ARTICLE IN PRESS

International Journal of Impact Engineering 35 (2008) 290–303

www.elsevier.com/locate/ijimpeng

Numerical predictions of ballistic limits for concrete slabs using a

modified version of the HJC concrete model
M. Polanco-Loriaa,, O.S. Hopperstadb, T. Børvikb,c, T. Berstada
a
SINTEF Materials and Chemistry, N-7465 Trondheim, Norway
b
Structural Impact Laboratory (SIMLab), Department of Structural Engineering, Norwegian University of Science and Technology (NTNU), N-7491
Trondheim, Norway
c
Research & Development Department, The Norwegian Defence Estates Agency, N-0103 Oslo, Norway

Received 12 October 2006; accepted 1 March 2007

Available online 18 March 2007

Abstract

Some modifications to the Holmquist–Johnson–Cook (HJC) model (1993) for concrete under impact loading conditions are proposed.
First, the pressure-shear behaviour is enhanced by including the influence of the third deviatoric stress invariant to take into account the
substantial shear strength difference between the tensile and compressive meridians. Second, the modelling of strain-rate sensitivity is
slightly changed so that the strain-rate enhancement factor goes to unity for zero strain rate. Third, three damage variables describing the
tensile cracking, shear cracking and pore compaction mechanisms are introduced. A critical review of the constitutive model with
alternative proposals for parameter identification is given. The model parameters are obtained for two concrete qualities, and perforation
of concrete slabs is considered numerically and compared with experimental results from the literature. Ballistic limit assessments with
deviations under 8% when compared to the experimental results are obtained, indicating that the modified version of the HJC concrete
model represents a good compromise between simplicity and accuracy for large-scale computations of concrete plates impacted by
projectiles.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: Concrete plates; Projectile perforation; Elastic-viscoplastic model; Isotropic damage; Numerical simulations

1. Introduction are available for the analysis and design of concrete

structures, as described in Refs. [12–14], among others.
Concrete is a material frequently used in protective However, still the development of reliable and robust
structures. This makes concrete of particular interest for material laws for concrete subjected to high-intensity
the nuclear industry and fortification installations for loading of short duration is a subject of current research.
defence purposes [1,2]. Many important studies of the Analysis of steel reinforced concrete structures impacted
behaviour of concrete targets impacted by projectiles have by free-flying projectiles is extremely difficult because of the
been of an empirical nature and most of the experimental complex states of stress acting inside the material. Under a
studies have focused on the determination of empirical multi-axial state of stress the damage mechanisms activated
formulae or simplified analytical models [3–5]. On the other are highly dependent on the loading path imposed. For
hand, the parallel development of numerical tools (e.g. finite concrete materials under static loading conditions, Mazars
element packages [6–8]) and new constitutive models for [15] showed that depending on the load path different
concrete has contributed to an increased use of numerical damage mechanisms are activated; namely cracking,
simulations for prediction of penetration and impact shearing (e.g. mode II cracking) and compaction, as shown
problems [9–11]. Different constitutive models for concrete in Table 1. In particular, for impact and penetration
problems these three mechanisms are always present, as
Corresponding author. Tel.: +47 93 28 04 35; fax: +47 73 59 29 31. indicated by Bourlion [16] and illustrated in Fig. 1.
E-mail address: [email protected] (M. Polanco-Loria). This requires the development of material models capable

0734-743X/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijimpeng.2007.03.001
 ARTICLE IN PRESS
M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303 291

Table 1
Main damage mechanisms according to Mazars [15]

Damage mechanism Classical loading

compaction
Type of loading A σ compression
Tension and/or compression with
low lateral pressure (uniaxial, biaxial
or tri-axial) tension cracking
Characteristics
Possibility of extension, at least in σ
Fig. 1. Schematic description of the mechanisms activated during impact
one principal direction [16].
Brittle/softening behaviour
Local damage mode ε
Cracking mainly in mode I
(debonding) and mode This paper describes the modifications proposed to the
I þ II ðdebonding þ branchingÞ HJC model in some detail. The MHJC model has been
implemented into LS-DYNA as a user-material model and
Type of loading B σ is validated against some experimental results for the
Compression with moderate to
ballistic perforation resistance of concrete slabs presented
high lateral pressure (tri-axial)
Characteristics in the literature.
σ1
No extension
‘‘Ductile’’ behaviour 2. Original HJC concrete model
σ2
Local damage mode
Cracking mainly in mode II and
III (branching) The HJC model was originally presented by HJC [17]
ε
with the purpose of developing a concrete model for impact
Type of loading C
computations where the material experiences large strains,
Hydrostatic pressure (tri-axial)
Characteristics p high strain rates and high pressures. The HJC concrete
No extension model is an elastic-viscoplastic model coupled with
Hardening and stiffening isotropic damage, where the response is separated into
Local damage mode hydrostatic and deviatoric contributions.
Consolidation of the micro-
The Cauchy stress tensor sij is decomposed into
porous structure
μ deviatoric and hydrostatic parts
sij ¼ Sij  Pdij , (1)
where S ij is the stress deviator, P ¼ skk =3 is the
to describe concrete behaviour under low and high hydrostatic pressure and dij is the Kronecker delta. The
confining pressures and the observed differences in rate von Mises equivalent stress is then defined by
sensitivity of these mechanisms. qffiffiffiffiffiffiffiffiffiffiffiffiffi
In [12] a literature review of existing experimental data seq ¼ 32Sij Sij . (2)
and constitutive models for concrete materials under
The deviatoric part D0ij of the rate-of-deformation tensor
impact loading conditions was undertaken. It was con-
Dij is defined as
cluded that the Holmquist, Johnson and Cook (HJC)
concrete model [17,18] represents a good compromise D0ij ¼ Dij  13Dkk dij (3)
between simplicity and accuracy for large-scale computa-
and it is decomposed into its elastic and plastic parts
tions, and the HJC model was implemented in LS-DYNA
according to
[7] for penetration simulations. Some improvements of the
original HJC model were recommended and evaluated in e p e 1 rJ p 3 S ij
D0ij ¼ D0ij þ D0ij ; D0ij ¼ S ; D0ij ¼ _ peq . (4)
[19], which forms the basis of the modified version of the 2G ij 2 seq
HJC model presented in this work. In particular, the
In Eq. (4), S rJ ij is the Jaumann rate of the stress deviator, G
modifications include a continuous function describing
is the shear modulus and _ peq is the equivalent plastic strain
the pressure-shear response, the influence of the third
rate
deviatoric stress invariant to take into account the qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
difference between the shear response of the tensile and _ peq ¼ 23D0ij p D0ij p . (5)
compressive meridians, a non-linear two-parameter equa-
tion to handle the strain-rate sensitivity and the introduc- In the HJC model, the deviatoric response is determined by
tion of three damage variables describing tensile cracking, the constitutive relation
shear cracking and pore compaction mechanisms. Because
seq ¼ ½Að1  DÞ þ BPN ½1 þ C ln _eq pSmax for P X0
of these modifications of the original model, the present
version will be referred to as the MHJC model. (6)
 ARTICLE IN PRESS
292 M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303

in which seq ¼ seq =f c is the normalized equivalent stress, The theoretical basis of the HJC model is relatively simple
P ¼ P=f c is the normalized pressure and _ eq ¼ _eq =_0 is the and it takes into account most of the important issues of
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
normalized strain rate, where _eq ¼ 23D0ij D0ij is the concrete behaviour; namely pressure dependency, rate depen-
dency, damage cracking and compaction (e.g. pressure–
equivalent deviatoric strain rate. The normalizing para- volume changes). Other important concrete behaviour me-
meters are the quasi-static uniaxial compressive strength f c chanisms are not included in the original formulation, for
and the reference strain rate _0 . Further, A is the cohesion instance the non-linear behaviour before the peak load, the
parameter, B is the pressure hardening coefficient, N is the stiffness degradation (Young’s modulus reduction), the
pressure hardening exponent, C is the strain-rate sensitivity stiffness recovery (crack closure), the induced anisotropy (loss
coefficient, and Smax is the normalized maximum strength of isotropy by cracking) and shear reduction influenced by the
that can be developed. Material degradation is described state of stress (difference between tensile and compressive
by the damage variable D, resulting in reduction of the meridians). It is believed, however, that for impact and
cohesive strength. In the negative pressure regime ðP o0Þ, penetration applications the compressive waves will reduce the
the normalized hydrostatic tension T  ¼ T=f c is intro- tensile cracking effect caused by the reflected tensile waves by
duced and a linear dependence between the normalized closing them. Consequently, as long as tensile cracking is not
equivalent stress and the normalized pressure is assumed. T considered as the main damage contribution with its
is the maximum hydrostatic tension the material can correspondent directional effects, the simple isotropic damage
withstand. The original article by HJC [17] does not formulation seems still attractive (this is not the case for static
describe in detail how the point of discontinuity ðP ¼ 0Þ is problems). In addition, the HJC concrete model can be largely
treated. The deviatoric response of the HJC model is improved by introducing the influence of the third stress
illustrated in Fig. 2. invariant, hence, differentiating the tensile and compressive
The HJC model includes a scalar damage formulation, meridians.
where the damage evolution is accumulated from both the In order to illustrate the importance of the third
equivalent plastic strain increment Dpeq (which is caused by invariant, some classical results from the literature [20]
plastic shear deformation) and the equivalent plastic related to the experimental behaviour of concrete under
volumetric strain increment Dmpeq (which is due to plastic static loading conditions are illustrated in Fig. 4. From this
crushing of the air voids in the concrete). The damage figure one can observe the curved shape of the tensile ðy ¼
evolution is expressed as 0 Þ and compressive ðy ¼ 60 Þ meridians, and how the
Dpeq þ Dmpeq shear capacity increases with increasing pressure. Thus,
DD ¼ . (7) material models like the HJC model, which only consider
fp þ mfp
the effect of pressure and shear (the latter in terms of the
The plastic strain to fracture fp þ mfp is a function of the von Mises equivalent stress) in its formulation, will not be
triaxiality level and is expressed as able to capture the actual shear variations in the deviatoric
plane. For demonstration we include in Fig. 4 the shear
fp þ mfp ¼ aðP þ T  Þb Xðfp ÞMIN , (8) strength prediction of the HJC model, using Eq. (6) with
A ¼ 0:75, B ¼ 1:65 and N ¼ 0:65 (values commonly used)
where a and b are material constants. The lower limit and excluding both damage and rate effects. Indeed, the
ðfp ÞMIN was introduced to allow for a finite amount of HJC model overestimates the shear response for low
plastic strain to fracture the material in order to suppress pressure levels ðP o1Þ for both meridians, while the tensile
fracture from low magnitude tensile waves. The compac- and compressive meridians are overestimated and under-
tion response of the HJC model is left to be presented in the estimated, respectively, for higher pressure levels ðP 42Þ.
next section because a quite similar formulation is adopted Based on observations like these a modified version of the
in this work, see Fig. 3. HJC model is proposed in the next chapter.

D = 0 (undamaged)
3. The modified HJC (MHJC) constitutive model
Smax

3.1. Pressure dependence

σ∗eq

D=1(fractured) In order to avoid the discontinuous description of the

original model and consequently circumvent the identifica-
tion of the cohesion parameter A, the MHJC model adopts
a simple continuous function defined by
A 8   N 
< B½P þ T ð1  DÞ F ð_eq ÞRðy; eÞpSmax
>
  
T*(1-D) P* seq ¼ for P X  T ð1  DÞ; (9)
>
:
Fig. 2. The shear-pressure strength response of the HJC model.
0 for P o  T  ð1  DÞ;
 ARTICLE IN PRESS
M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303 293

Plock Dc=1

Pcrush Η
μp
P Dc=0 μlock

2 3
P = K1μ + K2μ + K3 μ

Plock

Pcrush
K1
Kav ρ
K μ = − 1
ρ 0
μcrush μlock

T (1-D) [(1−DC )K + DCK1]

Fig. 3. Pressure–volume response of the HJC concrete model.

Richard_C
Richard_T
9
Balmer
8 θ = 60°
Imran
7
HJC, A = 0.75,B = 1.65,N = 0.6
6
MHJC, B = 2, N = 0.8(CM)
5
σeq
*

MHJC, B = 2, N = 0.8(TM)
4
3
θ = 0°
2
1
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
*
P

Fig. 4. Experimental concrete shear strength response and MHJC model prediction.

where B, P , T  , N and S max have the same meaning as in concrete qualities having compressive strengths between 22
the original model, while the new functions F ð_eq Þ and and 74 MPa, as illustrated in Fig. 5. Note that it seems to
Rðy; eÞ are defined below. Fig. 4 includes the shear strength be no experimental evidence of the existence of the shear
prediction of the MHJC model for the compression (CM) threshold S max , but this threshold is kept as an option in
and tension meridians (TM) using Eq. (9). For this the MHJC model and is included in Fig. 5 for illustration
response values of B ¼ 2:0, N ¼ 0:8 and a shape factor purposes.
e ¼ 0:7, see Eq. (11), were assumed. Both damage and rate Damage causes a reduction in strength by shifting the
effects were excluded. It is noted that in the MHJC model maximum strength curve to a damaged state (loss of
the normalized cohesive strength is automatically deter- cohesion), as in the original model. In a complete damaged
mined by state with D ¼ 1, the concrete behaves as a granular
material characterized by a certain residual strength
seq ¼ B½T  ð1  DÞN F ð_eq ÞRðy; eÞ for P ¼ 0. (10)
envelop. However, the identification of this residual state
By assuming 1:8pBp2:0 and 0:60pNp0:80, the MHJC is difficult and the authors are not aware of any available
material model agrees with experimental results for the information in the literature about the shear behaviour of
compressive meridian reported in the literature [19]. These concrete in a completely damaged state. However, one can
results apply for an undamaged state with D ¼ 0 and assume that the shear capacity in the damage state is
 ARTICLE IN PRESS
294 M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303

region. The lower limit of the shape factor e ¼ 0:5 leads

8
Smax to a normalized equi-biaxial compressive strength of 1.0,
7 whereas an upper value of e ¼ 1 increases dramatically the
B = 2.0 N = 0.8
6 IImran
mrantests normalized equi-biaxial compressive strength to 5.3. By
S eries2
Balmer
5 S eries7
Richard_C adopting the Willam-Warnke concrete model [21], Chen [20]
B = 1.8,N = 0.6 found that the shape factor changes from 0.684 to 0.705
σeq

4
*

when the pressure goes from 0.33 to 2.33. For moderate

3 pressure levels 1oP o3, Launay and Gauchon [23]
2 reported a shape factor around e ¼ 0:7. In addition, for
1 high pressure levels a highly ductile behaviour is expected
such that the deviatoric shape becomes circular. In order to
0
0 1 2 3 4 5 6 fit the results of Chen [20] and the classical plane stress
P* experimental results of Kupffer et al. [24] for the biaxial
compression case, the following linear function for e in
Fig. 5. Undamaged compressive meridian strength response of the MHJC
terms of P is adopted
model [19].
8
>
> 0:65 for P o0;
>
< 
P
reduced by loss of cohesion, i.e. the term T  ð1  DÞ in e ¼ 0:65 þ ð1  0:65ÞP for 0pP pPref ; (13)
Eqs. (9) vanishes. >
> ref
>
:  
1:0 for P 4Pref :
3.2. Influence of the third shear invariant A value of Pref ¼ 10 was adopted in this work. The
reduction factor Rðy; eÞ is introduced in a multiplicative
Tri-axial experiments on concrete clearly demonstrate way in Eq. (9) as proposed by Riedel [25].
the substantial difference of shear strength in conventional
tri-axial compression (compressive meridian) and tri-axial
3.3. Rate dependence
extension (tensile meridian) [20]. A reduction of the shear
strength on the compressive meridian can be considered by
In the original HJC concrete model the strain rate
introducing a function R depending on the deviatoric polar
influence F ð_ Þ is defined as a linear function on a
angle y and the normalized out-of-roundness parameter e,
logarithmic scale of the strain rate and characterized by
as proposed by Willam and Warnke [21], viz.
the slope C. However, in order to avoid negative values of
2ð1  e2 Þ cos y þ ð2e  1Þ½4ð1  e2 Þ cos2 y þ 5e2  4e1=2 F ð_ Þ for relative strain rate values _  o1, while keeping the
Rðy; eÞ ¼ ,
4ð1  e2 Þ cos2 y þ ð1  2eÞ2 two-parameter formulation, the following expression pro-
(11) posed by Camacho and Ortiz [26] and largely used by
Børvik et al. [27] for metals is adopted, viz.
where y is defined as
! F ð_eq Þ ¼ ½1 þ _eq C , (14)
1 1 27jS ij j
y ¼ cos (12) where the parameter C describes the non-linear character
3 2s3eq
of the rate effect. Practically the relative increase in
in which jS ij j is the determinant of the deviatoric stress compression strength, found in the literature [28,29], is
tensor Sij . The parameter e is a shape factor that describes always related to the static case which involves strain rate
the out-of-roundness of the deviatoric trace and it values of about 105 s1 . For this reason we assumed such
represents the portion of strength on the tensile meridian a value as the reference strain rate _0 .
y ¼ 0 compared to the state of stress on the compressive
meridian y ¼ 60 . The function R is only defined in the 3.4. Pressure–volume response
sector 0oyop=3, since the polar radius Rðy; eÞ extends to
all polar directions 0oyo2p using threefold symmetry. The pressure–compaction response is illustrated in Fig.
Along the tensile and compressive meridian Rðy; eÞ takes 3. The pressure P is defined as a function of the volumetric
the values Rð0; eÞ ¼ e and Rðp=3; eÞ ¼ 1, respectively. strain
Convexity and smoothness of this function requires that
r
0:5oeo1: The upper limit e ¼ 1 defines a circle, while the m ¼  1, (15)
r0
lower limit e ¼ 0:5 forms a triangle in the deviatoric plane.
The influence of the out-of-roundness parameter e in a where r0 and r are the initial and actual densities,
plane stress situation was largely commented by Menetrey respectively. In compression, the behaviour is divided into
[22]. He found that the failure envelope is insensitive to the three regions. The first region is linear elastic and limited
shape factor in the biaxial-tension and tension-compressive by ðmcrush ; Pcrush Þ. At this state the second region starts,
regions. However, it strongly affects the biaxial-compression which involves crushing of the concrete and production of
 ARTICLE IN PRESS
M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303 295

plastic volumetric strains, and it continues until modified volumetric strain is introduced
ðmlock ; Plock Þ. The air voids are then assumed to be fully m  mlock
compressed out of the concrete (compaction damage). In m̄ ¼ (24)
1 þ mlock
the third region, the concrete is fully dense, i.e. all air voids
are removed from the material. The first and second with
regions are modelled by a classical incremental elasto- rgrain
mlock ¼  1, (25)
plastic-damaging formulation with a linear strain hard- r0
ening and scalar damage (compaction) assumptions, while where rgrain is the grain density. Replacing Eqs. (15) and
the third region is modelled by assuming that concrete is (25) into Eq. (24), one gets
completely elastic (crushed material with no tensile
r
capacity). m̄ ¼ 1 (26)
In particular, the total volumetric strain increment rgrain
is separated into its elastic and plastic contributions, which shows that m̄ represents the volumetric strain of the
according to aggregate. The behaviour of this material follows a non-
Dm ¼ Dme þ Dmp . (16) linear elastic behaviour of type
P ¼ K 1 m̄ þ K 2 m̄2 þ K 3 m̄3 . (27)
The incremental pressure is calculated as
This equation is used as long as DC ¼ 1 (fully compacted
DP ¼ HDmp , (17) material). The modified volumetric strain m̄ is used in the
where the volumetric strain hardening modulus H, see constitutive relation so that the constants K 1 ; K 2 and K 3
Fig. 3, is defined by are equivalent to those for a material without voids.
Because no volumetric plastic strains can be generated
Plock  Pcrush for tensile (negative) pressures, the elastic predictor scheme
H¼ . (18)
mlock works also in this regime. Thus, Eq. (23) can be re-written
as
The volumetric plastic strain increment Dmp is found with a
classical elastic predictor–plastic corrector procedure. Pi ¼ maxðK av mei ; Tð1  DÞÞ. (28)
Given a total volumetric strain increment Dm, and the Consequently, hydrostatic tension (negative pressure) is
previous values of the pressure ðPi1 Þ, total ðmi1 Þ and always limited by the threshold value of Tð1  DÞ. It is
plastic ðmpi1 Þ volumetric strains, a trial pressure Ptrial is interesting to observe that in the fully compacted zone
calculated assuming an elastic behaviour as D ¼ DC ¼ 1, the tensile threshold vanishes (loss of cohe-
Ptrial ¼ K av ðDm þ mi1 Þ ¼ K av m, (19) sion by complete crushing).

where the averaged elastic bulk modulus K av is defined 3.5. Damage behaviour
according to the compaction damage value (DC , to be
defined later) as As previously explained, three basic damage mechanisms
K av ¼ ð1  DC ÞK þ DC K 1 . (20) are always present under impact loading and perforation of
concrete materials [12,15,16]. Specifically, tensile cracking
Thus, for a virgin material ðDC ¼ 0Þ, K av ¼ K and for a (when mode I and II dominate), shear cracking (when
fully compacted material ðDC ¼ 1Þ, K av ¼ K 1 . By compar- mode II and III dominate) and pore compaction are
ing the pressure in the previous step and the trial observed. The main idea in this work is to treat each
calculation one can deduce the incremental volumetric mechanism separately, and for this purpose three internal
plastic strains according to damage variables DT , DS and DC , representing the tensile,
8 shear and compaction damage, respectively, need to be
<0 if Ptrial pPi1 ;
introduced.
p
Dm ¼ Ptrial  Pi1 (21)
: if Ptrial 4Pi1 :
ðK av þ HÞ 3.5.1. Tensile damage (brittle cracking)
Tensile cracking is the main source of non-linearity in
An updating of the plastic volumetric strain and pressure is
concrete and is due to concrete’s low strength in tension
done according to
when compared to compression. Therefore, concrete
mpi ¼ Dmp þ mpi1 , (22) structures include some kind of reinforcement (commonly
steel) in these zones where tensile stresses are expected.
Pi ¼ K av ðmi  mpi Þ ¼ K av mei . (23) Prediction of tensile cracking has been largely studied and
basically three types of models are commonly used: the
The pressure–volume behaviour in the fully compacted fictitious crack model [30], the smeared crack approach [31]
region ðP4Plock Þ follows a non-linear elastic behaviour and the damage mechanics based models [13,15,16]. In
(crushed material with no tensile capacity). For this a these models the introduction of a crack formation
 ARTICLE IN PRESS
296 M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303

criterion is required. Two of the most popular crack compacted material). Thus, an internal compaction da-
formation criteria are the Rankine criterion that is based mage variable DC can be defined as
on the maximum principal stress and the Saint Venant Dmp
criterion that is based on the maximum principal strain. DDC ¼ . (33)
mlock
Alternatively, a critical hydrostatic tension or a critical
hydrostatic tensile strain can be considered as a fracture Here, Dmp is the incremental plastic volumetric strain
criterion. Without putting too much effort on deciding (defined by Eq. (21)) and mlock is the plastic volumetric
which criterion predicts the crack pattern best, it was strain of the fully compacted granular material (an input
decided to use the hydrostatic tensile strain as the main parameter). It should be pointed out that the compaction
indicator for crack formation. For this, the minimum value damage variable introduced in this work agrees with the
of the volumetric strain m (in the tensile regime) attained interpolation factor F defined in the original HJC model
during the loading history is assumed as the equivalent [17,18].
strain for crack formation and it reads
3.5.4. Total damage
tT ¼ minðt1 t
T ; minð0; m ÞÞ, (29)
The combination of three different damage mechanisms,
where the superscripts t and t  1 indicate the actual and defined by the damage variables DT , DS and DC , into one
previous increment, respectively. The tensile damage representative scalar value D is questionable. However, it is
criterion is simply defined by believed that by including only DS and DC in the averaging
( procedure, reasonable results can still be obtained. This
0 for tT 40 ;
DT ¼ (30) proposal is based on the fact that for moderate and high
1 for tT p0 ; confining pressure levels both DS and DC describe the
‘‘ductile’’ behaviour of concrete and in this way a shear
where 0 ¼ f t F ð_eq Þ=K is the volumetric tensile strain
strength reduction in Eq. (9) by void collapse can be
threshold for crack formation. The factor F ð_eq Þ, see
included. A similar averaging procedure was proposed in
Eq. (14), is included to take into account the strain rate
[25]. The total damage effect is calculated here as
effect in the tensile regime. When dealing with reinforced pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
concrete and penetration problems, tensile cracking is ð1  DÞ ¼ ð1  DS Þð1  DC Þ. (34)
assumed to be less important than shear cracking. Thus,
Still under the framework of the scalar damage theory the
the damage variable DT is used only as a tensile damage
exclusion of DT in the averaging proposal of Eq. (34) is
indicator, i.e. it has no influence on the stress–strain
based on the argument that the tensile cracking effect is
behaviour and is used only for post-processing purposes.
reduced by the presence of the steel reinforcement (if
present) and the competition between crack closure and
3.5.2. Shear damage opening caused by the compressive and tensile waves
The cumulative damage development proposed in the propagation.
HJC original model [17] is adopted here, but damage from
shear and volumetric straining is separated. The evolution 4. Behaviour of the MHJC model
of the shear damage variable DS is defined by
Dpeq The MHJC concrete model has been implemented in the
DDS ¼ . (31) finite element code LS-DYNA [7] and this section presents
fp
a description of the parameter identification and some
The plastic strain to fracture fp is here adopted in the form numerical tests performed to verify the numerical imple-
fp ¼ a½P þ T  b Xðfp ÞMIN , (32) mentation and the sensitivity of some of the parameters.
The model requires the identification of 19 parameters: the
where a and b are constants (as in the original model). It initial density r0 ; the two elastic constants E (Young’s
transpires that the plastic strain to fracture fp increases as modulus) and n (Poisson’s ratio); five shear-pressure
P increases. The third damage constant ðfp ÞMIN is strength constants defined by the uniaxial compressive
introduced to allow for a finite amount of plastic strain ðf c Þ and tensile ðf t Þ strengths, the hardening coefficient ðBÞ,
to fracture the material. the hardening exponent ðNÞ and the maximum shear
strength ðSmax Þ; two rate sensitivity parameters (_0 , C);
3.5.3. Compaction damage three damage constants to define the softening behaviour
Damage compaction due to plastic volumetric strain is ða; b; ðfp ÞMIN Þ; and seven constants to describe the pore
included in the equation of state defined by the Pressure– compaction behaviour where the elastic region is defined
volume law. Two consequences should be taken into by Pcrush and mcrush , the crushing region by Plock and mlock ,
account with this mechanism: the cohesive strength of the and K 1 , K 2 and K 3 define the behaviour of a fully
concrete is lost during air voids collapse (i.e. the pore compacted material. In the next section we present the tests
compaction contributes to damage) and the bulk stiffness required and calibration procedure for the model para-
increases (i.e. the bulk modulus approaches to that of a meters identification.
 ARTICLE IN PRESS
M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303 297

The sensitivity of the strength parameters was analysed 250

R=1
for the uniaxial and biaxial compressive tests, because
they are placed at the two extreme meridians; namely 200

Stress (MPa)
the compressive meridian ðy ¼ 60 Þ for the first case
and the tensile meridian ðy ¼ 0 Þ for the second case. The 150
possibility to obtain different softening responses in a
uniaxial compressive test is illustrated by the adequate 100
choice of the damage parameters. The pressure–volume law R = 0.65
is described with the classical hydrostatic state of stress 50
and, finally, the influence of the lateral confining pressure is
illustrated when comparing the response of the uniaxial 0
0 0.003 0.006 0.009 0.012
compressive strain vs. the uniaxial compressive stress. Strain

Fig. 7. Biaxial compressive ðy ¼ 0 Þ response of the MHJC model.

4.1. Uniaxial compression ðy ¼ 60 Þ
50
The following example illustrates the sensitivity of the p
MHJC model to the shear strength parameters. For this (εf )MIN = 1.0
40

Stress (MPa)
p
purpose a uniaxial compression test (compression mer- (εf )MIN = 0.01
idian) is used. The pressure–volume behaviour is assumed 30
p
(εf )MIN = 0.001
linear elastic and damage is inhibited (i.e. ðfp ÞMIN ¼ 1:0). 20
The strain rate effect is not included. Typical values of
B ¼ 2:0 and N ¼ 0:75 are assumed, see Section 3. 10
Considering a uniaxial compressive and tensile strength 0
value of f c ¼ 43 MPa and f t ¼ 2:4 MPa, respectively, 0 0.003 0.006 0.009 0.012 0.015
Eq. (9) yields seq ¼ 0:985, or in other words, the uniaxial Strain
compressive strength input f c is practically recovered. In Fig. 8. Influence of damage parameter ðfp ÞMIN on the softening behaviour.
the original model, the set of parameters A ¼ 0:75 (i.e. the
cohesive strength), B ¼ 1:6 and N ¼ 0:75 is frequently
assumed [17,24]. However, with such values in Eq. (6) the to characterize the compressive meridian. However, it was
uniaxial compressive strength is overestimated by about pointed out that the HJC model was limited to select an
45%. In the modified version presented here, the cohesive adequate value B representing a good compromise for both
strength is calculated automatically and the uniaxial meridians at the same time. This compromise is relaxed in
compressive strength is in general recuperated. Fig. 6 the modified version by including the reduction effect of
illustrates the main result of this analysis. the third shear invariant through the function Rðy; eÞ. The
numerical result of the biaxial compressive test is illustrated
in Fig. 7, where the influence of Rðy; eÞ is clearly observed.
4.2. Bi-axial compression ðy ¼ 0 Þ Indeed, a prediction of about five times the uniaxial
compressive strength is obtained when Rðy; eÞ equals unity
The following example is for biaxial compression (tensile (i.e. a circular shape of the deviatoric trace), while an
meridian) where the influence of the strength parameter B increase of only 20% of the uniaxial compressive strength
(i.e. the shear strength dependence on the meridian) is is seen when Rðy; eÞ ¼ 0:65 (i.e. more triangular shape of
studied. In [12], it was concluded that in the original HJC a the deviatoric trace). This result agrees well with the
value of B ¼ 1:2 seems more adequate to characterize the experimental findings of Kupffer et al. [24].
tensile meridian, while a value of B ¼ 2:0 was appropriated
4.3. Strain softening response
50

40 Fig. 8 illustrates the different softening responses for a

Stress (MPa)

uniaxial compression test predicted by the MHJC model.

30 Here, values of a ¼ 0:04 and b ¼ 1 were assumed. The
softening response shows the influence of the fracture
20
strain threshold ðfp ÞMIN , see Eq. (32), on the residual
10 strength. As one can see, the residual stress is reduced when
this parameter is increased. Larger values of ðfp ÞMIN delay
0 damage development.
0 0.005 0.01 0.015 0.02 0.025 0.03
The slope of the softening response is affected by the
Strain
parameters a and b in Eq. (32), and the softening response
Fig. 6. Uniaxial compressive ðy ¼ 60 Þ response of the MHJC model. is most sensitive to the exponent b. The influence of this
 ARTICLE IN PRESS
298 M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303

50 200

40  = 0.5 150

Stress (MPa)
Stress (MPa)

30  = 1.0
100
 = 2.0
20
50
10
0
0 0 0.005 0.01 0.015 0.02 0.025 0.03
0 0.003 0.006 0.009 0.012 0.015 Strain
Strain
Fig. 11. Uniaxial compression versus uniaxial strain response; influence of
Fig. 9. Influence of damage parameter b on the softening behaviour. lateral pressure.

concrete are characterized by its shear strength increase

350 with an increasing pressure. The positive effect of the
confining pressure on shear strength is illustrated in Fig. 11
300
by comparing two extreme loading situations: the classical
250 uniaxial stress (with no lateral confinement) versus the
Pressure (MPa)

200 uniaxial strain situation (varying high lateral pressure to

allow only vertical displacements). This result is qualita-
150
tively in agreement with the expected response for concrete
100 materials.
50
5. Parameter identification
0
-0.03 0 0.03 0.06 0.09 0.12 0.15
-50 Volumetric strain
As previously indicated the MHJC concrete model
requires the identification of 19 parameters. The required
Fig. 10. Pressure–volume response of the MHJC concrete model. tests and calibration procedure for a complete identifica-
tion of the material parameters involved are:
parameter for a uniaxial compression test is illustrated in
Fig. 9 where values of ðfp ÞMIN ¼ 0:001 and a ¼ 0:04 were (a) Uniaxial compression test for identification of E, n, r0
assumed. and f c .
(b) Uniaxial tensile test for identification of T (using the
4.4. Cyclic hydrostatic pressure tensile strength f t ).
(c) Triaxial compression tests with varying confinement
This example describes the response of the MHJC concrete pressures to identify B, N and S max .
model under an alternating hydrostatic state of stress. The (d) Hydrostatic compression test to identify the Pressure–
pore-compaction behaviour adopted follows the original volume response defined by the following set of
formulation of the HJC concrete model. Note that the main parameters: Pcrush , mcrush , Plock , mlock and K 1 ; K 2 ; K 3 .
parameters involved are chosen for illustration purposes and (e) The damage parameters a and b can be identified from
can deviate from representative values for actual concrete fracture strain curves for different pressure levels
materials. The compressive and tensile strengths are assumed (triaxiality effect). Hence, fracture strain values from
to be the same ðf c ¼ f t ¼ 43 MPaÞ. The elastic pressure– tests proposed in (a) and (c) can be used to calibrate the
volume region is defined by Pcrush ¼ 43 MPa and mcrush ¼ damage parameters. The fracture strain threshold
0:01 (this involves a bulk modulus of K ¼ 4300 MPa). The ðfp ÞMIN is more difficult to identify. However, to avoid
‘‘plastic region’’ is bounded by Plock ¼ 200 MPa and fracture from low magnitude tensile stress the values,
mlock ¼ 0:10. The fully compacted material is described by ðfp ÞMIN can be assumed as the fracture strain of a
the three constants K 1 ¼ 8500 MPa, K 2 ¼ 17 100 MPa and uniaxial compression test.
K 3 ¼ 20 800 MPa taken from the original paper. The cyclic (f) Finally, the rate sensitivity parameters _0 and C require
pressure–volume response is illustrated in Fig. 10. uniaxial compression tests for different strain rate
levels.
4.5. Influence of the confining pressure
It is believed that among the most important issues when
It is well known that for metals the yield properties are developing mathematical models for the characterization
quasi-independent of the hydrostatic pressure, contrary to of materials behaviour one should address, first, the
geo-materials. Indeed, pressure-dependent materials like compromise between accuracy and simplicity and, second,
 ARTICLE IN PRESS
M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303 299

the compromise between the material parameters and tests 8

required for identification. At first, a complete calibration 7
of the parameters involved in MHJC seems complex and 6
expensive, because one needs the stress–strain curve to 5
C_48
failure in uniaxial compression and at least two triaxial

σeq
*
4 MHJC: B=1.4, N = 0.65
compression tests with different confining pressure levels. 3 C_140
From these three tests E, n, r0 , f c , B, N, a, b and ðfp ÞMIN
2 MHJC: B =1.3, N=0.45
can be identified. There is no experimental evidence of the
1
existence of a shear strength threshold ðS max Þ, and if it does
0
exist its identification could demand extremely high lateral 0 2 4 6 8 10 12 14
pressure levels. Therefore, in this study the shear strength P*
threshold will be given a high value, S max X7, as is
commonly assumed in the literature [17,18]. Next, the Fig. 12. Shear-pressure response reported by Hanchak et al. [34] and the
identification of B and N parameters used in the MHJC concrete model.
splitting test can be used to identify the tensile strength f t .
Furthermore, the pressure–volume law can in principle be
extracted from one hydrostatic compression test. However,
the complete representation of this law will require 800

Residual velocity (m/s)

extremely high pressures, rendering difficult the identifica-
tion of the parameters Plock , mlock and ðK 1 ; K 2 ; K 3 Þ. Finally, 700 C_48, C = 0.04
at least three uniaxial compression tests with different C_140, C= 0.025
strain rates should be performed for identification of the 600
rate sensitivity parameters _0 and C.
500
As a result, a total of eight tests (i.e. five uniaxial and
three triaxial tests) are needed for the complete identifica- 400
tion of the parameters required in the MHJC concrete 0 0.0001 0.0002 0.0003 0.0004 0.0005
model. Another valid alternative consists of including a Time (s)
smaller number of tests complemented with inverse Fig. 13. Identification of the rate sensitivity parameter C by inverse
modelling and assumptions related to some values. For modelling.
instance, the material strength characterization requires the
uniaxial compressive strength f c and the initial density r0 ,
while the uniaxial tensile strength f p t can
ffiffiffiffiffi be related to the Pcrush ¼ f c =3 and mcrush ¼ f c ð1  2nÞ=E, while mlock can be
compressive strength, i.e. f t ¼ 0:54 f c , according to the assessed from the density of the gravel material ðrgrain Þ and
FIB/CEB [32]. Also Young’s modulus E can be related to the initial density of concrete ðr0 Þ using mlock ¼
the compressive strength, i.e. E ¼ 11 700ðf c Þ0:3 [33]. A rgrain =r0  1. Unfortunately, the identification of Plock
Poisson’s ratio of n ¼ 0:2 can be assumed. The shear- and ðK 1 ; K 2 ; K 3 Þ requires hydrostatic tests at very high
pressure behaviour can be represented by assuming values pressure levels. It is believed, based on results in the
of 1:8pBp2:0 and 0:60pNp0:80 according to the literature [17,18], that Plock X600 MPa and K 1 ¼
experimental results in the compressive meridian presented 8500 MPa, K 2 ¼ 17 100 MPa and K 3 ¼ 20 800 MPa are
in the literature [19]. In these results concrete qualities representative values for concrete materials. In order to
between 22 and 74 MPa compressive strengths were avoid the identification of ductility diagrams (fracture
reported. It should be pointed out that ‘‘special’’ concretes strain versus triaxiality), values of a ¼ 0:04 and b ¼ 1:0 can
can deviate largely from the B and N values previously be assumed to represent the concrete damage behaviour
proposed. As a matter of fact, in the experimental program with reasonable accuracy [17,18]. The fracture parameter
on perforation of concrete slabs, conducted by Hanchak ðfp ÞMIN with values of 0.01 (low strength quality) and 0.005
et al. [34], two concrete qualities C_48 and C_140 with (high strength quality) can be introduced to represent
special glassy gravel (called Steilacoon Glacial) were used. different embrittlement characteristics.
The hardness of this type of gravel is slightly lower than a In the original paper [17] the strain rate effect was
quartz mineral type. The shear-pressure response of a series assumed to be independent of the concrete strength, while
of triaxial tests reported in Hanchak et al. [34] is illustrated there is experimental evidence that high strength concretes
in Fig. 12. By a fitting procedure one finds values of are less sensitive to variation in strain rates than conven-
B ¼ 1:4, N ¼ 0:65 and B ¼ 1:30, N ¼ 0:45 for the concrete tional ones [28,29]. Based on this argument it was decided
qualities of C_48 and C_140, respectively, hence invalidat- to use inverse modelling in the high impact velocity region
ing the proposed B and N values for ‘‘normal’’ concretes. (i.e. from 500 to 1000 m/s) to identify the rate sensitivity
To sum up, two triaxial compression tests would be parameter C. In the inverse modelling procedure we
preferable for accurate identification of B and N. Next, identify the C value which gives the best numerical fit
some of the parameters involved in the pressure–volume (e.g. residual velocity) to one experimental point and use it
law can easily be identified. For instance, from elasticity to predict the entire residual velocity diagram. With this
 ARTICLE IN PRESS
300 M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303

procedure values of C ¼ 0:04 for the C_48 concrete and compaction curves, triaxial tests were performed under
C ¼ 0:025 for C_140 concrete were found based on the various confining pressure levels such that shear strength
penetration results reported by Hanchak et al. [34]. Indeed, versus pressure curves could be established, see Fig. 12.
for the C_48 concrete slab an experimental residual A 30-mm, smooth-bore powder gun was used to launch
velocity of 602 m/s was found when the impact velocity 0.50 kg ogival-nose steel projectiles with a length of
was 749 m/s while for the C_140 concrete slab with an 143.7 mm and a diameter of 25.4 mm. In the tests, initial
impact velocity of 743 m/s an experimental residual and residual projectile velocities were measured. These
velocity of 545 m/s was found. By using the MHJC values were used to construct initial versus residual velocity
concrete model in the numerical simulations these values curves for the two concrete qualities and from these
were practically recuperated (615 and 544 m/s), as illu- diagrams ballistic limits were deduced. Hanchak et al.’s
strated in Fig. 13. main conclusion was that even though the unconfined
compressive strength was increased by a factor of three, the
ballistic limit velocity only increased by 20%. This is due to
6. Numerical simulation of perforation of concrete slabs the fact that the shear strength ratios of the two concrete
materials studied approach each other for relatively small
6.1. Experimental study by Hanchak et al. [34] confining pressures.
The ballistic limit computations are based on the test
performed by Hanchak et al. [34], where square reinforced 6.2. Concrete slabs with f c ¼ 48 MPa
concrete plates of 610  610  178 mm3 were tested. Three
layers of square-pattern reinforcement steel rods with a Finite element analyses with 2D axisymmetric elements
diameter of 5.6 mm were used. Two concrete qualities C_48 were performed. A reduced integration scheme with
and C_140 with uniaxial compressive strength f c ¼ 48 and hourglass control was adopted. The finite element model
140 MPa, and uniaxial tensile strength f t ¼ 4 and 5 MPa, is illustrated in Fig. 14. For the concrete slab a total of 100
respectively, were used. In addition to the pressure– elements were used through the thickness and 50 elements
along the radius. The steel reinforcement was not included
CONCRETE PENETRATION (MHJC)
in the simulations since the effect on the perforation
Time = 0
resistance was found negligible [34]. The set of input
parameters assumed in the simulations are indicated in
Table 2. The steel projectile was modelled using a von
Mises material model (Mat_003 in LS-DYNA) with linear
isotropic hardening. The main data used for the projectile
are E ¼ 200 GPa, n ¼ 0:3, sY ¼ 1:72 GPa and E T ¼
15:0 GPa (tangent modulus). No strain rate effect was
considered. The original density of the projectile
(8020 kg=m3 ) was slightly modified to 8300 kg=m3 to obtain
the total launch package mass of 0.53 kg. In the present
calculations, we adopted the element erosion option of LS-
DYNA with a criterion based on the maximal principal
strain with a failure strain value of (1 ÞMAX ¼ 1:0 (based on
similar values used in [18]). The 2D_automatic_single_sur-
face contact option of LS-DYNA was used to define the
Y
contact behaviour between steel and concrete without
X friction. The rate sensitivity parameter value of C ¼ 0:04
Fig. 14. 2D axisymmetric finite element model used in the penetration was used based on the residual velocity predictions of
analysis (target radius is 305 mm and target thickness is 178 mm). Section 5. The number of elements used in this study, i.e.

Table 2
Material parameters for concrete with f c ¼ 48 MPa

r0 ðkg=m3 Þ E (MPa) n B N C Smax

2440 11 700 (f c Þ0:3 0.2 1.4 0.65 0.04 7

f c (MPa) f t (MPa) _0 ðs1 Þ a b ðfp ÞMIN

48 4 5 0.04 1.0 0.01
1  10
Pcrush (MPa) mcrush Plock (MPa) mlock K 1 (MPa) K 2 (MPa) K 3 (MPa)
16 f c ð1  2nÞ=E 800 0.1 8500 17 100 20 800
 ARTICLE IN PRESS
M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303 301

100 elements over the target thickness, was similar to that higher than 400 m/s. The predicted ballistic limit (325 m/s)
used in the original paper by HJC [17]. However, the deviates by less than 5% from the experimental value
influence of the mesh size, mesh bias, friction effects and (340 m/s). The tensile damage (cracking) evolution during
erosion strain parameters are not included in this work. the perforation process is illustrated in Fig. 16.
These issues will be considered in a forthcoming study.
The numerical predictions of the residual velocity are 6.3. Concrete slabs with f c ¼ 140 MPa
compared with the experimental findings [34] for the C_48
concrete in Fig. 15. The MHJC compares very well with the The uniaxial compressive and tensile strength of this
experimental values, in particular for impact velocities concrete was f c ¼ 140 and f t ¼ 5 MPa, respectively. A new
set of input parameters based on the tests performed by
Hanchanck et al. [34] is compiled in Table 3. We used the
1000
material data of the previous example for the projectile.
The element erosion option with a failure strain value of
Residual Velocity (m/s)

800 (1 ÞMAX ¼ 1:0 and the 2D_automatic_single_surface con-

tact option without friction of LS-DYNA were once again
600 Exp_C48 adopted. A rate sensitivity parameter value of C ¼ 0:025
MHJC
was found for this case (see Section 5). In addition, the
400 Exp_C140
MHJC plastic fracture strain limit is reduced to ðfp ÞMIN ¼ 0:004 in
200
order to account for the brittle nature of high strength
concrete. The numerical predictions of residual velocity are
0 compared with the experimental findings [34] in Fig. 15.
200 400 600 800 1000 1200 Also in this case the MHJC compares well with the
Impact velocity (m/s)
experiments, and the predicted ballistic limit (370 m/s)
Fig. 15. Ballistic limit predictions using the MHJC for C_48 and C_140 deviates by less than 8% from the experimental one
concrete qualities. (400 m/s).

Fig. 16. Tensile damage development during perforation of the C_48 concrete slab.
 ARTICLE IN PRESS
302 M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303

Table 3
Material parameters for concrete with f c ¼ 140 MPa

r0 ðkg=m3 Þ E (MPa) n B N C Smax

2520 11 700 (f c Þ0:3 0.2 1.30 0.45 0.025 7

f c (MPa) f t (MPa) _0 ðs1 Þ a b ðfp ÞMIN

140 5 1  105 0.04 1.0 0.004

Pcrush (MPa) mcrush Plock (MPa) mlock K 1 (MPa) K 2 (MPa) K 3 (MPa)

46 f c ð1  2nÞ=E 800 0.04 8500 17 100 20 800

6.4. Discussions
Acknowledgements
Both the original and the modified version of the HJC
model are able to reproduce the experimental findings of The authors gratefully acknowledge the financial sup-
the penetration tests of Hanchak et al. [34]. It is believed port of the Department of Applied Mechanics and
that during the penetration process the energy dissipated Corrosion at SINTEF, Materials and Chemistry and The
(dominated by shear damage) in the original HJC model is Norwegian Defence Estates Agency, Research & Develop-
of similar magnitude of that predicted by the MHJC ment Department.
model. So if the normalized pressure levels generated in the
slabs during the penetration test are ranging between 1 and
5 the shear response of the HJC represent an averaged References
response of the MHJC model, see Fig. 4. However, large
[1] Corbett GG, Reid SR, Johnson W. Impact loading of plates and
differences between these models are expected if the shells by free-flying projectiles: a review. Int J Impact Eng 1996;
normalized pressures in the target are less than 1 (thin 18(2):141–230.
slabs) or larger than 5 (massive blocks). On the other hand, [2] Børvik T, Langseth M, Hopperstad OS, Polanco-Loria MA, Ballistic
our predictions are in agreement with both concrete perforation resistance of high performance concrete slabs with
qualities used by Hanchak et al. [34]. To this end, the different unconfined compressive strengths. In: Brebbia CA, de
Wilde WP, editors. Proceedings of first international conference on
enhancements proposed to the original HJC concrete high performance structures and composites. Sevilla, Spain: WIT
model show more consistency with classical uniaxial press (ISBN 1-85312-904-6); 2002. p. 273–282.
compression and biaxial static tests reported in the [3] Kennedy RP. A review of procedures for the analysis and design of
literature, eliminate the cohesive parameter (difficult to concrete structures to resist missile impact effects. Nucl Eng Design
1976;37:183–203.
identify), introduce classical observed damage mechanisms.
[4] Ben-Dor G, Dubinsky A, Elperin T. Ballistic impact: recent advances
Further, the inverse modelling procedure used to identify in analytical modeling of plate penetration dynamics—a review. Appl
the strain rate parameter allows capturing the influence of Mech Rev 2005;58:355–71.
the concrete quality. The proposed modifications are [5] Li QM, Reid SR, Wen HM, Telford AR. Local impact effects of hard
expected to improve the numerical predictions of the missiles on concrete targets. Int J Impact Eng 2005;32:224–84.
original version. [6] Abaqus v 6.2 User’s Manual Hibbitt, Karlsson and Sørenson Inc.
Providence, RI, USA, 2001.
[7] LS-DYNA Keyword User’s Manual Ver. 950, Livermore Software
Technology Corporation, LSCT, 1999.
[8] AUTODYN, Theory Manual, Century Dynamics Ltd. Horsham,
7. Concluding remarks England, 1997.
[9] Teng TL, Chu YA, Chang FA, Chin HS. Numerical analysis of
oblique impact on reinforced concrete. Cement Concr Compos
The Holmquist–Johnson–Cook (HJC) model for con- 2005;27:481–92.
crete materials subjected to impact loading has been [10] Leppänen J. Concrete subjected to projectile and fragment impacts:
investigated in the current study, and some modifications modelling of crack softening and strain rate dependency in tension.
are proposed to the pressure-shear behaviour, the strain- Int J Impact Eng 2006;32:1828–41.
rate sensitivity term and the damage description. A critical [11] Tai YS, Tang CC. Numerical simulation: the dynamic behaviour of
reinforced concrete plates under normal impact. Theor Appl Fract
review of the constitutive model with alternative proposals Mech 2006;45:117–27.
for parameter identification is provided. Ballistic limit [12] Polanco-Loria M. Constitutive models for concrete materials under
assessments with deviations under 8% were found when impact loading conditions. SINTEF Report STF24 F00315, Trond-
compared to experimental results from the literature, heim, Norway, 2001.
indicating that the MHJC model represents a good [13] Polanco-Loria M. Numerical modelling of plain and reinforced
concrete by damage mechanics. PhD dissertation, The Norwegian
compromise between simplicity and accuracy for Institute of Technology, Norway, 1997.
large-scale computations of concrete plates impacted by [14] Isenberg J, editor. Finite element analysis of reinforced concrete
projectiles. structures II. Proceedings of the international workshop sponsored by
 ARTICLE IN PRESS
M. Polanco-Loria et al. / International Journal of Impact Engineering 35 (2008) 290–303 303

the U.S. National Science Foundation, Defence Nuclear Agency and [24] Kupffer H, Hilsdorf H, Rush H. Behaviour of concrete under biaxial
the U.S. Army Corps of Engineers. New York, USA: ASCE; 1991. stresses. ACI J 1969;23:656–65.
[15] Mazars J. Application de la mecanique de l’endommagement au [25] Riedel W. Ein makroskopisches, modulares Betonmodell für Hydro-
comportement nonlineaire et la rupture du beton de structure, These codes mit Verfestigung, Scädigung, Entfestigung, Drei-Ivariantenab-
de Doctorate d’Etat, LMT, Universite de Paris, France, 1984. hägigkeit und Kappe, Berich 7/98, Fraunhofer Institut für
[16] Burlion N. Compaction des betons: elements de modelisation et Kurzzeitdynamik-Ernst-Mach-Institut, Freiburg i. Br 1998.
caracterisation experimentale. PhD dissertation, LMT, ENS de [26] Camacho GT, Ortiz M. Adaptive Lagrangian modelling of ballistic
Cachan, France, 1997. penetration of metallic targets. Comput Methods Appl Mech Eng
[17] Holmquist TJ, Johnson GR, Cook WH. A computational constitu- 1997;142:269–301.
tive model for concrete subjected to large strains, high strain rates [27] Børvik T, Langseth M, Hopperstad OS, Malo KA.
and high pressures. In: Proceedings of 14th international symposium Ballistic penetration of steel plates. Int J Impact Eng 1999;18(2):
on Ballistics, Quebec, Canada, 1993. p. 591–600. 855–86.
[18] Johnson GR, Beissel SR, Holmquist TJ, Frew DJ. Computer radial [28] Bischoff PH, Perry SH. Compressive behaviour of concrete at high
stresses in a concrete target penetrated by a steel projectile. In: Jones strain rates. Mater Struct, RILEM 1991;24:425–50.
N, Talaslidis DG, Brebbia CA, Manolis GD, editors. Proceedings of [29] CEB. Concrete structures under impact and impulsive loading.
structures under shock and impact V, held at the Aristotle University CEB bulletin d’information No. 187, Laussanne, Switzerland,
of Thessaloniki, Greece (ISBN:1853125903). Southampton, UK: 1988.
Computational Mechanics Publications; 1998. p. 793–806. [30] Hillerborg A, Modeer M, Petersson PE. Analysis of crack formation
[19] Polanco-Loria M. Improvements to the HJC concrete model in LS- and crack growth in concrete by means of fracture mechanics and
DYNA SINTEF Report STF24 F01286, Trondheim, Norway, 2002. finite elements. Cement Concr Res 1976;6:773–82.
[20] Chen WF. Plasticity in reinforced concrete. New York: Mc Graw [31] Rots JG. Computational modeling of concrete fracture. PhD
Hill; 1982. dissertation, Delft University of Technology, Department of Civil
[21] Willam KJ, Warnke EP. Constitutive model for the triaxial behaviour Engineering, Delft, The Netherlands, 1988.
of concrete. In: International association of bridges and structural [32] FIB/CEB State of the art report: high strength concrete, Switzerland,
engineers, seminar on concrete structures subjected to triaxial 1990.
stresses, Paper III-1, IABSE Proceedings 19, Bergamo, Italy, 1975. [33] Sellevold E, Justnes H, Smeplass S, Hansen EA. Selected properties
[22] Menetrey P. Numerical analysis of punching failure in reinforced of high performance concrete. In: Grutzek MW, Sarkar SL, editors.
concrete structures. PhD dissertation, Ecole Polytechnique Federale Advances in cement and concrete. New England: ASCE; 1994.
de Lausanne, Switzerland, 1994. p. 562–609.
[23] Launey P, Gachon H. A study of the failure of concrete under [34] Hanchak SJ, Forrestal MJ, Young ER, Ehrgott JQ. Perforation of
combined compressive stress. University of Illinois, Engineering concrete slabs with 48 MPa and 140 MPa unconfined compressive
Experimental Station, Bulletin No. 185, 1970. strengths. Int J Impact Eng 1992;12(1):1–7.