International Journal of Solids and Structures 144–145 (2018) 230–247

Contents lists available at ScienceDirect

International Journal of Solids and Structures

journal homepage: www.elsevier.com/locate/ijsolstr

Simulating drop-weight impact and compression after impact tests on

composite laminates using conventional shell finite elements
E.V. González a,∗, P. Maimí a, E. Martín-Santos a,b, A. Soto a, P. Cruz b, F. Martín de la Escalera c,
J.R. Sainz de Aja c
a
AMADE, Mechanical Engineering and Industrial Construction Department, Universitat de Girona, Campus Montilivi s/n, Girona 17071, Spain
b
Applus IDIADA, Main Technical Center, Santa Oliva, Tarragona E-43710, Spain
c
Aernnova Engineering Division, Structural Integrity Department, 20 Manoteras Avenue - Building B, 5th Floor, Madrid 28050, Spain

a r t i c l e i n f o a b s t r a c t

Article history: Simulating polymer-based composite structures under low-velocity impact and sequencing compression
Received 12 November 2017 after impact loading, is a complex problem that requires using well-suited constitutive models and defin-
Revised 19 April 2018
ing advanced finite element capabilities. Therefore, developing simplified and efficient, but sufficiently
Available online 18 May 2018
accurate finite element models to solve such problems, is of interest. Here, a finite element modelling
Keywords: strategy is presented for simulating low-velocity impact and compression after impact tests on compos-
Composite materials ite laminates using Abaqus/Explicit software. The strategy is based on using conventional shell elements
Impact and cohesive surfaces. The proper out-of-plane structural response is solved by considering surface el-
Compression ements located on the bottom and top faces of the layers. The key parameters requested for defining
Finite element the models are concisely described and the values selected are well justified. The accuracy of the mod-
Shell elling strategy is proved by simulating monolithic and rectangular laboratory coupons. The results of the
simulations reveal good agreement with most of the experimental data reported.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction surement of the visually apparent damage area, measurement of

In order to assess the damage resistance and tolerance of a
polymer-based composite structure, the Low-Velocity Impact (LVI)
and sequenced Compression After Impact (CAI) tests are typically
performed. The large number of publications available in the litera-
ture evidences an intensive research dealing with the experimental
analysis of these tests, with the aim of knowing the complex pro-
gressive degradation of the material at different interacting failure
mechanisms (e.g. Chai et al., 1983; Byers, 1980).
The structural response in a LVI test is a function of the
structure parameters (thickness, in-plane size, lamina type, elas-
tic and fracture properties, density, stacking sequence, and bound-
ary conditions), impactor parameters (shape, size, elastic proper-
ties, mass, velocity, and incidence angle), and the environmental
conditions (Davies and Olsson, 2004). Reviews of studies address-
ing the analysis of composite structures under impact tests can
be found (Cantwell and Morton, 1991; Richardson and Wisheart,
1996; Abrate, 1998). In a LVI event, the damage may not be clearly
visible and the criteria for damage assessment may include mea-
∗
Corresponding author.
E-mail address: emilio.gonzalez@udg.edu (E.V. González).
dent depth, and non-destructive evaluation, such as C-Scan, for the
internal damage area (MIL-HDBK-17-3F, Military Handbook). Typi-
cally, matrix cracking is the first failure to occur at early stages of
the impact load. The matrix cracking induce delaminations which
grow as the load increases. Often, the appearance of notable de-
laminations can be detected by a noticeable drop in load (ASTM D
7136/D 7136M-12, 2012a). Following the growth of the delamina-
tion, perceptible fibre breakage may occur depending on the level
of impact energy. After a LVI test, permanent indentation may ap-
pear due to matrix plasticity and the disorder of broken fibres, and
it is used as an indication of the severity of the internal damage
induced by impact.
LVIs caused by large-mass impactors yield a type of response
which can be approximated as quasi-static (Olsson, 20 0 0). In this
sense, the impact event can be analyzed as a static indentation
problem and these tests can provide a meaningful indication on
the damage mechanisms occurring during LVIs at different dis-
placement values (Wagih et al., 2016a; 2016b).
Concerning the CAI test, the residual strength is function of the
local buckling of sublaminates, and propagation of impact-induced
matrix cracks, delaminations and fibre breakage. As noted in the
recent review found in Liv (2017), there is still a lot to do to under-
stand the sequence of failure modes that leads to the final failure,

https://doi.org/10.1016/j.ijsolstr.2018.05.005
0020-7683/© 2018 Elsevier Ltd. All rights reserved.
 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247
because there exist a dependence on the interaction among dif-
ferent failure mechanisms, including local buckling, resulted from
both material and structural properties of laminated composites.
The phenomenon of impact damage in laminated composite
structures is very complex and difficult to model analytically. Al-
ternatively, the numerical models based on the Finite Element (FE)
method represent a power tool for the prediction of the physi-
cal processes. The FE models allow the analysis of a large num-
ber of impact configurations and structures, specially in cases that
are too complex or expensive to analyze by purely empirical meth-
ods. Nonetheless, the more rigorous the model becomes, the more
demanding the computational power becomes. Therefore, an effort
must be done to develop FE models which are also time efficient.
Simulating LVI and CAI tests on polymer-based laminated com-
posite structures is a complex problem since contact and progres-
sive degradation of the material occur. Accordingly, the interlami-
nar (delamination) and the intralaminar failure mechanisms should
be suitably modelled in a 3D FE analysis, together with the con-
cise definition of several key parameters that can play a role in
the quality of the numerical predictions, i.e. material properties,
FE types and sizes, and model assembly and boundary conditions.
Modelling at the mesoscale level is a reasonable idealization of
the structure to capture the complexity of the problem and is the
scale commonly used in related work found in the literature. The
mesoscale describes each ply as a homogenised material, which
are separated from each other by the interfaces susceptible to de-
lamination, i.e. interfaces with mismatch fibre orientation of the
surrounding plies. Certainly, on this scale there are different al-
ternatives for modelling the laminate depending on the FE type
and the cohesive connection between plies and, in accordance with
the modelling approach selected, well-suited interlaminar and in-
tralaminar constitutive models should be formulated and imple-
mented.
A large number of recent investigations dealing with the sim-
ulation of LVI tests on laminated and monolithic specimens can
be found in the literature (Lopes et al., 2009; Bouvet et al., 2009;
2012; Hongkarnjanakul et al., 2013; Feng and Aymerich, 2014; Shi
et al., 2014; Schwab et al., 2016; Liu et al., 2016; Lopes et al.,
2016). Others dealing with the CAI test by assuming a damage
pattern (normally delamination), which may be based on exper-
imental inspections of impacted specimens in question, can also
be found (Pavier and Clarke, 1996; Suemasu et al., 20 08; 20 09;
Craven et al., 2010; Yan et al., 2010; Obdržálek and Vrbka, 2011).
Moreover, there are a significant number of studies published in
the last five years that report on the sequential simulation of both
LVI and CAI tests (González et al., 2012; Dang and Hallett, 2013;
Rivallant et al., 2013; Mendes and Donadon, 2014; Caputo et al.,
2014; Tan et al., 2015; Abdulhamid et al., 2016; Perillo et al., 2017;
Abir et al., 2017). These are the most interesting since neither LVI
nor CAI have to be tested experimentally. It is worth mentioning
that most of these numerical models are focused on simulating
common laboratory coupons, however, some other work focused
on predicting the LVI test on large composite structures has also
been developed (Johnson and Holzapfel, 2006; Faggiani and Fal-
zon, 2010; Riccio et al., 2016; Schwab and Pettermann, 2016; Sun
and Hallett, 2017). Certainly, the simulation of large structures is
more interesting from an industrial point of view, but even more
so if both LVI and CAI tests are simulated in order to assess a struc-
ture’s damage resistance and tolerance (Soto et al., 2017).
Within the studies cited above, there are different FE mod-
elling technologies able to model the plies and delaminations: 3D
solid FEs for plies, together with non-zero thickness cohesive el-
ements (Lopes et al., 2009; González et al., 2012; Dang and Hal-
lett, 2013; Lopes et al., 2016; Perillo et al., 2017; Sun and Hal-
lett, 2017), zero thickness cohesive elements (Bouvet et al., 2009;
2012; Rivallant et al., 2013; Hongkarnjanakul et al., 2013; Feng and
231
Aymerich, 2014; Shi et al., 2014; Liu et al., 2016; Abdulhamid et al.,
2016) or cohesive surfaces (Tan et al., 2015; Lopes et al., 2016);
continuum shell FEs for plies, together with non-zero thickness
(Faggiani and Falzon, 2010; Caputo et al., 2014; Riccio et al., 2016)
or zero thickness (Abir et al., 2017) cohesive elements; and finally,
conventional shell FEs for ply modelling together with non-zero
thickness cohesive elements (Mendes and Donadon, 2014; Schwab
et al., 2016; Schwab and Pettermann, 2016), zero thickness cohe-
sive elements (Soto et al., 2018) or cohesive surfaces (Johnson and
Holzapfel, 2006). In some studies, different FE technologies are
combined to simulate intralaminar failure mechanisms, as in the
work of Bouvet et al. (2009, 2012), where spring-based interface
and volumetric elements are used for matrix cracking and fibre
breakage modelling, respectively.
When modelling LVI and CAI events for laminated compos-
ite materials, using small size FEs is recommended to adequately
computate the energy dissipation for each failure mechanism
(Abir et al., 2017). Moreover, if the option to model most of the in-
terfaces susceptible for delamination is considered (Johnson et al.,
2001), together with the fact that composite materials have high
specific stiffness values, these features may lead to large FE mod-
els which require long computational analysis, especially when us-
ing explicit FE codes. Therefore, there is a need to develop accu-
rate FE modelling strategies to improve computational efficiency.
Because of their kinematic simplicity and their useful capability
to model a certain number of plies using a single shell element,
the use of conventional shell elements is an interesting modelling
approach. Johnson et al. (2001) used conventional shell elements
together with contact interface conditions in the PAM-CRASH FE
software (PAM-CRASH FE Code, 2004). The accuracy of the FE mod-
els was proved by simulating LVI on carbon fabric reinforced epoxy
plates for two impact velocities. A remarkable result of this work
is how different number of interfaces for delamination are mod-
elled, something which Soto et al. (2018) also considered for thin-
ply based laminates. Similarly, Mendes and Donadon (2014) sim-
ulated woven composite laminates, but in this case they used
Abaqus/Explicit FE code and simulated both the LVI and CAI tests
using two FE approaches. The first uses a single shell element in
which the whole laminate thickness is considered and the second
approach uses two shell layers with half of the laminate thick-
ness, so that the laminate is divided into two sub-laminates re-
lated by means of a thin layer of cohesive elements. They reported
predictions for two laminate thickness and different impact en-
ergy levels. More recently, Schwab et al. (2016) presented another
modelling approach in Abaqus/Explicit which used shell elements
for ply modelling and cohesive elements formulated with zero-
thickness but with a physical volume to join the separated sur-
rounding shell plies. The accuracy of the model was checked only
by simulating a fabric reinforced composite laminate under a high
impact energy, at which the structure response was highly local-
ized at the impact site and, therefore, the bending response was
not important as it is for LVI tests. In this last work, all interfaces
for the delamination of a laminate with eight plies were modelled.
Due to industrial interest in having predictive tools to deal with
analyzing advanced structures in reasonable calculation times, for
instance simulating both LVI and CAI tests on polymer-based lam-
inated structures, this present work purposes an efficient and ac-
curate modelling strategy to be used in Abaqus/Explicit FE code.
This strategy is based on using conventional shell elements to-
gether with cohesive surfaces, where the out-of-plane structural
response is checked and solved by considering surface elements
placed on the bottom and top faces of the layers and tied to the
shell elements. This modelling strategy and its validation are set
out as follows. In Section 2, the proposed modelling strategy is de-
scribed in detail. Section 3 details the materials, laminates and test
set-ups considered for the LVI and CAI experimental test campaign
232 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247
Fig. 1. FE modelling strategy based on using conventional shell elements. The pa-
rameter hp refers to the thickness of a ply (or of a sub-laminate if a set of plies are
defined in the same shell element).
to validate the numerical tool. Section 4 presents the key param-
eters required for the complete definition of the FE models using
the strategy presented here. In Section 5, the required calculation
times for each FE analysis are summarized and the accuracy of the
numerical models is checked by comparing them with the exper-
imental data. Finally, Section 6 presents the main conclusions and
some suggestions for future work.
2. FE modelling strategy
Fig. 1 illustrates the strategy proposed, where each ply is de-
scribed in the through-the-thickness direction using a conven-
tional shell element located at the mid-plane and surface ele-
ments on the bottom and top faces of the ply. These surface el-
ements are connected to the shell elements through rigid Tie con-
nectors, which transfer the kinematics from the shell elements to
the surface elements. Surface elements can be simply understood
as sheets of nodes which move according to the displacements and
the rotations of the shell elements without storing elastic energy.
On the other hand, the delamination between the plies is modelled
using cohesive surfaces, which describe the mechanical interaction
between two surfaces by means of a contact algorithm.
Defining the surface elements and Tie connectors of the mod-
elling strategy presented is necessary when using mid-plane lo-
cated conventional shell elements and cohesive surface interac-
tions. In fact, cohesive surfaces can also be used directly by re-
lating the corresponding surfaces of shell elements. However, co-
hesive surfaces not working properly due to a poor description of
the shell kinematics at the interfaces for delamination were de-
tected, i.e. the displacements on the top and bottom surfaces of
the shell thickness are not well predicted. This issue is illustrated
in Appendix A with the simulation of a three-point bending test
for an isotropic beam.
Surface elements are defined with a virtual thickness hst (see
Fig. 1), known by Abaqus as Surface Thickness (ABAQUS version
6.12-1, 2012). This thickness (i.e. the thickness that Abaqus con-
siders for contact with other elements), for shell elements rep-
resents the corresponding ply thickness hp assigned in the shell
section.1 For the correct contact prediction between surface el-
ements, a user-selected Surface Thickness must be defined for
1 The LVI and CAI tests were performed on rectangular-shaped spec-
This definition is given by default by ABAQUS version 6.12-1 (2012). However, if
the thickness of a shell element is greater than its in-plane dimensions, the original
surface thickness hp is automatically reduced by Abaqus to reach a suitable aspect
ratio between the thickness and the in-plane size of the element to keep the suit-
ableness of the shell theory (although not specified by Abaqus, the ratio is ≈ 0.5).
Using linked surface elements, the thickness of the shell elements is always guaran-
these elements: hst = α h p , where α is a small percentage of the
shell thickness modelled (typically ≈ 0.01). Accordingly, the con-
ventional shell elements should be physically modelled at the cor-
responding locations by taking into account the values of the Sur-
face Thickness of the surface elements in order to guarantee con-
tact at the real physical locations, or assume contact in slightly dif-
ferent locations otherwise.
The main advantage of the strategy presented here, is the elim-
ination of the detrimental effect of the FE thickness on the com-
putational analysis time, which appears when 3D solid elements
or continuum shell elements are used. The explicit integration al-
gorithm uses the stable time increment tst , which can be simply
approached using the condition described by Eq. (1):
min min
tst ≤ =  (1)
c E/ρ
where min is the minimum length of a critical finite element of the
mesh and c is the velocity of the sound wave propagated through
the material calculated by the Young modulus E and the density ρ .
Commonly, laminated composite materials have a small ply thick-
ness, and if the delamination has to be predicted, each ply is mod-
elled using a through-the-thickness single FE with small thickness.
Using the FE strategy described, the thickness of the plies does
not affect the stable time increment, thus it is more efficient com-
pared to conventional simulations based on 3D solid or continuum
shell elements. Therefore, the proposed solution is useful when the
modelled ply thickness is smaller than the in-plane FE dimensions.
Likewise, the use of cohesive surfaces avoids using interface ele-
ments which are often defined using 3D solid elements with very
small thickness (e.g. González et al., 2012), thus yielding to small
stable time increments.
It is worth remarking that Tie connectors do not affect the sta-
ble time increment. However, contact definitions carried out on
surface elements may have an effect on, or may even control,
the tst . When modelling contact on surface elements, Abaqus re-
quests a surface density to be defined, and thus a surface mass
which must be extracted from the mass of the corresponding mod-
elled ply or sub-laminate. Undoubtedly, this definition reduces the
density defined on conventional shell elements and it can therefore
decrease the stable time increment in accordance with Eq. (1).
3. Materials, laminates and test set-ups
Two different carbon fibre preform architectures have been con-
sidered to simulate the LVI and CAI tests: a Unidirectional Tape
TM
(UT) and a Woven Fabric (WF), both HexForce type reinforce-
®
ments of Hexcel , called G1157 (UD, 6 K, 270 gsm) and G0926
(5HS, 6 K, 370 gsm), respectively. These carbon fibre materials are
used with the HexFlow® RTM 6 mono-component epoxy system.
The mean cured ply thickness, hp , for each UT and WF material
are 0.247 mm and 0.353 mm, respectively (González et al., 2014)
and for both ply systems the fibre volume fraction is 58%. The cor-
responding material properties are listed in Tables 2 and 5 (see
Section 4).
Three different laminates, manufactured by the Resin Transfer
Moulding (RTM) process, have been analyzed: [(0, 45, 90, −45 )2 ]s
and [02 , 452 , 902 , −452 ]s using UT material, and [(0, 45)3 ]s using
WF material (from here on, these laminates are identified as LUT1,
LUT2 and LWF, respectively). The resulting thickness of the LUT1
and LUT2 laminates is 3.95 mm, and 4.24 mm for laminate LWF.
imens with 150 mm × 100 mm in-plane dimensions. It is worth
teed regardless of the in-plane element size used, and thus, the contact behaviour
is then well defined.
 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247 233

Table 1 scribed using a modified version of the constitutive model devel-

LVI configurations.
oped by Maimí et al. (2007a,b), while the WF material is described
Laminate LVI test Energy [J] Mass [kg] Velocity [m/s] by the model presented by Martín-Santos et al. (2014). Both mod-
LUT1 and LUT2 IC20 20 3 3.65 els were intended for use with shell elements and formulated in
IC30 30 3 4.47 a thermodynamically consistent framework, thus assuring the ir-
LWF IC30 30 3 4.47 reversibility of the different degradation processes under variable
IC40 40 5 4.00 loading conditions. In addition, the Crack Band Model (CBM) algo-
IC50 50 5 4.47
rithm formulated by Bažant and Oh (1983) was implemented in
Table 2 both models to guarantee mesh objectivity, i.e. regularization of
Ply properties of each composite material: Unidirectional Tape (UT) and Woven Fab- the energy dissipated at each material point according to the FE
ric (WF). NR refers to Not-Requested property.
size .
Property UT WF The main features of the modified version of the constitu-
Density ρ [kg/m ] 3
1510.0 1510.0 tive model for unidirectional plies presented in Maimí et al.
Elastic properties E1 [GPa] 116.73 59.54 (2007a,b) are: three in-plane damage variables associated with the
E2 [GPa] 8.31 54.95 longitudinal failure d1 (fibre breakage), the transverse failure d2
G12 [GPa] 4.67 5.21 (matrix cracking), and the in-plane shear damage variable d6 in-
ν 12 0.26 0.03
fluenced by longitudinal and transverse cracks are computed; the
Strength X1+ [MPa] 1477.09 804.06
X1− [MPa] 708.46 531.29 model accounts for in-plane shear plasticity prior to shear dam-
X2+ [MPa] 55.47 772.77 age; the damage activation functions are based on the LaRC failure
X2− [MPa] 170.00 534.11 criterion (Dávila et al., 2005; Pinho et al., 2005); all of the dam-
X6 [MPa] 117.45 NR
age law shapes are described by linear softening functions, albeit
Fracture toughness G1+ [N/mm] 248.50 91.00
G1− [N/mm] 1.0 × 104 1.0 × 104 with the exception of longitudinal tension and compression which
G2+ [N/mm] 0.62 91.00 are described by bi-linear softening functions in order to capture
G2− [N/mm] 3.02 1.0 × 104 the different failure mechanisms (for tension: fibre breakage and
G6 [N/mm] 1.82 NR bridging, and fibre pull-out; and for compression: kink-band initi-
ation and broadening). Accordingly, Fig. 2 shows the shapes of the
noting that the UT plies with 0° orientation are aligned with the constitutive laws for each loading case.
shortest in-plane dimension of the specimen. On the other hand, the main features of the constitutive model
Based on the ASTM D 7136/D 7136M-12 (2012a) and ASTM D used for woven fabric plies are (Martín-Santos et al., 2014): two
7137/D 7137M-12 (2012b) standards, the LVI and CAI tests were in-plane damage variables, d1 and d2 , are computed to account
carried out using a CEAST Fractovis Plus and an MTS-810 multi- for the degradation in the principal directions of the material (i.e.
purpose servo-hydraulic tester, respectively. The projected delam- weft and warp fibre tows, respectively); to minimise the computa-
ination areas of some of the impacted specimens were inspected tion time, the related damage activation functions are formulated
using the C-scan ultrasonic technique by means of an Olympus in a simplified way, based on a maximum stress criteria which is
OmniScan MX system. sufficient for woven fabric type plies (Arteiro, 2016); the degrada-
Different LVI configurations have been considered. The corre- tion of the tensile and compressive behaviour in both principal di-
sponding energy, mass and velocity for each test configuration rections are described using a bi-linear softening function (as in
are summarized in Table 1. For laminates with UT material, the Fig. 2(a)); and the in-plane shear behaviour is described by a non-
maximum impact energy was 30 J, which is close to the recom- linear isotropic hardening plasticity without damage. Additionally,
mended value defined by ASTM D 7136/D 7136M-12 (2012a) to in contrast to the model reported in Martín-Santos et al. (2014),
ensure delamination areas smaller than one-half of the smaller a damage degradation for in-plane shear loading based on the
unsupported specimen size. The resulting projected delamination damage in both principal directions, d6 = 1 − (1 − d1 )(1 − d2 ), has
area after impact for 30 J (obtained by C-scan inspections), was been included.
close to the specimens’ clamped boundaries (especially for lami- It is worth noting that the majority of the material proper-
nate LUT2); thus a second LVI configuration with a smaller energy ties required by these constitutive models are obtained through
was selected, i.e. 20 J. On the other hand, the laminates manufac- ply-based test methods. The corresponding material properties are
tured with WF material showed a better damage resistance at 30 J, listed in Table 2: ply elastic properties for each principal direction
resulting in a small delamination area concentrated around the im- and in-plane shear (E1 , E2 , G12 and ν 12 (ASTM D 3039/D 3039M-
pact point. Accordingly, two higher impact energies were consid- 0 0, 20 0 0; EN 2597:1998, 1998; AITM1-0 0 02, 1998), ply strengths
ered: 40 J and 50 J (González et al., 2014). (X1+ and X1− , longitudinal tension and compression, respectively
For laminates with UT material, a sample of three specimens (ASTM D 3039/D 3039M-0 0, 20 0 0; EN 2850, 1997); X2+ and X2− ,
was considered for each LVI configuration. For laminates with WF transverse tension and compression, respectively (EN 2597:1998,
material, two specimens for each LVI configuration were used. On 1998; EN 2850, 1997); X6 , in-plane shear (AITM1-0 0 02, 1998) and
the other hand, for each impact configuration only one specimen critical energy release rates (G1+ and G1− , longitudinal tension and
was tested under CAI for LUT1 and LUT2 laminates, and two speci- compression, respectively; G2+ and G2− , transverse tension and
mens for LWF laminate. As shown in Section 5, the mean values of compression, respectively; G6 , in-plane shear). For the elastic and
the experimental results obtained for each batch and configuration strength properties, a sample of five specimens was tested for each
were used for comparison with the numerical predictions. loading case and material. It is worth mentioning that X6 is con-
sidered as being the minimum between the maximum shear stress
4. Key FE definitions and the value of the shear stress at 5% of shear strain from the
in-plane shear tensile tests.
4.1. Constitutive models and material properties The values of the critical energy release rates, and even more
so, the shapes of the constitutive laws for fibre failure, must be
4.1.1. Ply modelling defined accurately because these material properties play an im-
For the ply materials presented in Section 3, two different user- portant role in the quality of the LVI and CAI predictions when
defined constitutive materials are used. The UT material is de- the selected impact energy exceeds a certain threshold. However,
234 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247

Fig. 2. Constitutive laws for unidirectional plies at each in-plane loading direction: (a) longitudinal, (b) transverse, and (c) shear.

these material properties are difficult to obtain. The critical en- Table 3
In-situ strengths of UT material for each ply clustering thickness and location.
ergy release rates G1+ and G1− are typically derived by using Com-
To calculate X6is , the values of X6P = 94.53 MPa and KP = 0.25 are considered. Al-
pact Tension (CT) and Compact Compression (CC) tests, respec- though the mid-plane cluster is modelled using two separated shell elements (see
tively, of cross-ply laminates (Pinho et al., 2006; Dávila et al., 2009; Section 4.2), the strength to be considered corresponds to that of the whole cluster.
Catalanotti et al., 2010; Laffan et al., 2012; Ortega et al., 2014; is
Number of plies Position X2+ [MPa] X6is [MPa]
Bergan et al., 2016), although it is worth mentioning that often
a large amount of undesirable dissipation by shear plasticity can 1 Outer 103.01 125.89
Embedded 163.69 156.64
develop when using cross-ply laminates. For the present work, ex-
2 Outer 72.84 117.45
perimental data for cross-ply laminates were not available and so Embedded 115.75 142.92
the quasi-isotropic laminates described in Section 3 were consid- 4 Embedded 87.86 142.92
ered for CT and CC tests instead. Using the method described by
Ortega et al. (2016), the translaminar critical energy release rate for
Table 4
each laminate, as well as the shape of the constitutive law for both Densities defined on surface and conventional shell elements for each material and
tension and compression, can be obtained (Ortega et al., 2017a). thickness in [kg/m2 ] and [kg/m3 ], respectively.
Based on these data, the corresponding fibre properties can be ad-
FE type UT (one ply) UT (two clustered plies) WF
justed. For more details, refer to Appendix B.
−2 −1
Surface element 9.32 × 10 1.86 × 10 1.33 × 10−1
As noted in Table 2, for UT material G2+ is approached to the
Shell element 755.00 755.00 755.00
corresponding interlaminar fracture toughness under pure mode I
propagation (see Table 5), and according to Maimí et al. (2007b),
GUT
G2− can be obtained by the equation cos (653◦ ) (without considering
In the present work, the transverse tensile and the in-plane
friction between crack faces). On the other hand, G2+ and G2− for shear strengths for UT material are adjusted to take into account
WF material are assumed equal to the values of the other prin- the in-situ effect (Camanho et al., 2006), thus X2+ is and X is are
6
cipal direction. The same criterion has been considered to define calculated based on the ply thickness and the test strength val-
the other parameters associated with the fibre in direction 2 ( f G2+ , ues summarized in Table 2. It is worth mentioning that the equa-
fG2− , fX2+ and fX2− ). tions used for calculating X6is correspond to a linearised hardening
In addition to the material properties summarized in Table 2, shape as shown in Fig. 2(c) (Soto et al., 2017), thus they are dif-
both constitutive models require the parameters for the in-plane ferent from those formulated using a non-linear law as reported
shear plastic behaviour to be defined. For unidirectional plies, this in Camanho et al. (2006). The values obtained are summarized in
is formulated with a linear plasticity prior to damage and defined Table 3.
by the yielding shear stress X6P and the incremental stiffness un- Additionally, the constitutive model for woven fabrics requires
der plastic flow KP (see Fig. 2(c)). These parameters are fitted using defining non-dimensional coupling factors that describe the biax-
the experimental curves σ 12 (ε 12 ) of tensile tests on symmetrical ial strengthening behaviour (i.e. the shape of the failure envelopes)
± 45° laminates following the test method reported in AITM1-0 0 02 (Martín-Santos et al., 2014): η1T , which couples σ 11 and compres-
(1998). Since these curves showed a non-linear evolution from a sion σ 22 ; η1S , which couples σ 12 and compression σ 11 ; η2T , which
linear elastic loading to a linear hardening, the property X6P is couples σ 22 and compression σ 11 ; and η2S , which couples σ 12 and
adjusted as the cross point between those elastic and hardening compression σ 22 . In the present work, these parameters are de-
lines. The resulting values for the UT material are X6P = 94.53 MPa fined equal to zero because no experimental data was available and
and KP = 0.25. On the other hand, the constitutive model for wo- it was observed that their effects on the predictions were negligi-
ven fabric plies is formulated with a non-linear in-plane shear ble.
plasticity described by the following hardening function (Martín- When using shell elements together with a user-defined con-
Santos et al., 2014): stitutive model, the transverse shear stiffnesses should be de-
 T p 
fined as requested by the Abaqus software (ABAQUS version 6.12-
σ12 = X6P + ςE 1 − e|ςE ε12 | + ςL ε12
p
(2) 1, 2012). Their values are typically approached as K11 = 56 G13 h p ,
K22 = 56 G23 h p and K12 = 0. In the simulations reported in Section 5,
where X6P is the yielding stress and ς E , ςET and ς L are fitting pa- a separated layer of shell elements is modelled for each fibre ori-
rameters of σ 12 (ε 12 ) experimental curves (AITM1-0 0 02, 1998). By entation, albeit with the exception of the mid-plane plies which
identifying a yielding stress of X6P = 50.0 MPa, the values of the are modelled by two separated layers of shell elements (see see
fitting parameters obtained by non-linear least square method are: Section 4.2). Therefore, all the potential interfaces susceptible to
ςE = 89.6 MPa, ςET = 36.9 and ςL = 500.0 MPa. delamination are modelled, including the interface at the sym-
 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247 235

Table 5 under any loading mode. In addition, the constitutive model has
Interface properties of each material: Unidirectional Tape (UT) and Woven Fabric
a unilateral behaviour for mode I propagation, so that compres-
(WF).
sive displacements have no degradation and cohesive surfaces be-
Property UT WF have as a usual contact constraint. However, for shear propaga-
Strength XI [MPa] 55.47 55.47 tion modes the behaviour is fully symmetric, thus the damage
XII [MPa] 117.45 116.99 in one direction degrades in the opposite direction in the same
Fracture toughness GIc [N/mm] 0.62 1.01 manner. On the other hand, a quadratic stress-based criterion is
GI I c [N/mm] 1.82 2.45 adopted for damage initiation (Ye, 1988) and the Benzeggagh and
G20%c [N/mm] 0.61 1.07 Kenane (BK) energy-based criterion is selected for damage propa-
G50%c [N/mm] 0.80 1.55 gation (Benzeggagh and Kenane, 1996).
G80%c [N/mm] 1.22 1.77
η 3.00 1.88
The interlaminar properties required to define the cohesive
surfaces are summarized in Table 5. The critical energy release
rates were obtained through standard tests: pure mode I GIc , us-
Table 6
ing the Double Cantilever Beam (DCB) test (AITM1-0 0 05 , 2014);
Number of elements used for the description of the fracture
process zone length for each laminate. pure mode II GIIc using Calibrated End-Loaded Split (C-ELS) test
(ISO15114 , 2014); and mixed propagation modes (at 20%, 50% and
Laminate Mode I Mode II
80%) using the Mixed-Mode Bending (MMB) test (ASTM D 6671/D
LUT1 0.23 1.01 6671M-13, 2007). A sample of five specimens was tested for each
LUT2 0.34 1.29
loading mode and material. The η parameter of the BK propagation
LWF 0.48 1.47
criterion for each material system was approached by the least-
square fit of the corresponding critical energy release rates. Follow-
Table 7 ing the standards, these tests were performed on laminates with a
Analysis computational time (data from Abaqus Status analysis file). The step time,
single orientation, thus on fracture planes with the same orienta-
the CPU time and the displacement reported for CAI simulations correspond to the
values at failure load. tion as the adjacent plies. However, the interfaces considered for
delamination in the predictions reported in Section 5 have a mis-
Laminate Case LVI CAI
match angle of 45°. Therefore, this issue may have an effect on the
Step [ms] CPU [min] Step [ms] Disp. [mm] CPU [min] quality of the LVI and CAI predictions when comparing them to the
LUT1 IC20 4.0 525 5.8 0.87 590 experimental data.
IC30 5.5 891 5.8 0.88 1124 On the other hand, since the same matrix is used for both UT
LUT2 IC20 5.0 297 4.9 0.72 214 and WF materials, the strengths for the onset of pure mode I de-
IC30 5.0 375 5.4 0.80 452
lamination XI are considered as being equal to the transverse ten-
LWF IC30 8.0 452 6.1 0.91 317
sion strength of the UT material X2+ UT . In addition, the strengths for
IC40 8.0 587 5.7 0.85 289
IC50 8.0 475 5.7 0.85 377 the shear modes II and III for each material are taken as being the
corresponding in-plane shear strength X6 . It is worth noting that
neither XI nor XII are defined with an in-situ effect adjustment.
metry plane. Accordingly, the transverse shear stiffnesses are: for According to Turon et al. (2010), some interlaminar models
E2UT available in the literature have not been validated correctly under
LUT1 plies, knowing GUT = , and assuming GUT = GUT and
23 2 (1+ν23
UT
) 13 12 mixed-mode loading conditions. A given selection of the interlami-
ν23
UT = 0.45, yield to K UT = 961.24 N/mm and K UT = 589.61 N/mm;
11 22
nar strengths can result in inaccurate simulations, yielding a wrong
for LUT2 plies, the transverse shear stiffnesses are twice the values prediction of the dissipation attributed to a mixed-mode propaga-
obtained for LUT1 plies; for WF plies, assuming GW 13
F = GW F and
23
tion. For the built-in cohesive surfaces of Abaqus/Explicit, the prop-
W F UT
G13 = G12 (the same matrix is used for UT and WF materials), re- agation of the delamination under mixed-mode conditions can be
W F = K W F = 1373.80 N/mm.
sult in K11 dependent on the interlaminar strength values, and inaccuracies in
22
Finally, according to the description given in Section 2, the def- the prediction of the delaminations may arise.
inition of the density on the surface elements is requested when For the complete definition of the CL, a penalty stiffness Ki
setting the contact on these elements. Certainly, the proper defini- (i = I, II, III) is requested for each propagation mode because defin-
tion of this property should yield to the corresponding total mass ing an infinite stiffness is not feasible in an FE model. For both
of the specimen, therefore, the density must be extracted from the UT and WF materials, and for any propagation mode, these stiff-
conventional shell elements. Although a significant effect on the nesses are approached by Ki = 10E2UT , thus Ki = 8.31 × 104 N/mm3
quality of the predictions resulting from changes in mass distri- (Turon et al., 2007).
bution percentages was not observed, the predictions reported in It is worth noting that the density and penalty stiffnesses of
Section 5 correspond to a definition of 25% of the mass for each of the surface elements should be defined carefully to avoid the asso-
the ply surface elements, and then 50% for the conventional shell ciated stable time increments being smaller than those computed
element. In accordance with the ply material densities defined in for conventional shell elements. In the present work, this issue was
Table 2, the corresponding FE densities are summarized in Table 4. checked by choosing the best model based on the computational
analysis time and the accurateness in predicting the impact elastic
4.1.2. Interface modelling loading response of a set of models with different values of penalty
As noted in Section 2, the interfaces for delamination are mod- stiffness (keeping the percentage of the mass distribution for sur-
elled using the cohesive surfaces of Abaqus/Explicit (ABAQUS ver- face elements and shell elements constant).
sion 6.12-1, 2012) which describe the mechanical interaction be- Additionally, a friction coefficient of μ = 0.3 is defined at any
tween two surfaces by means of a contact algorithm (based on a surface element. The friction behaviour is governed by Coulomb’s
penalty stiffness method and known as the General Contact algo- law and, once again, a penalty contact stiffness is requested. This
rithm). A linear Cohesive Law (CL) is selected to describe the pro- is herein defined as that considered for the cohesive behaviour. Al-
gressive degradation. This is considered accurate enough to pre- though the cohesive and friction models are formulated separately,
dict the delamination (Alfano, 2006). The formulation is based on it was checked that the shear cohesive behaviour (σIIcoh ) is coupled
an isotropic damage variable which accounts for the degradation with the shear friction when out-of-plane compression loadings
236 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247
Fig. 3. LVI predictions and comparisons with experimental data: laminate LUT1. All the results are shown without filtering.
are applied (σ I ). Accordingly, the whole shear response is given by
σII = σIIcoh − μσI .
4.2. Mesh of the specimens
In the present work, each fibre orientation, with the exception
of the mid-plane cluster of plies which is described by two sepa-
rated layers of shell elements, is modelled with a separated layer
of shell elements. Therefore, all the interfaces with mismatch ori-
ented surrounding plies are modelled and the plies of LUT1, LUT2
and LWF laminates are described by 16, 8 and 12 shell layers, re-
spectively.
4.2.1. FE types and sizes
A 50 mm × 80 mm rectangular in-plane area centred at the
impact site is modelled with a refined mesh of square-shaped el-
ements with  = 1.5 mm edges for any material system and thick-
ness (the dimension of 50 mm belongs to the direction of the
longest edge of the specimens, i.e. 150 mm). Moreover, each re-
fined mesh area is aligned according to the corresponding ply ori-
entation to mitigate possible dependences of the development of
damage patterns on the orientation of the finite elements. The
meshes outside of the refined area are described by rectangular-
shaped elements with a maximum edge size of  = 4.0 mm. The fi-
nite element chosen, called S4R, is a 4-node general-purpose shell
with reduced integration, hourglass control and finite membrane
strains (ABAQUS version 6.12-1, 2012). Additionally, S3R shell ele-
ments (3-node) are used in those locations where S4R elements
cannot be defined, i.e. at the boundaries of the refined mesh, ori-
ented at 45°, with the coarse mesh (always oriented at 0°).
The characteristic element size for an oriented mesh with the
√
corresponding fibre direction is calculated by  = 1 2 , where
1 and 2 are the in-plane dimensions of the finite element. The
maximum element size that prevents snap-back of the constitutive
softening branch at the corresponding loading direction is obtained
by equalling the elastic energy of the element with the energy dis-
sipated by fracture, giving the following expression:
i = 2Ei Gi Xi−2 (3)
where i = 1+, 1−, 2+, 2− and 6, and E6 = G12 (i = 6 only for UT
material). For the bi-linear constitutive law of the fibre, a maxi-
mum element size can be predicted for partial snap-back in the
first si 1 , and in the second si 2 , softening branches (where i =
1+, 1− for UT material; i = 1+, 1−, 2+ and 2− for WF material).
The corresponding expressions are:
 −1
si 1 = 2Ei fGi Gi Xi2 − fXi Xi2 (4)
 −1
si 2 = 2Ei (1 − fGi − fGi fXi )Gi fXi
2 2
Xi (5)
According to the material properties listed in Table 2 and the
in-situ strengths in Table 3, the smaller element size for UT ma-
terial is defined by the transverse tensile loading which gives a
value of 2+ = 0.39 mm and 2+ = 0.77 mm for embedded plies
of one and two clustered plies, respectively; and for four clustered
plies, the smaller element size is defined by in-plane shear failure
mode which yields to a critical size of 6 = 0.83 mm. As can be
observed, the element size used at the refined mesh area is signif-
icantly bigger than these critical sizes. Faced with that situation,
the implemented user-defined material subroutine automatically
 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247 237

Fig. 4. LVI predictions and comparisons with experimental data: laminate LUT2. All the results are shown without filtering.

Fig. 5. LVI predictions and comparisons with experimental data: laminate LWF. All the results are shown without filtering.
238 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247

Fig. 6. Projected delamination areas for LUT1 laminate. Axes in millimeters.

Fig. 7. Projected delamination areas for LUT2 laminate. Axes in millimeters.

Fig. 8. Projected delamination areas for LWF laminate. Axes in millimeters.

reduces the material strength for those critical failure modes but
keeps the corresponding
 critical energy release rate constant, so
that Xired = 2Ei Gi −1 . In the refined mesh area, the failure modes
affected for any ply clustering thickness are the transverse tension
and the in-plane shear; at the coarse meshes, all the failure modes
are affected except those of the fibre, i.e. 1+ and 1−.
On the other hand, since WF material is governed by fibre
breakage, the maximum element size that prevents complete or
partial snap-back is calculated by  ≤ min(i , si 1 , si 2 ) as done for
UT material in the longitudinal direction. In contrast with UT ma-
terial, the maximum element size is less critical since the resulting
maximum element size is 13.19 mm, which is considerably larger
than the FE size selected in both the refined and coarse mesh ar-
eas.
According to the modelling strategy presented, surface elements
are tied below and above the shell elements so that they define
the whole thickness of the corresponding ply or of a sub-laminate.
The surface element selected is called SFM3D4R, which is a 4-
node quadrilateral element with reduced integration. Additionally,
SFM3D3 type elements (3-node) are used in those locations where
SFM3D4R cannot be defined. The size of the surface elements is
defined here as equal to that of the linked shell elements. Us-
ing the closed-form solutions proposed by Soto et al. (2016) for
predicting the fracture process zone length for pure propagation
modes I and II of the delamination and the interface properties
summarized in Table 5 for each material system, the number of
elements that describe each propagation mode can be predicted.
In the refined mesh area, the corresponding fracture process zones
are described by the number of elements summarized in Table 6. It
is worth noting that the description is quite poor for both modes,
particularly for mode I. In a similar way to that explained for
intralaminar damage mechanisms, the corresponding interlaminar
strength can be reduced in order to increase the fracture process
zone, thus larger elements can be used for a suitable description
of the delamination. However, in the present simulations this ap-
proach was not taken since the predominant propagation mode in
 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247 239

Fig. 9. Damage field outputs of LUT1 laminate at 30 J: left, projected delamination area; middle, fibre damage; right, matrix cracking.

an LVI is rather mode II, and it was observed that using around
one element for the description of the fracture process zone was
accurate enough to simulate LVI and CAI tests.
4.2.2. FE deletion
When simulating in an explicit FE code, excessive distortion
of degraded elements can occur and Abaqus/Explicit automatically
stops the analysis. A usual practice is to remove the degraded ele-
ments from the mesh when a damage variable, commonly the vari-
able linked to the fibre, has reached a predefined threshold close
to the unit. However, as reported in Soto et al. (2018), these defini-
tions should be made carefully because the constitutive law shape
can be considerably modified and, in turn, the energy dissipated at
those FE material points.
For the LVI and CAI simulations of laminates manufactured with
UT material, no element deletion has been considered. Instead,
the damage variables have been defined with a maximum value
of 0.999 for longitudinal damage d1 , and 0.99 for transverse d2
and in-plane shear damage d6 . Accordingly, once the element has
been degraded a residual constant stiffness is used at each load-
ing mode, so that if the strain increases, the corresponding stress
also increases and some elastic energy is absorbed at those de-
240 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247

Fig. 10. Out-of-plane displacement field outputs of LUT1 laminate at 30 J (legend in millimeters): left, impacted face; right, non-impacted face.

graded elements. Alternatively, a permanent residual stress can be
considered once the element has been degraded, however this def-
inition gave rise to difficulties of excessive distortions and so was
discarded.
On the other hand, for laminates with WF material, the lon-
gitudinal damage variables d1 and d2 have been defined with a
maximum value of 0.9999 without element deletion, and thus
a residual stiffness is defined at each principal direction as was
done for the UT material. Accordingly, the maximum value of
the damage variable included for in-plane shear behaviour can be
closer to the unit as it is computed by d6 = 1 − (1 − d1 )(1 − d2 ).
Moreover, in the experimental tensile tests on ± 45° specimens
(AITM1-0 0 02, 1998), it was observed that all the specimens failed
around ε12 = 0.05. In the present work, the constitutive model
is defined such that maximum strain is reached, the stiffness
at any direction of the material is dropped to a residual value
and, once again, there is no element deletion. It was found that
these two considerations related with in-plane shear behaviour
considerably improved the quality of the LVI and CAI predictions
(Soto et al., 2018).
 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247 241

Fig. 11. Damage field outputs of LWF laminate at 30 J: left, projected delamination area; middle, fibre damage d1 ; right, fibre damage d2 .

4.3. LVI assembly and test boundary conditions
As done in previous simulation works by the authors
(González et al., 2012), the LVI and CAI FE models replicate the
experimental test set-ups which correspond to the definitions re-
ported in the standards ASTM D 7136/D 7136M-12 (2012a) and
ASTM D 7137/D 7137M-12 (2012b) standards, respectively. In the
LVI tests, the specimen is impacted with a 16 mm diameter hemi-
spherical impactor, and clamped using four rubber points and a
rectangular base support with a centred cut-out. The rubber points
have a 7.0 mm radius footprint, and the base support is mod-
elled with an external size of 200 mm × 150 mm and a centred
cut-out of 125 mm × 75 mm. All these parts are modelled us-
ing R3D3 (3-node triangular facet) and R3D4 (4-node rectangular
faced) 3D rigid elements and the corresponding boundary condi-
tions are: fully clamped for the base support; all degrees of free-
dom of the four clamping points are constrained except the vertical
displacement, which is prescribed with 2.5 × 10−3 mm against the
specimen (so that a preloaded clamping area is modelled); and the
impactor is prescribed with an initial vertical velocity just prior to
impact, defined according to the corresponding impact energy and
the impactor mass (see Table 1). The maximum FE edge size of
242 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247

Fig. 12. Out-of-plane displacement field outputs of LWF laminate at 30 J (legend in millimeters): left, impacted face; right, non-impacted face.

each these parts is 2.5 mm for the base support, 1.0 mm for the
rubber clamps, and 0.3 mm for the impactor.
As shown in Fig. 1, a surface thickness is defined at each sur-
face element. In the simulations reported in Section 5, the surface
thickness has been added to the real thickness of the specimens.
Therefore, knowing that the surface thickness is approached by
hst = 0.01h p for all laminates, the whole modelled laminate thick-
ness is slightly increased by 1%. Additionally, the same surface
thickness has been defined for all the rigid bodies in the LVI as-
sembly (impactor, base support and the four clamping points). It
was found that the definition of a surface thickness on those parts
gave less spurious oscillations in the predictions since the contact
between part surfaces is better tracked.
On the other hand, contact interactions between all the assem-
bly parts and on the specimen itself have been defined using the
General Contact algorithm from Abaqus/Explicit (ABAQUS version
6.12-1, 2012). This is the simplest way to define the contact do-
mains and requires defining a contact stiffness, which is selected
to be linear and with the same value as that of the cohesive inter-
actions. In addition, a friction coefficient was defined to model the
tangential behaviour between the contacting element faces, which
is again defined as that for cohesive surfaces: μ = 0.3.
 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247
Finally, the CAI test simulation is created by importing the as-
sembly and the results of the LVI simulation using the Restart com-
mand in Abaqus/Explicit (ABAQUS version 6.12-1, 2012), and the
corresponding boundary conditions are defined in the same man-
ner as done in (González et al., 2012). Prior to performing the CAI
simulation, the rigid parts are moved far away from the specimen
in a preliminary loading step, and afterwards new boundary condi-
tions are defined according to ASTM D 7137/D 7137M-12 (2012b).
5. FE predictions and comparison with experimental data
In this section, the numerical predictions of the LVI and CAI FE
models are compared with the experimental results. The FE mod-
els were generated automatically using a parameterized Python file
and the analyzes were run in Abaqus/Explicit v6.12 in single pre-
cision. The features of the computer used are: two socket work-
station (ASUS Z10PE-D16 WS Motherboard) with two 3.1 GHz CPU
Intel® Xeon® Processor E5-2687W v3 (10 cores, 160 W power, and
2133 DDR4 type memory), and a 512 GB solid-state disc and 32 GB
RAM for each CPU. Although the capabilities of the computer are
high, only 10 cores were used for each analysis. The resulting com-
putational times for each simulation are summarized in Table 7.
5.1. LVI predictions
Figs. 3–5, show the FE predictions for the impact load versus
the impactor displacement and the evolution of the energy ab-
sorbed by laminates LUT1, LUT2 and LWF, respectively. As can be
observed, most of the predictions are in good agreement with the
experimental data although some differences can be detected. For
most of the cases of LUT1 and LUT2 laminates, the dissipated en-
ergies (height of the energy evolution plateau) are underpredicted.
This trend is more obvious for LUT2 laminate and this can be re-
lated to the fact that the growth of a failure mechanism is not
well captured. If the projected delamination areas are also checked
as reported in Figs. 6 and 7 for LUT1 and LUT2, respectively, the
delaminations are also clearly underpredicted, and give circular-
shaped areas without capturing the asymmetry and the extension
of the experimental shapes. This scant prediction of the delamina-
tion means the numerical dissipated energies are smaller than the
experimental values. The main reason for these deviations can be
attributed to a limited prediction of the interaction between the
delamination and the intralaminar failure mechanisms (basically,
matrix cracking), and this is easily traceable since the delamina-
tion in the last interface, which gives its shape due to a long matrix
crack in the last ply, has not been captured correctly. In addition,
the deviations of the projected delamination area could also be re-
lated to other reasons such as: the value used for the interlaminar
shear strength XII , which is approached equal to X6 , may be inap-
propriate; the propagation of the delamination under mixed-mode
conditions may be dependent on the interlaminar strength values
(Turon et al., 2010); the value of the friction coefficient consid-
ered may have been defined too high so that the delamination in
mode II is significantly restrained; the interlaminar critical energy
release rates used were obtained from fracture tests on 0°/0° in-
terface specimens, whereas the interfaces modelled have 45° mis-
match angles of the surrounding plies. As an example, the effect
XII has is illustrated in Fig. 7(b) where the projected delamina-
tion area of an LVI simulation test with a 20% reduced interlam-
inar shear strength has been plotted (with legend XIIred ). Accord-
ingly, the material property X6 is also reduced and the correspond-
ing adjustment with the in-situ effect is accounted for. As can be
observed, the reduction of XII significantly improves the prediction
of the projected delamination area and, as will be seen later, also
the prediction of the CAI strength.
243
Table 8
Experimental and predicted post-impact residual compressive load for laminates
LUT1 and LUT2. The result with (∗ ) refers to the case with a reduced interlaminar
shear strength XIIred .
Laminate Test Load [kN]
Experimental Numerical
LUT1 IC20 96.75 105.44
IC30 88.15 103.85
LUT2 IC20 89.42 87.38
IC30 79.19 94.95; 84.79∗
For laminate LWF, the values of the impact load at the early
stages of the charts shown in Fig. 5 are higher than the experi-
mental values, so the bending stiffness of the experimental results
are smaller when a certain threshold is reached. This behaviour
can be related to the prediction of the delamination onset, which
is delayed at higher impact loads. Consequently, the apparition of
the following fibre breakage failure can be affected and can occur
at smaller impactor displacements. Nevertheless, the complete pro-
files of the impact load for each impact energy have been reason-
ably well captured in comparison with the experimental results.
On the other hand, Fig. 8 shows the projected delamination areas
which are quite rounded and have the same size for all impact en-
ergies. Although the same friction coefficient is defined as in the
laminates with UT material, the prediction of the dissipated ener-
gies and the projected delamination areas are in good agreement
with the experimental data. These results suggest that the appari-
tion of the failure mechanisms occur in quite a sequenced form.
Once the delaminations are onset, they grow until the maximum
size shown in Fig. 8 and afterwards the structure response is sim-
ply governed by the fibre breakage.
An important measurement used to define the allowable
strengths for the CAI of a composite laminate is the permanent in-
dentation, which is related to an impact energy and therefore has
a residual strength (see AITM1-0 010, 20 05). Since the intralami-
nar constitutive model implemented takes into account in-plane
shear plasticity and the FE model includes contacts with friction,
a permanent indentation can be read. As an example, the first two
images of Figs. 10 and 12 show, respectively, the out-of-plane dis-
placement of laminates LUT1 and LWF just after an LVI event of
30 J. The predicted dent depth located just at the impact site are
0.18 mm and 0.19 mm for LUT1 and LWF respectively, whereas the
experimental values measured within 30 min after the impact are
0.65 mm and 0.78 mm. Assuming that these experimental values
can be reduced by 70% due to material relaxation (AITM1-0010,
2005), the values of the predictions are similar to the experimental
ones. However, the finite elements used assume plane stress state
and often this is not enough for other laminates and impact con-
figurations. For an accurate prediction of the dent depth, the use of
a 3D stress state FE model with an adequate constitutive damage
model plus plasticity in certain loading directions is recommended
(e.g. Catalanotti et al., 2013; Vogler et al., 2013; Camanho et al.,
2013).
5.2. CAI predictions
The numerical and the experimental maximum loads for the
CAI tests are summarized in Tables 8 and 9 for laminates with
unidirectional tape and woven fabric materials, respectively. For
LWF laminates, where two experimental tests were performed, the
mean values are reported. As can be observed, most of the numer-
ical results overpredict the experimental values. The maximum dif-
ference found for woven fabric laminates is 26.0% (IC50), whereas
for LUT1 and LUT2 laminates these are 17.8% (IC30) and 19.9%
(IC30), respectively. In the case of laminate LUT2 (IC30) with a re-
244 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247
Table 9
Experimental and predicted post-impact residual compressive load for laminate
LWF.
Laminate Test Load [kN]
Experimental
LWF IC30 90.53
IC40 84.95
IC50 81.10
duced interlaminar shear strength, the improvement in the predic-
tion of the residual load is significant, where the difference with
the experimental load is reduced to 7.1%.
The differences detected can be associated to different reasons.
From an experimental point of view, only one or two specimens
were tested which are quite poor samples for CAI tests. In addi-
tion, the CAI boundary conditions of the support fixture side plates
(i.e. vertical side knife supports) were applied with an excessive
compressive load against the specimens by using fastening screws.
These prestress boundary conditions are not defined in the CAI
standard (ASTM D 7137/D 7137M-12, 2012b) and they can induce
a premature failure of the specimens. Moreover, the manufacturing
process used for the specimens presented can induce a significant
heterogeneity of the material properties which was not accounted
for in the FE models. Finally, in CAI tests there are often certain
load misalignments which are not modelled but they can signif-
icantly modify the CAI tests predictions. From a numerical point
of view, and as noted previously, the prediction of the delamina-
tion in the LVI tests was underpredicted for laminates manufac-
tured with UT material, so it makes sense that the CAI maximum
loads are overpredicted.
An interesting result that can be analyzed using the CAI simu-
lations performed is the sequence of damage that occurs, includ-
ing local/global buckling, right on and just after the CAI load col-
lapses. As illustrative examples, Figs. 9 and 11 depict the damage
field outputs with applied translucency of laminate LUT1 at 30 J
(projected delamination area, fibre breakage and matrix cracking)
and laminate LWF also at 30 J (projected delamination area and
fibre damage at both principal directions d1 and d2 ), respectively.
Additionally, Figs. 10 and 12 show the out-of-plane field output for
those tests. In both examples, the field outputs are shown for just
after the LVI test, and right on and just after the ultimate failure.
For laminate LUT1, a damage growth of matrix cracking and fibre
breakage can be observed. Once the maximum load is reached, a
sudden growth of all failure mechanisms over the entire width of
the specimen is observed. Local buckling concentrated around the
impact site (see Fig. 10) can also be observed, and grows at the
same time that the damage mechanisms evolve. On the other hand,
for laminate LWF a growth of fibre breakage (specially for d1 ) can
be appreciated just at the maximum load, with a simultaneous
small growth of the projected delamination area. Again, once the
structure fails, any damage occupies the whole width of the spec-
imen (because of that, it is recommendable to use a refined mesh
that includes the entire width of the specimen). However, in this
case significant local buckling is not appreciated.
It is worth noting that the predicted damage sequence observed
for each laminate type (compressive failure at the impact site cov-
ering the entire width of the specimen, with a noticeable local
buckling for LUT but not as pronounced for LWF), match the ex-
perimental observations.
6. Conclusions
An FE modelling strategy was presented to simulate LVI and
CAI tests on polymer-based reinforced composite laminates using
Abaqus/Explicit FE software. The strategy was based on using con-
ventional shell elements and cohesive surfaces so that the detri-
mental effect of the ply thickness on the stable time increment is
avoided. Therefore, the strategy presented improves the computa-
tional time of the analysis in comparison to common simulations
Numerical based on 3D solid elements, or continuum shell elements, together
113.70
with finite thickness cohesive elements. The proper out-of-plane
104.91 structural response was solved by considering surface elements lo-
102.16 cated on the bottom and top faces of the layers and tied to shell
elements modelled at the mid-plane layer thickness.
The key parameters required to define the FE models were con-
cisely described and the values selected were well justified. These
parameters are often not well explained in the literature but they
can have a significant role on the accurateness of the simulations.
As a novelty, the way to approach the constitutive law shape and
the corresponding critical energy release rates of the fibre has been
described for both unidirectional tape and the woven fabric mate-
rials considered.
The accuracy of the modelling strategy has been proved by sim-
ulating LVI and CAI tests on monolithic and rectangular laboratory
coupons with two different fibre material and architectures and
three different stacking sequences. The predictions of the impact
load and the displacement evolutions of laminates manufactured
with unidirectional tape revealed good agreement with the exper-
imental data. However, the evolution of the energy dissipated, to-
gether with the projected delamination area, showed an underpre-
diction of the delamination. Some justifications were pointed out,
where the main reason was linked to a limited interaction between
interlaminar and intralaminar damage mechanisms. On the other
hand, for the woven fabric material, the LVI predictions showed
a stiffer response at the early stage of the impact response which
was associated to a delayed prediction of the delamination onset at
higher impact loads. Finally, most of the predicted CAI maximum
loads for both type materials were above the experimental data.
Reasons to justify these not matching CAI results, basically derived
from the experimental tests, could be: boundary conditions, het-
erogeneity of the material properties, and/or load misalignments.
From the work presented, some future research can be sug-
gested. For the modelling strategy, an optimum mass distribu-
tion of surface elements and shell elements could be found, with
an equilibrium between the analysis’ computational time and the
accurateness of the predictions. In addition, a time performance
comparison with other modelling strategies based on 3D solid el-
ements or continuum shell elements for the same test configura-
tions would be of interest. On the other hand, from an experimen-
tal point of view, a better characterization of the friction coefficient
for interfaces with different oriented surrounding plies, as well as
a suitable characterization of the corresponding critical energy re-
lease rate, may be of interest. Finally, from the point of view of the
FE model, considering a statistical distribution of the heterogeneity
of the material properties (elastic and fracture), together with the
use of an improved Cohesive Law formulation so that the propaga-
tion under mixed-mode condition is not strength dependent, may
increase the accurateness of the predictions reported.
Acknowledgements
This work has been partially funded by European Regional De-
velopment Funds (ERDF), through the Spanish Government’s (Min-
isterio de Economía y Competitividad) Centro para el Desarrollo Tec-
nológico Industrial (CDTI) which in June 2010 approved the project
entitled Demostrador de un grid para actividades de simulación con
requerimientos extremos de cálculo. The first author would like to
thank the Spanish Government (Ministerio de Economía y Competi-
tividad) for their financial support under contract MAT2015-69491-
C3-1-R; the second and fourth author would like to thank the
Spanish Government (Ministerio de Economía y Competitividad) for
 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247 245

their financial support under contract MAT2013-46749-R. Finally, rotations θ pi (where i = x, y, z). To take these rotations into consid-
the fourth author would like to acknowledge the predoctoral grant eration the displacements at the points of the interfaces ui should
BES-2014-070374 from the Subprograma Estatal de Formación del be calculated according to Eq. (A.1) (Dávila et al., 20 07; 20 08). In
MICINN. the present strategy, this issue is solved using the Tie connectors
between the conventional shell and the surface elements.
Appendix A. Example: three point bending test
⎡ ⎤
u px
    u py ⎥
To illustrate the need to define surface elements and Tie con- ux 1 0 0 0 ±h p /2 0 ⎢
⎢u pz ⎥
nectors when using mid-plane located conventional shell elements uy = 0 1 0 ∓h p /2 0 0 ⎢
⎢θ ⎥
⎥ (A.1)
uz 0 0 1 0 0 0 ⎣ px ⎦
and cohesive surface interactions, a simple example of a three- θ py
point bending test of an isotropic beam is simulated. The beam θ pz
is modelled using two shell elements located in the mid-plane
thickness of each arm connected by a cohesive surface. The beam’s Appendix B. Fibre fracture toughness
properties are: h = 2 mm thick (thus h p = 1 mm), L = 50 mm long
and b = 10 mm wide, and the elastic properties are E = 120.0 GPa The method reported in Ortega et al. (2016) assumes that all
and ν = 0.3. Using simple beam
 theory,
 the
 analytical vertical dis- the dissipation mechanisms can be represented as if they are con-
placement is given by u3 = F L3 / 4Ebh3 , which yields a max- fined to a crack plane which still transfers stresses between its
imum displacement of 0.33 mm for a concentrated load of F = faces as the crack opens. The relationship between the transferred
100 N. Two models are simulated: with and without surface el- stress and the crack opening is known as the Cohesive Law (CL), i.e.
ements. In both cases the cohesive interaction is simply defined σ (w ). This method measures the CL, or the equivalent J -integral
by an elastic slope without damage. The maximum out-of-plane curve, from a single load displacement curve of in-plane quasi-
displacements predicted with and without surface elements are isotropic laminates in CT and CC specimens based on the follow-
0.335 mm and 1.308 mm, respectively. The result given without ing work: first, a semi-analytic model was developed to obtain
surface elements is completely wrong because the correct bend- the load displacement curve for any given CL shape (Ortega et al.,
ing response has been overpredicted by four times and this result 2017b) and second, the laminate CL was obtained with the semi-
corresponds to two separated boards that slide past each other (in analytic model, by best-fitting some selected points of the experi-
other words, without shear transfer among them). Therefore, the mental load displacement curve (Ortega et al., 2016). The data re-
FE software determines the relative displacements at the interfaces ductions from the CT and CC tests of the laminates considered in
using the displacements at the shell nodes upi , and not with their Section 3 are shown in Figs. B.13 and B.14, respectively (the mean

Fig. B13. Data reduction of Compact Tension tests for each laminate: (a) Cohesive Laws and (b) J -integral curves (Ortega et al., 2017a).

Fig. B14. Data reduction of Compact Compression tests for each laminate: (a) Cohesive Laws and (b) J -integral curves (Ortega et al., 2017a).
246 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247
values of the displacement and the load test data of a sample of
two specimens were used for each loading case and material).
As reported in Ortega et al. (2017a), the critical energy release
rate of the laminate can be defined as a combination of the corre-
sponding values of each ply. This is given by the following mixing
equation:
n
h p,i a p,i
J = J p,i (B.1)
h a
i compscitech.2004.12.024.
where J p,i is the dissipated energy of a single ply (i refers to
ply number), hp, i /h is the relative thickness of each ply with re-
spect to the whole laminate thickness, and ap, i /a is the ratio be-
tween the dissipated energy per unit crack length at the ply level
with respect to the laminate. For LUT laminates, the ratio ap, i /a
√ Compression Strength after Impact, Blagnac Cedex, France.
can be considered equal to 0 for 90° plies, 1 for 0° √ and 1/ 2
for ± 45°, thus the property G1+ is approached with 4( 2 − 1 )J .
As can be observed in Fig. B.13(b), the maximum value of J for
both laminates LUT1 and LUT2 is around 150.0 N/mm, then G1+ =
248.5 N/mm. Moreover, a change in the slope can be observed
around the points w = 0.10 mm and J = 90.0 N/mm for LUT1, and
w = 0.11 mm and J = 101.0 N/mm for LUT2. Using these points,
the ratio of the critical energy release rate to define the first soft-
ening branch in tensile fibre failure fG1+ is approached to 0.60
and 0.67 for one ply and for two clustered plies, respectively (see
Fig. 2(a)). For the results under compression (Fig. B.14), it can be
observed that the J -curves grow without reaching a plateau and
the corresponding CL’s show quite a permanent residual stress at
the largest crack openings (linked to the kink-band broadening),
thus G1− is infinite for both LUT laminates. Accordingly, G1− is
defined with a large value as noted in Table 2. In addition, the
ratio of the critical energy release rate under compression f G1−
is approached using the point w = 0.15 mm and J = 40.0 N/mm
for both LUT laminates, thus fG1− G1− = 66.3 N/mm. Finally, con-
sidering the stress after the first large drop in the laminate’s CL
(Figs. B.13(a) and B.14(a)), and the corresponding fibre strengths,
the strength ratios for a full definition of the inflection softening
points are 0.17 for tension fX1+ and compression fX1− , and for
both laminates LUT1 and LUT2.
For the WF material, the ratio ap, i /a is also considered equal
to 1√ for 0° plies but for 45° plies, it must be defined as equal
to 2. Therefore,√the tensile critical energy release rate is de-
fined by G1+ = 2( 2 − 1 )J thus giving G1+ = 91.0 N/mm using the
plateau value of the J -curve. To adjust the dissipated energy ra-
tio at the first tension softening branch,2 the point w = 0.1 mm
and J = 70.0 N/mm is used, thus fG1+ = 0.64. In addition, the crit-
ical energy release rate for fibre compression G1− is again infi-
nite, and the corresponding ratio which defines the area enclosed
by the first softening branch must be defined so that it satisfies
fG1− G1− = 33.0 N/mm. The ratios for the stress drop definition are
selected using the same criterion as noted for the UT material, thus
the resulting values are 0.19 for both tension and compression.
2
In Martín-Santos et al. (2014), the dissipated energies at the first softening
branch are defined using the slopes of these lines. Therefore, four slopes are re-
quested: Hi (where i = 1+, 1−, 2+ and 2−). In contrast to the model described in
Martín-Santos et al. (2014), the dissipated energies obtained by f Gi Gi are considered
here as the areas enclosed by the first triangle of the Cohesive Law constructed
using the line that goes from the inflection point of the two softening branches
to the origin of the Cohesive Law (instead of considering the area enclosed by the
first softening branch starting from the material strength until zero stress). There-
fore, the dissipated energies f Gi Gi reported here are related to those of Martín-
Santos et al. (2014) by multiplying them by Xi /(Xi − f Xi Xi ).
References
Abdulhamid, H., Bouvet, C., Michel, L., Aboissiere, J., Minot, C., 2016. Numerical sim-
ulation of impact and compression after impact of asymmetrically tapered lam-
inated CFRP. Int. J. Impact Eng. 95, 154–164. doi:10.1016/j.ijimpeng.2016.05.002.
Abir, M.R., Tay, T.E., Ridha, M., Lee, H.P., 2017. Modelling damage growth in com-
posites subjected to impact and compression after impact. Compos. Struct. 168,
13–25. doi:10.1016/j.compstruct.2017.02.018.
Abrate, S., 1998. Impact on Composite Structures. Cambridge University Press, Cam-
bridge, UK.
Alfano, G., 2006. On the influence of the shape of the interface law on the ap-
plication of cohesive-zone models. Compos. Sci. Technol. 66 (6). doi:10.1016/j.
Arteiro, A., 2016. Structural Mechanics of Thin-Ply Laminated Composites. Porto
Ph.D. thesis.
AITM1-0 0 02, 1998. Airbus Test Method. Fibre Reinforced Plastics: Determination of
In-plane Shear Properties (+− 45° Tensile Test), Blagnac Cedex, France.
ASTM D 3039/D 3039M-0 0, 20 0 0. Standard Test Method for Tensile Properties of
Polymer Matrix Composite Materials. ASTM International, West Conshohocken
PA, USA.
AITM1-0 010, 20 05. Airbus Test Method. Fibre Reinforced Plastics: Determination of
ASTM D 6671/D 6671M-13, 2007. Standard Test Method for Mixed Mode I-mode
II Interlaminar Fracture Toughness of Unidirectional Fiber Reinforced Polymer
Matrix Composites. ASTM International, West Conshohocken PA, USA.
ABAQUS version 6.12-1, 2012. USER’s Manual.
ASTM D 7136/D 7136M-12, 2012a. Standard Test Method for Measuring the Damage
Resistance of a Fiber-Reinforced Polymer Matrix Composite to a Drop-Weight
Impact Event. ASTM International, West Conshohocken PA, USA.
ASTM D 7137/D 7137M-12, 2012b. Standard Test Method for Compressive Residual
Strength Properties of Damaged Polymer Matrix Composite Plates. ASTM Inter-
national, West Conshohocken PA, USA.
AITM1-0 0 05, 2014. Airbus Test Method Carbon Fibre Reinforced Plastics Determina-
tion of Mode I Fracture Toughness Energy of Bonded Joints (G1c test), Blagnac
Cedex, France.
Bažant, Z.P., Oh, B.H., 1983. Crack band theory for fracture of concrete. Matériaux et
Construct. 16 (3), 155–177.
Benzeggagh, M.L., Kenane, M., 1996. Measurement of mixed-mode delamina-
tion fracture toughness of unidirectional glass/epoxy composites with mixed-
mode bending apparatus. Compos. Sci. Technol. 56 (4), 439–449. doi:10.1016/
0266-3538(96)0 0 0 05-X.
Bergan, A., Dávila, C., Leone, F., Awerbuch, J., Tan, T.M., 2016. A mode I cohesive
law characterization procedure for through-the-thickness crack propagation in
composite laminates. Compos. Part B Eng. 94. doi:10.1016/j.compositesb.2016.
03.071.
Bouvet, C., Castanié, B., Bizeul, M., Barrau, J.J., 2009. Low velocity impact modelling
in laminate composite panels with discrete interface elements. Int. J. Solids
Struct. 46 (14–15), 2809–2821.
Bouvet, C., Rivallant, S., Barrau, J.J., 2012. Low velocity impact modeling in compos-
ite laminates capturing permanent indentation. Compos. Sci. Technol. 72 (16),
1977–1988. doi:10.1016/j.compscitech.2012.08.019.
Byers, B., 1980. Behavior of Damaged Graphite/Epoxy Laminates under Compression
loading, NASA/CR-159293.
Camanho, P., Bessa, M., Catalanotti, G., Vogler, M., Rolfes, R., 2013. Modeling the
inelastic deformation and fracture of polymer composites part II: smeared crack
model. Mech. Mater. 59, 36–49. doi:10.1016/j.mechmat.2012.12.001.
Camanho, P.P., Dávila, C.G., Pinho, S.T., Iannucci, L., Robinson, P., 2006. Prediction of
in situ strengths and matrix cracking in composites under transverse tension
and in-plane shear. Compos. Part A Appl. Sci. Manuf. 37 (2), 165–176. doi:10.
1016/j.compositesa.2005.04.023.
Cantwell, W., Morton, J., 1991. The impact resistance of composite materials – a
review. Composites 22 (5), 347–362. doi:10.1016/0010-4361(91)90549-V.
Caputo, F., De Luca, A., Lamanna, G., Borrelli, R., Mercurio, U., 2014. Numerical study
for the structural analysis of composite laminates subjected to low velocity im-
pact. Compos. Part B Eng. 67, 296–302. doi:10.1016/j.compositesb.2014.07.011.
Catalanotti, G., Camanho, P., Marques, A., 2013. Three-dimensional failure criteria for
fiber-reinforced laminates. Compos. Struct. 95, 63–79. doi:10.1016/j.compstruct.
2012.07.016.
Catalanotti, G., Camanho, P.P., Xavier, J., Dávila, C.G., Marques, A.T., 2010. Measure-
ment of resistance curves in the longitudinal failure of composites using digital
image correlation. Compos. Sci. Technol. 70 (13), 1986–1993.
Chai, H., Knauss, W.G., Babcock, C.D., 1983. Observation of damage growth in
compressively loaded laminates. Exp. Mech. 23 (3), 329–337. doi:10.1007/
BF02319260.
Craven, R., Iannucci, L., Olsson, R., 2010. Delamination buckling: a finite element
study with realistic delamination shapes, multiple delaminations and fibre frac-
ture cracks. Compos. Part A Appl. Sci. Manuf. 41 (5), 684–692. doi:10.1016/j.
compositesa.2010.01.019.
Dang, T.D., Hallett, S.R., 2013. A numerical study on impact and compression after
impact behaviour of variable angle tow laminates. Compos. Struct. 96, 194–206.
doi:10.1016/j.compstruct.2012.10.006.
Davies, G.A.O., Olsson, R., 2004. Impact on composite structures. Aeronaut. J. 108
(1089), 541–563.
Dávila, C.G., Camanho, P.P., Rose, C.A., 2005. Failure criteria for FRP laminates. J.
Compos. Mater. 39 (4), 323–345.
 E.V. González et al. / International Journal of Solids and Structures 144–145 (2018) 230–247
Dávila, C.G., Camanho, P.P., Turon, A., 2007. Cohesive Elements for Shells, NASA/TP-
2007-214869.
Dávila, C.G., Camanho, P.P., Turon, A., 2008. Effective simulation of delamination in
aeronautical structures using shells and cohesive elements. J. Aircr. 45 (2), 663–
672. doi:10.2514/1.32832.
Dávila, C.G., Rose, C.A., Camanho, P.P., 2009. A procedure for superposing linear co-
hesive laws to represent multiple damage mechanisms in the fracture of com-
posites. Int. J. Fract. 158 (2), 211–223.
EN 2597:1998, 1998. Carbon Fibre Reinforced Plastics. Unidirectional Laminates.
Tensile Test Perpendicular to the Fibre Direction. BSI.
Faggiani, A., Falzon, B.G., 2010. Predicting low-velocity impact damage on a stiffened
composite panel. Compos. Part A Appl. Sci. Manuf. 41 (6), 737–749.
Feng, D., Aymerich, F., 2014. Finite element modelling of damage induced by low-
velocity impact on composite laminates. Compos. Struct. 108 (1), 161–171.
doi:10.1016/j.compstruct.2013.09.004.
González, E.V., Maimí, P., Sainz de Aja, J.R., Cruz, P., Camanho, P.P., 2014. Effects
of interply hybridization on the damage resistance and tolerance of composite
laminates. Compos. Struct. 108 (1). doi:10.1016/j.compstruct.2013.09.037.
González, E.V., Maimí, P., Camanho, P.P., Turon, A., Mayugo, J.A., 2012. Simulation
of drop-weight impact and compression after impact tests on composite lami-
nates. Compos. Struct. 94 (11). doi:10.1016/j.compstruct.2012.05.015.
Hongkarnjanakul, N., Bouvet, C., Rivallant, S., 2013. Validation of low velocity impact
modelling on different stacking sequences of CFRP laminates and influence of fi-
bre failure. Compos. Struct. 106, 549–559. doi:10.1016/j.compstruct.2013.07.008.
ISO 15114:2014, 2014. Determination of the Mode II Fracture Resistance for Unidi-
rectionally Reinforced Materials using the Calibrated End-Loaded Split (C-ELS)
Test and an Effective Crack Length Approach, Geneva, Switzerland.
Johnson, A.F., Holzapfel, M., 2006. Influence of delamination on impact damage in
composite structures. Compos. Sci. Technol. 66 (6), 807–815.
Johnson, A.F., Pickett, A.K., Rozycki, P., 2001. Computational methods for predict-
ing impact damage in composite structures. Compos. Sci. Technol. 61 (15),
2183–2192.
MIL-HDBK-17-3F (Military Handbook), 2002. Composite Materials Handbook: Poly-
mer Matrix Composites Materials Usage, Design, and Analysis, 3. Department of
Defense of USA.
Laffan, M.J., Pinho, S.T., Robinson, P., McMillan, A.J., 2012. Translaminar fracture
toughness testing of composites: a review. Polym. Test 31 (3), 481–489.
Liu, P.F., Liao, B.B., Jia, L.Y., Peng, X.Q., 2016. Finite element analysis of dynamic pro-
gressive failure of carbon fiber composite laminates under low velocity impact.
Compos. Struct. 149, 408–422. doi:10.1016/j.compstruct.2016.04.012.
Liv, Y., 2017. A Contribution to the Understanding of Compression After Impact of
Composite Laminates. Universitat de Girona.
Lopes, C.S., Camanho, P.P., Gürdal, Z., Maimí, P., González, E.V., 2009. Low-velocity
impact damage on dispersed stacking sequence laminates. Part II: numerical
simulations. Compos. Sci. Technol. 69 (7–8). doi:10.1016/j.compscitech.2009.02.
015.
Lopes, C.S., Sádaba, S., González, C., Llorca, J., Camanho, P.P., 2016. Physically-sound
simulation of low-velocity impact on fiber reinforced laminates. Int. J. Impact
Eng. 92, 3–17. doi:10.1016/j.ijimpeng.2015.05.014.
Maimí, P., Camanho, P.P., Mayugo, J.A., Dávila, C.G., 2007a. A continuum damage
model for composite laminates: part I – constitutive model. Mech. Mater. 39
(10), 897–908. doi:10.1016/j.mechmat.20 07.03.0 05.
Maimí, P., Camanho, P.P., Mayugo, J.A., Dávila, C.G., 2007b. A continuum damage
model for composite laminates: part II – computational implementation and
validation. Mech. Mater. 39 (10), 909–919. doi:10.1016/j.mechmat.20 07.03.0 06.
Martín-Santos, E., Maimí, P., González, E.V., Cruz, P., 2014. A continuum constitutive
model for the simulation of fabric-reinforced composites. Compos. Struct. 111
(1). doi:10.1016/j.compstruct.2013.12.024.
Mendes, P.A.A.E., Donadon, M.V., 2014. Numerical prediction of compression after
impact behavior of woven composite laminates. Compos. Struct. 113 (1), 476–
491. doi:10.1016/j.compstruct.2014.03.051.
Obdržálek, V., Vrbka, J., 2011. On the applicability of simple shapes of delamina-
tions in buckling analyses. Compos. Part B Eng. 42 (3), 538–545. doi:10.1016/j.
compositesb.2010.11.006.
Olsson, R., 20 0 0. Mass criterion for wave controlled impact response of composite
plates. Compos. Part A Appl. Sci. Manuf. 31 (8), 879–887.
Ortega, A., Maimí, P., González, E.V., Sainz de Aja, J.R., de la Escalera, F.M., Cruz, P.,
2017a. Translaminar fracture toughness of interply hybrid laminates under ten-
sile and compressive loads. Compos. Sci. Technol. 143, 1–12. doi:10.1016/j.
compscitech.2017.02.029.
Ortega, A., Maimí, P., González, E.V., Ripoll, L., 2014. Compact tension specimen
for orthotropic materials. Compos. Part A Appl. Sci. Manuf. 63. doi:10.1016/j.
compositesa.2014.04.012.
Ortega, A., Maimí, P., González, E.V., Trias, D., 2016. Characterization of the translam-
inar fracture cohesive law. Compos. Part A Appl. Sci. Manuf. doi:10.1016/j.
compositesa.2016.01.019.
Ortega, A., Maimí, P., González, E.V., Trias, D., 2017b. Specimen geometry and
specimen size dependence of the R-curve and the size effect law from a
cohesive model point of view. Int. J. Fract. 205 (2), 239–254. doi:10.1007/
s10704- 017- 0195- 1.
247
Pavier, M.J., Clarke, M.P., 1996. Finite element prediction of the post-impact com-
pressive strength of fibre composites. Compos. Struct. 36 (1–2), 141–153. doi:10.
1016/S0263-8223(96)0 0 073-6.
Perillo, G., Jørgensen, J.K., Cristiano, R., Riccio, A., 2017. A numerical/experimental
study on the impact and CAI behaviour of glass reinforced composite plates.
Appl. Compos. Mater. 1–23. doi:10.1007/s10443- 017- 9628- 2.
Pinho, S. T., Dávila, C. G., Camanho, P. P., Iannucci, L., Robinson, P., 2005. Failure
Models and Criteria for FRP Under In-Plane or Three-Dimensional Stress States
Including Shear Non-Linearity. NASA/TM-2005-213530.
Pinho, S.T., Robinson, P., Iannucci, L., 2006. Fracture toughness of the tensile and
compressive fibre failure modes in laminated composites. Compos. Sci. Technol.
66 (13), 2069–2079.
prEN 2850, 1997. Aerospace series - Carbon fibre thermosetting resin - Unidirec-
tional laminates - Compression test parallel to fibre direction.
PAM-CRASH FE Code, 2004. F-94578 Rungis Cedex, France.
Riccio, A., Ricchiuto, R., Saputo, S., Raimondo, A., Caputo, F., Antonucci, V., et al.,
2016. Impact behaviour of omega stiffened composite panels. Prog. Aerosp. Sci.
81, 41–48. doi:10.1016/j.paerosci.2015.11.004.
Richardson, M.O.W., Wisheart, M.J., 1996. Review of low-velocity impact properties
of composite materials. Compos. Part A Appl. Sci. Manuf. 27 (12), 1123–1131.
Rivallant, S., Bouvet, C., Hongkarnjanakul, N., 2013. Failure analysis of CFRP lami-
nates subjected to compression after impact: FE simulation using discrete in-
terface elements. Compos. Part A Appl. Sci. Manuf. 55, 83–93. doi:10.1016/j.
compositesa.2013.08.003.
Schwab, M., Pettermann, H.E., 2016. Modelling and simulation of damage and failure
in large composite components subjected to impact loads. Compos. Struct. 158,
208–216. doi:10.1016/j.compstruct.2016.09.041.
Schwab, M., Todt, M., Wolfahrt, M., Pettermann, H.E., 2016. Failure mechanism based
modelling of impact on fabric reinforced composite laminates based on shell el-
ements. Compos. Sci. Technol. 128, 131–137. doi:10.1016/j.compscitech.2016.03.
025.
Shi, Y., Pinna, C., Soutis, C., 2014. Modelling impact damage in composite laminates:
a simulation of intra- and inter-laminar cracking. Compos. Struct. 114 (1), 10–19.
doi:10.1016/j.compstruct.2014.03.052.
Soto, A., González, E.V., Maimí, P., de la Escalera, F., de Aja, J.R., Alvarez, E., 2018. Low
velocity impact and compression after impact simulation of thin ply laminates.
Compos. Part A Appl. Sci. Manuf. 109, 413–427. doi:10.1016/j.compositesa.2018.
03.017.
Soto, A., González, E.V., Maimí, P., Mayugo, J.A., Pasquali, P.R., Camanho, P.P., 2017. A
Methodology to Simulate Impact and Compression after Impact in Large Com-
posite Stiffened Panels. Compos. Struct..
Soto, A., González, E.V., Maimí, P., Turon, A., Sainz de Aja, J.R., de la Escalera, F.M.,
2016. Cohesive zone length of orthotropic materials undergoing delamination.
Eng. Fract. Mech. 159, 174–188. doi:10.1016/j.engfracmech.2016.03.033.
Suemasu, H., Irie, T., Ishikawa, T., 2009. Buckling and post-buckling behavior if com-
posite plates containing multiple delaminations. J. Compos. Mater. 43 (2), 191–
202. doi:10.1177/0021998308099217.
Suemasu, H., Sasaki, W., Ishikawa, T., Aoki, Y., 2008. A numerical study on com-
pressive behavior of composite plates with multiple circular delaminations con-
sidering delamination propagation. Compos. Sci. Technol. 68 (12), 2562–2567.
doi:10.1016/j.compscitech.2008.05.014.
Sun, X.C., Hallett, S.R., 2017. Barely visible impact damage in scaled composite lami-
nates: experiments and numerical simulations. Int. J. Impact Eng. 109, 178–195.
doi:10.1016/j.ijimpeng.2017.06.008.
Tan, W., Falzon, B.G., Chiu, L.N., Price, M., 2015. Predicting low velocity impact
damage and compression-after-impact (CAI) behaviour of composite laminates.
Compos. Part A Appl. Sci. Manuf. 71, 212–226. doi:10.1016/j.compositesa.2015.
01.025.
Turon, A., Camanho, P.P., Costa, J., Renart, J., 2010. Accurate simulation of delamina-
tion growth under mixed-mode loading using cohesive elements: definition of
interlaminar strengths and elastic stiffness. Compos. Struct. 92 (8), 1857–1864.
Turon, A., Dávila, C.G., Camanho, P.P., Costa, J., 2007. An engineering solution for
mesh size effects in the simulation of delamination using cohesive zone models.
Eng. Fract. Mech. 74 (10), 1665–1682. doi:10.1016/j.engfracmech.2006.08.025.
Vogler, M., Rolfes, R., Camanho, P., 2013. Modeling the inelastic deformation and
fracture of polymer composites part I: plasticity model. Mech. Mater. 59, 50–
64. doi:10.1016/j.mechmat.2012.12.002.
Wagih, A., Maimí, P., Blanco, N., Costa, J., 2016a. A quasi-static indentation test
to elucidate the sequence of damage events in low velocity impacts on com-
posite laminates. Compos. Part A Appl. Sci. Manuf. 82, 180–189. doi:10.1016/j.
compositesa.2015.11.041.
Wagih, A., Maimí, P., González, E., Blanco, N., De Aja, J., De La Escalera, F., et al.,
2016b. Damage sequence in thin-ply composite laminates under out-of-plane
loading. Compos. Part A Appl. Sci. Manuf. 87. doi:10.1016/j.compositesa.2016.04.
010.
Yan, H., Oskay, C., Krishnan, A., Xu, L.R., 2010. Compression-after-impact response of
woven fiber-reinforced composites. Compos. Sci. Technol. 70 (14), 2128–2136.
Ye, L., 1988. Role of matrix resin in delamination onset and growth in composite
laminates. Compos. Sci. Technol. 33 (4), 257–277.