International Journal of Impact Engineering 85 (2015) 5e19

Contents lists available at ScienceDirect

International Journal of Impact Engineering

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

Damage and risk assessment for reinforced concrete wall panels

subjected to explosive blast loading
Yufeng Shi*, Mark G. Stewart
Centre for Infrastructure Performance and Reliability, The University of Newcastle, New South Wales, 2308, Australia

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 13 May 2014
Received in revised form
19 April 2015
Accepted 4 June 2015
Available online 14 June 2015

The structural integrity of reinforced concrete (RC) structures in blast events is important for critical
facilities. In this paper, a structural reliability analysis is conducted to predict the damage and risk
reduction for RC wall panels subjected to explosive blast loading. Due to considerable uncertainties
associated with material properties, dimensions, structural response, blast loading, as well as expected
damage, probabilistic methods are used in quantifying the probability of damage for conventional and
blast-resistant RC precast cladding wall panels by incorporating spatial and non-spatial variables. The
variability of blast loading is also taken into consideration. Monte Carlo simulation and numerical
methods are utilized to predict damage of RC wall panels subject to various threat scenarios, based on a
physics-based computer programme LS-DYNA to estimate maximum support rotations. It was found that
spatial variability of concrete compressive strength and concrete cover has little effects on the structural
reliability for precast concrete panels, and the blast-resistant wall has 5%e100% lower probability of
hazardous failure than the corresponding value for a conventional wall.
2015 Elsevier Ltd. All rights reserved.

Keywords:
Probability
Reliability analysis
Explosive blast
Reinforced concrete
Terrorism

1. Introduction
Reinforced concrete (RC) is a principal construction material
used for civilian buildings and military constructions. These
structures might be subjected to explosive blast loading from
terrorist attacks, military ordnance, or other sources of explosions.
This may lead to severe damage to RC structures and result in
tremendous casualties and property loss. In recent years, signicant
worldwide efforts have been devoted to research on the performance of RC structures subjected to explosive blast loading.
RC precast wall panels are increasingly being used for residential
and commercial buildings, and they are vulnerable to explosive
blast loading. In recent years, much research has been conducted
on the dynamic response of RC wall panels to blast loading.
Remennikov and Kong [1] developed a numerical model to predict
the exural response of steel-concrete-steel panels subjected to
blast and high-speed vehicle impact loading. Lin et al. [2] investigated the effects of charge weight, standoff distance, panel thickness, and reinforcement ratio on the blast resistance of RC wall
panels by using LS-DYNA. Gareld et al. [3] conducted experiments
for wall panels constructed by normal concrete and ber reinforced

* Corresponding author. Tel.: 61 02 49216156.

E-mail address: yufeng.shi@uon.edu.au (Y. Shi).
http://dx.doi.org/10.1016/j.ijimpeng.2015.06.003
0734-743X/ 2015 Elsevier Ltd. All rights reserved.

concrete with a wide range of construction details under blast

loading. Pan and Watson [4] reported the interaction between RC
cladding panels and xings under blast loading by using DYNA3D.
However, the deterministic research has not incorporated
considerable uncertainties associated with material properties,
structural dimensions, blast loading, as well as expected damage.
Therefore, probabilistic and reliability theory methods are useful in
quantifying the probability of damage. Only very few probabilistic
and reliability analyses have been carried out for RC structures
subjected to explosive blast loading (eg. Refs. [5e11]). This is in
contrast to the approach that has been used very widely and successfully for other man-made and natural hazards (eg. [12]).
Moreover, many of these reliability analyses used simplied probabilistic blast loading models and nite element models, and all
assumed homogeneous material and dimensional properties. It is
acknowledged that material and dimensional properties of a concrete structure will not be homogeneous due to the spatial variability of work practices and environment. Shi et al. [13] developed
a procedure to assess the reliability of RC columns subjected to
explosive blast loading, considering spatial random variables.
Hence, a two-dimensional spatially variable reliability analysis is
developed to predict the likelihood and extent of damage for RC
wall panels subjected to blast loading. This model will consider the
random spatial variability of concrete compressive strength and
concrete cover.

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

In this paper, Monte Carlo simulation iterations and numerical

methods are used to estimate the damage of RC wall panels subjected to blast loading, based on the explicit nite element method
software LS-DYNA. The prediction of damage is based on the support rotation of the structure. The reliability analyses of both nonspatial and spatial models are compared. The variability of blast
loading is also taken into consideration. A terrorist Vehicle Borne
Improvised Explosive Device (VBIED) scenario and a military
Collateral Damage Estimation (CDE) scenario are considered to
reected the concerns and reality of current threats [14]. Expanded
to include the probabilistic blast loading model developed by
Netherton and Stewart [14] is the effect of charge shape. Hence, the
blast scenarios considered herein are 50 kg (small VBIED), 116 kg
(car-size) and 1000 kg (truck-size) of home-made Ammonium Nitrate Fuel Oil (ANFO), Mk82 General Purpose (GP) 500 lb bomb
(89 kg Tritonal), and Mk83 GP 1000 lb bomb (202 kg Tritonal)
detonated at various stand-off distance from a typical RC precast
wall panel. The shape of charge assumed herein is a cylindrical
explosive with aspect ratio of six to one. The TNT equivalent of
89 kg Tritonal and 116 kg ANFO are deterministically similar in peak
reected pressure. Additionally, the performance of conventional
and blast-resistant wall panels will also be compared. The reliability analysis allows Blast Reliability Curves (BRCs) to be generated - these represent damage and collapse risks as a function of
explosive mass and stand-off.
2. Finite element analysis
The RC precast wall panel is shown in Figs. 1 and 2, and Table 1
shows the design material and dimensional properties for conventional and blast-resistant wall panels. The hydro code LS-DYNA
is utilized to analyse the wall panel damage to blast loading. Eightnode solid hexahedron elements of 50 mm are used to represent
the concrete, and the Lagrangian formulation is employed in the
analysis. In the Lagrangian formulation, the nodes of the numerical
mesh are attached to the material, and they move and deform with
the material. Therefore, there is no material transportation between elements, and the boundary conditions and interface between materials can be precisely dened [15]. A numerical

Fig. 2. Cross-section of the wall panel.

convergence study shows that further decrease of the mesh size has
little effect on the numerical results but leads to a much longer
calculation time. Therefore, a mesh size of 50 mm is used in the
study.
The steel bars are modelled explicitly by 50 mm-long beam elements connected to the concrete mesh nodes. The bond-slip
interactive effect between main vertical steel and surrounding
concrete is also incorporated in this simulation to provide more
accurate RC structural performance [16,17]. The bond strength has
been investigated by pull out experiments and was found that the
bond stress due to static friction and chemical adhesion between
the concrete and reinforcing steel is 6.6 MPa for quasi-static
loading, 18.0 MPa for dynamic loading, and 22.0 MPa for impact
loading [18]. Shi et al. [19] used 18.0 MPa as the maximum bond
strength for RC column subjected to explosive blast loading, since in
blast event the bond strain reach the maximum value later than the
direct pull out test. Therefore, 18.0 MPa is also employed herein as
the maximum bond strength between concrete and reinforcement.
As for the boundary conditions of this RC precast wall panel, the
restraint of the top and bottom is modelled as a simple support,
since it is a cladding wall.
2.1. Material models
In this LS-DYNA model, concrete is represented by the Karagozian & Case Concrete Model-Release III (MAT72 R3) [20,21]. It is
a three-invariant model where three shear failure surfaces are used
with damage. Strain rate effects are incorporated in this model, and
different dynamic increase factors (DIFs) can be employed for
concrete in compression and tension to simulate the desired rate
effect. The model characterizes all aspects of concrete by a single
input parameter, namely uniaxial unconned compressive
strength. Previous studies have proved that this model is able to
provide robust representation of complex concrete laboratory
response and be employed in structural response analysis subjected to blast loading [21].
Reinforcing steel is modelled by the Plastic Kinematic Model
(MAT3) which is suited to simulate isotropic and kinematic hardening plasticity with the option of including rate effects. The
Cowper-Symonds model [15] is used to incorporate strain rate
Table 1
Material and dimensional properties for RC wall panel.

Fig. 1. Precast wall panel details.

Parameter

Design value
Conventional wall

Blast-resistant wall

Wall width
Wall height
Wall thickness
Main vertical steel
Minor vertical steel
Minor horizontal steel
Yield strength of
main steel
Yield strength of
minor steel
Fracture strain of hoops
and cross-ties
Cover (Cn)
Concrete Compressive
Strength F0 c

2400 mm
3600 mm
160 mm
8  12 mm diameter
N.A.
N.A.
413.7 MPa (Grade 60)

2400 mm
3600 mm
160 mm
8  18 mm diameter
8 mm @ 100 mm spacing
8 mm @ 100 mm spacing
517.1 MPa (Grade 75)

N.A.

413.7 MPa (Grade 60)

10%

10%

74 mm
40 MPa

71 mm
50 MPa

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

effect, which scales the yield stress by a strain-rate dependent

factor of 1 _=C1=P , where _ is the strain rate, and C and P are the
strain rate parameters for the Cowper-Symonds model.

2.2. Strain rate effects

Concrete and steel dynamic behaviour is strain-rate dependent.
The expected magnitude of strain rate for blast loading ranges from
100 to 1000 s1 [22]. At these high strain rates, the apparent
strength of both concrete and reinforcement can increase signicantly, by more than 50 percent for reinforcing steel, by more than
100 percent for concrete in compression, and 600 percent for
concrete in tension [23]. In MAT72 R3, the strain rate effect is
accounted for by using DIF which is the ratio of the dynamic-tostatic material strength versus strain rate. Several empirical formulations [23e26] have been proposed to estimate the strain rate
effects of concrete and steel material properties.
For concrete in compression, the values of DIF recommended by
CEB model [23] is employed in this research, and shown below:

fc =fcs

_
_ cs

1:026a

fc =fcs g_1=3

for _  30 s1

(1)

for _ > 30 s1

where fc is dynamic compressive strength at _ , fcs is the static

compressive strength at _ cs , fc/fcs is the compressive strength dynamic increase factor, _ represents strain rate, _ cs 30  106 s1,
Log g 6.156a e 0.49, a (5fcu/4)1, and fcu is the static cube
compressive strength in MPa.
A modied CEB formulation for DIF of concrete in tension was
suggested by Malvar and Ross [25], because the available data at
high strain rates seems to support a change in slope close to 1 s1
instead of 30 s1 as assumed by CEB. Therefore, this modied
formulation is also utilized in this research:


_ d
for _  1 s1
_ ts

1=3
_
ft =fts b
for _ > 1 s1
_ ts

ft =fts

(2)

where ft is the dynamic tensile strength at _ , fts is the static tensile

strength at _ ts , ft/fts is the tensile strength dynamic increase factor,
_ strain rate in the range of 106 to 160 s1, _ ts 106 s1,
Log b 6d e 2, d 1/(18f 0 c /f 0 co ), and f 0 co 10 MPa.
A DIF formulation proposed by Malvar [27] for steel is:

DIF

_
104

a
(3)

where a 0.074e0.040fy/414, _ is strain rate of the steel bar in s1,

and fy is the yield strength of a steel bar in MPa.
It is of importance to note that there is a model error between
data observed during actual tests and predictive value, dened as
Model error Observed value/Predicted value

(4)

Therefore, model error is analysed for DIF of concrete in

compression and tension, and reinforcing steels via Eqn. (4), using
the following.
(a) For concrete in compression, observed values from the test
data collected by Bischoff and Perry [22], and values predicted by the CEB [23] model.

(b) For concrete in tension, observed values from the test data
collected by Yan and Lin [28], Brara and Klepaczko [29], Schuler
et al. [30] and Asprone et al. [31]and values predicted by a
modied CEB formulation proposed by Malvar and Ross [25].
(c) For steel reinforcing bars, observed values from the test data
obtained by Wakabayashi et al. [32], and values predicted by
Malvar and Crawford's [26] model.
Model error statistics herein are assumed to be a normal distribution and shown in Table 2, where the results are corrected for
the COV of testing (0.04) and the COV of specimen (0.04) [33]. It is
indicated that model error for DIF of concrete shows much higher
COV than steel reinforcing, this is because concrete is a heterogeneous material consisting of cement paste and aggregates. Moreover, many test programmes used various techniques to load the
specimens, and different methods to analyse and interpret the results. Bischoff and Perry [22] suggested that care must be taken
when comparing the results from different researchers, and comparison of material properties may sometimes only be possible
within a particular test programme. Moreover, there are also differences in mix design, type of measurement, and in the size, shape,
aspect ratio, and curing conditions of specimens in the dynamic
tests of concrete. The model error for DIF of steel obtained from test
data is 0.05 [26,32]. However, when corrected for test and specimen
uncertainties [33], the COV of model error for DIF of steel is zero,
and so is taken as deterministic in this paper.
2.3. Validation of the numerical model
In order to validate the accuracy of this nite element model, a
numerical analysis of a RC precast wall subjected to explosive is
conducted and its results compared with the test of Cramsey and
Naito [34].
A series of experiments with three wall panel types have been
carried out to investigate the behaviour of precast RC wall panels
subjected to explosive blast loads [34]. The wall in Test 1 and 3 are
analysed herein. The wall panels were installed in a reaction
structure. The reaction structure consists of a heavily reinforced
precast reaction system which supports the panel at the top and
bottom of the walls. The sides of the walls were given a 6 mm gap
on either side to allow for unrestrained movement and a one way
action response. The gaps on the sides of the walls were covered
with metal ashing to limit pressure from entering the inside of the
reaction structure. Based on the connection details used, it is
assumed that the wall panels are simply supported; with the pin
support at the base and the roller support at the top end. The parameters of this wall panel are shown in Table 3.
The dynamic pressure and displacement history were recorded.
The impulse and rst positive duration are 293.15 kPa ms and
11.0 ms for Test 1, and 459.9 kPa ms and 7.0 ms for Test 3
respectively.
The comparison of the simulated and measured deection time
history at the middle height of the wall panel is shown in Fig. 3. It is
observed that the maximum and residual deection of numerical
simulation agrees well with the corresponding values recorded
from the experiments. Therefore, the present numerical simulation
gives accurate predictions of the RC wall panels subjected to
Table 2
Statistics of model error data.
Model error data

Sample size

Mean

COV (s/m)

DIF of concrete in compression

DIF of concrete in tension
DIF of steel reinforcing bars

44
59
109

0.89
0.95
1.02

0.24
0.29
0.00

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

Table 3
Material and dimensional properties for the RC wall panel eld test [34].

Table 5
Parameters for random elds.

Parameter

RC wall panel

Parameter

Concrete cover

Concrete compressive strength

Wall width
Wall height
Wall thickness
Main vertical steel
Minor vertical steel
Minor horizontal steel
Yield strength of main steel
Yield strength of minor steel
Cover
Concrete Compressive Strength F0 c

2438 mm (8 ft)
9144 mm (30 ft)
152 mm (6 in)
8  12.7 mm diameter
5.7 mm @ 100 mm spacing
5.7 mm @ 100 mm spacing
551.6 MPa (80 ksi)
689.5 MPa (100 ksi)
76 mm (3 in)
52.1 MPa

Correlation Function
Scale of Fluctuation (qx, qy)

Gaussian
1.0 m

Gaussian
0.5 m

explosive blast loading. Note that in stochastic nite element analyses few researchers consider model error. Therefore the model
error for the FE model is deterministic and taken a unity in this
study. However, an area for further study is an improved characterisation of FE model error.
2.4. Damage states
This RC wall panel is a non-axial load bearing cladding wall
simply supported top and bottom. Based on a single degree of
freedom structural response limits for antiterrorism design code

[35] for RC members in exure, the maximum support rotation (q)

obtained from the LS-DYNA analysis is used herein to estimate the
degree of damage of RC wall panels. Three damage limit states
based on test data and UFC 3-340-02 [43] are used [35]:
(a) q < 2 Repairable damage;
(b) 2  q  5 Heavy damage;
(c) q > 5 Hazardous failure.
Support rotation Arctan (mid height deection/mid-wall
height). Repairable damage means that the component has some
permanent deection. It is generally repairable, if necessary,
although replacement may be more economical and aesthetic.
Heavy damage denes that the component has not failed, but it has
signicant permanent deections causing it to be unrepairable.
Hazardous failure means the component has failed, and debris
velocities range from insignicant to very signicant. To be sure,
other damage criteria could be considered such as concrete damage
value, debris velocity or throw, or those tailored to the specic
requirements of the asset owner.
3. Probabilistic structural response model
3.1. Structural reliability theory
The probability of damage (D) to structures conditional on the
occurrence of a specic threat (T) is
Pr(DjT) Pr[G(X)  0]

(5)

The equality G(X) 0 denotes the limit state surface, or failure

surface, that separate the safe region and the failure region. Since
both resistance (R) and loads effects (S) are functions of random
variables, the limit state surface G(X) 0 is also a function of
random variables, where X (X1/Xn) is the vector of load and
resistance variables, eg. material properties, dimensions, and loads.
3.2. Random variables

Fig. 3. Numerical model validation.

The probabilistic structural response model considers variables

associated with RC structures, which includes uncertainty and
variability in concrete and reinforcing steel properties, and member
geometry. Statistical parameters for precast RC wall panels are
shown in Table 4. These statistics are representative of precast RC

Table 4
Statistical parameters for precast RC wall panel.
Parameter

Mean

COV

Distribution

Source

Cover (mm)
Yield Strength (MPa)
Fractural Strain of Steel
Elastic Modulus of Steel (GPa)
Concrete Compressive Strength
ME-DIF for concrete in compression
ME-DIF for concrete in tension

Cn
1.15Fy
0.1
202
0.99F0 c
0.92
0.95

0.09
0.05
0.15
0.04
0.09
0.24
0.29

Normal
Normal
Normal
Normal
Lognormal
Normal
Normal

Mirza and MacGregor [51]

Wisniewski et al. [52]
Wisniewski et al. [52]
Wisniewski et al. [52]
Wisniewski et al. [52]
Section 2.2
Section 2.2

Note: All parameters truncated at zero, Fy nominal (design) yield strength of reinforcing steel, F0 c nominal (design) concrete compressive strength, Cn nominal (design)
concrete cover.

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

structures in the United States, Canada, and Europe. The variability

of cross-section area of reinforcing bars, reinforcement spacing,
wall dimensions, and other variables are assumed as deterministic.
The yield strength for main vertical reinforcement and minor
reinforcement are assumed to be statistically independent. Due to
the material model of concrete used in LS-DYNA, other concrete
material properties, except concrete compressive strength, are
dependent variables.
3.3. Random eld analysis
Due to variations in work practice and environment, the nature
of various RC construction processes introduces the potential for
substantial variability of concrete quality. The spatial variability of
concrete is an important characteristic, which qualies the nonhomogeneity of mechanical and physical properties on structural
components [36]. Therefore, the spatial variability of concrete
compressive strength and concrete cover are incorporated in present study.
The Gaussian correlation function has been widely used in engineering applications, including spatial modelling of RC structures
[37,38], and as such it shall be used herein to model the spatial
correlation of elements within a random eld. Thus for random
elds with two dimensions (i.e. the surface of a RC element), the
Gaussian correlation function is described as:

rt exp

jtx j

dx

2

 2 !
ty

dy

where tx xi  xj and ty yi  yj are the distance between the

centroid of element i and j in the x and y directions respectively,
dx qx/p and dy qy/p where qx and qy are the scales of
uctuation.
The scale of uctuation denes the distance over which correlation persists in a random eld. The estimation of the scale of
uctuation can be calculated from existing data or estimated using
engineering judgement. The strength data of 18 RC columns [39]
are used to obtain the scale of uctuation for concrete compressive strength, which ranges from 0.27 m to 0.99 m. An average value
of 0.50 m is used in this paper. A value of qx qy 1.0 m [38] for
concrete cover is used herein. This means, for example, that concrete strengths at distances greater than 0.5 m have low correlation,
due to variation in concrete batching, curing, etc.
The size of the discretised element is usually dened based on
practical and analytical considerations [40]. Sterritt et al. [40]
suggested that the element size should be in the range of
0.1e0.75 m. Meanwhile, the element size of a random eld should
also be in conjunction with the mesh size of LS-DYNA. Therefore, an
element size of 0.2 m is used herein for random eld modelling.
Note that the mesh size for the FEA analysis remains at 50 mm so
that, for example, four adjacent meshes will have the same spatial
variable such as cover. The parameters for the random elds are
shown in Table 5.
4. Probabilistic blast loading model

(6)
It is clearly acknowledged that structural reliability is sensitive
to variability of loads. Therefore, it is necessary to understand and

Table 6
Statistical parameters for probabilistic blast loading model [14].
Parameter
Mass variability
Terrorist VBIED scenario
User factor (Wuser)
NEQ factor (WNEQ) for
Peak reected pressure (Pr)
Peak reected impulse (Ir)
Positive phase duration (td)
Military CDE scenario
User factor (Wuser)
NEQ factor (WNEQ) for
Peak reected pressure (Pr)
Peak reected impulse (Ir)
Positive phase duration (td)
Detonation location
Terrorist VBIED scenario
Target Location Error (TLE)
Weapon Delivery Error (WDE)

Military CED scenario

Target Location Error (TLE)
Weapon Delivery Error (WDE)
Ambient Air Temperature ( C)
Ambient Air Pressure (hPa)
Inherent Variability
Model error
Peak reected pressure (Pr)
Peak reected impulse (Ir)
0.59 m/kg1/3  Z < 6.0 m/kg1/3
6.0 m/kg1/3  Z < 40.0 m/kg1/3
Time of positive phase duration (td)
0.59 m/kg1/3  Z < 6.0 m/kg1/3
6.0 m/kg1/3  Z < 9.0 m/kg1/3
9.0 m/kg1/3  Z < 40.0 m/kg1/3

Mean

COV

Distribution

1.000

0.102

Normal

Mode 0.82
Mode 0.82
Mode 0.82

0.359
0.359
0.359

Triangular
Triangular
Triangular

1.000

0.001

Normal

1.070
0.960
1.015

0.025
0.025
0.025

Normal
Normal
Normal

xyz0
x0
y0
z0

s0m
sx 3.06 m
sy 1.53 m
sz 0 m

Deterministic
Normal
Normal
Deterministic

xy0
z0
xy0
z0
21.9  C
1015.0 hPa
1.000

s 8.49 m
s0m
s 11.04 m
s0m
0.356
0.014
0.010

Normal
Deterministic
Normal
Deterministic
Normal
Uniform
Normal

1.032

0.069

Normal

0.991
0.991

0.178e0.0236Z
0.036

Normal
Normal

0.43 0.596log10Z
0.43 0.596log10Z
1.00

C0 C1Z C2Z2 C3Z3

0.046
0.046

Normal

Note: C0 0.6267, C1 0.3510, C2 0.0713, C3 0.0048, scaled distance Z R/W1/3 (m/kg1/3).

Normal

10

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

quantify the variability of loads to estimate the reliability of

structures. It is observed form eld test that the blast load experienced by a target structure, for apparently similar circumstances,
will not always be the same [41]. The variability in blast loading can
be traced to parameter uncertainty, inherent variability, and model
error [14].
The probabilistic blast load model developed by Netherton and
Stewart [14], considers uncertainties for:
(a) User factor for mass of explosive (Wuser),
(b) Net equivalent quantity (NEQ) of an explosive in terms of a
mass of TNT (WNEQ)
(c) The range (R) and Angle of Incidence (AOI),
(d) Air temperature (Ta) and pressure (Pa), and
(e) Accuracy (model error) of load models.
Probabilistic models for model error and inherent variability
were obtained from eld data of repeatable tests. The polynomial
curves from the explosive blast loading model proposed by
Kingery and Bulmash [42] have been incorporated into widely
used and well respected blast load design model, such as: ConWep, UFC 3-340-02 [43], and LS-DYNA. Given such wide acceptance, the Kingery and Bulmash [42] polynomials are also
employed to predict blast load values. The probabilistic blast load
model assumes a hemispherical surface burst. In this research, the
modied Friedlander curve [44] is used to generate blast pressuretime histories.
4.1. Blast scenarios
The probabilistic blast loading model describes two circumstances that reect the concerns and reality of current world situation with respect to hazards as a result of explosive blast loads
[45]:
(a) Terrorist VBIED scenario. A terrorist-style vehicle borne
improvised explosive device (VBIED) containing home-made
ammonium nitrate fuel oil (ANFO). The size of explosive used
by a terrorist can vary, with mass ranging from truck-size
with devastating effect through to smaller suitcase-size
IEDs. As such, the following three types are considered: (i)
50 kg (small VBIED), (ii) 116 kg (car-size), (iii) 1000 kg (trucksize) of home-made ANFO.
(b) Military CDE scenario. The deployment of a military weapon
lled with Tritonal, where collateral damage estimation
(CDE) is a major consideration. The Mark-80 Series of General

Purpose (Mk80 GP) bombs are in signicant use in Western

air forces. Therefore, two types of Mark-80 series of GP
bombs are considered: (i) Mk82 GP 500 lb bomb (89 kg
Tritonal), and (ii) Mk83 GP 1000 lb bomb (202 kg Tritonal).
The United States Air Force (USAF) [46] describes the total
overall error related to the detonation location as a statistical
combination of the errors associated with determining the coordinates of the impact point, Target Location Errors (TLE), and
those associated with delivering the weapon to that point, Weapon
Delivery Errors (WDE). Due to the difference of the methods for
locating the target and delivering the weapon, the Military CDE and
Terrorist VBIED scenarios have different TLE and WDE. In that, the
Mark-80 series of GP bombs has to y to a remotely determined
detonation location, while a VBIED will be driven to its location
[45]. In the Military CDE scenario, the 50% Circular Error Probable
(CEP) is used to describe the accuracy of the weapon delivery. The
50% CEP gure is the radius of a circle around the target within
which 50% of the weapons should fall [46]. GPS guided GBU-38/B
JDAM GP bombs (CEP 13.0 m) are assumed in this probabilistic
blast loading model.
The types and mass of explosive for the Terrorist VBIED and
Military CDE scenarios are different. However, the explosives of
116 kg ANFO and 89 kg Tritonal can be considered deterministically
similar in terms of equivalent peak reected pressure. This is
because the explosive mass for the deterministic blast load model
have to be converted into an equivalent mass of TNT, which is
approximately 95 kg of TNT for both scenarios. Statistical parameters for the probabilistic blast model for a hemispherical charge
are shown in Table 6, and results are not sensitive to air temperature and pressure.

4.2. Effects of charge shape

For a non-spherical charge, the shock wave will not enter the
surrounding air as a spherical wave, nor at the same instant over
the entire charge surface. The shape and strength of the shock wave
entering the air will not depend upon charge geometry, and upon
the relative location at which initiation occurred. The blast parameters will be functions not only of radial standoff, but also of
azimuth [47]. Several experimental programs investigated the blast
eld around non-spherical explosives of regular geometries such as
cylinders, cubes, and cones. UFC 3-340-02 [43] also provides peak
pressure and scaled impulse of various non-spherical charge
shapes at various scaled distance, but azimuth is not specied.

Fig. 4. Australia's rst suicide bomber in Syria (Image from ABC News).

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

11

(Wcyl)  Equivalent Spherical Mass Ratio (Wsph/Wcyl)  Equivalent

Hemispherical Mass Ratio (Whem/Wsph).
4.2.1. Equivalent spherical mass ratio
The aspect ratio (length divided by diameter) of the explosive of
L
D 6 is considered since it is a typical ratio in VBIED scenarios
(Fig. 4), and for Mark-82 GP bombs.
Plooster [41] presented results of an experimental study of freeair blast wave properties for end-initiated cylindrical explosive
charges with various L/D ratios at several azimuth angles (Fig. 5).
Base on those data, the spherical equivalent mass ratio is determined in the similar way as TNT equivalency weights are generally
found. For each cylindrical peak pressure, an equivalent spherical
weight is determined that would produce the same peak pressure
at the same distance, R. Thus for peak pressure, the ratio of the
charge weight of a sphere Wsph to that of a cylinder Wcyl is

Wsph

Wcyl

Zcyl
Zsph

!3
(7)
Pcons tan t

Likewise, for impulse

Wsph

Wcyl

Fig. 5. Plan view of measurement lines for a free-air cylindrical charge.

In this research, a cylindrical explosive is considered, since it is a

common shape in VBIED and military scenarios. The probabilistic
blast model developed by Netherton and Stewart [14] employs a
hemispherical surface-bust. Thus, the cylindrical explosive mass
needs to be converted to on equivalent hemispherical mass. The
Equivalent Hemispherical Mass (Whem) Cylindrical Mass

mean

COV

of

of

Wsph
pressure
Wcyl

Wsph
pressure
Wcyl

0:025 0:610Z  0:072Z 2 ;

1:0;

0:462Z 1:549;
0:24;

Zcyl
Zsph

!3
(8)
Icons tan t

where Zcyl is the scaled distance R=Wcyl 1=3 and Zsph is the scaled
distance R=Wsph 1=3 . Based on Plooster's [41] data of Pentolite
charges, Wsph/Wcyl for pressure and impulse of free-air cylinders
with aspect ratio of 6/1 at ve different azimuth angles is presented
in Fig. 6. The average values of Wsph/Wcyl for all azimuth angles are
shown in Fig. 7. The Wsph/Wcyl of free-air cylinders is assumed the
same for surface burst because compared with free-air burst, only a
reection factor is needed to be multiplied for the charge weight to
estimate blast parameters for surface burst, and the reection
factor will be cancelled out in the calculation.
It is seen from Fig. 7 that the line of best t for mean value of
Wsph/Wcyl is best represented by a second degree polynomial line,
and COV by a rst degree polynomial line. It is assumed herein that
the distribution of Wsph/Wcyl is a normal distribution.
The lines of best t for pressure from Fig. 7 are:

1:24  Z  6:16 m=kg 1=3

Z > 6:16 m=kg 1=3

1:24  Z  2:83 m=kg 1=3

Z > 2:83 m=kg 1=3

(9)

(10)

Likewise, for impulse

mean

COV

of

of

Wsph
impulse
Wcyl
Wsph
impulse
Wcyl

0:383 0:711Z  0:093Z 2 ;

1:0;

0:907Z 2:098;
0:33;

1:24  Z  6:64 m=kg 1=3

Z > 6:64 m=kg 1=3

1:24  Z  1:95 m=kg 1=3

Z > 1:95 m=kg 1=3

(11)

(12)

12

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

Fig. 6. Equivalent spherical mass ratio for a free-air cylindrical charge of L/D 6.

where Z is the scaled distance R/W1/3 (m/kg1/3). The best t curve

for mean trends to 1.0, because the difference of shock wave
strength caused by the effects of charge shape could be negligible
when the scaled distance is large. As shown in Fig. 8, there are no
experimental data when the scaled distance is less than 1.24 m/kg1/
3
. Hence, it is assumed herein that the values of mean and COV of
the equivalent mass ratio for both pressure and impulse stay at the
values when Z 1.24 m/kg1/3 if Z < 1.24 m/kg1/3. Clearly, further
studies for explosions at small scaled distances are required to

8
3:039e0:365Z ;
<
Whem
pressure 1:63  0:15Z 0:007Z 2 ;
:
Wsph
1:0;

Fig. 7. Fitted curves for mean and COV of Wsph/Wcyl for L/D 6.

better quantify parameter uncertainties associated with the

equivalent spherical mass ratio.
4.2.2. Equivalent hemispherical mass ratio
Swisdak [48] investigated the peak pressure for spherical and
hemispherical TNT detonated on the surface. Based on his eld test
results, the equivalent hemispherical mass ratio (Whem/Wsph) is
shown in Fig. 8. In this research, this ratio is considered as deterministic. It is seen from Fig. 8 that the lines for best t for (Whem/
Wsph) of pressure and impulse are:

0:59  Z < 2:25 m=kg1=3

2:25  Z  15:76 m=kg 1=3
Z > 15:76 m=kg 1=3

(13)

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

13

1.2 106
Cylindrical
Spherical
Hemispherical

Reflected Pressure (Pa)

1 106

8 105

6 105

4 105

2 105

0.002

0.004

0.006

0.008

0.01

Time (sec)
Fig. 9. Pressure-time diagram for different explosive shapes.

phase duration is close to 1.0. Hence the equivalent mass ratio for
positive phase duration is assumed as 1.0 in this research.
The empirical relationships given by Eqns. (9)e(14) provide a
useful starting point for characterizing the statistical parameters for
the effect of charge shape on blast pressures, although more
research is clearly needed.
4.3. The variability of blast loading
The probability distribution of reected impulse (Ir) is shown in
Fig. 10 for (i) 50 kg VBIED, (ii) 116 kg VBIED, (iii) 1000 kg VBIED, (iv)
Mk82 (500 lb) GP bomb, and (v) Mk83 (1000 lb) GP bomb for
R 10 m and DL 6. It is observed that the variability of blast load is

0.0008
50 kg VBIED

Fig. 8. Fitted curve for Whem/Wsph.

0.0007

Mk82 GP bomb

VBIED
Mk-80 series GP bomb

8
< 1:275  0:422Z; 0:40  Z < 1:19 m=kg1=3
Whem
impulse
0:424 0:29Z; 1:19  Z  1:98 m=kg 1=3
:
Wsph
1:0;
Z > 1:98 m=kg 1=3
(14)
To illustrate the effects of charge shape, the blast-wave proles
for three different explosive shapes of W 116 kg ANFO detonated
at R 10 m are shown in Fig. 9. It is observed that in this scenario
peak reected pressure and peak reected impulse for cylindrical
explosives are 30% higher, compared with hemispherical explosives. This is because at this scaled distance both equivalent
spherical mass ratio (Wsph/Wcyl) and equivalent hemispherical
mass ratio (Whem/Wsph) for pressure and impulse are greater than
1.0.
The difference of positive phase duration (td) between cylindrical and spherical explosives is not explicitly discussed in the
broad literature, and the hemispherical mass ratio for positive

Probability Density

0.0006
116 kg VBIED

0.0005

Mk83 GP bomb

0.0004
0.0003
0.0002
1000 kg VBIED

0.0001
0
0

2000

4000

6000

8000

Reflected Impulse Ir (kPa-msec)

Fig. 10. Probability distribution of reected impulse for R 10 m.

1 104

14

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

Probability of Repairable Damage

0.8

(a)

Mk82 500 lb bomb

50 kg VBIED
Mk83 1000 lb bomb

0.6
1000 kg VBIED
0.4

116 kg VBIED

0.2
Non-spatial model
Spatial model
0
0

Fig. 11. Simulation histogram of support rotation for conventional wall for W 116 kg
VBIED and R 10 m.

5. Structural reliability analysis

The reliability analysis is complicated because the structural
system failure modes are neither statistically independent nor fully
dependent. Load and resistance are not statistically independent,
and structural response will be calculated from a non-linear FEA. So
a closed form solution is not readily tractable. Hence, the probability of damage will be obtained from a stochastic FEA model
which will use event-based Monte-Carlo simulation (MCS) analysis.
This will enable the progression of damage modes to be tracked and
reliability calculated for each limit states. The basic steps of this
approach are described in the following:
(a) Randomly generate peak reected pressure (Pr) and time of
rst positive phase duration (td) for known explosive mass
(W) and stand-off distance (R), and blast scenarios (Military

Probability of Hazardous Failure

considerable, with COVs of 0.40e3.30 for Ir for terrorist scenarios,

and COVs of 1.88e5.97 for Ir for the military scenarios. The higher
COV for the VBIED scenario occurs because VBIEDs are often located
exactly at a desired detonation location (TLE 0) with some small
variation in actual placement of the device related to the desired
location (WDE > 0). However, the aerial weapon of the Military CDE
scenario requires pre-loaded coordinates determined by remote
means (TLE > 0) with large WDE caused by the weapon's ight to
the desired location [45]. Hence, the standard deviation of detonation location in the military scenario is signicantly higher than
in the terrorist scenario (Table 6).
An understanding of a blast wave's angle of incidence (AOI) e
relative to the plane of a target's surface is very important. There
are signicant variations in the pressure and impulse imparted by a
blast wave to a target, where such variations will be a function of
the incident pressure and the incident angle between the blast
front and the target's surface [45]. The effect of AOI is also incorporated in the probabilistic blast loading model. The peak pressure
and positive impulse varies over the height and width of the wall
with the highest value at the bottom centre.

10

20

30

Range (m)

(b)

40

50

Non-spatial model
Spatial model

0.8

0.6
116 kg VBIED
1000 kg VBIED
0.4

50 kg VBIED

0.2

Mk82 500 lb bomb

Mk83 1000 lb bomb

0
0

10

20

30

Range (m)

40

50

Fig. 12. Probability of (a) repairable damage and (b) hazardous failure for non-spatial
and spatial blast-resistant wall models.

(b)

(c)
(d)
(e)
(f)

CDE or Terrorist VBIED) using the modied probabilistic blast

loading model for cylindrical explosive with aspect ratio of 6/
1.
Randomly generate six statistically independent variables:
yield strength of main vertical steel, yield strength of minor
reinforcement, fractural strain, elastic modulus of steel, and
model errors for DIF of concrete in tension and compression.
Generate spatial variables: cover and f 0 c from stationary
random eld.
Estimate mid-height displacement and support rotation q
from FEA using simulated variables as input parameters.
Repeat Steps (a)e(d) so that q is estimated for each MonteCarlo simulation realization of random variables.
Extract probabilistic information from N such realisations.

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

Fig. 13. Blast reliability curves for Terrorist VBIED scenarios.

15

16

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

Fig. 14. Blast reliability curves for Military CDE scenarios.

The procedure of MCS for non-spatial RC precast wall panel is

identical with the above procedure, except using non-spatial variables in Step (c).
The probability of damage states conditional on threat T is
Pr(DjT):

nq < 2+ 
N
n2+  q  5+ 
Prheavy damagejT
N
nq > 5+ 
Prhazardous failurejT
N
Prrepairable damagejT

(15)

where n[] is the number of realisations when D matches the

damage criteria, and N is the number of simulation runs.
6. Results
The blast scenarios considered are three VBIEDs (50 kg, 116 kg,
1000 kg) and two Mark-80 series GP bombs (Mk82 500 lb bomb

and Mk83 1000 lb bomb) detonated from R 2.5 m to R 50 m

from the front face of a 3.6 m high precast RC cladding wall panel.
Table 1 and Figs. 1 and 2 show that the RC wall is 2.4 m long and
160 mm thick. The conventional wall has reinforcing steel diameter
of 12 mm, yield strength of 413.7 MPa, and concrete compressive
strength of 40 MPa. While the blast-resistant wall has larger steel
diameter and higher grades of steel and concrete. Due to high
computational demand associated LS-DYNA, N 100 simulation
runs for each scenario were used to generate distribution of support rotation of the walls, and estimates of probability of damage
and collapse. Stochastic analyses were conducted for (i) non-spatial
conventional wall model, (ii) non-spatial blast-resistant wall
model, and (iii) spatial blast resistant wall model.

6.1. Spatial vs. non-spatial model

Fig. 11 shows the histogram of support rotations for the conventional wall for W 116 kg VBIED and R 10 m. Then the
probability of three damage levels will be calculated from the histograms. The probability of repairable damage and hazardous

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

17

Fig. 15. Risk reduction curves.

failure for non-spatial and spatial models are shown in Fig. 12, for
the blast-resistant wall. It is found that the probability of repairable
damage decreases and the probability of hazardous failure increase
as the stand-off distance increases for both Terrorist VBIED and
Military CDE scenarios. The results of the spatial model are near
identical with the non-spatial model, which suggests that the
spatial variability of concrete cover and concrete compressive
strength has little effect on structural reliabilities. This is not surprising because due to horizontal casting of the wall panel the
random eld of concrete strength is modelled as a stationary
random eld where the mean and covariance are constant over the
entire random led. If it is a vertical casting, the mean value of
concrete compressive strength will vary with height due to selfweight compaction [49]. Secondly, the COV of concrete strength
is very small because of the high quality of manufacturing and
plant-cast construction, rather than in-situ construction. Therefore,

Fig. 16. Sensitivity analysis of wall panel parameters.

the spatial variability of concrete compressive strength is quite

small in the precast RC wall panel. Thirdly, the sensitivity analysis in
Section 6.4 suggests that the concrete cover has little effect on
structural response. Hence, the difference of the performance between spatial and non-spatial models in the reliability analysis is
insignicant. All results to follow are for the non-spatial model.

6.2. Conventional vs. blast-resistant walls

The conventional wall panel exhibits a larger support rotation
than the blast-resistant wall panel for the same blast loading. This is
because the exural resistance of blast-resistant wall is higher than
a conventional wall, due to higher reinforcement ratio, yield
strength, and grade of concrete. Blast Reliability Curves (BRCs) are
shown in Figs. 13 and 14 for conventional and blast-resistant walls
for VBIED and military scenarios, respectively. The 90% condence
bounds are also shown e more simulation runs would reduce the
90% condence intervals, but those are sufcient to infer the BRCs.
The difference between the probability of damage for the conventional wall and the blast resistant wall is signicant. For
example, if the 1000 kg VBIED is presumed to detonate at a
Designated Mean Point of Impact (DMPI) that is 30 m directly in
front of the wall panel, the probability of hazardous failure for the
conventional wall is 54%, and reduces to 21% for the blast-resistant
wall (Fig. 13 (e) and (f)). If the Mk82 500 lb bomb is presumed to
detonate at a DMPI that is also 10 m, the probability of hazardous
failure for the conventional wall is 8%, and reduces to 1% for the
blast-resistant wall (Fig. 14 (a) and (b)). For the blast-resistant wall
the probability of repairable damage is 0e28% higher, and the
probability of hazardous failure is 3%e100% lower than the corresponding values of the conventional wall.
It is also found that the difference between structural responses
is also signicant when subjected to 116 kg VBIED and Mk82 GP
explosions when detonated at the same stand-off distance, even
though the explosives of 116 kg ANFO and 89 kg Tritonal can be
considered deterministically similar in terms of equivalent peak
reected pressure. This suggests that the structural response is
sensitive to the simulated blast scenarios. Table 6 shows the standard deviation of target location and weapon delivery errors in the
Military CED scenario is much higher than it in the Terrorist VBIED
scenario, which result in higher COV of reected impulse (Fig. 11).

18

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

6.3. Risk reduction

The reduction of probability of hazardous failure for conventional and blast-resistant walls in various blast scenarios are shown
in Fig. 15. This gure indicates how much risk will be reduced, if
protective measures such as bollards or other perimeter security
measures allow the stand-off to be increased. For example, if the
stand-off distance of a 116 kg VBIED increase from 2.5 to 10 m, the
probability of hazardous failure is reduced by 64% and 87% for
conventional and blast-resistant walls, respectively. This analysis
will allow comparisons to be made between the relative effectiveness of security measures, explosive delivery method or other
mitigation measures.
6.4. Sensitivity analysis
A sensitivity analysis is performed to study the relative importance of the variability of each random variable on the calculated
structural response. An approach developed by Pan et al. [50] is
herein to show the effects of variations in each of the parameters.
The blast scenario is deterministic with W 116 kg ANFO and
R 15 m. The method changes the parameters of each variable by
10% of the mean value, and considers the effects of the change on
the output for each of the random variables. The non-spatial wall
model is used in the sensitivity analysis.
The results are shown in Fig. 16, where the parameters and the
support rotations are normalized with respect to the mean value of
parameters and corresponding support rotations. It is observed that
the support rotation is most sensitive to standoff distance, followed
by explosive mass. Compared with the input parameters for blast
loading, parameters for structural resistance have signicantly
smaller inuence on the output.

As shown in Fig. 17, the probability of damage for deterministic

resistance where the random variables for structural are taken as
deterministic present nearly identical results for probabilistic
resistance. This also indicates that the parameters for structural
resistance have little effects on the output.
7. Future work
In this research, various parameters values and their statistical
characterisations are calculated, surmised and assumed from
limited data sets. There are several areas for future work in the
improvement and validation of these values. Firstly, there have
been many tests for DIF over many decades, but few of them were
repeated. Repeatable tests are required to provide more accurate
estimation of model error of concrete in compression and tension
at high strain rates. Secondly, repeated experimental tests could be
conducted to statistically characterise the variability of model error
for FE analyses used to simulate structural response to blast loads.
Thirdly, the charge shape considered herein is the cylinder with L/
D 6. More charge shapes or length-to-diameter ratios could be
incorporated to investigate the effects of charge shapes on structural response.
8. Conclusions
This paper stochastically modelled the damage to precast RC
wall panels when subjected to explosive blast loading. This probabilistic analysis incorporates considerable uncertainties associated
with structural response, blast loading, and spatial variability of
concrete compressive strength and concrete cover. A simply supported non-loading bearing cladding wall panel was subjected to
Terrorist VBIED and Military CDE blast scenarios. In blast scenarios,

Fig. 17. Probability of damage for deterministic and probabilistic resistance.

Y. Shi, M.G. Stewart / International Journal of Impact Engineering 85 (2015) 5e19

the uncertainty of the support and casing effect of the explosives

are not considered in this paper. A comprehensive and detailed 3-D
FE structural model was developed using LS-DYNA. Combining this
model with Monte-Carlo methods, the probability of damage of RC
wall panels is obtained. The structural reliability analysis calculated
damage and collapse risks for RC wall panels subjected to various
explosive threat scenarios for non-spatial and spatial models, as
well as conventional and blast-resistant wall panels. The spatial
variability of concrete compressive strength and concrete cover has
little effects on the structural reliabilities for this precast structure.
It is also found that the blast-resistant wall has up to 27% higher
probability of repairable damage, and the probability of hazardous
failure is 5%e100% lower than the corresponding values for conventional walls. It is observed that the support rotation is most
sensitive to standoff distance, followed by explosive mass.
Compared with blast loading input parameters, statistical parameters for structural resistance have a signicantly smaller inuence
on the output.
Acknowledgement
The authors would like to acknowledge the support of the
Australian Research Council. The rst author also appreciates the
nancial support provided by the China Scholarship Council and
the University of Newcastle.
References
[1] Remennikov AM, Kong SY. Numerical simulation and validation of impact
response of axially-restrained steel-concrete-steel sandwich panels. Compos
Struct 2012;94:3546e55.
[2] Lin X, Zhang YX, Hazell PJ. Modelling the response of reinforced concrete
panels under blast loading. Mater Des 2014;56:620e8.
[3] Gareld TT, Richins WD, Larson TK, Pantelides CP, Blakeley JE. Performance of
RC and FRC wall panels reinforced with mild steel and GFRP composite in blast
events. Proc Eng 2011;10:3534e9.
[4] Pan YG, Watson AJ. Interaction between concrete cladding panels and xings
under blast loading. Cem Concr Compos 1996;18:323e32.
[5] Low HY, Hao H. Reliability analysis of reinforced concrete slabs under
explosive loading. Struct Saf 2001;23(2):157e78.
[6] Low HY, Hao H. Reliability analysis of direct shear and exural failure modes
of RC slabs under explosive loading. Eng Struct 2002;24(2):189e98.
[7] Netherton MD, Stewart MG. The effects of explosive blast load variability on
safety hazard and damage risks for monolithic window glazing. Int J Impact
Eng 2009;36(12):1346e54.
[8] Stewart MG, Netherton MD. Security risks and probabilistic risk assessment of
glazing subject to explosive blast loading. Reliab Eng Syst Saf 2008;93(4):
627e38.
[9] Kelliher D, Sutton-Swaby K. Stochastic representation of blast load damage in
a reinforced concrete building. Struct Saf 2012;34:407e17.
[10] Stewart MG, Netherton MD, Shi Y, Grant M, Mueller J. Probabilistic terrorism
risk assessment and risk acceptability for infrastructure protection. Aust J
Struct Eng 2012;13(1):1e17.
[11] Stewart MG, Shi Y, Zhi X. Structural reliability analysis of reinforced concrete
columns subject to explosive blast loading. In: 9th international conference on
shock & impact loads on structures; 2011.
[12] Stewart MG, Melchers RE. Probabilistic risk assessment of engineering system.
London, UK: Chapman & Hall; 1997. p. 20e5.
[13] Shi Y, Stewart MG. Spatial reliability analysis of explosive blast load damage to
reinforced concrete column. Struct Saf 2015;53(3):13e25.
[14] Netherton MD, Stewart MG. Blast load variability and accuracy of blast load
prediction model. Int J Prot Struct 2010;1:543e70.
[15] Hallquist JO. LS-DYNA theory manual. Livermore, California: Livermore Software Technology Corporation; 2006.
[16] Salari MR, Spacone E. Finite element formulations of one-dimensional elements with bond-slip. Eng Struct 2001;23:815e26.
[17] Casanova A, Jason L, Davenne L. Bond slip model for the simulation of reinforced concrete structures. Eng Struct 2012;39:66e78.
[18] Weatherby JH. Investigation of bond slip between concrete and steel reinforcement under dynamic loading conditions. The Department of Civil and
Environmental Engineering, Mississippi State University; 2003.
[19] Shi YC, Hao H, Li ZX. Numerical derivation of pressure-impulse diagrams for
prediction of RC column damage to blast loads. Int J Impact Eng 2008;35(11):
1213e27.

19

[20] Malvar LJ, Crawford JE, Wesevich JW, Simons D. A plasticity concrete material
model for DYNA3D. Int J Impact Eng 1997;19:847e73.
[21] Malvar LJ, Crawford JE, Morrill K. K&C concrete material model release IIIeautomated generation of material model input. In: K&C technical report: TR99-24-B1, Glendale; 2000.
[22] Bischoff PH, Perry SH. Compressive behavior of concrete at high-strain rates.
Mater Struct 1991;24(144):425e50.
[23] CEB-FIP MC. Comite Euro-International du Beton. Wiltshire, UK: Redwood
Books; 1993.
[24] Zhou XQ, Kuznetsov VA, Hao H, Waschl J. Numerical prediction of concrete
slab response to blast loading. Int J Impact Eng 2008;35(10):1186e200.
[25] Malvar LJ, Ross CA. Review of strain rate effects for concrete in tension. ACI
Mater J 1998;95(6):735e9.
[26] Malvar LJ, Crawford JE. Dynamic increase factors for steel reinforcing bars.
Orlando, U. S.: Naval Facilities Engeering Service Centre; 1998.
[27] Malvar LJ. Review of static and dynamic properties of steel reinforcing bars.
ACI Mater J 1998;95(5):609e16.
[28] Yan D, Lin G. Dynamic properties of concrete in direct tension. Cem Concr Res
2006;36(7):1371e8.
[29] Brara A, Klepaczko JR. Experimental characterization of concrete in dynamic
tension. Mech Mater 2006;38(3):253e67.
[30] Schuler H, Mayrhofer C, Thoma K. Spall experiments for the measurement of
the tensile strength and fracture energy of concrete at high strain rates. Int J
Impact Eng 2006;32(10):1635e50.
[31] Asprone D, Cadoni E, Prota A. Experimental analysis on tensile dynamic
behavior of existing concrete under high strain rates. ACI Struct J 2009;106(1):
106e13.
[32] Wakabayashi M, Nakamura T, Yoshida N, Iwai S, Watanabe Y. Dynamic
loading effects on the structural performance of concrete and steel materials
and beams. In: Proceedings of the seventh world conference on earthquake
engineering6; 1980. p. 271e8.
[33] Ellingwood B, Galambos TV, MacGregor JG, Cornell CA. Development of a
probability based load criterion for American national standard A58. Washington DC, USA: U. S. Government Pringting Ofce; 1980.
[34] Cramsey N, Naito C. Analytical assessment of the blast resistance of precast,
prestressed concrete components. Bethlehem, U. S. A.: Lehigh University;
2007.
[35] Single degree of freedom structural response limits for antiterrorism design.
U.S. Army Coprs of Engineers Protective Design Center; 2008.
[36] Nguyen NT, Sbartai ZM, Lataste JF, Breysse D, Bos F. Assessing the spatial
variability of concrete structures using NDT techniques e laboratory tests and
case study. Constr Build Mater 2013;49:240e50.
[37] Stewart MG, Mullard JA. Spatial time-dependent reliability analysis of corrosion damage and the timing of rst repair for RC structures. Eng Struct
2007;29(7):1457e64.
[38] Sudret B, Defaux G, Pendola M. Stochastic evaluation of the damage length in
RC beams submitted to corrosion of reinforcing steel. Civ Eng Environ Syst
2007;24(2):165e78.
teborg: Elanders
[39] Petersons N. Strength of concrete in nished structures. Go
boktr.; 1964.
[40] Sterritt G, Chryssanthopoulos MK, Shetty NK. Reliability-based inspection
planning for RC highway bridges, safety, risk, reliability e trends in engineering. 2001.
[41] Plooster MN. Blast effects from cylindrical explosive charges: experimental
measurements. Denver Research Institute; 1982.
[42] Kingery CN, Bulmash G. Airblast parameters from TNT spherical air burst and
hemispherical surface burst. Maryland, U. S. A.: US Army Armament Research
and Development Centre; 1984.
[43] UFC3-340-02. Structures to resist the effects of accidental explosions. Washington DC, USA: U. S. Army Corps of Engineers; 2008.
[44] Borenstein E, Benaroya H. Sensitivity analysis of blast loading parameters and
their trends as uncertainty increases. J Sound Vib 2009;321(3e5):762e85.
[45] Netherton MD. Probabilistic modelling of structural and safety hazard risks for
monolithic glazing subject to explosive blast loads. PhD thesis. Newcastle,
Australia: School of Engineering, the University of Newcastle; 2012.
[46] USAF. USAF intelligence targeting guide. Air Force Pamphlet 14-210 Intelligence; 1998.
[47] Esparza ED. Spherical equivalency of cylindrical charges in free-air. Southwest
Research Institute; 1992.
[48] Swisdak MMJ. Explosive effects and properties, part I e explosive effects in air.
Naval Surface Weapons Centre; 1975.
[49] Zhu WZ, Gibbs JC, Bartos PJM. Uniformity of in-situ properties of selfcompacting concrete in full scale structural elements. Cem Concr Comp
2001;23:57e64.
[50] Pan Y, Agrawal AK, Ghosn M. Seismic fragility of continuous steel highway
bridges in New York State. J Bridge Eng 2007;12(6):689e99.
[51] Mirza SA, Macgregor JG. Variations in dimensions of reinforced concrete
members. J Struct Div 1979;105(4):751e66.
[52] Wisniewski DF, Cruz PJS, Henriques AAR, Simoes RAD. Probabilistic models for
mechanical properties of concrete, reinforcing steel and pre-stressing steel,
structure and infrastructure engineering: maintenance, management. Life
Cycle Des Perform 2012;8(2):111e23.