Composites Science and Technology 72 (2012) 1781–1790

Contents lists available at SciVerse ScienceDirect

Composites Science and Technology

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

The low velocity impact response of foam-based sandwich panels

Jin Zhou, Mohamad Zaki Hassan, Zhongwei Guan ⇑, Wesley J. Cantwell
School of Engineering, University of Liverpool, Liverpool L69 3GH, UK

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

Article history: This paper describes the results of a combined experimental/numerical study to investigate the perfora-
Received 21 March 2012 tion resistance of sandwich structures. The impact response of plain foam samples and their associated
Received in revised form 29 June 2012 sandwich panels was characterised by determining the energy required to perforate the panels. The
Accepted 8 July 2012
dynamic response of the panels was predicted using the finite element analysis package ABAQUS/Explicit.
Available online 16 July 2012
The experimental arrangement, as well as the FE model were also used to investigate, for the first time,
the effect of oblique loading on sandwich structures and also to study the impact response of sandwich
Keywords:
panels on an aqueous support.
A. Glass fibres
PVC foam sandwich
Testing has shown that the perforation resistance of the plain foams and their sandwich panels is
B. Impact behaviour strongly dependent on the properties of the foam core. For example, increasing the density of the cross-
C. Finite element analysis (FEA) linked PVC foams from 60 to 200 kg/m3 yielded an eight fold increase in the perforation resistance of the
sandwich panels. At intermediate and higher densities, the crosslinked PVC foams and their associated
sandwich structures offered a superior perforation resistance to their linear PVC counterparts. The FE
analysis reasonably predicted the impact load–displacement responses and the perforation energies of
both the plain foams and the sandwich panels. Finally, it has been shown that sandwich panels impacted
in an aqueous environment offer a lower perforation resistance than those tested in air.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction the specific perforation energy of the sandwich panels by over

twenty percent for a comparable areal density. Dear and Brown
As a result of their favourable specific properties, particularly [5,6] undertook research on the impact toughness of reinforced
when subjected to flexural loading, composite sandwich structures polymeric materials and various lightweight sandwich panels.
are finding increasing use in a wide range of lightweight load-bear- The purpose of the study was to satisfy the energy absorption
ing engineering structures. Typically, sandwich structures, such as and structural integrity requirements of the next generation of ad-
those used in high-performance aerospace components, are based vanced aircraft, marine-craft, road and rail vehicles. Zhou and Hill
on thin composite skins bonded to a low density foam or honey- [7] investigated the parameters affecting damage development and
comb core. The fact that the skins are usually thin, often renders energy absorption in sandwich panels subjected to localised im-
these lightweight panels highly susceptible to damage by a hard pact loading. They argued that changing the skin thickness not only
projectile, such as that associated with runway debris or hail. A changed the flexural rigidity of the panels, but also changed the
number of researchers have investigated the response of sandwich mechanisms of load transfer between the upper and lower skins.
structures to localised impact loading and much of this work has This in turn influenced the development of damage within the
been summarised in detailed reviews on the topic [1,2]. Mines structure.
et al. [3] investigated the low velocity impact response of glass A number of workers have attempted to model the impact re-
fibre/vinyl ester sandwich panels based on both aluminium honey- sponse of sandwich structures [8–13]. Lee et al. [8] used a refined
comb and Coremat cores. Their findings showed that the energy- formulation of a composite sandwich panel to study the impact re-
absorbing capacity of the sandwich panels increased with loading sponse of a sandwich plate. In their study, the face skins were mod-
rate, this being due to strain-rate effects in the core crush stress elled as two separate Mindlin plates, and the core was considered to
and the skin failure stress. Reyes and Cantwell [4] conducted high offer both normal and shear stiffnesses. Good agreement was ob-
velocity impact tests on sandwich panels based on fibre–metal tained between the predicted and measured impact responses.
laminate skins and an aluminium foam core and noted that the Yang and Qiao [9] used a higher-order impact model to simulate
introduction of metal layers into the composite skins increased the dynamic response of sandwich beams based on soft cores. They
investigated the effects of impactor mass, initial velocity, core
⇑ Corresponding author. depth and core stiffness on the impact response of the beams and
E-mail address: [email protected] (Z. Guan).
showed that varying the core depth has a significant influence on

0266-3538/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compscitech.2012.07.006
1782 J. Zhou et al. / Composites Science and Technology 72 (2012) 1781–1790

the impact response of the structure. Palazotto and Gummadi [10] plies of woven E glass fibres in a thermosetting triazine resin (sup-
used a displacement-based, plate bending finite element analysis to plied by Gurit Ltd.). The composite skins were laid up on the foam
model impact damage in sandwich plates based on a Nomex hone- core and placed in a picture-frame mould. It should be noted that
comb core. A comparison with low velocity impact and quasi-static the bonding cross the skin–core interface was assured by the resin
indentation data highlighted the model’s ability to predict impor- from the composite prepreg, i.e. no additional adhesive was used.
tant features of the behaviour of the sandwich panels. Aktay et al. The mould was then placed in Meyer hot press and cured for
[11] used the finite element package PAM-CRASH to predict impact 1 hr at a temperature of 125 °C. During the curing, a pressure of
damage in sandwich panels based on both PEI foam and Nomex 70 kPa was applied. The panels were not post-cured following
honeycomb cores and showed that the model was capable of repro- manufacturing.
ducing the physical behaviour of the two types of structure. In addi- Low velocity impact tests were conducted on both the plain
tion, Arora et al. [14,15] undertook experimental and numerical foams as well as the sandwich structures. Here, the 150 mm square
work on glass-fibre reinforced polymer (GFRP) sandwich panels panels (containing 10 bolt holes along their edges) were placed on
and laminated tubes subjected to blast in air and underwater envi- a 100 mm long steel cylinder fixed to a square steel base and
ronments. The authors highlighted the importance of support and clamped using a steel ring and 10 bolts. Both the cylindrical sup-
boundary conditions with regards to blast-resistant design. port and the clamping ring had an internal diameter of 100 mm.
Few attempts have been made to investigate the perforation The panels were impacted at their centres by a carriage with a
behaviour of sandwich structures subjected to impact loading. 10 mm diameter hemispherical head. The mass of the impactor
Lin and Hoo Fatt [12] developed an analytical model to predict was 5.56 kg and the release height of the impact carriage was var-
the low velocity impact response of glass/epoxy-aluminium hon- ied between 0.2 and 1.4 m. The impact force and displacement of
eycomb panels. A three-stage perforation model was used that the impactor were measured using a piezoelectric load cell and a
considered failure of the top composite skin, the honeycomb core high-speed video camera respectively. Previous unpublished work
and the lowermost composite skin. The perforation model pre- by the authors has shown that the total impulse matches the
dicted the perforation velocity of the panel under low velocity im- change in impact momentum as witnessed by the high speed video
pact conditions. Buitrago et al. [13] used finite element techniques camera. Further details of impact rig, including a photograph, are
to model the process of impact perforation in sandwich panels given in Ref. [16]. Tests were also conducted on sandwich panels
based on carbon fibre/epoxy skins and an aluminium honeycomb placed on a water support. These tests were undertaken to simu-
core. They showed that most of the incident energy of the projec- late the effect on an internal impact on an immersed sandwich
tile was absorbed by the 2 mm thick composite skins, with the core structure, such as a boat hull. Here, the circular ring support was
absorbing between 10% and 20% of the impact energy. filled with water and the panel was clamped as discussed
The aim of this paper is to model the low velocity impact perfo- previously.
ration response of lightweight sandwich structures based on a range After testing, the panels were sectioned through the impacted
of polymeric foams. Here, the finite element analysis package region, ground, polished and photographed in order to elucidate
ABAQUS/Explicit is used to predict the response of the foam core the failure processes associated with the perforation process.
panels and sandwich panels during impact by a steel projectile.
Particular attention is given to establishing the effect of varying the 3. Finite element modelling
properties of the foam core on the perforation resistance of the sand-
wich panels. Once validated, the model is used to predict the effect of Numerical models were developed to simulate the dynamic re-
oblique impact on the perforation resistance of the sandwich panels, sponse of the foam panels and the foam core sandwich panels sub-
a loading condition that is difficult to conduct experimentally. Finally, jected to projectile impact. The following sections outline the
the model is also used to investigate the perforation response of sand- techniques used to model the various constituents of the sandwich
wich panels in an aqueous environment. panels.

2. Experimental study 3.1. Modelling of the PVC foam

Nine different types of polymer foam were investigated in this Each core was modelled as a crushable foam using hardening
study. Five of the foams were based on a crosslinked PVC, two on curves obtained following compression tests on square blocks.
a linear PVC and two on a PET foam. Details of the various foams Deshpande and Fleck [17] proposed a phenomenological yield sur-
are given in Table 1. All of the foam panels had a thickness of face for a closed-cell foam, given by:
20 mm. Sandwich structures based on these nine foams were pre- 1  
uh i q2 þ a2 r2m  r2y 6 0 ð1Þ
pared by bonding thin composite skins using a two part epoxy
1 þ ða Þ2
3
adhesive. The 0.25 mm thick composite skins were based on two

Table 1
Mechanical properties of the foams investigated in this study.

C60 C80 C100 C130 C200 PET105 PET135 L90 L140

Density (kg/m3) 60 80 100 130 200 105 135 90 140
Poisson’s ratio 0.32 0.32 0.32 0.32 0.32 0.32 0.32 0.32 0.32
Compressive modulus (MPa) 69 97 125 160 280 90 140 56 110
Compressive strength (MPa) 0.9 1.3 1.9 2.6 4.8 1.4 2.4 0.9 1.6
Compressive fracture strain 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7
Tensile modulus (MPa) 45 66 84 110 175 110 175 50 90
Tensile strength (MPa) 1.3 2 2.7 3.8 6 2.3 3.3 1.4 2.4
Shear modulus (MPa) 22 30 38 47 75 21 30 21 37
Shear strength (MPa) 0.8 1.2 1.6 2.3 3.5 0.9 1.3 1 1.85
Shear fracture strain 0.16 0.23 0.27 0.3 0.3 0.15 0.12 0.75 0.8
 J. Zhou et al. / Composites Science and Technology 72 (2012) 1781–1790 1783

where ry is the uniaxial tensile or compressive yield strength of the transverse moduli of elasticity were assumed to be equal since
foam, q is the Von Mises stress, rm is the mean stress. The parame- the GFRP skins were based on a plain weave. The material proper-
ter a defines the shape of the yield surface, which is related to the ties are E1 = E2 = 23 GPa, E3 = 5 GPa, G12 = G13 = G23 = 5 GPa, t12 ¼
ratios of the initial uniaxial yield stress roc and the hydrostatic ten- t13 ¼ t23 ¼ 0:15, as determined both experimentally and analyti-
sile yield stress pt to the hydrostatic compressive yield stress poc , cally in Ref. [19].Damage initiation was modelled using Hashin’s
respectively. The yield stress in hydrostatic compression pc provides failure criteria [20] which assume four damage initiation mecha-
the evolution of the size of the yield surface and can be expressed nisms, namely fibre tension, fibre compression, matrix tension
as: and matrix compression. Using the longitudinal, transverse and
 h i shear effective stress tensor components within the plane of the
  rc evplol rc ðeplv ol Þða12 þ 19Þ þ p3t GFRP, the damage initiation criteria can be determined [21].
pc ev ol
pl ¼   ð2Þ The following parameters are required in Hashin’s failure criteria,
rc evplol
pt þ the tensile and compressive strengths (XT = 480 MPa and
3
XC = 432 MPa) in the longitudinal direction, the tensile and
v ol is the plastic volumetric strain for the volumetric harden-
where epl compressive strengths (YT = 480 MPa and YC = 432 MPa) in the
ing model, which is equal to eaxial
pl the uniaxial compressive plastic transverse direction, the longitudinal and transverse shear
strain. Therefore, pc should be determined from a uniaxial compres- strengths (SL = 480 MPa and ST = 480 MPa).
sion test on the foam. The damage elastic matrix, which controls degradation of the
For a rate-dependent material, the response follows the uniaxial material stiffness, can be expressed as:
flow rate definition as: 2 3
ð1  df Þ=E1 ð1  df Þð1  dm Þv 21 =E1 0
e_ pl ¼ hðq; epl ; hÞ ð3Þ 16 ð1  df Þð1  dm Þv 12 =E2 ð1  dm Þ=E2 0 7
CD ¼ 6 7 ð6Þ
D4 0 0 ð1  ds ÞGD
5
where h is a strain hardening function, epl is the equivalent plastic
strain, and h is the temperature. The rate-dependent hardening
where G is the shear modulus and D is an overall damage variable,
curves can be expressed as:
which can be expressed as:
 
e epl ; e_ pl ¼ ry ðepl ÞRðe_ pl Þ ð4Þ D ¼ 1  ð1  df Þð1  dm Þv 12 v 21 ð7Þ
e_ pl and R are the equivalent plastic strain-rate and stress ratio Here, df, dm, and ds reflect the current state of fibre, matrix and shear
(= r  =ry ) respectively. damage, respectively.Damage evolution is modelled by the negative
Damage initiation in the PVC foam was modelled by applying a slope of the equivalent stress–displacement relation after damage
ductile damage criterion in conjunction with shear damage crite- initiation is achieved. The fracture energies for fibre and matrix ten-
rion [18]. The fracture strains corresponding to the initiation of sion (GFft ¼ GFmt ¼ 110 kJ=m2 ) as well as fibre and matrix compression
ductile damage and shear damage and the related strain-rate need (GFfc ¼ GFmc ¼ 120 kJ=m2 ) failure modes are specified to determine
to be specified. Damage development, associated with ductile and the energy dissipated during damage development.
shear failure is controlled by the fracture energy in terms of the en-
ergy required for failure development. A linear softening law was 3.3. Modelling of the resin layer and other contact conditions
used, specifying a linear relationship between the softening stress
and the displacement after the onset of the damage for elasto-plas- The resin layer between the GFRP skin and the foam core was
tic materials. modelled using cohesive elements (0.25 mm thick) available in
A tensile failure criterion was specified in the lower half of the ABAQUS [22]. The mechanisms of initiation and progression of
central region of the panel to model failure of the foam. The crite- damage were explicitly incorporated in the formulation of the ele-
rion assumes that failure occurs when the pressure stress becomes ment, and were based on the maximum nominal stress and the
more tensile than the specified hydrostatic cut-off stress. In this effective displacement following the linear softening law.
model, the hydrostatic pressure stress was used as a measure of Damage initiation in the cohesive elements was evaluated using
failure to model either dynamic spall or a pressure cut-off. the ratios of the predicted nominal stress components based on the
During impact, it was noted that the PVC foam panels exhibited elastic traction–separation behaviour to the corresponding critical
large recoverable deformations. In order to simulate such deforma- values in tension (debonding) or shear (sliding) at the interface.
tion prior to failure, a hyperelastic model was applied to the exter- A general contact interaction was defined between the GFRP
nal region of the panel, i.e. that which is located outside the central skin and the foam to include the case where the cohesive layer is
40 mm region. In the hyperelastic model, it is necessary to define damaged by projectile. In addition a surface-to-surface contact
the strain energy potential, which describes the strain energy interaction allows for sliding between the projectile surface and
stored in the material per unit of reference volume (i.e. the volume the individual nodes located in the central region of the target.
in the initial configuration) as a function of the strain at a given
point. The Mooney-Rivlin model was selected as a strain energy 3.4. Geometrical model and mesh generation
potential [18], which is:
1 el Here, fully-clamped foam panels and foam core sandwich pan-
U ¼ C 10 ðI1  3Þ þ C 01 ðI2  3Þ þ ðJ  1Þ2 ð5Þ els (100 mm in diameter with skin and core thickness as 0.25 and
D1
20 mm respectively) subjected to low velocity impact loading were
where C10, C01 and D1 are temperature-dependant material param- simulated using ABAQUS/Explicit. Given that the panels were cir-
eters, I1 and I2 are the first and second deviatoric strain invariants cular, only one quarter of the panel was modelled. Fig. 1 shows
and Jel is the elastic volume ratio. the geometric, boundary and loading conditions for the sandwich
panels. Meshing is also shown in the same figure. Modelling of
3.2. Modelling of the GFRP skin the plain foam panel was carried out by removing the skins and
the cohesive layers in the sandwich panel. The cylindrical projec-
Prior to damage initiation, the GFRP skin was modelled as an tile had a mass of 5.56 kg, which is assumed to be rigid relative
orthotropic elastic material. The in-plane longitudinal and to the sandwich panel.
1784 J. Zhou et al. / Composites Science and Technology 72 (2012) 1781–1790

r=5 mm

(v)
h=19 mm

Fixed
Fixed

H=20 mm 0.25 mm GFRP

0.25 mm resin

20 mm core

R=50 mm

Fig. 1. Geometrical, mesh, boundary and loading conditions of the sandwich model.

Mesh sensitivity effects were studied by varying the mesh den- to over-estimate the initial stiffness of the higher density foams.
sity within the plane and through the thickness of the panel. The In spite of these discrepancies, agreement between the numerical
foam was meshed using eight-node reduced integration elements and experimental data is good. Fig. 2b compares both the experi-
(C3D8R) and the skin using continuum shell elements (SC8R). mental and predicted perforation zones for these three foam pan-
els. The lowest density C80 foam (as well as the C60 and C100
foams) exhibited a cylindrically-shaped shear zone, similar in size
4. Results and discussion to the diameter of the projectile. Perforation in the intermediate
(130 kg/m3) and high (200 kg/m3) density foams resulted in a
4.1. Perforation resistance of the plain foams mixed mode of failure, with a cylindrical shear region in the upper
half of the sample and a frustrum-shaped zone in the lower portion
Fig. 2a shows typical load–displacement traces following im- of the test panel. It has been shown that locally-high tensile stres-
pact tests on three of the crosslinked PVC foams. Included in the ses under the indentor give rise to this low energy mode of fracture
figure are the predictions offered by the finite element analyses. [23]. The cross-sections suggest, therefore, that the 60, 80 and
All three experimental traces display similar trends. The initial por- 100 kg/m3 foams fail in the higher energy shearing mode, whereas
tions of the load–displacement trace exhibit some oscillatory the 120 and 200 kg/m3 fail as a result of a lower energy, mixed ten-
behaviour, due to dynamic effects in the plate and ringing in the sile/shear mechanism. This evidence suggests the perforation resis-
load cell, with its slope reflecting the elastic modulus of the foam. tances of these foams are likely to be determined by the tensile and
With increasing load, a knee is reached, beyond which the slope of shear fracture properties of the foams. The fracture properties of
the traces decreases as the projectile starts to crush the foam and the present range of foams, subjected to these two modes of load-
penetrate the top surface of the panel. Finally, the load drops shar- ing, have been investigated in a previous study [24]. In this earlier
ply as the steel projectile perforates the target. investigation, the work of fracture (Wf) values of these foams in
An examination of Fig. 2a indicates that the FE model captures tension and shear were measured using a recently-developed shear
the fundamental features apparent in the experimental load– rig and the single edge notch bend (SENB) specimen geometry
displacement traces. The models exhibit more pronounced respectively. The resulting values are listed in Table 1 [24]. An
oscillatory effects in the initial response of the plates and tends examination of the data indicates that the shear values of Wf are

3000
C80 Test

2500
C200 C80 FE
2000
Load (N)

1500 C130 Test

C130
1000 C130 FE
C80
500
C200 Test
0
0 5 10 15 20 25 30
C200 FE
Displacement (mm)
(a) load-displacement traces (b) cross-sections
Fig. 2. Comparison of load–displacement traces and cross-sections of crossed-linked PVC foams. The solid lines correspond to the experimental data and the dashed lines to
the predictions.
 J. Zhou et al. / Composites Science and Technology 72 (2012) 1781–1790 1785

800
PET105 Test
700

600
PET 135
500 PET105 FE
Load (N)
400

300

200 PET 105 PET135 Test

100

0
0 5 10 15 20 25
PET135 FE
Displacement (mm)

(a) load-displacement traces (b) cross-sections

Fig. 3. Comparison of load–displacement traces and cross-sections of T92 PET PVC foams. The solid lines correspond to the experimental data and the dashed lines to the
predictions.

1400
L90 Test
1200

1000 L140
L90 FE
Load (N)

800

600 L90

400 L140 Test

200

0
0 10 20 30 40 L140 FE
Displacement (mm)

(a) load-displacement traces (b) cross-sections

Fig. 4. Comparison of load–displacement traces and cross-sections of R63 linear PVC foam. The solid lines correspond to the experimental data and the dashed lines to the
predictions.

significantly higher than their corresponding tensile values. This PVC foams. Once again, increasing the density of the foam serves to
evidence suggests that those samples exhibiting a cone cracking increase the initial stiffness of the plate, the maximum load and the
mode of failure are not achieving their full potential in terms of maximum measured displacement.
their energy-absorbing capability. Fig. 5 presents a comparison of the predicted (solid lines) and
Fig. 3a shows the load–displacement traces for the two plain measured perforation energies of the plain core materials. The
PET foam materials. In both cases, agreement between the pre- predictions for both the linear PVC and PET foams are extended
dicted and measured response is excellent, with the former to include notional materials with densities of 60 and 200
accurately predicting the initial slope, the maximum load and the kg/m3. The properties of these foams were obtained by extrapo-
maximum displacement. Increasing the density from 100 to lating the data from those foams within the same family (Table 1
130 kg/m3 has a significant influence on the general shape of the indicates that the majority of mechanical properties vary in a
load–displacement trace. Fig. 3b shows the resulting cross-sections linear manner). An examination of the figure shows that the
for this foam, where it is clear that the lowest density system again model predicts the experimental perforation energies with some
exhibits a mixed type of failure with the upper part of the plate success. It is evident that the linear PVC foams offer a superior
failing in the higher energy shear mode and the lower region of perforation resistance to their crosslinked counterparts. Interest-
the target failing in a tensile mode. Failure in the higher density ingly, an examination of the Mode II work of fracture data in
system is less distinct, although there is some evidence of a transi- Table 2 suggests that the higher density crosslinked foams would
tion region at the mid-plane of the sample. Again, agreement be- have out-performed their linear counterparts if failure had
tween the predicted and observed failure modes is good, occurred in a pure shear mode rather than a mixed shear/tensile
indicating that the FE model accurately predicts the perforation mode. Finally, it is evident that the PET foams offer the lowest
process in these panels. Finally, Fig. 4 shows the load–displace- perforation energies, with values approximately one half those
ment traces and cross-sections following impact on the two linear offered by the crosslinked foams.
1786 J. Zhou et al. / Composites Science and Technology 72 (2012) 1781–1790

50 an initial peak resulting from fracture of the top skin followed by

45 a steadily increasing force up to final perforation. Here, a small fi-
40
nal peak in the load is apparent as the projectile fractures the lower
surface skin. Finally, the highest density foam system offers a stea-
Predicted energy (J)

L
35
dily increasing load–displacement curve, in which failure of the
30 composite skins is difficult to discern. A comparison of Figs. 2a
25 and 6a suggests that the core plays an increasingly dominant role
C
20 as the density of the foam is increased. Indeed, at higher densities,
15 the load–displacement traces closely resemble those of the plan
PET core panels. Interestingly, the failure processes were similar in all
10
of the sandwich structures, with the projectile shearing a relatively
5
clean hole through the target, Fig. 5b. Given that there was little
0 evidence of the distinct frustrum-shaped fracture zone observed
0 30 60 90 120 150 180 210
in the higher density plain crosslinked cores, it is evident that
Density (kg/m3)
the core in these sandwich structures is absorbing more energy
Fig. 5. Perforation energy versus density for the linear PVC (circles), crosslinked than their plain foam counterparts.
PVC (triangles) and PET PVC (squares) foam panels. Fig. 7a shows the load–displacement traces for the sandwich
structures based on the 105 kg/m3 PET and the 140 kg/m3 linear
PVC foams. The former exhibits two well-defined peaks, resulting
4.2. Perforation resistance of the sandwich panels from fracture of the two skins as well as a region of constant force
in which the projectile passes though the foam core. In contrast,
Fig. 6 presents typical load–displacement traces following low the response of the higher density linear PVC foam is once again
velocity impact on sandwich structures based on the three crosss- largely dominated by the fracture behaviour of the core material.
linked foams presented in Fig. 2. Once again, agreement between The cross-section of the experimental PET-based system displays
the FE predictions and the experimental data is good, with the a frustrum-shaped fracture zone similar to that observed in the
model capturing most of the features in the experimental data. plain foam, Fig. 7b. The model predicted a smaller conical zone
The lowest density foam exhibits a classic load–displacement trace to that observed experimentally, indicating that it has not fully
[25], with pronounced peaks associated with fracture of the upper predicted the failure process.
and lower skins and a relatively smooth plateau resulting from the The energies required to perforate the sandwich structures are
projectile passing through the foam core. The FE model fails to summarised in Fig. 8. Once again additional structures based on
identify the first peak, although the remainder of the trace is linear PVC and PET foams were modelled by extrapolating the
accurately predicted. The trace for the 130 kg/m3 foam exhibits mechanical properties given in Table 1. Agreement between the

Table 2
Perforation energies of the foams and sandwich structures.

Foam Work of fracture in tension (kJ/ Work of fracture in shear (kJ/ Eperf core Eperf sandwich
m2) m2) experimental experimental
C60 0.26 6.48 3.93 10.69
C80 0.44 12.6 10.16 15.46
C100 0.62 18.4 8.76 18.34
C130 0.76 27.6 17.7 27.06
C200 1.33 44.2 32.92 55.58
PET105 2.3 7.38 5.33 9.12
PET135 2.5 18.2 8.32 19.03
L90 6.06 21.2 9.57 20.07
L140 12.1 27.3 26.1 25.56

4000
C80 Test
3500
C200
3000 C80 FE

2500
Load (N)

C130 Test
2000
C130
1500 C130 FE
1000
C80
500 C200 Test

0
0 5 10 15 20 25 30
C200 FE
Displacement (mm)
(a) load-displacement traces (b) cross-sections
Fig. 6. Comparison of load–displacement traces and cross-sections of sandwiches made with C70 foam cores. The solid lines correspond to the experimental data and the
dashed lines to the predictions.
 J. Zhou et al. / Composites Science and Technology 72 (2012) 1781–1790 1787

1800
PET105Test
1600
1400
L140
1200 PET105FE

Load (N)
1000
800
600
L140Test
PET105
400
200
0 L140FE
0 5 10 15 20 25 30
Displacement (mm)

(a) load-displacement traces (b) cross-sections

Fig. 7. Load–displacement traces and cross-section of sandwich panels made with T92105 and R63140 cores. The solid lines correspond to the experimental data and the
dashed lines to the predictions.

70 4.3. The effect of angle of obliquity on the perforation resistance

60 Normal impacts, such as those reported above, rarely occur in

C real engineering situations. Instead, components are more fre-
Predicted energy (J)

50
quently loaded at some oblique angle. Such tests are difficult to
undertake experimentally, given the need to guide the projectile,
40
L coupled with the presence of a horizontal force component that
30 applies a load to the required guide rails.
In this investigation, the process of oblique impact on three
20 types of sandwich structure was modelled using the procedures
PET
outlined above. Fig. 9 shows the variation of perforation with angle
10 of obliquity for sandwich panels based on two crosslinked PVC
cores and a PET foam. Here, the impact angle refers to the angle be-
0 tween the axis of the projectile and the normal to the panel. From
0 30 60 90 120 150 180 210
Density (kg/m3) the figure, it is evident that the perforation energy increases with
impact angle, for example, passing from approximately 29 J for a
Fig. 8. Perforation energy versus density for the linear PVC (circles), crosslinked normal impact, to over 35 J for 30° loading. All three traces exhibit
PVC (triangles) and PET-based PVC (squares) sandwich structures. The solid or similar trends with the rate of increase in perforation energy
broken lines represent the predictions of the finite element model.
increasing with impact angle. Fig. 10a shows the predicted load–
displacement traces for the C130 and PET105 sandwich panels fol-
lowing oblique impact. With increasing the impact angle, the final
peak force shifts to the right, associated with the increase in perfo-
model (solid lines) and the experimental data is generally good. It
ration energy. Fig. 10b presents the cross-sections resulting from
is interesting to note that in certain cases, the addition of the com-
the FE predictions. The crosslinked foam again exhibits a clear
posite skins has an important effect on the energy-absorbing
capacity of the core. This is most pronounced in the highest density
crosslinked PVC foam which exhibited a significant enhancement
in impact resistance following the addition of skins. Here, a change 40
in failure mode was observed, passing from a mixed tensile–shear
in the plain foam to a pure shear mode in the sandwich. As shown 35 C130
in Table 2, the fracture energy associated with tensile failure is rel-
30
atively low, and one would therefore expect a significant improve-
Perforation energy (J)

ment in perforation resistance as the failure mode shifts to pure 25

shear. From the figure it is also evident that the perforation resis-
tance of the lowest density linear PVC foam is superior to that of 20 C80
its crosslinked counterpart, whereas the converse is true for the
higher density systems. Interestingly, the shear work of fracture 15
values exhibits similar trends as illustrated in Table 2. Here, the
10 PET105
evidence suggests that the shear fracture properties of the linear
and crosslinked foams cross at approximately 125 kg/m3, a value 5
similar to that observed in the perforation data in Fig. 8. This evi-
dence again supports the conclusion that the fracture properties 0
0 5 10 15 20 25 30 35
of the foam strongly influence the perforation resistance of these
Incident angle (degree)
thin-skinned sandwich structures. Finally, it is clear that the PET-
based sandwich structures clearly offer the lowest perforation Fig. 9. The variation of perforation energy with impact angle for three sandwich
resistance reflecting the lower fracture properties of this foam. structures.
1788 J. Zhou et al. / Composites Science and Technology 72 (2012) 1781–1790

2500
0°
0° C130 T105
2000 10°
20°
C130
30°

Load (N)
1500
10°

1000

500 T105 20°

0
0 5 10 15 20 25 30 35 40
30°
Displacement (mm)

(a) load-displacement traces (b) cross-sections

Fig. 10. Predicted load–displacement traces and cross sections of sandwich panels made with C70.130 cross-linked PVC and PET 92.105 PVC core subjected to oblique impact
at incident angles of 10°, 20° and 30°.

cylindrical-shaped perforation zone, similar to that observed under 2000

normal impact. The PET foam also exhibits a well-defined cylindri- Test L140 air
cal zone, with there being no evidence of the small conical region 1600
Test L140 water
observed following normal impact, Fig. 7b. FE L140 water Air
A simple geometric analysis shows that the ratio of the surface Load (N)
area of an elliptic oblique cylinder to that of a right cylinder (the 1200
projectile creates an elliptical entrance hole on the top surface of
the target) is given by: 800
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 2
ðr þ R2 Þ=r cosðIÞ 400
2
where r is the radius of the right cylinder, R is the long radius of the
0
elliptical oblique cylinder and I is the angle between the right cylin- 0 5 10 15 20 25 30
der and the elliptical oblique cylinder. Similarly, the ratio of the vol- Displacement (mm)
ume an elliptic oblique cylinder to that of a right cylinder is given
by: Fig. 11. Comparison of load–displacement traces of sandwiches made with L140
foam cores between sandwiches supported on water and without support.
1= cosðIÞ:

Given that energy is dissipated in shearing both the composite and based on the L140, the PET135 and the C130 foams. Fig. 11 shows
the foam around the perimeter of the projectile, as well as crushing load–displacement traces for the L140 sandwich structure. In-
the foam ahead of the impactor, the change in perforation energy cluded in each figure are the associated FE predictions and the cor-
associated with increasing impact angle will reflect a combination responding experimental trace for the previous impact tests (i.e. in
of both equations. Applying these expressions to a 30° impact sug- the absence of water). An examination of the traces for the linear
gests that the surface area increases by 24.8% and the volume by PVC sandwich structures (L140) in Fig. 10 indicates that the panels
15.5%. Interestingly, the relative increase in the perforation energy supported on the water foundation exhibit reasonably similar
in Fig. 9 increases with foam toughness, with the PET105 system traces up to the point at which the projectile approaches the rear
increasing by approximately 18% and the C130 foam by approxi- surface. Here, the force rises rapidly in the ‘wet’ system before
mately 23%, in passing from normal impact to a 30° impact. This dropping rapidly as the rear surface fractures. It is clear that the
suggests that the C130 system may be absorbing more energy in rear surface peak is much higher in the fully-supported wet panel
(the area-based) shear mechanism, whereas the PET system absorbs than in its dry counterpart. In addition, the wet panel exhibits vir-
a substantial amount of energy in volumetric crushing under the tually no out of plane deflection, in contrast to the relatively flex-
impactor. ible dry panel. The FE model predicts the experimental data with
some success, although the final drop in force is not as abrupt.
4.4. Impact tests on sandwich panels supported on water Here, the water was meshed as a three-dimensional body in the
FE model, with typical properties (shear viscosity, density and
As a result of their superior specific properties and non-mag- wave speed) assigned.
netic characteristics, sandwich structures are finding increasing The predicted and measured perforation energies for the three
use in the design of boats hulls and other primary marine struc- wet sandwich structures are compared with their corresponding
tures. In order to investigate the influence of an aqueous environ- dry samples in Fig. 12. From the figure, it is clear that the model
ment on the dynamic response of a sandwich structure, a series of accurately predicts the perforation thresholds of all of the sand-
low velocity impact tests were undertaken on panels supported on wich panels. In five out of the six cases, the model slightly over-
a combined circular ring/water base. Here, the ring support used estimates the experimental value, although the differences are
for the earlier perforation tests was filled with water and panels not significant. It is interesting to note that the ‘wet’ panels offer
were placed on the water base and clamped using a steel ring, as a lower resistance to perforation than their dry counterparts. This
before. Three types of sandwich structure were tested, these being reduction is largely associated with the inability of the former to
 J. Zhou et al. / Composites Science and Technology 72 (2012) 1781–1790 1789

35 the second peak loads are reduced with increasing water depth. For
Experiment example, the first peak load decreases from 650 to 570 N in passing
30 from air to a water depth of 18 m. Similarly, the energy required to
FE model perforate the sandwich structures decreases with increasing depth
25 of immersion, with the perforation energy of the 18 m panel being
approximately 10% below that measured in air.
Energy (J)

20

15 5. Conclusions

10 The low velocity perforation resistance of a range of plain foams

and their sandwich structures has been investigated both experi-
5 mentally and numerically. Agreement between the predictions of
the finite element models and the experimental data was very
0 good across the range of foams investigated here. The linear cross-
linked PVC foams offered a higher perforation resistance than the
linear PVC foams at low densities, whereas the converse was true
at higher densities. The perforation resistances of the sandwich
Fig. 12. Comparison of energy absorption of sandwiches made with C130, PET135 structures follow the trends in the shear fracture properties of
and L140 foam cores between sandwiches supported on water and without support. the foams, suggesting that this mode of failure is important in
(A = air, W = water).
determining the perforation resistance of thin-skinned sandwich
structures.
deflect out-of-plane. This is demonstrated clearly in Fig. 11, where Finite element analyses simulating oblique impact on a number
the projectile displacement at the instant that the lower skin fails of the foams have shown that the energy required to perforate the
is equal to the initial thickness of the panel (i.e. the flexural re- targets increased with the angle of obliquity, reflecting the increase
sponse of the target has been largely suppressed). in surface area of material failing in shears as well as the larger vol-
In the final part of this investigation, the FE model was used to ume of foam that is crushed during failure. It has also been shown
predict the response of a sandwich structure in which the aqueous that sandwich structures impacted in a marine environment offer a
environment is located on the impact surface, with the distal sur- lower perforation resistance than those tested in air.
face being in contact with air. This can be considered to reflect the
case of a boat hull that is in collision with an immersed object. The Acknowledgement
impact velocity was fixed at 5.6 m/s. Under such a low velocity im-
pact conditions, cavitation is highly unlikely to be an issue. How- The authors are grateful to Airex AG for supplying the polymer
ever, the model allows for interaction between the projectile and foams.
the water. Clearly, the object striking the hull will be in a pressur-
ised aqueous environment, with the distal surface forming the in- References
ner wall of the boat (atmospheric conditions). Such tests would be
[1] Abrate S. Localized impact on sandwich structures with laminated facings.
difficult to replicate in practise and FE modelling therefore repre- Appl Mech Rev 1997;50:69–82.
sents an ideal technique for investigating this problem. Fig. 13 pre- [2] Chai GB, Zhu S. A review of low-velocity impact on sandwich structures. Proc
sents the load–displacement traces of sandwiches based on Inst Mech Eng, Part L: J Mater Des Appl 2011;225:207–30.
[3] Mines RAW, Worrall CM, Gibson AG. Low velocity perforation behaviour of
PET105 foam cores placed under water to depths of 6 and 18 m polymer composite sandwich panels. Int J Impact Eng 1998;21:855–79.
(the trace for a panel impacted in air is also included and marked [4] Reyes VG, Cantwell WJ. The high velocity impact response of composite and
as 0 m). It can be clearly seen that there is an initial displacement FML-reinforced sandwich structures. Compos Sci Technol 2004;64:35–54.
[5] Dear JP, Brown SA. Impact damage processes in reinforced polymeric
associated with the panels immersed under water, due to the pre-
materials. Composites Part A 2003;34(5):411–20.
vailing hydrostatic pressure loading. This initial loading also shifts [6] Dear JP, Lee H, Brown SA. Impact damage processes in composite sheet and
the first peak load to the right. In addition, both the first peak and sandwich honeycomb materials. Int J Impact Eng 2005;32(1–4):130–54.
[7] Zhou G, Hill M. Investigation of parameters governing the damage and energy
absorption characteristics of honeycomb sandwich panels. J Sandwich Struct
Mater 2007;9:309–42.
[8] Lee LJ, Huang KY, Fann YJ. Dynamic responses of composite sandwich plate
1000
impacted by a rigid ball. J Compos Mater 1993;27:1238.
[9] Yang M, Qiao P. Higher-order impact modeling of sandwich structures with
PET105_0m flexible core. Int J Solids Struct 2005;42:5460–90.
800 [10] Palazotto EJ, Gummadi LNB. Finite element analysis of low-velocity impact on
PET105_6m
composite sandwich plates. Compos Struct 2000;49:209–27.
PET105_18m [11] Aktay L, Johnson AF, Kroplin B-H. Numerical modelling of honeycomb crush
600 behaviour. Eng Fract Mech 2008;75:2616–30.
Load (N)

[12] Lin C, Hoo Fatt MS. Perforation of sandwich panels with honeycomb cores by
hemispherical nose projectiles. J Sandwich Struct Mater 2005;7(2):133–72.
[13] Buitrago BL, Santiuste C, Sánchez-Sáez S, Barbero E, Navarro C. Modelling of
400 composite sandwich structures with honeycomb core subjected to high-
velocity impact. Compos Struct 2010;92:2090–6.
[14] Arora H, Hooper PA, Dear JP. Dynamic response of full-scale sandwich
200 composite Structures subject to air-blast loading. Compos A 2011;42:1651–62.
[15] Arora H, Hooper PA, Dear JP. The effects of air and underwater blast on
composite sandwich panels and tubular laminate structures. Exp Mech
2012;52:59–81.
0 [16] McKown S, Cantwell WJ, Jones N. Investigation of scaling effects in fibre–metal
0 5 10 15 20 25 30 35
laminates. J Compos Mater 2008;42(9):865–88.
Displacement (mm) [17] Deshpande VS, Fleck NA. Multi-axial yield behavior of polymer foams. Acta
Mater 2001;49:1859–66.
Fig. 13. Comparison of load–displacement traces of sandwiches made with PET105 [18] ABAQUS. Theory manual. Version 6.7. Pawtucket: Hibbitt, Karlsson & Sorensen
foam cores between sandwiches under water depth of 0, 6 and 18 m. Inc.; 2008.
1790 J. Zhou et al. / Composites Science and Technology 72 (2012) 1781–1790

[19] Fan J. Investigation of the behaviour of fibre metal laminates subjected to low [23] Ball A. On the bifurcation of cone cracks in glass plates. Phil Mag
velocity impact. PhD thesis. University of Liverpool; 2010. 1996;73:1093.
[20] Hashin Z, Rotem A. A fatigue failure criterion for fiber reinforced materials. J [24] Zaki M, Cantwell WJ. The influence of loading rate on the mechanical
Compos Mater 1973;7:448. properties of polymeric foams. Int J Polym Technol, submitted for publication.
[21] Fan J, Guan ZW, Cantwel WJ. Numerical modelling of perforation failure in [25] Wen HM, Reddy TY, Reid SR, Soden PD. Indentation, penetration and
fibre metal laminates subjected to low velocity impact loading. Compos Struct perforation of composite laminates and sandwich panels under quasi-static
2011;93:2430–6. and projectile loading. Key Eng Mater 1998;501:141–3.
[22] ABAQUS/Explicit. User’s manual. Version 6.7. Pawtucket: Hibbitt, Karlsson &
Sorensen Inc.; 2008.