc.; 1998.
[15] Standard test method for measuring the damage resistance of a bre-
reinforced polymer matrix composite to a drop-weight impact event, Tech.
rep. West Conshohocken (PA, USA): American Society for Testing and
Materials (ASTM), ASTM D 7136/D 7136M-05; 2005.
[16] Fibre reinforced plastics-determination of compression strength after impact,
Tech. rep. Blagnac (France): Airbus Test Method (AITM), AITM1-0010; 2005.
[17] Standard test method for compressive residual strength properties of damaged
polymer matrix composite plies, Tech. rep. West Conshohocken (PA, USA):
American Society for Testing and Materials (ASTM), ASTM D 7137/D 7137M-
05; 2005.
[18] Sanchez-Saez S, Barbero E, Zaera R, Navarro C. Compression after impact of
thin composite laminates. Compos Sci Technol 2005;65:19119.
[19] Freitas M, Reis L. Failure mechanisms on composite specimens subjected to
compression after impact. Compos Struct 1998;42:36573.
[20] Hashin Z. Failure criteria for unidirectional bre composites. J Appl Mech
1980;47:234329.
[21] Lopes CS, Camanho PP, Grdal Z, Maim P, Gonzlez EV. Low-velocity impact
damage on dispersed stacking sequence laminates. Part II: Numerical
simulations. Compos Sci Technol 2009, doi:10.1016/j.compscitech.
2009.02.015.
936 C.S. Lopes et al. / Composites Science and Technology 69 (2009) 926936