Available online at www.sciencedirect.

com

Composite Structures 83 (2008) 413–428

www.elsevier.com/locate/compstruct

Failure analysis of low velocity impact on thin composite

laminates: Experimental and numerical approaches
a,*
Volnei Tita , Jonas de Carvalho b, Dirk Vandepitte c

a
Department of Materials, Aeronautical and Automobilist Engineering, Engineering School of São Carlos, University of São Paulo, Av.
Trabalhador São-Carlense 400, 13566-590 São Carlos, SP, Brazil
b
Department of Mechanical Engineering, Engineering School of São Carlos, University of São Paulo,
Av. Trabalhador São-Carlense 400, São Carlos 13566-590, SP, Brazil
c
Department of Mechanical Engineering, PMA Division, Katholieke Universiteit Leuven, Belgium

Available online 26 June 2007

Abstract

The dynamic behavior of composite laminates is very complex because there are many concurrent phenomena during composite lam-
inate failure under impact load. Fiber breakage, delaminations, matrix cracking, plastic deformations due to contact and large displace-
ments are some effects which should be considered when a structure made from composite material is impacted by a foreign object. Thus,
an investigation of the low velocity impact on laminated composite thin disks of epoxy resin reinforced by carbon fiber is presented. The
influence of stacking sequence and energy impact was investigated using load–time histories, displacement–time histories and energy–
time histories as well as images from NDE. Indentation tests results were compared to dynamic results, verifying the inertia effects when
thin composite laminate was impacted by foreign object with low velocity. Finite element analysis (FEA) was developed, using Hill’s
model and material models implemented by UMAT (User Material Subroutine) into software ABAQUSTM, in order to simulate the fail-
ure mechanisms under indentation tests.
 2007 Elsevier Ltd. All rights reserved.

Keywords: Composite laminates; Low velocity impact; NDE; Indentation tests; Finite element analysis

1. Introduction should change the material and/or architecture of the com-

During the last years, criteria of automobile and aircraft
projects have been more and more rigorous for component
developed in order to absorb impact energy. Research on
the development of structural components with high crash-
worthiness has been carried out not only by the automobile
and aeronautics industries, but also by naval, trains and
elevators industries. The project concept for structural
components with high crashworthiness depends on the
crash resistance concept described by Kindervater and
Georgi [1]. The crash resistance concept is based on the
energy absorption capacity and structural integrity. For
developing a project that reaches these requirements, it
*
Corresponding author. Tel.: +55 16 3373 8612; fax: +55 16 3373 9590.
E-mail address: voltita@sc.usp.br (V. Tita).
ponent. However, changes in the architecture can cause
increase of costs and/or of weights, reducing the perfor-
mance of the structure. The weight increase is not attractive
for the aircraft development, because it reduces the aircraft
performance. Nowadays, many kinds of components have
been made using composite materials, because these mate-
rials can absorb a high amount of impact energy and can
guarantee the survival of the passengers.
However, the dynamic behavior of composite laminates
is very complex, because there are many concurrent phe-
nomena during composite laminate failure under impact
load. Fiber breakage, delaminations, matrix cracking, plas-
tic deformations due to the contact and large displacements
are some effects which should be considered when a struc-
ture made from composite material is impacted by a for-
eign object. Therefore, it is very common to find research

0263-8223/$ - see front matter  2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2007.06.003
414 V. Tita et al. / Composite Structures 83 (2008) 413–428

works about this issue at the literature, for example: Oñate

et al. [2]; Cairns and Lagace [3]; Farley and Jones [4]; Haug
and De Rouvray [5]; Belingardi et al. [6]; Collombet et al.
[7]; Gottesman and Girshovich [8]; Vicente et al. [9]; Kin-
dervater et al. [10]; Kostopoulos et al. [11]; Lopresto and
Caprino [12]; Naik and Shrirao [13]; De Morais et al.
[14]; Mitrevski et al. [15]; Mikkor et al. [16]; Mamalis
et al. [17]; Kim and Chung [18]; Zhao and Cho [19].
This work shows an experimental investigation and
numerical results of the low velocity impact on thin com-
posite laminates plates of epoxy resin reinforced (matrix)
by carbon fiber (reinforcement). This study describes the
technology to manufacture the specimens and the set-ups
of equipments to make dynamic experimental tests (using
a drop-tower instrumented) and to realize non-destructive
evaluation (NDE) (using ultrasonic C-scan), as well as
the finite element model and UMAT subroutine imple-
mented into software ABAQUSTM to simulate the failure
mechanisms of the composite laminate. Afterwards, several
issues are discussed:

• Experimental results as load–time, displacement–time
and energy–time histories of the laminate plate impacted
under low velocity, considering different stacking
sequence and different impact energies;
• Experimental results as images from non-destructive
evaluation (NDE) of laminate damaged after impact
test, using NDE by ultrasonic C-scan technique;
• Experimental results as load–displacement of the lami-
nate under indentation test (quasi-static loading), con-
sidering the same stacking sequence used during
impact test and the same maximum force level reached
at impact tests;
• Comparison between experimental results from impact
and indentation tests;
• Comparison between indentation tests results and finite
element model results, using Hill’s model and two mate-
rial models implemented by UMAT.
2. Composite failure mechanisms
Composite laminate structures were made from the
stacking of plies, which contains a polymeric matrix rein-
forced by fibers. Therefore, composite laminate shows
two failure modes:
1. Intra-ply failure mode: damages at fibers, polymeric
matrix and/or interface between fibers and matrix.
2. Inter-ply failure mode: delaminations between plies.
The intra-ply damage at fibers is shown by mechanism 4
(Fig. 1) that is the fiber rupture. However, the fiber failure
mode depends on the type of loading, because compression
loads can induce micro-buckling, but tensile loads can
induce rupture of fibers. The intra-ply damage at the
matrix depends on the ductility of the polymer, as well as
Fig. 1. (a) Composite failure mechanisms: intra-ply and inter-ply failures
and (b) Intra-ply damages (by Anderson [20]).
on the in-service temperature. Thus, the polymeric matrix
can present a fragile or a plastic behavior (mechanism 5).
Other intra-ply failure mechanisms are shown in Fig. 1.
Mechanism 1 is called ‘‘Pull-Out’’ and occurs when the
interface between fiber and matrix is weak. So, the fiber
is pulled out of the matrix after the debonding mechanism
(mechanism 3) occurs. If the interface between fiber and
matrix is strong, the fiber is not pulled out of the matrix
and mechanism 2 called ‘‘Fiber Bridging’’ occurs.
The inter-ply failure called delamination occurs, in
many cases, after intra-ply damages, i.e., the evolution of
intra-ply damages propagates the delaminations, because
the regions damaged at the ply propagates when the load
increases and the cracks at two adjacent plies (with differ-
ent orientation angle) join for creating a discrete failure
between them. At that moment, the interlaminar shear
increases strongly and the delamination process initiates.
This failure mechanism is very common to occur during
impact events. According to Abrate [21], there is a charac-
teristic shape for the plate thickness when the laminate suf-
fers impact loading. For thick laminates, the damage
occurs at the outer plies, which contact to the impacter,
because the structure is less flexible. The damages at those
plies occur due to the high contact stress values and prop-
agate to the inner plies, promoting delaminations. The del-
aminations increase the flexibility of the structure and new
intra-ply damages occur close to the delaminations. The
intra-ply damage evolution increases the interlaminar shear
that promotes new discrete damage between adjacent plies,
forming a ‘‘pine tree’’ shape at the plate thickness. This
progressive failure process stops when the impact energy
is insufficient to activate any failure mechanism. For thin
 V. Tita et al. / Composite Structures 83 (2008) 413–428 415

laminates, the damages occur at the opposite plies that con- Temperature Pressure Vacuum
tact to the impacter, because the structure is more flexible. 4 0.0
The damage at those plies occurs due to the high flexural 120

stress values and propagate to the inner plies, promoting

delaminations. The delaminations increase the flexibility 100 3 -0.2

Pressure [bar]
of the structure and new intra-ply damage occurs close to

Vaccum [bar]
Temperature [°C]
the delaminations. The intra-ply damages evolution 80
-0.4
increases the interlaminar shears that promotes new dis- 2
crete damage between adjacent plies, forming a ‘‘reverse 60
pine tree’’ shape at the plate thickness. This progressive -0.6
failure process stops when the impact energy is insufficient 40 1
to activate any failure mechanism. Besides, according to
Abrate [21], the delaminated plies show the damage shape 20
-0.8

like a ‘‘peanut shape’’ oriented in the same direction of the 0

fibers in each ply. 0 30 60 90 120 150
time [minutes]

3. Failure analysis approaches Fig. 2. Cure cycle.

3.1. Experimental analysis

American Society for Testing and Materials (ASTM)
standards were followed to manufacture the specimens
and to realize the experimental tests. It is important to note
that the fabrication of the specimens and the experimental
tests was executed at Leuven Composites Processing Cen-
tre (LCPC) of Katholieke Universiteit Leuven (Belgium).
The prepreg M10 (epoxy resin reinforced by carbon fiber
unidirectional) from Hexcel was used for manufacturing
of the composite plates.
After the hand-lay-up process used to stack the plies, the
composite plate was put in the auto-clave with vacuum sys-
tem set to 0.8 Bar (0.08 MPa). According to Hexcel,
the complete cure cycle for M10 occurs when this material
is processed at 120 C, under a pressure with range from
0.3 to 5.0 Bar for 60 min (Fig. 2). After the curing process,
the composite plates were cut on square shapes (length and
width of 120 mm and thickness of 1.8 mm), using diamond
saw in order to guarantee the tolerances specified by
standards.
For the impact tests under low velocity, drop-tests were
realized following the specifications by FD Method
described on the ASTM D5628-96 [22] and using a drop-
tower. Table 1 shows the values for dimensions of square
specimen, as well as, impact mass (m), height of falling
impact mass (h), impact velocity (mo) and impact energy
(Ei). The specimens were fixed by two circular steel disks
that had a hole with a diameter of 80 mm. Thus, in fact, cir-
cular composite plates with boundary clamped by two cir-
cular steel disks attached (using four bolts under torque
equaling 27 Nm) were evaluated under impact conditions
specified in Table 1. The geometry of the aluminium dart
(hemisphere; radius equal to 8 mm) with a mass of
0.0513 kg is according to ASTM D5628-96 [22] (FD
Method) summarised in Table 1. The support of mass
and the load cell of the drop-tower with the aluminium dart
have a total mass of 1.205 kg shown in Table 1. The instru-
mented drop-tower has an optic sensor fixed at the base
and a LED (‘‘light emitting diode’’) attached to the support
of the mass, which permits to measure the displacement of
the impactor support in function of the time (si(t)) shown in
Fig. 3. Thus, numerical methods to derive the displacement
measured were applied in order to obtain the velocity (mi(t))
and acceleration of the support during the impact event.
On the other hand, the load cell measures the force during
the impact event (Fexp(t)) shown in Fig. 3. The load cell was

Table 1
Specifications for drop-tests
Stacking sequence Length (mm) Thickness (mm) m (kg) h (m) mo (m/s) Ei (J)
[0]10 120 1.8 1.205 0.5 3.13 5.91
120 1.8 1.205 0.2 1.98 2.36
[0/90/0/90/0]s 120 1.8 1.205 0.5 3.13 5.91
120 1.8 1.205 0.2 1.98 2.36
[+45/45/+45/0/90]s 120 1.8 1.205 0.5 3.13 5.91
120 1.8 1.205 0.2 1.98 2.36
120 1.8 2.205 0.5 3.13 10.82
120 1.8 2.205 0.2 1.98 4.33
Impactor material Mass (kg) Nose shape Radius of impactor (mm)
Aluminium 0.0513 Hemispherical 8.0
416 V. Tita et al. / Composite Structures 83 (2008) 413–428

Fig. 3. Drop-tower at LCPC of Katholieke Universiteit Leuven: kinematic energy (impact energy level), absorbed energy and elastic energy.

plugged on a KistlerTM amplifier (model 5007) that sends
signal to the computer. In the computer, there is a system
for data acquisition with 11 bits, three channels for input
data and sampling frequency set to 19 kHz. Thus, the
acceleration of the load cell was obtained dividing the force
measured by the impact mass. Numerical methods to inte-
grate the acceleration measured were applied in order to
obtain the velocity and displacement of the load cell during
the impact event. In fact, not only the dimensions of the
support and numerical methods to derive, but also the
measurements of the load cell and numerical methods to
integrate produce errors on the calculated displacement,
velocity and acceleration of the dart. Therefore, in order
to minimize the errors, both measures (from load cell and
from support) were used to determine the displacement,
velocity and acceleration of the dart.
If an object with mass m impacts a composite plate with
a velocity mo, the impact energy of the impacter Ei can be
expressed by
mv2o
Ei ¼ : ð1Þ
2 amount and the type of failure mechanisms activated will
Besides, the kinematic energy KE(t) transferred from
object to the target (composite laminate plate) can be
expressed by
2
mv2o mðvi ðtÞÞ
KEðtÞ ¼  ;
2 2
where the velocity of the impacter mi(t) can be obtained by
Z • Type of fiber and/or matrix used for manufacturing of
1 t
vi ðtÞ ¼ vo  F exp dt: ð3Þ
m 0
The experimental impact force Fexp is measured during
the impact event. Thus, it is possible to evaluate the impact
energy, which reaches the composite plate, as well as the
absorbed energy and the elastic energy (Fig. 3).
The absorbed energy could be understood as ‘‘released
energy’’, because the failure mechanisms activated during
the impact event release energy. However, the literature
considers these release energies as a fraction of impact
energy, which is absorbed by the structure and is not trans-
formed on elastic vibrations (elastic energy).
Fig. 3 shows that the absorbed energy from measure-
ments can be calculated using
KEðtÞ ¼ Ee þ Ea ; ð4Þ
where KEi is the kinematic energy; Ee is the elastic energy
and Ea is the absorbed energy.
In other words, the impact energy E(t) transferred from
the object to the target (composite laminate plate) is
absorbed by the failure mechanisms activated. Thus, each
failure mechanism, i.e., intra-ply failures and/or delamina-
tions absorb a fraction of impact energy. Therefore, the
affect the absorbed energy values. However, the amount
and the type of failure mechanisms activated depend on
some factors:
• Mass and velocity of the impacter (impact energy level);
ð2Þ
• Geometry of the impacter;
• Geometry of the structure;
the composite structure;
• Stacking sequence of the plies.
 V. Tita et al. / Composite Structures 83 (2008) 413–428 417

For this work, the influence of the impact energy level
and the stacking sequence was verified.
The indentation tests were realized using a universal
machine equipment in order to produce a quasi-static load-
ing under a velocity of 10 mm/min on the specimens with
the same geometry and the same boundary conditions used
for drop-tests. Thus, each circular composite plate was
loaded from zero to the maximum force value reached at
respective drop-test. After that, the load was removed till
zero under a velocity of 10 mm/min.
According to Abrate [21], non-destructive evaluation
techniques can be used to determine the principal failure
mechanisms that occur when a composite structure is
impacted. However, the destructive evaluation techniques
are used in order to verify more details about failure
mechanisms. In fact, non-destructive and destructive tech-
niques are used in conjunction to improve the information
quality, increasing the accuracy of inspection results, but
this approach also increases the costs. In this work, only
the NDE by ultrasonic C-scan technique was used,
because this technique is applied on large scale by aero-
nautic industries. The C-scan equipment was set with
the frequency of 5 MHz and set to cover a square area
(length is equal to 30 mm) of the composite plate
impacted by the dart, using the increment for scanning
of 298.4 lm.
Some C-scan equipment can be used to inspect the
extension of damage at a specific layer. Thus, it is possible
to determine with accuracy the location and the extent of
the delaminations. However, the ultrasonic C-scan equip-
ment at LCPC of Katholieke Universiteit Leuven (Bel-
gium) did not offer that tool. Therefore, the image
captured is a superposition of the damaged regions on
each layer, i.e., all failure mechanisms are projected in
the same plane. Despite this limitation, it was possible
to identify the principal failure mechanisms at composite
plates impacted.
3.2. Description of material models
First, a material model, defined as Material Model I,
was developed in order to simulate only the intra-ply dam-
age mechanisms, using laminated shell elements and plane
stress state. After that, a more complex material model,
defined as Material Model II, was developed in order to
simulate the intra-ply and inter-ply damage mechanisms,
using laminated solid elements and 3D stress state. Both
material models were implemented as a UMAT subroutine
into software ABAQUSTM [23].
Table 2 shows the Material Model I used finite element
analysis with shell elements. It is important to note that a
failure criteria associated to a type of failure constitutes
the Material Model. Table 3 shows an explicit definition
for all symbols.
The failure criteria for the Material Model I is based on
works published by: Yamada and Sun [24]; Hashin [25,26];
Chang and Chang [27,28].
According to the failure (fiber or matrix failure), there
is a degradation law in order to reduce the material prop-
erties of the failure ply based on the works published by:
Chang and Chang [27,28]; Chang et al. [29]; Chang and
Lessard [30]. For the matrix and fiber failure, E22 and
m12 are reduced to zero. However, E11 is not reduced,
considering the matrix failure, but this property and

Table 2
Material Model I for shell element
Material Model I
Failure Criteria Type of failure Degradation law
 2     
r1 eF > 1 ! damage Fiber compression A
¼ e2F Edf
11 ! E11 exp  H
Xc eF 6 1 ! no damage Ao
   
A
Gdf
12 ! G12 exp  H
 2 Ao
r1 ð2r212 =G12 Þ þ ð3ar412 Þ Fiber tension
þ ¼ e2F ; i:e:; Edf
22 ! 0
XT ð2S 212 =G12 Þ þ ð3aS 412 Þ
 2  mdf
12 ! 0
r1 F1 eF > 1 ! damage
þ ¼ e2F
XT F2 eF 6 1 ! no damage

 2 " 2 # 
r2 YC r2 F 1 eM > 1 ! damage Matrix compression Edm
11 ! E11
þ 1 þ ¼ e2M
2S 23 2S 23 YC F2 eM 6 1 ! no damage Edm
22 ! 0

mdm
12 ! 0
 2 
r2 F1 eM > 1 ! damage Matrix tension
þ ¼ e2M
YT F2 eM 6 1 ! no damage
418 V. Tita et al. / Composite Structures 83 (2008) 413–428

Table 3 more for Eq. (5). Thus, it is not verified that the Young
Definition for the symbols of Material Models I and II modulus reduces suddenly, which is very reasonable for
Symbol Definition indentation tests.
r1 Component stress acting at fiber direction However, G12 degradation law is also updated according
r2 Component stress acting at normal to the fiber to the non-linear behavior of the shear stress at planes 1–2.
r3 Component stress acting at normal to the ply plane Thus, G12 is the updated shear modulus, which is calculated
r12 Shear stress acting at the ply plane
r13 Shear stress acting transverse to the ply plane
using the tangent of the curve shear stress–shear strain:
r23 Shear stress acting transverse to the ply plane 1
XT Strength value parallel to the fiber under tension G12 ¼ 1
; ð6Þ
XC Strength value parallel to the fiber under compression G12
þ 3ar212
YT Strength value normal to the fiber under tension
YC Strength value normal to the fiber under compression where a represents the value provided by the stress tensor
ZT Strength value normal to the ply plane under tension component of the fourth order shown by Hahn and Tsai
S12 Strength value for shear at the ply plane [32].
S13 Strength value for shear transverse to the ply plane
S23 Strength value for shear transverse to the ply plane
Table 4 shows the Material Model II used for simulation
eF Failure index for the fiber finite element analysis with solid elements. It is important
eM Failure index for the matrix to note that a failure criteria associated to a type of failure
edela Failure index for delamination constitutes the Material Model. The Failure Criteria
E11 Young modulus at fiber direction applied is based on works published by: Hashin [25,26];
E22 Young modulus at normal to the fiber
m12 Poisson’s ratio
Liu et al. [33]; Kermanidis et al. [34]; Dávila and Camanho
df Modulus degradeted after fiber damage [35]; Dávila et al. [36]; Kostopoulos et al. [11]. Table 3
dm Modulus degradeted after matrix damage shows an explicit definition for all symbols.
dela Modulus degradeted after delamination According to the failure (delamination, fiber or matrix
failure), there is a degradation law in order to reduce the
material properties of the failure ply based on the works
G12 are reduced for fiber failure, based on an exponential published by: Chang and Chang [27,28] and Kermanidis
decaying et al. [34]. For the matrix and fiber failure, intra-ply failure,
   
df A the degradation law is the same applied for the Material
P ¼ P exp  H ; ð5Þ Model I. For delamination, inter-ply failure, the degrada-
Ao
tion factors are the following:
where P is the property to be reduced, A is the area of the
damage zone; Ao is the area of the interaction zone of the d dela
1 ¼ d dela
2 ¼ d dela
3 ¼ d dela
4 ¼ d dela
5 ¼ 0:01: ð7Þ
fiber failure that was obtained from Chang and Chang
[27,28] and from Shahid and Chang [31]. Finally, H is a Finally, Table 5 shows the material data used during the
parameter, which controls the level of the property degra- finite element analysis.
dation and it is obtained using reverse analysis.
Fig. 4 shows the E11 degradation law for Chang and 3.3. Finite element analysis
Chang [27,28] and for the present work, where b is the Wei-
bull distribution parameter. The parameter H is similar to The finite element analysis, using UMAT subroutines,
parameter b, but the degradation phenomenon is slightly was realized to evaluate the performance of Material
Model I and II to predict the composite structure behavior.
Numerical results as force–displacement graphs were com-
pared to experimental results from indentation tests.
E11 Chang-Chang
β =2.0 Material Model I (Table 2) and II (Table 4) were imple-
1.00E+011
β =3.5 mented by UMAT subroutine into software ABAQUSTM
β =5.0 [23], because some failure criteria adopted for these material
8.00E+010 β =8.0
Present work models were used by other researchers, but the combination
6.00E+010 H=0.5 of the failure criteria is a new investigation and the degrada-
H=1.0 tion law proposed in Eq. (5) is a new investigation, too.
E11 [Pa]

H=1.5
The stacking sequences investigated are shown in Table
df

4.00E+010
1, in order to evaluate the influence of the fiber orientation
2.00E+010 on the composite structure behavior, and, to evaluate the
material models performance for predicting the composite
0.00E+000 structure behavior.
First of all, a synthesis of the contact problem will be
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
shown. After that, details of the finite element models will
A/A o
be shown, like the geometry, the mesh, and boundary con-
Fig. 4. E11 degradation laws. ditions and mainly the material model.
 V. Tita et al. / Composite Structures 83 (2008) 413–428 419

Table 4
Material Model II for solid element
Material Model II
Failure Criteria Type of failure Degradation law

      
r1 eF > 1 ! damage Fiber compression A
¼ e2F Edf
11 ! E 11 exp  H
XC eF 6 1 ! no damage Ao
   
A
Gdf
12 ! G12 exp  H
 2  2  Fiber tension Ao
r1 r13 F1 eF > 1 ! damage
þ þ ¼ e2F Edf
22 ! 0
XT S 12 F2 eF 6 1 ! no damage
V df
12 ! 0

" 2 #
1 2 YC ðr2 þ r3 Þ 1 Matrix compression Edm
11 ! E11
ðr2 þ r3 Þ þ 1 þ 2 ðr223  r2 r3 Þ
4S 223 2S 23 YC S 23 Edm
22 ! 0
 2 
r13 F1 eM > 1 ! damage V dm
12 ! 0
þ þ ¼ e2M
S 12 F2 eM 6 1 ! no damage

 2 Matrix tension
1 2 1 2 r13
ðr2 þ r3 Þ þ 2 ðr23  r2 r3 Þ þ
Y 2T S 23 S 12

F1 eM > 1 ! damage
þ ¼ e2M
F2 eM 6 1 ! no damage

 2  2  2  Delamination dela
r3 r13 r23 2
edela > 1 ! failure Edela
33 ! d 1 E 33
þ þ ¼ edela
ZT S 13 Z 23 edela 6 1 ! no damage Gdela dela
23 ! d 2 G23
dela
Gdela
13 ! d 3 G13
dela
V dela
23 ! d 4 V 23
dela
V dela
13 ! d 5 V 13

Table 5 ABAQUSTM provides a type of contact model called ‘‘expo-

Material data used for simulations
nential contact’’, where the contact forces are activated
Elastic properties Strength values when the hemisphere penetrates the disk for a distance of
E11 127 GPa XT 1400 MPa Co, providing a pressure Po. For this work, Co was taken
E22 = E33 10 GPa XC 930 MPa to be 104 mm and Po was 102 N/m2.
G12 = G13 5.4 GPa YT 47 MPa
G23 3.05 GPa YC 130 MPa
For the finite element analysis, two types of finite ele-
m12 = m13 0.34 ZT 62.3 MPa ment models were developed. The first model was created
m23 0.306 S12 = S13 53 MPa using shell elements (S4) with four nodes and four integra-
S23 89 MPa tion points [23] in order to simulate the plane stress at the
laminate. The nodes at the edge of the model were fixed in
order to simulate the circular disk clamped (Fig. 5a). The
For the contact problem, it is very important to evaluate second model was created, using solid elements (C3D8)
the work produced by contact forces Fc on the contact sur- with eight nodes and four integration points in order to
face Sc in the equilibrium equations. Thus, during each simulate the 3D state of stress at laminate. The solid had
iteration i, the contact algorithm of software ABAQUSTM three elements per thickness and the nodes at the edge of
verifies the contact occurrence in order to evaluate the con- the model were fixed (Fig. 5b). It is important to note that
tact forces Fc. In this work, a circular disk was used in a for the shell model, as well as for the solid model, the exter-
clamped set-up under a load applied at the centre, using nal surface of the indention hemisphere was simulated as a
a hemisphere, which contacts the composite disk. Software rigid surface, which contacted the central region of the
420 V. Tita et al. / Composite Structures 83 (2008) 413–428

Fig. 5. Finite element models: (a) Shell element S4 and (b) Solid element
C3D8.
disk. There was a node associated to the rigid surface where
the boundary conditions were imposed in order to simulate
the displacement at the z-direction of the hemisphere in
two steps (normal to the disk). For the first step, the load-
ing phase was simulated, i.e., the hemisphere was displaced
from the initial position (very close to the disk) to the final
position (maximum displacement verified at the experimen-
tal test). For the second step, the unloading phase was sim-
ulated, i.e., the hemisphere was displaced from the initial
position, which was equal to the final position of the first
step, to the final position, which was equal to the initial
position of the first step.
Two material models were implemented as UMAT sub-
routine into software ABAQUSTM. Model I was used for
the shell element model (Table 2) and Model II was used
for the solid element model (Table 3). Fig. 6 shows that
the software ABAQUSTM manages all calculation proce-
dures, and the user can recover components of the stress
tensor for each iteration i. After that, the stress tensor for
each integration point is read by the UMAT subroutine
in order to be used in a failure criteria. If failure does not
occur then the updated stress tensor is equal to the original
stress tensor. Thus, this tensor comes back to the principal
calculus procedure in order to form the internal forces vec-
tor, which will be compared to the external forces vector.
The result of this comparison will produce residues (R) that
will be compared to a specified tolerance. If R is higher
than the tolerance, a new iteration will be necessary.
However, if failure occurs then a degradation law is
applied on the elastic properties of the ply, according to
the type of failure (Tables 2 and 3). Therefore, the consti-
tutive tensor must be corrected and the stress tensor will
be updated before to return to software ABAQUSTM. Then,
R is calculated with the internal and external forces vector
in order to be compared to the threshold that is set to indi-
cate convergence. If there is convergence, then a new incre-
ment of load occurs, otherwise, the software ABAQUSTM

New Increment

BEGIN

Calculus of Stress Tensor

Equilibrium at iteration i
for the integration point

Aplication of Calculusof Residual (R)

Damage Criteria

Yes
R < Tolerance?
No
Failure? Original Stress Tensor
No
Yes
Yes
Aplication of the Convergence?
Updated Stress Tensor
Degradation Law
No

Corrected END
Constitutive Tensor
Subroutine UMAT ABAQUS ®

Fig. 6. UMAT subroutine and software ABAQUSTM.

 V. Tita et al. / Composite Structures 83 (2008) 413–428 421

checks the convergence. During the iterations for only one a 2500
load step, it is expected that R decreases. If R increases, Stacking Sequence: [0]10
then the solution produces divergence, and the procedure
Ei =5.91J
stops. Frequently, the divergence occurs, because the mate- 2000
rial properties have high degradation, and the structure
does not have enough stiffness to support the loads
1500
imposed. Other critical parameter, which can cause diver-

Force [N]
gence in non-linear finite element analysis, is the time step
applied on the simulations. For the initial time increment, a
1000
value of 0.05 was used, which would be modified as
required, because the automatic time stepping scheme Ei =2.36J
was set on. For the time period of the step, 7 was used. 500
For the minimum and maximum time increment allowed,
0.0001 and 0.5, respectively were chosen. Finally, for the
maximum number of increments in a step, 100 was chosen. 0
0 1 2 3 4 5 6 7
time [ms]
4. Results and discussion
b
First, the impact test results are shown (force–time, dis- 6
placement–time and energy–time graphics) and images
from ultrasonic C-scan technique. Second, the indentation 5
test results are shown, comparing the force–displacement
graphs between quasi-static and impact loading. Finally, 4
the finite element results are shown, comparing the force–
E(t) [J]

Impact Level = 5.91J

displacement graphs between numerical and experimental 3 Impact Level = 2.36J
results.
The stacking sequences investigated are shown in Table 2
1, in order to evaluate the influence of the fiber orientation
on the composite structure behavior, and, to evaluate the
1
material models performance for predicting the composite
structure behavior.
0
0 1 2 3 4 5 6 7 8
4.1. Impact test results and NDE images time [ms]

Fig. 7. Experimental impact results for laminate [0]10.

4.1.1. Experimental impact results for laminate [0]10
Fig. 7a shows force–time graphics for composite plates
with stacking sequence [0]10, considering two impact energy
levels (5.91 J and 2.36 J). There is a region with oscillations
of high frequency (in the initial phase of contact from 0 to
1 ms) due to initial contact between specimen and dart at
force–time graph (Fig. 7a). From 2 ms, there are some
oscillations with more intensity that show damage
occurred, not only for specimens under Ei of 5.91 J, where
maximum impact force is 2250 N, but also for specimens
under Ei of 2.36 J, where maximum impact force is
1300 N. Due to the stacking sequence [0]10, the principal
failure mechanism is matrix rupture, slightly reducing the
global stiffness of the structure. Therefore, there is no
abrupt drop of the impact force value after the progressive
damage process initiation (Fig. 7a). The contact time
depends on the impact energy level; so, the contact time
is longer (6 ms) for specimens impacted by impact energy
of 2.36 J than by impact energy of 5.91 J. Thus, for high
contact time, the mechanical behavior of the specimen
tends to quasi-static response.
Fig. 7b shows the energy–time graph for composite
plates with stacking sequence [0]10, considering the impact
energy of 5.91 J. Based on Fig. 3 and Eq. (4), it is verified
that specimens have absorbed energy equal to 4.42 J, so
only 25% of the impact energy is converted to elastic
vibrations.
However, the specimens impacted by 2.36 J have just
1.2 J for absorbed energy (Fig. 7b), converting 50% of
the impact energy to elastic vibrations. Thus, the specimens
impacted by 2.36 J show less failure mechanisms. This
observation can be confirmed, using the images from ultra-
sonic C-scan technique.
Fig. 8 shows the ultrasonic C-scan images for composite
plates with stacking sequence [0]10, considering the impact
energies of 5.91 J and 2.36 J. Specimens impacted by 5.91 J
show damage more concentrated close to the impact region
(Fig. 8a). However, specimens impacted by 2.36 J show
damage more distributed at the matrix represented by
cracks oriented to the fibers (Fig. 8b). Besides, it was
observed that some fibers fail at the opposite side of the
impact for the specimens impacted at 5.91 J. This observa-
tion can explain why these specimens absorbed 75% of the
impact energy, where the failure mechanisms of fibers
422 V. Tita et al. / Composite Structures 83 (2008) 413–428

a 3500 Stacking Sequence: [0/90/0/90/0]s

3000 Ei =5.91J

2500

2000

Force [N]
1500

Ei =2.36J
1000

500

0
0 1 2 3 4 5
time [ms]

b
6

Fig. 8. Experimental impact results for laminate [0]10: (a) energy–time for 4
5.91 J and 2.36 J and (b) C-scan images.
E(t) [J]

3
Impact Level = 5.91J
release more energy than failure mechanisms of the matrix. Impact Level = 2.36J
Because the energy to rupture the fibers is higher than the 2
energy to crack the polymer matrix.
1
4.1.2. Experimental impact results for laminate [0/90/0/90/
0]s 0
0 1 2 3 4 5 6
Fig. 9a shows force–time graphics for composite plates
time [ms]
with stacking sequence [0/90/0/90/0]s, considering two
impact energy levels (5.91 J and 2.36 J). Fig. 9. Experimental impact results for laminate [0/90/0/90/0]s.
There is a region with oscillations of high frequency
(from 0 to 1.2 ms) due to initial contact between specimen
and dart in the force–time graph (Fig. 9a). From 1.8 ms,
there are some small oscillations that show damage, not
only for specimens under Ei of 5.91 J (where maximum
impact force is 3000 N), but also for specimens under Ei
of 2.36 J (where maximum impact force is 1750 N). Due
to the stacking sequence [0/90/0/90/0]s, the main failure
mechanisms are the matrix rupture and delaminations,
reducing the global stiffness of the structure. Therefore,
there is no abrupt drop of the impact force value after
the progressive damage process initiation (Fig. 9a). The
contact time depends on the impact energy level; so, the
contact time is higher (4.5 ms) for specimens impacted by
impact energy of 2.36 J than by impact energy of 5.91 J.
Thus, for high contact time, the mechanical behavior of
the specimen tends to quasi-static response. Fig. 10. Experimental impact results for laminate [0/90/0/90/0]s: (a)
Fig. 9b shows the energy–time graph for composite energy–time for 5.91 J and 2.36 J and (b) C-scan images.
plates with stacking sequence [0/90/0/90/0]s, considering
the impact energy of 5.91 J. It is verified that specimens energy (Fig. 9b), converting 75% of the impact energy to
have absorbed energy of 3.96 J, so only 33% of the impact elastic vibrations. Thus, the specimens impacted by 2.36 J
energy is converted to elastic vibrations. However, the show less failure mechanisms. This observation can be con-
specimens impacted by 2.36 J have just 0.58 J for absorbed firmed, using the images from ultrasonic C-scan technique.
 V. Tita et al. / Composite Structures 83 (2008) 413–428
Fig. 10 shows the ultrasonic C-scan images for compos-
ite plates with stacking sequence [0/90/0/90/0]s, considering
the impact energy of 5.91 J (Fig. 10a) and 2.36 J (Fig. 10b).
Due to the stacking sequence [0/90/0/90/0]s, there are more
layers oriented at 0 than at 90, so, there are more cracks
aligned to 0. Besides, there are ‘‘peanut shapes’’, which
represent delaminations, at both directions (0 and 90).
The area damaged for the specimens impacted by 5.91 J
(340 mm2) is larger than the specimens impacted by
2.36 J (150 mm2). This calculus can explain why the speci-
mens impacted by 2.36 J absorbed just 25% of the impact
energy. Many failure mechanisms cannot be activated for
the stacking sequence equal [0/90/0/90/0]s by this impact
energy level, and the mechanical behavior of the specimen
tends to elastic response.
4.1.3. Experimental impact results for laminate [+45/45/
+45/0/90]s
Fig. 11 shows force–time graphs for composite plates
with stacking sequence [+45/45/+45/0/90]s, considering
four impact energy levels (2.36 J, 4.33 J, 5.91 J and 10.82 J).
The dynamic responses for the specimens impacted by
5.91 J and by 10.82 J are very similar. They reach a maxi-
mum impact force of 3300 N and 4500 N, respectively.
However, dynamic responses are very different for the spec-
imens impacted by 2.36 J and by 4.33 J, which reaches a
maximum impact force of 1900 N at 2.36 J and 2500 N at
4.33 J (Fig. 11). Besides, the specimens impacted by
5.91 J and by 10.82 J show many oscillations close to the
maximum impact force that represent many types of failure
mechanisms, for example: cracks at the matrix, delamina-
tions between layers oriented at 0/90 and at 45/0, as
well as, some fiber rupture. However, there are just some
oscillations close to the maximum impact force for speci-
mens impacted by 2.36 J and by 4.33 J (Fig. 11) that repre-
sent a small amount of damage. In fact, the specimens
impacted by lower energy have a response that tends to
quasi-static response. However, the specimens impacted
5000 Stacking Sequence: [+45/-45/+45/0/90]s
4500 Ei =10.82J
4000 E i =5.91J
3500
3000
Force [N]
2500 E i =4.33J
2000
1500
Ei =2.36J
1000
500
0 0
0 1 2 3 4 5
time [ms]
Fig. 11. Experimental impact results for laminate [+45/45/+45/0/90]s.
423
by a mass of 2.205 kg with an impact energy of 4.33 J
and 10.82 J show that the contact time is higher than the
specimens impacted by a mass of 1.205 kg (Fig. 11).
Fig. 12 shows the energy–time graph for composite
plates with stacking sequence [+45/45/+45/0/90]s, con-
sidering the impact energy equal to 5.91 J. Specimens have
absorbed 4.0 J of energy, so only 32% of the impact energy
is converted to elastic vibrations. The specimens impacted
by 10.82 J have the highest absorbed energy, because
8.8 J was absorbed by many failure mechanisms. On the
other hand, the specimens impacted by 2.36 J have a very
low absorbed energy level (0.8 J), converting 66% of the
impact energy to elastic vibrations. Finally, the specimens
impacted by 4.33 J show that the absorbed energy is
2.6 J, converting only 40% of the impact energy to elastic
vibrations. Thus, the specimens impacted by 2.36 J show
less failure mechanisms. This observation can be con-
firmed, using the images from ultrasonic C-scan.
Fig. 13 shows the ultrasonic C-scan images for compos-
ite plates with stacking sequence [+45/45/+45/0/90]s,
considering the impact energy levels of 5.91 J (Fig. 13a),
2.36 J (Fig. 13b), 10.82 J (Fig. 13c) and 4.33 J (Fig. 13d).
Due to the stacking sequence [+45/45/+45/0/90]s,
there are more layers oriented at +45 than at other direc-
tion, so there are more cracks aligned to +45. Besides,
there are ‘‘peanut shapes’’, which represent delaminations
at four directions (0, +45, 45 and 90). The areas dam-
aged for the specimens impacted by 5.91 J (360 mm2) and
by 10.82 J (450 mm2) are larger than the specimens
impacted by 2.36 J (165 mm2) and by 4.33 J (211 mm2).
This explains why the specimens impacted by 10.82 J
absorbed 81% of the impact energy, where many failure
mechanisms are activated for the stacking sequence [+45/
45/+45/0/90]s by this impact energy level. The mechani-
cal behavior of the specimen tends to a completely inelastic
response where the dart is almost retained by the composite
plate.
12
11
10
9
8 Impact Level = 5.91J
7 Impact Level = 2.36J
E(t) [J]
6 Impact Level = 10.82J
5 Impact Level = 4.33J
4
3
2
1
0
1 2 3 4 5 6 7
6 time [ms]
Fig. 12. Experimental impact results for laminate [+45/45/+45/0/90]s
Energy–time for 5.91 J; 2.36 J; 10.82 J; 4.33 J.
424 V. Tita et al. / Composite Structures 83 (2008) 413–428

indentation curves do not show the oscillations inherent

for the dynamic response obtained from drop-tests. Besides,
when the load reaches the maximum force, which is equal to
the maximum force obtained at drop-test, the indentation
curves do not show oscillations because failure mechanisms
of the composite plate are activated. Thus, for the same
level of load, the indentation curves show lower displace-
ment than impact curves, because, there are more failure
mechanisms activated during the impact event than quasi-
static event. Besides, the failure mechanisms for composite
plate under quasi-static loading concentrate closer to the
region where the dart contacts the specimen. However,
the failure mechanisms shown by plates impacted are more
distributed.
Therefore, the failure mechanisms for a composite plate
under impact loading reduce the global stiffness of the
structure more than the concentrated failure mechanisms.
This observation explains why the slope of indentation
curve is very different of the slope presented by the impact
curve, in the unloading step.

4.2.2. Experimental indentation results for laminate [0/90/0/

Fig. 13. Experimental impact results for laminate [+45/45/+45/0/90]s:
C-scan images.
4.2. Indentation test results
4.2.1. Experimental indentation results for laminate [0]10
Fig. 14 shows the force–displacement graphs for com-
posite plates with stacking sequence [0]10. In general, in
the initial phases of loading, the indentation and the impact
curves have about the same slope that represents non-linear
behavior due to the contact phenomenon. However, the
90/0]s
Fig. 15 shows the force–displacement graphics for com-
posite plates with stacking sequence [0/90/0/90/0]s. In gen-
eral, in the initial phases of the loading step, the
indentation and the impact curves have about the same
slope that represents non-linear behavior due to the contact
interaction between the disks and the aluminium dart
(hemisphere). However, the indentation curves do not
show the oscillations inherent for the dynamic response
obtained from drop-tests. Besides, when the load reaches
the maximum force, which is equal to the maximum force
obtained at drop-test, the indentation curves do not show
oscillations due to intra-ply failures and delaminations
activated.

Indentation Test (Fmax = 2,250 N)

2500 Indentation Test (Fmax = 3,000 N)
Indentation Test (Fmax = 1,300 N) 3500
Indentation Test (Fmax = 1,750 N)

3000
2000

2500

1500
2000
Force [N]

Force [N]

1000 Impact Test 1500

(Fmax = 2,250 N) Impact Test
1000 (Fmax = 3,000 N)
500
500
Impact Test Impact Test
(Fmax = 1,300 N) (Fmax = 1,750 N)
0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5
displacement [mm] displacement [mm]

Fig. 14. Experimental indentation results for laminate [0]10 maximum Fig. 15. Experimental indentation results for laminate [0/90/0/90/0]s
force at: 2250 N and 1300 N. maximum force at: 3000 N and 1750 N.
 V. Tita et al. / Composite Structures 83 (2008) 413–428
Fig. 15 shows for the same level of load that the inden-
tation curve has lower displacement than impact curves,
because there are more failure mechanisms activated dur-
ing the impact event than during the quasi-static event.
However, Fig. 15 shows that the slope of indentation curve
is similar to the slope presented by the impact curve not
only in the loading step, but also in the unloading step.
The specimens impacted by 2.36 J (maximum impact force
equal to 1750 N) just absorbed 25% of the impact energy
and the mechanical behavior of the specimen tends to elas-
tic response. Therefore, many failure mechanisms are not
activated for the stacking sequence which is equal to [0/
90/0/90/0]s by neither impact event nor indentation test.
4.2.3. Experimental indentation results for laminate [+45/
45/+45/0/90]s
Fig. 16 shows the force–displacement graphs for com-
posite plates with stacking sequence [+45/45/+45/0/
90]s. Fig. 16 shows that the indentation and the impact
curves have different slopes. The specimens with stacking
sequence [+45/45/+45/0/90]s impacted show many types
of failure mechanisms that are distributed at composite
plate.
However, the failure mechanisms for a composite plate
under quasi-static loading are concentrated close to the
region where the dart contacts the specimen. These failure
mechanisms reduce the structure global stiffness less than
the distributed failure mechanisms.
However, Fig. 16 shows that the slope of the indentation
curve is similar to the slope of the impact curve not only in
the loading step, but also in the unloading step. The spec-
imens impacted by 2.36 J (maximum impact force equal to
1900 N) converted 66% of the impact energy to elastic
vibrations. Although the slopes of the two curves are sim-
ilar, there is an offset between both curves due to the differ-
ence between the failure mechanisms shown by impact
Indentation Test (Fmax = 3,300 N)
Indentation Test (Fmax = 1,900 N)
3500
3000
2500
Force [N]
2000
1500 Impact Test
(Fmax = 3,300 N)
1000
500 Impact Test
(Fmax = 1,900 N)
0
0 1 2 3 4 5
displacement [mm]
Fig. 16. Experimental indentation results for laminate [+45/45/+45/0/
90]s maximum force at: 3300 N and 1900 N.
425
event (more distributed) and by quasi-static event (more
concentrated).
4.3. Finite element results
4.3.1. Numerical results for laminate [0]10
Fig. 17a shows that the Material Model I can simulate
the indentation test, during the loading phase when param-
eter H is 1.2.
However, the maximum load simulated is higher than
the experimental maximum load, and, the initial phase of
unloading is not simulated very well. Material Model I
can simulate the final of phase unloading, but the perma-
nent strain is not represented when the polymer matrix
exhibits plastic deformation under the rigid hemisphere.
This phenomenon can be verified in Fig. 17b, that shows
a stress concentration at the centre of the disk, because
the contact interaction between the disk and the rigid hemi-
sphere causes contact press and contact stress.
In order to simulate the permanent strain, Hill’s Crite-
rion, implemented in software ABAQUSTM [23], provides
good results at the final phase of unloading. However,
the loading phase cannot be represented, because Hill’s
2500
Material Model I (H=1.2)
Hill's Criterion
2000
Indentation Test
1500
Force [N]
1000
500
0
0 1 2 3 4 5 6
displacement [mm]
Fig. 17. Numerical indentation results for laminate [0]10: (a) force–
displacement and (b) stress distribution.
426 V. Tita et al. / Composite Structures 83 (2008) 413–428
Criterion cannot represent the failure mechanisms shown
by the composite structure. Besides, the maximum load
simulated is much lower than the experimental maximum
load (Fig. 17a).
It is important to note for finite element models on shell
elements, the maximum number of iterations is equal to 10.
For solid elements, the maximum number of iterations is
equal to 9.
4.3.2. Numerical results for laminate [0/90/0/90/0]s
Fig. 18 shows that the Material Model I can simulate
the indentation test, during the loading and unloading
phases when parameter H is equal to 1.6. Besides, the
maximum load simulated is quite similar to the experi-
mental maximum load. However, the simulated load val-
ues for 1.5 mm up to 3.0 mm for the displacement is
very different from the experimental load values, because
delaminations occur for the laminate [0/90/0/90/0]s and
the Material Model I cannot simulate this phenomenon.
Any material degradation law is not activated and the
simulated load values are higher than the experimental
load values. The model correlates with the experimental
test when displacement reaches 3.5 mm, because fiber fail-
ures occur and the degradation law decreases the material
properties. In order to improve the results during the
loading phase, the Material Model II is applied to simu-
late the delaminations. Fig. 18 shows that Material Model
II can simulate the indentation test during the loading and
unloading phases, when parameter H is set to 0.4. Corre-
lation with experimental results is better than with Mate-
rial Model I.
Material Model I cannot simulate the final unloading
phase due to the permanent strain at the disk centre when
the polymer matrix exhibits plastic deformation under the
rigid hemisphere. In order to simulate the permanent
strain, de Hill’s Criterion was used, implemented in soft-
ware ABAQUSTM [23], provides poor results both in load-
3500 Material Model I (H=1.6)
Material Model II (H=0.4)
3000 Hill's Criterion
Indentation Test
2500
2000
Force [N]
1500
1000
500
0 0
0 1 2 3 4 5
displacement [mm]
Fig. 18. Numerical indentation results for laminate [0/90/0/90/0]s.
ing and in the unloading phases (Fig. 18). In fact, the
permanent strain for the laminate [0/90/0/90/0]s is lower
than the laminate [0]10 due to the stacking sequence.
It is important to note for finite element models on shell
elements, the maximum number of iterations is equal to 8.
For solid elements, the maximum number of iterations is
equal to 8, too.
4.3.3. Numerical results for laminate [+45/45/+45/0/90]s
Fig. 19 shows that Material Model I cannot simulate the
indentation test very well, during the loading and unload-
ing phases when parameter H is 1.2. However, the maxi-
mum load simulated is quite similar to the experimental
maximum load. The simulated load values for 1.5 mm up
to 3.2 mm for the displacement are very different from
the experimental load values, because delaminations occur
for the laminate [+45/45/+45/0/90]s and the shear stres-
ses increase due to the stacking sequence, providing a
non-linear behavior. Any material degradation law is not
activated and the simulated load values are higher than
the experimental load values. The model correlates with
the experimental test when displacement is close to
3.5 mm, because fiber failures occur and the degradation
law decreases the material properties. In order to improve
the results during the loading phase, the Material Model II
is applied to simulate the delaminations. Fig. 19 shows that
Material Model II cannot simulate very well the indenta-
tion test during the loading and unloading phases, when
parameter H is equal to 0.3, because the non-linear behav-
ior caused by the shear stresses is not simulated with accu-
racy. Finally, in order to simulate the permanent strain,
Hill’s Criterion, implemented in software ABAQUSTM
[23], provides poor results too (Fig. 19).
It’s important to note for finite element models on shell
elements, the maximum number of iterations is equal to 7.
For solid elements, the maximum number of iterations is
equal to 8.
Material Model I (H=1.2)
3500 Material Model II (H=0.3)
Hill's Criterion
3000
Indentation Test
2500
2000
Force [N]
1500
1000
500
0 1 2 3 4 5
displacement [mm]
Fig. 19. Numerical indentation results for laminate [+45/45/+45/0/90]s.
 V. Tita et al. / Composite Structures 83 (2008) 413–428
5. Conclusions
For the experimental approach, it is verified that stack-
ing sequence and impact energy level can influence on the
dynamic response of composite plates. The graphs of
force–time and energy–time, as well as the images from
ultrasonic C-scan technique are used in order to compare
the mechanical behavior of the specimens, which is repre-
sented by graph of the absorbed energy versus impact
energy level. Fig. 20 can be divided into three regions:
• Region 1: the specimens have a quasi-elastic behavior,
because the fraction of absorbed energy is very low
(under 35%) as failure mechanisms are not activated;
• Region 2: the specimens show some failure mechanisms,
which are matrix crack and delaminations; so, the frac-
tion of absorbed energy is intermediate between 35%
and 75%;
• Region 3: the specimens show many types of failure
mechanisms, for example: fiber rupture, matrix crack
and delaminations; so, the fraction of absorbed energy
is very high (over 75%).
In general, the indentation test can be used to represent
a drop-test when the impact energy level is low and the
specimen has a quasi-elastic behavior. Because the indenta-
tion curves do not show the oscillations inherent for the
dynamic response obtained from drop-tests and there are
more failure mechanisms activated during the impact event
than quasi-static event. Besides, the failure mechanisms
shown by the impact event are more distributed. With a
quasi-static event they are more concentrated. Thus, the
structural global stiffness reduces with more intensity for
drop-test. Finally, it is very important to comment that
Stacking Sequence
[0]10
85
[0/90/0/90]s
80 [+45/-45/+45/0/90]s
75
70
65
Absorbed Energy [%]
60
55
50
45
40
35
30
25
20
0 1 2 3 4 5 6 7 8 9
Impact Energy Level [J]
Fig. 20. Absorbed energy · impact energy.
427
the experimental test results for this work were used to val-
idate and calibrate a composite material failure model
developed by the authors and this failure model will be
shown in a future publication.
For the numerical approach, Material Model I for shell
elements can simulate the indentation test of three stacking
sequences, if the parameter H is adjusted according to the
experimental results. However, Material Model I cannot
represent the delaminations, because any material degrada-
tion law is not activated by this model. Thus, for some
regions of the curve force–displacement, the numerical
result does not converge to the experimental result. Mate-
rial Model II for solid elements can simulate the delamina-
tions, during the indentation test. However, Material
Model II does not simulate the non-linear behavior caused
by the shear stresses with accuracy. Besides, both material
models cannot represent the permanent strain after the
final of the unloading phase. In fact, both material models
can be improved in order to simulate the phenomena with
more accuracy.
Acknowledgments
The authors would like to thank Dr. Ir. Mieke Lossie
and technician Paul Van Cauwenbergh (Mechanical Engi-
neering Department of Katholieke Universiteit Leuven) for
the technical support during the realization of this work.
The authors would like to thank Professor Ignaas Verpoest
(Material Engineering Department of Katholieke Universi-
teit Leuven (Belgium)) for allowing the realization of this
work at LCPC (Leuven Composites Processing Centre).
The authors would like to thank FAPESP for the funds
to realize this work.
References
[1] Kindervater CM, Georgi H. Composite strength and energy absorp-
tion as an aspect of structural crash resistance. In: Norman J, Tomaz
W, editors. Structural crashworthiness and failure. London: Elsevier
Science; 1993. p. 189–235.
[2] Oñate E et al. Métodos avanzados de cálculo de estructuras de
3 materiales compuestos – monografia 3. Barcelona: Centro Internac-
ional de Métodos Numéricos en Ingenierı́a; 1991.
[3] Cairns DS, Lagace PA. A consistent engineering methodology for the
treatment of impact in composite materials. J Reinf Plast Compos
1992;11(4):395–412.
[4] Farley GL, Jones RM. Prediction of the energy absorption capability
of composites type. J Compos Mater 1992;26(3):388–404.
2 [5] Haug E, De Rouvray A. Crash response of composite structures. In:
Norman J, Tomaz W, editors. Structural crashworthiness and
failure. London: Elsevier Science; 1993. p. 237–94.
[6] Bellingardi G, Gugliotta A, Vadori R. Fragmentation of composite
material plates submitted to impact loading: comparison between
numerical and experimental results. Key Eng Mater 1998;144:75–88.
1 [7] Collombet F, Lalbin X, Lataillade JL. Damage prediction of
laminated composites under heavy mass–low velocity impact. Key
Eng Mater 1998;141–143(1):743–76.
10 11 [8] Gottesman T, Girshovich S. Impact damage assessment and mechan-
ical degradation of composites. Key Eng Mater 1998;141(1):3–18.
[9] Vicente JLS, Béltran F, Martı́nez F. Simulation of impact on com-
posite fuselage structures. In: European congress on computational
428 V. Tita et al. / Composite Structures 83 (2008) 413–428
methods in applied sciences and engineering, Proceedings of the
ECCOMAS 2000, Barcelona, 11–14 September 2000.
[10] Kindervater CM et al. Crash and impact simulation of aircraft
structures-hybrid and FE based approaches. In: European congress
on computational methods in applied sciences and engineering,
Proceedings of the ECCOMAS 2000, Barcelona, 11–14 September
2000.
[11] Kostopoulos V et al. Finite element analysis of impact damage
response of composite motorcycle safety helmets. Composites: Part B
2002;33:99–107.
[12] Lopresto V, Caprino G. Elastic response of circular CFRP plates
under low-velocity impact. In: European conference on composite
materials, Proceedings of ECCM 10, Brugge, 3–7 June 2002.
[13] Naik NK, Shrirao P. Composite structures under ballistic impact.
Compos Struct 2004;66:579–90.
[14] De Morais WA, Monteiro SN, d’Almeida JRM. Evaluation of
repeated low energy impact damage in carbon–epoxy composite
materials. Compos Struct 2005;67:307–15.
[15] Mitrevski T et al. The effect of impactor shape on the impact
response of composite laminates. Compos Struct 2005;67:139–48.
[16] Mikkor KM et al. Finite element modelling of impact on preloaded
composite panels. Compos Struct 2006;75:501–13.
[17] Mamalis AG et al. The static and dynamic axial collapse of CFRP
square tubes: finite element modelling. Compos Struct
2006;74:213–25.
[18] Kim J-S, Chung S-K. A study on the low-velocity impact response of
laminates for composite railway bodyshells. Compos Struct
2007;77:484–92.
[19] Zhao GP, Cho CD. Damage initiation and propagation in composite
shells subjected to impact. Compos Struct 2007;78:91–100.
[20] Anderson TL. Fracture mechanics – fundamentals and applica-
tions. New York: CRC Press; 1995.
[21] Abrate S. Impact on composite structures. London: Cambridge
University Press; 1998.
[22] ASTM D5628-96. Standard test method for impact resistance of flat,
rigid plastic specimens by means of a falling dart (tup or falling mass).
Philadelphia: American Society for Testing and Materials, 1996.
[23] ABAQUS. Standard user’s manual: theory. Pawtucket: Hibbitt,
Karlsson & Sorensen, 2002.
[24] Yamada SE, Sun CT. Analysis of laminate strength and its
distribution. J Compos Mater 1978;12:275–84.
[25] Hashin Z. Analysis of properties of fiber composites with anisotropic
constituents. J Appl Mech 1979;46(3):543–50.
[26] Hashin Z. Failure criteria for unidirectional fiber composites. J Appl
Mech 1980;47(2):329–34.
[27] Chang FK, Chang KY. Post-failure analysis of bolted composite
joints in tension or shear-out mode failure. J Compos Mater
1987;21(9):809–33.
[28] Chang FK, Chang KY. A progressive damage model for laminated
composites containing stress concentrations. J Compos Mater
1987;21(9):834–55.
[29] Chang KY, Liu S, Chang FK. Damage tolerance of laminated
composites containing an open hole and subjected to tensile loadings.
J Compos Mater 1991;25(3):274–301.
[30] Chang FK, Lessard LB. Damage tolerance of laminated composites
containing an open hole and subjected to compressive loadings. Part I
– analysis. J Compos Mater 1991;25(1):2–43.
[31] Shahid I, Chang FK. An Accumulative damage model for tensile and
shear failures of laminated composite plates. J Compos Mater
1995;29(7):926–81.
[32] Hahn HT, Tsai SW. Nonlinear elastic behavior of unidirectional
composite laminae. J Compos Mater 1973;7:102–18.
[33] Liu S, Kutlu Z, Chang FK. Matrix cracking and delamination in
laminated composite beams subjected to a transverse concentrated
line load. J Compos Mater 1993;27(5):436–70.
[34] Kermanidis T et al. Finite element modeling of damage accumulation
in bolted composite joints under incremental tensile loading. In:
European congress on computational methods in applied sciences and
engineering, Proceedings of the ECCOMAS 2000, Barcelona, 11–14
September 2000.
[35] Dávila CG, Camanho PP. Decohesion elements using two and three-
parameter mixed-mode criteria. In: International structures specialists
meeting, Proceedings of international structures specialists meeting
2001, Williamsburg, 2001.
[36] Dávila CG, Camanho PP, Moura MF. Progressive damage analyses
of skin/stringer debonding. In: American Society of Composites
Annual Technical Conference. Proceedings of ASC/16, Blacksburg,
2001.