International Journal of Impact Engineering 85 (2015) 67e82

Contents lists available at ScienceDirect

International Journal of Impact Engineering

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

An analytical model of the dynamic response of circular composite

plates to high-velocity impact
A. Schiffer a, b, *, W.J. Cantwell a, V.L. Tagarielli b
a
Khalifa University of Science, Technology and Research (KUSTAR), Abu Dhabi, United Arab Emirates
b
Department of Aeronautics, Imperial College London, SW7 2AZ, UK

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

Article history: Analytical models are developed to predict the transient elastic response of fully clamped circular
Received 12 January 2015 composite plates subject to high-velocity impact by a rigid spherical projectile. The models are based on
Received in revised form first-order shear deformation plate theory and account for the effects of large deformations as well as
10 May 2015
propagation and reflection of flexural waves. Analytical predictions of plate deflection history and peak
Accepted 14 June 2015
Available online 22 June 2015
strain in the plates are found in good agreement with those obtained from detailed explicit FE simula-
tions. The dynamic response is found to be governed by four non-dimensional parameters and two
characteristic regimes of behaviour are identified. The models are used to construct maps to design
Keywords:
Ballistic limit
impact-resistant composite plates.
Finite element © 2015 Elsevier Ltd. All rights reserved.
Wave propagation
Plate theory
Impact damage

1. Introduction velocities below the ballistic limit, matrix cracking and delamina-
tion have been recognised to be the main energy dissipation
Fibre-reinforced composite materials are progressively mechanisms. Takeda et al. [2] conducted impact tests on glass-
employed as structural materials in light-weight ships, road vehi- fibre/epoxy laminates and used high-speed photography to
cles, aircraft components and armour systems, due to their low observe the growth of delamination cracks propagating in the
weight, high stiffness and excellent corrosion resistance. The samples, concluding that delamination growth was associated with
resistance of composite plates to high-velocity impact is a concern flexural wave propagation. Post-impact matrix cracks and de-
in many industrial applications. In the last decades, significant laminations were also observed by Heimbs et al. [3], who con-
effort has been devoted to foster understanding of the dynamic ducted an experimental and numerical study of the impact
response of composite laminates consequent to localised impact behaviour of CFRP composites subject to compressive and tensile
loading. In composite materials, energy absorption due to plastic preloads, concluding that tensile preloading leads to a reduction in
deformation is very limited and their response to localised trans- delamination while compressive preloading facilitates delamina-
verse impact leads to deformation modes dictated by propagation tion. At impact velocities near the ballistic limit, they also observed
of longitudinal, shear and flexural waves travelling in the material fibre failure in addition to delaminations and matrix cracks. Other
at different velocities [1]. authors [4,5] employed theoretical modelling approaches to study
The damage and failure modes of composites upon impact delamination of laminates subject to transverse impact. Some au-
depend on the plate geometry, impact velocity as well as on the thors have investigated the mechanism of plate spalling induced by
shape and mass of the projectile. The impact resistance of a reflection of through-thickness stress waves, see e.g. Ref. [6].
structure is often quantified by the limit velocity (or ballistic limit), Studies investigating the deformation and failure mechanisms
defined as the velocity required for a projectile to penetrate a given of laminates impacted above the ballistic limit are extensively
material at least 50% of the time. When laminates are impacted at described in the literature and a comprehensive review of existing
work on this subject can be found in Abrate [7]. For example,
Cantwell and Morton [8] observed the mechanisms of perforation
* Corresponding author. Khalifa University of Science, Technology and Research of thin CFRP beams and noted that plate failure involved a shear-off
(KUSTAR), Abu Dhabi, United Arab Emirates. Tel.: þ971 (0)2 4018204. penetration in the upper half of the plate (impact side) and tensile
E-mail address: [email protected] (A. Schiffer).

http://dx.doi.org/10.1016/j.ijimpeng.2015.06.010
0734-743X/© 2015 Elsevier Ltd. All rights reserved.
68 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82

breakage of plies in the bottom half. The effect of projectile ge- transversely by a cylindrical projectile, predicting that the struc-
ometry on the perforation resistance of fibre-reinforced composites tural response comprised propagation of tensile waves and ‘cone
was investigated by Wen [9], who derived a simple empirical waves’ emanating from the impact point, with the cone wave
relationship for the ballistic limit by assuming that the resistance travelling at lower speed. The theory was used to predict the bal-
provided by the laminate is composed of a static and a dynamic listic resistance of composite systems and predictions were found
term, with the latter depending on the nose shape of the projectile. in agreement with Cunniff's scaling theory [12].
Mines et al. [10] conducted ballistic tests on woven, z-stitched and In this study we derive an analytical model for the dynamic
through-thickness reinforced glass/polyester laminates, varying response of a fully-clamped, circular composite plate subject to high
laminate thickness as well as mass and geometry of the projectile. velocity impact by a rigid projectile. Effects of transverse shear de-
Their results showed only small differences in the impact behaviour formations, large deflections and flexural wave propagation will be
of the different composite systems investigated. taken into account. In addition, the effect of higher order vibrational
While carbon-fibre (CFRP) and glass-fibre reinforced laminates modes, activated upon reflection of flexural waves at the bound-
(GFRP) are the most widely used material systems in engineering aries, will also be modelled. The model is based on a linear elastic
applications, the recent development of new fibres with extremely material response but accounts for the geometric non-linearities in
high stiffness to weight ratios has greatly improved the ballistic the problem and, to some extent, for material anisotropy.
performance of fibre-composites. They include Nylon, aramids (e.g. It is clear that the prediction of the ballistic limit of arbitrary
Kevlar®), ultra-high molecular weight polyethylene (e.g. Spectra®, composite plates is beyond the scope of the present study, which
Dyneema®) and PBO (e.g. Zylon®). Zhu et al. [11] performed dy- does not attempt modelling the complex damage mechanisms
namic perforation tests on Kevlar/polyester laminates and found activated in composite laminates upon impact. On the other hand
that they outperform aluminium plates of equal weight in terms of the model presented here provides, for a certain class of composite
impact resistance. They also tested laminates with deliberately plates and for an arbitrary projectile, the critical impact velocity at
introduced delaminations and the results showed that the ballistic the onset of tensile ply failure; this information is readily used in
limit was not greatly affected by such defects. the design of components exposed to a substantial threat of impact
In an attempt to relate the ballistic performance of a given lami- loading (e.g., impact of runway debris or similar on aircraft struc-
nate to the velocity and geometry of the projectile, Cunniff [12] tures). The model allows identifying the four main governing non-
proposed a set of non-dimensional parameters and argued that the dimensional groups of the impact problem and predicts two
ballistic limit of fibre composites scales with a characteristic velocity possible, distinctive regimes of behaviour.
determined by the material properties of the fibres. However, for The outline of this paper is as follows: in Sections 2 and 3 we
some types of laminate, the characteristic velocity introduced by derive the analytical models and describe the FE scheme employed;
Cunniff does not accurately capture the experimental data. For in Section 4 we validate the analytical models by comparing
example, Karthikeyan et al. [13] recently measured the ballistic per- analytical and FE predictions; in Section 5 the validated analytical
formance of Dyneema® plates (ultra-high molecular weight poly- model is used to compare the damage resistance of glass-fibre and
ethylene fibre composite) and found that the characteristic velocity carbon-fibre reinforced composite plates, and non-dimensional
required to normalise the perforation data cannot be deduced from design maps are constructed for both types of laminate.
the fibre properties. Their observations showed that the propagation
of flexural wave fronts followed an almost square-like pattern, due to 2. Analytical modelling
the extremely low shear strength of this type of laminate, whereas
those observed on CFRP plates were almost circular. The elastic response of composite plates to high-velocity impact
A considerable body of literature exists on numerical and by rigid projectiles is dictated by propagation of flexural waves,
theoretical predictions of the elastic response of composite plates shear waves and extensional waves travelling in the material at
subject to various dynamic loading conditions. A possible analytical different velocities. In fibre-reinforced composites, wave speeds are
treatment of impact on elastic plates follows that given in Zener different when measured along different axes or directions due to
[14] who expressed the transient response of thin simply- the anisotropic behaviour of the material, see e.g. Sierakowski and
supported isotropic plates in terms of mode shapes and natural Chaturvedi [22] for a comprehensive account of the dynamic
frequencies. A similar approach was used by Olsson [15] who behaviour of fibre-reinforced composites. Due to the complexity of
extended the theory of Zener [14] to the case of orthotropic plates. the problem, exact solutions are restricted to the use of numerical
Sun and Chattopadyay [16] employed a similar technique to methods which require high computational effort, especially when
investigate the central impact of a mass on a simply-supported parametric studies are being conducted.
laminated composite plate under initial stress by employing a The objective of this study is to develop an approximate
plate theory that accounts for transverse shear deformations [17]. analytical model able to provide, in a computationally efficient way,
They also noted that rotary inertia has only a minor effect on the reliable predictions of plate deformation associated with the dy-
dynamic response. Dobyns [18] also used plate theory [17] to namic elastic response a circular composite laminate subject to
analyse the dynamic response of composite plates subject to high-velocity impact. In this section, we employ an approach
loading by pressure pulses of various shapes, in order to mimic similar to that of Schiffer and Tagarielli [23] to derive the equations
different types of blast loading. Finite strain solutions for the impact of motion in form of non-dimensional ODEs, and to identify the
behaviour of elastic plates with fully-clamped boundaries are ob- governing non-dimensional parameters; various assumptions
tained in the published literature via approximate techniques, since concerning plate deformation and material behaviour will be
closed-form solutions are not available in this case. For example, explained and discussed in detail. Finally, we define two charac-
the Rayleigh-Ritz method was employed by Qian & Swanson [19] teristic deformation regimes and construct a regime map.
for the case of impacted rectangular carbon/epoxy plates. A
reduced model for predicting the dynamic deformation modes is 2.1. Governing equations
presented in Hoo Fatt and Palla [20] for the case of composite
sandwich plates subject to loading by a prescribed pressure history. 2.1.1. Material modelling
Phoenix and Porwal [21] derived a theoretical model for the 2D In this study we consider symmetric composite laminates
response of an initially unloaded elastic membrane impacted comprised of a stack of transversely isotropic plies with equally
 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82 69

spaced fibres (quasi-isotropic layups such as [0/±60 ]S, [0/45 /90 /

45 ]S, etc.). In a first approximation, we neglect the directionality
of material stiffness in the circumferential direction of the plate (in
case of a cross-ply laminate for example) and adopt the concept of
effective (or average) laminate stiffness. In doing so, we define four
effective elastic constants

1 A0 12 1
Er ¼ nr4 ¼  12 Efr ¼ Gr ¼ (1)
hA011 A011 h3 D011 hS011

where Er, nr4, Efr and Gr are the effective in-plane modulus, Poisson's
ratio, flexural modulus and transverse shear modulus, respectively,
and A011 is the first element of the laminate's compliance matrix,
A0 ¼ inv(A). It is important to mention that the stiffnesses A011 , A011 ,
D011 and S011 in eq. (1) are, in general, dependent on the choice of
reference system and vary along the circumferential direction 4. In
order to obtain effective averages, the properties were evaluated n
times (typically n ¼ 8) in different reference systems obtained by
rotating an arbitrary cylindrical system about the z-axis by in- Fig. 1. Sequence of deformation profiles associated with the response of a clamped
crements 4j ¼ 2pj/n (j ¼ 0,1,…,n) and the corresponding properties elastic plate subject to impact of a rigid projectile: (a) projectile impinges on the target,
Erj , Gjr and njr4 were averaged to obtain the effective laminate (b) propagation of a flexural wave (Phase 1 response) and (c) excitation of higher order
mode shapes due to boundary effects (Phase 2 response).
properties as

X
n
1Xn X
n
~r ¼ 1
E Erj ~nr4 ¼ njr4 ~r ¼ 1
G Gjr : (2) front reaches the boundary of the plate, i.e. z ¼ R (see Fig. 1c), the
n n n plate deflection profile is affected by the boundary conditions. We
j¼0 j¼0 j¼0
denote as ‘Phase 1’ the response ranging from t ¼ 0 to the instant
The response of the composite is isotropic in the plane of the when the flexural wave reaches the plate's centre point, t1, i.e.
plate but has bending and shear moduli independent of the in- z(t1) ¼ R, while ‘Phase 2’ represents the response at subsequent
plane properties. For the composites modelled in this work, the times.
very small differences between Efr and Er allowed to assume E ~ zE ~r .
fr
It should be mentioned here that the assumption of axisym-
metric plate deformation is not ideal for all types of composites; for 2.1.3. Phase 1 response 0  t  t1
example experiments [13] have shown that impact on composite We proceed to derive the governing equations for the plate's
systems with very low shear strength (e.g. Dyneema®) results in response during Phase 1. In plate theory, it is convenient to intro-
flexural waves propagating outwards with square wave-fronts. The duce stress resultants in terms of the forces and moments applied
models developed in this paper are not adequate for this type of to the plate's middle surface (per unit length of laminate side)
composites. which are defined as
0 1 0 1 0 1
Nrr Zh=2 srr Mrr
2.1.2. Problem geometry and modelling approach @ N44 A ¼ @ s44 Adz @ M44 A
Consider a fully clamped circular plate of thickness h and Nr4 sr4 Mr4
h=2
radius R made from a composite laminate (see Section 2.1.1) of
0 1
areal mass m ¼ rh, as sketched in Fig. 1 (r denotes the average Zh=2 srr   Zh=2  
@ s44 Azdz Qr trz
density of the laminate). The plate is subject to dynamic trans- ¼ ¼ dz (3)
Q4 t4z
verse loading by impact of a rigid spherical projectile of mass M h=2
sr4 h=2
and radius Rs, travelling at a velocity v0 perpendicular to the plate
surface (Fig. 1a). Here, attention is restricted to impact of rela- where Ni, Mi and Qi are the in-plane forces, bending/twisting mo-
tively large projectiles, i.e. 2Rs > h, on plates with small to ments and transverse shear forces, respectively.
moderate aspect ratios, 0.02 < h/R < 0.15. For these ranges it is Here, we employ the first-order shear deformation theory of
reasonable to use the thin plate assumption which neglects local plates (i.e. Mindlin plate theory) based on the von Ka rman strain
plate indentations. relations, which account for non-linear terms in the in-plane strain
As sketched in Fig. 1b, we assume that the initial phase of response due to stretching of the plate's mid-surface. Note that for
response is dictated by propagation of a flexural wave, emanating transverse loading cases, it is widely accepted that the radial and
from the impact point and propagating towards the boundary. tangential mid-plane displacements are vanishingly small
Despite the dispersive behaviour of flexural waves [1], impact compared to the transverse deflections; then, the kinematic re-
experiments on composite plates [13,24] have shown that the lations are
shape of the dynamic disturbance does not appreciably change
during this phase and that the flexural wave front can be idealised uðr; z; tÞ ¼ zqr ðr; tÞ wðr; z; tÞ ¼ wðr; tÞ (4)
by an elastic wave front propagating at a velocity z_ in the positive
r-direction, see Fig. 1b. In the light of these observations, we as- where u and w are the displacements in the r and z directions,
sume a simple axisymmetric polynomial displacement field to respectively, and qr denotes the rotation of the cross-section in the
describe the initial deformation response of the composite plate, rz-plane, with reference to the coordinate system shown in Fig. 1.
w(r,t), in terms of two degrees of freedom: the centre deflection Employing eq. (4) and imposing axisymmetric deformations, the
w0(t) and the flexural wave position z(t). When the flexural wave von Karman strain-displacement relations can be written as
70 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82

  ~ h3
1 vwðr; tÞ 2 vqr ðr; tÞ E
εrr ¼ þz ε44 ¼ εr4 ¼ gr4 ¼ 0 grz D¼  r : (14)
2 vr vr 12 1  ~nr4 2
vwðr; tÞ
¼ þ qr ðr; tÞ (5)
vr Now write the total elastic energy of the plate as the sum of the
strain energies associated to bending, membrane and transverse
and the corresponding curvatures of the mid-surface with respect shear
to the angle of rotation, qr(r,t), are
U ¼ Ub þ Um þ Us
vqr 1
krr ¼ k44 ¼ qr kr4 ¼ 0: (6) ZzðtÞ ZR
vr r 1   1  
¼ 2p Mrr krr þ M44 k44 rdr þ 2p Nrr εrr þ N44 ε44 rdr
Let us now assume that plate deformation within the portion 2 2
0 0
0  r  z(t) can be approximated by an axisymmetric polynomial
shape function that satisfies the boundary conditions of the prob- ZzðtÞ
1
lem. Such a function may be written as þ 2p Qrz grz rdr:
2
" # 0
3r 2 2r 3 (15)
wðr; tÞ ¼ w0 ðtÞ 1  2
þ (7)
zðtÞ zðtÞ3
Substituting eqs. (5)e(8) and (12) in eq. (15) and evaluating the
integral terms gives
where w0(t) denotes the plate's centre deflection and z(t) is the
flexural wave position (Fig. 1b); for r > z(t), the plate is assumed to
pD h i ~r h
pE w4
remain straight during Phase 1, hence w(r,t) ¼ 0. A shear defor- U¼ 2
26:38g2rz0 z2 þ 72w0 ðgrz0 z þ w0 Þ þ   0
2
mation profile that is compatible with the boundary conditions and 8z 8 1  ~n2r4 z
symmetry requirements is given by ~rh
5pG
 2 þ g2rz0 z2 :
rp 32
grz ðr; tÞ ¼ grz0 ðtÞsin (8) (16)
zðtÞ
The derivation of the membrane energy, represented by the
with grz0(t) the shear deformation amplitude. second term of eq. (16), merits some further comment. The
The constitutive description for the composite laminate is nonlinear term in εrr, as defined in eq. (5), provides the relationship
treated as follows. For a symmetric laminate, the relationship be- between the transverse displacement w0 and the membrane strain
tween the in-plane forces Ni and the corresponding strains εi is in the radial direction, and predicts zero strain within the straight
 T portion of the plate, r > z(t). However, this contradicts experimental
ð Nrr N44 Nr4 ÞT ¼ A$ εrr ε44 gr4 (9) observations [25] and theoretical models [21] which suggest that a
tensile precursor wave emanates from the impact point and
where A denotes the in-plane stiffness matrix of the laminate. propagates radially towards the plate boundary, at a velocity higher
Similarly, the bending and twisting moments Mi can be related to than that of the flexural wave speed, thus inducing tensile radial
the corresponding curvatures ki as stresses in the portion r > z(t). The presence of such precursor
waves have also been detected in our FE calculations, as detailed in
ð Mrr M44 Mr4 ÞT ¼ D$ð krr k44 kr4 ÞT (10) Section 3. For the range of geometries considered here (see Section
2.1.2), preliminary FE simulations have shown that the propagation
with D the bending stiffness matrix of the laminate, and for velocity of the tensile precursor wave is much higher than that of
transverse shear deformations the stressestrain relationship is the flexural wave and it is therefore reasonable to neglect the
given by propagation of the precursor wave and to assume that the mem-
 T  T brane strain, εmrr , is uniform in the radial direction and approxi-
Qrz Q4z ¼ k S$ grz g4z (11) mately equal to the average strain induced in an ideal membrane
with a vanishing curvature, krr ¼ 0. Hence, we write
where S is the shear stiffness matrix of the laminate and the con-
stant k denotes the shear correction factor; note that in Mindlin's 1 w20
plate theory, k ¼ 5/6 for rectangular cross-sections. εm
rr ¼ : (17)
2 Rz
Employing first-order shear deformation theory of plates, the
constitutive equations are Previous studies [16] have shown that the effect of rotary inertia
plays only a minor role in the impact response of composite plates
 
Nrr ¼ C εrr þ ~nr4 ε44 N44 ¼ C~nr4 εrr Nr4 ¼ 0 ;  and is therefore neglected in our analysis. Then the total kinetic
Mrr ¼ D krr þ ~nr4 k44 M44 ¼ D k44 þ ~nr4 krr Mr4 ¼ 0 ; energy of the system is given by
~ r hg
Qrz ¼ kG Q 4z ¼ 0:
rz
ZzðtÞ 2  
(12) 1 vw 1 vw0 2
T¼ 2pm rdr þ M
2 vt 2 vt
after defining the axial rigidity C as 0
 2
 1
~r h
E ¼ pm 0:214w20 z_ þ 0:171w0 w_ 0 zz_ þ 0:0857w_ 20 z2 þ M w_ 20 :
C¼ (13) 2
1  ~nr4 2 (18)

and the bending rigidity D as Here, the over-dots denote derivatives with respect to time.
 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82 71

The EulereLagrange equations are now employed to solve for where v0 is the non-dimensional impact velocity. The initial value
the time history of the degrees of freedom w0(t), z(t) and grz0(t). The problem can be numerically integrated to obtain the histories
Lagrangian function w0 ðtÞ, zðtÞ and grz0 ðtÞ.
It can be seen from eqs. (16) and (18) that the Lagrangian
L¼T U (19) function (19) is independent of the plate radius R and therefore the
Phase 1 solutions are unaffected by the boundary conditions, as for
is obtained by combining eqs. (16) and (18), and is used to derive the case of a plate with infinite radius. The Phase 1 solutions cease
the equations of motion of the system via the EulereLagrange being valid when the flexural wave reaches the plate boundary,
equations zðt 1 Þ ¼ 1, as different deformation modes are induced by the
      interaction with the plate's supports. Correspondingly we modify
d vL vL d vL vL d vL vL
 ¼0  ¼0  ¼ 0: the shape functions for Phase 2 response, as detailed in the
dt vw_ 0 vw0 dt vz_ vz dt vg_ rz0 vgrz0 following section.
(20)
Evaluating the derivatives in eq. (20) and writing the equations 2.1.4. Phase 2 response t > t1
of motion in non-dimensional terms, we find that the non- We proceed to derive the governing equations for the ensuing
dimensional generalised coordinates Phase 2 response. The reflection of flexural waves at the clamped
plate boundary, commencing at t ¼ t1, gives rise to transverse os-
w0 z cillations at higher frequencies which affect the response of the
w0 ¼ z¼ grz0 (21)
R R plate. Such phenomena are not addressed in the current literature
pffiffiffiffiffiffiffiffi [14,15,21]; here we construct an approximate model able to predict
are functions of the non-dimensional time t ¼ t E=r=R and of the
the effects of the supports upon the dynamic response of the
following set of non-dimensional parameters
laminates. In order to obtain a tractable system of governing
rffiffiffiffiffi ~r equations for the Phase 2 response, only a limited number of mode
h M r G
h¼ M¼ v0 ¼ v0 G¼ : (22) shapes can be considered. In a first approximation, we add two
R rhR2 p ~
Er ~r
E sinusoidal shape functions to the polynomial shape function
considered for Phase 1, as
Here, w0 and z represent the normalised centre deflection and
flexural wave position, respectively, while the parameters rp2
3r 2 2r 3
h; M and v0 denote aspect ratio, mass ratio and non-dimensional wðr; tÞ ¼ w0 ðtÞ 1  þ þ w1 ðtÞsin
R2 R3 R
impact velocity, respectively.  2
After some algebraic manipulation the non-dimensional gov- 3rp
þ w2 ðtÞcos (27)
erning equations for the Phase 1 response of the plate are obtained 2R
as
where w1(t) and w2(t) are the vertical displacement amplitudes
 2
0:318 0:171w0 z_ þ 0:171w0 z€ _ zz_ þ 0:171w
z þ 0:514w0
€ z2
0
corresponding to the first and second sinusoidal terms, respec-
tively. It should be mentioned that only the first and last terms in
 0:159 2 
€  0:857w0 z z_ 1  ~nr4 2
2 eq. (27) contribute to the centre deflection, w(r ¼ 0,t), therefore
þ Mw 0 ¼ 2
z 1  ~nr4 2
wtot ðtÞ ¼ w0 ðtÞ þ w2 ðtÞ (28)
 
2 grz0 z
þ 3h þ w0 þ w30 is the total centre deflection in Phase 2. Accordingly, the shear
2
deflections in Phase 2 are modified to
(23)

"
h  i 1   5G  
4 0:1071w20 z€ _ zz_ þ 0:0428z2 w w _2 ¼ 2 4
z þ 0:257w0 w €  _ z
 0:343w 1  ~nr4 2 
4
g2 z 1  ~nr4 2
0 0 0 þ w0 0
2z
2
1  ~nr4 2 8 rz0
  #
2 grz0 w0 z w2
þ 3h þ w20 þ 0 (24)
2 2

rp2 g ðtÞ  
2rp
grz ðr; tÞ ¼ grz0 ðtÞsin þ rz1 sin
  R 2 R
2  3  
0 ¼ h 0:549grz0 z þ 0:75w0 þ 0:3125Gg2rz0 z 1  ~n2r4 : (25) g ðtÞ 3rp
þ rz2 sin (29)
2 R
The impact event can be mathematically described by the
following initial conditions where grz1(t) and grz2(t) are the shear deformation amplitudes
associated to the first and second sinusoidal terms, respectively.
        Fig. 2 shows the shape functions (27) and (29) at arbitrary ampli-
w0 t ¼ 0 ¼ 0 _ t¼0 ¼v
w0 0 z t¼0 ¼0 z_ t ¼ 0 ¼ 0 tudes. These satisfy the boundary conditions and symmetry re-
(26) quirements of the problem
72 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82

where TP1
t1 represents the kinetic energy of the plate at the end of
wðr ¼ 0Þ ¼ wtot wðr ¼ 1Þ ¼ 0 qr ðr ¼ 0Þ ¼ 0 q r ðr ¼ 1Þ ¼ 0 : Phase 1, t ¼ t1 , and is given by
(30)  2
P1
Tt1 ¼ pm 0:214w20;t1 z_ t1 þ 0:171w0;t1 w_ 0;t1 zt1 z_ t1
where w ¼ wtot =R and qr ¼ grz  vw=vr (see eq. (3)). The four  1
additional DOFs introduced in Phase 2 are the shear deformation þ 0:0857w_ 20;t1 z2t1 þ M w_ 20;t1 (34)
amplitudes associated with the first and second higher-order mode 2
shapes, grz1 and grz2, respectively, as well as the corresponding non-
while, at the beginning of the ensuing Phase 2 response, the kinetic
dimensional deflection amplitudes defined as deflections
energy reads
w1 w2
w1 ¼ w2 ¼ : (31) R2 m h  
R R P2
Tt1 ¼ 3
6480p4 w_ 20;t1 þ 113400p4 w_ 0;t1 v1;0 þ v2;0
75600p
We now employ the EulereLagrange equations (eq. (20)) and re-
   
þ 14175p4 v21;0 þ v22;0 þ 18900v1;0 p4 v2;0 þ p2 w_ 0;t1
write the governing eqs. (23)e(25) in terms of eqs. (27) and (29),
 
resulting in six non-dimensional equations of motion (given in
pffiffiffiffiffiffiffiffi  8400p2 v2;0 v1;0 þ v2;0 þ 30912p2 v1;0 v2;0
Appendix A) with respect to t ¼ t E=r =R which can be solved i
þ 226800w_ 0;t1 v1;0 þ 11200w_ 0;t1 v2;0
numerically for the histories w0 ðtÞ, w1 ðtÞ, w2 ðtÞ, grz0 ðtÞ, grz1 ðtÞ and
grz2 ðtÞ after imposing the following set of initial conditions 1  
þ M w_ 0;t1 þ v2;0 :
2
     
w0 t 1 ¼ w0;t1 w1 t 1 ¼ 0 w2 t 1 ¼ 0 (35)
_ t  ¼ w
w _ _ t ¼v
w
 _ t ¼v
w

0 1 0;t1 1 1  1;0 2 1  2;0 Now assume identical initial velocities for both higher-order
(32)
grz0 t 1 ¼ grz0;t1 grz1 t 1 ¼ 0 grz2 t 1 ¼ 0
      modes
g_ rz1 t 1 ¼ 0 g_ rz2 t 1 ¼ 0 g_ rz0 t 1 ¼ g_ rz0;t1
v1;0 ¼ v2;0 ¼ v12;0 : (36)
Here, the indices including ‘t1’ denote variables evaluated from the
Phase 1 solutions at time t ¼ t 1 . Combining eqs. (33)e(36) allows one to solve for the common
The initial (non-dimensional) velocities v1;0 and v2;0 (eq. (32)), initial velocity v12,0.
corresponding to the initial amplitudes of the higher-order Experiments and simulations show that, upon reaching the
deflection modes, were determined such to satisfy conservation plate boundary, the flexural wave is reflected towards the plate
of kinetic energy between the two phases of response centre; we do not model explicitly such reflection but we introduce
additional degrees of freedom to represent the excitation of high-
P1 P2
order vibration modes, enforcing conservation of kinetic energy
Tt1 ¼ Tt1 (33) across the two phases of response.

2.2. Deformation modes

In this section we proceed to examine the deformation behav-

iour during both phases of the response (see Section 2.1). With the
assumed deflection profiles, eqs. (7) and (27), it can be shown that
the peak tensile strain at the centre of the plate
" #
1 w20 h vqr ðr; tÞ
ε1;max ¼ εrr ðr ¼ 0; z ¼ h=2Þ ¼ max þ
0t < ∞ 2 Rz 2 vr r¼0

(37)

increases monotonically when the ratio between centre deflection

and flexural wave position increases; thus we define the parameter

w0 ðtÞ=zðtÞ t  t1
bðtÞ ¼ (38)
wtot ðtÞ=R t > t1

While such parameter is not meant to quantify the peak strain

ε1,max, it will have the same order of magnitude and can be taken as
a qualitative estimate of ε1,max. It can be seen from eq. (38) that the
time histories of b(t) are strongly affected by the flexural wave
propagation process: if the flexural wave speed z_ is low compared
to the transverse plate velocity w_ 0, bmax ¼ max[b(t)] is likely to be
reached in the early phase of the response, indicating that the peak
strain of ε1,max (eq. (37)) may occur during Phase 1, i.e. t  t1, thus
promoting early penetration of the plate in the wave-controlled
phase; on the other hand, if z_ > > w_ 0 , the peak strains are more
Fig. 2. Various mode shapes used to describe the Phase 2 response of the plate: (a)
likely to occur in the boundary-controlled Phase 2 response (t > t1).
normalised plate deflection wi ¼ wi =R ði ¼ 0; 1; 2Þ, and (b) transverse shear defor- Accordingly, two characteristic deformation modes can be identi-
mation grzi (i ¼ 0,1,2) as functions of the normalised radius r ¼ r=R. fied in this context:
 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82 73

- Mode 1: bmax is reached in Phase 2, i.e. in the boundary- gradients at the impact point, a finer mesh (element size 0.5 mm)
controlled phase of the response; the peak strains and failure was used for a central patch of radius 6 mm surrounding the
mechanisms will depend strongly on the boundary conditions impact point. The projectile was modelled as a spherical, rigid
(e.g. plate size). surface with its centre of mass located such to coincide with the z-
- Mode 2: bmax is reached in Phase 1, i.e. in the wave-controlled axis.
phase of the response; the peak strains and failure mecha- Two different composite laminates were modelled:
nisms will be less sensitive to boundary conditions.
3.1. Carbon-fibre/epoxy laminates (CFRP)
It merits comment that the above classification is based on the
assumption that the composite fails by a tensile fibre failure We considered cross-ply and quasi-isotropic CFRP laminates
mechanism. While in the case of extreme impact velocities, local- comprising of unidirectional AS4/epoxy plies [26] each of thickness
ised transverse shear failure is often the dominant failure mode [8], hl ¼ 0.125 mm and density r ¼ 1580 kg m3; the elastic lamina
the two deformation modes defined above are more relevant to properties were taken from Ref. [26] and are listed in Table 1. The
problems at the lower end of the high-velocity impact range layups of the cross-ply laminates were chosen to be symmetric,
(approximately 50e300 ms1). [0,90]ns, where n was either 2, 5 or 10 to obtain laminates of total
In Fig. 3 we plot a mode transition map in the M  h space for thicknesses h ¼ 1 mm, 2.5 mm and 5 mm, respectively. The quasi-
the case of elastic, isotropic plates with n ¼ 0.25, hence isotropic laminate had a total thickness of h ¼ 5 mm and stacking
G ¼ 1=½2ð1 þ nÞ ¼ 0:4; the contours in Fig. 3 denote the mode sequence [0,45,90,45]5s.
transitions for fixed values of non-dimensional impact velocity, v0 .
While extremely low mass ratios M cause Mode 2 to dominate, the 3.2. Glass-fibre/epoxy laminates (GFRP)
effect of increasing M is to extend the Mode 2 domain to smaller h
values; it can also be seen that if M > 1:2, Mode 2 deformation is The GFRP laminates considered here comprised unidirectional
fully suppressed for any choice of v0 . If M < 1:2, an increase in v0 E-glass/epoxy plies, each of thickness hl ¼ 0.125 mm as for the CFRP
promotes a Mode 2 response. While the transition map presented plates but with a higher density of r ¼ 2030 kg m3. The me-
in Fig. 3 is valid for plates of arbitrary stiffness E, strictly speaking, it chanical properties of the GFRP laminae (also taken from Ref. [26])
is limited to the choice of n ¼ 0.25. However, our calculations are included in Table 1. The same stacking sequences as for the CFRP
suggest that the sensitivity of the map in Fig. 3 to variations of the are analysed in the case of the GFRP.
Poisson's ratio is small in the practical range 0 < n < 0.5. The laminate was modelled in ABAQUS as a stack of transversely
isotropic laminae by using the built-in composite shell section; this
3. Finite element models approach is convenient because ABAQUS automatically computes
the laminate's stiffness matrix from the specified ply thickness,
Three-dimensional dynamic FE simulations were performed stacking sequence and material properties according to laminate
using ABAQUS/Explicit to validate the analytical model derived theory. In this study, attention is restricted to the elastic response of
above. The FE models consisted of two components, a spherical composites and therefore, damage and failure mechanisms are not
rigid projectile of radius Rs and mass M and a circular orthotropic considered in the FE calculations.
plate of radius R and thickness h; unless otherwise stated, Simulations were also performed on isotropic plates with den-
R ¼ 50 mm and Rs ¼ 5 mm. sity, Young's modulus and Poisson's ratio chosen as
The circular plate was discretised using four-noded quadrilat- r ¼ 1700 kg m3, E ¼ 50 GPa and n ¼ 0.25, respectively. In these
eral shell elements with reduced integration (S4R in ABAQUS). In simulations, an axisymmetric modelling approach was employed:
the radial direction, the element size was approximately 1.5 mm, the plate was meshed with 150 elements in the radial direction and
while 60 elements were used to discretise the plate along the 20 elements in the through-thickness direction using 4-noded
circumferential direction. In order to accurately resolve the strain axisymmetric elements with reduced integration (CAX4R in
ABAQUS).
In both 3D and axisymmetric FE models the plate was fully
clamped along its periphery, with all DOFs constrained to zero.
Impact loading was performed by imparting an initial velocity v0 to
the projectile. Contact between the plate and projectile was
assumed to be frictionless and was modelled in ABAQUS using a
surface-to-surface contact based on the penalty contact method
with a finite sliding formulation; both plate and projectile were
permitted to move independently subsequent to contact
separation.
In the following we compare the numerical and analytical pre-
dictions and explore the two characteristic deformation mecha-
nisms defined in Section 2.2.

4. Comparison of analytical and FE predictions

In this section, analytical and FE predictions of centre deflection

versus time histories are compared in order to validate the
analytical models, to explore the two characteristic deformation
modes and to examine the sensitivity of the dynamic response to
Fig. 3. Non-dimensional chart in the M  h space showing the transitions between the
the governing non-dimensional parameters (eq. (22)). In addition,
two characteristic deformation modes for the case of elastic isotropic plates with the sensitivity of plate deflection to variations of projectile size will
n ¼ 0.25 (G ¼ 0:4); contours of non-dimensional impact velocity v0 are included. be explored and discussed. Initially, we focus on isotropic material
74 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82

Table 1
Elastic properties and tensile failure strains of the CFRP and GFRP laminae considered in this study, as reported in Soden et al. [26].

E1 (GPa) E2 (GPa) n12 n21 n23 (GPa) G12 ¼ G13 (GPa) G23 (GPa) ε*1T (%)

CFRP 126 11 0.28 0.024 0.4 2.6 3.9 1.38

GFRP 45.6 16.2 0.28 0.1 0.4 6.6 5.8 2.8

behaviour, in order to compare the two types of predictions in h, M and v0 , corresponding to the two deformation modes
absence of any inaccuracy caused by the axisymmetric idealisation described in Section 2.2.
employed in the analytical models. Subsequently we probe the In Fig. 4a, analytical and FE predictions of the normalised local
accuracy of our analytical models to predict the elastic response of deformation parameter b ¼ w0/z (eq. (38))
ffi are plotted as functions
pffiffiffiffiffiffiffi
the CFRP and GFRP laminates (see Section 3) by comparing their of the non-dimensional time, t ¼ t E=r =R, for the choice h ¼ 0:1,
predictions to those obtained from detailed dynamic FE M ¼ 0:6 and v0 ¼ 0:04. We also include in this figure predictions
simulations. obtained from a reduced analytical model that ignores the excita-
tion of the additional sinusoidal mode shapes during the Phase 2
4.1. Response of isotropic plates response; the governing equations for this reduced model can be
readily obtained by setting w1 ¼ w2 ¼ grz1 ¼ grz2 ≡ 0 in eqs.
4.1.1. Deflection versus time histories (39)e(44), see Appendix A. According to the mode transition chart
Analytical and FE predictions of centre deflection versus time presented in Fig. 3, the above choice of non-dimensional parame-
histories are compared in this section for the case of fully isotropic ters should give rise to a Mode 1 behaviour, as indicated by the
elastic plates. Two axisymmetric FE simulations were performed respective marker, and the predictions presented in Fig. 4a confirm
with E ¼ 50 GPa and n ¼ 0.25 (G ¼ 1=½2ð1 þ nÞ ¼ 0:4), and the plate this, with b reaching its peak during the Phase 2 response, t > t 1 .
thickness h, the projectile mass M and the impact velocity v0 were In Fig. 4b, the same predictions are plotted in a normalised
chosen to obtain two different sets of non-dimensional parameters centre deflection w0 ¼ w0 =R versus t chart, showing excellent

Fig. 4. Mode 1 behaviour: comparison of analytical and FE predictions performed for the case of an elastic isotropic plate (n ¼ 0.25, G ¼ 0:4) with h ¼ 0:1, M ¼ 0:6 and v0 ¼ 0:04: (a)
local deformation b ¼ w0/z, (b) normalised centre deflection w0 ¼ w0 =R and (c) normalised flexural wave position z ¼ z=R as functions of non-dimensional time; results obtained
from a reduced model are included in (a) and (b) for comparison.
 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82 75

agreement between the FE and analytical predictions. On the other response, t < t 1 . The corresponding w0 ¼ w0 =R versus t traces are
hand the reduced model significantly under-predicts the deflection illustrated in Fig. 5b. While the agreement between analytical and
of the plate (this reduced model does not enforce conservation of FE predictions is satisfying in the Phase 1 response, larger dis-
kinetic energy across the transition from Phase 1 to Phase 2). crepancies occur in the ensuing Phase 2 response, in which the FE
Fig. 4c presents the corresponding predictions of normalised model predicts rapid transverse oscillations which are not picked
flexural wave position, z ¼ z=R, as functions of t. Note that the FE up by the analytical model due to the limited number of mode
predictions of zðtÞ were determined from the deformed plate shapes considered in this phase of response. However, the peak
contours by tracing the position of the elastic hinge, defined at the centre deflections, wmax 0 ¼ maxðw0 =RÞ, as predicted by the FE and
point where the slope of the deformed middle-surface in r-direc- analytical model, respectively, are found in good agreement, while
tion is zero. It can be seen from Fig. 4c that the analytical and FE the reduced analytical model, again, substantially under-predicts
predictions are in good agreement, suggesting that the proposed the FE results of wmax 0 . However recalling that, for the case of a
analytical model adequately captures the details of the flexural Mode 2 response, the onset of fibre failure is achieved during Phase
wave propagation mechanism associated with Mode 1 behaviour. 1 (t < t 1 ), these discrepancies are not relevant for the prediction of
The analytical calculations also showed that, short after the impact damage initiation.
had occurred, the speed of the flexural wave, z, _ was approximately In Fig. 5c, we present the corresponding analytical and numer-
equal to the shear wave speed in the elastic solid, cs ¼ G/ ical predictions of z ¼ z=R as functions of t. Good correlation be-
r ¼ 3430 ms1, and quickly dropped to reach a constant speed of tween both types of predictions is achieved for the initial phase of
z_ ¼ 950 ms1 . response, the flexural wave speed starts to slow down at tz1:5 in
In Fig. 5a we compare FE and analytical predictions of the the FE results, while the analytical model predicts a nearly constant
b ¼ w0/z versus t response corresponding to the choice h ¼ 0:05, wave speed of z_ ¼ 960 ms1 until the end of the Phase 1 response,
M ¼ 0:05 and v0 ¼ 0:04. It can be seen from Fig. 3 that this set of at t ¼ t 1 . However, the agreement between the two types of pre-
non-dimensional parameters results in a Mode 2 deformation dictions is still reasonably satisfactory.
response, and both types of predictions included in Fig. 5a show Analytical and FE predictions of normalised deflection profiles
that this is the case, with bmax reached clearly in the Phase 1 are compared in Fig. 6a and b for the two cases presented in Figs. 4

Fig. 5. Mode 2 behaviour: comparison of analytical and FE predictions performed for the case of an elastic isotropic plate (n ¼ 0.25, G ¼ 0:4) with h ¼ 0:05, M ¼ 0:05 and v0 ¼ 0:055:
(a) local deformation b ¼ w0/z, (b) normalised centre deflection w0 ¼ w0 =R and (c) normalised flexural wave position z ¼ z=R as functions of non-dimensional time; results obtained
from a reduced model are included in (a) and (b) for comparison.
76 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82

Fig. 6. Snapshots of normalised deflection profiles for elastic isotropic plates (n ¼ 0.25,
G ¼ 0:4): (a) Mode 1 behaviour with h ¼ 0:1, M ¼ 0:6 and v0 ¼ 0:04; (b) Mode 2
behaviour with h ¼ 0:05, M ¼ 0:05 and v0 ¼ 0:055; analytical and FE predictions are
compared.

and 5, respectively; the two snapshots presented in each figure are

taken before and after the flexural wave had reached the plate
boundary (i.e. transition between Phase 1 and 2). Fig. 6a shows that
the analytical model adequately captures the FE predictions of the
deflection profile during both phases of this Mode 1 type of
response. For Mode 2 behaviour, analytical and FE predictions are in
good agreement during the initial phase of response (i.e. for
t ¼ 0:76, Phase 1), as seen from Fig. 6b. On the other hand, during
Phase 2, the FE predictions show that plate deflection is dictated by
higher order mode shapes which are not accurately captured by the
analytical model due to the limited vibrational modes considered in
eq. (27). However, for this Mode 2 type of response, the peak strains
are expected to occur early in Phase 1 (see Section 2.2) and
therefore, from a failure perspective, the discrepancies in plate
deflection for t ¼ 6:2 are of minor relevance.

4.1.2. Sensitivity of the dynamic response to non-dimensional

Fig. 7. Analytical predictions of the maximum normalised centre deflection (dashed
parameters
lines), wmax ¼ maxðw0 =RÞ, and local deflection (solid lines), bmax ¼ max(w0/z), for the
In this section, the analytical model is employed to examine, for case of elastic isotropic plates (n ¼ 0.25, G ¼ 0:4) with M ¼ 0:05 (a) and M ¼ 0:6 (b);
the case of isotropic material behaviour (E ¼ 50 GPa, n ¼ 0.25, contours of aspect ratio h ¼ h=R are plotted for three different values and FE pre-
G ¼ 0:4), the sensitivity of the plate's deflection response to the dictions of bmax are included for comparison.
governing non-dimensional parameters (eq. (22)).
In Fig. 7a, analytical predictions of the normalised peak de-
flections, wmax ¼ maxðw0 =RÞ and bmax ¼ max(w0/z), are presented excellent agreement and show that the effect of increasing v0 is to
pffiffiffiffiffiffiffiffi monotonically increase both w0 and b for each choice of h, and the
as functions of the non-dimensional impact velocity v0 ¼ v0 r=E
for the choice M ¼ 0:05; contours of aspect ratio h ¼ h=R are slope of the h-contours in the b - v0 space increases with decreasing
included for three different values 0.02, 0.05 and 0.1, respectively. value of h, while in the w0  v0 space, the h-contours are almost
For comparison purposes, we also include in Fig. 7a predictions of parallel. In addition, Fig. 7a shows that larger aspect ratios lead to
b ¼ w0/z obtained from corresponding FE simulations (indicated by smaller values of w0 and b.
triangles, diamonds and circles). It can be seen from Fig. 7a that for Similar information is presented in Fig. 7b for the case of a much
some ranges, the analytical predictions of the peak values of w0 and higher non-dimensional projectile mass, M ¼ 0:6. For this choice of
b coincide, indicating Mode 1 behaviour, while those with w0 sb M the analytical predictions of w0 ¼ w0 =R and b ¼ w0/z coincide
are associated with a Mode 2 response, in line with the transitions almost over the entire range shown in Fig. 7b, indicating that Mode
plotted in Fig. 3. Both FE and analytical predictions are found in 1 behaviour is predominantly active. It can be seen from Fig. 7b that
 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82 77

Mode 2 behaviour is active only in case of thin plates, h ¼ 0:02, 4.2.1. Deflection versus time histories
impacted at very high velocities, v0 > 0:022; this transition can be With reference to Fig. 9, consider clamped circular composite
attributed to (i) the slowing flexural wave associated with a plates of radius R ¼ 50 mm and thickness h ¼ 2.5 mm, made from
decrease of plate thickness, and (ii) the increase in the ratio b ¼ w0/ a CFRP laminate. The plates are subject to impact loading by a
z when the impact velocity is increased, effectively promoting a rigid spherical projectile of mass M ¼ 3.1 g and radius Rs ¼ 5 mm,
Mode 2 response. Note that these observations are in line with the travelling at velocity v0 ¼ 86 ms1. The corresponding analytical
regime transitions plotted in Fig. 3. In Fig. 7b, the observed trends and FE predictions of centre deflection versus time histories are
are similar to those obtained for the M ¼ 0:05 case (compare included in Fig. 9a for two different layup choices of equal
Fig. 7a), and the analytical predictions are again in excellent thickness and areal mass, [0,45,90,45]2s (lamina thickness of
agreement with those obtained from the FE simulations. 0.156 mm, dotted curves) and [0,90]5s (lamina thickness of
0.125 mm, continuous curves). In the analytical calculations, eq.
4.1.3. Influence of projectile dimensions on the deflection response (2) was employed to calculate, for both types of laminate, the
effective Young's modulus E ~r , Poisson's ratio ~nr4 and transverse
We proceed to examine the sensitivity of plate deformation to
~
shear modulus Gr , as listed in Table 2. Fig. 9a shows that the peak
variations of projectile dimensions, as quantified here by the nor-
malised projectile radius, Rs ¼ Rs =R. First, it should be clarified that centre deflection, wmax, is found to be in good agreement be-
the effects of Rs are not accounted for in our analytical models and tween analytical and FE predictions. For the quasi-isotropic
therefore we shall restrict this study to the use of the FE method. laminate, the FE model predicts a slightly lower peak deflection
Axisymmetric FE simulations were performed on isotropic compared to the cross-ply layup, and the analytical predictions
plates (E ¼ 50 GPa, n ¼ 0.25, G ¼ 0:4) and the normalised projectile follow this trend.
radius, Rs , was varied between 0.04 and 0.5, while the parameters The corresponding predictions of flexural wave position versus
M ¼ 0:6 and v0 ¼ 0:022 were held fixed; the aspect ratio h ¼ h=R time histories z(t) are presented in Fig. 9b; the dashed and solid
was either 0.05 or 0.1. Fig. 8 reports the predicted sensitivity of the curves in this figure represent the analytical predictions, and the FE
normalised peak centre deflection wmax ¼ maxðw0 =RÞ to variations results are indicated by full and empty circles, respectively. Note
of the normalised projectile radius, Rs , for the two choices of h; that the FE results of z(t) were obtained by tracing the position of
analytical predictions corresponding to the chosen sets of non- the flexural wave front along the 0 -direction of the laminate. The
dimensional parameters are included for comparison and show analytical predictions show that the flexural wave speed associated
no sensitivity to Rs , as expected. It can be seen from the FE pre- to the response of the cross-ply laminate is slightly lower than that
dictions that variations in Rs only play a minor role in the elastic of the quasi-isotropic laminate owing to the lower effective Young's
modulus, E ~r , of the cross-ply layup (see Table 2), and the FE pre-
deformation response, justifying the assumption of vanishing
contact indentations in our analytical model. dictions confirm this scenario.
The predictions presented in Fig. 9a and b were used to plot
time histories of b ¼ w0/z, as illustrated in Fig. 9c. Both types of
4.2. Response of composite plates
predictions show that the initial peak in b(t) during the Phase 1
response (t < t1) is smaller than that found in Phase 2, hence this
Having demonstrated that our analytical models are accurate for
response is associated to Mode 1 behaviour (see Fig. 3). It can be
the isotropic case, we now proceed to examine the response of
seen from Fig. 9c that the analytical predictions of both peaks
composite plates. The sensitivity of the predicted response to var-
are found in good correlation with those predicted by the FE
iations of composite layup are investigated.
models.

4.2.2. Sensitivity of the dynamic response to non-dimensional

parameters
We proceed to examine the effects of the non-dimensional pa-
rameters (22) on the deflection response of laminated composites.
For this study, analytical calculations were conducted on CFRP
laminates with lay-up [0,90]ns and radius R ¼ 50 mm (see Section 3
for mechanical properties), and the laminate thickness h, the
impact velocity v0 and the projectile mass M were varied in order to
construct non-dimensional response maps similar to those pre-
sented in Fig. 7 for the case of isotropic plates.
Fig. 10 presents two such maps in which analytical predictions
of non-dimensional peak deflections, wmax ¼ wmax =R (dashed
curves), and maximum local deformation, bmax ¼ wmax/z (solid
curves), are plotted as functions of the non-dimensional impact
velocity v0 for the choices M ¼ 0:05 (Fig. 10a) and M ¼ 0:1
(Fig. 10b), with contours of aspect ratio h ¼ h=R included. Also
included in this figure are FE predictions of bmax obtained for
selected points within the range considered here, in order to
provide further validation of the analytical models. The two
charts illustrated in Fig. 10 show similar response characteristics
as those presented in Fig. 7 for the isotropic case, and the
analytical predictions are found in excellent agreement with the
results obtained from the FE simulations, which gives us confi-
Fig. 8. Analytical and FE predictions of the maximum normalised centre deflection,
wmax ¼ maxðw0 =RÞ, as functions of the normalised projectile radius Rs ¼ Rs =R for the
dence that our analytical models are adequate to represent the
choices M ¼ 0:6 and v0 ¼ 0:022; contours of aspect ratio are included for h ¼ 0:05 and dynamic deformation response of composite laminates subject to
h ¼ 0:1; the material properties were taken as E ¼ 50 GPa and n ¼ 0.25. ballistic impact.
78 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82

Fig. 9. Analytical and FE predictions of the deformation response for the case of CFRP plates with aspect ratio h ¼ 0:05 (see Table 2) subject to ballistic impact by a rigid ball
projectile of mass M ¼ 3.1 g and velocity v0 ¼ 86 ms1; both cross-ply [0,90]5s and quasi-isotropic [0,45,90,45]2s layups are considered: (a) centre deflection w0, (b) flexural wave
position z, and (c) normalised local deformation b ¼ w0/z as functions of time.

5. Onset of failure Before studying the failure behaviour of the laminated plates, it
is necessary to assess the accuracy of the strain predictions pro-
Having established the accuracy of the analytical predictions, vided by our analytical model. To do this we first compare analytical
the models are now employed to determine the onset of tensile strain predictions to those obtained from detailed FE simulations
failure. Impact experiments on fibre-reinforced composite plates and explore their sensitivity to variations of projectile dimensions.
(see e.g. Heimbs et al. [3]) have shown that fibre failure can initiate Then, a strain-based damage initiation criterion is stated to
at the distal face of the laminate at impact velocities much below compare the impact resistance of typical CFRP and GFRP plates, and
the ballistic limit, leading to a degradation of stiffness which can finally, we construct non-dimensional design charts which can be
appreciably affect the plate's dynamic response. The analytical used to determine the onset of tensile failure for both types of
models presented herein do not account for such failure processes; laminates.
however, they can be used to predict their first occurrence by
considering a strain-based failure criterion, which is the scope of 5.1. Time histories of fibre strain
this section. In the case of impact from a projectile with general
shape it is likely that composite plates may suffer localised damage Finite element simulations were performed on a CFRP plate of
in the proximity of the contact point. In what follows, we assume thickness h ¼ 2:5 mm, radius R ¼ 50 mm and lay-up [0,90]5s (see
that such localised damage does not appreciably affect the stiffness Table 2) subject to impact at v0 ¼ 86 ms1 by a ball projectile of
of the composite plate, i.e., contact damage is sufficiently restricted mass M ¼ 3.1 g, giving M ¼ 0:1, v0 ¼ 0:016 and h ¼ 0:05; the nor-
to a small area around the impact point. malised projectile radius Rs ¼ Rs =R was either 0.05, 0.1 or 0.2 in

Table 2
Stacking sequences, thicknesses, densities and effective elastic properties (eq. (2)) of selected GFRP and CFRP laminates.

Lamina material Layup h (mm) r (kg m3) ~r (GPa)

E ~nr4 ~ r (GPa)
G

CFRP [0,45,90,45]2s 2.5 1580 51.06 0.29 5.26

CFRP [0,90]5s 2.5 1580 45.6 0.15 5.26
GFRP [0,90]5s 2.5 2030 24.4 0.26 5.81
 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82 79

Fig. 10. Analytical predictions of the maximum normalised centre deflection (dashed Fig. 11. Analytical and FE predictions of strain induced at the tensile face of a CFRP
lines), wmax , and local deflection (solid lines), bmax, for the case of cross-ply CFRP plates plate (stacking sequence [0,90]5S, aspect ratio h ¼ 0:05) below the impact point for the
(stacking sequence [0,90]ns, see Table 2) with M ¼ 0:05 (a) and M ¼ 0:1 (b); contours of choice M ¼ 0:1: (a) fibre strain versus time histories, ε1(t), for the case v0 ¼ 0:016, and
aspect ratio h ¼ h=R are plotted for three different values, and FE predictions of bmax (b) sensitivity of the peak tensile fibre strain, ε1,max ¼ max[ε1(t)], to variations of non-
are included for comparison. dimensional velocity, v0 .

these calculations. Fig. 11a presents the predicted time histories of The FE predictions plotted in Fig. 11a also show that ε1,max is
radial fibre strain induced in the distal face of the laminate below reached approximately 15 ms after the first contact with the pro-
the impact point, ε1(t); we also include in this figure the corre- jectile has occurred. This raises questions whether the time his-
sponding analytical predictions of ε1(t) which are insensitive to Rs . tories presented in Fig. 11a are affected by propagation of the
The obtained predictions show that the strain rapidly rises and tensile precursor wave or whether radial equilibrium has been
soon reaches a peak value, ε1,max, followed by a more moderate achieved, as assumed in our analytical model (see Section 2.1.3).
decay. The FE predictions in Fig. 11a show that the variations of the This issue was clarified by tracing the propagation of the tensile
normalised plate radius, Rs , have small effect on the strain response, precursor wave in radial direction of the composite plate, as pre-
ε1(t), for the range of Rs considered here; the corresponding dicted by the dynamic FE simulations. The results showed that the
analytical prediction follows a similar trend and its peak value, tensile wave reached the plate boundary 7 ms after the impact had
ε1,max, is found in good correlation with the FE results. With the occurred (average wave speed of 6900 ms1); hence, the tensile
above loading conditions, the predicted ε1,max values are higher wave only reflected once from the fixed boundary before ε1,max was
than the quasi-static tensile ductility, ε*1T ¼ 1:38 % [26], shown in reached in Fig. 11a. However, further examination of the stress field
Fig. 11a and b as a reference. in the plate revealed that the magnitude of the tensile wave during
80 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82

0 < t < 40 ms was negligible compared to the bending stresses laminate and that the response of the GFRP plates is associated to
induced through propagation of the flexural wave; this can be larger values of ε1,max compared to CFRP plates impacted at equal
justified by the vanishingly small membrane stresses induced in velocities; this can be justified by the lower equivalent stiffness, E~r ,
the plate during this phase of response (the plate deflection was of the GFRP laminates, see Table 2. However, taking into account
less than 1.5 mm). the higher quasi-static ductility of the GFRP laminate, ε*1T ¼ 2:8 %
qffiffiffiffiffiffiffiffiffiεffi1,max
Fig. 11b shows analytical and FE predictions of peak strain (Table 1), it follows that the critical velocities, v*0 , of the GFRP
as functions of non-dimensional impact velocity v0 ¼ v0 r=E ~r and laminates are approximately 30% higher than those predicted for
reveals a nearly linear relationship between both quantities. It can the stiffer (and more brittle) CFRP plates (ε*1T ¼ 1:38 %, see Table 1),
also be seen from this figure that the effect of the projectile which allows concluding that GFRP composites generally outper-
dimension, represented by Rs ¼ Rs =R, is more pronounced when form CFRP laminates in terms of damage resilience.
the impact velocity v0 is higher. We now employ the analytical model to determine damage-
If we assume that the composite shows signs of failure at the resistant plate designs and construct design charts in the h-M-
distal face when ε1;max > ε*1T , we find a critical impact velocity at the space, for the practical ranges 0:01  M  1 and 0:02  h  0:14.
inception of failure of v*0 ¼ 43 ms1 (or v*0 ¼ 0:008 in non- Fig. 13a presents such a map for the case of CFRP cross-ply lam-
dimensional terms); this, as expected, is significantly lower than inates ([0,90]ns, G ¼ G ~ r =E
~r ¼ 0:12, ~nr4 ¼ 0:15, see Table 2); we
the ballistic limit vL, reported in the literature for a similar type of
laminate (e.g. Cunniff [12] reports vLz150 ms1 for the case of a
CFRP laminate).
The choice of identifying the limiting tensile strain with the
measured quasi-static tensile ductility ε*1T of the composite is
obviously only a first approximation; such limiting strain is ex-
pected to be influenced by the applied strain rate, the details of the
strain field and the sensitivity of the measured tensile ductility to
gauge size. Our calculations provide effective predictions of the
peak tensile strain in the laminate; these may be used as inputs for
more complex failure models, but this is not pursued in this study.

5.2. Damage resistant design of CFRP and GFRP plates

The validated analytical model is now used to explore the

damage resistance of CFRP and GFRP laminates, as quantified by the
critical velocity, v*0 , at which the failure strain, ε*1T , is reached at the
tensile face. In Fig. 12 we present analytical predictions of
maximum fibre strain, ε1,max, induced in CFRP and GFRP plates with
equal mass and layup [0,90]ns, as functions of v0 for the choice
M ¼ 0:05; contours of aspect ratio h ¼ 0:04 and h ¼ 0:08 are
included for each laminate. The predictions show that the choice of
h has vanishing effect on the magnitude of ε1,max for a given type of

Fig. 13. (a) Design chart in the M  h space for the case of CFRP laminates ([0,90]ns,
G ¼ 0:12, ~nr4 ¼ 0:15,
qffiffiffiffiffiffiffiffiffisee
ffi Table 2) with contours of non-dimensional critical impact
Fig. 12. Analytical predictions of maximum tensile fibre strain, ε1,max ¼ max[ε1(t)], as a velocity, v*0 ¼ v*0 r=E ~r , at the onset of tensile failure, ε *
1;max ¼ ε1T ¼ 1:38%; (b) similar
function of impact velocity, v0, for CFRP and GFRP plates with equal mass and layup design chart for GFRP laminates ([0,90]ns, G ¼ 0:24, ~nr4 ¼ 0:26, see Table 2) with higher
[0,90]ns; contours of aspect ratio h ¼ h=R are included for the choice M ¼ 0:05. ductility, ε*1T ¼ 2:8%.
 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82 81

include in Fig. 13a contours of non-dimensional critical impact Acknowledgements

velocity, v*0 , defined as the velocity at which the maximum tensile
strain in the laminate reaches the failure strain of the CFRP ma- The authors would like to thank EPSRC for funding through the
terial, ε1,max ¼ ε*1T ¼ 1:38 % (Table 1). Note that the area left to Doctoral Prize Fellowship scheme.
each v*0 - contour represents the design space {h,M} in which the
laminate can safely withstand ballistic impact without fibre
damage, i.e. ε1;max < ε*1T . In Fig. 13b we present a similar design Appendix A. Governing equations for the phase 2 response
map for the case of GFRP cross-ply laminates ([0,90]ns, G ¼ 0:24,
~nr4 ¼ 0:26, see Table 2) with higher ductility, εmax 1 ¼ ε*1T ¼ 2:8% The governing equations for the Phase 2 response (t > t1) are
(Table 1). obtained in terms of eqs. (27) and (29) by employing the
EulereLagrange equations (eq. (20)), and are given by

6. Conclusions 1 h 
€ þw €
 
4 € €

4€
3780p4 M w 0 2 þ 567p w1 þ w2 þ 648p w0
945p3
We developed and validated a physically based model for pre- i
€  420p2 w
þ 945p2 w € þ 11340w € þ 560w
€
dicting the dynamic deformation of fully clamped, circular elastic 1 2 1 2
composite plates subject to impact by a rigid projectile. The 1 h 2
 2
 
mathematical framework is based on first-order shear deformation ¼   3ph grz0 h þ 4w0 þ 2w2 þ 2pw0 5w20
theory of plates and takes into account large deformation, propa- 1  ~n2r4
gation of flexural waves as well as higher-order vibrational modes  2
i
emerging in the boundary-controlled phase of the response; local þ 9w22 þ h ð2grz2  3grz1 Þ ;
indentation and damage at the contact point are not explicitly (39)
modelled, which limits the applicability of the model to thin plates
impacted by relatively blunt projectiles. The constitutive response h i
1:086 0:19w€  0:345w € þ 0:268w €
of the composite was linear elastic with effective stiffness deduced 0 1 2
from the stiffness matrix of the laminate; the mathematical 0:0833 h 2
formulation of plate deflection was based on axisymmetric shape ¼   h ð2:82grz0  35:9grz1 þ 17grz2 þ 243:6w0
functions assumed a-priori. This approach yields a set of nonlinear 1  ~n2r4
ODEs which can be solved using common numerical integration i
methods. þ 159:9w2 Þ þ 120w30 þ 216w0 w22 ;
The dynamic response was found to be governed by only four (40)
non-dimensional parameters, namely h, M, v0 and G, representing
aspect ratio, mass ratio, non-dimensional impact velocity and  
0:1403w€  0:291w € þ 0:352w € þM w € þw €
transverse shear stiffness, respectively. Two characteristic defor- 0 1 2 0 2
mation modes were identified and non-dimensional transition 0:0833 h 2
maps were constructed. ¼   h ð  35:85grz0  25:45grz1 þ 112:9grz2
The analytical models were validated by comparing their pre- 1  ~n2r4
dictions to those of detailed dynamic FE simulations and a good i
correlation was found for a wide range of plate geometries, pro- þ 177:9w0 þ 1064:4w2 Þ þ 486w32 þ 270w2 w20 ;
jectile masses and impact velocities. It was shown that neglecting (41)
additional vibrational modes during the boundary-controlled
phase of the response can lead to under-predictions of the plate's 2
centre deflection and peak strain. In addition, detailed FE simula- 0 ¼ h ð35:85w2  11:82w0  6:59grz0 þ 0:449grz1 þ 3:804grz2 Þ
   
tions showed that the deflection response of elastic plates is only
 3:75Ggrz0 1  ~n2r4 þ 0:597Ggrz1 1  ~n2r4
mildly sensitive to the projectile radius.
 
The sensitivity of peak tensile strain to variations of the (non- þ 0:424Ggrz2 1  ~n2r4 ;
dimensional) projectile velocity was examined for two types of
composites, (i) carbon-fibre/epoxy (CFRP) and (ii) glass-fibre/epoxy (42)
(GFRP) laminates. It was shown that tensile failure of plies may
initiate early, in the wave-controlled phase of response, and the 2
0 ¼ h ð38:76w0  25:45w2  0:449grz0 þ 5:71grz1  2:70grz2 Þ
critical velocities associated to the inception of damage were found,    
as expected, below the measured limit velocities for full penetra-  0:597Ggrz0 1  ~n2r4 þ 1:25Ggrz1 1  ~n2r4
tion. It was found that the GFRP plates can sustain higher impact  
velocities at the inception of failure compared to plates made from  0:486Ggrz2 1  ~n2r4 ;
CFRP of equal mass, concluding that GFRP composites outperform
(43)
stiffer and more brittle CFRP laminates in terms of impact damage
resilience. 2
0 ¼ h ð18:88w0 þ 112:9w2  3:804grz0  2:70grz1 þ 11:98grz2 Þ
The critical velocity v*0 provided by the calculations presented    
here can be interpreted as (1) a lower bound on the ballistic limit of  0:424Ggrz0 1  ~n2r4  0:486Ggrz1 1  ~n2r4
the plate or (2) as an upper bound on the maximum impact velocity  
which a certain plate can sustain with no damage. The analytical þ 1:25Ggrz2 1  ~n2r4 :
models were used to construct design maps for both CFRP and GFRP
(44)
laminates in order to aid the selection of damage-resistant plate
geometries.
82 A. Schiffer et al. / International Journal of Impact Engineering 85 (2015) 67e82

References [14] Zener C. The intrinsic inelasticity of large plates. Phys Rev 1941;59:669e73.
[15] Olsson R. Impact response of orthotropic composite plates predicted from a
one-parameter differential equation. AIAA J 1992;30:1587e96.
[1] Graff KF. Wave motion in elastic solids. Oxford: Clarendon Press; 1975.
[16] Sun CT, Chattopadhyay S. Dynamic response of anisotropic laminated plates
[2] Takeda N, Sierakowski RL, Ross CA, Malvern LE. Delamination-crack propa-
under initial stress to impact of a mass. J Appl Mech 1975;42:693e8.
gation in ballistically impacted glass/epoxy composite laminates. Exp Mech
[17] Whitney JM, Pagano NJ. Shear deformation in heterogeneous anisotropic
1982;22:19e25.
plates. J Appl Mech 1970;37:1031e6.
[3] Heimbs S, Bergmann T, Schueler D, Toso-Pentecote N. High velocity impact on
[18] Dobyns AL. Analysis of simply-supported orthotropic plates subject to static
preloaded composite plates. Compos Struct 2014;111:158e68.
and dynamic loads. AIAA J 1981;19:642e50.
[4] Chang FK, Chang KY. A progressive damage model for laminated composites
[19] Qian Y, Swanson SR. Experimental measurement of impact response in car-
containing stress concentrations. J Compos Mater 1987;21:834e55.
bon/epoxy plates. AIAA J 1990;28:1069e74.
[5] Matsumoto H, Ogawa Y, Adachi T. Stress analysis of an orthotropic laminated
[20] Hoo Fatt MS, Palla L. Analytical modeling of composite sandwich panels under
slab subject to a transverse load. JSME Int J I Solid M 1991;35:165e9.
blast loads. J Sandw Struct Mater 2009;11:357e80.
[6] Wu E, Tsai CZ. Impact behaviour and analysis of CFRP laminated plates. In:
[21] Phoenix SL, Porwal PK. A new membrane model for the ballistic impact
Reid SR, Zhou G, editors. Impact behaviour of fibre-reinforced composite ma-
response and V50 performance of multi-ply fibrous systems. Int J Solids Struct
terials and structures. Cambridge: Woodhead Publishing Ltd.; 2000. p. 212e38.
2003;40:6723e65.
[7] Abrate S. Impact on laminates composites: recent advances. Appl Mech Rev
[22] Sierakowski RL, Chaturvedi SK. Dynamic loading and characterization of fiber-
1994;47:517e44.
reinforced composites. New York, USA: Wiley-Interscience; 1997.
[8] Cantwell WJ, Morton J. Impact perforation of carbon fibre reinforced plastic.
[23] Schiffer A, Tagarielli VL. The dynamic response of composite plates to un-
Compos Sci Technol 1990;38:119e41.
derwater blast: theoretical and numerical modelling. Int J Impact Eng
[9] Wen HM. Predicting the penetration and perforation of FRP laminates struck
2014;70:1e13.
normally by projectiles with different nose shapes. Compos Struct 2000;49:
[24] Latourte F, Gregoire D, Zenkert D, Wei X, Espinosa HD. Failure mechanisms in
321e9.
composite panels subjected to underwater impulsive loads. J Mech Phys
[10] Mines RAW, Roach AM, Jones N. High velocity perforation behaviour of
Solids 2011;59:1623e46.
polymer composite laminates. Int J Impact Eng 1999;22:561e88.
[25] Takeda N, Sierakowski RL, Malvern LE. Wave propagation experiments on
[11] Zhu G, Goldsmith W, Dharan CKH. Penetration of laminated kevlar by pro-
ballistically impacted composite laminates. J Compos Mater 1981;15:157e74.
jectiles - I. Experimental investigation. Int J Solids Struct 1992;29:339e420.
[26] Soden PD, Hinton MJ, Kaddour AS. Lamina properties, lay-up configurations
[12] Cunniff PM. Dimensionless parameters for optimization of textile based body
and loading conditions for a range of fibre-reinforced composite laminates.
armor systems. In: Proceedings of the 18th International Symposium on
Compos Sci Technol 1998;58:1011e22.
Ballistics. San Antonio: Texas; 1999. p. 1303e10.
[13] Karthikeyan K, Russell BP, Fleck NA, Wadley HNG, Deshpande VS. The effect of
shear strength on the ballistic response of laminated composite plates. Eur J
Mech A Solids 2013;42:35e53.