Defence Technology 17 (2021) 1514e1530

Contents lists available at ScienceDirect

Defence Technology
journal homepage: www.keaipublishing.com/en/journals/defence-technology

Experimental and numerical investigation on penetration of clay

masonry by small high-speed projectile
Cheng-zong Wang a, *, Ai-jun Chen a, b, **, Zi-qing Li a, Chao-an Gong b, Shu Wang b,
Wen-min Yan b
a
School of Science, Nanjing University of Science and Technology, Nanjing, 210094, China
b
Science and Technology on Transient Impact Laboratory, Beijing, 102202, China

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

Article history: This study investigates a kind of masonry consisting of clay-fired brick (fc ¼ 10 MPa; r ¼ 1:38 g=cm3 )
Received 2 February 2020 and mortar (fc ¼ 10 MPa; r ¼ 1:8 g=cm3 ). Clay-fired brick masonry connotes a traditional construction
Received in revised form material of old architecture and public buildings. We carried out penetration experiments in which four
8 September 2020
clay-fired brick walls employing two different patterns were subjected to impact from small high-speed
Accepted 16 September 2020
Available online 5 October 2020
projectile, i.e. 12.7 mm armor-piercing explosive incendiary projectile and material tests in which the
static and dynamic compressive strengths of clay-fired brick and mortar were determined by quasi-static
and SHPB (Split Hopkinson Pressure Bar) tests. The experimental data include hit and exit velocities,
Keywords:
Clay-fired brick
damage configuration of clay brick masonry and mechanical properties of material at low and high strain
Penetration of masonry rates, though which influence of thickness and bonding patterns of wall on kinetic loss of bullet, the
RHT model damage patterns of masonry observed experimentally and dynamic increase of material strengths are
Impact analyzed. To keep minimum boundary inconsistency with reality, full 3D detailed finite element model
High strain rate consisting of two different material is established. Sharing common nodes and employing automatic
tiebreak contact are combined to reduce computational time usage of large-scale model. For description
of clay-fired brick and mortar RiedeleHiermaiereThoma (RHT) material model is employed. Material
parameter set is derived based on experimental data, available literature and engineering assumptions.
The numerical simulations study the mesh resolution dependency of material model, reproduce the
crucial phenomena of masonry in experiment acceptably and offer more time-resolved insight into
motion of bullet in the process of penetration. The feasibility of means of constructing finite element
model and applying RHT model to the masonry herein and analogous constructions is explored through
numerical investigation.
© 2020 China Ordnance Society. Publishing services by Elsevier B.V. on behalf of KeAi Communications
Co. Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

1. Introduction people around the world have relied on the structure in addition to
its characteristics mentioned for it is robust enough for human
Masonry is one of the most common building materials, which beings to obtain a safe and tidy environment. Masonry walls consist
is used on account of low cost, good sound and heat insulation of bricks and adhesive joints by which the bricks are connected. The
properties, locally available material [1]. It has a long history. The bricks and joints are of varying types. Fired-clay bricks (fc ¼ 10 MPa;
first bricks were based on dried mud and were used for the first r ¼ 1:38 g=cm3 ) and mortar (fc ¼ 10 MPa; r ¼ 1:8 g=cm3 ) are of
time in 8,000 BC in Mesopotamia [2]. Ever since its invention, relevance in this paper. Even though the use of clay bricks has been
shrank because of environmental impacts brought by fired-clay
brick industry, there are still considerable amount of clay brick
* Corresponding author. masonry buildings existing such as churches, historical construc-
** Corresponding author. School of Science, Nanjing University of Science and tions and civilian infrastructure in the world. The predictive
Technology, Nanjing, 210094, China.
E-mail addresses: wdwcz@njust.edu.cn (C.-z. Wang), chenaijun@njust.edu.cn
response of clay brick masonry to seismic load and dynamic im-
(A.-j. Chen). pacts is instructive for design configuration [3]. The ballistic
Peer review under responsibility of China Ordnance Society response of masonry is of interest because of some military value

https://doi.org/10.1016/j.dt.2020.09.017
2214-9147/© 2020 China Ordnance Society. Publishing services by Elsevier B.V. on behalf of KeAi Communications Co. Ltd. This is an open access article under the CC BY-NC-
ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530 1515

[4]. In the past decades the characteristic of concrete under dy- interrelationships of brick and mortar; it is not easily available
namic loading conditions such as penetration or explosive impact because of limitation of experimental tests and variation in prop-
has been studied thoroughly but the similar research with respect erties and proportion of material [1]. The research mentioned
to brick, mortar and masonry was relatively much less. previously has investigated quasi-static and dynamic behaviors of
What has been historically employed to investigate the dynamic brick, mortar and masonry in different perspectives. Some focused
properties of concrete include theoretical study, experimental on the constitutive formula of brick, mortar and bond between
investigation and numerical simulation. It has been shown that them, some detailed ways of constructing model of simulation,
numerical simulation can provide more information of process some investigated the mode of failure of masonry subjected to
history for sophisticated analysis than experimental investigation. quasi-static pressure, explosive blast impacts or seismic loads,
Nevertheless, the data from experiments are crucial because on the which all take us further on the road of understanding masonry as a
one hand any modern analysis and numerical computation should composite building structure, properties of its anisotropic constit-
necessarily be validated by those from real shooting range tests, on uents, namely brick and mortar, and interaction between them.
the other hand the material parameters which are of vital impor- Still, the knowledge and code with regard to masonry and relevant
tance to obtain a reasonable and compelling result must be fur- material is far less consensual than concrete or steel [25].
nished by or refer to experimental data [5]. The experimental In this paper experimental and numerical investigation on
exploration indicates that the masonry constituents, namely clay response of clay masonry subjected to bullet impact is presented. It
brick and mortar joint, act similarly to concrete which is brittle addresses the damage model of this class of construction consisting
material as well when encountering static loadings and dynamic of clay-fired brick and mortar in experiment. Through simulation
impacts [6e8]. Many material models in simulation are proposed to the reliable finite element model and material model are developed
describe the behavior of concrete, which are able to be applied to and explored.
solution of simulating behavior of masonry reasonably. Li et al. [9] In the following, the experimental campaigns including pene-
investigated the response mode of unreinforced clay brick masonry tration and material tests are elaborated and the results and dis-
subjected to vented gas explosion through experiment and simu- cussion are reported in Section 2. The material model and finite
lation, in which the brittle damage material model developed in element model are described in Section 3. Numerical results and
Ref. [10] for concrete has been adopted. The potential of assessment are presented in Section 4.
Holmquiste JohnsoneCook (HJC) material model was explored
profoundly in Ref. [11] to apply in simulation of penetration of 2. Experimental investigation
masonry, which was initialled by Holmquist and Johnson in 1993
[12] for concrete. Based on phenomenology of concrete Rie- To study the weapon resistance of masonry and the capability of
deleHiermaiereThoma (RHT) [13e15] material model describing the bullet to penetrate clay masonry, the shooting range tests were
macroscopic properties of concrete was used even in simulation of conducted with 12.7 mm armor-piercing explosive incendiary
adobe [16e19]. bullet as penetrator and clay masonry employing two different
To better represent the performance of material under dynamic bonding patterns as target. After range tests, the parametric tests of
loadings the effect of strain rate must be taken into consideration. materials were carried out to measure mechanical properties of
Experimental investigation is normally a way of studying clay brick brick and mortar individually.
and mortar at different strain rates and DIF (dynamic increase
factor) is defined to describe the increment to properties of mate- 2.1. Test configuration
rial [3,6,7]. Numerical simulation was adopted to explore the effect
of strain rate in Ref. [20]. Most of literature focused on the dynamic The Type MU15 bricks were made according to GB5101-2017
characteristic of brick or mortar under compressive impact. The [28] possessing nominal dimensions of 240mm 
tensile DIF of masonry joint was investigated in Ref. [21] and the DIF 115mm  53mm. Four walls with two kinds of bonding patterns
3.1 relative to reference strain rate of 11 was concluded. It was were built. The clay-fired bricks were connected by Type M5 mortar
found that the increment to material properties results not only with thickness of 1~2 cm. With employing different bonding
from intrinsic properties of material but also structural effect pattern, the masonry is called “24 wall” or “37 wall”, which is
associated with dynamic experiments, such as inertial effect [22] named after their thickness of masonry respectively though the
and end friction confinement [23]. The contribution of fibres and actual values vary slightly. Fig. 1 shows the maximal units to
water content in adobe to DIF has been illustrated in Ref. [24], construct a wall by connecting the top and bottom of same units,
which provides a reasonable explanation for dynamic strength- the dark thick lines representing the mortar. The unit of 37 wall,
ening phenomenon for adobe but also an appropriate interpreta- showed in Fig. 1 (b), consists of two layers of bricks, one of which
tion for other analogous material. can exchange the front row of bricks with the rear one to obtain the
The strategies to numerically model masonry wall fall into three same pattern as the other. Each layer has one or two partial bricks
categories in scale [8]. First method is macro-model in which the to avoid the continuous mortar across the whole wall without any
inhomogeneous composite material masonry, consisting of bricks corners, which decreases the clay brick masonry’s capacity to bear
and mortar, is homogenized as one single material, which is used to loading. It should be stated that realistic constructions are not
capture the behavior of the whole wall and circumvent computa- completely same as examples in Fig. 1 but follow similar method-
tional expense [25]. The second approach is simplified micro- ology. The experimental specimens are shown in Fig. 2. The charge
model which simplifies the mortar joint as a surface of no thick- format of the bullet is 16 g. Different initial velocities of the bullets
ness and the dimensions of the masonry units should be adjusted were achieved by altering the charge to 14 g and 12 g.
properly, as in Ref. [20,26,27]. The last one is micro-model which To ensure safety of shooter and collaborators, a rope was con-
constructs masonry with an assemblage of bricks connected by nected to trigger of machine gun. Fig. 3 shows the setup by which
mortar joints and retains the most detail. The model can examine the shooter hiding behind the shelter can press the trigger through
the failure of brick, mortar joint and the bond between the brick the rope. After the bullets were fired with zero attack angle, the
and mortar, as in Ref. [9]. velocities of which were first measured by laser velocimeter which
Analytical modeling of masonry demands the properties and is 8 m from muzzle. Then the bullets arrived at the masonry and
high-speed camera (resolution: 1024  1008; frames rate: 13,500
1516 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530

Fig. 1. Two different bonding patterns: (a) 24 wall; (b) 37wall. Fig. 3. Experimental delineation.

frames/s) recorded the process of the bullets penetrating the not measured due to difficulty to break clay brick wall.
masonry. Supposing that the mass of the bullet is a constant over the
process, i.e. m ¼ 48 g, define E by Eq. (1) as energy loss character-
2.2. Test results and discussion izing the capacity of bullet to penetrate clay masonry (or the
resistance of the masonry against impact of bullet).
The hit and exit velocities were estimated from high-speed
videos by reading key frames and considering the distance be- E ¼ EI  ER (1)
tween bullets and high-speed camera, which were calibrated by the
measured velocities from laser velocimeter. The laser velocimeter is where EI denotes the initial energy of the bullet, ER represents the
not used to measure the velocities directly because the block of the residual energy after penetration. The energy herein is considered
diffusing smoke and flying debris will interference with the results as energy of motion of bullet only. Using hit and exit velocities
considerably, as Fig. 4 (a) showed. Fig. 5 (a) presents the cartridges. derived from tests yields EI and ER respectively. According to test 1
The projectile consists of four components: steel core, lead sheath, and 4, the energy loss of bullets in 37 wall and 24 wall are E37 z1:
incendiary agent and jacket. The materials of steel core, lead sheath 689  104 J and E24 z1:077  104 J respectively, which demon-
and jacket are T12A stool steel, Pb and F11 copper clad steel strates 37 wall consumed more energy than 24 wall. The ratio of
respectively. As presented in Fig. 10 right, the incendiary agent is energy loss in 37 wall to that in 24 wall is 1.568 that was little
located in front of steel core, however which is omitted for greater to ratio of 37 wall’s thickness to 24 wall’s that is
simplification. The steel core shown in Fig. 5 (b) is retrieved from 37:5=24z1:563, which reveals that the bonding patterns to some
the masonry specimen, geometric properties of which are given in extent improve resistance of brick wall against impact. However,
Table 1. The experimental results are shown in Table 2. The masonry the improvement is insignificant. Therefore thickness of wall is a
in test 2 and 3 were not perforated. The depth of penetration was more relevant factor predicting the resistance of clay brick masonry

Fig. 2. 24 wall and 37 wall (left: 24 wall; right: 37 wall).

 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530 1517

Fig. 4. (a) The diffusing smoke and flying debris; (b) tilt of masonry wall.

Fig. 5. (a) The cartridges of projectiles; (b) the steel core retrieved after experiment.

Table 1 2.3. Material test

Dimensions of steel core illustrated in Fig. 5 (b).

Dimensions of steel core After the range tests, the densities of the materials were
Nose Shape Ogival
measured first. It has reached a consensus that the building ma-
Overall length 52.2 mm terials under dynamic loadings usually behave differently from
Maximum diameter 10.8 mm quasi-static loadings. Most building materials show an increase of
Length of tail 9.5 mm strength and stiffness with an increase of the strain rate. To
Length of cylinder 25.77 mm
determine dynamic increase of material properties the quasi-static
Curvature radius of head 32.57 mm
Cone angle of tail 8 and Split-Hopkinson-Pressure-Bar (SHPB) tests were employed to
investigate the properties of materials under static and dynamic
conditions.
A number of fragments of brick and mortar were collected from
than bonding patterns. Moreover, it is clear the energy loss over
masonry by breaking the walls after the range tests. The properties
thickness is approximately constant and with assumption of con-
of fragments of brick or mortar were assumed to be the same,
stant mass of projectile, the energy loss is proportional to difference
because they should have similar properties though realistically,
of square of two velocities, which leads to constant average
the properties of brick and mortar vary slightly from fragment to
deceleration with any given thickness of wall then constant average
another fragment. Some of them were used to measure densities,
drag force in the process of penetration, which is consistent with
were weighed and measured volume through Archimedes method.
the inference from qualitative analysis in Refs. [29].
The densities of materials were calculated according to Eq. (2) [30].

Table 2
Test results.

Test number Type of wall Charge Perforation or not Hit velocity=ðm,s1 Þ Exit velocity=ðm,s1 Þ
ðAccuracy: ±10Þ ðAccuracy: ±10Þ

1 37 wall 16 g P 842.4 78.6

2 37 wall 16 g NP 807.9 /
3 37 wall 14 g NP 768.5 /
4 24 wall 12 g P 684.1 139.1
1518 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530

three samples of brick and mortar as well. Strain values are in a

wd rw
rb ¼ (2) state of changing rapidly in dynamic tests, which can be captured
ws  wA by the semiconductor strain gauges. The high dynamic strain in-
dicator recording the data through transferring to computer was
where wd is dry weight, ws represents the saturated weight, wA is
connected to the wheatstone bridges consisting of semiconductor
the weight of the saturated specimen suspended in a container of
strain gauges and the basic circuit configuration. Utilizing Eq. (3)
the saturating liquid (water), rw denotes the density of water and rb
yields the transient strain.
is the bulk density of the material. The average bulk densities of
brick and mortar are 1:38g=cm3 and 1:8g=cm3 , respectively. The U
experimental investigation shows that geometry of the specimen εi ¼ (3)
E,k,n,s
has influence on the compressive and the tensile strength [8]. To
circumvent resonances, inertial and end friction effects in the types where εi represents strain at t ¼ iDt, U is the voltage value
of system at high rates, the dimensions of the specimens must be a measured from the wheatstone bridge, E denotes voltage of bridge
compromise between (1) maximizing the size of the specimen to box, k is sensitivity coefficient of semiconductor strain gauges, n is
have a comprehensive characterization of the materials, (2) magnification of high dynamic strain indicator and s is a value
appropriate L/D (length versus diameter) ratio to reduce the friction depends on bridge method, s ¼ 1/2 if the half-bridge has been
effects at two ends (the ratio is recommended to be 1 or more than adopted. According to one-dimensional wave propagation theory
1 [23]), (3) minimizing the size of the specimen to reduce inertial and two-wave method, integrating the reflected strain component
effect and non-uniform stress and strain distribution. Some of εr over time yields the average strain ε in a specimen.
fragments of brick and mortar were used to prepare the specimens
for quasi-static and SHPB tests, which were split from the frag- ðt
2C
ments of brick and mortar by electrodrill then sanded to reduce εðtÞ ¼  εr ðtÞdt (4)
l0
friction. Moreover, the Vaseline was smeared on end of specimens 0
to reduce end friction (Fig. 6 (b)). The specimens for quasi-static
tests were made into cubes with size of 10  10  10 mm and The strain rate and stress are obtained by
cylinders with 10 mm diameter, 5 mm high for SHPB tests. Because
of difficulty to split specimens from the fragments of brick and 2C
ε_ ðtÞ ¼  εr ðtÞ (5)
mortar the numbers of specimens were limited. The surfaces of l0
specimens are not perfectly smooth due to lack of appropriate
apparatuses, which introduces error to some extent. The results are A
indicative and as reference when determining the material pa- sðtÞ ¼ Eεt ðtÞ (6)
A0
rameters of simulation though.
Quasi-static tests were conducted on three samples of brick and where εr , εt represent reflection and transmission strain respec-
mortar using MTS machine, as shown in Fig. 6 (a). The displacement tively. C, E, A denotes the elastic wave velocity, elastic modulus and
controlled upper plate was moving with a constant velocity of cross-sectional area of compression bar respectively. A0, l0 are
0.8~0.9 mm/s. Displacement and load values were captured by cross-sectional area and length of specimen. Table 3 summarizes
digital controller unit. Displacement was measured by crosshead the static and dynamic test results. What should be stated is that
movement, which may introduce error because deformation of the strain rate is not averaged but selected when the maximum
loading unit itself and engagement between platen and sample also strength is reached, which is reasonable because the strain rate
contribute to platen displacement. SHPB tests were performed on varied instantly in experiment and the material properties are re-
flected by corresponding strain rate at the point of maximum
strength. All strain and stress values were transferred to true ones
by considering the cross-sectional changes, which is normally
negligible. Nevertheless this transverse change presenting in dy-
namic tests leads to inertial force and is responsible for partial
increment of DIF, which is characterized by Poisson’s ratio and like
an effect of refusing to be flattened, is structural and inevitable and
known as inertial effect. Therefore the ratio of specimen length to

Table 3
Compressive strengths of brick and mortar at low and high strain rates. (Q repre-
sents quasi-static or H is split Hopkinson bar; B for brick or M for mortar).

Test No. Strain rate /(s-1) Ultimate strength /MPa

QB1 0.009 3.0

QB2 0.009 10.9
QB3 0.009 8.7
QM1 0.005 11.3
QM2 0.003 8.1
QM3 0.006 5.2
HB1 991.1 27.3
HB3 719.6 23.7
HB4 703.2 16.3
Fig. 6. The specimens of (a) quasi-static test; (b) SHPB test. HM1 747.6 10.5
HM2 774.0 27.4
HM3 692.5 27.0
 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530 1519

diameter should be close to 1 because inertial effect is size

dependent [31]. The enhancement of compressive strength is
obvious for both brick and mortar in Table 3.
Benefited from research on concrete, the dynamic increase of
concrete-like material such as clay-fired brick, adobe and mortar
has been accepted resulting from viscoelastic effect normally
related to the hardened material by water and rate dependent crack
evolution [24]. Normal force will appear when two plates between
which is a thin viscous layer are separated or approaching, known
as Stefan effect. The force is proportional to the velocity of moving
plates. In dynamic tests the specimens are loaded rapidly, which
results in greater force between soil particles and reflects higher
strength in macro scale. Researchers have found heated or frozen
(attempt to exclude the effect of water) concrete is less rate sen-
sitive and saturated one is double in tensile strength, which sup-
ports the theory [32]. Another factor is evolution of crack. When Fig. 7. Three limit surfaces of RHT model [37].

subjected to slow loading the energy accumulated in specimens has

enough time to release in a path defined by minimum energy,
avoiding stiffer areas and connecting cracks that are pre-existent
and weaker areas. Instead in dynamic scenario the velocities of
particles are high and they will encounter coarse aggregates (stiffer I is the second-order identity tensor. The first term in Eq. (7) is
areas) but they are too fast and it is crucial that they have the en- illustrated in this section and description of EOS will be in Section
ergy to break them, which is illustrated from the fact from concrete 3.1.2.
experiments that there are increasing amount of broken coarse The strength model defines three limit surfaces in principle
aggregates along the facture surface with increasing tension load stress space (i.e. the failure surface, elastic limit surface, and re-
rate [32]. The similar phenomena were reproduced numerically in sidual friction surface, shown in Fig. 7) to describe the behavior of
Ref. [23,31]. The high confining stress solicited by impact condition material, which are defined using pressure p, effective stress seff
contributes to strain rate strengthening effect as well, which keeps and lode angle q(stress values are always positive).
the interfaces under overall loading so as to retard emerging of
cracks and causes deformation to occur both in weaker areas and 1 1
stronger aggregates. The conclusion is that the aggregates are p ¼  trðsÞ ¼  ðs1 þ s2 þ s3 Þ (8)
3 3
broken in dynamic situation to withstand much more stress and
dissipate the energy, which is not the case in static condition [33].
rffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3 1h i
3. Material model seff ¼ trðsÞ ¼ ðs1  s2 Þ2 þ ðs2  s3 Þ2 þ ðs3  s1 Þ2
2 2
(9)
The simulations were conducted with LS-DYNA hydrocode. The
class of hydrocode decomposes the dynamic problem into partial
differential equations which are solved by time integration based pffiffiffi detðsÞ 27detðsÞ
cos 3 q ¼ 3 6  3=2 ¼ (10)
on conservation of mass, momentum, energy coupled with state of tr s2 2s3eff
equation and constitutive model [34]. The state of equation deter-
mining the relation between density, pressure and energy while For a given stress state and strain rate, the failure surface is given
the constitutive model catering for behavior resulting from devia- as
toric mechanical loading, are specific for certain material. The
crucial mechanical phenomena such as transition from elastic to     
plastic regime and the onset of failure for material are included in sf p; q; ε_ p ¼ fc s*f p* ; Fr ε_ p ; p* R3 ðq; p* Þ (11)
constitutive model. RHT material model is applied in this paper
[13e16]. which consists of fc and two functions. The first demonstrates the
pressure dependence for principle stress conditions s1 ¼ s2 > s3
3.1. RHT model
and is expressed in terms of the failure stress s*f

RHT material model consists of equation of state and strength 8 "  1=n #n
characterization. The validation of RHT model for adobe, clay bricks >
> Fr A
>
> A p *
 þ 3p*  Fr
and lightweight adobe masonry was explored in Refs. [16e19,26]. >
> 3 Fr
>
>
In the following the RHT material model is reviewed. >
>
>
> 
>
> * *
>
< Fr fs þ 3p* 1  fs
3.1.1. Strength characterization Fr > 3p*  0
s*f ðp* ; Fr Þ ¼ Q1 Q1 (12)
The material model treats hydrostatic pressure p and deviatoric >
>
>
>  
stress portion s, which are both split from stress tensor s, sepa- >
> Fr fs* 1 f*
>
>  3p*  s * 0 > 3p* > 3p*t
rately [35,36]. Thus two groups of formula are developed: an EOS >
>
>
> Q1 Q2 Q1 ft
(see Section 3.1.2) to associate pressure with thermodynamical >
>
>
:
state variables density and internal energy; and a strength model to 0 3p*t > 3p*
deal with the deviatoric stress tensor
the superscript asterisk means the values are normalized with the
s ¼ sðp;ε; ε_ Þpðr;eÞI (7) compressive strength, in which Fr is dynamic increment factor and
1520 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530

Fr Q2 fs* ft* 2
p*t ¼   (13) bt ¼ (21)
3 Q1 ft*  Q2 fs* 20 þ fc
The formula allows steeper increment when strain rate exceeds
p*t is the normalized failure cutoff pressure. fs* and ft* represent the c=t c=t
ε_ p . With ε_ p ¼ 30s1 recommended by the CEB-FIP Model Code
normalized shear and tensile strength, respectively. A and n are
[38], gc=t is obtained in combination of continuity requirements
material constants, which are determined from triaxial compres-
sion tests. Q1 and Q2 in Eqs. (12) and (13) respect the effect of the
failure surface on the third invariant of the stress tensor. Fig. 8 log gc ¼ 6bc  0:492 log gt ¼ 7bt  0:492 (22)
shows the loge angle q dependence of failure stress. For a given
c=t
relatively low pressure the failure stress is increasing when tensile By default ε_ p is set 3  1022 s1 thus the second branch is left out
meridian rotates to the compressive meridian, in the process the and the bi-functional approach requires users’ modification.
fraction of tensile (shear) over compressive failure stress is The initial elastic limit surface is expressed with failure surface,
expressed as scaling function Fe and cap function Fc

  rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ffi
2 1  Q 2 cos q þ ð2Q  1Þ 4 1  Q 2 cos 2 q þ 5Q 2  4Q
r
R3 ðq; p* Þ ¼ ¼   (14)
rc 4 1  Q 2 cos 2 q þ ð1  2Q Þ2

The experimental observation demonstrated that the tensile     

p*
meridian gets close to the compressive meridian when pressure are sel p; q; ε_ p ¼ fc s*f ; FR ε_ p ; p* R3 ðq; p* ÞFe ðp* ÞFc ðp* Þ (23)
increasing, which is characterized by parameter Q Fe

0:5  Q ðp* Þ ¼ Q0 þ Bp*  1 (15) Fe is calculated interpolating between yield surface parameters gc*
and gt* determined by uniaxial material experiments.
Q is approaching 1 with increasing pressure, which leads to a
shape change from triangle in Fig. 8 to circle and reflects the brittle- 8 gc*
to-ductile transition of material. The Q1 and Q2 are given as >
>
>
>
> 3p*  Frc gc*
p  >
>
< 3p*  Frc gc*  * 
Q1 ¼ R 3 ; 0 (16)
6 Fe ðp* Þ ¼ gc*  gt  gc* Frc gc* > 3p*  Frt gt* ft*
>
> Frc gc* t * *
þ Fr gt ft
>
>
>
>
Q2 ¼ R3 ð0; p* Þ ¼ Q ðp* Þ (17) >
: Frt gt* ft* > 3p*
gt*
Fr characterizes the dependence of strain rate.
(24)
8 Frc
>
> The cap function is used to set the elastic limit to 0 when pressure
>
>
> 3p*  Frc
>
> goes beyond the pore crush pressure, through which deviatoric
< 3p*  Frc  t 
Fr ¼ Frc  Fr  Frc Frc > 3p*  Frt ft* (18)
>
>
> Frc þ Frt ft*
>
>
>
> Frt ft* > 3p*
:
Frt

with
8
!bc=t
ε_ p >
>
  f dynamic >
> c=t
c=t
< ε_ p  ε_ p
c=t
Frc;t ε_ p ¼ ¼ ε_ 0 (19)
fc=t >
>
>
: g p
> ffiffiffiffiffi c=t
c=t ε_ p ε_ p > ε_ p
3

Subscript or superscript c/t means compression or tension, hence

RHT model respects same formula for strain rate dependence in
compression and expansion but the bc and bt are derived slightly
differently from

4
bc ¼ (20)
20 þ 3fc
Fig. 8. Typical deviatoric plane of strength surfaces for low pressures [37].
 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530 1521

stress is consistent with inelastic volumetric stress built into EOS, as over particle velocity) curve distinctive for porous material,
shown in Fig. 7 and Eq. (25). compaction path is achieved by using variable a which is con-
structed as
8 0 p* > p*c
>
> 8 8 !N 99
>
> s ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi < < ==
>
<  *  pcomp  p
p  p*u 2 a ¼ max 1; min a0 ; 1 þ ða0  1Þ (34)
Fc ¼ 1  p*c > p*  p*u (25) : : pcomp  pel ;;
>
> p*c  p*u
>
>
>
:
1 p*u > p* in which pel is the current pore crush pressure and pcomp denotes
solid compaction pressure. N is compaction exponent. These pa-
The upper limit of the cap p*c equals pel in Eq. (34) and p*u is initial rameters are determined by dynamic inverse planar-plate impact
Frc gc* G* εp tests recommended in Ref. [35].
pressure expressed as p*u ¼ 3 þ fc
. Once the initial elastic sur-
face is reached, plastic strain initiates. The current elastic-plastic 3.2. Determination of material parameters for mortar and clay
yield surface is obtained interpolating between the initial yield
brick
surface (_ε*p ¼ 0) and failure surface (_ε*p ¼ 1), as P1 to P2 in Fig. 7.
     In this section the material parameters determination of brick
p*
sy p; q; ε_ p ¼ fc s*f ; Fr ε_ p ; p* R3 ðq; p* Þg (26) and mortar is illustrated respectively. The compressive strength is
g average of experimental data (the value that is too small has been
! excluded), which for brick and mortar are very close therefore they
εp sf ðp* Þð1  Fe Fc Þ   are both set as 10 MPa in parameter set for simulation. The pa-
ε*p ¼ min h ; 1 ; εhp ¼ ; g ¼ ε *
þ 1  ε*
Fe Fc rameters of material are shown in Table 4. The source of data is
εp g3G* p p
illustrated in Fig. 9.
(27) Brick. The density of brick is 1.8 g/cm3. The shear modulus is
determined through Young’s modulus from Ref. [6] and Poisson’s
G* is plastic shear modulus and calculated with original shear ratio n ¼ 0:15. The compressive strain rate exponent is derived
modulus of material multiplied by reduction factor x characterizing from ratio of dynamic compressive strength to static one. Only one
the hardening behavior. relation is activated by remaining default break compressive strain
rate as 3  1022 s1 . Utilizing Eq. (19) yields the compressive strain
1 dseff
G* ¼ xG ¼ (28) rate exponent.
3 dεp
lnðfcd =fc Þ
The residual surface is defined as bc ¼    (35)
ln ε_ p ε_ c0
* nf *
Af ðp Þ p >0
s*r ðp* Þ ¼ (29) Experimentally determined fcd and fc are 25.6 MPa and 10 MPa
0 p*  0
respectively. ε_ p and ε_ c0 are 991.07s-1 and 0.009s-1 respectively. Thus
Once the failure surface is reached the damage initiates. With compressive strain rate exponent bc ¼ 0:081. The tensile strain rate
further inelastic loading damage is accumulated and reflected by exponent and gt* are from Ref. [26]. The yield surface parameter gc*,
plastic strain. The damage parameter is defined as failure surface parameters and residual surface parameters are
εðp adopted from Refs. [16] but with minor adjustment. Residual sur-
dεp face parameter nf is smaller than that in Ref. [16], so as failure
D¼ (30) surface parameter n, which are based on the parameter set for
εfp
εhp standard concrete (C30/37). The normalized shear, tensile strength
and lode angle dependence factor are derived from Refs. [16] as
n
D o well. Moreover, default value of D1 which is equivalent to 0.04, is
εfp ¼ max D1 p*  ð1  DÞp*t 2 ; εm
p (31) not used and the value 0.015 in Ref. [39] is chosen. The influence of
this parameter and also residual surface parameters is illustrated in
The resulting damage surface is interpolated between failure
Ref. [39,40]. The parameters of equation of state are obtained from
surface and residual surface from D ¼ 0 to D ¼ 1 (as P2 to P3 in
default material parameters for the standard concrete (C30/37) by
Fig. 7).
scaling those figures with the ratio of respective porous densities
(The standard concrete possesses the density of 2.31 g/cm3). The
3.1.2. Equation of state
procedure of determining parameters of EOS is adopted from
In the RHT model, the pressure is described by Mie-Gruneisen
Ref. [19]. In addition, initial compaction pressure pel slightly below
form through a polynomial Hugoniot curve and p-a compaction
uniaxial compressive strength is selected, as is for standard con-
relation. For compression (h > 0) or expansion (h < 0), it is given as
crete. The solid compaction pressure pcomp is calculated by retain-
ing the ratio of solid compaction pressure and initial compaction
1 ðB0 þ B1 hÞare þ A1 h þ A2 h2 þ A3 h3 h>0
p¼ (32) pressure for standard concrete. The volume faction porosity of brick
a B0 are þ T1 h þ T2 h2 h<0 24% is chosen according to Ref. [30], hence initial porosity of brick is
1.32. The compaction exponent remains unchanged.
together with
Mortar. Some parameters of brick are benchmarks of those of
ar mortar. The shear modulus of mortar is set approximately half of
hðrÞ ¼ 1 (33)
a0 r0 that of brick based on engineering assumption. For mortar, fcd and fc
are 27.0 MPa and 10 MPa respectively. ε_ p and ε_ c0 are 692.53 s-1 and
in which Ai, Bi, Ti are material constants. a0 denotes the initial 0.005 s-1 respectively. Hence, the compressive strain rate exponent
porosity and equals to rmatrix =r0 . To reflect a drop in us-up (shock bc ¼ 0:084 is determined. The initial porosity of mortar is retained
1522 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530

Table 4
Material parameters applied for mortar and for clay brick arranged as format in LS-DYNA. The distinct data of clay brick and mortar for identical entries are divided by forward
slash symbol in arrangement of “brick/mortar”. If there is no forward slash symbol they share the parameters. Back flash symbols are used to separate different parameters for
one material.

RHT model parameters for mortar/brick

Equation of State Strength

Porous density r0 1.38/1.8(g,cm-3) Shear strength fs* (fs/fc) 0.07/0.06

Parameter B0 1.22 Shear modulus G 2.8/1.13(GPa)
Parameter B1 1.22 Compressive strength (fc) 10(MPa)
Parameter T1 21/27(GPa) Tensile strength ft* (ft/fc) 0.04/0.05
Parameter T2 0 Failure surface parameter A 1.6/1.67736
Parameter A1 21/27(GPa) Failure surface parameter n 0.61
Parameter A2 23.64/31(GPa) Lode angle dependence factor Q0 0.58/0.6
Parameter A3 5.38/7(GPa) Loge angle dependence factor B 0
Initial compaction pressure Pel 9.5/9.9(MPa) Shear modulus reduction factor x 0.5/0.4
Compaction exponent N 3 Reference [Comp.\Tens.] strain rate ε_ 0
½c\t 0.009/0.005(s-1)
Solid compaction pressure Pcomp 300/400(MPa) Break [Comp.\Tens.] strain rate
½c\t
ε_ p 3  1022 ðs1 Þ
Initial porosity a0 1.32/1.184 [Comp.\Tens.] strain rate exponent b½c\t [0.081\0.059684]/[0.084\0.09]
Failure *
[Comp.\Tens.] yield surface parameter g½c\t [0.53\0.2222]/[0.61\0.7]
Damage constant D1 0.015/0.018 Residual surface parameter Af 1.67736
Damage constant D2 1.0 Residual surface parameter nf 0.7/0.745
Minimum strain to failure εm
p 0.01 Friction bullet and mortar/brick 0.15
Erosion strain 2.0 Friction interface between Area II & III 0.30

Fig. 9. Acquisition of material parameters.

from standard concrete material parameters because they are 3.3. Material model and parameters for projectile
similar material. Moreover, other parameters of mortar are ob-
tained by using similar methods as brick and based on the standard 12.7 mm armor-piercing explosive incendiary projectile consists
concrete parameters. Some minor adjustments are included of steel core, lead sheath, incendiary agent and jacket. The influence
considering the experiment data or different characterization be- of incendiary agent on penetration is very small, which is validated
tween them though overall similarity is assumed of their through experiments [41]. Thus it is omitted in the simulation. The
properties. numerical model is shown in Fig. 10. The material of steel core, lead
sheath and jacket are T12A stool steel, Pb and F11 copper clad steel,
all of which are simulated using simplified Johnson-Cook material
model. Johnson-Cook material model is capable of capturing
 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530 1523

Fig. 10. The geometric model and mesh for 24 wall and projectile. (a): The material location is illustrated by respective color. Area I of 24 wall has finest mesh, area II takes the
second place and area III has the coarsest; (b): from outside to inside are F11, Pb and T12A.

behavior of metal subjected to large deformation, high strain rate included in the finite element model. As shown in Fig. 10 (a), the
and high temperature. The thermal effects and damage are ignored masonry wall is divided into three areas I, II and III. Area I and II
in the simplified Johnson-Cook, hence the maximum stress is share common nodes but area II is discretized with coarser mesh,
limited to compensate thermal softening. The damage here is not of which is as a transition from Area I to Area III. The interaction be-
paramount importance on account of completeness of bullets after tween Area II and Area III is reflected by automatic tiebreak contact
penetration (Fig. 5 (b)). The model is suitable for situation where algorithm, which is chosen because of high compatibility of
the strain rates change over a large range. The constitutive formula different mesh resolution on interface being contacted. The exis-
is given as tence of Area II contributes to avoiding much too sharp mesh
  inconsistency between Area I and Area III and itself is helpful to
sy ¼ A þ Bεp
n
ð1 þ c ln_ε* Þ (36) reduce computing time with gradually larger element size. Within
three Areas sharing common nodes is employed to represent
interaction of brick and mortar. The three components of 12.7 mm
where sy is the flow stress; A is the yield stress under reference armor-piercing explosive incendiary projectile are constructed
strain rate ε_ 0 ; B and n are parameters characterizing strain hard- individually sharing common nodes. The gravity is loaded in global
ening of material; c is the material constant reflecting the strain model.
rate dependence of the material; εp denotes the effective plastic
strain and is obtained by removing the elastic strain from the total
4.1. Contact algorithm applied in simulation
strain; ε_ is the strain rate; and ε_ * is the normalized effective strain
rate calculated from ε_ =ε0 , where ε0 ¼ 1=s. The material parameters Choosing appropriate contact algorithm is crucial to obtain
of projectile are adopted from Refs. [41], shown in Table 5. more accurate and realistic results for simulation. The masonry
consists of brick and mortar, interaction between which is very
4. Simulation complicated. Sharing common nodes and utilizing contact algo-
rithm are combined to solve the problem. The contact between
The simulations are performed using hydrocode LS-DYNA. To Area II and Area III are simulated with automatic TIEBREAK surface-
keep minimum boundary inconsistency with reality, full 3D to-surface contact algorithm, which allows the transmission of both
detailed finite element model consisting of two different material is tensile and compressive forces resulting in a TIE. The separation of
established. The base of masonry wall is a block of wood and also the slave node from the master is resisted by a linear contact spring
for both tensile and compressive forces until failure after which the
tensile coupling is removed. Post failure in all TIEBREAK contacts
Table 5 allows the node to interact with the segment as in traditional
Simplified Johnson-Cook constitutive material parameters used in simulation for
compression only contacts. The debonding of contact is assumed to
bullet components ([41]; http://www.varmintal.com/aengr.htm). The parameters
are arranged in LS-DYNA format. be governed by failure criterion [36].

Material T12A stool steel Pb F11 copper clad steel    

jsn j 2 jss j 2
-3
Density /(g,cm ) 7.85 11.349 7.92 þ 1 (38)
NFLS SFLS
Young’s modulus /GPa 218 13.79 114.38
Poisson’s ratio 0.33 0.42 0.33
A/MPa 1540 3.4856 300 where sn and ss are normal and shear stress on the interface,
B/MPa 477 21.089 275 respectively; NFLS and SFLS are normal and shear strength of the
n 0.16 0.132066 0.15 interface, which are calculated from minimum of compressive and
C 0 0 0.022 shear strength of brick and mortar due to weaker strength of
Failure plastic strain default 1.9214 1.5
interface (see Section 4.3.2). The eroding surface-to-surface contact
1524 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530

Fig. 11. Damage configuration of 24wall (test 4): (a) front view; (b) rear view.

algorithm is selected to simulate the interaction between pene- structure and lower demand for computer power, three different
trator and target, namely three components of bullet and masonry mesh resolutions are investigated using 24 wall compared to test 4
wall consisting of brick and mortar, which is crucial to define first. The steel core of bullet and Area I which are directly subjected
erosion of elements reaching material failure and allows the to impact are discretized with the finest mesh varying with
remaining interior elements to continue to contact after the failure different mesh resolutions. The Area III remains the element edge
of outer elements during impact process. The friction coefficient of size same as thickness of mortar with the lead sheath and jacket
components of bullets with brick and mortar is considered as 0.15 keeping element edge size as 0.1 cm on account of their small
[16], and value 0.3 is used between Area II and Area III. thickness. It should be stated that the bullet in simulation impacts

the target with estimated pitch 2:0375 though the bullets were
fired with zero attack angle, which is based on observation from
4.2. Investigation on mesh resolution with 24 wall (test 4)
high-speed camera that the masonry wall in test 4 is placed aslant
(Fig. 4(b)). Fig. 11 shows the damage picture of 24 wall in front view
The hydrocode simulation in this section is performed with
and rear view. The results of simulation are listed in Table 6 and
parameter set, finite element model and contact algorithm pre-
Fig. 12. For the sake of convenience in writing, simulations using
sented above. The RHT model shows mesh resolution dependency,
minimum element edge size 2, 1.8 and 1.5 mm are denoted by S2,
which is illustrated in Ref. [13,16]. Considering relative simpler

Table 6
The exit velocities and damage using three different mesh resolutions.
 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530 1525

Fig. 12. The upward deviation of trajectory.

Fig. 13. Damage configuration of 37wall (test 1): (a) front view; (b) rear view.

S1.8 and S1.5. not the case because the exit velocity experiences a drop from S1.8
The exit velocities of S2, S1.8 and S1.5 are 146, 161and 158 m/s to S1.5. This phenomenon will be explained later. The damage
respectively compared to 139.1 m/s in experiment. The RHT ma- configuration is analyzed first. The front and rear views with
terial model displays softer characteristic when the mesh is finer damage scalar are shown in Table 6. Comparing simulation results
[13]. Therefore greater exit velocity and less energy loss of bullet are to Fig. 11 shows a similar failure mode to experiment in case of front
predictable if higher mesh resolution is adopted. Nonetheless, it is view of masonry wall. The scope of crater is comparable. In addition

Fig. 14. Damage configuration of 37wall (test 2): (a) front view; (b) rear view.
1526 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530

to the overall similarity in simulation using different mesh reso- snapshots are not in an identical plane due to lateral deflection of
lutions, with increase of mesh resolution the delicate cracks are bullet. The trajectories show a small increase of slope when mesh
appearing. The depth of crater is not measured in experiment thus resolutions becomes higher, as S2 to S1.8 then S1.8 to S1.5. In RHT
it is not examined. As for rear view, an obvious deviation is material model, tensile failure will occur when the tensile stress
observed in Fig. 11. The hit position at Fig. 11 (a) is at mortar but the falls under a specific value, meanwhile minimum strain to failure is
exit position of bullet occurs at middle of the upper brick, which predefined in tension, which demonstrates failure stress and strain
means that the bullet experienced an upward displacement. The are independent from element size therefore the fracture energy is
front and rear views are only a reflection of initial and final states of mesh dependent [37]. The phenomenon can be illustrated by
bullet, however, the detailed movement during the process is un-
known in experiment. To investigate the trajectory, side view of 2  GF
et;fmin ¼ (39)
masonry wall is obtained by removing corresponding part. Fig. 12 st  Leq
presents the initial and final states of bullet in simulations, these

Table 7
Results of simulation of 37 wall (test 1 and 2).
 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530 1527

and mortar are arranged by hand. The size of brick and mortar
varies in reality, which is hard to synchronize with in simulation.
The precise hit position is not easy to locate due to acute damage of
masonry wall. Therefore the deviation of bullet in reality is elusive
to some extent and of many possibility especially for clay brick
masonry wall that is not well standardized and consists of site
dependent material [24]. Considering these facts, the agreement of
simulation to experiment is remarkable.

4.3. Simulation of 37 wall

Fig. 15. The trajectory of bullet.
The finite element model with minimum element edge size
1.5 mm presented above and parameter set in Table 4 are used to
where et,fmin and st is tensile failure strain and stress respectively.
simulate experiments of 37 wall. The charge of test 1 and test 2 are
Leq denotes characteristic length of element and GF is fracture en-
16 g and their initial velocities are close (Table 2). Their hit positions
ergy. For given failure stress and strain the fracture energy of bigger
are at mortar layer located at center of the front of wall (Fig. 13 (a)
element will be higher than that of smaller one because charac-
and Fig. 14 (a)). Nonetheless, their results are different. In contrast
teristic length is highly dependent on size of element, which can
with perforation of test 1, the masonry wall in test 2 is not perfo-
partially explain why RHT model shows mesh dependency. As
 rated. The comparison of Figs. 13 (a) and Fig. 14 (a) reveals that the
shown in Fig. 12, with pitch of 2:0375 and finer mesh resolution
damage configuration of front of masonry walls is distinctive,
the exit position is getting closer to the middle of the upper brick,
which can partially attribute to variation of material properties in
which means more deflection of the bullet. More deflection of
different masonry walls. It appears that the masonry wall in test 1 is
bullet make it lose more energy when travelling through the ma-
more brittle and of weaker strength thus the impact energy can
sonry wall because of more contact area with target. This effect
concentrate on penetration efficiently by pulverizing the brick and
coupled with less energy loss of bullet introduced by mesh reso-
mortar in front of bullet and cohesion of the area being impacted
lution results in a drop in S1.5. The dominant factor is less energy
and its periphery can be broken rapidly, which result in smaller
loss of bullet introduced by mesh resolution from S2 to S1.8, hence
crater diameter and perforation. Conversely, the material proper-
the exit velocity increases a bit. However from S1.8 to S1.5, more
ties in test 2 seem stronger though they are from same lot, however,
deflection introduced by mesh resolution takes the power, hence
small quantitative can cause a qualitative change. The comparison
the velocity drops a little. There are more tiny visible debris behind
of simulation with experiment includes damage configuration and
the masonry wall when mesh resolution is higher. Though it is far
trajectory of bullet.
behind the phenomenon shown in Fig. 4(a) it makes simulations
more realistic. The higher mesh resolution is deemed necessary.
4.3.1. Test 1 and 2
Nonetheless, only three cases of mesh resolution are included
Only one simulation is carried out because of the high similarity
because full 3D finite element model is very expensive and on ac-
of test 1 and test 2. The mesh of wall is graded similarly as shown in
count of computer power limitation. The finest mesh resolution
Fig. 10. The initial velocity in simulation is 842.4 m/s. As shown in
herein is kept in following simulations. This phenomenon that the
Table 7, the size of crater of front view of masonry wall is compa-
numerical results (deviation of bullet) are more similar to what was
rable to that in Fig. 14 (a) and a hairline crack is observed in the top
observed experimentally seem to be obtained by applying higher
right-hand corner, which is observed in Fig. 14 (a) as well.
resolution should be taken care of, what is demonstrated from
But the tiny cracks in bottom right-hand corner in simulation
which is the influence of mesh resolution on RHT model. The
are not observed in experiment. Concerning trajectory plot in
phenomenon does not necessarily mean that the numerical simu-
Table 7 bullet is not shown because of lateral deflection of bullet,
lation is more accurate because there are many other factors, two of
which makes it difficult to show more complete trajectory and
which are small yaw angles of projectiles and slight tilt of brick wall
meanwhile bullet. The rear view of experiments in Figs. 13 (b) and
undetected in experiment thus not being included in simulation
Fig. 14 (b) both demonstrate pulverization of big block of brick
but having quite an influence on trajectory of bullet. Besides, the
though the position is not exactly same. The hit positions of test 1
deviation of bullet is clear to be affected by some random factors,
and test 2 are very similar at mortar but the exit position in test 1 is
such as irregular voids in masonry walls, variation in material
upper brick and that of test 2 is nether brick, which seems to be
properties and uneven size distribution of mortar because brick
symmetrical. The hairline cracks in Figs. 13 (b) and Fig. 14 (b) are

Fig. 16. Damage configuration of 37wall (test 3): (a) front view; (b) rear view.
1528 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530

more or less captured by simulation. The pulverization of brick is (a). The crater diameter becomes relatively smaller and the depth of
reflected by outline of damage in rear view in Table 7 but un- crater is deeper compared to test 2 and 4. The damage model of rear
derestimates the experimental results. The shape of trajectory is view, i.e. pulverization of big block of brick, is validated once more,
like S, which is close to that in Fig. 15, along which the energy of which attributes to the stress state of brick wall during impact. The
bullet will be consumed more quickly than a straight one. rear of masonry wall is subjected to tensile stress before bullet
arrives at the rear of masonry wall because of spreading of shock
4.3.2. Test 3 wave. The failure emerges on the interface of brick and mortar
In test 3, the hit velocity in experiment is 768.5 m/s. The damage because the spreading velocities of stress wave in two material are
configuration in experiment is presented in Fig. 16. The hit position not same, which causes complex stress state on interface and the
is interface of brick and mortar, which can be observed from Fig. 16 strength of interface is much weaker (we can know the fact from

Table 8
Results of simulation of 37 wall (test 3).
 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530 1529

Figs. 11, Fig.13, Figs. 14 and 16 that all the damage configurations are Acknowledgement
with hairline cracks on the interface of brick and mortar). Then the
brick is impacted by concentrated power from relatively small The work presented in this paper is funded by Opening Project
bullet. Combined with brittle property of brick the phenomenon of Science and Technology on Transient Impact Laboratory (Grant
occurs. The results of simulation are shown in Table 8. The simu- No. 614260601010517). This support and the good cooperation are
lation predicts approximate damage configuration of front view to gratefully acknowledged.
experiment. Regarding rear view, the pulverization of brick is not
captured exactly. It seems that the simulation overestimates the
References
strength of the material. However both in front and rear views, the
hairline cracks in simulation, which are similar to experiment, are [1] Kumavat HR. An experimental investigation of mechanical properties in clay
outlined along the edge of brick or mortar similar in experiment. brick masonry by partial replacement of fine aggregate with clay brick waste.
J Inst Eng: Ser A 2016;97(3):199e204. https://doi.org/10.1007/s40030-016-
Fig. 16 demonstrates that bullet has an upward deflection, which is
0178-7.
not captured by simulation. The trajectory of simulation shows [2] Torgal FP, Jalali S. Masonry units. In: Pacheco Torgal F, Jalali S, editors. Eco-
downward deflection in Table 8. Still considering the reason stating efficient construction and building materials. Springer London; 2011.
in end of Section 4.2, the numerical results are acceptable. p. 131e42. https://doi.org/10.1007/978-0-85729-892-8_6.
[3] Hao H, Tarasov B. Experimental study of dynamic material properties of clay
brick and mortar at different strain rates. Aust J Struct Eng 2008;8:117e32.
https://doi.org/10.1080/13287982.2008.11464992.
[4] Fremont-Barnes G. A history of modern urban operations. Switzerland:
5. Conclusions Springer Nature; 2020. p. 58.
[5] Lin SL. Experimental and simulation of projectile penetrating concrete target.
Master’s thesis of Beijing University of Technology; 2016.
The response of construction consisting of clay-fired brick and [6] Zhang X, Chiu Y-W, Hao H, Hsieh A, Salter N, Cui J. Dynamic compressive
mortar subjected to quasi-static or dynamic loading is important to material properties of clay bricks at different strain rates. Construct Build
ensure safety in some historic sites, public buildings and so on. To Mater 2018;192:754e67. https://doi.org/10.1016/j.conbuildmat.2018.10.150.
[7] Pereira Jo~ao M, Lourenço Paulo B. Experimental characterization of masonry
this end, the penetration experiments were carried out using four and masonry components at high strain rates. J Mater Civ Eng 2017;29(2):
clay brick masonry walls with two different patterns as target and 04016223. https://doi.org/10.1061/(ASCE)MT.1943-5533.0001755.
12.7 mm armor-piercing explosive incendiary bullet as penetrator. [8] Gebbeken N, Linse T, Araujo T. Masonry under dynamic actions - experimental
investigations, material modeling and numerical investigations. Adv. Protect.
The experimental investigation shows that the average drag force
Struct. Res. 2012:131e62. https://doi.org/10.1201/b12768-6.
in the process of penetration is constant, which is basic assumption [9] Li Z, Chen L, Fang Q, et al. Experimental and numerical study of unreinforced
of some theoretical ballistic formula. Then static and dynamic clay brick masonry walls subjected to vented gas explosions. Int J Impact Eng
2017;104:107e26. https://doi.org/10.1016/j.ijimpeng.2017.02.002.
material tests were conducted on specimens of brick and mortar,
[10] Govindjee S, Kay GJ, Simo JC. Anisotropic modelling and numerical simulation
which demonstrates the clay-fired brick and mortar show an in- of brittle damage in concrete. Int J Numer Methods Eng 1995;38.
crease of compressive strength at high strain rate. [11] Meyer CS. Development of brick and mortar material parameters for nu-
The simulations are performed using LS-DYNA hydrocode. The merical simulations. In: Proulx T, editor. Dynamic behavior of materials, ume
1. New York: Springer; 2011. p. 351e9.
RHT model incorporating strength model in which pressure, lode [12] Holmquist TJ, Johnson GR, Cook WH. A computational constitutive model for
angle dependency of material and damage are described in com- concrete subjected to large strains, high strain rates, and high pressures. In:
bination with porous equation of state is adopted. With assistance Proceedings of the fourteenth international symposium on ballistics; 1993.
[13] Riedel W, Thoma K, Hiermaier S, et al. Penetration of reinforced concrete by
of some references and experimental results the parameter set is BETAB- 500-numerical analysis using a new macroscopic concrete model for
obtained. To keep minimum boundary inconsistency with reality, hydrocodes. In: Proceedings of 9th international symposium on interaction of
full 3D detailed finite element model consisting of two different the effect of munitions with structures; 1999. p. 315e22. Berlin.
[14] Riedel W. 10 years RHT: a review of concrete modelling and hydrocode
material is established. To deal with the problem of computational application. In: Hiermaier S, editor. Predictive modeling of dynamic processes
time usage of large-scale model, sharing common nodes and e a tribute to professor Klaus Thoma. Springer; 2009.
employing automatic tiebreak contact are combined. Using devel- [15] Riedel W, Kawai N, Kondo K. Numerical assessment for impact strength
measurements in concrete materials. Int J Impact Eng 2009;36:283e93.
oped parameter set and finite element model, the influence of
[16] Sauer C, Heine A, Riedel W. Developing a validated hydrocode model for
mesh resolution on RHT model is investigated with test 4, which adobe under impact loading. Int J Impact Eng 2017;104:164e76. https://
reveals that deviation of bullet in simulation is affected by less doi.org/10.1016/j.ijimpeng.2017.01.019.
[17] Sauer C, Heine A, Riedel W. Comprehensive study of projectile impact on
energy loss and more reflection of bullet introduced by mesh res-
lightweight adobe masonry. Int J Impact Eng 2019;125:56e62. https://
olution. The experimental observation demonstrates: (1) much doi.org/10.1016/j.ijimpeng.2018.10.010.
weaker strength of the interface of brick and mortar; (2) damage [18] Sauer C, Heine A, Weber KE, Riedel W. Stability of tungsten projectiles
model of crater approximating to circle in front of clay brick ma- penetrating adobe masonry - combined experimental and numerical analysis.
Int J Impact Eng 2017;109:67e77. https://doi.org/10.1016/
sonry; (3) damage model of pulverization of big block of brick in j.ijimpeng.2017.06.001.
rear. The numerical results reproduce the experimental phenom- [19] Sauer C, Bagusat F, Heine A, Riedel W. Shock response of lightweight Adobe
ena with respect to exit velocity, damage configuration and tra- masonry. J Dyn Behav Mater 2018;4(2):231e43. https://doi.org/10.1007/
s40870-018-0151-9.
jectory acceptably. The way of constructing the finite element [20] Wei X, Hao H. Numerical derivation of homogenized dynamic masonry ma-
model and the material model are validated suitable for this class of terial properties with strain rate effects. Int J Impact Eng 2009;36(3):522e36.
construction. We perceive our work contributes to the under- https://doi.org/10.1016/j.ijimpeng.2008.02.005.
[21] Burnett S, Gilbert M, Molyneaux T, Tyas A, Hobbs B, Beattie G. The response of
standing of clay-fired brick wall that is a complicated structure and masonry joints to dynamic tensile loading. Mater Struct 2007;40(5):517e27.
relevant exploration in experiment and simulation is of value for https://doi.org/10.1617/s11527-006-9160-6.
future investigation. [22] Hao Y, Hao H, Zhang XH. Numerical analysis of concrete material properties at
high strain rate under direct tension. Int J Impact Eng 2012;39(1):51e62.
https://doi.org/10.1016/j.ijimpeng.2011.08.006.
[23] Hao Y, Hao H, Li ZX. Influence of end friction confinement on impact tests of
concrete material at high strain rate. Int J Impact Eng 2013;60:82e106.
Declaration of competing interest https://doi.org/10.1016/j.ijimpeng.2013.04.008.
[24] Piani L, Weerheijm J, Peroni M, et al. Dynamic behaviour of adobe bricks in
The authors declare that they have no known competing compression: the role of fibres and water content at various loading rates.
Construct Build Mater 2020;230. https://doi.org/10.1016/j.con-
financial interests or personal relationships that could have buildmat.2019.117038. UNSP 117038.
appeared to influence the work reported in this paper. [25] Medeiros P, Vasconcelos G, Lourenco PB, Gouveia J. Numerical modelling of
1530 C.-z. Wang, A.-j. Chen, Z.-q. Li et al. / Defence Technology 17 (2021) 1514e1530

non-confined and confined masonry walls. Construct Build Mater Apr [34] Benson DJ. Computational methods in langrangian and eulerian hydrocodes.
2013;41:968e76. Comput Methods Appl Mech Eng 1992;99(2e3):235e94. https://doi.org/
[26] Michaloudis G, Gebbeken N. Modeling masonry walls under far-field and 10.1016/0045-7825(92)90042-I.
contact detonations. Int J Impact Eng 2019;123:84e97. https://doi.org/ [35] Grunwald C, Schaufelberger B, Stolz A, Riedel W, Borrvall T. A general concrete
10.1016/j.ijimpeng.2018.09.019. model in hydrocodes: verification and validation of the Riedel-Hiermaier-
[27] Su Y, Wu C, Griffth MC. Modelling of the bond-slip behavior in FRP reinforced Thoma model in LS-DYNA. Int J Protect Struct 2017;8(1):58e85. https://
masonry. Construct Build Mater 2011;25(1):328e34. https://doi.org/10.1016/ doi.org/10.1177/2041419617695977.
j.conbuildmat.2010.06.021. [36] Ls-Dyna. Keyword user’s manual. Livermore Software Technology Corporation
[28] Ministry of Construction of the PR China. Code for design of masonry struc- (LSTC); 2018.
tures (GB5101-2017). Beijing, China: China Architecture and Building Press; [37] Hu G, Wu J, Li L. Advanced concrete model in hydrocode to simulate concrete
2017. structures under blast loading. In: Advances in civil engineering; 2016.
[29] Robins B. New principle of gunnery. 1974. p. 7540151. https://doi.org/10.1155/2016/7540151. Published online.
[30] Hall C, Hamilton A. Porosityedensity relations in stone and brick materials. [38] Comite Euro-International du Be ton. CEB-FIP model code. Lausanne: Thomas
Mater Struct 2015;48(5):1265e71. https://doi.org/10.1617/s11527-013- Telford; 1990.
0231-1. [39] Tu Z, Lu Y. Evaluation of typical concrete material models used in hydrocodes
[31] Hao Y, Hao H, Zhang XH. Numerical analysis of concrete material properties at for high dynamic response simulations. Int J Impact Eng 2009;36(1):132e46.
high strain rate under direct tension. Int J Impact Eng 2012;39(1):51e62. https://doi.org/10.1016/j.ijimpeng.2007.12.010.
https://doi.org/10.1016/j.ijimpeng.2011.08.006. [40] Abdel-Kader M. Modified settings of concrete parameters in RHT model for
[32] Yan DM, Lin G. Dynamic properties of concrete in direct tension. Cement predicting the response of concrete panels to impact. Int J Impact Eng
Concr Res 2006;36(7):1371e8. https://doi.org/10.1016/ 2019;132:103312. https://doi.org/10.1016/j.ijimpeng.2019.06.001.
j.cemconres.2006.03.003. [41] Kuo B, Xianfeng Z, Mengting T, Beibei C, Haiyang W. Ballistic test and nu-
[33] Grote DL, Park SW, Zhou M. Dynamic behavior of concrete at high strain rates merical simulation on penetration of a boron-carbide-ceramic composite
and pressures: I. experimental characterization. Int J Impact Eng 2001;25(9): target by a bullet. Explos Shock Waves 2019;39(12):1001e455 (2019)39:
869e86. https://doi.org/10.1016/S0734-743X(01)00020-3. 12<ZDZJTH>2.0.TX;2-#.