See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/232389874

Modeling of impacts on composite structures

Article  in  Composite Structures · February 2001

DOI: 10.1016/S0263-8223(00)00138-0

CITATIONS READS

351 1,294

1 author:

S. Abrate
Southern Illinois University Carbondale
134 PUBLICATIONS   5,237 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Development of an aeronautical crashworthy seat View project

Dynamics of Multilayer Structures View project

All content following this page was uploaded by S. Abrate on 19 February 2019.

The user has requested enhancement of the downloaded file.

 Composite Structures 51 (2001) 129±138
www.elsevier.com/locate/compstruct

Modeling of impacts on composite structures

Serge Abrate
Department of Technology, College of Engineering, Southern Illinois University, Carbondale, IL 62901-6603, USA

Abstract
Impacts of foreign objects on composite structures can create internal damage that reduces the strength of the structure sig-
ni®cantly. The study of such impacts requires understanding the dynamics of the event, predicting the extent of the induced damage,
and estimating the residual properties of the structure. The impact event involves the motion of the target, the motion of the
projectile, and the local indentation in the contact zone. A large number of parameters a€ect the impact dynamics, many types of
responses can be obtained, and many models have been proposed in the literature. These models can be classi®ed into three cat-
egories: (1) energy-balance models that assume a quasi-static behavior of the structure; (2) spring-mass models that account for the
dynamics of the structure in a simpli®ed manner; (3) complete models in which the dynamic behavior of the structure is fully
modeled. Simple models can bring insight into the problem and be ecient but have limited applicability. Complex models may have
wider applicability but require signi®cantly higher modeling and computational e€ort. There is a need for a general understanding of
the impact dynamics and for a method for developing ecient and accurate models. Ó 2001 Elsevier Science Ltd. All rights
reserved.

Keywords: Impact dynamics; Models; Laminated composite materials

1. Introduction analysis. Some impacts produce deformations in a small

zone surrounding the point of impact while others in-
Laminated composite materials are used extensively, volve deformations of the entire structure. In some
but their behavior under impact is of concern since cases, a major portion of the impact energy is trans-
damage which can be introduced reduces the strength of ferred to the plate and in other cases most the energy is
the structure signi®cantly. Comprehensive literature re- restituted to the projectile. For some problems, the in-
views on the e€ect of impact on composite materials dentation absorbs a signi®cant portion of the impact
were presented by Abrate [1±4]. The large number of energy so that it must be modeled adequately in the
papers published is an evidence of the interest generated analysis. In other cases, the e€ect of indentation are
by this topic and the high level of research activity. Most negligible. Sorting out these di€erent types of behavior is
of the work reported is concerned with low-velocity necessary for the interpretation of experimental results
impact and both the damage resistance and damage and for the selection of an appropriate mathematical
tolerance aspects of it. model.
A ®rst step towards understanding the e€ect of im- The objectives of this article are to study the various
pacts is to develop a model for predicting the contact models available for analyzing the impact dynamics and
force history and the overall response of the structure. It to present an approach for selecting an appropriate
involves modeling the motion of the projectile, the dy- model for each particular case. The many models used
namics of the structure, and the local indentation of the to study the impact dynamics are classi®ed here ac-
structure by the projectile. Experimental modal analysis cording to how the structure is modeled: spring-mass
[5] showed that low velocity impact damage has only models, energy balance models, complete models, and a
minor e€ects on the dynamic properties of laminated model for impact on in®nite plates. Simpli®ed models of
plates. Small shifts in the natural frequencies of higher- the overall deformation of the structure lead to simple
order bending modes are observed which con®rms that more ecient models for the impact dynamics analysis
damage needs not modeled in the impact dynamics and provide insights into the impact response. Section 2
discusses the mechanics of contact between a smooth
indentor and a composite material. An energy-balance
E-mail address: [email protected] (S. Abrate). model is used in Section 3 to analyze impacts on a half

0263-8223/01/$ - see front matter Ó 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 2 6 3 - 8 2 2 3 ( 0 0 ) 0 0 1 3 8 - 0
130 S. Abrate / Composite Structures 51 (2001) 129±138

place and in Section 4, a method for analyzing impacts

P ˆ Pm ‰…a ÿ a0 †=…am ÿ a0 †Š3=2 : …6†
on in®nite plates is recalled. Spring-mass models and
complete models are discussed in Sections 5 and 6. Ex- In some cases, only a small fraction of the impact energy
amples are presented in Section 7 and a discussion of the is used in the local indentation process and therefore it is
procedure followed for selecting an appropriate model is not necessary to distinguish between loading and un-
presented in Section 8. loading branches, and Eq. (1) is used throughout the
indentation process. Sometimes the indentation process
need not be modeled at all.
2. Contact mechanics

Local deformations in the contact zone are not 3. Energy-balance models

modeled with beam, plate or shell theories since those
theories usually assume that the structure is inextensible One approach for analyzing the impact dynamics is
in the transverse direction. However, in many cases, to consider the balance of energy in the system. The
local indentation has a signi®cant e€ect on the contact initial kinetic energy of the projectile is used to deform
force history and must be accounted for in the analysis. the structure during impact. Assuming that the structure
The contact phenomenon is recognized as being rate behaves quasi-statically, when the structure reaches its
independent for most laminated composite materials maximum de¯ection, the velocity of the projectile be-
and statically determined contact laws are used by most comes zero and all the initial kinetic energy has been
investigators. During the loading phase of the impact, used to deform the structure. Therefore, the energy-
the contact force P is related to the indentation a by balance equation can be written as
1
P ˆ ka3=2 : …1† MV 2 ˆ Eb ‡ Es ‡ Em ‡ Ec ; …7†
2
The contact sti€ness is given by where the subscripts b, s, m refer to the bending, shear,
4 and membrane components of the overall structural
k ˆ ER1=2 ; …2† deformation and Ec is the energy stored in the contact
3
region during indentation.
where the parameters R and E are de®ned as When the overall de¯ections of the structure are
1 1 1 1 1 ÿ m21 1 ÿ m22 negligible compared to the local indentation, the prob-
ˆ ‡ and ˆ ‡ ; …3† lem is reduced to that of an impact on a half-space and
R R1 R2 E E1 E2
the maximum contact force, and the contact duration
where R1 and R2 are the radii of curvature of the two are given by
bodies. The Young's moduli and Poisson's ratios of the  3=5
5  3 6 2 1=5
two bodies are E1 ; m1 and E2 ; m2 , respectively. Subscripts Pˆ M V k ; …8a†
4
1 denotes properties of the indentor, while subscript 2
identi®es properties of the target. Eq. (1) is usually re- 1=5
M2

ferred to as the Hertzian law of contact. Tc ˆ 3:2145 : …8b†
Permanent indentation occur even at relatively low Vk 2
loading levels, and the unloading phase of the process is These simple expressions show the e€ect of the projectile
signi®cantly di€erent from the loading phase. During mass and velocity and the contact sti€ness on the con-
unloading, the contact law is tact force.
2:5
P ˆ Pm ‰…a ÿ a0 †=…am ÿ a0 †Š ; …4†
4. Impacts on in®nite composite plates
where Pm is the maximum force reached before un-
loading, am the maximum indentation, and a0 is the
Olsson [7] presented an approximate solution for
permanent indentation. a0 is zero when the maximum
impacts on in®nite plates. With an Hertzian contact law
indentation remains below a critical value acr . When (Eq. (1)), non-dimensional indentation and time vari-
am > acr ,
ables can be de®ned as
h i
a0 ˆ am 1 ÿ …acr =am †2=5 : …5† a t
aˆ ; tˆ ; …9†
TV T
During subsequent reloading, the reloading curve is
distinct from the unloading curve but always returns to where
the point where unloading began [6]. The unloading
2=5
T ˆ M kV 1=2
 ÿ
curve is modelled by : …10†
 S. Abrate / Composite Structures 51 (2001) 129±138 131

The non-dimensional indentation is governed by the

single, nonlinear, ordinary di€erential equation
d2 a 3 da
2
‡ k a1=2 ‡ a3=2 ˆ 0; …11†
dt 2 dt
which depends on a single non-dimensional parameter
h pi
k ˆ kc2=5 V 1=5 M 3=5 = 8 mD …12†

called the inelasticity parameter. The equivalent bending

rigidity in Eq. (12) is de®ned as
A‡1
D ˆ …D11 D22 †1=2 ;
2 …13†
1=2
A ˆ …D12 ‡ 2D66 †=…D11 D22 † : Fig. 1. (a) Two-degree-of-freedom model. (b) Single-degree-of-free-
dom model.
Eq. (11), with the initial conditions
da…0†
a…0† ˆ 0; ˆ1 …14† linear sti€ness of the structure (Kbs ), another spring Km
dt
for the nonlinear membrane sti€ness, a mass M2 repre-
must be solved numerically. The contact force is given senting the e€ective mass of the structure, the nonlinear
by contact sti€ness, and M1 the mass of the projectile. If the
1=5
P ˆ k 2 M 3 V 6 a3=2 :

…15† e€ect of shear deformation is negligible, the spring
constant Kbs is replaced by Kb which account for bend-
After contact ceases, the de¯ection at the point of im- ing deformations only. From the free body diagrams of
pact remains constant while the deformation propagates the two masses M1 and M2 , the equations of motion of
outward. During an impact on an in®nite plate, the the system can be written as
wavefront is nearly elliptical. The distance a between the
M1x1 ‡ P ˆ 0; …18a†
wavefront and the impact point in the x-direction, can
be estimated using M2x2 ‡ Kbs x2 ‡ Km x32 ÿ P ˆ 0; …18b†
p D11 1=4 p
 
1=8 where P is the contact force which is a highly nonlinear
aˆ2 p ‰2…A ‡ 1†Š t: …16†
m function of the indentation x1 ÿ x2 .
The dynamics of the system described by Eqs. (15)
In the y-direction, the wavefront is approximately lo-
and (16) and the initial conditions
cated at a distance
1=4
x_ 1 …0† ˆ V ; x_ 2 …0† ˆ 0; x1 …0† ˆ x2 …0† ˆ 0 …19†
b ˆ a=…D11 =D22 † : …17†
can be studied numerically.
Eqs. (16) and (17) provide an estimate of the size of the The sti€ness used in the spring-mass models can be
deformed zone during impact. determined from formulas available in many handbooks
In [4], this model has been extended for other contact or, numerically, using the ®nite element method for ex-
laws and it was shown that for a linear contact law, Eq. ample [8]. For a completely clamped, isotropic, circular
(11) becomes identical to the equation of motion of a plate, the bending and membrane sti€nesses are given by
single-degree-of-freedom system with viscous damping.
In that case, the non-dimensional parameter is analo- 4pEh3
Kb ˆ ; …20†
gous to the damping coecient and the physical inter- 3…1 ÿ m2 †a2
pretation is that the deformation of the plate dissipates
…353 ÿ 191m†pEh
some of the impact energy. Km ˆ ; …21†
648…1 ÿ m†a2
where E is the elastic modulus, m the Poisson's ratio, h
5. Spring-mass models the thickness, and a is the radius of the plate. The ef-
fective mass of the plate is taken as one-fourth of the
5.1. General case total mass of the plate [9].

Spring-mass models are simple and provide accurate 5.2. Linear two-degree-of-freedom model
solutions for some types of impacts often encountered
during tests on small size specimens. The most complete In many cases, the transverse de¯ections are small
model (Fig. 1) consists of one spring representing the and membrane-sti€ening e€ects are negligible and the
132 S. Abrate / Composite Structures 51 (2001) 129±138

Table 1
Approximate formulas for natural frequencies of two-degree-of-free-
dom system
k2 k2
1 1
k1 k1
Case I r Case IIIr
m1 m 1 k2 m 1 k2
1 x1 ˆ x1 ˆ
m2 m 2 k1 m 2 k1
x2 ˆ 1
r
m1
x2 ˆ 1 ‡
m2
Fig. 2. Linear two-degree-of-freedom spring-mass impact model. Case IIr Case IV
m1 k2 x1 ˆ 1
1 x1 ˆ
m2 k1
sti€ness of the structure can be represented by the linear r r
spring k2 . In order to understand the dynamics of the m1 m 1 k2
x2 ˆ 1 ‡ x2 ˆ
impact, we also assume that the local indentation can be m2 m 2 k1
represented by a linear spring k1 (see Fig. 2). The motion
of the linear two-degree-of-freedom system is governed
by veering'' phenomenon for the two natural frequencies as
m1 =m2 increases for a ®xed k2 =k1 ratio.
m1x1 ‡ k1 …x1 ÿ x2 † ˆ 0;
…22† For case I ……m1 =m2 †  l; …k2 =k1 †  1†, the natural
m2x2 ‡ k2 x2 ‡ k1 …x2 ÿ x1 † ˆ 0 frequencies are approximated by
with the initial conditions r r
k2 k1
x1 …0† ˆ x2 …0† ˆ 0; x_ 1 …0† ˆ V ; x_ 2 …0† ˆ 0: …23† x1 ˆ ; x2 ˆ …28†
m2 m1
Introducing the non-dimensional variables and the free vibration modes are as shown in Fig. 4(a).
With mode I, the two masses move together but m1 is
r p
k1 k1 =m1
sˆt and yi ˆ xi : …24† negligible compared to m2 . For mode II, the heavy mass
m1 V
remains stationary and oscillates as a single-degree-of-
Eqs. (22) can be written as freedom system with sti€ness k1 .
y100 ‡ y1 ÿ y2 ˆ 0; For case II ……m1 =m2 †  1; …k2 =k1 †  1†, the mass of
the target is small compared to the mass of the impactor
…25†
 
00 m1 m1 k2 and the sti€ness of the target is small compared to the
y2 ÿ y1 ‡ 1‡ y2 ˆ 0;
m2 m2 k1 contact sti€ness and the natural frequencies are ap-
and the initial conditions become proximated by

y10 …0† ˆ 1; y20 …0† ˆ 0:

r r
y1 …0† ˆ y2 …0† ˆ 0; …26† k2 k1
x1 ˆ ; x2 ˆ : …29†
The behavior of the system depends on the two non- m1 m2
dimensional parameters m1 =m2 and k2 =k1 . In mode I, the indentation is negligible and the projectile
The non-dimensional natural frequencies of the and the target form a SDOF system (Fig. 4(b)). In
system are obtained by solving the bi-quadratic equa- Model II, the projectile remains stationary, the sti€ness
tion of the target is negligible, and the target oscillates as a
SDOF system.
  
4 2 m1 k2 m1 k2
x ÿx 1‡ 1‡ ‡ ˆ 0: …27† For case III ……k2 =k1 †  1; …m1 =m2 †  1†, the sti€ness
m2 k1 m2 k1
of the target is much larger than the contact sti€ness, the
The natural frequencies can be approximated as indi- mass of the projectile is small compared to that of the
cated in Table 1 in which we distinguish four special target. Then
cases. Fig. 3(a) shows that as k2 =k1 becomes large, the
®rst natural frequency tends to the limiting value of one.
r r
k2 k1
The convergence rate is strongly a€ected by the mass x1 ˆ ; x2 ˆ : …30†
m2 m1
ratio when m1 =m2  1. When k2 =k1 is larger than one,
the second
p
 non-dimensional frequency increases with The mode shapes are shown in Fig. 4(c). For case IV
k2 =k1 and the validity of the approximations given in ……k2 =k1 †  1; …m1 †=…m2 †  1†, the sti€ness of the target
Table 1 is shown in Fig. 3(b). Fig. 3(c) shows a ``curve is much larger than the contact sti€ness and
 S. Abrate / Composite Structures 51 (2001) 129±138 133

Fig. 3. Free vibrations of two-degree-of-freedom system: (a) ®rst non-dimensional frequency; (b) second non-dimensional frequency (solid line:
exact; dashed line: approximate); (c) Curve veering phenomenon for k2 =k1 ˆ 10 (solid line: exact; dashed line: approximate).

Fig. 4. Approximate mode shapes of linear two-degree-of-freedom system.

r r For Mode I, the target remains stationary and the
k1 k2 projectile oscillates (Fig. 4(d)). For mode II, the motion
x1 ˆ ; x2 ˆ : …31†
m1 m2 of the plate is una€ected by the presence of the projectile
134 S. Abrate / Composite Structures 51 (2001) 129±138

due to the weak coupling provided by the contact initial velocity of the projectile is suciently large,
sti€ness. damage can be introduced plate motion is established.
In that case, a three-dimensional analysis is required.
5.3. Linear single-degree-of-freedom (SDOF) spring- Once an appropriate theory is selected, all the vibration
mass model modes participating in the response have to be predicted
accurately and must be retained in the model. For a
A signi®cant simpli®cation occurs when membrane- simply supported plate, for example, an analytical so-
sti€ening e€ects are negligible and the indentation is lution can be found for the natural frequencies and
small compared with the overall deformation of the mode shapes. The transient response is then expressed in
structure. In that case, a linear SDOF system (Fig. 1(b)) terms of these mode shapes and all participating modes
can be used and the contact force is then given by can be included. For other geometries or boundary
p p  conditions, variational or ®nite element models must be
P ˆ V Kbs M sin Kbs =M t : …32† used. With such approximate methods, a suciently
large number of degrees of freedom must be selected so
Eq. (32) predicts that the maximum contact force is di- that the participating modes are predicted accurately.
rectly proportional to the initial velocity of the projec- If N equations are needed to describe the motion of
tile. Similarly, the maximum contact force is the structure and one equation for the projectile, the
proportional to the square root of the kinetic energy. N ‡ 1 di€erential equations can be written in matrix
The contact force increases with the square root of the form as
sti€ness of the structure and the square root of the mass
of the impactor. A sti€er structure will cause a harder ‰MŠfX g ‡ ‰KŠfX g ˆ fF g …35†
impact and, for the same initial velocity, a larger mass
and integrated using Newmark's step-by-step time in-
will have a larger kinetic energy which will also increase
tegration method. The contact force is unknown and is a
the contact force.
nonlinear function of the indentation. Therefore, the
force vector in Eq. (35) is assumed to be known at the
5.4. Nonlinear SDOF spring-mass models end to the nth time step but its value at the end of step
n ‡ 1 is unknown. In order to determine the displace-
There are two situations for which a nonlinear SDOF ments at the end of step n ‡ 1, we start assuming that
can provide accurate predictions of the contact force fP gn‡1 ˆ fP gn and solve equation (35) for a ®rst esti-
history. In the ®rst instance, the overall de¯ection of the mate of fX gn‡1 . A new estimate of fP gn‡1 can be cal-
structure is negligible compared to the local indentation. culated from these displacements and a new iteration
In that case, the spring in Fig. 1(b) represents the con- can be performed. After several iterations, the solution
tact sti€ness, and the equation of motion is converges so that both the equations of motion and the
3=2
M1x1 ‡ kx1 ˆ 0: …33† contact law are satis®ed and the process is repeated for
the following time steps.
The second situation for which nonlinear a SDOF
model can yield accurate predictions of the contact force
history is when the local indentation is negligible but the
7. Examples
de¯ections of the structure becomes large and mem-
brane sti€ening is signi®cant. The equation of motion of
7.1. Wave controlled impacts
the single-degree-of-freedom model is
Mx ‡ kb x ‡ km x3 ˆ 0: …34† Qian and Swanson [10] studied the impact of a steel
sphere on graphite±epoxy laminates with the following
A numerical solution of these nonlinear equations of
material properties:
motion (Eqs. (33) and (34)) yields the dynamic response
and contact force histories. E1 ˆ 120 GPa; E2 ˆ 7:9 GPa; G12 ˆ 5:5 GPa;
3
m12 ˆ 0:30; q ˆ 1580 kg=m :
6. Complete models The diameter of the impactor is 12.7 mm, its mass is
8.537 g, and the initial velocity is 3.0 m/s. The Hertzian
3=2
With a complete model, the dynamic behavior of the contact sti€ness is k ˆ 8:394  108 N=m . The plates
structure is described accurately. This means that the studied where simply supported and had a size of 200 
appropriate structural theory is used. For example, in 200 mm. Three plates with a ‰0; 90; 0; 90; 0ŠS layup were
many cases the classical plate theory can be used but, in considered. Plate A had a total thickness of 2.69 mm
some cases, transverse shear deformations become sig- (0.269 mm/layer), plate B is 5.38 mm thick (0.538 mm/
ni®cant and higher-order theories must be used. If the layer), and plate C is 10.76 mm thick (1.076 mm/layer).
 S. Abrate / Composite Structures 51 (2001) 129±138 135

The inelasticity parameter k is equal to 2.055, 0.5139, the elliptical deformed zone is a ˆ 143 mm if the plate is
and 0.1285 for plates A, B, and C, respectively. There- in®nite so in this case again the impact can be consid-
fore, di€erent behaviors are expected for these three ered to be in®nite and the e€ect of shear deformation is
plates with identical layups but di€erent thicknesses. negligible. The central de¯ection at the end of the con-
For plate A, the numerical solution to Eq. (11) gives tact duration is 0.02371 mm which is very small and
the contact force history shown in Fig. 5 with a maxi- geometric nonlinearities are negligible. The maximum
mum force of 285.4 N and a contact duration of contact force predicted in [10] is 562.8 N in good
257:4 ls. The compressive wave that propagates agreement with the present analysis. The maximum in-
through the thickness of the laminate in the early stages dentation is 0.07921 mm and the plate de¯ection cannot
of the impact has a velocity be neglected.
p For plate C, the contact force history predicted by the
c ˆ E2 =q: …36† impact on in®nite plate model (Fig. 5) gives a maximum
Here c ˆ 2236 m/s, so it takes just 1:2 ls for the wave to of 802.7 N and a contact duration of 105:5 ls. Eq. (16)
travel through the thickness of the laminate. Therefore gives a ˆ 192 mm and therefore, waves are close to be
multiple re¯ections occur and the bending motion is re¯ected back to the impact point but shear deforma-
established. Since k ˆ 2:055, the contact force history is tions and membrane sti€ening are still negligible. The
asymmetric with the maximum occurring 37:7 ls after maximum indentation is 9:706  10ÿ2 mm, while the
the start of the impact event. After 257:4 ls, the size of maximum central plate de¯ection is 7:489  10ÿ3 mm.
the deformed zone predicted by Eq. (16) is a ˆ 150 mm As plate de¯ections become small, the problem tends to
which means that the disturbance initiated by the impact that of an impact on a half-space. If plate de¯ections are
at the center of the plate would travel 150 mm in the x- neglected, the energy balance approach predicts a
direction and similarly 150 mm in the ÿx-direction in an maximum contact force of 910 N and a contact duration
in®nite plate. For a 200  200 mm plate, the disturbance of 103:4 ls. While the plate de¯ection is small, relative
reaches the boundary and is re¯ected back, but does not to the indentation in this case, the maximum contact
have sucient time to travel back to the point of impact. force cannot be accurately predicted when plate de¯ec-
Therefore, the in®nite plate model is still valid and the tions are neglected. For plate C, [10] gives a maximum
maximum contact force of 286.8 N predicted by Qian contact force of 759.3 N which again agrees well with
and Swanson [10] is identical to that predicted here. This the prediction from the impact on in®nite plate model.
model is based on the classical plate theory while [10]
used the ®rst order shear deformation theory. Because 7.2. Boundary controlled impacts
the size of the deformed zone is large compared to the
thickness of the plate, the e€ect of shear deformation is Choi and Hong [8] studied impact on ‰0; 90Š4S
negligible. graphite±epoxy laminates with the following material
For plate B, the contact force history predicted by the properties:
impact on in®nite plate model is also shown in Fig. 5.
E1 ˆ 135:4 GPa; E2 ˆ 9:6 GPa;
The maximum force predicted is 591.7 N and the con-
3
tact duration is 116:5 ls. The length of the semi-axis of m12 ˆ 0:31 GPa; G12 ˆ 4:8 GPa; q ˆ 1580 kg=m ;

with a ply thickness of 0.1125 mm. The size of the plate

is 100  100 mm and the four edges are fully clamped.
The radius of the steel impactor is 6 mm and its modulus
of elasticity and Poisson's ratio are E ˆ 207 GPa and
m ˆ 0:3. The mass of the plate is mP ˆ 28:5 g and the
mass of the impactor M is varied so that n ˆ M=mP ˆ 1;
3:5; 10, and 35.
In this case, the inelasticity parameter k ˆ 10:39 when
the mass ratio is equal to one and k ˆ 87:72 when the
mass ratio is equal to 35. Therefore, for all the cases
considered in [8], the impact event is dominated by the
deformation of the plate. For k ˆ 10:39, the impact on
in®nite plate model predicts a contact duration of 5.03
ms. During that predicted contact duration, the wave-
front emanating for the impact point at the center of the
plate would travel 546 mm in the x-direction if the plate
Fig. 5. Example 1: Contact force histories for impacts on plates A, B, were of in®nite extent. In this ®nite size plate, waves
C. reach the boundary and are re¯ected back and forth
136 S. Abrate / Composite Structures 51 (2001) 129±138

several times during the predicted contact duration.

Therefore, this is a boundary-controlled impact and
spring-mass model with successfully predict the contact
force history. A single-degree-of-freedom spring-mass
model with a spring constant calculated assuming that
the plate is circular and has a radius of 50 mm, so that
Eq. (20) can be used. The results shown in Fig. 6 are in
excellent agreement with the ®nite element results in [8].
For an impact velocity of 2.76 m/s and a mass ratio of
35, the maximum de¯ection of the plate predicted by the
spring mass model is 4.29 mm, more than twice the
thickness of the plate. A rule of thumb is that, when
de¯ections become larger that the thickness of a plate,
geometrical nonlinearities play an important role.
Modeling the plate as a clamped circular plate again, the
membrane sti€ness can be estimated using Eq. (21) and Fig. 7. Example 2: E€ect of membrane sti€ening on contact force
the motion of the single-degree-of-freedom model is history (solid line: linear, dashed line: nonlinear).
governed by Eq. (34). The nonlinear e€ect due to
membrane sti€ening cause a signi®cant reduction of the
contact duration and a substantial increase in the max- 7.3. Impact on aluminum plate
imum contact force (Fig. 7).
The energy-balance approach predicts a maximum The example studied by Shivakumar et al. [9] deals
contact force of 1773 N when only plate bending is with the impact of a 19-mm radius steel sphere on a
considered …Km ˆ 0; Ec ˆ 0† which is the same value 3.2-mm thick, aluminum, circular plate of 38-mm
predicted by the linear spring-mass model. Including the radius. For steel, the modulus of electricity E, Poisson's
e€ect of indentation, the maximum contact force drops ratio m, and density q are taken as E ˆ 199:95 GPa,
to 1748 N. Considering bending and membrane defor- m ˆ 0:33; q ˆ 7971:8 kg=m3 . For aluminum, E ˆ 68:95
mations but neglecting, Ec , the energy balance approach GPa, m ˆ 0:33; q ˆ 2768 kg=m3 , and the shear modulus
yields a maximum contact force of 3315 N which is G ˆ 25:92 GPa. The initial velocity of the projectile is
nearly identical to that obtained using the nonlinear 2.54 m/s, the indentation is assumed to follow Hertz
spring-mass model. When bending, membrane and contact law, and the impact occurs in the center of the
contact deformations are considered together, the plate.
maximum contact force drops to 3127 N. The e€ect of In this case, k ˆ 14:70 which implies that most of the
indentation is much larger when geometric nonlinear energy is used in the deformation of the plate. A nu-
behavior of the plate is considered. Since, for this plate, merical solution of Eq. (11) gives a contact duration of
the side length to thickness ratio is 56, shear deforma- the impact on an in®nite plate of 8.69 ms. During the
tion e€ects are negligible. duration for impact, Eq. (16) predicts that the radius of
the deformed zone in an in®nite plate is 27.47 m, many
times the radius of the ®nite size plate considered here.
Therefore, the deformation of the plate is expected to be
quasi-static. The bending rigidity of the clamped circular
plate is calculated using Eq. (20). Neglecting shear de-
formation and membrane sti€ening, the energy balance
approach gives a maximum contact force P ˆ 3167 N.
Only 7.7% of the initial kinetic energy of the projectile is
used in the indentation process. The bulk of the initial
impact energy is used for the quasi-static bending of the
target. Using a linear, single-degree-of-freedom spring-
mass model, the maximum contact force is 3296 N and
the contact duration 5.54 ms. This provides a good
approximation to the contact force history even though
it does not capture the small oscillations predicted by a
complete model using a ®nite element discretization of
the plate [11].
Fig. 6. Example 2: Contact force histories for impacts on a ‰0; 90Š4S Refs. [9,11] indicate that, as the thickness of the plate
graphite±epoxy plate as a function of the mass ratio. increases past 15 mm, the contact duration decreases
 S. Abrate / Composite Structures 51 (2001) 129±138 137

and tends to a limit. The maximum contact force pre-

dicted by the energy balance approach increases with the
plate thickness and reaches a limit (Fig. 8). As the
thickness becomes large, plate de¯ections become neg-
ligible and we have essentially an impact on a half-space.
The inelasticity parameter k is inversely proportional
to the square of the thickness h and for h ˆ 6 mm;
k ˆ 4:18. The energy balance approach predicts a
maximum contact force of 7109 N, a maximum dis-
placement of the projectile of 0.2232 mm a maximum
plate de¯ection of 0.1465 mm and that plate bending
absorbs 70.5% of the impact energy. The two-degree-of-
freedom spring-mass model con®rms these predictions
but also reveals the presence of oscillations in the mo-
tion of the plate (Fig. 9). These oscillations are re¯ected
in the contact force history (Fig. 10) If the mass of the
plate is neglected …M2 ˆ 0†, the two-degree-of-freedom
Fig. 9. Impact of a steel sphere on a 6-mm thick aluminum plate (solid
model predicts a smooth contact force history. A single- line: projectile displacement, dashed line: plate de¯ection).
degree-of-freedom spring mass model can be used to
predict the contact force history in this case. The
equivalent spring constant is determined by dividing the
maximum contact force by the maximum displacement
of the projectile predicted by the energy balance ap-
proach. The contact force history predicted by Eq. (22)
is in excellent agreement with that obtained from the
two degree of freedom model.
When h ˆ 15 mm, k ˆ 0.6690 and the indentation
process is expected to absorb a major portion of the
kinetic energy of the projectile. The energy balance
model predicts a maximum contact force of 13,320 N
and that 84% of the initial kinetic energy of the pro-
jectile is used in the indentation process. The maximum
de¯ection of the plate is small compared to the maxi-
mum indentation. The two degree of freedom spring±
mass model will provide accurate prediction of the
contact force history in this case.
Fig. 10. Impact of a steel sphere on a 6-mm thick aluminum plate:
contact force history (solid line: two-degree-of-freedom model, dashed
line: two-degree-of-freedom with M2 ˆ 0, dotted line: single degree of
freedom model).

The limit case is when the de¯ection of the plate is

negligible and then, the energy balance approach (Eqs.
8) predict a maximum contact force of 14,777 N and a
contact duration of 0.148 ms respectively, in good
agreement with the results given in [9,11].

8. Selection of an impact dynamics model

The selection of an appropriate impact dynamics

model starts by neglecting the de¯ections of the plate
and using the energy±balance model (Eqs. 8) to obtain a
Fig. 8. Example 3: Impact of a steel sphere on an aluminum plate: ®rst estimate of the maximum contact force and of the
maximum contact force as a function of plate thickness. contact duration. The in®nite plate model is used to
138 S. Abrate / Composite Structures 51 (2001) 129±138

determine the e€ect of plate de¯ections and the type of the overall deformation of the plate. Then we can de-
response of the target. The inelasticity parameter k de- termine whether the deformation reached the plate
termines how much of the impact energy is absorbed by boundaries during the duration of the impact. If bend-
the deformation of the plate. When k  1, the defor- ing waves travel from the impact point to the edge of the
mation of the plate is negligible and the energy balance plate and back many times during the predicted contact
approach is satisfactory. Otherwise, the dynamics of the duration, we have a boundary controlled impact and a
plate play an important role. If the deformation front spring mass model or an energy balance approach might
has not reached the boundary of the plate, the in®nite be adequate because in that case, the plate behaves in a
plate model is a very ecient and accurate way of an- quasi-static manner. If the deformation never reaches
alyzing the impact. When the deformation front prop- the edges of the plate, we have a wave-controlled impact
agates many times more than the distance from the and the approximate solution provides very good re-
impact point to the boundary, the ®nite size of the target sults. For intermediate cases, the in®nite plate model
must be accounted for. might be adequate initially but re¯ected waves will a€ect
In the examples discussed here as well as in many the contact force history. Then, a complete model taking
examples in the literature, the contact deformation is into account the full dynamic behavior of the plate and
modeled using Hertz's law for both loading and the boundary conditions will be necessary. Examples
unloading. However, contact laws have been investi- demonstrating this procedure also show that in many
gated at length and some authors account for di€erent cases a simple model can provide accurate predictions of
behavior in the unloading and reloading phases (Eqs. the contact force history.
4, 6). When k is small, when the contact force reaches
its maximum, most of the energy is used to indent the
target so that energy losses due to indentation can be References
signi®cant. Therefore, when k is large, the indentation
[1] Abrate S. Impact of composite laminates. Appl Mech Rev
uses only a small fraction of the impact energy so that
1991;44(4):155±90.
it is not necessary to model the contact behavior [2] Abrate S. Impact of laminated composites: recent advances. Appl
carefully. Mech Rev 1994;47(11):517±44.
[3] Abrate S. Localized impact on sandwich structures with laminated
facings. Appl Mech Rev 50(2):69±82.
[4] Abrate S. Impact on composite structures. Cambridge: Cam-
9. Conclusions
bridge University Press; 1998.
[5] Tracy JJ, Dimas, DJ, Pardoen GC. The e€ect of impact damage
This article presents an overview of mathematical on the dynamic properties of laminated composite plates. In:
models used for the analysis of the dynamics of impacts Proceedings of the Fifth International Conference on Composite
between a foreign object and a composite structure. Materials, ICCM-V, 29 July±1 August; 1985. San Diego, CA;
1985. p. 111±25.
Currently available models are classi®ed into four cat-
[6] Yang SH, Sun CT. Indentation law for composite laminates.
egories: spring±mass models, energy±balance models, ASTM STP 1982;787:425±49.
complete models, and an impact on in®nite plate model. [7] Choi IH, Hong CS. New approach for simple prediction of impact
Simple models are easy to use and ecient but have force history on composite laminates. AIAA J 1994;32(10):2067±
limitations due to the simplifying assumptions on which 72.
[8] Shivakumar KN, Elber W, Illg W. Prediction of Impact force and
they are based. Selecting an appropriate model requires
duration due to low-velocity Impact on circular composite
on understanding of the e€ects the many factors a€ect- laminates. J. Appl Mech 1985;52:674±80.
ing the impact dynamics. In this paper, a procedure is [9] Olsson R. Impact response of orthotropic composite plates
presented in order to determine the type of impact to be predicted from a one-parameter di€erential equation. AIAA J
expected and to select an appropriate model. The pro- 1992;30(6):1587±96.
[10] Qian Y, Swanson SR. A comparison of solution techniques for
cess starts by assuming that the plate is in®nite. In this
impact response of composite plates. Comput Struct 1990;14:
case, a simple approximate solution requiring to solve a 177±92.
single nonlinear ordinary di€erential equation is avail- [11] Wang CY, Yew CH. Impact damage in composite laminates.
able. This model predicts the contact force history and Comput Struct 1990;37(6):967±82.

View publication stats