You are on page 1of 15

International Journal of Impact Engineering 68 (2014) 41e55

Contents lists available at ScienceDirect

International Journal of Impact Engineering


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

Numerical study of concrete spall damage to blast loads


Jun Li b, *, Hong Hao a, b
a
Tianjin University and the University of Western Australia Joint Research Centre of Protective Structures, Tianjin, China
b
School of Civil and Resource Engineering, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia

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

Article history: Spall damage is a typical damage mode of concrete structures under blast or high velocity impact loads.
Received 20 July 2013 Blast and impact loads generate a stress wave propagating in the structure. At the opposite side from
Received in revised form which the structural element is impulsively loaded, spall will occur if the net primary stresses over an
24 November 2013
area exceed the concrete dynamic tensile strength and the resistance force such as the material dynamic
Accepted 1 February 2014
Available online 8 February 2014
bond and interlock. Fragments of structural element generated from spall damage could eject with large
velocities, and impose significant threats to equipment and personnel even it does not necessarily greatly
reduce the load carrying capacity of the structural components. In the present study, spall damage of
Keywords:
Concrete column
generic reinforced concrete columns subjected to blast loads is investigated numerically. Three-dimen-
Spall damage sional numerical models are developed to predict the concrete spalling under blast loads. The accuracy of
Damage curve the numerical simulations is verified with blast testing data reported by other researchers. Intensive
numerical simulations are then carried out to investigate the influences of the column dimensions and
reinforcement mesh on concrete spall damage. Based on numerical simulation data, empirical relations
are suggested to predict concrete spall damage based on explosion scenarios, column dimensions and
reinforcement conditions.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction the trapped impulse is large enough to overcome the resistant


forces such as the bond, shear around the periphery of the cracked
The analysis of concrete structures against short duration dy- portion, and the mechanical interlocking, the cracked off parts will
namic loads, such as those induced by air blast detonation, is a topic separate from the backside of the structure at some velocity. Even
of extensive study in recent several decades. The dynamic analysis though it may not generate obvious overall structural failure, spall
of the structural performance is a complex issue because short and fragmentation can bring threat to the personnel and equip-
duration high amplitude blast loading often exhibits strong time ment around the area. Such damage is usually not considered in
variation which results in a varying strain rate for the concrete normal protective designs of concrete structures. In extreme cases,
material as well as the steel reinforcement. spall damage can cause severe damage and lead to significant
Reinforced concrete components are commonly used in the reduction on load-carrying capacities of the structural element
protective design against potential blast loading conditions. When although the damage could be limited to a small area. Therefore
concrete components are subjected to close-in blast loads, local investigation upon such damage is important and essential.
failure can be observed. On the surface facing the detonation, the Theoretical analysis on the spall damage is quite complex.
concrete experiences triaxial compression and may fail under high Concrete spall is dependent on the stress wave shapes, magnitudes,
compressive force. When the compressive shock wave propagates and angles of incidence of the applied blast loads and bomb frag-
within the concrete and interacts with the free surface, it will be ment impacts. Spall is also dependent upon the stress change
reflected and converts to a tensile wave. Under this condition, due during the stress wave propagation and the dynamic properties of
to the low tensile resistance of concrete, crack will form if the net the concrete material under high strain rate. In order to give ac-
stress exceeds concrete dynamic tensile strength. Furthermore, if curate spall damage analysis, the resistant force from the material
dynamic bond, shear and mechanical interlock should also be
carefully considered. Until now, there are still uncertainties about
* Corresponding author.
these parameters. The theoretical analysis methods discussed in
E-mail addresses: 20675901@student.uwa.edu.au, alexsolomen@gmail.com Refs. [1,2] were based on some simplified assumptions and their
(J. Li). application scope is limited to light and moderate spall damage. In

http://dx.doi.org/10.1016/j.ijimpeng.2014.02.001
0734-743X/Ó 2014 Elsevier Ltd. All rights reserved.
42 J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55

the research work conducted by McVay [3], a theoretical method Generally speaking, correlation of the experimental data with
for determining if and where a stress wave, induced by air blast predictive method is difficult. It was also concluded that spall
with peak reflected pressures below the HEL (Hugoniot Elastic damage from large scale detonations is worse than would be pre-
Limit) of the concrete, would cause the concrete to crack in tension dicted based on spall damage from small-scale detonation tests
was derived. In his method, the changes in the stress caused by [13], because of the different strain rates causing different dynamic
stress waves travelling at different velocities, wave attenuation, and strengths in the different scale tests.
dispersion were neglected. The only change in the stress wave Numerical study is another commonly used method for inves-
propagation that was taken into consideration was divergence. As a tigating the structural dynamic response to blast loads. With veri-
result spall damage might not be accurately predicted because the fied material model and element model, numerical simulation
forces resisting the concrete spall behind the crack could not be could be powerful supplement for the experimental tests and it is
accurately obtained. widely used for predicting responses and damage of different
Compared with the theoretical methods, experimental study can structures under various blasting scenarios.
provide valuable data for locating damage and establishing the Nowadays with advanced computer hydro-code such as LS-
damage criteria. Dennis et al. [4] conducted a series of experimental DYNA [14] which contains specialized contact algorithm and
tests to determine the response of a one-way ¼ scale CMU wall to the erosion techniques, it is possible to reproduce the blast scenario
detonation of an explosive charge. Ohtsu et al. [5] experimentally and investigate the structural dynamic response through high fi-
and analytically investigated the dynamic failure of fibre-reinforced delity three-dimensional FE models. Zhou et al. [15] employed a
concrete (FRC) slabs, and it was observed that the averaged di- dynamic plastic damage model for concrete material to estimate
ameters and the volumes of the spall failure remarkably decreased responses of an ordinary reinforced concrete slab and a high
with the increase in the flexural toughness of FRC concrete. Ghani strength steel fibre concrete slab subjected to blast loading. Later,
et al. [6] investigated the blast loading response of reinforced con- Zhou and Hao [16] conducted mesoscale modelling and analysis of
crete panels with externally bonded GFRP laminates. In their ex- damage and fragmentation of concrete slab under contact deto-
periments, blast wave characteristics, including incident and nation. Using meshfree method, Rabczuk et al. [17] investigated the
reflected pressures and impulses, as well as panel central deflection dynamic concrete failure. A model which can capture important
and strain in steel and on concrete/FRP surfaces were measured. All effects when concrete is subjected to high dynamic loading con-
these panels were inspected for cracking, spalling and scabbing. It ditions and large deformations is proposed. Comparisons between
was determined that overall the GFRP retrofitted panels performed the experimental tests and numerical simulations were made and
better than the companion control panels without GFRP retrofit. good agreement was observed. Xu and Lu [18] studied general
Leppänen [7] conducted experimental and numerical analyses of behaviour of concrete plates subjected to air blast loading using
blast and fragment impacts on concrete, and the depths of pene- numerical approach and particularly focused on the spall damage.
tration and spalling were measured in these tests. Empirical spallation criteria considering the three-dimensional
Recently, Wang et al. [8] conducted experimental study on ex- concrete response were proposed. Uenishi et al. [19] developed a
plosion resistance of a one-way square reinforced concrete slab full three-dimensional finite difference code to simulate the wave
under a closed-in explosion. In their tests, six slabs (in two groups) and fracture propagation within solids and it was used to study the
with different scale-down factors were investigated, and two major blast demolition of RC concrete structures. The reliability of the
damage levels, i.e., spallation with a few cracks and moderate code was demonstrated through a demolition analysis of a road
spallation were identified. In the subsequent study, they numeri- bridge pier.
cally modelled the spallation process and the numerical results It is noticed from the above literature review, studies of spall
were compared with the test results [9]. Yamaguchi et al. [10] damage are mainly focused on concrete slabs. RC column is another
performed experimental study to investigate the applicability of important structural component, but the spall damage investiga-
polyethylene fibre reinforced concrete (PEFRC) for use in blast- tion on the reinforced concrete column is rather limited. Although
resistant RC structures. It was shown that PEFRC was effective in spall damage is a local phenomenon that depends only on the
reducing the spall damage to contact detonation as compared with structural material properties, dimensions and reinforcement
normal concrete. Moreover, an equation for estimating the damage confinement, the available methods for predicting concrete slab
depth of the PEFRC slab subjected to contact detonation was spall damage might not be applicable to predict concrete column
derived based on the test results. spall damage. In the present study, a numerical investigation is
In order to properly define and evaluate the spall damage of carried out to study the spall damage of concrete columns sub-
concrete structures, an extensive review of spallation process and jected to air blast loading. The finite element model is calibrated
compiled experimental data was presented by McVay [3]. Based on first through simulation of a previous experimental test reported in
the review, concrete spall damage levels were classified and the the literature [8] and comparison of the numerical and experi-
respective prediction curves for concrete damage to blast loads mental data. Intensive simulations are then carried out using the
were derived. In UFC guideline [11], data from spall tests have been calibrated numerical model to investigate the influences of
compiled and damage curves are given to predict the concrete spall different column and explosion parameters on RC column spall
damage. In these tests, a cylindrical charge, cased or bare, is ori- damage. Based on the numerical simulation results, analytical
ented side-on at a standoff distance from a wall slab and oriented formulae are developed to predict spall damage of concrete
end-on in contact with the ground. Recently Foglar and Kovar [12] columns.
plotted their experimental results on these spall and breach pre-
diction curves, and they concluded that the observed spall damages 2. Numerical model calibration
in RC specimens agree with the spall and breach prediction curves
according to UFC guideline. However, they also noted the spall and In the present study, the numerical simulations are carried out
breach prediction curves according to UFC are not suitable for in commercial hydro-code LS-DYNA 971. LS-DYNA, which is based
predicting the spall damage in fibre reinforced concrete. Moreover, on explicit numerical algorithm, has been widely used in the dy-
the spall damage severity is not clearly defined in UFC guideline. namic simulations of RC structures to blast loads, and it has been
Therefore it can only predict the occurrence of spall damage in the proven yielding reliable numerical predictions of structural dy-
wall slab under a blast load, but cannot quantify the damage levels. namic responses.
J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55 43

Fig. 1. Experiment setup [8].

To validate the accuracy and reliability of the numerical 2.1. Numerical model
model developed for simulating concrete structure response and
damage to blast load, a reinforced concrete slab tested by other 2.1.1. Material model
researchers is modelled first according to the testing conditions In LS-DYNA, a number of material models can be used to
[8]. The dynamic response of the slab to the blast loads is simulate the dynamic behaviour of concrete materials, such as
simulated. The results are compared with those recorded in the Pseudo Tensor (MAT_16), Brittle Damage (MAT_96), Johnson
test [8] to verify the numerical model. The modelled slab is a Holmquist Concrete (MAT_111) and Concrete Damage Rel3
one-way square reinforced concrete slab subjected to close-in (MAT_72_REL3). In the present study, Concrete_Damage_Rel3 is
blast loadings. This slab is selected in this study to calibrate used. It is a plasticity-based model, and it uses three shear failure
the numerical model only because its data is readily available in surfaces which change shape, depending on the confinement
the open literature. Ideally testing data on RC columns should be pressure. The damage and strain rate effect is included in this
used to calibrate the model since the model will be used to model. The major advantage of this model is that it is based on a
simulate spall damage of RC columns. Unfortunately, testing data single user input parameter, i.e., the unconfined compressive
of RC column spall damage are not available in the open litera- strength. The remaining model parameters are automatically
ture. It should be noted that spall damage is a local phenomenon. generated using a built-in algorithm and can also be modified by
It depends on concrete material properties, reinforcement con- the user.
finements, and cross-section thickness. It does not depend on In this material model, the stress tensor is expressed in com-
structural type and boundary conditions as will be numerically bination of the hydrostatic stress tensor and the deviatoric stress
proven in this study. Moreover, a calibrated numerical model tensor. The hydrostatic tensor changes the concrete volume and the
that can reliably predict RC slab damage should also be able to deviatoric stress tensor controls the shape of deformation. For the
predict RC column damage. Therefore the calibrated model based hydrostatic stress tensor, the compaction model is a multi-linear
on testing data on a RC slab is considered reliable for simulating approximation of internal energy. Pressure is defined as:
RC column responses.
The test setup of the RC slab is shown in Fig. 1. Both the length
and width of the slab are 1250 mm, and the thickness is 50 mm. The p ¼ cð3 v Þ þ gTð3 v ÞE (1)
reinforcements within the specimen are constructed using 6 mm
diameter bar. The slab is reinforced in two directions and the dis- where E is the internal energy per initial volume, g is the ratio of
tance between each bar is 75 mm. The reinforcement ratio in the specific heats. The volumetric strain 3 v is given by the natural log-
both directions is r ¼ 1.43%. The cylindrical TNT charge of 0.64 kg arithm. As shown in Fig. 2, when tension stress is greater than the
was detonated at a standoff distance of 0.5 m above the specimen in hydrostatic tension cut-off, tensile failure occurs. When the volu-
the test. metric strain exceeds the elastic limit, compaction occurs and the
The concrete has an average compressive strength of 39.5 MPa, concrete becomes a granular kind of material. The bulk unloading
as measured using three normal 150 mm  150 mm  150 mm modulus is a function of volumetric strain. Uploading occurs along
concrete cubes; a tensile strength of 4.2 MPa; and a Young’s the uploading bulk modulus to the tension cut-off. Reloading al-
modulus of 28.3 GPa. The reinforcement has the yield strength of ways follows the unloading path to the point where uploading
600 MPa and a Young’s modulus of 200 GPa. A steel frame was built began and continues on the loading path.
on the ground to support the slab on two sides as shown in Fig. 1. For the deviatoric stress analysis, a three-curve model is adop-
The RC slab is clamped on the two sides to prevent uplifting during ted. As shown in Fig. 3, the upper curve represents the maximum
the test. Wooden bars of the same width and length are placed strength curve, the middle curve is the initial yield strength curve
between the specimen and the frame, to provide uniform sup- and the lower curve is the residual strength curve for the failed
porting conditions and prevent direct impact damage on the concrete material.
specimen edges. More detailed information of the tested slab can In LS-DYNA, this material model is used in conjunction with an
be found in Ref. [8]. equation of state EOS_TABULATED_COMPACTION, which gives the
44 J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55

8  
> d
>
< 3_
for 3_  1 s1
ft 3_ s
DIF ¼ ¼  1=3 (3)
fts >
>
: b 3_ for 3_ > 1 s1
3_ s

where ft is the dynamic tensile strength at 3_ ; fts is the static tensile


strength at 3_ s ; 3_ is the strain rate in the range of 106 to 160 s1; 3_ s is
the static strain rate 106 s1; log b ¼ 6d-2; d ¼ 1/(1 þ 8 fcs/fco);
fco ¼ 10 MPa.
For reinforcement:
 a
3_
DIF ¼ (4)
104

where for the yield strength, a ¼ afy ¼ 0.074-0.04 fy/60; and for the
Fig. 2. Pressure versus volumetric strain.
ultimate stress, a ¼ afu ¼ 0.019-0.009 fy/60.
The above material properties have been widely used to model
RC structures including both the slabs and columns. Previous
current pressure as a function of the volumetric strain as defined in studies [23e26] have shown that they gave reliable predictions of
Eq. (1). structural responses to blast loads.
For the steel reinforcement, the material model Piecewise Linear
Plasticity (MAT_24) with strain rate effects is used. This model al- 2.1.3. Blast load modelling
lows the definition of arbitrary stress versus strain curve and In the present numerical simulations, the blast load is simu-
arbitrary strain rate curve. Also, failure based on a plastic strain or a lated with the Load_Blast function in LS-DYNA. This function is
minimum time step size can be defined. In the present study, the developed based on a report by Randers-Pehrson and Bannister
input parameters for this material model are the steel yield [27] and is mainly based on empirical relations derived from
strength, Young’s modulus and the stain rate curve that will be blasting tests. The utilization of this function avoids the detailed
defined below. modelling of the explosive charge and shock wave propagation in
air, thus it can save the calculation cost. The disadvantage of this
function is that it cannot model the shock wave and structure
2.1.2. Strain rate effect interaction. The reliability of this function in simulating blast
The strain rate effect is taken into consideration in this study. loads on structures has been proven and it is very commonly
Dynamic Increase Factor (DIF) for the concrete compressive used in numerical simulations of structural responses to blast
strength is defined according to the empirical function provided by loads [28].
CEB-FIP Model Code 1990 [20], that for the concrete tensile
strength is defined according to [21], and that for steel according to 2.1.4. Finite element model and erosion algorithm
[22]. In this study, solid element is used to model concrete and
For concrete compressive strength: reinforcement bars are modelled by beam elements. In order to
capture the localized damage with high fidelity, the mesh size
8 
> 1:026a used for the concrete and beam element is 5 mm, and the total
>
< 3_3_s for 3_  30 s1
fc number of elements is 630,000 for the tested slab. This mesh size
DIF ¼ ¼  1=3 (2)
fcs >
> is determined by conducting the convergence test. It is found that
: g 3_ for 3_ > 30 s1
s 3_ s the use of smaller elements for concrete and reinforcement re-
sults in similar simulations but increases the calculation time
where fc is the dynamic compressive strength at 3_ ; fcs is the static enormously. In order to simulate the physical fracture, shear
compressive strength at 3_ s ; 3_ is the strain rate in the range of failure, cratering, spalling, and crushing of the concrete material
30  106 to 300 s1; 3_ s is the static strain rate 30  106 s1; log under the blast or impulsive loads, the erosion algorithm is
gs ¼ 6.156 a 2; a ¼ 1/(5 þ 9 fcs/fco); fco ¼ 10 MPa. usually adopted. In the explicit simulation, when the dynamic
For concrete tensile strength: response of an element reaches a critical value defined by the

Fig. 3. Strength model for concrete material.


J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55 45

Fig. 4. Comparison of the numerical and experimental results of the RC slab damage contour (a) experimental results of the upper surface; (b) numerical results of the upper
surface; (c) experimental results of the bottom surface; (d) numerical results of the bottom surface.

user, the element is immediately erased from the model. The 2.2. Model calibration and discussion
advantage of this function is to smooth the simulation and
simulate the element failure which is not defined for some solid According to the experimental study that is used to calibrate the
element models in LS-DYNA. However, it is worth noting that, present numerical model, the blast load is generated from a 0.64 kg
such erosion process is irreversible, which means if the user TNT explosion at a standoff distance of 0.5 m. According to the UFC
defined criteria is set too low, the conservation of mass and en- code, the peak pressure in this blast scenario is 20.8 MPa, while in
ergy will not be maintained and the simulation results are no the numerical simulation, the predicted peak pressure is 21.0 MPa.
longer reliable. In the previous study, the erosion algorithm has The two results agree well.
been widely used in simulating the concrete response under high Fig. 4 shows the comparison of the damaged areas obtained
velocity and low velocity impacts [29e32] and blast loads [33e from the numerical simulation and blasting tests. The numerically
35]. Considering the strain rate effect on the concrete tensile simulated plastic strain contour is shown for comparison. As dis-
strength, Xu and Lu [18] adopted an erosion criterion based on played in Fig. 4a, no clear compressive crushing damage can be
the principle strain. It has been proved that with this erosion observed on the upper surface of the concrete slab except a small
criterion, the spall damage of concrete materials can be area at the slab centre. However, a number of concrete cracks are
adequately simulated. Other researchers used different erosion obvious. The numerical simulation well captures these damages as
criteria. For example, based on intensive numerical simulation of shown in Fig. 4b. Fig. 4c and d shows the slab rear surface. As shown
blast damage to a RC bridge structure, Tang and Hao [34] sug- some extensive cracks as well as spall damage can be clearly
gested a criterion based on both the tensile strength and prin- observed in the experimental study. In the numerical simulation,
cipal strain of concrete material. In this study, intensive the spall damage is also well captured, and the major cracks are also
simulations are also carried out with different erosion criteria, it reasonably reproduced.
is found that using principle tensile strain of 0.01 as the erosion The residual deflection at the slab centre recorded in the test is
criterion, which is also used in Ref. [18], leads to reliable pre- 19.0 mm, and the numerical simulation gives a value of 19.7 mm.
dictions of RC slab responses. The spall radius in the test is about 80 mm while the numerical
46 J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55

Table 1
Spallation classification.

From no
change in the column condition to a
few visible cracks; From a few cracks
Threshold spall damage and a hollow sound to a large bulge
in the concrete with a few small piece on the floor.

Medium damage From a very shallow spall


to a third of the column thickness.

Severe spall damage From just over one third the column thickness to breach

simulation gives a radius about 75 mm. These results demonstrate three categories as displayed in Table 1. In the numerical
that the above numerical model leads to reasonable predictions of simulations conducted in the following sections, the spall
the structural response to blast loads. damage in the RC column is similarly classified according to
these criteria.
3. Numerical simulations

The above calibrated numerical model is used in this study to


simulate dynamic responses of RC columns to blast loads. It should
be noted that although the model is calibrated based on test data of
a RC slab, it is used here to simulate RC column responses. This is
acceptable because 1) spall damage is a local damage depending on
the concrete material properties, thickness and reinforcement
confinement, but relatively independent of the global structural
response modes and boundary conditions; 2) material models of RC
slab and RC columns are the same.

3.1. Spall damage classification

According to the research work conducted by McVay [3], the


concrete spallation damage can be classified into the following

Table 2
Material model in LS-DYNA.

Material LS-DYNA model Input parameters Magnitude

Concrete MAT_Concrete_ Concrete strength 40 MPa


Damage_Rel3
Erosion criterion Principle tensile strain 0.01
Reinforcement MAT_Piecewise_ Mass density 7800 kg/m3
Linear_Plasticity Elastic modulus 2.00Eþ11
Fig. 5. Typical “threshold spall damage” scenario: (a) 2.4 m column 5 kg TNT @ 0.5 m;
Yield stress 335 MPa
(b) 3 m column 5 kg TNT @ 0.5 m; (c) 4.5 m column 5 kg TNT @ 0.5 m.
J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55 47

Fig. 8. Critical charge weight-standoff distance for the state of “threshold spall
damage”.

of the structural stiffness, a previous study on RC slab spall damage


revealed possible interactions between local and global responses
[5]. To investigate if RC column spall damage also depends on its
flexural stiffness, RC columns of different heights are therefore
Fig. 6. Typical “medium spall damage” scenario: (a) 2.4 m column 1 kg TNT @ 0.2 m;
(b) 3 m column 1 kg TNT @ 0.2 m; (c) 4.5 m column 1 kg TNT @ 0.2 m.
considered in the study. Column width is not considered as a
parameter in the analysis because blast pressure is assumed to be
uniformly distributed cross the column width. Increasing the col-
3.2. Column configurations
umn width simultaneously increases the blast loads acting on the
column, therefore column width is not considered as a variable in
In the following numerical simulations, generic reinforced
this study. Reinforcement mesh is another parameter that affects
concrete columns of H (height)  D (depth)  W (width) with two
the RC column spall damage because it provides confinement to
fixed ends are considered. In the simulations, the column width is
concrete. In the present study the influence of the reinforcement
kept to be 300 mm, while three different heights with
spacing is investigated by considering columns with different
H ¼ 2400 mm, 3000 mm and 4500 mm, and three different column
number of longitudinal and hoop rebar in numerical simulations. It
depths with D ¼ 200 mm, 300 mm and 400 mm are considered to
should be noted that spall damage is usually generated by blast
investigate the influences of column flexural stiffness and thickness
loading induced stress waves propagating in the structure. It occurs
on spall damage. The selection of varying these parameters in the
before the structure experiences large global (shear and flexural)
study is because column thickness obviously affects the spall
responses. Therefore deformations of the reinforcement bars are
damage. Although spall damage induced by a close-in detonation is
usually small when spall damage occurs. The function of the rein-
a localized phenomenon and is commonly considered independent
forcement mesh is mainly to provide confinement to concrete
material. For this reason, unlike to global structural responses, the
spacing between reinforcement bars is a more important param-
eter than reinforcement ratio to spall damage. To concentrate on
investigating the confinement effect from reinforcement mesh on
concrete, in numerical simulations, the longitudinal and stirrup
reinforcement ratios are kept to be approximately 1.5% and 0.2%,
but different numbers of reinforcement bars with different sizes are

Fig. 7. Typical “severe spall damage” scenario: (a) 2.4 m column 1 kg TNT @ 0.1 m; (b)
3 m column 1 kg TNT @ 0.1 m; (c) 4.5 m column 1 kg TNT @ 0.1 m. Fig. 9. Critical charge weight-standoff distance for the state of “medium damage”.
48 J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55

Fig. 10. Critical charge weight-standoff distance for the state of “severe spall damage”.

considered because more bars with smaller diameter and spacing


between any two of them provide better confinement than less Fig. 12. Similar spall damage in three columns of different depths corresponding to
bars with larger diameter although the reinforcement ratio is different blast scenarios: (a) 4 kg @ 0.3 m; (b) 5 kg @ 0.3 m; (c) 6 kg @ 0.3 m.
approximately the same. It should be noted that in this study the
reinforcement mesh is modelled in detail by using beam elements.
3.3. Results and discussion
In the section discussing the influence of the reinforcement layouts
on spall damage, the dimension of the reinforcement is rounded to
3.3.1. Columns with different heights
an integer. This makes the reinforcement ratio not exactly the same
In this section, numerical simulations are conducted to study
in the cases considered, but they are all approximately the same to
the effect of column flexural stiffness by varying its height on
study the influences of reinforcement layouts on confining the
concrete spall damage. All the columns considered have the same
concrete.
depth and width of 300 mm. There are four B20 mm longitudinal
It should also be noted that in the present study, the explosives
rebar placed on the four corners of the cross-section, the stirrup
are placed at locations directly face the mid-span of the column in
reinforcement rebar has a diameter of 10 mm, and spaced at
all the cases considered. To consider the influences of axial loading
300 mm between each stirrup reinforcement along the column
in the column, prior to the application of blast load, axial load from
height. The concrete cover thickness is 25 mm.
the upper structure is applied to the column. The axial load is
Fig. 5 shows typical numerical simulation results (plastic strain
assumed to be 20% of the column loading capacity to represent the
contour) corresponding to the state of “threshold spall damage”. It
actual loading condition.
can be observed that, under the same blast loading condition of
The material properties used in the simulations are the same as
5 kg TNT @ 0.5 m, the spall depth and length are almost the same in
those used to model the RC slab and are listed in Table 2.
these three columns.
Fig. 6 shows the state corresponding to the “medium spall
damage”. Again it can be observed that the spall depth and frag-
mentations at the end of the simulation time are quite similar in
these three columns. The spall depth is about 1/3 of the column
thickness.
Fig. 7 shows the numerical simulation results corresponding to
the state of “severe spall damage”. In these cases, perforation is
about to occur owing to significant spall and crushing damage. It
can be noted that even under such a severe damage scenario, the
failure is still highly localized although such failures may lead to
total loss of load carrying capacities of the column.
From the above numerical simulation results, it can be
concluded that the blast induced concrete spall damage is inde-
pendent of the column height or flexural stiffness. This is because
the concrete spall damage induced by close-in detonations occurs
very fast before the column experiences any significant structural
responses. Therefore the structural stiffness and boundary condi-
tions have little effect on localized spall damage.
Fig. 8 shows the results obtained from the numerical simula-
tions corresponding to the state of “threshold spall damage” with
respect to the explosive weight and standoff distance. The best
fitted curve which is derived using DATAFIT [36] is also given in the
figure. This curve defines the boundary of occurrence of spall
damage. If the standoff distance and explosion weight combination
Fig. 11. Column response under 1 kg TNT explosion @ 0.2 m: (a) 200 mm depth col- falls on the left and above the curve, no spall damage, otherwise
umn; (b) 300 mm depth column; (c) 400 mm depth column. spall damage occurs. In this plot, the explosive weight more than
J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55 49

Fig. 13. Critical charge weight-standoff distance relation corresponding to the threshold spall damage state for columns with different depths.

Fig. 14. Critical charge weight-standoff distance relation corresponding to the medium spall damage state for columns with different depths.

Fig. 15. Critical charge weight-standoff distance relation corresponding to the severe spall damage state for columns with different depths.
50 J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55

Fig. 16. Reinforcement of the columns (a) longitudinal rebar with 210 mm spacing; (b)
longitudinal rebar with 100 mm spacing; (c) longitudinal rebar with 65 mm spacing.

Fig. 18. Column response under a 2 kg TNT explosion @ 0.2 m: (a) longitudinal rebar
15 kg is not included. This is because with an explosive weight more with 210 mm spacing; (b) longitudinal rebar with 100 mm spacing; (c) longitudinal
rebar with 65 mm spacing.
than 15 kg, the corresponding critical standoff distance for
threshold spall damage also increases, and the blast induced col-
umn response is no longer dominated by localized spall damage. In
simulated data and the best fitted curve for the state of the “me-
such cases, global flexural and/or shear failure of RC columns are
dium spall damage”.
the more critical damage modes. Such damage modes are not
The best fitted damage curve for “medium spall damage” can be
investigated in the present study.
expressed as
The best fitted curve for the threshold spall damage of RC col-
umns under consideration is y ¼ 0:1110 þ 0:1048  w0:7509 ð0:5 kg  w < 15 kgÞ (6)

y ¼ 0:1482 þ 0:1264  w0:6959 ð0:5 kg  w < 15 kgÞ (5) Fig. 10 shows the simulated data and the bested fitted curve for
the state of the “severe spall damage”.
Similarly, based on intensive numerical simulations, the data The best fitted damage curve can be expressed as
corresponding to the other two spall damage levels and the cor-
responding best fitted curves are also derived. Fig. 9 shows the y ¼ 0:3085 þ 0:4014  w0:2799 ð0:5 kg  w < 15 kgÞ (7)
These curves can be used to predict spall damage of RC columns
having similar dimensions and reinforcement conditions as
considered in the present study.

3.3.2. Columns with different depths


In this section, numerical simulations are conducted to study
the effect of column depth on spall damage. In these case studies,
the column height is 3000 mm and width is 300 mm. The concrete
cover thickness is 25 mm. The reinforcement ratio and arrange-
ment remain the same as defined in the last section.
Fig. 11 shows the structural response of the three columns under
the same blast loading scenario (1 kg TNT @ 0.2 m). It can be easily
observed that with the increase of the column depth, less spall
damage is experienced by the column. For column with 200 mm
depth, more than one third of the column depth has spall damage,
corresponding to the “severe spall damage” category. For column
with 300 mm depth, a shallow spall area forms under the blast
loads, and the damage level more closely matches the “medium
spall damage” according to the above definition. For column with
400 mm depth, only a few visible cracks can be observed on the
column rear surface, and the damage level fits in the “threshold
spall damage” category.
Fig. 12 shows three similar structural spall damage cases which
Fig. 17. Column response under a 3 kg TNT explosion @ 0.3 m: (a) longitudinal rebar
fit closely to the severe spall damage category. In these three cases,
with 210 mm spacing; (b) longitudinal rebar with 100 mm spacing; (c) longitudinal about half of the column depth experiences the spall damage. In
rebar with 65 mm spacing. case one, the blast scenario is 4 kg @ 0.3 m; in case two, the blast
J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55 51

Fig. 19. Critical charge weight-standoff distance relation corresponding to the threshold spall damage state for columns with different longitudinal reinforcement layouts.

Fig. 20. Critical charge weight-standoff distance relation corresponding to the medium spall damage state for columns with different longitudinal reinforcement layouts.

Fig. 21. Critical charge weight-standoff distance relation corresponding to the severe spall damage state for columns with different longitudinal reinforcement layouts.
52 J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55

Fig. 14a shows the numerical results corresponding to the state


of “medium spall damage” for the three columns with different
depths. The best fitted surface is shown in Fig. 14b and its analytical
expression is

y ¼ 2:09  D0:446  w0:609 ð200 mm  D < 400 mm; 0:5 kg


 w < 15 kgÞ
(9)
The results corresponding to the severe spall damage are shown
in Fig. 15. The best fitted surface can be expressed as:

y ¼ 1:871  D0:481  w0:566 ð200 mm  D


< 400 mm; 0:5 kg  w < 15 kgÞ (10)

From the results shown above, it is obvious that a thicker col-


umn experiences less spall damage under the same explosion
condition. This result is expected because an increase in the column
depth results in an increase in the distance the stress waves have to
Fig. 22. Reinforcement cages of the columns with different stirrups, (a) B10 mm @ travel before they can cause spall damage to the RC column. The
0.3 m spacing; (b) B8 mm @ 0.2 m spacing; (c) B7 mm @ 0.15 m spacing.
further the stress waves travel, the more their magnitudes and
shapes may change. These changes are due to the attenuation,
scenario is 5 kg @ 0.3 m; and in case three, the blast scenario is 6 kg divergence, and dispersion.
@ 0.3 m. It can be generally concluded that in order to achieve the
same spall damage level, larger blast load is needed for thicker
columns. 3.3.3. Columns with different longitudinal reinforcement spacing
Fig. 13a shows the numerical results corresponding to the state Besides the column dimensions, the reinforcement of the col-
of the “threshold spall damage”. The best fitted boundary surface is umn could also have significant influence on the spall damage
shown in Fig. 13b, and the corresponding equation is given below. because reinforcement cage provides confinement to the concrete.
In this section, three columns with approximately the same per-
y ¼ 2:77  D0:428  w0:494 ð200 mm  D < 400 mm; 0:5 kg centage of reinforcement ratio (1.5% for longitudinal reinforcement
and 0.2% for stirrup reinforcement), but different sizes of longitu-
 w < 15 kgÞ dinal reinforcing bars and spacing between the longitudinal bars,
(8) are numerically tested. The columns are constructed by C40 con-
crete and reinforcement with yield strength of 335 MPa. The col-
where D is the column depth. umn height is 3000 mm while its depth and width are 300 mm. The
It can be observed that under the same explosion condition, the stirrup reinforcement rebar has a diameter of 10 mm, spaced at
standoff distance decreases as the column depth increases in order 300 mm along the column height. The concrete cover is 25 mm the
to generate the same spall damage. column is fixed at the both ends. The reinforcement of the three
columns is displayed in Fig. 16. They are: a) 4B20 mm longitudinal
rebar with spacing between bars 210 mm; b) 6B16 mm longitu-
dinal rebar with spacing between bars 100 mm; and c) 8B14 mm
longitudinal rebar with spacing between bars approximately
65 mm. Fig. 16 shows the reinforcement cage of the three columns
considered in the study.
Fig. 17 shows the results from a 3 kg TNT explosion at 0.3 m
standoff. It can be clearly observed that with the increase of the
number of longitudinal rebar and decrease of the rebar spacing, the
level of spall damage reduces, indicating the confinement of rebar
mitigates concrete spall damage. In the column with 210 mm lon-
gitudinal rebar spacing, the spall depth is about 18 cm, whereas the
spall depth in the column with 100 mm longitudinal rebar spacing
and 65 mm longitudinal rebar spacing reduce to 15 cm and 13 cm,
respectively.
Results from 2 kg TNT detonated at 0.2 m from the column is
shown in Fig. 18, similar observations can be drawn. In the column
with 210 mm longitudinal rebar spacing, the spall depth is about
19 cm, whereas the spall depth in the column with 100 mm lon-
gitudinal rebar spacing and 65 mm longitudinal rebar spacing are
16 cm and 14 cm, respectively.
The reason that more reinforcement bars gives better spall
mitigation can be attributed to that the longitudinal reinforcement
Fig. 23. Response of columns with different stirrup reinforcements under a 2 kg TNT
with close spacing causes more significant wave dispersion of the
explosion @ 0.2 m, (a) B10 mm @ 0.3 m spacing; (b) B8 mm @ 0.2 m spacing; (c) stress waves, and thus reduces the peak net stress which can
B7 mm @ 0.15 m spacing. generate spall. Moreover, when the spall damage happens, the
J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55 53

Fig. 24. Numerical results and the best fitted boundary surface for threshold spall damage of RC columns with different stirrup spacing.

Fig. 25. Numerical results and the best fitted boundary surface for medium spall damage of RC columns with different stirrup spacing.

Fig. 26. Numerical results and the best fitted boundary surface for severe spall damage of RC columns with different stirrup spacing.
54 J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55

cracked off concrete must break into smaller parts to pass through
the small spaced reinforcement bars. The smaller spaced rein- y ¼ 0:238  Sh 0:134  w0:532 ð0:1 m  Sh  0:3 m; 0:5 kg
forcement bars provides more uniform resistance force to better  w < 15 kgÞ
confine the concrete. These results indicate that it is better to use
(15)
more numbers of smaller size rebar than less numbers of larger size
rebar in structural design. This observation is consistent with the The corresponding results for the “severe spall damage” state
common understanding that more numbers of small-size rebar are shown in Fig. 26a. The best fitted boundary surface is shown in
provides better confinement and crack control of concrete in RC Fig. 26b, and its analytical expression is
structures. However, using more numbers of small-size rebar in-
creases the construction cost. Therefore in a real situation a detailed y ¼ 0:148  Sh 0:164  w0:587 ð0:1 m  Sh  0:3 m; 0:5 kg
analysis is needed to find the optimal solution.
 w < 15 kgÞ
For the three columns with different spacing of longitudinal
rebar (Sl), the numerical results corresponding to the state of (16)
“threshold spall damage” are shown in Fig. 19a. The best fitted
boundary surface which is shown in Fig. 19b can be expressed as
4. Conclusions
y ¼ 0:112  Sl 0:155  w0:481 ð65 mm  Sl  210 mm; 0:5 kg
This paper presents a 3D numerical model to predict spall
 w < 15 kgÞ
damage of RC columns. The validity of the numerical model is
(11) calibrated against blasting test. Intensive numerical simulations are
carried out to investigate the influence of RC column parameters on
The corresponding results for the “medium damage” state are
spall damage under close-in explosion loads. It is found that spall
shown in Fig. 20a. Fig. 20b shows the best fitted boundary surface,
damage generated by close-in explosion is limited to a local area
and it can be expressed as
and is insensitive to the column stiffness and boundary conditions.
However, column depth and reinforcement layouts have profound
y ¼ 0:0747  Sl 0:168  w0:568 ð65 mm  Sl
effects on RC column spall damage. Increasing column depth, and/
 210 mm; 0:5 kg  w < 15 kgÞ (12) or using more dense reinforcement bars that provides better
confinement to concrete are very effective in mitigating spall
Fig. 21a shows the numerical results corresponding to the se-
damage. Based on intensive numerical simulation data, analytical
vere spall damage state. The best fitted boundary surface which is
expressions are derived to predict RC column spall damage in terms
shown in Fig. 21b can be expressed as
of the explosion weight, standoff distance, column depth, and
longitudinal and stirrup reinforcement spacing.
y ¼ 0:045  Sl 0:185  w0:587 ð65 mm  Sl  210 mm; 0:5 kg
 w < 15 kgÞ
References
(13)
[1] Kot CA. Spalling of concrete walls under blast load. SMIRT 4, J10/5. San
Francisco, CA: Nuclear Engineering and Design; 1977.
3.3.4. Columns with different stirrup reinforcement spacing [2] Kot CA, Valentin RA, McLennan DA, Turula P. Effects of air blast on power plant
In this section, three columns with four longitudinal reinforce- structures and components. IL, USA: Argonne National Lab; 1978.
[3] McVay MK. Spall damage of concrete structures. DTIC Document; 1988.
ment bars of reinforcement ratio 1.5% and varying stirrup rein-
[4] Dennis S, Baylot J, Woodson S. Response of 1/4-scale concrete masonry unit
forcement bar sizes and spacing are considered. The stirrup (CMU) walls to blast. J Eng Mech 2002;128:134e42.
reinforcement ratio is approximately 0.2% for all the three columns, [5] Ohtsu M, Uddin FAKM, Tong W, Murakami K. Dynamics of spall failure in fiber
reinforced concrete due to blasting. Constr Build Mater 2007;21:511e8.
other column parameters and material properties are the same as
[6] Ghani Razaqpur A, Tolba A, Contestabile E. Blast loading response of rein-
those given in the last section. The reinforcement cages of the three forced concrete panels reinforced with externally bonded GFRP laminates.
columns are displayed in Fig. 22. Compos Part B Eng 2007;38:535e46.
Fig. 23 shows the results generated by a 2 kg TNT explosion @ [7] Leppänen J. Experiments and numerical analyses of blast and fragment im-
pacts on concrete. Int J Impact Eng 2005;31:843e60.
0.2 m. As shown with the decrease of the stirrup spacing, less [8] Wang W, Zhang D, Lu F, Wang S-C, Tang F. Experimental study on scaling the
cracked off pieces eject from the column, indicating using denser explosion resistance of a one-way square reinforced concrete slab under a
stirrups can mitigate spall damage. close-in blast loading. Int J Impact Eng 2012;49:158e64.
[9] Wang W, Zhang D, Lu F, Wang S-C, Tang F. Experimental study and numerical
Similar to the observations made above on spall damage of simulation of the damage mode of a square reinforced concrete slab under
columns with different longitudinal reinforcement layouts, the close-in explosion. Eng Fail Anal 2013;27:41e51.
closely spaced stirrup reinforcement can provide better confine- [10] Makoto Yamaguchi KM, Takeda Koji, Mitsui Yoshiyuki. An experimental study
on blast resistance of polyethylene fiber reinforced concrete. J Temporal Des
ment to the concrete therefore mitigate spall damage. Archit Environ 2009;9:158e61.
The numerical results corresponding to the state of threshold [11] US Army, US Navy, U.A. Force. Structures to resist the effects of accidental
damage are shown in Fig. 24a. The best fitted boundary surface explosions. US Government Printing Office; 2008.
[12] Foglar M, Kovar M. Conclusions from experimental testing of blast resistance
shown in Fig. 24b can be expressed as
of FRC and RC bridge decks. Int J Impact Eng 2013;59:18e28.
[13] Nash PT, Vallabhan C, Knight TC. Spall damage to concrete walls from close-in
y ¼ 0:303  Sh 0:121  w0:471 ð0:1 m  Sh  0:3 m; 0:5 kg  w cased and uncased? Explosions in air. ACI Struct J 1995;92.
[14] LS-DYNA. LS-DYNA keyword user’s manual, version 971. Livermore, USA:
< 15 kgÞ Livermore Software Technology Corporation; 2007.
[15] Zhou XQ, Kuznetsov VA, Hao H, Waschl J. Numerical prediction of concrete
(14) slab response to blast loading. Int J Impact Eng 2008;35:1186e200.
[16] Zhou XQ, Hao H. Mesoscale modelling and analysis of damage and fragmentation
where Sh is the stirrup reinforcement spacing. of concrete slab under contact detonation. Int J Impact Eng 2009;36:1315e26.
The corresponding numerical results for the “medium damage” [17] Rabczuk T, Eibl J. Modelling dynamic failure of concrete with meshfree
methods. Int J Impact Eng 2006;32:1878e97.
state are shown in Fig. 25a. The best fitted boundary surface which [18] Xu K, Lu Y. Numerical simulation study of spallation in reinforced concrete
is shown in Fig. 25b can be expressed as plates subjected to blast loading. Comput Struct 2006;84:431e8.
J. Li, H. Hao / International Journal of Impact Engineering 68 (2014) 41e55 55

[19] Uenishi K, Takahashi H, Yamachi H, Sakurai S. PC-based simulations of [29] Farnam Y, Mohammadi S, Shekarchi M. Experimental and numerical in-
blasting demolition of RC structures. Constr Build Mater 2010;24:2401e10. vestigations of low velocity impact behavior of high-performance fiber-rein-
[20] Béton CE-Id. CEB-FIP model code 1990: design code. Thomas Telford; 1993. forced cement based composite. Int J Impact Eng 2010;37:220e9.
[21] Malvar LJ, Crawford JE. Dynamic increase factors for concrete. DTIC Docu- [30] Beppu M, Miwa K, Itoh M, Katayama M, Ohno T. Damage evaluation of con-
ment; 1998. crete plates by high-velocity impact. Int J Impact Eng 2008;35:1419e26.
[22] Malvar L, Crawford J. Dynamic increase factors for steel reinforcing bars. In: [31] Lian YP, Zhang X, Zhou X, Ma ZT. A FEMP method and its application in
28th DDESB seminar; 1998. Orlando, USA. modeling dynamic response of reinforced concrete subjected to impact
[23] Li J, Hao H. Influence of brittle shear damage on accuracy of the two-step loading. Comput Methods Appl Mech Eng 2011;200:1659e70.
method in prediction of structural response to blast loads. Int J Impact Eng [32] Nyström U, Gylltoft K. Comparative numerical studies of projectile impacts on
2013;54:217e31. plain and steel-fibre reinforced concrete. Int J Impact Eng 2011;38:95e105.
[24] Li J, Hao H. Numerical study of structural progressive collapse using sub- [33] Coughlin AM, Musselman ES, Schokker AJ, Linzell DG. Behavior of portable
structure technique. Eng Struct 2013;52:101e13. fiber reinforced concrete vehicle barriers subject to blasts from contact
[25] Bi K, Hao H. Numerical simulation of pounding damage to bridge structures charges. Int J Impact Eng 2010;37:521e9.
under spatially varying ground motions. Eng Struct 2013;46:62e76. [34] Tang EK, Hao H. Numerical simulation of a cable-stayed bridge response to
[26] Zhao CF, Chen JY. Damage mechanism and mode of square reinforced blast loads, part I: model development and response calculations. Eng Struct
concrete slab subjected to blast loading. Theor Appl Fract Mech 2013;63e 2010;32:3180e92.
64:54e62. [35] Wu K-C, Li B, Tsai K-C. The effects of explosive mass ratio on residual
[27] Randers-Pehrson G, Bannister KA. Airblast loading model for DYNA2D and compressive capacity of contact blast damaged composite columns. J Constr
DYNA3D. DTIC Document; 1997. Steel Res 2011;67:602e12.
[28] Chen W, Hao H. Numerical study of a new multi-arch double-layered blast- [36] User manual, DataFit. Oakdale, PA, USA: Oakdale Engineering.
resistance door panel. Int J Impact Eng 2012;43:16e28.

You might also like