International Journal of Impact Engineering 58 (2013) 66e75

Contents lists available at SciVerse ScienceDirect

International Journal of Impact Engineering

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

Design and numerical simulation of a new sandwiched sphere

structure for ballistic protection
Yibin Fu, Jun Zhou, Xiaosheng Gao*
Department of Mechanical Engineering, The University of Akron, Akron, OH 44325, USA

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

Article history: Sphere structures for ballistic protection have rarely been studied. In this paper a novel unit cell, called
Received 3 October 2012 Square Based Pyramid Unit Cell (SBPUC), is created as a fundamental building block of sphere structures
Received in revised form for the purpose of ballistic protection. Although it has been proved by the other authors that the
5 March 2013
arrangement of spheres in a tapered chain is an efficient way for impact energy absorption (Sokolow
Accepted 6 March 2013
A, Pfannes JM, Doney RL, Nakagawa M, Agui JH. Absorption of short duration pulses by small, scalable,
Available online 19 March 2013
tapered granular chains. Appl Phys Lett 2005;87:1e3), it is very difficult to convert this 1D concept to 3D
application in reality with arbitrary tapering angle. The adoption of SBPUCs may provide a good solution.
Keywords:
Sphere structure
In this research, the core of a sandwiched sphere structure is build with such SBPUCs and is characterized
Ballistic protection with increasing sphere diameter, decreasing sphere thickness, and increasing sphere cap thickness. From
Energy absorption the numerical simulation results with ABAQUS/Explicit, the sandwiched sphere structure can absorb at
Structure design least 11.6% more impact energy compared with the monolithic plate with the same area density. This
Numerical simulation study might provide a potential armor design platform which makes further performance improvement
possible by optimizing the large parameter space associated with material selection and structure design.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction absorption ability of a tapered chain can be dramatically enhanced

by placing small interstitial grains between the regular grains in the
The objective of this research is to study sphere structures for tapered chain systems. But to the authors’ knowledge, this 1D
ballistic protection. A sphere is a perfectly round geometrical object concept has not yet been converted into 3D application in reality for
in three-dimensional space and it is known as having the smallest armor design. The arrangement of spheres in 1D with any desired
surface area among all surfaces enclosing a given volume. As one of degree of tapering is easy to realize, however, the 3D arrangement
the natural shapes, sphere has been extensively studied and various of spheres with variable diameter in different layers presents a
sphere structures have been widely employed in engineering ap- challenge.
plications. However, sphere structure for ballistic protection has In the study conducted by Zeng et al. [4] on the impact behavior
rarely been studied. Andrews [1] might be the first one who gave a of hollow sphere structures, where spheres are randomly arranged,
detail description of 3D multiple-layer sphere structures in his ar- the functionally graded design concept was introduced, i.e. the
mor design. It is indicated that, in contrast to a solid steel plate, the density of hollow spheres varies gradually within the structure.
primary advantage provided by the sphere structures (energy This study indicates that placing the hardest layer of spheres as the
dispersion objects) is that the energy associated with an incoming first impacted layer and the weakest layer of spheres as the last
ballistic projectile is at least partially dispersed toward the perim- layer has some benefits in terms of maximizing energy absorption.
eter, rather than directing all of the energy straight through the However, in another study conducted by these authors [5], it is
protecting layers. indicated that the best gradient profile in terms of the energy ab-
The important research by Sokolow et al. [2] demonstrates that sorption capacity should be a weaker first layer of spheres and a
a small alignment of progressively shrinking spheres (Tapered progressively enhanced core. In the current design, these two
Chain) of a strong, light-mass material, placed horizontally in an concepts are combined to make the first and last layers harder and
appropriate casing, can absorb about 80e90% of the incident en- the middle layer weaker. Layers of ceramic spheres embedded in
ergy. Further research by Doney et al. [3] indicates that the shock epoxy for composite armor design can be found in [6], but with
little description about the sphere structure.
* Corresponding author. Tel.: þ1 330 972 2415; fax: þ1 330 972 6027. Shell or hollow sphere may have excellent energy-absorbing
E-mail address: [email protected] (X. Gao). capacity during high speed impact if design properly. Dadrasi [7]

0734-743X/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijimpeng.2013.03.005
 Y. Fu et al. / International Journal of Impact Engineering 58 (2013) 66e75 67

studied the energy-absorbing capacity of thin-walled shells element (type S4R) were successfully used to model the steel plate
impacted by a rigid plate. Fig. 1 shows the loadedeformation curve under ballistic impact [9]. Guo et al. [8] presented a preliminary
and the deformation modes. The energy absorption ability is related criterion to decide whether a panel-like target can be modeled by
to the wall thickness, shell diameter, impact mass and impact ve- shell elements.
locity. With the propagation of the deformation, three modes of In this study, a novel unit cell, called Square Based Pyramid Unit
deformation are observed, namely axisymmetric ring, stationary Cell (SBPUC), is created as a fundamental building block of sphere
plastic hinges, and inversion. The slope of the loadedeformation structures for the purpose of ballistic protection. The perforation of
curve is not constant. The formation of stationery plastic hinge ab- the new sphere structure with sandwich core built with SBPUCs by
sorbs most impact energy among the three failure modes. a flat-nose projectile is numerically simulated. Three representative
Experimental study could be overly expensive. Analytical study impact locations are considered and the effect of the yield stress of
may be impossible for such a complex impact problem. Numerical the target material is examined. This study may provide a potential
modeling can provide insights to understand the penetration armor design platform, which makes further performance
process. In this paper, all perforations are simulated by using a improvement possible by optimizing the large parameter space
commercial finite element code ABAQUS/Explicit. Due to the lim- associated with material selection and structure design, such as the
itation of the computational resources, most FEA perforation sim- adoption of woven fabric among spheres, polymer-coated or fiber-
ulations in the literature were performed using 2D models. In this encapsulated spheres, or the adoption of gradient at various levels
study the 3D model is used due to non-symmetric impact loadings in sphere diameter and density.
in most cases. To control the parametric study within an acceptable
time schedule, the geometry of the 3D model is scaled down.
2. Material models and validation study
Therefore, the projectile considered in this study is only 2 mm in
diameter, which is probably the smallest size of the bullet for a
2.1. Material model of the target
riffle air gun. Furthermore, the adoption of shell elements for some
3D hollow spheres can reduce simulation time significantly. Ac-
In this paper, the JohnsoneCook model (JC), which accounts for
cording to the research done by Guo et al. [8] the major difference
strain rate sensitivity and thermal softening, is used to describe the
between the capability of a 3D solid element model and that of a
behavior of the target material under high velocity impact. Ac-
shell element model lies on the stress state of the material under
cording to the JC plasticity model [10], the equivalent stress s is
consideration. Unlike the 3D stress state in a solid element, the
given as a function of the equivalent plastic strain εpl , plastic strain
normal stress along the thickness direction is neglected in a shell pl
rate ε_ and temperature T
element. As a result, shell elements are not capable of accounting
!
for the stress wave propagation in the target thickness direction.
 h
n i  _ pl 
ε
When the influence of the normal stress on the target failure pl
s εpl ; ε_ ; T ¼ A þ B εpl 1 þ Cln bm
1T (1)
cannot be ignored, the solid elements have to be employed. But ε_ 0
when the perforation of a projectile through a target is dominated
by certain failure modes, shell elements are able to model the where A is the initial yield stress, B is a material constant, n is the
material failure of the targets with good accuracy. For example, in strain hardening parameter, ε_ 0 is a reference strain rate, C is the rate
the perforation simulation of a thin (0.4 mm) steel plate impacted sensitivity parameter, and m is the temperature sensitivity
by a projectile at velocities between 200 and 600 m/s, shell b is the non-dimensional homologous tem-
parameter. In Eq. (1), T
perature defined as

b ¼ T  T0
T ; T0  T  Tmelt (2)
Tmelt  T0

where T is the current temperature, Tmelt is the melting tempera-

ture, and T0 is the room temperature.
The fracture model proposed by Johnson and Cook [11] takes
into account the effect of plastic strain rate, temperature and stress
triaxiality s*. Failure is assumed to occur when the damage
parameter D, as defined by Eq. (3), reaches unity.
0 1

 X Dεpl
pl _ pl * @ A
D ε ; ε ; T; s ¼ (3)
pl  pl
εf ε_ ; T; s*

where Dεpl is the plastic strain increment, εpl f

is the equivalent
strain at failure, and the summation is performed over the defor-
mation increments of the loading history. The strain at failure, εplf
, is
assumed to be dependent on a non-dimensional plastic strain rate,
pl
ε_ =_ε0 , a non-dimensional stress triaxiality ratio, s* ¼ sm =s (where
sm is the mean stress), and the non-dimensional temperature, Tb ,
defined earlier. The dependencies are assumed separable and are of
the form given below

 h
i  _ pl 
ε

pl
εf
pl
ε_ ; T; s* ¼ D1 þ D2 exp D3 s* 1 þ D4 ln b
1 þ D5 T
Fig. 1. The loadedeformation curve and the deformation modes [7]. (A) Formation of ε_ 0
axisymmetric ring. (B) Formation of stationery plastic hinge. (C) Formation of
inversion. (4)
68 Y. Fu et al. / International Journal of Impact Engineering 58 (2013) 66e75

where D1 to D5 are material constants. When this failure criterion is Table 2

met, the “element killing” algorithm is invoked to delete the failed Material constants of the hardened Arne tool-steel for the projectile [13].

elements from the mesh. E (MPa) n r (kg/m3) s0 (MPa) Et (MPa)

The target material in this study is the Weldox 460 E steel. The 204,000 0.33 7850 1900 15,000
material constants were identified by Borvik et al. [12] and are
listed in Table 1.

each simulation the target plate is fully clamped at the support,

2.2. Material model of the projectile
while the projectile is given an estimated initial velocity which is
close to the corresponding ballistic limit velocity provided by Bor-
A simplified constitutive model is adopted for the projectile [13].
vik et al. [14]. The initial size of the smallest element in the impact
The projectile is modeled as a bilinear elasticeplastic von Mises
region is 0.125  0.1 mm2 in all simulations for target thicknesses
material with isotropic hardening, i.e.
equal to or less than 12 mm, resulting 60 (for 6 mm thick target

plate) to 120 (for 12 mm thick target plate) elements throughout
Eε; ε  ε0
s¼ (5) the plate thickness. For thicker targets, the element size in the
s0 þ Et ðε  ε0 Þ; ε > ε0
impact region is increased to 0.25  0.2 mm2 in order to reduce the
where E is the Young’s modulus, s0 is the yield stress, and ε0 ¼ s0/E computational time. This leads to 80 (for 16 mm thick target plate)
is the yield strain. The material constants for the projectile are given to 100 (for 20 mm thick target plate) elements through the thick-
in Table 2. The strain-rate effect and possible damage are neglected ness. The contact between the projectile and the target plate is
due to the higher yield stress (1900 MPa) compared to the target modeled using surface-to-surface contact with finite sliding
material (490 MPa). formulation. The frictional effect is neglected for blunt projectiles in
accordance with experimental observations by Borvik et al. [16]. For
each simulation, the steel target plate is meshed with 4-node,
2.3. Validation study
bilinear, axisymmetric, quadrilateral elements with reduced inte-
gration (CAX4R). Adaptive meshing is employed due to large
Borvik et al. [12e16] conducted an extensive study on perfora-
deformation with strong element distortion in high speed perfo-
tion of the monolithic plate under various impacting speed and
ration problems. The projectile is meshed with the same type of
target thickness. The relation they got between the ballistic limit
elements (CAX4R) having 0.4  0.4 mm2 element size. According to
velocity and the target thickness can be used for this validation
the mesh sensitivity study conducted by Borvik et al. [14], the nu-
study. In the study, the length and diameter of the blunt-nosed
merical results tend to converge toward a limit solution when the
cylindrical projectile is 80 mm and 20 mm respectively. The
number of elements over the target thickness is sufficiently large.
target plate is clamped on a rigid circular frame with an inner clamp
The mesh sensitivity is most distinct close to the ballistic limit,
diameter of 500 mm. The thickness of the target plate varies from 6
while at higher velocities the results are far less affected by mesh
to 30 mm. In the numerical simulation it is assumed that the target
size. In order to reduce the computational time, the element size is
is fully clamped at the support. This assumption has been proved
gradually enlarged toward the boundary.
reasonable in that the boundary conditions are of minor impor-
The ballistic limit velocities, obtained from experiments [14],
tance in ballistic penetration by small mass projectiles in the
from LS-DYNA simulations [14], and from ABAQUS/Explicit simu-
ordnance velocity regime (500e1300 m/s) as long as the target
lations (current study), are given in Table 3 and plotted against the
diameter is more than just a few projectile diameters.
target plate thickness in Fig. 3. It is worth mentioning that in the
In the validation study, a series of simulations need to be carried
LS-DYNA simulations performed by Borvik et al. [14] the modified
out to obtain the final ballistic limit velocity for each target thick-
JohnsoneCook model is used and the plasticity and damage
ness. Fig. 2 shows the finite element model for the blunt projectile
models are coupled. In ABAQUS/Explicit, the original Johnsone
and a 6 mm thick target plate. Because of symmetry, the finite
Cook model is used and the plasticity and damage models are
element simulations are performed using axisymmetric models. In
not coupled. The numerical values from the current study are
within 10% of the corresponding experimental values and nu-
Table 1
merical values from [14] for all target thicknesses. The agreement
Material constants of the Weldox 460E steel for the target [12]. is especially good for target thickness less than 10 mm. The
equivalent thickness of the new structures to be described in detail
Elastic constants and density E (GPa) 200
n 0.33
in the following sections is only 1.84 mm. Therefore, the ABAQUS/
r (kg/m3) 7850 Explicit built-in JohneCook model with the material parameters
Yield stress and strain hardening A (MPa) 490 given in Table 1 should be able to provide valid numerical results
B (MPa) 383 for the perforation simulations of the new sandwiched sphere
n 0.45
structures.
Strain-rate hardening ε0 (s1) 5  104
C 0.0114
Damage evolution εd 0
Dc 1.0
Adiabatic heating and temperature softening Cp (J/kgK) 452
c 0.9
a (K1) 1.1  105
Tm (K) 1800
T0 (K) 293
m 0.94
Fracture strain constants D1 0.0705
D2 1.732
D3 0.54
D4 0.015
D5 0
Fig. 2. Finite element model of the projectile and target plate for the validation study.
 Y. Fu et al. / International Journal of Impact Engineering 58 (2013) 66e75 69

Table 3
Ballistic limit velocities versus target thickness.

Target thickness (mm) Ballistic limit velocities (m/s)

Experimental Numerical Numerical

valuesa valuesb valuesc
6 145.5 149.1 143
8 154.3 157.6 149
10 165.3 161.1 153
12 184.5 174.3 164
16 236.9 234.4 218
20 293.9 291.9 264
a
Values obtained from [14].
b
Values obtained from [14].
c
Values from current study.

3. Configuration of the sandwiched sphere structure

3.1. Square based pyramid unit cell (SBPUC)

The square based pyramid unit cell (SBPUC) is the basic building
block of the core of the sandwiched sphere structure. Fig. 4 shows
an image of the SBPUC. An SBPUC is consists of four small spheres
with radius r and one big sphere with radius R. The four centers of Fig. 4. The square based pyramid unit cell (SBPUC).
the small spheres form a square which serves as the base of an
imaginary pyramid with the top vertex at the center of the big
rear plate, the three caps, the boundary, and spheres in layer 2 and
sphere. Here R is two times of r (R ¼ 2  r). The five spheres are
layer 3 are meshed with 4-node linear shell elements with reduced
closely packed such that they are in touch with each other.
integration (S4R) and with the smallest element size of 0.1 mm in
Therefore, the length of the four base edges of the imaginary pyr-
the impact region. The spheres in layer 1 are meshed with 4-node
amid is 2  r, the length of the four edges from a base vertex to the
linear tetrahedron elements (C3D4) and with the smallest element
pffiffiffithe pyramid is R þ r, and the height of the imaginary
top vertex of
size of 0.1 mm in the impact region. The projectile is meshed with
pyramid is 7*r.
8-node linear brick elements with reduced integration (C3D8R) and
with 0.25 mm element size. Adaptive meshing is used to handle the
3.2. Configuration of the sandwiched sphere structure issue of large deformation and element distortion associate with
high speed perforation.
The core of the sandwiched sphere structure is characterized
with increasing sphere and cap diameters, decreasing sphere
3.3. Other possible configurations of the sandwiched sphere
thickness, and increasing cap thickness. The dimensions of the
sandwiched sphere structure are shown in Table 4. The length and structure
radius of the projectile are 4 mm and 1 mm respectively. The as-
sembly of the structure is shown in Figs. 5 and 6 (section view) and Fig. 8 shows other possible configurations of the sandwiched
sphere structure. It must emphasize that this figure is only to show
the finite element mesh is shown in Fig. 7. The front plate, rear plate
and boundary are fully clamped along their periphery. The that there are other design options. For simplicity and clarity of the
paper, these configurations are not analyzed. The simulation results
boundary is a simplified representation of outside spheres, not
shown in the figure, contacting with inside spheres, shown in the presented in this paper are for the sandwiched sphere structure
shown in Fig. 5.
figure. All spheres are bounded together with polymer, but the
bounding forces are neglected in the simulation. The front plate,
3.4. Equivalent monolithic plate

The equivalent thickness of the monolithic plate is 1.84 mm for

the sandwiched sphere structure based on the same area density.
Instead of adopting a square plate, a more convenient round plate is

Table 4
Dimensions of the sandwiched sphere structure.

Radius Height Thickness Number of Length Width Depth

(mm) (mm) (mm) spheres (mm) (mm) (mm)
Front plate 1/4 12 12
Layer r1 0.75 1/4 64
Layer r2 1.5 1/8 16
Layer r3 3 1/16 4
Layer cap1 6 6 1/32
Layer cap2 6.2 4.3 1/16
Layer cap3 7 3 1/8
Rear plate 1/4 12 12
Boundary 1/8 12.25 12.25 14.64
Fig. 3. Ballistic limit velocities versus target thickness.
70 Y. Fu et al. / International Journal of Impact Engineering 58 (2013) 66e75

Fig. 5. Isometric full view of the sandwiched sphere structure (without boundary).

Fig. 6. Isometric section view of the sandwiched sphere structure (with boundary).

simulated because the boundary shape is of minor importance in

high speed ballistic penetration by small mass projectiles. The
radius of the plate is 12.5 mm. Fig. 9 shows a quarter of the full
assembly. Fig. 10 shows the finite element mesh of the assembly.

Fig. 8. Other possible configurations of the sandwiched sphere structure.

The plate is fully clamped along its outer periphery and has two
symmetric boundaries. The projectile is assigned an initial impact
velocity of 600 m/s. The plate is meshed with 4-node linear tetra-
hedron elements (C3D4) and with the smallest element size of
Fig. 7. Finite element mesh (section view) for the sandwiched sphere structure 0.1 mm in the impact region. The projectile is meshed with 8-node
assembly. linear brick elements with reduced integration (C3D8R) and with
 Y. Fu et al. / International Journal of Impact Engineering 58 (2013) 66e75 71

first layer of the sandwich core are selected as the entry point of the
projectile. One is in the center of a sphere. The second is located at
the mid-point between two neighboring spheres. And the last is at
the central point among four adjacent spheres. The locations are
depicted in Fig. 11.

4.1.2. Yield stress of the target material

According to the US patent (US8096223 B1) [1], the spheres in
the sandwich core should have a hardness that optimizes their
energy dispersion properties. In other words, if the spheres are too
hard, the strike from a projectile will simply shatter the spheres and
a minimal amount of energy will be dispersed outwards. On the
other hand, if the spheres are too soft, the spheres will deform
around an incoming projectile rather than moving against each
other and a minimal amount of energy will be dispersed outwards.
Therefore, sphere hardness should be an important factor to be
considered in the simulation. According to Oyen’s comment [17] on
the relationship between hardness and elastic modulus, elastic
modulus is an intrinsic material property and fundamentally
related to atomic bonding. Hardness is an engineering property and
for metals it can be related to the yield strength. Therefore, a yield
Fig. 9. Isometric view of the monolithic plate and the projectile. stress ratio is defined as a ¼ sy/sy0, where sy0 ¼ 490 MPa is the
original yield stress of the material. The yield stress ratios of 1.0, 1.1,
1.2, 1.3, 1.4, 1.5, and 1.6 are considered in the numerical simulations
respectively.
0.25 mm element size. Adaptive meshing mechanism is used due to
large deformation with strong element distortion in high speed
perforation problem. 4.2. Simulations of the sandwiched sphere structure and the
monolithic plate

4. Numerical simulations and simulation results 4.2.1. The sandwiched sphere structure
Totally 21 cases (three locations and seven yield stress ratios) are
4.1. Variables to be studied simulated for the sandwiched sphere structure. Fig. 12 shows the
example simulation for the case of location 1, a ¼ 1.5. The pene-
The effects of two control variables, the impact location and the tration process is recorded every 20 ms to help understanding the
yield stress of the target, are investigated in this study. The pro- tumbling effect of the projectile. At t ¼ 60 ms the projectile hits the
jectile’s initial impact velocity is set at 600 m/s and its residual rear plate and it has tumbled significantly and thus increases the
velocity needs to be determined for all cases. projectile’s effective contact area with the rear plate. At t ¼ 200 ms
the projectile has already penetrated through the structure and it
4.1.1. Impact location tumbles to the opposite direction, which is also helpful if another
It is important that the performance of the system is similar protection layer or subsystem is arranged within a distance just a
wherever the projectile hits. Three representative locations in the few projectile lengths behind the existing structure. For entry lo-
cations 1 and 2, after the projectile hitting the first layer of spheres
the projectile does not tumble since the symmetric reaction forces

Fig. 10. Finite element mesh of the monolithic plate and the projectile. Fig. 11. Three representative entry locations.
72 Y. Fu et al. / International Journal of Impact Engineering 58 (2013) 66e75

Fig. 12. Side section view of penetration for the case of location 1, a ¼ 1.5.
 Y. Fu et al. / International Journal of Impact Engineering 58 (2013) 66e75 73

it received. But after hitting the second layer of spheres the pro-
jectile start to tumble because of the non-symmetric reaction force
received from the impacted curved sphere surface. Theoretically
speaking the projectile should not tumble after hitting the entry
location 3 because the projectile keeps receiving symmetric reac-
tion forces as it penetrates through all layers of spheres. But from
the simulation result, at the entry location 3, the projectile still
tumbles slightly after penetrating through the whole sandwiched
sphere structure. This is because the 3D assembly mesh is not
perfectly axisymmetric.
The density of hollow spheres and sphere caps varies gradually
within the structure - the adoption of functionally graded design
concept. Placing the hardest layer of spheres or sphere cap as the
first and the last impacted layer and the weakest layer of spheres or
sphere cap as the middle layer may have some benefits in terms of
maximum energy absorption. The energy associated with the
incoming projectile is partially dispersed toward the perimeter,
which can be seen by observing the boundary deforming as the
projectile proceeding. The energy associated with the incoming
projectile is partially dispersed to those spheres and sphere caps in
different layers; this can be seen by observing the motion of
deforming, translating, or rotating of some spheres and sphere
caps. In this design, the sphere diameters increase gradually in
different layers, such that the impacted smaller spheres in the
anterior layer can be obstructed gradually and completely from the
bigger spheres in the posterior layers. The sphere diameter in the
last layer of spheres should be a little bit larger than the length of
the projectile to provide enough room to allow the projectile
rotating after penetrating through the first and second layers of
spheres. In this design, the diameter of the spheres in the third layer
Fig. 14. Penetration at t ¼ 50 ms for the case of a ¼ 1.5.
is 6 mm, which is 1.5 times bigger than the length of the projectile.
Fig. 13 shows the velocity of the projectile, measured at pro-
jectile head, center, and tail respectively, versus time. From this case of a ¼ 1.5 at t ¼ 50 ms. No tumbling effect is observed from this
graph, the energy absorption in different phases can be roughly figure. Fig. 15 shows the velocity of the projectile, measured at
estimated. At t ¼ 20 ms, the velocity of the projectile drops from projectile head, center, and tail respectively, versus time. From the
600 m/s to about 340 m/s, the projectile reaches the third layer of velocity curve, at t ¼ 5 ms the velocity of the projectile drops from
spheres, and about 67.9% of the total kinetic energy has been 600 m/s to about 230 m/s, most part of the projectile has pene-
absorbed by the structure. At t ¼ 100 ms, the velocity of the pro- trated through the monolithic plate, and about 85.3% of the total
jectile drops to about 40 m/s, which is close to the residual velocity, kinetic energy has been absorbed by the plate. At t ¼ 10 ms, the
the projectile almost penetrates through the rear plate, and about velocity of the projectile drops to the average residual velocity
99.5% of the total kinetic energy has been absorbed by the struc- 208 m/s, the projectile completely penetrates through the rear
ture. Before t ¼ 100 ms, i.e., before the projectile completely pene- plate, and about 87.9% of the total kinetic energy has been absorbed
trates through the target, the three velocity curves, measured at by the plate. The three velocity curves, measured at projectile tail,
projectile tail, center, and head, have different fluctuating deceler- center, and head, fluctuate first at the time just penetrating through
ating rates, which reflects the tumbling effect observed in Fig. 12. the plate and then tend to converge to a fixed value. The translation
of stress wave, back and forth within the projectile in the impacting
4.2.2. The monolithic plate direction, makes the velocities different at projectile tail, center,
Totally seven cases (seven yield stress ratios) are simulated for and head. It is worth noting that before t ¼ 5 ms, i.e, the projectile
the monolithic plate. Fig. 14 shows the example simulation for the has not completely penetrated through the target, the three

Fig. 13. Projectile velocity versus time for the case of location 1, a ¼ 1.5 (measured at Fig. 15. Projectile velocity versus time for the case of a ¼ 1.5 (measured at projectile
projectile head, center, and tail). head, center, and tail).
74 Y. Fu et al. / International Journal of Impact Engineering 58 (2013) 66e75

Table 5
Residual velocities of different cases for the sandwiched sphere structure and the monolithic plate.

Residual velocity vr (m/s)

a ¼ 1.0 a ¼ 1.1 a ¼ 1.2 a ¼ 1.3 a ¼ 1.4 a ¼ 1.5 a ¼ 1.6

Sandwiched sphere structure Location 1 229 214 167 191 126 42 39
Location 2 203 192 147 123 31 0 0
Location 3 230 186 154 176 183 0 32
Monolithic plate Center 268 261 249 234 223 208 193

velocity curves show little difference, which reflects no tumbling

effect ever happened.

4.3. Summary of the simulation results

The simulation results are summarized in Table 5 and are

plotted in Fig. 16. Twenty-one results are for the sandwiched sphere
structure and seven results are for the monolithic plate. The new
structure performs best at location 2 and worst at location 1. The
new structure performs better than the monolithic plate at all lo-
cations. The effect of yield stress to the new structure is stronger
than that to the monolithic plate. For the monolithic plate, the re-
sidual velocity decreases monotonically with the increasing of the
Fig. 18. Projectile velocity versus time for the case of location 1, a ¼ 1.0, with enlarged
yield stress ratio. However, for the new sandwiched sphere struc- boundary size.
ture, there exists an optimized yield stress value at a ¼ 1.5. At

a ¼ 1.5, the projectile cannot penetrate through the new structure

300 at locations 2 and 3. Although it can penetrate through at location 1,
Residual Velocity (m/s)

250 the residual velocity is only 42 m/s. At a ¼ 1.5, the projectile can
200 New Structure, Location 1 penetrate through the monolithic plate with a high residual ve-
150
New Structure, Location 2 locity of 208 m/s. At a ¼ 1.5, the new structure can absorb about
New Structure, Location 3 99.5% of the total impact energy at location 1 (the worst location),
100 Monolithic Plate, Center
while the monolithic plate can absorb about 87.9% of the total
50
impact energy. Thus, the new structure can absorb 11.6% more
0 impact energy with a ¼ 1.5.
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7
Yield Stress Ratio
4.4. Boundary effect
Fig. 16. Residual velocities of different cases for the sandwiched sphere structure and
the monolithic plate. The size of the sandwiched sphere structure is
12.25 mm  12.25 mm for all simulations conducted in the last
section. To confirm that this region is large enough that the
boundary effect is minimal, A new simulation is conducted with the
boundary size of 36.75 mm  36.75 mm. Fig. 17 shows the pene-
tration process at t ¼ 80 ms for the new simulation (Location 1,
a ¼ 1.0). Fig. 18 shows the variation of the projectile’s velocity with
time. The residual velocity from the new simulation is 225 m/s,
which is almost the same as the result, 229 m/s, in the previous
section, although the analysis region is nine times larger. Therefore,
the original boundary size of 12.25 mm  12.25 mm is sufficient.
This test confirms that the boundary conditions are of minor
importance in ballistic penetration by small mass projectiles in the
ordnance velocity regime (500e1300 m/s) as long as the target
diameter is more than a few projectile diameters.

5. Discussions and conclusions

In this research, a three-dimensional unit cell e square based

pyramid unit cell (SBPUC), is created as the basic building block of a
new sandwiched sphere structure for ballistic protection. The
perforation of the new sphere structure with sandwich core built
with SBPUCs by a flat-nose projectile is numerically simulated.
Fig. 17. Side section view of penetration at t ¼ 80 ms for the case of location 1, a ¼ 1.0, Three representative impact locations are considered and the effect
with enlarged boundary size. of the yield stress of the target material is examined. The
 Y. Fu et al. / International Journal of Impact Engineering 58 (2013) 66e75 75

performance of the new structure is compared with the monolithic or by adding more layers of spheres. The diameter ratio of the
plate based on the same area density by numerical simulation. The spheres in the first layer of the sphere structure to the projectile
energy absorption ability of the new structure during high speed should also be an important factor that affects the performance of
impact is related to many factors. Besides those previous findings the sphere structure. Because the energy dispersion toward the
from a single sphere impact test in the literature, such as perimeter and to other spheres is directly related to the direction of
sphere thickness and diameter, impact mass and impact velocity, the reaction force received by the projectile when impacting the
the formation of stationery plastic hinge of hollow spheres, etc., curved sphere surface and the direction of the reaction force is
the significant findings in the current design are summarized as further related to this ratio. These will be investigated in the future
follows study.

1. The tumbling effect of the new structure is significant. The

projectile tumbles significantly during and after the pene- References
trating process at most entry locations and thus increases the
[1] Andrews MD. Multi-layer composite armor and method. US patent 8096223
projectile’s effective contact area with the target or with the B1; Jan. 17, 2012.
object behind the target. [2] Sokolow A, Pfannes JM, Doney RL, Nakagawa M, Agui JH. Absorption of short
2. There exists an optimum yield stress ratio (around a ¼ 1.5) for duration pulses by small, scalable, tapered granular chains. Appl Phys Lett
2005;87:1e3.
the sphere structure to have maximum ballistic protection. [3] Doney RL, Agui JH, Sen S. Energy partitioning and impulse dispersion in the
3. With a ¼ 1.5, the effectiveness of the new structure is almost decorated, tapered, strongly nonlinear granular alignment: a system with
uniform regarding different impact locations. Even at the worst many potential applications. J Appl Phys 2009;106:1e13.
[4] Zeng HB, Pattofatto S, Zhao H, Girard Y, Fascio V. Impact behaviour of hollow
entry location (location 1), the new structure still performed sphere agglomerates with density gradient. Int J Mech Sci 2010;52:680e8.
much better than the monolithic plate. The new structure can [5] Zeng HB, Pattofatto S, Zhao H, Girard Y, Fascio V. Perforation of sandwich
absorb at least 11.6% more impact kinetic energy than the plates with graded hollow sphere cores under impact loading. Int J Impact Eng
2010;37:1083e91.
monolithic plate at a ¼ 1.5. [6] Jovicic J, Zavaliangos A, Ko F. Modeling of the ballistic behavior of gradient
4. The energy associated with the incoming projectile is partially design composite armors. Compos Part A Appl Sci Manuf 2000;31:773e84.
dispersed toward the perimeter and to some spheres and [7] Dadrasi A. Energy absorption of semi-spherical shells under axial loading. Aust
J Basic Appl Sci 2011;5:2052e8.
sphere caps in different layers; this can be seen by observing
[8] Guo J, Shi G, Wang Y, Lu C. Efficient modeling of panel-like targets in perfo-
the boundary deforming as the projectile proceeding and by ration simulation. Comput Struct 2003;81:1e8.
observing the motion of deforming, translating, or rotating of [9] Dean J, Dunleavy CS, Brown PM, Clyne TW. Energy absorption during pro-
some spheres and sphere caps. jectile perforation of thin steel plates and the kinetic energy of ejected frag-
ments. Int J Impact Eng 2009;xxx:1e9.
5. The density of hollow spheres and sphere caps varies gradually [10] Johnson GR, Cook WH. A constitutive model and data for metals subjected to
within the structure. Placing the hardest layer of spheres or large strains, high strain rates and high temperatures. In: Proceedings of
sphere cap as the first and the last impacted layers and the seventh international symposium on ballistics, The Hague, The Netherlands;
1983. p. 541e7.
weakest layer of spheres or sphere cap as the middle layers [11] Johnson GR, Cook WH. Fracture characteristics of three metals subjected to
may have some benefits in terms of maximum energy various strains, strain rates, temperatures and pressures. Eng Fract Mech
absorption. 1985;21:31e48.
[12] Borvik T, Hopperstad OS, Dey S, Pizzinato EV, Langseth M, Albertini C. Strength
6. The sphere diameters increase gradually in different layers, and ductility of Weldox 460 E steel at high strain rates, elevated temperatures
such that the impacted smaller spheres in the anterior layer can and various stress triaxialities. Eng Fract Mech 2010;72:1071e87.
be obstructed gradually and completely from larger spheres in [13] Borvik T, Hopperstad OS, Berstad T, Langseth M. A computational model of
viscoplasticity and ductile damage for impact and penetration. Eur J Mech A
the posterior layers.
Solids 2001;20:685e712.
7. The sphere diameter in the last layer of spheres should be set at [14] Borvik T, Hopperstad OS, Langseth M, Malo KA. Effect of target thickness in
a value a little larger than the length of the projectile to provide blunt projectile penetration of weldox 460 E steel plates. Int J Impact Eng
2003;28:413e64.
enough room to allow the projectile tumbling after penetrating
[15] Borvik T, Langseth M, Hopperstad OS, Malo KA. Ballistic penetration of steel
through the first and second layers of spheres. plates. Int J Impact Eng 1999;22:855e86.
[16] Borvik T, Langseth M, Hopperstad OS, Malo KA. Perforation of 12 mm thick
According to the one-dimensional model in the literature, the steel plates by 20 mm diameter projectiles with flat, hemispherical and
conical noses, part I: experimental study. Int J Impact Eng 2002;27:19e35.
performance of this new sphere structure can be further increased [17] Relationship between hardness and elastic modulus. M.L. Oyen. http://
by introducing a series of smaller beads between existing spheres imechanica.org/node/2285; September 21, 2012.