International Journal of Impact Engineering 75 (2015) 140e149

Contents lists available at ScienceDirect

International Journal of Impact Engineering

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

Impact mechanics of topologically interlocked material assemblies

Yuezhong Feng a, Thomas Siegmund a, *, Ed Habtour b, Jaret Riddick b
a
b

School of Mechanical Engineering, Purdue University, West Lafayette, IN 47907, USA

Vehicle Technology Directorate, US Army Research Laboratory, Aberdeen Proving Ground, MD 21005, USA

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 3 January 2014
Received in revised form
24 June 2014
Accepted 6 August 2014
Available online 23 August 2014

The objective of the present study is to provide understanding of potential energy absorption and
dissipation mechanisms and benets of topologically interlocked material (TIM) assemblies under
impact loading. The study is motivated by earlier ndings that TIMs under low rate loading demonstrated
attractive properties including the capability to arrest and localize cracks and to exhibit a quasi-ductile
response even when the unit elements are made of brittle materials. It is hypothesized that TIMs due to
their modularity would possess advantageous impact characteristics. In order to test this hypothesis, a
series of computational experiments on the dynamic loading of TIMs are conducted. Results obtained in
this study are presented for a planar TIM conguration based on a dense packing of tetrahedral unit
elements to form an energy absorption layer. Finite element models are calibrated on samples fabricated
using fused deposition modeling (FDM) additive manufacturing (AM). With employing the Lambert
eJonas formula to interpret the numerical data, it is demonstrated that TIMs can absorb more impact
energy than conventional solid plates. An extended LamberteJonas model is dened such that accurate
description of the impact response of TIMs is obtained.
2014 Elsevier Ltd. All rights reserved.

Keywords:
Topologically interlocking material
assemblies
Impact loading
Ballistic limit
Residual velocity
Modeling and simulation

1. Introduction
In the design of multifunctional structural components, it is
essential to go beyond the selection of materials with the appropriate properties. Developing logical and hybrid structures can be
effective in solving some of the critical and challenging engineering
problems that may require multifunctionality without developing
new and exotic material composition. Hybrid material systems are
understood to provide efcient alternatives to conventional materials and ll holes in material property spaces [1e3]. One of the
methods relevant for the creation of hybrid material systems is that
of segmentation. Thereby, a hybrid material system is created from
the bottom-up by the ordered assembly of unit elements. Hybrids
created by this approach allow for the exploration of important
material mechanics concepts such as topological toughening and of
size effects. Past studies on hybrid materials created by assembly
have predominately considered quasi-static loading conditions. The
aim of the present study is to explore the response of such material
systems to high rate loading and to impact. In particular, the study
is concerned with one class of segmented hybrids, Topologically
Interlocked Material assemblies, TIMs. In a TIM system, each

* Corresponding author.
E-mail address: siegmund@purdue.edu (T. Siegmund).
http://dx.doi.org/10.1016/j.ijimpeng.2014.08.003
0734-743X/ 2014 Elsevier Ltd. All rights reserved.

Downloaded from http://www.elearnica.ir

individual unit element is topologically conned by its neighbors

due to the oblique angles of the basic unit elements; therefore no
binders are necessary to maintain the integrity of the assembly
[4e7]. We hypothesize that the topological interlocking of the unit
(or fragment) elements would provide for an enhancement in the
impact resistance compared to monolithic solids.
The concept of topological interlocking assemblies has attracted
some interest in the recent past. Proposed by Glickman [8] with the
name of G-Block for the reduction of sub-structure and diminished
maintenance cost for paving system, expanded, generalized and
termed with the current name by Dyskin et al. [4,5,9,10], TIMs can
be considered as novel mechanical meta-materials as these derive
their unique properties not from composition but structure. Forces
and energy can only be transmitted through the contact surfaces;
accordingly, there are no tensile forces developed when TIMs carry
loads. These unique characteristics of TIMs provide unique properties that can be utilized to create adaptive and congurable
structures to harsh conditions such as random and harmonic vibrations, thermal loads, repetitive shocks and acoustic attenuation.
For example, Dyskin et al. [4,5,9,10] introduced the general concept
of TIM, and demonstrated the requirements for basic elements to
assemble a TIM tile. Estrin et al. [11] assembled TIMs with identical
cubes and tested their responses with point load tests. Schaare et al.
[12] investigated the response of cube-based TIMs to point loading
both experimentally and numerically. Brugger et al. [13]conducted

Y. Feng et al. / International Journal of Impact Engineering 75 (2015) 140e149

Symbols and abbreviations

a
a1,a2
D
E
K
K*
L
m0
M
N
p
p1,p2
s0
t

parameter in the LamberteJonas model

parameters in the extended LamberteJonas model
projectile diameter
Young's modulus of tetrahedron material
contact stiffness
normalized contact stiffness
monolayer span of TIM
mass of a single tetrahedron
projectile mass
number of tetrahedra in one row/column of TIM
parameters in the LamberteJonas models
parameters in the extended LamberteJonas model
edge length of regular tetrahedron
thickness of TIM monolayer

indention experiments on TIMs assembled by osteomorphic ice

blocks and plaster made cubes. Their results showed that TIM
composed of cube elements can experience negative stiffness when
unloading. Dyskin et al. [14] revealed that TIMs are damage tolerant
in that cracks inside of one unit element will not propagate across
the contact interface into adjacent units. Khandelwal et al. [15]
experimentally examined the quasi-static responses of TIMs
assembled with regular tetrahedra, and proposed an analytic model
to predict the quasi-static behaviors of TIMs via thrust line theory.
Mather et al. [16] studied the remanufacturability of TIMs, illustrating that only a small portion of units could not be reused after
failure of the TIMs. And for the applications of TIMs, Estrin et al. [17]
proposed the potential application of TIM as protective tiles for the
space shuttle. Carlesso et al. [18] illustrated TIM composed by
porous osteomorphic blocks enhances sound absorption because of
the gaps among basic blocks. Tessmann [19] explored TIMs with
more complex geometries for potential architecture applications.
The present investigation is aimed at understanding the
response mechanism of TIMs due to impact loading, and the potential benets of TIMs in energy absorption, as well as mechanical
and structural applications. TIMs consisting of dense packing of
tetrahedral unit elements made of Acrylonitrile-Butadiene-Styrene
(ABS) material, which were fabricated using fused deposition
modeling (FDM) additive manufacturing (AM) motivate the material system under investigation [15]. Limited exploratory research
has been done in understanding the mechanical characteristics of
these novel structures, more specically TIMs subject to low velocity impact. Plates subjected to impact are of great research interest for protection, and extensive work has been reported on
various types of systems subjected to ballistic impact [20e28]. Most
recently, Wadley et al. [29] studied sandwich structures with
different shapes of core material under ballistic impact. Their results demonstrated that sandwich structure performance could be
improved with topological optimization of core materials. In the
present study, the impact velocities considered are much lower
than the dilatational wave speed of the material in the unit elements. The nite element method was employed to conduct numerical experiments. First, the model is calibrated on experimental
data of low and constant velocity loading. Then, the response under
constant applied velocity is considered, and the contributions of
inertia, elastic resistance, and contact and friction to the mechanical resistance are explored. Subsequently, impact loading is
considered and the formulation of Lambert and Jonas [30] is
employed to characterize the velocity response. The model is
reformulated to provide specic insight into the impact response of

141

V0
constant velocity
Vbl
ballistic velocity
Vimp
impact velocity
Vres
residual velocity
Vtrs,1,Vtrs,2transition velocities in the extended LamberteJonas
model
VV
virtual ballistic velocity in the extended Lambert
eJonas model
d0,df
damage initiation separation and failure separation in
CZM
m
coefcient of friction
n
Poisson's ratio of tetrahedron material
r0 ; b
r
densities of tetrahedron and projectile, respectively
smax
cohesive strength
CZ, CZM cohesive zone and cohesive zone model
TIM
topologically interlocked material

the TIM assemblies. A comparison to monolithic structures under

impact is made.
2. Model denition and model calibration
2.1. Model denition
Fig. 1 shows the model system under consideration with (a) a
regular tetrahedron as the basic unit of the TIM assembly, (b) a
boundary element (termed abutment), (c) the indenter/projectile
and the center ve tetrahedra onto which the indenter transfers the
load, (d) the TIM tile composed with identical regular tetrahedra,
and (e) the overall TIM assembly together with the cylinder shaped
indenter/projectile. In the model assembly, a total of N
N 49
identical tetrahedra are assembled in a square lattice pattern into a
monolayer TIM assembly [4,8,15], and conned with abutments
along the edges of the assembly. The edge length of an individual
regular tetrahedron is s0. The other dimensions of the TIM assembly
p
can be derived as the thickness of the monolayer t s0 = 2, the
dimension of the mid-plane of the TIM assembly, L s0(N/2). Here,
the TIM assembly is characterized by N 7 and s0 25 mm, such
that t 17.7 mm and L 87.5 mm.

Fig. 1. Model of Topologically Interlocked Material (TIM) assembly: (a) Unit element
regular tetrahedron, (b) Abutment (c) Indenter/projectile and center tetrahedra, (d)
TIM tile, (e) Overall assembly.

Y. Feng et al. / International Journal of Impact Engineering 75 (2015) 140e149

Two loading conditions were considered: (1) constant velocity

loading by an indenter, and (2) impact loading by a projectile. In
both cases the load was applied to the center of the TIM assembly
such that the indenter/projectile interacts directly with tetrahedra
#1, #2, and #3 of the assembly. The diameter D of the indenter/
projectile is D L/5.
Tetrahedra were considered to be isotropic and elastic. Motivated by prior experiments on TIM assemblies in Ref. [15], the
material properties of ABS were employed (Young's modulus
E 1.827 GPa, Poisson's ratio n 0.35, density r0 950 kg/m3). One
tetrahedron thus possesses a mass of m0 1.75
103 kg and the
overall assembly weighs 8.58
102 kg. The abutments as well as
the indenter/projectile were considered as rigid bodies. The density
of the projectile was b
r 1470 kg/m3. The contact properties, i.e.,
the contact pressureeoverclosure relation and the friction condition were not directly available. The overclosure is the penetration
distance of the nodes on contact surfaces in the FEA models. A
Coulomb friction model was employed which introduces the coefcient of friction m. For the normal contact behavior, it was
assumed that contact pressure pc is linearly dependent on the
overclosure Dn as pc KcDn, where Kc K*E*/(ps0) is the contact
stiffness with K* a normalized contact stiffness, and E* E/
[2(1  n2)]. This type of interaction was adopted to describe the
contact behaviors for the inter-tetrahedra, tetrahedron-abutment
and tetrahedra-indenter contacts. Experimental data on the low
velocity impact response of TIM assemblies presented in Ref. [15]
was used to calibrate the model parameters m and K*.
Fragmentation commonly occurs for brittle materials subject to
impact loading [31e33]. To compare the impact resistance of TIMs
made by brittle materials with and without fragmenting, a cohesive
zone model (CZM) was further developed. Motivated by the TIM
material used in experiments [15], in which the tetrahedra were
made with additive manufacturing, the zero thickness cohesive
zone (CZ) layers were created parallel to the mid-plane of TIM assembly, i.e., perpendicular to the impact direction, which was
meshed with cohesive element COH3D8. In each tetrahedron, eight
equally spaced CZ layers were inserted. Therefore, each tetrahedron
contains 10 solid layers and 8 CZ layers. A bilinear tractionseparation law was used to describe the constitutive response of
CZM, with maximum strength smax E/100, failure separation
df s0/1000 and damage initiation separation d0 df/10. Since
ABAQUS/Explicit [34] method was used to solve the dynamic
problem, the density of the CZ layer needs to be dened. In the
simulations for TIMs with CZ layers, the density of the CZ layer is
rcoh 1.0 kg/m3, which does not increase the inertia of the structure signicantly, as well as the computational difculty [35]. Aside
from the CZ properties, all the other properties and conditions are
the same as those used in the impacting simulations of TIMs with
isotropic tetrahedra.
In the constant velocity cases, a predetermined displacement is
applied on the indenter and the time interval of loading is varied in
order to achieve a constant velocity V0. In the impact cases, a projectile of a dened mass M with initial velocities Vimp strikes the
TIM assembly. The mass of the projectile is chosen as 10% of the
total mass of the 7
7 tetrahedra. The abutments are xed in space
in all cases.
Finite element models of the TIM assembly were constructed in
the general purpose nite element code ABAQUS (Version 6.12). The
nite element mesh employs, 3D linear solid element (C3D8R in the
ABAQUS code) with each tetrahedron is comprised of 500 elements
such that the overall model consisted of 24,500 elements. The
model was solved using the ABAQUS/Explicit solver, and the
general contact algorithm implemented in the ABAQUS software
was employed. In all the simulations, one analysis step was dened.
The automatic time increment algorithm was employed and the

global stable increment estimator was selected. The analysis step

time for constant velocity loading varies dependent on loading
velocities while for impact loading, an analysis interval of 5 ms is
considered in all the cases.
2.2. Parameter calibration
In the experiments presented in Ref. [15], a drop mass of 6.21 kg
and drop height of 36.58 mm were employed. Since the drop mass
signicantly exceeds that of the TIM assembly, a near constant
velocity loading condition (V0 0.925 m/s) was achieved in the
experiments. The data of the forceedeection record were presented in Ref. [15], and this data was used to calibrate the parameters m and K*.
The computational study provided the forceedeection curves
of TIM for contact stiffness (K* 0.5) and different coefcients of
friction. Fig. 2 shows predicted forceedeection curves from
computations considering a range of values of m (0.1 < m < 1.0). In
these calculations the contact stiffness was K* 0.5. In all cases a
forceedeection response typical of TIM assemblies was predicted
[15]. The force initially increases with the deection; a peak force
value is reached, followed by gradual softening until the resistance
of the TIM assembly vanishes. The resistance of the TIM assembly to
the applied load was found to increase with an increase in m. The
predicted peak force increased from 48.5 N to 513.0 N when m was
changed from m 0.1 to 1.0. The results also indicate an increase of
the deection to nal collapse of the TIM with an increase in m. For
the lowest value of m considered (m 0.1), the load carrying capacity of the TIM vanishes at a deection of 22.0 mm but this
limiting deection more than doubles for cases considering higher
values of m. From the summary of these simulation results it is
evident that connement and interlocking alone is not the only
relevant mechanism in TIMs, at least for systems with lower elastic
modulus in the solid as in the present. In addition to connement,
inter-element slip can indeed play a substantial role in the model
response.
The next set of computations considered a range of contact
stiffness values and constant coefcient of friction. Fig. 3 shows
reaction the predicted forceedeection records for computations
considering variations in contact stiffness (0.1 < K* < 1.0). In these
calculations m 0.3 was considered. It is found that the magnitude
of the predicted forces increases with increased contact stiffness.

600

= 0.10
= 0.20
= 0.30
= 0.40
= 0.50
= 0.60
= 0.80
= 1.00

500
400

Force [N]

142

300
200
100
0
0

10

20

Deflection [mm]

30

40

Fig. 2. Predicted forceedeection curves for a range of values of the coefcient of

friction m.

Y. Feng et al. / International Journal of Impact Engineering 75 (2015) 140e149

450

K = 0.10
K* = 0.20
K* = 0.30
*
K = 0.40
K* = 0.50
*
K = 0.60
*
K = 0.70
*
K = 0.80
K* = 0.90
*
K = 1.00

400
350
300

Force [N]

143

250
200
150
100
50
0
0

10

15

20

Deflection [mm]

25

30

35

Fig. 3. Predicted forceedeection curves for a range normalized contact stiffness

values.K*

The peak force was found to increase from approximately 52.0 N to

513.0 N as K* was increased from K* 0.1 to 1.0. However, the
deection to loss of load carrying capacity did not vary signicantly,
and for all the cases was about 35.0 mm.
The results of the parametric computations indicate that the
coefcient of friction affects both the magnitude of the force and
the deection to nal failure, while the contact stiffness plays a role
only in the force magnitude. Therefore, the following calibration
procedure was followed. First, the coefcient of friction was calibrated with respect to the experimental results of the deection to
nal failure. Then, the contact stiffness is selected to provide a nal
forceedeection response which ts that of the experiments presented in Ref. [15]. The calibrated parameters are m 0.2 and
K* 0.38. Model predictions of the forceedeection response obtained using the best-t parameters for m and K* together with the
experimental data are shown in Fig. 4, and a good quantitative
agreement is found. The residual deformations of the TIM assemblies after indention are shown for the experiment in Fig. 5(a) [15],
and the model simulation in Fig. 5(b). A good qualitative agreement
is obtained.

140

3. Results

FEA prediction
Experiment observation

120

Force [N]

100
80
60
40
20
0
0

Fig. 5. Deformed congurations of TIM assemblies after loading: (a) Experimental

observation [15], (b) Model prediction with color contours depicting the residual stress
state as characterized by the von Mises equivalent stress. (For interpretation of the
references to colour in this gure legend, the reader is referred to the web version of
this article.)

10

15

20

Deflection [mm]

25

30

Fig. 4. Force deection curves: experimental data [15] and best t prediction.

3.1. Constant velocity

Simulations of TIM subjected to constant velocity loading were
conducted with applied velocities ranging from 0.1 m/s to 10 m/s.
Fig. 6 depicts the predicted forceedeection responses. For the
lowest velocity case, V0 0.1 m/s, the force increases smoothly with
the increase of deection, reaches its peak at a deection of
11.0 mm then decreases gradually to zero. As the applied velocity
increases, multiple peaks develop in the forceedeection curves.
When the loading velocity is greater than 1 m/s but less than 5 m/s,
the value at the rst peak increases dramatically with increasing
velocity, while the second peak value remains constant. When the
loading velocity is greater than 5 m/s, both peaks increase with
increasing velocity.
Fig. 7 shows the forceedeection relations for V0 0.1 m/s with
connement, and V0 4 m/s with and without connement,
respectively. A simulation for unconned TIM means the abutments, Fig. 1(d), are taken away in the computation. In this case, the

144

Y. Feng et al. / International Journal of Impact Engineering 75 (2015) 140e149

450
400

Center tetrahedron(V0 = 4 m/s)

Center tetrahedron(V = 10 m/s)

V0 = 1.0 m/s

350

V = 2.0 m/s
V0 = 4.0 m/s

250

V0 = 5.0 m/s

200

V0 = 6.0 m/s
V0 = 8.0 m/s

150

V = 10.0 m/s
0

100

Indenter(V0 = 4 m/s)

15

300

Velocity [m/s]

Force [N]

20

V0 = 0.1 m/s

Indenter(V = 10 m/s)
0

10

50
0
0

10

20

Deflection [mm]

30

40

Fig. 6. Predicted forceedeection curves for various constant velocities.

V0 = 0.1 m/s (Quasistatic)

120

V = 4.0 m/s (Unconfined)

0

V = 4.0 m/s (Confined)

0

Force [N]

100
80
60
40
20
0
0

10

20

Deflection [mm]

30

10

20

30

40

Indenter displacement [mm]

50

Fig. 8. Predicted translational velocities (in load direction) of the center tetrahedron
for loading velocities of V0 4 m/s and 10 m/s.

force increases quickly to its peak then drops precipitously. There is

no conning force acting on the TIM tile, therefore, the force shown
in the gure is due to the inertia of the TIM tile. Since there no such
phenomenon occurred for the V0 0.1 m/s case, the V0 0.1 m/s
case can be considered as quasistatic loading case. Therefore, the
forceedeection curve for the V0 4 m/s with connement can be
considered as the sum of the other two cases: quasistatic response
and unconned model with same impact velocity, Fig. 7.
Fig. 8 depicts the velocities of center tetrahedra during the
indentation for relatively low and high loading velocities, V0 4
and 10 m/s, respectively. The velocity of the center tetrahedron is
dened on the center of the center tetrahedron, and the center
tetrahedron is the tetrahedron that the indenter directly strikes on,
T1 as shown in Fig. 1(c). The velocities increase with time at the
beginning of indention, then decrease and increase alternatively,
and nally they keep constant. Overshoots occur during indention.
For V0 10 m/s, after its peak value, the velocity of center tetrahedron drops far below the loading velocity because of the interaction with adjacent tetrahedra through contact. The indenter
accelerates it again, which results in a second peak force in the
forceedeection curves. For V0 4 m/s, one also observes a time

140

0
0

40

Fig. 7. Predicted forceedeection curves for quasistatic loading with conned model,
V0 4 m/s conned and unconned models.

interval in which the velocity of the center tetrahedron is lower

than that of indentation velocity. However, the velocity difference is
not signicant so the re-acceleration is not observed in the forceedeection curve.
3.2. Impact
A series of computations with impact velocities ranging from
1 m/s to 80 m/s was performed, all employing the calibrated model
parameters. The ballistic limit for the TIM assembly is determined
as Vbl 34 m/s, such that at the end of this impact event the velocity of the projectile is zero. Below Vbl the projectile rebounds and
above Vbl the TIM monolayer is penetrated. The central variable to
characterize the resistances of plates and plate-like systems to
impact, is the residual velocity of the projectile after passing
through the systems. When the impact velocity is greater than the
ballistic limit velocity, the projectile penetrates the plates and
keeps moving forward with a non-zero kinetic energy.
Fig. 9 depicts the velocity histories of the projectile and the
center ve tetrahedra, Fig. 1(c), for four values of the impact velocity Vimp. In all cases Vimp  Vbl. At the ballistic limit, Fig. 9(a), both
velocities of the tetrahedra and the projectile reach zero. For impact
velocity slightly higher than ballistic limit, Vimp 45 m/s, the velocities of the center tetrahedron and the projectile velocity are
only slightly larger than those of the surrounding tetrahedra,
Fig. 9(b). This indicates that the projectile strongly interacts not
only with the center tetrahedron, but due to the topological interlocking, also with the surrounding elements. For relatively high
impact velocity cases, Fig. 9(c) and (d), this situation changes. In
both cases shown, i.e., Vimp 55 and 80 m/s, the center tetrahedron
gains much higher velocity than its neighbors. Still, for Vimp 55 m/
s the projectile velocity is similar to those of the surrounding unit
elements, indicating that the TIM assembly continues to provide
some level of resistance. Only at much higher impact velocity,
Vimp 80 m/s, is such an interaction lost and the projectile velocity
increases beyond that of the surrounding tetrahedra. The predicted
in-plane conning forces, Fig. 10, conrm these different levels of
interaction between projectile and the TIM monolayer. Fig. 10(a)
depicts the conning forces in x-direction while Fig. 10(b) shows
those in the y-direction. The forces are normalized with respect to
their peak values.

Y. Feng et al. / International Journal of Impact Engineering 75 (2015) 140e149

0.8
0.6
0.4
0.2
0
0
1.2

Normalized velocity

Projectile
T1
T2
T3
T4
T5

Normalized velocity

1.0

1.2

(a)

1.0

1.0

2.0

3.0

4.0

Time [ms]

Projectile
T1
T2
T3
T4
T5

0.6
0.4
0.2
1.0

2.0

3.0

4.0

Time [ms]

0.6
0.4
0.2
1.0

2.0

3.0

4.0

Time [ms]

5.0

(d)

1.0

Projectile
T1
T2
T3
T4
T5

0.8
0.6
0.4
0.2
0
0

5.0

Projectile
T1
T2
T3
T4
T5

0.8

1.2

(c)

0.8

0
0

(b)

1.0

0
0

5.0

Normalized velocity

Normalized velocity

1.2

145

1.0

2.0

3.0

4.0

Time [ms]

5.0

Fig. 9. Predicted projectile velocity during impact compared to the velocities of the ve central tetrahedra units in the TIM assembly: (a) Vimp 34 m/s, (b) Vimp 45 m/s, (c)
Vimp 55 m/s, (d) Vimp80 m/s.

Vres

8
<0
1=p


: a Vp  Vp
imp
bl

Vimp  Vbl ;

(1)

Vimp > Vbl ;

Normalized reaction force on frame

in which a an p are constants to be tted with the residual velocity

data. Fig. 11 depicts the t to the residual velocity data using the
LamberteJonas models. A nonlinear least square method was
employed to obtain the parameters, which was conducted with
MATLAB built-in function lsqnonlin. The tted parameters are
a 0.343, p 1.487 and the coefcient of determination (R2) is
0.9570.
The tted LamberteJonas model can provide an overall good
description of the numerical results. Still, but for higher impact
velocities, the model ts are less accurate than for velocities close to
1.0
0.5

(a)

V0 = 34 m/s
V = 45 m/s
0

V0 = 55 m/s

0.0

V = 80 m/s
0

1.0
0.5
0.0
0.0

1.0

2.0
3.0
Time [ms]

4.0

5.0

the ballistic limit. A re-examination of the data, Fig. 12 leads to the

conclusion that the residual velocities solution can be separated
into at least two regimes (below 48 m/s and above 56 m/s). A
transient regime exists at intermediate velocities. This transition
regime is of particular interest as for that domain the residual velocity becomes rather independent of the impact velocity.
A better description of the residual velocity data can be provided
by an extended LamberteJonas model of form:

Vres

8
0
>
>

1=p1
>
>
>
< a1 V p1  V p1
imp
bl

const:
>
>
>

1=p2
>
>
: a V p2  V p2
2
V
imp

Vimp  Vbl ;
Vbl  Vimp  Vtrs;1 ;
Vtrs;1 < Vimp  Vtrs;2 ;

(2)

Vimp > Vtrs;2 ;

with Vtrs,1 and Vtrs,2 the transition velocities separating the two
limiting regimes and dening the transient impact velocity regime,
Normalized reaction force on frame

For conventional metal plates or plate-like systems, Lambert

and Jonas [30] proposed a formula to predict the residual velocity:

1.0
0.5

(b)

V0 = 34 m/s
V = 45 m/s
0

V0 = 55 m/s

0.0

V = 80 m/s
0

1.0
0.5
0.0
0.0

1.0

2.0
3.0
Time [ms]

4.0

5.0

Fig. 10. Predicted in-plane conning forces on frame normalized with respect to their peak values: (a) Conning force in x-direction, (b) Conning force in y-direction.

146

Y. Feng et al. / International Journal of Impact Engineering 75 (2015) 140e149

Residual velocity [m/s]

30
25

FEA results
LambertJonas model: a = 0.343, p = 1.487
Asymptotic limit

20
15
10
5
0
0

20

40

Impact velocity [m/s]

60

80

Fig. 11. Predicted residual velocities in dependence of the impact velocity and data t
by the LamberteJonas model, m 0.2.

VV a virtual ballistic velocity tted on the high impact velocity

regime, and a1, a2, p1 and p2 the characterizing parameters. Fig. 12
depicts the residual velocity data tted to the extended model with
t parameters Vbl 34 m/s, a1 0.514, a2 0.407, p1 1.278,
p2 1.672, VV 48 m/s, Vtrs,1 48 m/s, and Vtrs,2 57 m/s,
respectively. The coefcient of determination for low impact velocity regime is 0.9852 while for high impact velocity regime is
0.9871, which indicates a better t to the numerical data than
provided by the basic LamberteJonas model.
For quasistatic loading, Fig. 2 indicated that friction can have a
relevant inuence on the material system response. For impact, the
response characteristics were thus computed also for values of
m 0.15 and m 0.25, and analyzed with the extended LamberteJonas model, Figs. 13 and 14. For m 0.15 the parameter for
the extended LamberteJonas model were determined as
Vbl 28 m/s, a1 0.397, a2 0.375, p1 1.394, p2 1.646,
VV 45 m/s, Vtrs,1 45 m/s, and Vtrs,2 55 m/s, while for m 0.25
these parameter are Vbl 39 m/s, a1 0.434, a2 0.370, p1 1.419,
p2 2.016, VV 52 m/s, Vtrs,1 52 m/s, and Vtrs,2 65 m/s. For all
cases it is observed that Vtrs,1 VV.

Residual velocity [m/s]

25
20

Fig. 13. Predicted residual velocities and the extended LamberteJonas model, m 0.15.

Considering the three sets of simulation with m 0.15,0.20,0.25

one determines that the ballistic velocity The multiple individual
response domains in the residual velocity can be traced back to the
velocity history of projectile and tetrahedra, Fig. 9, and is identied
with the differences in interaction of the projective with the unit
elements in the TIM assembly, Fig. 9, and the development of the
in-plane connement forces during the impact event, Fig. 10. Two
limiting conditions can be dened: (i) low velocity regime with
strong persisting projectile-monolayer interaction and a substantial in-plane connement forces throughout the impact event, and
(ii) a high velocity regime with weak projectile-monolayer interaction and a rapid loss of connement forces during the impact
event. As the value of m is increased, the simulations predict that the
ballistic velocity is increased, yet the residual velocity at high
impact velocities changes little, Vres z23 m/s for Vimp 80 m/s
independent of m. This observation leads to the argument that the
failure in high velocity failure mode is little dependent of friction
and slip, but that failure the low velocity regime depends on m. Such
an argument would be supported by the fact that in the high velocity regime primarily the interaction between across few contact
surfaces are relevant while in the low velocity regime many of the
interactions surfaces are activated.

FEA results
a1 = 0.514, p1 = 1.278, Vbl = 34 m/s,
(V V
48 m/s)
bl
imp
a2 = 0.407, p2 = 1.672, VV = 48 m/s,
(57 m/s Vimp 80 m/s)

15
10
5
0

30

40

50

60

Impact velocity [m/s]

70

80

Fig. 12. Predicted residual velocities and the extended LamberteJonas model, m 0.20.

Fig. 14. Predicted residual velocities and the extended LamberteJonas model, m 0.25.

Y. Feng et al. / International Journal of Impact Engineering 75 (2015) 140e149

Simulations for four impact velocity cases (Vimp 34, 45, 55 and
80 m/s) were performed with the model enhanced by cohesive
zone elements. Table 1 lists the terminal velocities for isotropic TIM
and CZ enhanced TIM models. Fig. 15 compares the projectile velocities with Vimp 55 m/s, and similar comparison can be obtained
for other impact velocity cases. Fig. 16 shows the fragmentation of
CZ enhanced TIMs for different impact velocities. The results show
that the CZ enhanced model resists impact better with lower projectile residual velocity, but the difference is not signicant when
impact velocity is high. Energy distributions show that the CZ
enhanced TIM has higher kinetic energy at the end of impact event,
which is resulting from the fragmentation of the TIM.
Finally, a comparison of the impact response of the TIM relative
to monolithic systems is being shown for the two limiting velocities
considered in the prior table. The monolithic system corresponds to
the TIM system in all aspects, except that the contact surfaces between all tetrahedra elements are fused. Fig. 17 depicts the temporal evolution of the energy dissipated in damage for a low and a
high impact velocity for the TIM and the corresponding monolithic
system, respectively. These simulations indicate that the monolithic system is more damage tolerant than the TIM at low impact
velocities. This, however, changes as the impact velocity is
increased. At the high impact velocity, the damage to the TIM is
found to be substantially lower than that in the monolithic system.
Considering Fig. 18 it can be found that for the monolithic system
damage is conned to the domain under impactor at low impact
velocities but becomes widespread at high impact velocities. In
TIMs on the other hand, damage remains a mostly local event,
conned to the impact site, see Fig. 16.

4. Discussion
The simulations of the TIM response under constant loading
velocity, Figs. 6e8, provide insight into the relevance of the material parameters in the TIM system. The responses can be divided
into three categories according to loading velocities: quasi-static
response at low velocities, combination of quasi-static and dynamic response, and dominant dynamic. For quasi-static case, a
quasi-ductile response can be observed, which agrees with the
experimental observation [15]. For moderate loading velocities, the
forceedeection response can be considered as the combination of
quasi-static response and unconned TIM under same loading velocity. The transition velocity is determined by the density of the
TIM material and the normal contact behavior, while coefcient of
friction does not affect the inertia responses signicantly.
The dynamic characteristics of TIMs are further investigated
with projectile impact loading conditions. The projectile residual
velocity is used as the parameter to illustrate TIMs' impact resistance, with LamberteJonas models being employed to describe the
numerical results. To have a more accurate description of the residual velocity results, an extended LamberteJonas model is
developed. For conventional plates, Brvik et al. [36] tted the
LamberteJonas formula with a 0.76 and p 2.36 based on
experimental data for ballistic penetration of steel plates. Chi et al.
[37] applied the model to a bi-layer metal system and obtained
ranges of the parameters for different thicknesses of metal plates:

147

Fig. 15. Projectile velocity for isotropic and CZ enhanced TIM models (Vimp 55 m/s).

0.93 < a < 0.97 and 1.72 < p < 2.14. Other experimental and numerical results [22,38e44] also indicate that for homogeneous
plates and composite plates, a is close to 1.0 and p is close to 2.0.
Comparing to the parameters in the LamberteJonas model for TIMs
with conventional plates, the parameter p is close to that for conventional plates while a is signicantly smaller. According to
LamberteJonas formalism, a smaller a indicates lower residual
velocity, therefore, more kinetic energy absorption. Friction was
found to play a role in determining the impact tolerance at low
impact velocities while such an effect is less at higher impact velocities. This nding is attributed to the competition between a
local deformation at high values of Vimp and a distributed deformation at low values of Vimp.
Residual velocity comparison was performed between TIMs
with isotropic tetrahedra or CZ enhanced tetrahedra. When the
impact velocity is close to ballistic velocity of TIM with isotropic
tetrahedra, the residual velocities for TIM with CZ layers are
smaller. The increasing of TIMs' stiffness is due to the fragmentation. At the end of impact events, energy distributions of the entire

Table 1
Terminal velocities of TIMs with and without CZ enhancement [m/s].

Isotropic TIM
CZ enhanced
TIM

Vimp 34 m/s

Vimp 45 m/s

Vimp 55 m/s

Vimp 80 m/s

0
2

8
2

10
8

23
18

Fig. 16. Fragmentation of CZ enhanced TIMs at time 5 ms: (a) Vimp 34 m/s, (b)
Vimp 45 m/s, (c) Vimp 55 m/s, (d) Vimp 80 m/s.

148

Y. Feng et al. / International Journal of Impact Engineering 75 (2015) 140e149

Fig. 17. Predicted evolution of energy dissipated in material separation during impact, comparison of TIM to monolithic system: (a) Vimp 34 m/s, (b) Vimp 80 m/s.

system show that, the kinetic energy of the tetrahedra and fragments of TIMs with CZ layers is higher than TIMs with isotropic
tetrahedra. These results can be conrmed with the result of Fig. 16.
For example, when the impact velocity is 34 m/s, every tetrahedron
is almost motionless for TIMs without CZ layers (Fig. 9(a)) at the
end of impact; while for TIMs with CZ layers (Fig. 16(a)), fragments
are created and move with non-zero velocity. Therefore, because of
fragmentation, the implementation of CZ layers into TIMs actually
increases their dynamic stiffness. The increment in dynamic stiffness is remarkable when impact velocities are close to ballistic
velocity. For high impact velocities (Vimp 55 and 80 m/s), the
difference of kinetic energies of tetrahedra/fragments for both
models is not signicant, resulting in similar residual velocities,
Table 1. Fig. 16 also illustrates that damage can be localized for
TIMs: only the tetrahedra directly impacted and their neighbors are
damaged, while the tetrahedra away from the impact center keep
their integrity. This processes is rather independent of the impact
velocity considered. In a monolithic system constructed for comparison, on the other hand, this is not the case. Damage is
concentrated in the vicinity of the indenter only at low impact
velocities while at high impact velocities damage is widespread. In
the TIM the presence of the contact surfaces clearly plays a role in
inhibiting widespread crack propagation.
5. Conclusions
The present work numerically investigated the dynamic responses of TIM under constant velocity and projectile impact
loading with nite element method. The FEA model is calibrated

with published experimental data. The resistances of TIM under

constant velocity and projectile impact were examined. Numerical
results indicate that: (i) TIM responses to constant velocity loading
can be categorized into three groups according to the loading velocities: quasi-static response, combination of dynamic and quasistatic response, and dominant dynamic response. At low velocities, the response of TIMs emerges as the sum of the inertia (unconned TIM) and the quasi-static response. For higher velocity
cases, the indenter imparts multiple inertia responses. (ii) The
terminal velocities of projectile for impact velocities from 1 m/s to
80 m/s are obtained. Investigation of the residual velocities of
projectile after penetrating the TIM monolayer shows projectile
velocities following an extended LamberteJonas model. The parameters in the extended LamberteJonas model are tted on obtained numerical results. The parameter p is close to conventional
solid metallic plates for high velocity regime but is smaller for the
low velocity regime, while the parameter a is much smaller for both
stages. For a given p in the LamberteJonas formalism, a smaller
value of a indicates lower residual velocity, hence higher kinetic
energy absorption. (iii) Comparisons between TIMs with and
without CZ layers show that TIMs with CZ layers can resist damage
better when the impact velocity is close to ballistic velocity. When
the impact velocity is higher, the difference of the two TIM models
in resisting impact is not signicant. TIMs emerge as more damage
tolerant than monolithic systems at high impact velocities due to
the localization of damage to the impact domain.
Acknowledgment
The authors Y. Feng and T. Siegmund acknowledge funding by
the Army Research Ofce grant Mechanics of Multiscale Energy
Dissipation (ARO Grant No. W911 NF-12-1-0370) in Topologically
Interlocked Materials-11.1 STIR, (Larry Russell Jr, Solid Mechanics,
Program Manager). T. Siegmund also acknowledges support from
the National Science Foundation (N.S.F.) while working at N.S.F.
The information presented in this paper is the sole opinion of
the authors and does not necessarily reect the views of any
sponsoring agency or NSF.
References

Fig. 18. Predicted spatial distribution of damage following impact (t 5.0 ms) in the
monolithic system: (a) Vimp 34 m/s, (b) Vimp 80 m/s.

[1] Ashby M, Brchet Y. Designing hybrid materials. Acta Mater 2003;51(19):

5801e21.
[2] Ashby M. Hybrids to ll holes in material property space. Philos Mag
2005;85(26e27):3235e57.
[3] Ashby M. Hybrid materials to expand the boundaries of material-property
space. J Am Ceram Soc 2011;94:s3e14.
[4] Dyskin AV, Estrin Y, Kanel-Belov AJ, Pasternak E. Topological interlocking of
platonic solids: a way to new materials and structures. Philos Mag Lett
2003;83(3):197e203.

Y. Feng et al. / International Journal of Impact Engineering 75 (2015) 140e149

[5] Dyskin A, Estrin Y, Kanel-Belov A, Pasternak E. Interlocking properties of
buckyballs. Phys Lett A 2003;319:373e8.
[6] Dyskin A, Estrin Y, Pasternak E, Khor H, Kanel-Belov A. The principle of topological interlocking in extraterrestrial construction. Acta Astronaut
2005;57(1):10e21.
[7] Estrin Y, Dyskin A, Pasternak E. Topological interlocking as a material design
concept. Mater Sci Eng C 2011;31(6):1189e94.
[8] Glickman M. The g-block system of vertically interlocking paving. In: Proceedings of 2nd International Conference on Concrete block paving. Delft, The
Netherlands: Delft University of Technology, American Society for Testing and
Materials; 1984. p. 345e8.
[9] Dyskin A, Estrin Y, Kanel-Belov A, Pasternak E. A new concept in design of
materials and structures: assemblies of interlocked tetrahedron-shaped elements. Scr Mater 2001;44(12):2689e94.
[10] Dyskin A, Estrin Y, Kanel-Belov A, Pasternak E. A new principle in design of
composite materials: reinforcement by interlocked elements. Compos Sci
Tech 2003;63(3e4):483e91.
[11] Estrin Y, Dyskin A, Pasternak E, Schaare S, Stanchits S, Kanel-Belov A. Negative
stiffness of a layer with topologically interlocked elements. Scr Mater
2004;50(2):291e4.
[12] Schaare S, Dyskin A, Estrin Y, Arndt S, Pasternak E, Kanel-Belov A. Point
loading of assemblies of interlocked cube-shaped elements. Int J Eng Sci
2008;46(12):1228e38.
[13] Brugger C, Brchet Y, Fivel M. Experiments and numerical simulations of
interlocked materials. Adv Mat Res 2008;47:125e8.
[14] Dyskin AV, Estrin Y, Kanel-Belov AJ, Pasternak E. Toughening by fragmentationdhow topology helps. Adv Eng Mater 2001;3(11):885e8.
[15] Khandelwal S, Siegmund T, Cipra R, Bolton J. Transverse loading of cellular
topologically interlocked materials. Int J Solids Struct 2012;49(18):2394e403.
[16] Mather A, Cipra R, Siegmund T. Structural integrity during remanufacture of a
topologically interlocked material. Int J Struct Integr 2012;3(1):61e78.
[17] Estrin Y, Dyskin A, Pasternak E, Khor H, Kanel-Belov A. Topological interlocking of protective tiles for the space shuttle. Philos Mag Lett 2003;83(6):
351e5.
[18] Carlesso M, Giacomelli R, Krause T, Molotnikov A, Koch D, Kroll S, et al.
Improvement of sound absorption and exural compliance of porous
alumina-mullite ceramics by engineering the microstructure and segmentation into topologically interlocked blocks. J Eur Ceram Soc 2013;33(13e14):
2549e58.
[19] Tessmann O. Topological interlocking assemblies. In: Achten H, editor. Digital
physicality: proceedings of eCAADe 2012. Czech Technical University in Prague, Faculty of Architecture; 2012. p. 211e9.
[20] Richardson M, Wisheart M. Review of low-velocity impact properties of
composite materials. Compos Part A 1996;27(12):1123e31.
[21] Abrate S. Impact on composite structures. Cambridge University Press; 2005.
[22] Ben-Dor G, Dubinsky A, Elperin T. A model of high speed penetration into
ductile targets. Theor Appl Fract Mech 1998;28(3):237e9.
[23] Zhu F, Lu G. A review of blast and impact of metallic and sandwich structures.
EJSE Spec Issue Load Struct 2007:92e101.
[24] Goldsmith W, Sackman JL. An experimental study of energy absorption in
impact on sandwich plates. Int J Impact Eng 1992;12(2):241e62.

149

[25] Wu H-YT, Fu-Kuo C. Transient dynamic analysis of laminated composite plates

subjected to transverse impact. Comput Struct 1989;31(3):453e66.
[26] Duan Y, Keefe M, Bogetti T, Cheeseman B, Powers B. A numerical investigation
of the inuence of friction on energy absorption by a high-strength fabric
subjected to ballistic impact. Int J Impact Eng 2006;32(8):1299e312.
[27] Simonsen BC, Lauridsen LP. Energy absorption and ductile failure in metal
sheets under lateral indentation by a sphere. Int J Impact Eng 2000;24(10):
1017e39.
[28] Dean J, Dunleavy C, Brown P, Clyne T. Energy absorption during projectile
perforation of thin steel plates and the kinetic energy of ejected fragments. Int
J Impact Eng 2009;36(10):1250e8.
[29] Wadley H, O'Masta M, Dharmasena K, Compton B, Gamble E, Zok F. Effect of
core topology on projectile penetration in hybrid aluminum/alumina sandwich structures. Int J Impact Eng 2013;62:99e113.
[30] Lambert J, Jonas G. Towards standardization in terminal ballistics testing:
velocity representation. Tech. Rep., DTIC Document. Aberdeen Proving
Ground, Maryland, USA: Ballistic Research Laboratory; 1976.
[31] Camacho G, Ortiz M. Computational modelling of impact damage in brittle
materials. Int J Solids Struct 1996;33(20):2899e938.
[32] Denoual C, Hild F. Dynamic fragmentation of brittle solids: a multi-scale
model. Eur J Mech e A Solids 2002;21(1):105e20.
[33] Grady D. Local inertial effects in dynamic fragmentation. J Appl Phys
1982;53(1):322e5.
[34] Dassault Systmes. Abaqus/explicit user's manual 6.12; 2012.
vila CG, Rose CA. Guidelines and parameter selection for the
[35] Song K, Da
simulation of progressive delamination. In: Abaqus user's conference, Newport, Rhode Island. USA: Hibbit Karlsson and Sorensen; 2008. p. 1e15.
[36] Brvik T, Langseth M, Hopperstad O, Malo K. Ballistic penetration of steel
plates. Int J Impact Eng 1999;22(9):855e86.
[37] Chi R, Serjouei A, Sridhar I, Tan GE. Ballistic impact on bi-layer alumina/
aluminium armor: a semi-analytical approach. Int J Impact Eng 2013;28(2):
37e46.
[38] Mileiko S, Sarkisyan O. Phenomenological model of punch-through. J Appl
Mech Tech Phys 1981;22(5):711e3.
[39] Larsson F. Damage tolerance of a stitched carbon/epoxy laminate. Compos
part A 1997;28(11):923e34.
[40] Dey S, Brvik T, Teng X, Wierzbicki T, Hopperstad O. On the ballistic resistance
of double-layered steel plates: an experimental and numerical investigation.
Int J Solids Struct 2007;44(20):6701e23.
[41] Brvik T, Dey S, Clausen A. Perforation resistance of ve different highstrength steel plates subjected to small-arms projectiles. Int J Impact Eng
2009;36(7):948e64.
[42] Xue L, Mock Jr W, Belytschko T. Penetration of dh-36 steel plates with and
without polyurea coating. Mech Mater 2010;42(11):981e1003.
[43] Garca-Castillo SK, S
anchez-S
aez S, Barbero E. Behaviour of uniaxially preloaded aluminium plates subjected to high-velocity impact. Mech Res Commun 2011;38(5):404e7.
[44] Seyed Yaghoubi A, Liaw B. Thickness inuence on ballistic impact behaviors of
glare 5 ber-metal laminated beams: experimental and numerical studies.
Compos Struct 2012;94(8):2585e98.