Composite Structures 153 (2016) 278–296

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Three-dimensional Voronoi model of a nacre-mimetic composite

structure under impulsive loading
Abdallah Ghazlan, Tuan D. Ngo, Phuong Tran ⇑
Department of Infrastructure Engineering, The University of Melbourne, Australia

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

Article history: Nacre, the inner layer of mollusk shells holds the key to the development of an effective composite
Received 9 January 2016 system for protecting structures from extreme loads due to its superior fracture toughness, despite its
Revised 23 May 2016 brittle constituents. It is known that the hard mineral tablets provide structural rigidity, while the soft
Accepted 7 June 2016
organic polymer matrix provides the mechanisms to mitigate damages and dissipate energy uniformly
Available online 7 June 2016
across the structure. Nacre’s composite structure is arranged to have multiple laminates and three
dimensional polygonal tablets bonded with organic adhesives to maximize its load sharing capability.
Keywords:
This paper presents a novel 3D Voronoi model of an Aluminum/Vinylester composite structure that
Bio-inspired composite
Nacre
closely mimics multilayer nacre’s tablet. Vinylester cohesive and adhesive layers are introduced between
Blast resistance nacre-mimetic polygonal Aluminum tablets and layers, respectively, to simulate the bonding and
Voronoi delamination process. The performances of nacre-like composite structures under blast loading are
Biomimetic evaluated in terms of maximum deformation, damage distributions as well as dissipated energy. The
Finite element influences of size and shape of the nacre-mimetic tablets, as well as the number of composite laminates
on the blast resistance of the composite are also investigated. Results reveal the importance of tablet size
and number of laminates as opposed to the insignificant influences of tablet overlapping.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction high hydrostatic pressure in the ocean. Better understanding of

Accidental and deliberate loads on civil and military structures
continue to cause severe damage worldwide, along with catas-
trophic losses of human life [1,2]. Consequently, the demand for
lightweight high-performance materials has increased signifi-
cantly in protective structural engineering. Biological structures,
meta-materials, woven fabrics, nano-polymers, composite sand-
wich panels and many others have recently been investigated by
structural research engineers because of their unique mechanical
characteristics, which make them suitable under a range of
extreme applications, namely blast, ballistics, fire and so on [3–10].
In particular, the two-layer armour system (Fig. 1b) found in
mollusk shells [11] is believed to be the most efficient armour sys-
tem, particularly because it is composed mainly of brittle minerals
but boasts a fracture toughness which is several orders of magni-
tude greater. This biological composite system has been perfected
by the animal over millions of years of evolution to protect its soft
tissues from loads that may arise from predator bites or extremely
⇑ Corresponding author at: Department of Infrastructure Engineering, The
University of Melbourne, Victoria 3010, Australia.
E-mail address: Phuong.tran@unimelb.edu.au (P. Tran).
nacre’s load sharing mechanism will lead to the development of
a superior composite structure for protective applications.
The armour system found in red abalone shells (and other
bivalves and gastropod species) consists of a hard brittle outer cal-
cite layer and a tough nacreous layer at its inner surface [12]. The
inner layer (nacre), which is mainly composed of aragonite, a brit-
tle mineral that accounts for 95% of its volume [13,14], exhibits
remarkable toughness. Moreover, nacre shows a hierarchical struc-
ture over several length scales (macro to nano). On the most ele-
mentary level, many have observed that nacre’s structure
resembles that of a brick wall at the micro-scale [15–21], with
polygonal aragonite platelets stacked over several layers and
bonded together by a soft organic matrix (Fig. 1c and d). This
organic matrix serves as both adhesive and cohesive bonds
between nacre’s layers and polygonal tablets, respectively. Most
studies in the literature, however, focus only on the influences of
the adhesive layer rather than the cohesive bonds between nacre’s
grains. Other features believed to contribute to nacre’s remarkable
toughness are nano-asperities on the tablet surfaces providing
additional sliding friction [22–25]; mineral bridges at the interface
as reinforcements between tablets [17,22,26] and waviness on the
surfaces of the tablets for strain hardening [11,27]. The aspect ratio

http://dx.doi.org/10.1016/j.compstruct.2016.06.020
0263-8223/Ó 2016 Elsevier Ltd. All rights reserved.
 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296
Fig. 1. (a) Red abalone shell; (b) two-layer armour system; (c) brick and mortar microstructure of nacre; (d) Voronoi-shaped polygonal architecture found in each nacreous
layer; (e) staggered platelet configuration in nacre (adapted from [6,7]).
of the platelets is also believed to have a certain influence on the
strength and stiffness of the nacreous composite [28,29].
Other investigations on nacre’s microstructural features (vol-
ume fractions, tablet aspect ratio, overlap length etc.) have been
conducted in an attempt to link them with its mechanical proper-
ties. Dutta et al. [30] claimed that nacre chooses its overlap length
to minimize crack driving forces at the interface, thereby delaying
crack initiation. Kotha et al. [31] concluded that composites with
high toughness can be manufactured from platelets with low
aspect ratios through a shear-lag modelling approach. Gao et al.
[24] employed Griffith’s fracture criterion to show that mineral
platelets become insensitive to flaws at small length scales. Barthe-
lat et al. [32] found that nacre does not achieve steady state crack
propagation due to toughness amplification from tablet pullout
and subsequent process zone toughening mechanisms. Other
investigations have also found that nacre has tremendous ability
to arrest crack propagation due to intrinsic and extrinsic toughen-
ing mechanisms that operate in front of and behind a developing
crack tip, respectively [33]. Flores-Johnson et al. [34] claimed that
the performance of nacre-like plates under blast loading is
explained by the hierarchical structure, which facilitates globalized
energy absorption by interlayered interlocking and delamination.
In summary, the geometric parameters such as tablet aspect
ratio, interfacial waviness, overlap length and interlayer interac-
tions of platelets have been found to influence nacre’s toughness.
The aforementioned investigations have focused mainly on the
localized load sharing mechanisms of nacre activated by tablet
sliding under uniaxial tension or pure bending. This raises the
question of whether the same mechanisms are activated under
transverse loadings such as hydrostatic pressure or blast impulse.
There are, however, very limited studies in the literature focusing
on this topic both numerically and experimentally. This paper
develops a novel nacre-mimetic composite model for simulating
the Voronoi-shaped tablets, multilayer structures, grain cohesion
and interfacial bonding to address: (1) The influences of the multi-
layered hierarchical structure of the nacre-mimetic composite on
its resistance to impulsive loadings; (2) The impact of laminate
staggering on the toughness of the composite; and (3) The size
and shape effect of the platelets on fracture resistance. Specifically,
the crack propagation patterns in the adhesive/cohesive layers and
the energy dissipated via fracture, delamination and plastic defor-
mation will be captured and analyzed. The model consists of a
Voronoi-like platelet arrangement resembling red abalone nacre
and the nacre-like laminates are bonded together with different
overlapping configurations. The composite platelets are modelled
279
with amour-graded Aluminum AA5083-H116, which are adhered
together by a Vinylester matrix. A rate-dependent material model
is used to simulate the transient responses and plastic deformation
of the Aluminum tablets under impulsive loading.
2. Voronoi model mimicking nacre’s tablet structure
2.1. Assembling the nacre mimicking geometry
Barthelat et al. [11] observed that the arrangement of the
tablets in each layer of nacre is similar to that of a Voronoi dia-
gram, through optical images of a red abalone specimen (Fig. 1d).
Based on these optical images, they have generated a model con-
sisting of two layers of nacre’s tablet structure for finite element
analysis. This procedure offers limited control over the geometry
and arrangement of the tablets in each layer, making it challenging
to develop nacre-mimetic composite systems. Each nacre laminate
is characterized by the random distribution of mineral tablets and
bonded to other nacreous layers. There is limited study on how
these nacreous tablet structures are chosen, as well as the influ-
ence of the shape and size parameters of the mineral tablets on
the load sharing capability of nacre. In this work, a novel technique
for developing a model that closely mimics the geometry of nacre
is described. This technique could be effectively used to fabricate
nacre-mimetic composites using 3D additive manufacturing such
as 3D-printing technology.
Voronoi diagrams are well-known in computational geometry.
Briefly, a Voronoi diagram comprises of sites (or points, illustrated
in Fig. 2a), where every point inside the polygon enclosing the site
is closest to that site. Mathematically, this can be represented by
the following equation:
Rk ¼ fx 2 Xjdðx; Pk Þ 6 dðx; Pj Þ8j–kg
where Rk is the set of all points in the Voronoi diagram X such that
the distance dðx; Pk Þ between each point x and a site P k is less than
or equal to the distance dðx; Pj Þ between that point and any other
site P j . Note that the distance function can be arbitrary but the
Euclidean distance is normally used to give a more uniform distri-
bution of the Voronoi regions, which happens to be closer to the
tablet profiles found in nacre.
The simplest possible configuration for a Voronoi diagram is
shown in Fig. 2a, where the sites are arranged in a grid to form
square Voronoi regions of length b. The sites could be shifted diag-
onally to generate an overlapping gap (a) between two adjacent
layers (Fig. 2b). A Voronoi diagram generated from the newly
280 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296

Fig. 2. (a) Plan view of a simple square Voronoi region with grid arrangement of Voronoi sites; (b) diagonally shifted Voronoi sites to obtain initial overlap between two
adjacent layers and; (c) elevation view showing staggered arrangement in nacre.

Fig. 3. (a) Voronoi diagram in a grid formation; (b) each site is chosen randomly within a circled region; (c) new Voronoi diagram is generated from the new site system and
clipped with a rectangle cut to avoid the infinite Voronoi region.

shifted sites serves as the nacre-mimetic architecture for staggered
composite laminates. Fig. 2c shows an elevation view of the stag-
gered layer arrangement, with layer thickness tn and the thickness
of the composite t. By randomly changing the locations of the sites,
a new Voronoi diagram could be generated, which includes poly-
gons of random shapes.
The technique of generating random Voronoi structures mim-
icking the nacreous tablets is described in Fig. 3. From the original
configuration (Fig. 3a), the sites are allowed to move within the
white circle (with radius R) randomly. This way, a new random
Voronoi diagram could be generated for each composite layer. A
representative infinite Voronoi diagram generated from the ran-
domly oriented sites is illustrated in Fig. 3c. Part of the diagram
consists of an infinite region, which is indicated by the whitespace
shown in Fig. 3c. To address such limitation, a computer script has
been used to create a rectangle cut, which is indicated by the
white-dashed line in Fig. 3c. By generating additional sites outside
this clipping rectangle, additional polygons are in turn generated
beyond the boundary. The polygons outside this boundary are then
clipped (or discarded) to achieve the finite nacreous layer shown in
Fig. 4a. As mentioned earlier, to generate a random Voronoi
diagram, the sites (Fig. 3a) can be shifted randomly inside the
vicinity of a circle (Fig. 3b). The positions of the sites ðx; yÞ can be
controlled by using the following simple equations based on polar
coordinates:
x ¼ x0 þ rcosðhÞ; y ¼ y0 þ rsinðhÞ ð1Þ
where:
ðx0 y0 Þ is the reference position of the site i.e. in a grid formation
(Fig. 3);
r is a random value in the range ½0; R, R is the radius of the circle
and;
h is a random angle in the range ½0; 2p.
The radius R is chosen such that the platelets are reasonably
polygonal. As R ! 0, the general polygons return to normal squares
(Fig. 2a).
2.2. The Baseline model
The baseline model comprises of five 1 m  1 m  3 mm-thick
layers of armour graded Aluminum (AA5083-H116) platelets,
bonded together by Vinylester adhesive layers of negligible thick-
ness. Each layer consists of approximately 100 Voronoi polygonal
tablets as illustrated in Fig. 4a. The models used in the parametric
studies in the following section have the same arrangement as the
 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296 281

Fig. 4. (a) The resulting finite Voronoi diagram representing a nacreous layer; (b) meshed region of the nacreous layer.

Fig. 5. (a) Cohesive element insertion between the grains of two 3D volumetric elements. (b) a simple process of cohesive element insertion at a junction of three volumetric
elements. All the cohesive elements have zero thickness, but thickened in the diagram for illustration.

baseline model, but with different layer thickness t n , shifting (stag-
gering) length a and grain size b (see Fig. 2) to maintain similar
overall thickness of the composite (t ¼ 30 mm). In this way, the
total mass of each model remains identical.
As indicated earlier, to generate two adjacent nacre-like lami-
nates (upper and lower) with different Voronoi staggering arrange-
ments, the original sites in the adjacent layers are first shifted
diagonally by a distance a (Fig. 2b) to obtain the desirable overlap
length. The sites are then allowed to take random locations within
the vicinity of a circle of radius R (Fig. 3b) and Voronoi diagrams
are generated for both layers. In the baseline model, each layer
consists of 100 platelets with dimensions of approximately
100 mm by 100 mm. Starting with the square grid arrangement
shown in Fig. 3, the sites are shifted diagonally by a dis-
tancea ¼ 33 mm. The table below presents specific values of
design parameters for various nacre-mimetic composite models
investigated in this work.
The above process is repeated to generate a composite panel of
five staggered laminates. The Voronoi diagram of each layer is
exported to a CAD format as trimmed parametric surfaces [35]
(Fig. 4a), by which they can be imported into a finite element pro-
gram and meshed (Fig. 4b). The trimmed parametric surface is the
extension to a well-known computational geometry technique
based on Bezier curves and non-uniform rational B-splines
(NURBS) [36]. A representative meshed region of the Voronoi layer
is illustrated in Fig. 4b.
The next step is to employ a graphics processing script to auto-
matically introduce cohesive elements of zero thickness at the
boundaries of the composite tablets (Fig. 4b). Fig. 5 illustrates an
algorithm to introduce the cohesive elements based on the existing
282 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296
finite element mesh. Here, each meshed layer (Fig. 4) is processed
by the graphics processing engine, which takes into account the
Voronoi geometry. A sufficiently fine finite element mesh of
3  3 mm-elements (Fig. 4b) is chosen to capture the transient
responses of the composite tablets and the decohesion processes
along the grain boundaries. The elements located along edges of
two connected Voronoi polygons are marked for cohesive element
insertion. The nodes shared between elements along the tablet
boundaries are split and connected by three dimensional cohesive
elements with zero thickness. To ensure the robustness and mesh
independence of the cohesive element insertion algorithm, a list of
geometrical and finite element mesh features are recorded
including:
 The line segments of each edge, which is formed by two points.
 The index of the first Voronoi polygon that shares the edge.
 The index of the second Voronoi region that shares the edge.
 A list of nodes belonging to the edge.
After inserting the new nodes associated with the cohesive
elements, the indices of the edge elements of the two neighboring
polygons are updated accordingly to point to these nodes. Fig. 5
illustrates a representative process of inserting zero-thickness
cohesive elements and the node numbering algorithm. In
Fig. 5a, a cohesive element is introduced between two volumetric
elements along the grain boundary, while Fig. 5b describes the pro-
cess of embedding cohesive elements at the junction. The thickness
of each cohesive element is set to zero, but thickened in Fig. 5 for
visualization purposes. The following steps illustrated in Fig. 5b are
taken to insert the zero-thickness cohesive elements:
 Starting from the shared edge between polygon 1 and polygon
2, nodes 2 and 3 are duplicated to create nodes 8 and 9,
respectively.
 Insert a new 4-node cohesive element between nodes 3-2-8-9
in the counter-clockwise order.
Fig. 6. Summary of the process to develop a FE model for single layer of nacre-mimetic platelets.
 Repeat the same steps for edges between polygons 2 and 3, and
polygons 3 and 1 to obtain the other two cohesive elements.
The order of nodes describing the cohesive elements is impor-
tant in identifying the debonding surface and to ensure the stabil-
ity of the simulation. From the two-dimensional model
constructed from the above steps, nodes and elements are copied
and extruded in the out-of-plane direction, and connected to form
three-dimensional elements. It is also important to have suitable
node numbering of the 3D cohesive elements. For example, in
Fig. 5a, the 3D volumetric element 1 has the node arrangement
1-2-3-4 for the top face, and 5-6-7-8 for the bottom face, both in
the counter-clockwise direction. This node number arrangement,
however, leads to the inaccurate interpretation of the through-
thickness direction of the embedded 3D cohesive element, which
^ . Hence, the stacking direction must
is parallel with normal vector n
be reconsidered.
The node sequence of the cohesive element is then reordered to
have the new top face formed by the node sequence 4-3-7-8 and
the bottom face formed by node sequence 1-2-6-5 in the
counter-clockwise direction. This way, the through-thickness
direction of the cohesive element is also coincident with the nor-
mal vector k,^ which is defined by the node sequence. Finally, the
node sequence for each cohesive element must be specified to have
the nodes belonging to the bottom face and followed by those of
the top face i.e. 1-2-6-5-4-3-7-8. A summary of entire process for
generating the overall finite element model has been summarized
by the flowchart in Fig. 6.
Once all five layers have been generated, they are bonded
together by thin adhesive layers to form the composite panel,
which is illustrated in as a brick and mortar structure in
Fig. 7a and b.
A representative FE mesh of a zero-thickness adhesive layer
between the Voronoi nacre-mimetic laminates is shown in
Fig. 7c and d. The three dimensional 2 mm  2 mm cohesive
elements (COH3D8) are employed to simulate the cohesive and
 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296 283

Table 1
Specific design parameters for the nacre-mimicking models (general design param-
eters are defined in Fig. 2).

Baseline Other models designed for parametric

model studies on the effects of
Number Staggering Voronoi
of layers arrangement grain size
Number of layers, n 5 10 and 2 5 5
Layer thickness, t n (mm) 6 3 and 15 6 6
Staggering length, 33 33 25 and 50 33
a (mm)
Average tablet length, 100 100 100 50 and 200
b (mm)

( )2 ( )2 ( )2
tn ts tt
0
þ 0 þ 0 ¼1 ð2Þ
tn ts tt
Fig. 7. (a) 3D view of a five-layer nacre-mimicking composite model; (b) Voronoi
polygons are bonded by zero-thickness cohesive elements and composite laminates
are bonded by adhesive layers; (c) the FE mesh of the adhesive layer; and (d) a where tn0, ts0, tt0 represent the peak values of the nominal stress
close-up view of the FE mesh for the adhesion layer. when the deformation is either purely normal to the interface, or
purely in the first or the second shear direction, respectively. The
power law form was adopted to describe the rate of stiffness
degradation:
adhesive elements. Tie constraints are applied to the interfaces
between the adhesive layers of the associated upper and lower
( )2 ( )2 ( )2
Gn Gs Gt
Voronoi platelets. In summary, the following element types are þ þ ¼1 ð3Þ
used in the above 3D model: GCn GCs GCt

 Continuum shell (SC8R 8-node brick and SC6R 6-node wedge),
with an average size of 3  3  3 mm for the platelet
elements.
 Cohesive elements (COH3D8 hexahedron), with zero thickness
for the cohesive bonds and adhesive layers between the
Voronoi tablets and nacre-mimetic composite laminates,
respectively.
where GCn, GCs , GCt refer to the fracture energy required to cause fail-
ure in the normal and shear directions, respectively. The cohesive
material model adopted in this work is representative of Vinylester
resin and its properties are tabulated in Table 2.
Table 2
Vinylester cohesive material model.

t0n ; t0s ; t0t 80 MPa

2.3. Cohesive damage model for bonding between the tablets and
GCn ; GCs ; GCt 1 mJ=mm2
composite laminates q 1850 kg=m3
Enn 4 GPa
Delamination is an important failure mode in composite mate- Ess ; Ett 1.5 GPa
rials subjected to transverse loads. It can cause a significant reduc-
tion in the load-carrying capacity of a structure. The debonding
behavior of the composite can be simulated by the Cohesive Zone
Model (CZM) [37–39]. Table 3
This research employs a bilinear cohesive model illustrated in Mechanical properties of Aluminum AA5083-H116.
Fig. 8b for modelling delamination in the cohesive and adhesive Property Value Property Value
bonds (see Fig. 7). The cohesive elements with finite thickness con- 3
Density, q 2750 kg/m Plastic
nect two volumetric elements as shown in Fig. 8a with traction- Elastic, A 215 MPa
separation laws, which relate the cohesive traction vectors E 70 GPa B 280 MPa
T ¼ ftn ; ts ; t t g and the displacement jump d ¼ fdn ; ds ; dt g, where m 0.3 N 0.404
the subscripts n, s and t, respectively, denote the normal and tan- m 0.859
C 0.0085
gential components. We adopt a simple bilinear quadratic cohesive
e_ 0 0:001 s1
law for damage initiation:

Fig. 8. (a) 3D Eight-node cohesive element; (b) rate-independent bilinear softening model.
284 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296
Fig. 9. Blast pressure–time history obtained from CONWEP for a 0.6 kg TNT charge at a standoff distance of 0.5 m.
2.4. Rate-dependent model for nacre-like tablets
A rate-dependent material model is adopted to simulate the
behavior of the Aluminum polygonal tablets. The Johnson–Cook
constitutive law for ductile metals [40] was utilized to predict this
behavior as follows:
r ¼ ½A þ B ½1 þ Cln ½1  T m
n _ 
ð4Þ
where: A represents the yield stress;  is the equivalent plastic
strain; B and n account for the effects of strain hardening;
_  ¼ _ =_ 0 is the dimensionless strain rate for the reference strain
rate _ 0 ¼ 0:001s1 ; the constant C is obtained from experiment
(tension, torsion, etc.); and the temperature T m is ignored, assum-
ing isothermal conditions. The material employed in this work is
Aluminum AA5083-H116, with properties listed in Table 3.
2.5. Blast model and boundary conditions
The model is clamped at its boundary and subjected to a TNT
charge of weight W ¼ 0:6 kg at a stand-off distance of R ¼ 0:5 m
from the front face. These blast parameters were to chosen to
ensure adequate delamination in the cohesive and adhesive
bonds in order to conduct the parametric studies outlined below.
The blast pressure–time history is plotted in Fig. 9 below. The
Conventional Weapons Effects (CONWEP) program, which
determines the air blast parameters based on empirical data from
explosive tests conducted by Kingery and Bulmash was used to
generate the blast pressure profile [41,42]. This model is also
adopted in the Unified Facilities Criteria UFC 3-340-02 in graphical
Fig. 10. Snapshots of crack propagation patterns in the adhesive layers of the
baseline nacre-like Voronoi composite model.
form for designing structures to resist explosions [43]. The blast
pressure profile is modelled using the modified Friedlander
equation, which is representative of an exponentially decaying
pressure as follows:
 
t tbt
Ps ðtÞ ¼ Ps0 1  e d ð5Þ
t0
Where Ps is the peak overpressure; t d is the positive phase
duration; t is the time and b is the blast waveform parameter,
which typically varies with standoff distance.
3. Numerical results and discussions
Fig. 10 illustrates the crack propagation patterns in the adhesive
layers of the baseline composite model. A middle cross-section
snapshot of the adhesive layers is captured at different times of
the dynamic event, showing the damage development history.
The failed adhesive elements are hidden to highlight the delamina-
tion zone, while they are still kept active to prevent tablet penetra-
tion. It can be observed that the crack initiates from the edges of
the layer near the blast source and propagates toward the central
region over time. Complete delamination is observed at the center
of the composite panel by the end of the simulation, while some
other adhesive zones remain completely intact.
The cohesive debonding patterns in the Voronoi composite lam-
inates presented in Fig. 11 exhibit similar trends, where the cohe-
sive bonds between the grains fracture in a similar pattern among
each layer. The debonding of Voronoi polygon tablets are shown
for three representative layers: front laminate (facing the blast
load), middle and back laminates. All undamaged cohesive ele-
ments are highlighted by the continuous black lines, while the
failed elements are marked in red. As can be seen from Fig. 11,
the cohesive failures of the tablets in the top layer are only con-
tained near the edge regions, which is consistent with the adhesive
failure pattern in Fig. 10. The cohesive damages become more
extensive in the middle layer, where the debonding occurs at the
center of the laminate from 1.5 ms and does not grow thereafter.
It is also noticed that the cohesive failures near the edges of the
middle layer are less severe compared to the frontal layer. The back
composite laminate, on the other hand, shows no cohesive debond-
ing at the edges, but early and extensive damages are observed at
the center of the layer. The central damages start as early as 0.9 ms
and continue to grow until 1.8 ms. While the cohesive failure pat-
terns of each laminate are different from each other, the total
debonding zones seem to spread out evenly among the layers. This
observation could be further confirmed in studies on the dissipated
energy history below.
 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296 285

Fig. 11. Snapshots of crack propagation patterns in the cohesive bonds between the Voronoi tablets of three representative layers of the baseline composite model. The
continuous black lines indicate intact cohesive elements, while the elements highlighted in red indicate completely damaged cohesive bonds.

Fig. 12. Time histories of damage dissipation energy associated with debonding of (a) the adhesive layers and (b) the cohesive bonds of the baseline composite model.

The time histories of damage energy associated with debonding are presented for each layer for comparison, thereby quantifying
of the adhesive layers and cohesive elements are plotted in the levels of damage related to cohesive and adhesive debonding,
Fig. 12a and b, respectively. The dissipated energy components as well as their distributions. The results appear to be consis-
286 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296
tent with the interlayer crack propagations observed in
Figs. 10 and 11 i.e. most of the energy imparted from the blast
shockwave is dissipated through delamination of the central adhe-
sive layers of the composite panel. On the other hand, the dissi-
pated energy associated with the cohesive failures between the
Voronoi polygons appear to be maximized in the back layer, which
is furthest from the blast source. The time evolutions of cohesive
damage energy curves are fairly similar for the inner laminates.
In general, the damages are distributed evenly among the cohesive
and adhesive layers, while the top and bottom layers carry the
largest damages. Thus, it may be reasonable to assume that by
introducing more composite layers or decreasing the size of the
Voronoi tablet grains, more impact energy could be dissipated.
Specifically, this could help to spread damages more uniformly
throughout the composite layers.
In nacre, the tablet sliding mechanism has been found to be a
key factor in activating intrinsic and extrinsic toughening mecha-
nisms that arrest crack propagation, namely crack bridging and
process zone toughening [33]. The mechanisms induce cohesive
forces and residual plastic strains, respectively to close a crack. This
sliding-induced stiffening mechanism is not dominant in this par-
ticular blast loading problem due to the transverse direction of the
impulsive load. Significant damages are attributed to cohesive and
adhesive debonding. Under impulsive loading, the energy dissi-
pated through damage and deformations is more prevalent than
Fig. 13. (a) Time histories of plastic dissipated energy of different Aluminum platelets of the baseline model and (b) snapshots of Von Mises stress fields in several layers of
the baseline composite model.
frictional contacts. The fact that nacre or similar mollusk shells
developed their bio-composite structures under constant
hydrostatic pressure poses several questions: Whether building
several hundred organized tablet layers is the key protective
mechanism of nacre to mitigate the transverse impulsive loadings
and how the size of the nacreous polygon tablets affect the load
sharing and energy absorption mechanisms. Hence, parametric
studies are conducted later in this work to quantify the effects of
several mechanical characteristics of the nacre-mimetic composite
structure, which are dependent mostly on the geometrical factors
such as the composite layup, overlap length and Voronoi tablet
size.
It is also worth noting that the levels of plastic dissipation
energy in the composite laminates presented in Fig. 13a are much
lower compared to the fracture energy dissipated due to
debonding of the cohesive/adhesive layers (Fig. 12). We can also
see from Fig. 13a that the top and bottom layers undergo
considerably larger plastic deformation compared to the middle
layers. This observation is consistent with the Von Mises stress
fields shown in Fig. 13b, where the yielding of the platelets is
relatively localized. This localized plastic deformation is clearly
visible at the center of the blast-facing layer at approximately
1.5 ms, and at the edges of the other layers further from the
blast source. There is, however, negligible plastic deformation in
the middle layer of the baseline composite model and this may
 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296 287

Fig. 14. Mesh convergence study for the baseline model. The top Voronoi layer and adhesive layer near the blast source are illustrated with several mesh densities: (a)
2  2 mm; (b) 3  3 mm; and (c) 6  6 mm elements; (d) total damage dissipation energy; and (e) total plastic dissipation energy. The element size in the adhesive layer was
kept constant at 1  1 mm. All the cohesive elements have zero thickness, but are thickened in the diagram for illustration.

be attributed to the energy released due to distributed fracture in
the cohesive bonds.
Fig. 14d and e illustrates the results of a convergence study in
terms of the total damage and plastic dissipation energy in the
nacre composite. The size of the tablet and cohesive elements were
varied as shown in Fig. 14a–c, such that compatibility was main-
tained at the interface between the tablets and cohesive zones.
The results show that the fracture process zone ahead of the crack
tip was simulated as accurately as possible.
4. parametric studies
4.1. Influence of the number of nacre-mimetic composite layers
Analysis from the previous section has demonstrated the rela-
tively uniform distribution of damage and energy absorption
among the different layers of the composite panel. Nacre’s compos-
ite structure (Fig. 1) is actually composed of hundreds of mineral
layers to maximize damage mitigation and energy absorption
capabilities. A parametric study is conducted in this section to
investigate the performances of two (n = 2), five (n = 5) and ten
(n = 10) layered composite panels of the same thickness. These
models are illustrated in Fig. 15, where the total thickness
t ¼ 30 mm is maintained for all three cases, as well as a Voronoi
polygon size (b) of approximately 100  100 mm. Performances
of the composite panels in terms of back face displacements and
velocities, as well as damage dissipation energy are compared.
The overall mass and thickness of the composite panel is main-
tained (see Table 1).
Firstly, the cross-section snapshots of crack propagation pat-
terns in the adhesive layers are compared among the multi-
layered models as shown in Fig. 16. It can be observed that the
cracks at mid-span do not propagate through the thickness of the
ten-layer model (n = 10). Interestingly, a notch is formed as early
as 0.9 ms and does not seem to grow further thereafter. The crack
arrest of the ten-layer composite structure could be attributed to
the activations of more damage sites mitigating the blast
shockwave. This behavior is quite different to the five (n = 5) and
two-layer (n = 2) models, where integrity of the composite is
completely lost due to catastrophic fracture in all of the adhesive
layers. Thus, the multi-layered architecture in the nacre-mimetic
composite appears to influence its crack arresting capability. In
these structures, cracks grow through more complex paths that
distribute damage transversely and horizontally, thereby increas-
ing the uniformity of damage energy dissipation throughout the
composite.
The ten-layer Voronoi model (n = 10) shows a significant reduc-
tion in the peak transverse displacements (Fig. 17a). At approxi-
mately 0.9 ms, the displacements among the three models begin
to deviate from each other and the two-layer (n = 2) model appears
to be stiffer. This is possibly attributed to the larger aspect ratio of
the platelets. Although the ten-layer model achieves a slightly lar-
ger peak transverse velocity in comparison with the other two
models at approximately 1.2 ms (Fig. 17b), it still shows signifi-
cantly higher stiffness. This may be attributed to the bonding
288 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296

Fig. 15. 30 mm-thick nacre-mimetic composite panels with a different number of tablet layers (n = 2, 5 and 10) corresponds to three cases a–c. (d) Plan view showing Voronoi
polygon structures of similar size used for all three cases.

Fig. 16. Crack propagation patterns in the adhesive layers of the Voronoi model.

between several adhesive layers, which remain intact such that the
transverse velocity diminishes noticeably towards the end of the
simulation. At this point, the five-layer (n = 5) model appears to
exhibit slightly less stiffness degradation compared to the two-
layer (n = 2) model, whereas the stiffness in the two-layer model
continues to degrade linearly because the adhesive layer has com-
pletely failed.
The energy dissipated via damage is relatively similar among
the multi-layered models (Fig. 18a). The energy dissipated in each
adhesive layer is broken down in Fig. 18b for the baseline model
(5 layers). A large variation can be observed, where the middle
adhesive layers (2 and 3) take the majority of damage, with a max-
imum dissipated energy of approximately 180 J in the adhesive
layer 3. Thus, the baseline model appears to be ineffective in dis-
tributing damage uniformly across the entire composite. It is noted
that the energy dissipated via damage in the cohesive elements
between the Voronoi grains is insignificant compared with that
of the adhesive layers (see Fig. 12). Fig. 18c illustrates the damage
dissipation energy in all adhesive layers of the ten-layer model
(n = 10). Maximum and minimum damage dissipation energies of
 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296
Fig. 17. (a) Peak transverse displacements; and (b) peak transverse velocities at the back facet of the multi-layered composite models.
approximately 90 J and 25 J, respectively, can be observed in the
adhesive layers adh-layer6 and adh-layer4 (indicated by an open
circle on each respective curve). With the baseline model (n = 5),
a maximum damage dissipation energy of approximately 180 J is
seen in the adhesive layer adh-layer3, with a minimum of 80 J in
adh-layer4 near the back facet. Compared with the baseline model
(n = 5), the adhesive damage in the ten-layer model (n = 10) is
spread more uniformly across the entire composite, where the
interior layers (adh-layer6, adh-layer3, adh-layer7 and so on) take
the most damage. This explains why the ten-layer model ulti-
mately shows higher stiffness, as the integrity of the inter-layer
bonding is maintained. Thus, by introducing double the amount
of Voronoi layers, the maximum energy dissipated via damage in
each layer is reduced by half, approximately. In contrast, the max-
imum energy dissipated in the two-layer model (Fig. 18d) is
approximately double that of the baseline model. In terms of
designing protective structures, this is very important as the
kinetic energy delivered to a critical structural component will
be reduced substantially by activating more damage sites to dissi-
pate energy, whilst maintaining the integrity of the protective
system.
The total energy dissipated in the cohesive bonds, adhesive lay-
ers and tablets are compared between the multi-layered nacre
composite panels (n = 2, 5, 10 layers) and a monolithic Aluminum
plate of equal areal mass in Fig. 19. It can be observed that an
increase in the number of composite layers increases the damage
dissipated in the adhesive bonds, whilst reducing damage in the
cohesive bonds between the tablets. This explains why the 10-
layer model retains stiffness as shown in Fig. 17 as opposed to
the stiffness degradation observed in the two-layer and baseline
models. Consequently, the ten-layer model exhibits larger energy
dissipation via plastic deformation in the tablets as shown in
Fig. 19d. This is important insight into the reason why the ten-
layer model (n = 10) is superior over the other models in terms of
289
deformation and energy dissipation. By spreading damage to the
adhesive bonds, the energy imparted from the blast shockwave is
dissipated more uniformly throughout the composite, as indicated
by the uniform energy distribution in the adhesives shown in
Fig. 18. In comparing the energy dissipated by the nacre composite
and a monolithic plate of equal areal mass (Fig. 19a), it is clear that
the nacre composite is superior because most of the energy
absorbed by the monolithic plate is in the form of recoverable
strain energy.
4.2. Influence of overlapping length (staggering parameter a)
A parametric study is conducted to investigate the influence of
the staggering length (a) between the tablets of adjacent layers on
the performances of the nacre-like composite models under blast
loading. It is important to note that in the nacreous composite
structure, due to the randomness, it is not feasible to establish a
uniform overlap length across the layer. Thus, in this research, an
initial staggered length (a) is used to diagonally shift one layer rel-
ative to its neighboring layers (see Fig. 2b) to define an initial over-
lap, and the sites are then randomly chosen to obtain randomly
shaped Voronoi tablet patterns as explained in the baseline model.
An average staggering length (a) of one-third of the tablet length
(b) is selected for the baseline model, which is based on the
average overlap observed in nacre (Fig. 20a). This configuration is
compared against other cases with average staggering lengths of
one-half and one-quarter of the tablet length, respectively
(Fig. 20b and c). A section view is presented in Fig. 20d, where
the baseline five-layer (n ¼ 5) composite panel is illustrated with
a constant thickness t ¼ 6 mm for each layer. The staggering length
(a) is also indicated. The central back facet displacements, as well
as the damage dissipation energies in the adhesive bonds, respec-
tively are compared among three cases with a ¼ b=3; a ¼ b=2 and
a ¼ b=4.
290 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296

Fig. 18. Damage dissipation energy in: (a) all multi-layered models; (b) the baseline model (five-layers, n = 5); (c) the ten-layer model (n = 10); and (d) the two-layer model
(n = 2). The circles mark the layers that incur minimum and maximum damage for comparing the uniformity of damage distribution throughout the composite.

It can be observed from Fig. 21 that the average staggering
length has a marginal influence on the deflection of the composite
panels. Various simplified two-dimensional unitcell analytical
models have been developed to elucidate why nacre prefers a cer-
tain overlap length to maximize its load bearing capacity under
pure tension. However, these models often ignore the three dimen-
sional effects and the randomness of the nacreous tablet arrange-
ments, which weaken the theory of optimum overlapping length.
It can be seen that this baseline overlap length (a = b/3) does not
give the smallest back face displacement under impulsive loading
(Fig. 21). A larger overlap length tends to give a smaller transverse
displacement at the mid-span of the panel, which suggests that the
2D analytical studies are inconclusive in pointing out the signifi-
cance of nacre’s overlap length. Furthermore, these models have
been studied for planar tensile loads, while loadings are transverse
in this study. This poses the question, what are the other mecha-
nisms that could elucidate why the nacreous composite chooses
a specific average overlap length? Is tablet sliding really prominent
to nacre’s activation of toughening mechanisms under transverse
loads?
The average overlap length found in columnar nacre has been a
focus of many researchers in the literature, who utilized the
shear-lag approach to model a unitcell under uniaxial tension,
and investigated its influence on the load transfer in nacre. Under
out-of-plane impulsive loading, it can be observed that the overlap
length is insignificant in terms of dissipating the energy imparted
from the blast shockwave. This may be attributed to the inter-
layer load transfer mechanism, which does not occur via shear at
the interface, as well as rapid delamination in the cohesive and
adhesive bonds. In the aforementioned 2D models, shear plastic
deformation is assumed at the interface, which is believed to
activate the relative sliding between the tablets and toughen the
nacreous composite structure.
It can be seen in Fig. 22a that the three models dissipate a sim-
ilar amount of energy via total damage. By comparing the damage
dissipation energy of the individual adhesive layers (Fig. 22b–d), it
can be observed that the model with one-half overlap (a = b/2)
gives the most uniform damage distribution among layers. This
may be attributed to the larger overlap area, which shields trans-
verse crack growth, thereby enforcing the redirection of the shock-
wave to propagate to other weaker adhesive regions. However, the
maximum dissipated energy of 175 J is similar among the models,
meaning that the overlap length is not influential in terms of mit-
igating the energy throughout the volume of the composite. This is
further reinforced by the component-wise energy dissipation
shown in Fig. 23, where the difference in energy dissipation in
the adhesive layers, cohesive bonds and tablets is negligible.
This poses a question on whether the overlap length is a significant
design parameter, based on the 2D planar unitcell models.
However, it could very well be significant when coupled with other
parameters controlling the tablet geometry. Generally, in nacre-
mimicking composites, there will most likely be fracture at the
 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296 291

Fig. 19. Total energy dissipated in: (a) the entire composite vs. a monolithic plate of equal areal mass; (b) the adhesive layers; (c) the cohesive bonds; and (d) the tablets.

Fig. 20. Effects of the staggering length (a) on three composite staggering configurations: (a) the baseline model (a = b/3); (b) a = b/2 and (c) a = b/4. (d) Elevation view
showing five layers of the baseline model with a constant layer thickness of 6 mm.
292 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296

Fig. 21. Transverse displacement at the rear layer furthest from the blast source for several staggered lengths.

Fig. 22. Damage dissipation energy in: (a) all models; (b) the adhesive layers in the baseline model (a ¼ b=3); (c) the adhesive layers in the model with one-half tablet
overlap; and (d) the adhesive layers in the model with one-quarter tablet overlap.

adhesives under impulsive loading, and it is important to under-
stand how the crack propagates and arrests.
4.3. Influence of grain size
A parametric study is conducted to investigate the influence of
the grain size (b) of the nacre-like composite models under blast
loading. The baseline model (b ¼ 100 mm) consists of 10  10 Vor-
onoi polygons. The performance of the baseline composite panel is
then compared with two other models (see Fig. 24) with tablet
sizes of b ¼ 50 mm (20  20 polygons) and b ¼ 200 mm (5  5
polygons). Specifically, the peak transverse displacements and
the damage dissipation energies in the individual layers of each
model are compared to determine the influence of the size-effect
under blast.
It can be observed from Fig. 25 that a significant reduction in
the peak transverse displacement occurs in the model with tablet
size b ¼ 50 mm when compared with the baseline model. This
indicates that smaller grains with more cohesive surfaces are more
favorable under impulsive loading. It has also been suggested in an
 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296 293

Fig. 23. Total energy dissipated in: (a) the entire composite; (b) the adhesive layers; (c) the cohesive bonds; and (d) the tablets.

Fig. 24. Section view of tablet (grain) distributions in (a) the baseline model b ¼ 100 mm (10  10 polygons); (b) the model with b ¼ 50 mm (20  20 polygons) and (c) the
model with b ¼ 200 mm (5  5 polygons). (d) Elevation view showing five layers of constant thickness (6 mm) used in the above models.

earlier parametric study that the ten-layer composite model with are seen in both models (the 10-layer and the model with the
more adhesive bonds shows more uniform distribution of damage smallest grain size b ¼ 50 mm). The ten-layer model with larger
throughout the composite. Similar peak transverse displacements Voronoi polygons but thinner grains (100  100  3 mm) slightly
294 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296

Fig. 25. Central transverse displacement histories of the rear layers away from the blast load for three models with different grain sizes.

Fig. 26. Comparing damage dissipation energy of (a) all models; (b) individual adhesive layers in the baseline model b ¼ 100 mm (10  10 polygons); (c) model with
b ¼ 50 mm (20  20 polygons); and (d) model with b ¼ 200 mm (5  5 polygons).

outperforms the five-layer model with smaller grain areas but
thicker tablet layers (50  50  6 mm).
The total damage dissipation energy components are compared
in Fig. 26a. The model with double-sized polygons (b ¼ 200 mm) is
found to exhibit more overall damage than the baseline model.
Upon comparing the damage dissipation energy of individual
adhesive layers (Fig. 26b–d), it can be observed that the
composite with b ¼ 50 mm is more effective in terms of spread-
ing damage throughout its volume, with most of the damage
occurring in the middle layers. This is similarly observed in
the baseline model, which takes approximately two-thirds more
damage (180 J as opposed to 110 J). The model with
b ¼ 200 mm tends to focus damage in the lower adhesive
layers, which leads to massive stiffness degradation as shown
in Fig. 25. Thus, smaller grain sizes appear more favorable in
terms of mitigating damage whilst retaining the stiffness of
the composite. However, contrary to the multi-layered (n = 2,
5 and 10 layers) parametric study, decreasing the grain size
reduces damage in the adhesive layers whilst increasing
damage in the cohesive bonds (see Fig. 27).
 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296
Fig. 27. Total energy dissipated in: (a) the entire composite; (b) the adhesive layers; (c) the cohesive bonds; and (d) the tablets.
5. Conclusions
Several nacre-mimicking composite models were developed for
investigating the influence of nacre-mimetic microstructural fea-
tures on the blast resistance of composite panels. The following
parameters were assessed in terms of the damage mitigation effi-
cacy of the multilayer nacre-like composite structure: (a) number
of staggering composite layers, (b) staggering lengths, and (c)
tablet grain sizes. The results showed that by adding more layers
to the composite, the integrity of the system was maintained such
that a significant reduction in the peak displacement was observed,
which indicated the stiffness of the multilayer composite was
enhanced under impulsive loadings. Furthermore, the damages
caused by decohesion and delamination of the layers and tablets
were spread more uniformly throughout the volume of the com-
posite. The maximum damage dissipation energy was reduced by
approximately half compared with the baseline model. The crack
propagation patterns indicated that the ten-layer model was better
in arresting and redirecting crack growth, thereby enabling more
damage sites to be activated to dissipate energy. This study gave
some insights into the dynamic behaviors of nacre-like multilay-
ered composite structure under impulsive loads. In contrast, the
staggering length parameter was found to have little influence in
terms of improving stiffness and energy dissipation uniformity in
the composite. Smaller grain sizes showed significant potential in
dissipating energy whilst retaining stiffness of the composite. This
indicated that the superior toughening mechanisms found in nacre
occur at the micro-scale and its unique microstructure holds the
key to the design and development of a sacrificial layer for protect-
ing critical structural elements from blast loads.
295
Acknowledgements
This research was sponsored by the Australian Research Council
linkage Grant LP150100906 and by the University of Melbourne
Early Career Researcher Grants.
References
[1] Ngo T, Mendis P, Gupta A, Ramsay J. Blast loading and blast effects on
structures – an overview. Electron J Struct Eng 2007(Special Issue):76–91.
[2] Thomet L. Four dead, 16 injured in fiery blast at Mexico oil rig; 2015.
[3] Imbalzano G, Ngo T, Tran P. A numerical study of auxetic composite panels
under blast loadings. Compos Struct 2015;135:339–52.
[4] Nguyen QT, Ngo T, Tran PA, Mendis P, Battacharyya D. Influences of clay and
manufacturing on fire resistance of organoclay/thermoset nanocomposites.
Compos A Appl Sci Manuf 2015;74:26–37.
[5] Yang CC, Ngo T, Tran P. Influences of weaving architectures on the impact
resistance of multi-layer fabrics. Mater Des 2015;85(11):282–95.
[6] Tran P, Ngo T. Bio-inspired composite structures subjected to underwater
impulsive loading. Comput Mater Sci 2014.
[7] Imbalzano G, Tran P, Ngo TD, Lee PV. Three-dimensional modelling of auxetic
sandwich panels for localised impact resistance. J Sandwich Struct Mater 2015.
[8] Mohotti D, Ali M, Ngo T, Lu J, Mendis P. Strain rate dependent constitutive
model for predicting the material behaviour of polyurea under high strain rate
tensile loading. Int J Mater Des 2014;53:830–7.
[9] Raman SN, Ngo T, Lu J, Mendis P. Experimental investigation on the tensile
behavior of polyurea at high strain rates. Mater Des 2013;50:124–9.
[10] Mohotti D, Ngo T, Mendis P, Raman SN. Polyurea coated composite aluminium
plates subjected to high velocity projectile impact. Mater Des 2013;52:1–16.
[11] Barthelat F, Tang H, Zavattieri P, Li CM, Espinosa HD. On the mechanics of
mother-of-pearl: a key feature in the material hierarchichal structure. J Mech
Phys Solids 2007;55:306–37.
[12] Currey JD. Mechanical properties of mother of pearl in tension. Proc R Soc
London Ser B Biol Sci 1977;196(1125):443–63.
[13] William Pro J, Lim RK, Petzold LR, Utz M. GPU-based simulations of fracture in
idealized brick and mortar composites. J Mech Phys Solids 2015;80:68–85.
296 A. Ghazlan et al. / Composite Structures 153 (2016) 278–296
[14] Zhang N, Yang S, Xiong L, Hong Y, Chen Y. Nanoscale toughening mechanism of
nacre tablet. J Mech Behav Biomed Mater 2016;53:200–9.
[15] Sarikaya M. An introduction to biomimetics: a strutural viewpoint. Microsc
Res Tech 1994;27:360–75.
[16] Shao Y, Zhao HP, Feng XQ, Gao H. Discontinuous crack bridging model for
fracture toughness analysis of nacre. J Mech Phys Solids 2012;60:1400–19.
[17] Song F, Soh AK, Bai YL. Structural and mechanical properties of the organic
matrix layers of nacre. Biomaterials 2003;24:3623–31.
[18] Sun J, Ling M, Wang Y, Chen D, Zhang S, Tong J, et al. Quasi-static and dynamic
nanoindentation of some selected biomaterials. J Bionic Eng 2014;11:144–50.
[19] Tang H, Barthelat F, Espinosa HD. An elasto-viscoplastic interface model for
investigating the constitutive behavior of nacre. J Mech Phys Solids
2007;55:1410–38.
[20] Wilbrink DV, Utz M, Ritchie RO, Begley MR. Scaling of strength and ductility in
bioinspired brick and mortar composites. Appl Phys Lett 2010;97:1–3.
[21] Ghazlan A, Ngo TD, Tran P. Influence of interfacial geometry on the energy
absorption capacity and load sharing mechanisms of nacreous composite
shells. Compos Struct 2015;132:299–309.
[22] Barthelat F. Nacre from mollusk shells: a model for high-performance
structural materials. Bioinspiration Biomimetics 2010;5.
[23] Evans AG, Suo Z, Wang RZ, Aksay IA, He MY, Hutchinson JW. Model for the
robust mechanical behavior of nacre. J Mater Res 2001;16(9):2475–84.
[24] Song F, Bai YL. Effects of nanostructures on the fracture strength of the
interfaces in nacre. J Mater Res 2003;18(8):1741–4.
[25] Wang RZ, Suo Z, Evans AG, Yao N, Aksay IA. Deformation mechanisms in nacre.
J Mater Res 2001;16(9):2485–93.
[26] Meyers MA, Chen PY, Lin AYM, Seki Y. Biological materials: structure and
mechanical properties. Prog Mater Sci 2008;53:1–206.
[27] Blank S, Arnoldi M, Kohoshnavaz S, Treccani L, Kuntz M, Mann K, et al. The
nacre protein perlucin nucleates growth of calcium carbonate crystals. J
Microsc 2003;212(3):280–91.
[28] Gao H, Ji B, Jager IL, Arzt E, Fratzl P. Materials become insensitive to flaws at
nanoscale: Lessons from nature. Proc Natl Acad Sci USA 2003;100
(10):5597–600.
[29] Wegst UGK, Ashby MF. The mechanical efficiency of natural materials. Phil
Mag 2004;84(21):2167–81.
[30] Dutta A, Tekalur SA. Crack tortuosity in the nacreous layer – topological
dependence and biomimetic design guideline. Int J Solids Struct
2014;51:325–35.
[31] Kotha SP, Li Y, Guzelsu N. Micromechanical model of nacre tested in tension. J
Mater Sci 2001;36(8):2001–7.
[32] Barthelat F, Rabiei R. Toughness amplification in natural composites. J Mech
Phys Solids 2011;59:829–40.
[33] Rabiei R, Bekah S, Barthelat F. Failure mode transition in nacre and bone-like
materials. Acta Biomater 2010;6(10):4081–9.
[34] Flores-Johnson EA, Shen LM, Guiamatsia I, Nguyen GD. A numerical study of
bioinspired nacre-like composite plates under blast loading. Compos Struct
2015;126:329–36.
[35] The Initial Graphics Exchange Specification (IGES) Version 6.0; 2001. p. 15–34,
63–181.
[36] Schneider P, Eberly DH. Geometric tools for computer graphics. In: Barsky BA,
editor. The Morgan Kaufmann series in computer graphics and geometric
modeling. US: Morgan Kaufmann Publishers; 2003.
[37] Tran P, Kandula SS, Geubelle PH, Sottos NR. Comparison of dynamic and quasi-
static measurements of thin film adhesion. J Phys D 2011;44(3).
[38] Tran P, Kandula SSV, Geubelle PH, Sottos NR. Hybrid spectral/finite element
analysis of dynamic delamination of patterned thin films. Eng Fract Mech
2008;75(14):4217–33.
[39] Tran P, Kandula SSV, Geubelle PH, Sottos NR. Dynamic delamination
of patterned thin films: a numerical study. Int J Fract 2010;162(1–2):77–90.
[40] Johnson GR, Cook WH. A constitutive model and data for metals subjected to
large strains, high strain rates and high temperatures. In: Proceedings of the
7th international symposium on ballistics.
[41] Hyde D. ConWep – application of TM 5-855-1. In: Fundamentals of protective
design for conventional weapons. Vicksburg, MS: USACE Waterways
Experiment Station; 1992.
[42] Kingery CN, Bulmash G. Airblast parameters from TNT spherical air burst and
hemispherical surface burst. Aberdeen Proving Ground. Aberdeen,
MD: Ballistic Research Laboratory; 1984.
[43] Engineers, U.A.C.o., UFC 3-340-02: structures to resist the effects of explosions,
U.D.o. Defense, Editor; 2008, Washington, DC.