International Journal of Impact Engineering 75 (2015) 162e173

Contents lists available at ScienceDirect

International Journal of Impact Engineering

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

Experimental and finite element analysis of doubly reinforced

concrete slabs subjected to blast loads
Ganesh Thiagarajan a, *, Anirudha V. Kadambi a, Stephen Robert b, Carol F. Johnson b
a
Department of Civil Engineering, University of Missouri Kansas City, 352 Flarsheim Hall, 5100 Rockhill Road, Kansas City, MO 64110, USA
b
U.S. Army Engineer Research and Development Center, 3909 Halls Ferry Rd, Vicksburg, MS 39180, USA

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

Article history: This paper presents research on the response and behavior of both high strength concrete (107 MPa) and
Received 28 August 2013 normal strength concrete (27.6 MPa) slabs doubly reinforced with high strength low alloy vanadium
Received in revised form (HSLA-V) reinforcement (VR) and conventional steel reinforcing bars (NR) subjected to explosive loads.
24 July 2014
Four types of reinforced concrete (RC) slabs namely High Strength Concrete (HSC) with HSLA-V Steel
Accepted 28 July 2014
Available online 7 August 2014
Reinforcing bars (HSC-VR), High Strength Concrete with Conventional Steel Reinforcing bars (HSC-NR),
Normal Strength Concrete (NSC) with HSLA-V Steel Reinforcing bars (NSC-VR), and Normal Strength
Concrete with Conventional Steel Reinforcing bars (NSC-NR) have been studied and compared both
Keywords:
Blast loading
experimentally and numerically. The slabs were subjected to blast loads using a shock tube capable of
Reinforced concrete slab generating both positive and negative phase pressures. Data collected during the dynamic experiments
Experimental data consisted of reflected pressure obtained from several pressure gages arranged along the perimeter of the
Finite element simulation test article and mid-span deflections captured from an accelerometer, a laser device, and high speed
High strength materials video. The numerical analysis was performed with the commercial program LS-DYNA using two material
models. The concrete material models considered were Winfrith Concrete Model (WCM) and Concrete
Damage Model Release 3 (CDMR3). Results from the numerical simulation are compared with the
experimental values to determine material parameters and other finite element model related con-
straints. Mesh sensitivity and crack propagation studies were also conducted. From this study it was
observed that CDMR3 and WCM can be used over a wider range of concrete compressive strengths. The
advantages and disadvantages of using high strength materials are discussed.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction individual components towards dynamic loadings by providing

more reliable predictions.
Blast and impact events on structures such as the Oklahoma City Experimental and numerical analysis can be performed on steel
bombing in 1995 and the September 11, 2001 attacks, have led reinforced concrete elements. However, blast experimental efforts
researchers to probe into the aspects of making buildings and other require specialized equipment, labor, and could be fairly expensive.
socio economically vital structures strong enough to withstand Numerical analysis of the dynamic behavior of steel reinforced
extreme loadings. Furthermore, it becomes important to under- concrete when subjected to the extreme loadings can be studied
stand the response of reinforced concrete as a structural material using the non-linear finite element software such as ABAQUS® and
when subjected to large stresses and strain rates through explosive LS-DYNA®. LS-DYNA® has a number of features that make it suitable
loadings. The study of the dynamic nonlinear responses of indi- for blast load simulations and has been used in this study.
vidual structural components like beams, slabs, and columns of an
entire building system has recently become an important topic. 2. Objective
Advances in finite element modeling and analysis have further
enhanced interest in studying the behavior and response of these The objective of this research was to numerically and experi-
mentally study the composite response and behavior of both high
strength concrete 107 MPa and normal strength concrete 27.6 MPa
slabs doubly reinforced with high strength low alloy vanadium
* Corresponding author. Tel.: þ1 816 235 1288; fax: þ1 816 235 1260. (HSLA-V) reinforcement and conventional steel reinforcing bars
E-mail addresses: [email protected], [email protected] (G. Thiagarajan). (NR) when subjected to explosive loads. Four types of reinforced

http://dx.doi.org/10.1016/j.ijimpeng.2014.07.018
0734-743X/© 2014 Elsevier Ltd. All rights reserved.
 G. Thiagarajan et al. / International Journal of Impact Engineering 75 (2015) 162e173 163

concrete (RC) slabs namely High Strength Concrete (HSC) with model was not significant due to the large standoff distance. Xu
HSLA-V Steel Reinforcing bars (HSC-VR), High Strength Concrete et al. [12] used the Pseudo-Tensor concrete material model [13]
with Conventional Steel Reinforcing bars (HSC-NR), Normal with a principal strain based erosion criterion to study concrete
Strength Concrete (NSC) with HSLA-V Steel Reinforcing bars (NSC- spallation in reinforced concrete slabs under various blast loading
VR), and Normal Strength Concrete with Conventional Steel Rein- and structural conditions. They obtained numerical relationships
forcing bars (NSC-NR) have been studied and compared both between critical charge-standoff and three different damage
experimentally and numerically. The slab construction and exper- criteria. Tanapornraweekit et al. [14] performed an experiment on
imental work was performed at the U.S. Army Corps of Engineers e an RC slab subjected to 5000 kg of TNT explosive and compared
Engineer Research and Development Center (USACE-ERDC) in numerical simulation results using CDMR3. The maximum deflec-
Vicksburg, MS. The 1652 mm  857 mm  101.6 mm slabs, rein- tion obtained from LS-DYNA was seventeen percent less than
forced longitudinally on the front and back face (referred to as experimental values. Zhou et al. [15] used a plastic damage model
double mat), were subjected to dynamic air blast loads using the to study the dynamic behavior of both RC slabs and steel fiber
Blast Load Simulator (BLS) located in Vicksburg. Pressure gages at concrete slabs and compared the results with explosive tests. They
multiple locations around the perimeter of the test article docu- derived critical charge weight-standoff distance curves for four
mented the reflected pressure, an accelerometer and a laser device different damage levels. Schenker et al. [16] performed full scale
were used to characterize the mid-span deflection, and still photos field tests on concrete slabs in order to validate the model using a
and high speed video were used to document the final damage hydrocode based program. They also studied the effectiveness of
patterns. Results from the experimental study are used to deter- using aluminum foam as a blast energy dissipater but the results
mine the characteristics that improved the performance of the were inconclusive. Hao et al. [17] studied the influence of concrete
slabs subjected to blast loads the most. strength ratio, slab thickness, steel reinforcement ratio on RC slabs
Numerical analysis of the dynamic behavior of steel RC slabs subjected to the blast loading using LS-DYNA and the John-
subjected to the extreme loadings associated with explosive events soneHolmquist model and compared it with a single experimental
can be studied using the non-linear finite element software such as study of an open air near field explosion of 5 kg TNT on an RC slab.
LS-DYNA [1]. The explicit finite element analysis discussed in this Sangi et al. [18] compared the experimental and numerical
document was conducted using LS-DYNA and two concrete ma- behavior of the reinforced concrete slabs with WCM and CDMR3,
terial models, the Winfrith Concrete Model (WCM) [2] and Con- when subjected to drop weights. They observed that the damage
crete Damage Model Release 3 (CDMR3) [3,4]. Results from the pattern and the impact force histories obtained from the WCM
numerical simulation are compared with the experimental values were in agreement with the experimental values. Morales et al. [19]
to determine the primary numerical features, the strengths and have presented a new setup and tested four RC panels subjected to
weaknesses of the models studied, to identify other constraints each open air blast loading and have performed numerical analysis
related to finite element models, and to identify important mate- using the Winfrith Concrete Model and Brittle Damage model for
rial parameters and their effect on the slab response. Mesh sensi- concrete. The authors concluded that reasonably good simulation
tivity studies were performed using a 25.4 mm and 12.7 mm mesh results can be achieved using simplified material models. Thiagar-
size. A 6.35 mm mesh size was used for crack propagation studies ajan et al. [20] performed a preliminary study on the numerical
to qualitatively compare crack development in the RC slab. The simulation of high strength steel and high strength and normal
results indicate that the mesh size plays a significant role in the strength concrete slabs and compared it with experimental results.
predictions. Results presented include a comparison of experi- In that study the CDMR3 showed a poor response for high strength
mental and numerical results including deformation histories, concrete parameters that were generated internally by LS-DYNA.
crack propagation and damage patterns. Conclusions are drawn The work presented here extends the study to compare and
from the numerical and experimental program regarding the ef- contrast the behavior for four different slab combinations and also
ficacies of using conventional and high strength concrete and steel has material parameters for the CDMR3 that have been input
materials. manually based on concrete mixes developed by Neeley et al. [21]
The yield strength of HLSA-V steel reinforcement is 572 MPa and for a similar high strength concrete mix. A brief comparison of
the failure strain is slightly higher than that of conventional rein- the two material models used in this study is presented below.
forcing bar. The introduction of vanadium into the chemical
composition of a steel reinforcement bar has the advantages of 1. CDMR3 is a three invariant model that uses three shear failure
increased strengths without compromising on ductility or form- surfaces and includes damage and strain-rate effects, while the
ability and has good fracture toughness and weldability. Winfrith model is a smeared crack model (sometimes known as
pseudo crack) and smeared rebar model that is implemented in
3. Literature survey and material models the 8 node single integration point continuum element.
2. In CDMR3 the equation of state is required to define the pres-
Several material concrete models are available in LS-DYNA; such sureevolumetric strain curve, while the Winfrith model will
as, the Winfrith Concrete Model [2], Holmquist Johnson Concrete automatically use a scaled curve if it is omitted from the input.
Model [5], Continuous Surface Cap Model for Concrete [6] and the 3. Both models are capable of material strain rate effects with the
Karagozian and Case Concrete Damage Model Release 3 [3,7]. Based strain rate being defined by a load curve in CDMR3, while the
on two previous studies [8,9] on concrete material models, the Winfrith model uses the CEB [22] strain rate enhancement and
Concrete Damage Model Release 3 (CDRM3) and Winfrith Concrete enhances the elastic, shear and bulk moduli as well as the
Model (WCM) have been chosen for use in this study. strength in tension and compression.
Several research studies have addressed the issue of the 4. In CDMR3 model the most significant user improvement pro-
behavior of RC slabs subjected to explosive loading. Ågårdh et al. vided by Release III is a model parameter generation capability,
studied the behavior of RC slabs subjected to blast loading using based solely on the unconfined compression strength of the
finite elements [10,11] and used the Winfirth model to compare concrete, while the Winfrith model does not generate parame-
with experimental studies on fiber reinforced slabs. They have re- ters from uniaxial confined strength alone.
ported a good comparison of peak displacement results for low 5. CDMR3 model is not capable of crack generation, while Winfrith
charge weights and noted that the strain rate effect of the material model has a crack generation capability with up to three
164 G. Thiagarajan et al. / International Journal of Impact Engineering 75 (2015) 162e173

orthogonal crack planes per element and viewing the cracks is 60 or HSLA-V reinforcement in combination with 27.6 MPa or
possible using a post processing capability. 107 MPa concrete. Fig. 1d shows the concrete slab mounted in the
target vessel.
4. Experimental program Tests were performed using relatively similar blast pressures
and impulses to determine the performance of the different RC slab
The experimental program consisted of ten 1/3 scale RC slabs combinations. The mid-span deflection, average reflected blast
evaluated in the BLS at USACE-ERDC. The objective of the experi- pressure and impulse were recorded.
ments was to investigate the potential improvements in the pro-
tection level resulting from the substitution of high-performance 4.2. Experimental data
materials for conventional materials when the RC slabs are subject
to blast loads. The basic input data used for the numerical simulations were
the reflected pressure vs. time plots. Fig. 2aed shows the reflected
4.1. Experimental setup pressure and impulse histories recorded for the slabs under
consideration in this study. Table 1 shows the average peak re-
The BLS (Fig. 1a) is a mechanical device developed to apply dy- flected pressure and impulse values recorded from four of the
namic air blast loads with peak pressures and impulses considered experiments.
representative of actual high-explosive events. It consists of a driver The dimensions of the rectangular RC slab evaluated in the
pressure vessel, followed by a vented expansion cone (Fig. 1a), experimental program were 1652 mm  857 mm  101.6 mm. The
transition rings and a target vessel. The RC slab is located at the front conventional steel and HSLA-V reinforcement bars used in the slab
end of the target vessel and is mounted on a steel reaction frame were both 9.5 mm diameter. The main steel reinforcement was
(Fig. 1b). The BLS uses compressed air/helium in the pressure vessel, spaced 101.6 mm on center. However, the spacing was reduced to
which is released by puncturing a metal diaphragm, and is designed 50.8 mm at the ends in order to reduce edge effects. The shrinkage
to simulate the air-blast environment for explosive yields up to steel reinforcement was spaced 304.8 mm on center in the short
9000 kg of TNT. A two piece steel reaction frame was used to confine direction. Fig. 3a shows the geometry details in plan view, while the
the slab in the target vessel face (Fig. 1bed). As shown in Fig. 1d, the typical cross section of the experimental specimen is shown in
second piece of the frame has two 75.3 mm  75.3 mm  12.7 mm Fig. 3b.
steel tubes welded to the steel picture frame. The tubing acts as a
bearing surface for the slabs and the frame allows for translation in 5. Numerical modeling
the direction of the blast loading. A 6.35 mm cardboard layer was
placed between the bottom of slab and the steel frame as a Four slabs were considered for the numerical studies performed
compressible bearing pad (Fig. 1c). The slab was not supported in and are presented in this paper. The numerical model of the RC slab
the lateral direction resulting in a one way action for the slab. The RC consists of eight noded hexahedron elements. The constant stress
slabs used for the study consisted of double mat conventional Grade solid element formulation was used with three mesh sizes

Fig. 1. (a) Blast Load Simulator (BLS) ‘Shock Tube’ used to simulate air pressure blast load. (b) First piece of frame bolted into target vessel face. (c) Concrete slab installed on top of
cardboard spacers and against structural tubing on the first frame piece. (d) Second piece of target frame sandwiched RC slab between the steel tubes.
 G. Thiagarajan et al. / International Journal of Impact Engineering 75 (2015) 162e173 165

Fig. 2. Reflected pressure and impulse histories for a typical slab a) HSC-NR, b) HSC-VR, c) NSC-NR and d) NSC-VR.

consisting of 25.4 mm, 12.7 mm and 6.35 mm. The constant stress The choice of the 25.4 mm mesh size corresponds well with the
element is a reduced 1-point integration formulation element. For study by Coughlin et al. [23] who used 30  30  30 mm size el-
the reduced integration elements, the FlanaganeBelytschko based ements in their blast study on portable concrete vehicle barriers.
hourglass control option available in LS-DYNA was used with the After this study was conducted, Foglar and Kovar [24] reported the
hourglass coefficient set to 0.03. usage of 30 mm mesh size for their experimental and numerical
HugheseLiu beam elements were used to model steel rein- studies on the blast resistance of fiber reinforced concrete and
forcement in two layers. Longitudinal and transverse steel were reinforced concrete bridge decks. They reported that the usage of
modeled to replicate the reinforcement locations in the experi- smaller elements did not increase the quality of the results. Nam
mental slab. The 25.4 mm mesh size model has 11,376 nodes, 8704 et al. [25] showed that the propagation of the blast wave in a ma-
solid elements and 1560 beam elements while the 12.7 mm mesh terial depends on the hourglass control type used.
size model consists of 83,151 nodes, 69,632 solid elements and 3120
beam elements. The 6.35 mm mesh size model consists of 598,554 5.1. Model boundary conditions
nodes, 557,056 solid elements and 6240 beam elements.
The boundary conditions used in the model replicated the
experimental conditions (Fig. 4a) by restraining the top and bottom
Table 1 nodes of the slab in the model from movement in the vertical Y-
Peak reflected pressure and reflected impulse values for the four experiments. direction. The horizontal steel tubes welded in the first piece of the
Slab ID Peak reflected Peak reflected target frame (Fig. 1b) were used to restrain the slab from movement
pressure (MPa) impulse along the top and bottom of the back face were represented by a
(MPa-ms) 152.4 mm restraint in the Z-direction (direction of the applied
HSC-NR 0.39 7.71 pressure load which is normal (90 ) to the bearing surface) as
HSC-VR 0.36 7.52 illustrated in Fig. 4b. The slab was secured in the target vessel frame
NSC-NR 0.35 6.79 by the steel tube on piece two on the blast face (Fig. 1d). This tube
NSC-VR 0.39 7.36
was represented in the models by placing one strip of nodes to
166 G. Thiagarajan et al. / International Journal of Impact Engineering 75 (2015) 162e173

Fig. 3. Concrete slab a) plan view and b) cross section showing loading direction and reinforcement details.

restrain movement in the Z-direction only (Fig. 4c) within the

pffiffiffiffiffi
surface area of the 75.3 mm steel bar that was used experimentally. AS2 l S2 BI1
þ þ 1¼0 (1)
s2c sc sc
5.2. Concrete and steel material models
where,
Two material models namely the Winfrith Concrete Model
(WCM) and the Concrete Damage Model Release 3 (CDMR3) were  
1
chosen for the study. l ¼ K1 cos  cos1 ðK2 cos 3qÞ for cos 3q  0
3
hp i
5.2.1. Winfrith concrete model l ¼ K1 cos  1=3cos1 ðK2 cos 3qÞ for cos 3q < 0
3
The theoretical basis and input parameters of the WCM, also pffiffiffi
known as the smeared rebar model, is presented in Broadhouse and 3 3
cos 3q ¼ 1:5 S3
Attwood [26]. Schwer [27] has given a detailed description of the 2S2
constitutive equations and the behavior of the WCM. The hydro-
static stress state in the model is determined from a volume and I1 is the first stress invariant, S2 and S3 are the second and third
compaction curve. The deviatoric stress state in the concrete is deviatoric stress invariants respectively and q is the lode angle. The
incremented elastically, using a locally rate dependent modulus. constants A, B, K1 and K2 are called the shape parameters and
The yield surface expands with increasing hydrostatic stress, and its control the meridional shape of the shear failure surface and are all
radii at the compressive and the tensile meridian are determined by functions of sc =st , where sc and st are the compressive and tensile
the locally rate sensitive compressive and tensile strengths. The strengths respectively. The constants K1 and K2 define the shape of
yield surface is described analytically by a function of the stress and the shear failure surface in the octahedral plane. The material
stress deviator tensors in the equations (Eq. (1)) below. property values used for concrete are a concrete mass density of
 G. Thiagarajan et al. / International Journal of Impact Engineering 75 (2015) 162e173 167

Fig. 4. Boundary conditions imposed on the (a) slab in the y-direction, (b) slab back face and (c) slab pressure (loading) face.

2400 kg/m3; concrete tangent modulus of 24.8 GPa, Poisson's ratio implements the strain rate by a user defined strength enhancement
of 0.18, uniaxial compressive strength of 27.6 MPa, uniaxial tensile versus strain rate curve in the program as shown in Table 3. Larcher
strength of concrete of 3.3 MPa and fracture energy/crack width of [28] studied the strength enhancement versus strain rate and based
0.0002 J/mm2. The material parameters for high strength concrete on comparison of experimental work shown in Bischoff and Perry
and regular strength concrete were the same with the exception of [29] and numerical models proposed by Gebbeken and Rupert [30],
the compressive and tensile strength values. The parameters that used a hyperbolic function to represent the strength enhancement
were related to the reinforcement were taken as zero, since a from 1 to 2.5 for strain rates up to 106. Table 3 shows the strength
different material model for steel was used. The default volume enhancement versus strain rate that has been used. The values from
compaction curve (pressure vs. volumetric strain) for 27.6 MPa the table are used in LSDYNA and is based on Malvar and Ross [31].
concrete was used [27]. For high strength concrete another volume The uniaxial compressive strength and the tensile strength were
compaction curve was used and a comparison of the two input as 27.6 MPa and 3.3 MPa respectively, and the rest of the
compaction curves is shown in Fig. 5. parameters were taken as zero, allowing the model to generate
damage function values on its own for normal strength concrete
5.2.2. Concrete Damage Model Release 3 simulations. For high strength concrete simulations these param-
The CDMR3 model is a three-invariant model, which uses three eters were generated from experimental data available from
shear failure surfaces and includes damage and strain-rate effects. USACE-ERDC for a high strength concrete developed for use in
The model has the inbuilt ability to generate the required model another study. Neeley et al. [21] developed the high strength con-
input parameters in LS-DYNA based on providing the unconfined crete mix used in the dynamic experiments and modeled in the
compressive strength alone. Model details and its applicability to numerical simulations. The material strain rate enhancement ef-
blast simulations are described in Malvar et al. [3]. The model fects in this model were not considered.
CDMR3 model [4] is a continuum plasticity model for concrete,
accounting for the three behaviors listed above and consists of
separate evolutions for the hydrostatic and deviatoric stress com-
ponents. The hydrostatic pressure is a function of the volumetric
strain and internal energy with a tension and compression cut off.
The model uses an equation of state functionality in LS-DYNA to
provide input pairs of pressure and volumetric strain data. The
deviatoric stress evolution is described using three failure surfaces.
From triaxial compression tests on concrete the von Mises stress,
which in this case is also the stress difference between the axial and
lateral stress, is related to the mean stress using three parameters
as shown in Eq. (2).

Table 2
CDMR3 failure surface parameters for normal strength concrete.

a0 (MPa) a1 (MPa) a2 (MPa)

Maximum failure surface 13.42 0.4463 0.00178

Yield surface 10.13 0.625 0.005672
Residual failure surface 0 0.4417 0.002606
Fig. 5. Pressure versus volumetric strain curves for normal and high strength concrete.
168 G. Thiagarajan et al. / International Journal of Impact Engineering 75 (2015) 162e173

Table 3 5.3. Crack propagation simulation

Strain rate versus strength enhancement factor.

Strain rate (/s) Strength Strain rate (/s) Strength The WCM has an option to depict the propagation of cracks on
enhancement enhancement the concrete surface. When a concrete element fails with a tensile
3.0Eþ01 9.7 3.0E-08 1.00 stress component, a crack is flagged in a plane normal to the
3.0E-01 9.7 1.0E-07 1.03 maximum principal stress. To simulate the physical cracks in con-
1.0E-01 6.72 1.0E-06 1.08 crete in a material model such as the CDMR3, an additional erosion
3.0E-02 4.50 1.0E-05 1.14
model called Mat_Add_Erosion is used. This erosion model is based
1.0E-02 3.12 1.0E-04 1.20
3.0E-03 2.09 1.0E-03 1.26 on the concept that the concrete element is deleted when the
1.0E-03 1.45 3.0E-03 1.29 material response in an element reaches a certain critical value. A
1.0E-04 1.36 1.0E-02 1.33 typical strain at peak tensile stress is approximately 0.0002.
1.0E-05 1.28 3.0E-02 1.36
Assuming a static cracking strain of five times the peak tensile
1.0E-06 1.20 1.0E-01 2.04
1.0E-07 1.13 3.0E-01 2.94
strain magnitude and with dynamic strain enhancement effects
1.0E-08 1.06 3.0Eþ01 2.94 due to material and structural effects accounting for about ten
0.0Eþ00 1.00 times the static cracking strain [12]. The principal tensile strain of
0.01 is adopted as the governing criterion in the implementation of
the erosion algorithm in the numerical simulation. The experi-
mental crack pattern shown in the respective figures are at the end
smean of the experiment. In order to make a uniform comparison the
svm ¼ ðsaxial  slateral Þ ¼ a0 þ (2) deflection patterns from the numerical simulations are also shown
a1 þ a2 smean
at the end of the simulation, which is 145 ms in most cases. The
The behavior is elastic up to the initial yield surface and in- post peak deflections are fairly close the to the peak deflection in
creases in an elastoplastic manner until the maximum yield surface most numerical simulations.
is reached. After the maximum stress is reached softening occurs
till the residual surface is reached. For each of the three surfaces the 6. Numerical results and experimental comparisons
model defines a set of three parameters. The model has been cali-
brated and validated by Schwer and Malvar [4] using test data for This section presents the comparison of experimental data with
concrete strength of 45.6 MPa. Table 2 shows the values of the those from numerical simulations. Only the deflection history and
parameters used for the three surfaces. The model has the capa- damage data, in the form of images taken after the blast load was
bility to model strain rate effects using a dynamic increase factor applied, are available from experimental results and are compared
and two parameters, b1 and b2, are used to incorporate differing to the numerical simulations. Experimentally measured deflections
strain rate effects in tension and compression. The damage effects are compared with predictions for two concrete models and two
are incorporated using volumetric damage accumulation using mesh sizes, namely 25.4 mm and 12.7 mm. The 6.35 mm mesh size
another input parameter, b3. A user defined damage accumulation was only used to compare crack propagation between the models.
function, h, as a function of the effective plastic strain parameter l,
is used to locate the current yield surface relative to the maximum 6.1. Normal strength concrete with HSLA-V steel reinforcement
failure surface. The model, in this study, used a fractional shear (NSC-VR)
dilatancy value of 0.5 (for parameter U) allowing for enhanced
plastic volume with increasing confinement leading to a higher 6.1.1. Deflection comparison
mean stress and therefore an increased shear strength. The mass The deflection results obtained are compared in Fig. 6. The
density of concrete is set to 2320 kg/m3, the Poisson's ratio to 0.15, average deflection obtained from the experiment was 129.5 mm.
the uniaxial compressive and tensile strength to 45.6 MPa and From Fig. 6, varying the mesh size from 25.4 mm to 12.7 mm
4.75 MPa respectively, and the maximum aggregate size to 19 mm. showed an increase in deflection by 76.2 mm in both CDMR3 and
The aggregate size is used to determine the fracture energy used in WCM. It is observed that a 25.4 mm mesh size was adequate and
the model. Coarse aggregate has the effect of higher tensile best in comparing the deflection responses. Purely from a magni-
strengths and hence higher fracture energies [7]. tude perspective the CDMR3 with a 25.4 mm mesh size gave the
closest prediction of deflection when compared to the experi-
5.2.3. Steel material model mental deflection, but the WCM with 25.4 mm mesh showed a
In order to simulate the response of steel reinforcement repre- better history response and had a better prediction of the time to
sented as beam elements in the model, a plastic kinematic model maximum deflection. It can also be seen that the predicted times of
was chosen in order to consider the effects of isotropic and kine- peak deflections can vary as much as 30e40 ms between the
matic hardening of beam elements. The parameters used for the models and the mesh sizes. The difference in the predicted de-
model were: mass density of 7830 kg/m3; Young's modulus of flections between the two mesh sizes is indicative of the difficulty
200 GPa; Poisson's ratio of 0.3; yield strength of HSLA-V steel of in achieving mesh convergence.
572 MPa; tangent modulus of 20 GPa. Kinematic hardening effect
was taken into consideration and the strain rate effects in the 6.1.2. Crack propagation
model were not considered. Several researchers have used this Fig. 7a is a posttest view of the experimental slab illustrating the
material model in blast simulations of reinforced concrete slabs actual crack pattern observed during the experiment at the end of
[14,17,32,33]. the test. The numerical simulation of these cracks for the WCM
In this study, the reinforcement is modeled as separate ele- with a 25.4 mm mesh size is shown in Fig. 7b, which is a good
ments. Since the loads under consideration are explosive in nature match with the experimental results. The numerical simulation
and the rate of loading is very high, bond-slip has been neglected showed that the cracks first start appearing at the center of the slab
for blast and impact studies [12]. Concrete and steel elements are and propagate to the periphery of the slab along with the appli-
tied by coupling the reinforcing beam elements to concrete ele- cation of the load. The maximum crack width that was observed on
ments through the constrained Lagrange in solid formulation. the slab was 4.2 mm. Fig. 7c depicts the damage pattern obtained
 G. Thiagarajan et al. / International Journal of Impact Engineering 75 (2015) 162e173 169

Fig. 6. Deflection history of normal strength concrete with high strength bar (NSC-VR
Fig. 8. Deflection history of normal strength concrete with normal strength bar (NSC-
slab).
NR slab).

from the CDMR3 with an erosion governing criteria of maximum

principal strain of 0.01 for a 6.35 mm mesh. The damage and crack
patterns obtained using the models from all three mesh sizes were 6.2.2. Crack propagation
compared with crack patterns obtained from the experiment. The Fig. 9a shows the actual cracks that were observed during the
erosion criterion results in loss of stiffness and with a larger experiment at maximum deflection. The numerical simulation of
element size the loss is artificially higher. Hence, we have observed these cracks for the WCM with a 25.4 mm mesh size has been
that the 6.35 mm model gave a good prediction. In the higher mesh depicted in Fig. 9b, which is a good match with the experimental
size models e once an element erodes we observed that deflections results. The maximum crack width that was observed on the slab
increased resulting in further erosion. All the crack patterns are was 4.2 mm. Fig. 9c shows the damage pattern obtained from the
reported at the end of the respective simulations. 6.35 mm mesh size in the CDMR3 model. It is observed that while
the overall pattern is captured the experimental slab had significant
spallation that is not seen in the simulation.
6.2. Normal strength concrete with conventional steel
reinforcement (NSC-NR)
6.3. High strength concrete with conventional steel reinforcement
6.2.1. Deflection comparison (HSC-NR)
Fig. 8 shows the comparison of the deflection results obtained
from the two LS-DYNA concrete models and the experiments. The 6.3.1. Deflection comparison
average deflection obtained from the experiment was 221 mm. The The deflection results obtained for the HSC-NR category are
deflections obtained from the WCM and the CDMR3 were lower compared in Fig. 10. The average deflection obtained from the
than the experimental deflection by 63.5 mm and 78.8 mm experiment was 139.7 mm. The peak deflection obtained from the
respectively. In these simulations the 12.7 mm models gave results WCM with 25.4 mm mesh size was 124.2 mm for both mesh sizes.
closer to the experimental deflection. Difficulties in achieving mesh In this simulation, the Winfrith model appears to be giving a good
convergence have been observed in this case also. mesh convergence.

Fig. 7. Final crack patterns from experiments and simulations (NSC-VR) a) experimental pattern, b) WCM with 25.4 mm mesh, c) CDMR3 with 6.35 mm mesh.
170 G. Thiagarajan et al. / International Journal of Impact Engineering 75 (2015) 162e173

Fig. 9. Final crack patterns from experiments and simulations (NSC-NR) a) experimental pattern, b) WCM with 25.4 mm mesh, c) CDMR3 with 6.35 mm mesh.

6.3.2. Crack propagation values. Difficulty in achieving mesh convergence can also be
Fig. 11a depicts the actual cracks that were observed during the observed in this figure.
experiment at maximum deflection. The numerical simulation of
these cracks for the WCM with a 25.4 mm mesh size are shown in 6.4.2. Crack propagation
Fig. 11b, which is a good match with the experimental results. Fig. 13a shows the cracked specimen from the experiments.
Fig. 11c depicts the damage pattern obtained from the CDMR3 used Fig. 13b and c shows the damage patterns from the simulations.
in conjunction with an erosion governing criteria of maximum Although the exact damage pattern was not captured, the overall
principal strain of 0.01. Though the 6.35 mm mesh model depicts major cracking patterns are shown in the figures.
the crack patterns in the HSC-NR category, the patterns are not a
very close match to the experimental patterns and as effective as
7. Analysis of results
the ones seen in the normal strength concrete categories.
Table 4 shows the maximum deflections from the experimental
6.4. High strength concrete with HSLA-V steel reinforcement (HSC- and numerical simulations. Based on the experimental and nu-
VR) merical deflection data and the percent difference between the
experimental and numerical values presented in Table 5, the
6.4.1. Deflection comparison behavior of the two models can be summarized as follows.
The deflection results obtained for the HSC-VR category are
compared in Fig. 12. The peak deflection obtained from the exper- e In the normal strength concrete (NSC-VR and NSC-NR) cate-
iment was 121.9 mm. Although a continuation of increased de- gories with 25.4 mm mesh size, both the WCM and the CDMR3
flections was observed the first peak has been considered for had deflection predictions that were higher up to 6% (for NSC-
comparisons. The peak deflection obtained from the WCM for both VR) and ranged from 27.5% to 35.6% less (for NSC-NR)
mesh sizes were similar and matched closely with experimental compared to the experimental values.
e In the high strength concrete (HSC-VR and HSC-NR) categories
with 25.4 mm mesh size, the WCM and CDMR3 provided
deflection values that under predicted 37.5% and 17% (for HSC-
VR) and 10.9e12.5% (for HSC-NR) compared to the experi-
mental values.
e In the NSC-VR and NSC-NR categories with 12.7 mm mesh size,
both the WCM (58.8% for NSC-VR and 13.8% for NSC-NR) and the
CDMR3 (68.6% for NSC-VR and 2.3% for NSC-NR) showed
deflection values that were higher than that of the experimental
values.
e In the HSC-VR and HSC-NR categories with 12.7 mm mesh size,
the WCM provided deflection values that were less (10.4% for
HSC-VR and 10.9% for HSC-NR) than experimental values;
however, the CDMR3 deflection values were much higher (48.0%
for HSC-VR and 29.1% for HSC-NR) than the experimental values.
e Table 4 summarizes the peak deflection values from all the
numerical simulations and compares them to the experimental
values. The WCM provided a good response in terms of both
Fig. 10. Deflection history of high strength concrete with normal strength bar (HSC-NR deflection and crack depictions. In this research work it is found
slab). that the usage of high strength concrete provided the greatest
 G. Thiagarajan et al. / International Journal of Impact Engineering 75 (2015) 162e173 171

Fig. 11. Final crack patterns from experiments and simulations (HSC-NR) a) experimental pattern, b) WCM with 25.4 mm mesh, c) CDMR3 with 6.35 mm mesh.

decrease in the level of response and enhanced resistance to

blast effects. The usage of HSLA-V bars followed in effectiveness.

Table 5 compares the percent difference in the experimental

deflections between various combinations of slabs used in this
study. The following salient observations can be made from this
table. It must be noted that in comparison to NSC-NR as a base line
all the other slab combinations had approximately 10% higher
pressure-impulse loading applied to them.

e Peak deflection dropped by approximately 41% for NSC-VR

compared to NSC-NR.
e Peak deflection dropped by approximately 45% for HSC-VR
compared to NSC-NR.
e When using high strength concrete, the change in reinforce-
ment from NR to VR lowered the deflections by 12.7%.
e When using normal reinforcement, the change in concrete
Fig. 12. Deflection history of high strength concrete with high strength bar (HSC-VR
slab).
strengths from NSC to HSC resulted in a deflection drop of 36.8%.

Fig. 13. Final crack patterns from experiments and simulations (HSC-VR) a) experimental pattern, b) WCM with 25.4 mm mesh, c) CDMR3 with 6.35 mm mesh.
172 G. Thiagarajan et al. / International Journal of Impact Engineering 75 (2015) 162e173

Table 4
Analytical and experimental deflection summary.

Slab ID Deflections (mm) (% difference)

Mesh size 25.4 mm 12.5 mm

Concrete model WCM CDMR3 WCM CDMR3 Experimental

NSC-VR 137 (þ5.9%) 129 (0) 205 (þ58.8%) 218 (þ68.6%) 129
NSC-NR 160 (27.5%) 142 (35.6%) 251 (þ13.8) 226 (þ2.3%) 221
HSC-VR 76 (37.5%) 101 (16.7%) 109 (10.4%) 180 (þ48.0%) 122
HSC-NR 124 (10.9%) 121 (12.7%) 124 (10.9%) 180 (þ29.1%) 140

The crack patterns obtained from the WCM with a 25.4 mm concrete strengths with the help of the volume compaction
mesh size was closer to that of the experiment and did not require curve.
an additional material model to initiate cracks. In the case of 5. Both the models performed reasonably well in predicting the
CDMR3, it requires an additional material model to generate cracks peak deflection, as well as the deflection histories for all the
and a 6.35 mm mesh or lower to generate cracks effectively. The simulations.
primary reason for the two different mesh sizes for the CDMR3
model and the WCM model is that WCM has an option to model
References
cracks while CDMR3 models crack using an additional erosion
model. A smaller e 6.35 mm e mesh size for CDMR3 was used since [1] LS-DYNA V971, keyword manual, vols. 1 and 2. Livermore, CA: Livermore
the erosion of an element results in loss of stiffness and a larger Software Technology Corporation; 2009.
[2] Broadhouse B. The Winfrith concrete model in LS-DYNA3D. Report: SPD/D
mesh size erosion can artificially reduce the stiffness of the struc-
(95); 1995. p. 363.
ture to a much greater extent compared to a smaller mesh size, [3] Malvar L, Crawford J, Morrill K. K&C concrete material model release III e
after cracking occurs. automated generation of material model input, Karagozian and case structural
engineers. Technical Report TR-99-24.3; 2000.
[4] Schwer L, Malvar L. Simplified concrete modeling with MAT_CONCRETE_
DAMAGE_REL3. In: JRI LS-DYNA User Week. Bamberg: LS-DYNA Anwender-
8. Conclusions forum; 2005.
[5] Holmquist TJ, Johnson GR, Cook WH. A computational constitutive model for
The following conclusions can be drawn from the detailed concrete subjected to large strains, high strain rates, and high pressures. In:
The 14th International Symposium on Ballistics, Quebec; 1993. p. 591e600.
experimental and numerical study presented in this paper. [6] Schwer L, Murray Y. Continuous surface cap model for geomaterial modeling:
a new LS-DYNA material type; 2002. p. 16e35.
1. From Table 4 it is evident that using high strength materials, [7] Magallanes JM, Wu Y, Malvar LJ, Crawford JE. Recent improvements to release
III of the K&C concrete model. In: 11th International LS-DYNA Users Confer-
both steel and concrete, improved the level of protection. NSC- ence. Livermore, CA: Livermore Software Technology Corporation; 2010.
NR had the lowest combination of pressure and impulse [8] Brannon RM, Leelavanichkul S. Survey of four damage models for concrete.
applied but the highest deflection. Comparing HSC-NR and NSC- Albuquerque, New Mexico 87185 and Livermore, California 94550: S.N. Lab-
oratories; 2009.
NR the experimental deflection dropped by 37%, although the [9] Yaramada VKR. Numerical response of steel reinforced concrete slab subjected
pressure and impulse loading was only about 10% higher. Hence, to blast and pressure loadings in LS-DYNA. Kansas City: Civil and Mechanical
it can be observed that high strength concrete was very effective Engineering, University of Missouri Kansas City; 2010.
[10] Ågårdh L. Finite element modeling of fibre reinforced concrete slabs subjected
in reducing the level of response.
to blast load. Le Journal de Physique IV 1997;7:723e8.
2. From Table 5 comparing HSC-NR and HSC-VR the high strength [11] Ågårdh L, Laine L. 3D FE-simulation of high-velocity fragment perforation of
steel usage reduced deflections by 12.7%, while comparing NSC- reinforced concrete slabs. Int J Impact Eng 1999;22:911e22.
[12] Xu K, Lu Y. Numerical simulation study of spallation in reinforced concrete
NR and NSC-VR the reduction was 45%. Hence, the change in
plates subjected to blast loading. Comput Struct 2006;84:431e8.
using high strength bars instead of normal strength bars was [13] Malvar LJ, Crawford JE, Wesevich JW, Simons D. A plasticity concrete material
more effective in normal strength concrete than high strength model for DYNA3D. Int J Impact Eng 1997;19:847e73.
concrete. However, given the limited range of testing and ana- [14] Tanapornraweekit G, Haritos N, Mendis P, Ngo T. Modelling of a reinforced
concrete panel subjected to blast load by explicit non-linear finite element
lyses presented in this paper, the advantages provided by the code. In: Proceedings of the Earthquake Engineering Australia; 2007.
advanced materials are an observation rather than a conclusion. [15] Zhou X, Kuznetsov V, Hao H, Waschl J. Numerical prediction of concrete slab
3. From Table 4, it can be seen that mesh size sensitivity is sub- response to blast loading. Int J Impact Eng 2008;35:1186e200.
[16] Schenker A, Anteby I, Gal E, Kivity Y, Nizri E, Sadot O, et al. Full-scale field tests
stantially higher when normal strength concrete is used with of concrete slabs subjected to blast loads. Int J Impact Eng 2008;35:184e98.
high strength steel reinforcement. Mesh sensitivity was lower [17] Hao DU, Zhongxian L. Numerical analysis of dynamic behavior of RC slabs
for WCM in the high strength concrete category. In general, the under blast loading. Trans Tianjin Univ 2008;15:61e4.
[18] Sangi A, May I. High-mass, low-velocity impacts on reinforced concrete slabs;
models, while they may achieve convergence in static loading 2009.
cases, were not very successful in showing convergence in [19] Morales-Alonso G, Cendo  n DA, Erice B, S anchez-Galvez V. Blast response
extreme dynamic loading situations and further focused analysis of reinforced concrete slabs: experimental procedure and numerical
simulation. Trans ASME J Appl Mech 2011;78.
research on mesh convergence studies are recommended.
[20] Thiagarajan G, Vasudevan AK, Robert S. Numerical modeling of concrete slabs
4. The WCM has the ability to show the propagation of cracks on reinforced with high strength low alloy vanadium steel bars subjected to blast
the surface of the reinforced concrete slab, without being very loads. American Concrete Institute Special Publication, SP-281; 2011.
[21] Neeley BD, Hammons MI, Smith DM. The development and characterization of
sensitive to mesh sizes. It can also be used for a wide range of
conventional-strength and high-strength Portland cement concrete mixtures
for projectile penetration studies. Vicksburg, MS: USACE Waterways Experi-
mentation Station; 1991.
Table 5 [22] CEB-FIP model code 1990. Trowbridge, Wiltshire, UK: Comite Euro Interna-
Experimental deflection summary and comparison. tional du Beton, Redwood Books; 1993.
[23] Coughlin A, Musselman E, Schokker A, Linzell D. Behavior of portable fiber
Reference slab NSC-VR HSC-VR HSC-NR reinforced concrete vehicle barriers subject to blasts from contact charges. Int
NSC-NR 41.4% 44.8% 36.8% J Impact Eng 2010;37:521e9.
[24] Foglar M, Kovar M. Conclusions from experimental testing of blast resistance
HSC-NR 12.7%
of FRC and RC bridge decks. Int J Impact Eng 2013;59:18e28.
 G. Thiagarajan et al. / International Journal of Impact Engineering 75 (2015) 162e173 173

[25] Nam J, Kim J-H, Kim S, Yi N, Byun K. A study on mesh size dependency of finite [29] Bischoff P, Perry S. Compressive behaviour of concrete at high strain rates.
element blast structural analysis induced by non-uniform pressure distribu- Mater Struct 1991;24:425e50.
tion from high explosive blast wave. KSCE J Civ Eng 2008;12:259e65. [30] Gebbeken N, Ruppert M. A new material model for concrete in high-dynamic
[26] Broadhouse B, Attwood G. Finite element analysis of the impact response of hydrocode simulations. Arch Appl Mech 2000;70:463e78.
reinforced concrete structures using dyna3d. In: Kussmaul K, editor. SMiRT- [31] Malvar LJ, Ross CA. Review of strain rate effects for concrete in tension. ACI
12. San Francisco, CA, B.V.: Elsevier Science Publishers; 1993. p. 339e44. Mater J 1998;95.
[27] Schwer L. The Winfrith concrete model: beauty or beast? insights into the [32] Huang F, Wu H, Jin Q, Zhang Q. A numerical simulation on the perforation of
Winfrith concrete model. In: 8th European LS-DYNA Users Conference; 2011. reinforced concrete targets. Int J Impact Eng 2005;32:173e87.
[28] Larcher M. Development of discrete cracks in concrete loaded by shock waves. [33] Du H, Li Z. Numerical analysis of dynamic behavior of RC slabs under blast
Int J Impact Eng 2009;36:700e10. loading. Trans Tianjin Univ 2009;15:61e4.