International Journal of Impact Engineering 69 (2014) 55e68

Contents lists available at ScienceDirect

International Journal of Impact Engineering

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

Tensile behavior of concrete under high loading rates

_
Jo
sko O

zbolt, Akanshu Sharma*, Bars Irhan,
Emiliano Sola
Institute for Construction Materials, University of Stuttgart, 70569 Stuttgart, Germany

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 17 May 2013
Received in revised form
7 February 2014
Accepted 8 February 2014
Available online 22 February 2014

The experimental and theoretical studies show that the inuence of loading rate on tensile behavior of
concrete is relatively strong. Dynamic tensile resistance of concrete is difcult to measure by direct
tensile test. Therefore, the indirect tensile tests such as split Hopkinson bar tests are used. The evaluation
of experimental measurements shows that after reaching a certain critical strain rate, tensile resistance
progressively increases with increasing strain rate. In this paper, the authors attempt to investigate and
discuss: (i) the reason for progressive increase of tensile resistance beyond a certain strain rate and (ii)
whether the dynamic resistance can be attributed only to material strength or whether some other
factors also contribute towards the same. To answer these questions, numerical analysis on two different
types of examples is carried out: (i) Simple elastic-cohesive nite element (FE) model subjected to direct
tension and (ii) FE model of indirect tension test on modied split Hopkinson bar. The results are
evaluated in terms of apparent and true strength and compared with experimental results. It is found
that under static loads, the true and apparent strengths are always equal, while under dynamic loads
they are different. The true strength is controlled by the rate dependent constitutive law and the
apparent strength is signicantly inuenced by the size of the fracture process zone and the size of the
specimen. Evaluation of numerical results shows that concrete fracture energy is approximately a linear
function of strain rate (semi-log scale) and is controlled by the rate dependent constitutive law. It is
concluded that the results of any indirect tension test such as split Hopkinson bar test need careful
interpretation, i.e. due to the fact that concrete specimen is damaged, and not elastic, the results of
measurement need to be corrected.
2014 Elsevier Ltd. All rights reserved.

Keywords:
Concrete
Dynamic fracture
Rate sensitivity
True and apparent strength
Microplane model

1. Introduction
Several experimental [1e9], theoretical [10e13] and numerical
[14e23] studies have shown that loading rate signicantly inuences the resistance and failure mode of concrete structures. In
linear elastic materials or within linear elastic range no rate
sensitivity can be observed, while, in case of materials that exhibit
damage and fracture phenomena, such as concrete, there is signicant inuence of loading rate on strength and structural
response. This indicates that rate sensitivity might be closely
related to damage and softening of the material, i.e. more damage,
the stronger will be the inuence of loading rate on structural
response. This is conrmed by experimental results which show
that concrete-like materials exhibit the highest rate sensitivity

* Corresponding author.
E-mail addresses: akanshu.sharma@iwb.uni-stuttgart.de,
yahoo.co.in (A. Sharma).
http://dx.doi.org/10.1016/j.ijimpeng.2014.02.005
0734-743X/ 2014 Elsevier Ltd. All rights reserved.

akanshusharma@

whereas brittle materials (e.g. glass) are much less sensitive to the
strain rate [3].
The response of concrete structures depends on time dependent
loading through three different effects [17,18]: (1) through the rate
dependency of the growing micro-cracks (inuence of inertia at the
micro-crack level), (2) through the viscous behavior of the bulk
material between the cracks (creep of concrete or viscosity due to
the water content) and (3) through the inuence of inertia forces,
which can signicantly change the state of stresses and strains of
the material. When modeling concrete in the framework of mesoor macro-continuum, the rst two effects can be accounted for by
the constitutive law and the third effect should be automatically
accounted for through dynamic analysis where the constitutive law
interacts with inertia. Depending on the material type and loading
rate, the rst, second or third effect may dominate. For quasi-brittle
materials, such as concrete, the rst two effects are important for
relatively low and medium strain rates. For higher strain rates
(impact) the last effect dominates, however, the rate dependency
cannot be neglected.

56

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

The experimental evidence shows that concrete exhibits the

strongest inuence of loading rate under tensile load [24]. The
evaluation of experimental results, under assumption that concrete
before tensile cracking is linear elastic, show that after reaching a
certain critical strain rate tensile resistance progressively increases
with increasing strain rate. To experimentally study the uni-axial
tensile behavior of concrete is difcult even under static load.
Therefore, indirect methods such as Brazilian test, compact tension
tests, etc. are often employed. Moreover, it is very difcult to
evaluate the direct dynamic uni-axial tensile behavior of concrete
even numerically, since at high loading rates the failure always
occurs locally, near the loading points. Therefore, under dynamic
loads, the problem is studied through indirect tests such as split
Hopkinson bar tests [6,21,25].
Typically, in the split Hopkinson bar test, the measurement of
concrete strength is based on the theory of uni-axial (one dimensional) wave propagation through elastic media and measurements
of strain and strain rates in the bar. The results of experiments [24]
show that for strain rates lager than approximately 1/s the resistance increases progressively with the increase of strain rate. The
questions that arise are: (i) what is the reason for such an increase
and (ii) whether the experimentally measured resistance can be
attributed only to the material strength or there are some other
effects that need to be considered when evaluating results of
experimentally measured data.
Though the split Hopkinson bar tests are widely used for
determining the dynamic compressive and tensile strength for
concrete, there are certain issues that are still unclear to the researchers. Wu et al. [26] commented that the SHPB results are
reliable only for ductile materials such as metals, whereas the results may contain signicant errors when measurements are made
using SHPB on concrete-like materials. Li and Meng [27] reported
that the apparent dynamic strength enhancement beyond the
strain rate of 100/s is strongly inuenced by the hydrostatic stress
effect due to the lateral inertia connement in an SHPB test. They
further strongly commented that this apparent dynamic strength
enhancement has been wrongly interpreted as strain-rate effect
and has been adopted in both dynamic structural design and
concrete-like material models for analytical and numerical simulations, which may lead to over-prediction on the dynamic strength
of concrete-like materials [27]. Based on their study, which was
primarily based on investigating the dynamic compressive strength
of concrete, they recommended further experimental and numerical studies to understand the genuine strain rate effects.
The contribution of inertia in leading to the rise in concrete
compressive strength under dynamic loading is well acknowledged
and generally acceptable [26e35]. Though, the inuence of inertia
on tensile strength is not so well-accepted as that on compressive
strength, recent experimental and numerical studies show a very
high contribution of inertia on the tensile strength beyond a certain
strain rate [18,36e38]. Based on numerical work supported by
theoretical equations, Cusatis [36] reported that the contribution of
the inertia forces to the tensile DIF is about ten times bigger than
the contribution of the inertia forces to the compressive DIF. He also
commented that experimental data showing signicant difference
in the compressive and tensile DIF are tted very well by the numerical model if inertia effects (as well as other structural features
of the tests such as specimen geometry and boundary conditions
effects) are included in the numerical simulations [36]. Weerheijm
and Van Doormaal [37] reported that it is possible that even in case
of notched specimens, at the high loading rates the inertia effects
dominate the failure process and stress concentrations have a minor effect on the observed specimen strength. They commented
that detailed numerical modeling is needed to quantify the effect.
Cotsovos and Pavlovic [38] suggested that under dynamic tests the

concrete specimen must be viewed as a structure since its behavior

is directly linked to the inertia effect of its mass and the boundary
conditions. It has been shown by several examples [18,36,38,40]
that the approach where the material properties is not articially
modied to consider the steep rise in DIF and considering the inertial effects leads to realistic simulation of the dynamic tensile
behavior of the concrete. Lu and Li [39] concluded from their study
that micro-crack inertia is one of the mechanisms responsible for
the strain-rate sensitivity of the tensile strength of concrete-like
materials and that the observed increase of tensile strength with
strain rate from dynamic tensile tests can be largely attributed to
the inertia effects of micro-cracks.
In general, when subjected to impact loads, the body is under
dynamic equilibrium and the applied load and reactions are
essentially not equal due to the presence of inertial component. The
applied load is partly balanced by the inertial forces and partly by
transmitted reactions. Up to a certain strain rate, the inertial
component is insignicant and both applied load and reactions are
more or less same, though both of them are higher than the corresponding values at static loading rates. This is termed as rate
sensitivity and is totally attributed only to material behavior. Once
that critical strain rate is crossed and inertial forces become signicant, the applied load becomes signicantly higher than the
reaction. The relative magnitude of applied load with respect to the
reactions keeps on increasing with increased loading rate. This is
termed as inuence of inertia and cannot be attributed to the
material resistance. Obviously, inertia is dependent on size and
shape of the specimen/component. This suggests that the
measured applied load during impact does not provide the true
material strength but inherently has an inertial component and
thus it can be referred to as apparent strength. Inertia in dynamics
is always present, even if the material behavior is linear elastic.
However, it is important to note that in quasi-brittle materials,
which undergo damage and softening, inertia is activated also
because of the material softening and change of the failure mode.
This is especially true in case of high strain rates, for both, dominant
compression and tension.
This work aims at clearly bringing out the difference in true and
apparent tensile strength of concrete subjected to high loading
rates and to understand and nd out the reason for this difference.
In the paper, after discussing typical test methods for measuring
material strength at high strain rates, the rate sensitive microplane
model, which is used in the subsequent numerical simulations, is
briey discussed. The possible effect of inertia is demonstrated
through consideration of dynamic equilibrium on a simple
example. This consideration is also conrmed by numerical examples in which static and dynamic analyses of a simple elasticcohesive FE model is subjected to direct tension. Three different
sizes of elements are loaded at varying loading rates. The band of
numerical results for tensile strength is compared with that of the
experimental results available in literature. Finally, in order to
investigate the validity of interpretation of measured dynamic
tensile strength and fracture energy in the experiments, 3D nite
element simulation of recently performed experiments on the so
called modied split Hopkinson bar [8,41] is carried out for
different loading rates and for two different materials, quasi-brittle
(concrete) and brittle. The results are evaluated, compared with the
experimental results and the corresponding conclusions are drawn
out.
2. Indirect tension tests on concrete specimens
Hopkinson [42] experimentally tested behavior of various materials at high strain rates. On a long bar he generated compressive
pulse by explosive charge or impacting bullet. At the end of the bar

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

the compressive loading wave reected as a tensile loading wave

and caused fracture of brittle materials such as concrete, rock etc.
Later, Kolsky [43] make the method operable for the practical
measurements. The method is known as split Hopkinson (pressure)
bar (SHB) schematically shown in Fig. 1 for compressive testing. The
bar can be modied such that the specimen can be also tested for
different loading conditions, e.g. compression. In this kind of
experimental set-up the specimen should be in the mechanically
balanced state.
The theory of uni-axial wave propagation through elastic media
[43] leads to the following relations for the evaluation of strain, s
and strain rate, _ s :

s 

2c0
L

Zt
r dt and _ s 

2c0 r
L

(1)

1
ft rc0 vpb with c0
2

57

s
E

(3)

The main assumption in Eqs. (1)e(3) is that before the formation

of discrete crack(s) there is linear elastic behavior of the material.
However, as will be shown later, for concrete-like materials this
assumption is critical because at high strain rates before localization of damage in single or more discrete cracks the entire specimen suffer signicant damage. This means that longitudinal wave
velocity c0 must account for this damage, otherwise the predicted
material strength, using for instance Eq. (3), can be signicantly
overestimated. Moreover, in the case of SHB the inertia forces can
possibly have inuence on the measured stress.
3. Three-dimensional numerical analysis

where c0 is the wave propagation velocity, r is the strain due to

reected pulse in the incident bar and L the specimen length. The
stress in the specimen can be calculated as:

ss Et

A
As

(2)

where E, A and t are Youngs modulus, cross-sectional area and

strain due to transmitted pulse in the transmitter bar, respectively,
while As is the cross-sectional area of the specimen. High strain
rates are reached if r is large, which can be achieved by using a
specimen diameter that is small compared to incident bar or in case
when the bar material is much stiffer than the specimen.
The reected and transmitted strain can be measured using
strain gauges and using Eqs. (1) and (2), the stress in the specimen
can be calculated as a function of strain rate. By recording stresses
ss at different strains s it is also possible to calculate fracture energy of concrete.
Using SHB for tensile loading, strain rates up to 10/s can be
obtained [41]. For higher strain rates usually modied SHB (MSHB)
is used, which is schematically shown in Fig. 2. In this test set-up
the specimen is loaded in a quasi-uni-axial stress state and it is
unbalanced since one end is in contact with the Hopkinson (incident) bar whereas the other end is free. Typical strain rates in the
range of 20e200/s can be reached with this technique [41].
With MSHB it is necessary to measure so called pull-back velocity (vpb) at the free end of the specimen. Based on the theory of
uni-axial wave propagation through elastic media tensile strength
(ft) can be calculated as:

3.1. Transient nite element analysis

In this work, both static and dynamic numerical analysis is
performed. The static FE analysis is performed using an implicit 3D
FE code, which is based on the incremental secant stiffness
approach [44]. The dynamic analysis is performed in the framework
of 3D transient multi body dynamic FE analysis, where the body
modeling projectile is allowed to impact at the body modeling
target with a specied impact velocity. The system of unknown
displacements in each time step Dt is calculated by solving the
following system of equations (Voigt notation) using an explicit
direct integration scheme [44]:

t Cut
_
Mu
GT u; tl  ft 0
Gu; t 0

(4)

where M mass matrix, C damping matrix,

nodal accelerations, u_ nodal velocities, f(t) resulting
u
nodal forces, G is the element contact displacement constrain
matrix and l is the vector of unknown Lagrange multipliers [44].
The resulting nodal forces are calculated as:

ft f ext t  f int t
with : f ext t external nodal forces;
f int t internal nodal forces

(5)

The external nodal forces are known nodal loads. The internal
nodal forces are unknown and they are calculated by the integration of the stresses over the nite elements. In the dynamic FE code
used, the mass and damping matrices are assumed to be diagonal.
In static and dynamic analysis, standard eight or four node solid
elements are used. To obtain results objective with respect to the
element size, crack band approach is employed as a regularization
method [45].
3.2. Constitutive law e rate sensitive microplane model for concrete

Fig. 1. Schematic of split Hopkinson bar (SHB).

In the microplane model the material response is calculated

based on the monitoring of stresses and strains in different predened directions. The macroscopic stress tensor is obtained by
integrating microplane stresses in a thermodynamically consistent
way. The constitutive framework is similar to discrete type of the
models (e.g. random particle model), however the model is
formulated in the framework of continuum. The physical concept
behind the microplane model was discussed at the beginning of last
century by Mohr [46] and Taylor [47]. In this model, the material is
characterized by uni-axial relation between stress and strain
components on planes of various orientations. The microplane

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

58

Fig. 2. Schematic of split modied Hopkinson bar (MSHB).

model used in this work is based on the so called relaxed kinematic

constraint concept [48], which is a modication of the M2 microplane model proposed by Ba

zant and Prat [49]. In a number of
numerical applications it has been demonstrated that the model is
able to realistically predict concrete response for rather complex
stressestrain histories [48,50,51].
Each microplane is dened by its unit normal vector components ni (Fig. 3). The microplane strains are assumed to be the
projections of macroscopic strain tensor ij. On each microplane,
one normal (sN; N) and two shear stressestrain components (sL,
sK; L, K) are considered. To realistically model the concrete,
normal microplane stress and strain components are further
decomposed into volumetric and deviatoric parts (sN sV sD,
N V D). To prevent unrealistic model response for dominant
tensile load, for strong localization of strains the kinematic
constrain is relaxed [48]. Based on the microemacro work conjugancy and using pre-dened microplane constitutive laws, the
macroscopic stress tensor is calculated as an integral over all
possible microplane orientations (indicial notation):

sij sV dij

3
2p

Z 

sD ni nj 

dij

sL 
2

sM 
2

mi nj mj ni




li nj lj ni


(6)

dS

where S denotes the surface of the unit radius sphere and dij denotes Kronecker delta. The integration is performed by an optimal
Gaussian integration formula for a spherical surface with 21 integration points (symmetric part of the sphere, see Fig. 3, left). It has
been shown that this is an efcient formula that yields acceptable
accuracy [52]. To account for large strains and displacements,
GreeneLagrange nite strain tensor is used. To account for the
loading history of concrete, co-rotational stress tensor is employed.
Detailed discussion of the formulation of the microplane model can
be obtained from Refs. [48,53].
At the constitutive level the rate dependency consists of two
parts: (1) rate dependency related to the growing micro-cracks
(effect of inertia forces at the level of micro-crack tip), and (2)
rate dependency due to the viscosity of concrete (bulk material)
between the micro-cracks. In this model both effects are modeled

z
microplane

z
microplane
integrationpoint

T
M

in the integral sense based on the rate process theory, similar to the
concept proposed by Mihashi and Wittmann [13]. The inuence of
inertia forces on the rate effect is not a part of the constitutive law
and this effect is automatically accounted for through dynamic
analysis in which the constitutive law interacts with inertia forces.
The rate dependency for each microplane component is modeled
based on the rate process theory [53e55]. The rate dependency for
each microplane component reads [17,54]:

sM M s0M M

r

 
2g_
1
c
_
_ _ c 3
with g
1c2 ln
2 ij ij 1 scr
c1

where, c2 and c3 are material rate constants, which are calibrated by

tting of test data, scr is assumed crack spacing and
_ ij components of the macroscopic strain rate tensor.
The rate magnitude is measured on the macro-scale. Eq. (7)
applies to all microplane components except to volumetric
compression, which is assumed to be rate insensitive. It must be
kept in mind that Eq. (7) denes the inuence of rate of loading on
the true material strength only, i.e. without any consideration of
inertia. The inertial forces are automatically considered through
dynamic analysis. The model parameters for Eq. (7) are calibrated
based on the uni-axial compressive tests performed by Dilger et al.
[56], where the rate of applied loading was kept within the range
when inertial forces do not play a major role (less than 1/s). It may
be noted that though the material model is calibrated against
compression test data, a recent study on dynamic behavior of
concrete through compact tension specimen [40] has proven the
suitability of application of the model also for predominant tension
cases. The numerical analysis technique along with the constitutive
law as discussed above has also been widely validated against
several experiments [17,18,50,57,58]. For more detail see O

zbolt
et al. [17,18].
4. Role of damage and inertia in dynamics of quasi-brittle
materials
To illustrate the effect of inertia forces in case of dynamic
equilibrium of damaged concrete, Fig. 4 schematically shows concrete (dashed area) and linear elastic specimen loaded in uni-axial
tension. Fig. 4 essentially represents a free body diagram of the
portion cut by sections 1e1 and 2e2 in Fig. 1. In Fig. 4, section 1e1 is
considered as xed while section 2e2 is considered to move
(displacement control). It is assumed that the entire concrete
specimen undergoes softening (damage). Since the linear elastic

y
microplane
FEintegration
point

Fig. 3. Decomposition of the macroscopic strain vector into microplane strain components e normal (volumetric and deviatoric) and shear.

(7)

Fig. 4. State of equilibrium of forces when concrete softens.

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

specimen moves at a constant rate, no inertia forces are activated in

it, while due to softening (damage) in the concrete, inertia forces
get activated as shown in Fig. 4.
From the simplied equilibrium condition it follows:

ss As Mc u c st At or st

ss As Mc u c

(9)

At

This can be re-written as

st st;s st;I with st;s

ss As
At

st;I

c
Mc u
At

(10)

where Mc and c are equivalent mass and acceleration of the concrete softening zone, respectively. As a consequence of damage of
concrete the contact plane between concrete and elastic bar
(transmitter) moves faster than the section 2e2 of the elastic bar,
where the displacement is assumed to be applied. Consequently, if
the material gets damaged in a non-negligible part of the volume
(quasi-brittle material), the measured stress st in elastic part of the
set-up cannot be attributed only to material strength. It consists of
the material resistance st,s and the contribution of inertia st,I activated due to damage. Larger the damaged zone of the material the
tested specimen, higher will be the contribution of inertia. It must
be emphasized that this inertia is activated only due to the damage
of concrete and is different from the inertia activated in case of high
speed compression test which results in an equivalent connement. Therefore, in the case of elastic material or for quasi-static
loading, for constant strain rate applied in section 2e2 no acceleration (inertia) would be generated.
The above theoretical consideration is illustrated on a simple 3D
FE model. The model consists of two 3D eight-node solid nite
elements (see Fig. 5) and principally corresponds to the geometry

Fig. 5. Simple cohesive-elastic FE model.

59

shown in Fig. 4. The rst element is cohesive and represents

damaged concrete, which after reaching its specied tensile
strength undergo softening and the second element is considered
as linear elastic. To simulate the same boundary conditions, as
discussed above (section 1e1 xed and section 2e2 moving), the
nodes on the upper surface of concrete are xed in vertical direction. In other orthogonal directions, only minimum constraints
required for stable FE analysis are provided. The nodes on the lower
surface of linear elastic elements are pulled, uni-axially, under
displacement control. The displacement is applied at a constant
rate with respect to time (ramp loading) and the rate of application
of displacement controls the strain rate in the cohesive element.
The strain in the cohesive element was considered as the
displacement divided by the length of the element, i.e. the strain in
the linear elastic portion was ignored.
The properties of cohesive element (concrete) are taken as:
Youngs modulus Ec 30,000 MPa, Poissons ratio n 0.18, tensile
strength ft 3.5 MPa, uni-axial compressive strength fc 40 MPa,
concrete fracture energy GF 90 J/m2 and weigh density
rc 2400 kg/m3. It is assumed that the elastic element has elastic
properties and weight density the same as cohesive element. The
model is loaded by controlling displacement rate (velocity) of
nodes at the free surface of elastic element in axial direction (see
Fig. 5). For all loading cases the loading rate was kept constant
during the entire loading history (ramp loading). Measured are
both applied loading force and reaction. The analysis is performed
for three different specimen (element) sizes, a 25, 50 and
100 mm. In this simple model the size (volume) of the cohesive
zone is equal to the element size. As a constitutive law the above
discussed rate sensitive microplane model is employed.
The results of the analysis in terms of loadetime and reactione
time responses for relatively low strain rate of (static) 0.2/s and very
high strain rate of 200/s are shown in Fig. 6. The strain rates are
calculated as prescribed displacement rate over the length of the
cohesive element in the elastic response phase. In the static analysis the load and reaction must be the same. This should also be the
case in dynamic analysis with constant strain rate, but only if both
elements are linear elastic. However, the question is whether the
load and the reaction are the same in the case of dynamic analysis,
where concrete element after reaching tensile strength undergoes
softening. If they are not equal, the maximum stress in elastic
element is not true material strength it is apparent strength, which
consists of the contribution of concrete strength and the contribution of stress due to inertia.
As can be inferred from Fig. 6, for relatively low strain rate, the
load and reaction are almost the same however, for very high strain
rate the reaction is much smaller than the load. This implies that
because of damage and softening of the cohesive material the
inertia is activated in loading direction. The forces generate in
elastic element stresses that are much higher than the stresses in
cohesive element. Consequently, the stress measured in the elastic
element cannot be true stress, i.e. there is signicant contribution
of inertia forces due to the softening of cohesive element. Therefore
strength (maximum stress) measured in elastic element is apparent
strength, which consists of true material strength and contribution
of inertia forces, which are generated as a consequence of damage
of concrete.
Due to the fact that inertia is related to the size of the damage
zone (FPZ) and the level of damage before localization of crack, it
can be expected that for brittle materials (e.g. glass) the inuence of
inertia on the apparent strength is relatively small. To investigate
this effect the above analysis is repeated assuming extremely brittle
behavior of cohesive element. Fracture energy of cohesive material
is taken as GF 0.9 J/m2. All other properties are taken the same as
before. In Fig. 7 are again plotted reactione and loadetime

60

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

Fig. 6. Typical reactione and loadetime response (a 50 mm): (a) Strain rate 0.2/s and (b) Strain rate 200/s.

response (a 50 mm) for strain rate of 200/s (a) for cohesive and
(b) for brittle material. It can be seen that for brittle material, even
at very high strain rates, the difference between the peak reaction
(12.2 kN) and peak load (13.5 kN) is not signicant compared to the
difference between the peak load and peak reaction obtained for
the cohesive material. This conrms that for brittle material the
strength measured in the elastic element can be considered as
approximately the true material strength.
In the case of relatively low strain rate the initial response in
both cases (Fig. 7(a) and (b)) should be the same. However, if the
strain rate is high the difference is due to the fact that before the
loading wave reaches supports, a portion of the cohesive element
already undergoes softening. The softening of the material generates inertia forces that contribute to the loading part but not to the
reaction part. In the case of brittle material, there is no softening
and consequently no change of strain rate over the specimen
length, i.e. no acceleration. The strain rate is approximately constant over the entire specimen and there is no inertia with the
result that the reaction is approximately the same as the loading.
Similar to the brittleness of the material, the size of damage
zone should also have an effect on inertia forces. For instance,
constant fracture energy of cohesive element but large softening
volume (larger element) should lead to the stronger inuence of
inertia on apparent strength. This also shows the experiments, e.g.
assuming the same evaluation procedure of experimental results,
dynamic increase factor (DIF) of low quality concrete is larger than
that of high quality concrete [59,60]. The same is the case with mass
density, i.e. larger mass density would generally cause larger inertia
forces assuming the same fracture energy.

In Fig. 8 are plotted calculated DIF for true (reaction) and

apparent (load) strength for three different element sizes (a 25,
50, 100 mm) and for different strain rates. In all three cases the
material properties including quasi-static fracture energy are
considered the same with the values as those mentioned earlier. It
turns out that the true and apparent strengths are approximately
the same for the strain rates up to approximately 2/s. However, for
higher strain rates the true strength increases approximately linear
in semi-log scale (constitutive law) whereas the apparent strength
exhibit progressive increase. As expected, the increase is stronger if
the element is larger. In contrary to this, the true strength shows
almost no sensitivity on the size of the element. Note that the true
strength is in the here used constitutive law (microplane model)
controlled by the energy activation theory. It is very interesting to
note that Cusatis [36] reported an almost identical plot of DIF v/s
strain rate for concrete specimen loaded in tension.
Fig. 9 presents numerical results for apparent strength
compared with the strength measured in experiments [61]. All test
results are evaluated assuming elastic concrete response before
localization of crack, i.e. no damage is accounted for. The experimental data represents measurements on specimens of different
sizes and shapes, different concrete qualities and different content
of water. Therefore, it is not surprising that there are relative large
scatter of the test data with respect to the measured strain rate.
However, it is surprising that in spite of very simple numerical
model, the numerical results for apparent strength nicely t the
band of experimental results. The comparison suggests that the
experimental results correspond to apparent strength rather than
true tensile strength of concrete.

Fig. 7. Typical reactione and loadetime response (a 50 mm, strain rate 200/s) for: (a) cohesive element and (b) brittle element.

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

61

5. Indirect tensile test on MSHB: numerical analysis vs.

experiment

Fig. 8. Calculated DIF for true (reaction) and apparent (load) strength.

Apart from the inertia due to damage and softening, in dynamic

fracture of concrete there are also other important effects related to
inertia. For instance, in uni-axial compression tests lateral inertia
(connement) is probably generated before the specimen undergo
softening [18]. This contribution of inertia to dynamic compressive
strength of concrete is generally well acknowledged [26e35].
Furthermore, when crack propagates, at the crack tip inertia forces
are activated, which tends to prevent crack opening, i.e. with
increasing crack speed stress intensity factor is decreasing. The
consequence of this is crack branching [18]. Moreover, it was
demonstrated experimentally [2] and numerically [58,62] that
inertia is also responsible for the change of failure mode, i.e. with
increasing loading rate mode-I failure tends to switch to mixedmode or shear type of failure. This has been proven by theoretical
studies as well [10,11]. Finally, the micro-crack inertia is acknowledged as one of the mechanisms responsible for the strain-rate
sensitivity of the tensile strength of concrete-like materials [40].

To investigate tensile strength and fracture energy of concrete

for strain rates in the range between 10 and 200/s, in the past a
number of experiments were carried out using MSHB test set-up.
The evaluation of experimental results, which are based on the
loading wave propagation through elastic continuum, showed that
for the strain rates higher than approximately 10/s there is a progressive increase of DIF for tensile strength and fracture energy. To
bring more light into the tensile behavior of concrete loaded under
high strain rates, the experiments performed by Schuler et al. [41]
are numerically simulated using the above discussed 3D FE code
based on the rate dependent microplane model. The geometry of
the test set-up is shown in Fig. 10. The total length of the incident
bar (aluminum) is 5.50 m with the concrete specimen of 250 mm
length that was glued at the end of the incident bar. The crosssection diameter of the incident bar and concrete specimen is
74.20 mm. The incident bar was loaded by the impact of steel
hammer with mass of 1.95 kg. Three impact velocities were
applied: 4.10, 7.60 and 11.10 m/s (cases: A, B, and C). According to
the experimental results, the average strain rates in the concrete
specimen were 32, 68 and 80 1/s, respectively.
In the numerical simulation the MSHB test set-up is discretized
with 3D four node standard nite elements (constant strain). Discretized are projectile (steel), incident bar (aluminum) and concrete specimen. In the numerical model the loading impulse is
generated by the impact velocity of the projectile on the incident
bar. The concrete specimen is modeled by the rate sensitive
microplane model. The projectile (steel) and incident bar
(aluminium) are assumed to be linear elastic with the following
properties: Youngs modulus Es 210 GPa and Ea 70 GPa, Poissons ratio ns 0.33 and na 0.30, respectively. The concrete material properties for quasi-static loading are the same as in the
experiments: Youngs modulus Ec 29.75 GPa, Poissons ratio
nc 0.18, uni-axial compressive strength fc 43.2 MPa, tensile
strength ft 2.92 MPa, fracture energy GF 107 J/m2 and mass
density rc 2400 kg/m3.

Fig. 9. Comparison of numerically and experimentally observed apparent strength.

62

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

Fig. 10. Geometry of the MSHB test set-up (Schuler et al., 2006), all dimensions in [mm], and FE discretization of concrete specimen, incident bar and projectile.

As discussed before, the strain rate effect is accounted for

through the microplane constitutive law and it is based on the rate
process theory. According to the theory, the material strength increases approximately linear in semi-log scale [55]. Typical tensile
stressestrain relationships for quasi-static and rate sensitive
constitutive law are plotted in Fig. 11. The curves correspond to the
unit element size and are obtained by loading single 8-node nite
element in uni-axial tension or compression, respectively. Note that
in the analysis, the crack band approach is employed [45]. In Table 1,
the rate dependent concrete properties for the strain rate range up
to 1000/s are summarized. The values for fracture energy are obtained by multiplying the area under the stressestrain curves
(Fig. 11) with the element size, which was in the FE model for
concrete specimen constant h 5 mm. The secant Youngs modulus
is calculated from the uni-axial compressive stressestrain curves
for the compressive strain of 0.0002.
As mentioned earlier, in this work, the experiments performed
by Schuler et al. [41] were numerically simulated. The schematic of
the experiment is shown in Fig. 10. Exactly the same test was
simulated numerically by modeling the incident bar and the
specimen and impacting with three different velocities. The
experimentally measured and evaluated data (average values) are
summarized in Table 2 [41].
The experimentally reported and corresponding numerically
predicted failure patterns of concrete specimen are shown in
Fig. 12. It can be observed that the predicted failure patterns are

very similar to those experimentally obtained. It shall be noted that

with the increase in the loading rate, the number of cracks increases both in experiment and analysis. Same as in the experiments, for the highest loading rate even crack branching was
observed numerically.
To calculate concrete tensile strength and fracture energy, in the
experiments the strains at the end of the incident bar and pull-back
velocity at the end of the concrete specimen were measured. The
same was therefore evaluated numerically as well. The experimental and numerical results for the impact velocity of 4.10 m/s
(case A) are presented in Fig. 13. The comparison shows remarkably
good agreement, i.e. the evolution of axial strain (surface of the
specimen) and velocity (center of the specimen) are almost the
same as measured in the experiment.
This comparison between experimental and numerical results
shows that the 3D FE analysis is able to realistically and quite
accurately replicate experiments of Schuler et al [41]. Therefore, to
better understand the inuence of strain rate on tensile strength
and fracture energy, in the following, the results of the analysis are
evaluated in two different ways: (i) the evaluation based on the
direct measurement (monitoring) of stress, strain and fracture energy in the nite elements and (ii) the same as evaluated by Schuler
et al [41], which is based on the 1D propagation of loading waves
through concrete, with and without accounting for damage of
concrete.
5.1. Rate dependent tensile strength
The typical evolution of axial stresses and strains, with time, in
cracked (FE in the crack) and un-cracked part (FE between free
surface and crack) of the concrete sections (case A) are plotted in
Fig. 14. The corresponding axial stressestrain histories in the
cracked section are shown in Fig. 15. The plotted data are measured
in the center of the specimen. Note that since the analysis is threedimensional, strains and stresses vary over the cross-section, i.e.

Table 1
Concrete properties from the rate dependent constitutive law.

Fig. 11. Rate dependent tensile stressestrain response e constitutive law.

Strain rate
[1/s]

Youngs modulus Compressive

Tensile
Fracture
[GPa]
strength [MPa] strength [MPa] energy [J/m2]

Quasi-static
1
10
100
1000

29.75
34.35
35.15
35.95
36.70

43.2
51.1
52.5
53.9
55.4

2.92
4.20
4.43
4.67
4.90

107
178
192
206
221

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

63

Table 2
Experimentally measured and evaluated data assuming elastic behavior of concrete
before crack localization [41].
Case Pull-back
Strain
Youngs
Tensile
Fracture
velocity [m/s] rate [1/s] modulus [GPa] strength [MPa] energy [J/m2]
A
B
C

2.63
3.36
3.67

33
68
80

40.1
38.7
39.9

12.9
16.2
18.0

288.5
334.5
e

the specimen cannot be exactly under uni-axial stressestrain state,

as assumed in the evaluation of experimental results. In Fig. 15(b),
the recorded stressestrain curve is compared with the constitutive
law. It can be observed that the recorded response for strain rate of
34/s (see Table 4, case A) shows lower tensile strength and fracture
energy than the corresponding uni-axial constitutive law. This can
be attributed to the three-dimensional stressestrain state. Namely,
the considered element is in tension in lateral direction also.
From the evaluation of axial stresses and strains in the uncracked part of the specimen, relatively high level of tensile
strains are observed before crack localization (damage, see Figs. 14
and 15). Moreover, the analysis shows that before localization of
crack the entire concrete specimen exhibit stress level that is very
close to tensile strength and some parts even undergo softening.
The typical evolution of stresses and strains for other two cases (B
and C) is principally the same and therefore not shown.
The typical record of acceleration and displacement in the uncracked part of the specimen is shown in Fig. 16(a). It is interesting to observe that the maximum acceleration is recorded
after localization of crack. The generated inertia forces are partly
in self equilibrium (kinetic energy), however, their unbalanced
portion can cause distributed damage in concrete, which could
be nicely seen in experiments of Schuler et al [41] (see Fig. 16(b)).
However, it must be noted that for heterogeneous material such
as concrete, material strength is not constant and can also result
in multiple cracking. Nevertheless, multiple fractures zones/
macro cracks can develop supported by the mentioned inertia
effect.
After localization of cracks in the critical sections the unlocalized parts of the specimen are unloaded. As mentioned
before, in these (unloaded) parts signicant level of damage is
observed, i.e. it turns out that concrete is far from being elastic. This
is shown in Table 3 where are summarized maximum strains and
stresses reached in the part of the specimen between free concrete
surface and the rst crack, measured at the surface and in the
center of the cross-section.

Fig. 13. Experimentally measured (top) and numerically predicted (bottom) strains
and particle velocities.

The level of damage is estimated from the corresponding

stressestrain curves shown in Fig. 15. Damage parameter (1  u) is
calculated by subtracting the ratio between tensile secant modulus
and initial modulus (u) from unity. The evaluation shows that in
un-cracked part of the concrete specimen, the level of damage is
very high, approximately equal to 0.70. The reason for this is
probably due to damage induced acceleration and related inertia
forces. According to the evaluation of numerical results (Fig. 16(a)),
the acceleration is in the order of 105 m/s2. The same order of acceleration was recently estimated from experiments [37,63].

Fig. 12. Crack patterns for three impact velocities, left experiment and right numerical analysis with fragmentation (top 4.1 m/s, middle 7.6 m/s and bottom 11.1 m/s).

64

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

Fig. 14. Time evolution of stress (a) and strain (b) in the nite element (center of the cross-section).

In Table 4 are summarized numerically predicted data in the

crack that is the closest to the free specimen surface. Shown are
numerically predicted pull-back velocities, strain rates of the
cracked section at the specimen surface, the corresponding rate
sensitive tensile strength (constitutive law) and numerically predicted (nite element) and calculated tensile strength using Eq. (3).
It should be noted that predicted data varies over the cross-section
as well as over the length of the specimen. The strain rate represents the average value in the cracked section of the specimen
measured before localization (see Fig. 14(b)) of crack. The numerically predicted pull-back velocity is for all three impact velocities
similar. This suggests that for the investigated strain rate range
(34e83/s) tensile strength does not change signicantly. According
to the constitutive law, for the three strain rates the tensile strength
varies in the range from 4.56 to 4.67 MPa, i.e. there is approximately
linear increase in semi-log scale but for the considered range of the
strain rate obviously not signicant. The predicted tensile strength
measured in the nite elements, in the center and at the surface of
the cracked section, respectively, shows in principle, the same
tendency. The average tensile strength is close to the strength
predicted by the constitutive law. It varies over the section due to
the fact that the strain rate and the state of stressestrain state is not
exactly uni-axial. Moreover, the strain rates shown in Table 4 are
calculated as an average value over a multiple time steps and do not
exactly corresponds to predicted strengths. Note that the fact that
the overall response predicted by the nite element analysis (crack
pattern and pull-back velocity, see Figs. 12 and 13) shows very good
agreement with the experimental measurements, implicitly

conrms that rate dependent tensile strength and fracture energy,

as discussed below, should follow rate dependent constitutive law.
The constitutive law gives linear and not progressive increase of
concrete strength, as not correctly evaluated from the experimental
data.
Further, in Table 4 is also shown tensile strength calculated using wave Eq. (3). Longitudinal wave velocity c0 is once calculated
accounting for damage of concrete before localization of crack and
once assuming no damage (elasticity). Damage is accounted for by
setting numerically predicted secant stiffness from Table 3 into Eq.
(3). Assuming elasticity, c0 is calculated using Youngs modulus
from Table 1. It can be seen that the wave equation results in similar
prediction as predicted by numerical analysis and by the constitutive law (true strength). However, when assuming elastic
behavior of concrete in evaluation of numerical results, tensile
strength becomes signicantly overestimated (apparent strength),
i.e. much higher than that predicted numerically. These values are
in the same range as evaluated by Schuler et al [41]. (see Table 2).
Obviously, when assuming elastic behavior of concrete before
fracture the predicted tensile strength increases progressively with
the increase of strain rate and does not follow the constitutive law.
The different evaluation of normalized tensile strength (Dynamic increase factor) is also summarized in Fig. 17. It can be
observed that the numerical prediction and evaluation based on
the wave equation, which accounts for damage of concrete, follow
the rate dependent constitutive law. On the contrary the prediction
based on the wave equation and elasticity signicantly overestimates the true material strength. The comparison with the

Fig. 15. Stressestrain relationships (a) in the nite element (center of the cross-section) and (b) comparison between stressestrain relationship in the crack and constitutive law.

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

Fig. 16. Records of acceleration and displacement in the un-cracked part of the concrete specimen (a) and distributed damage observed in the experiments of Schuler
et al. [41].

experimental data from the literature (Fig. 17) also shows that the
experimental results, that are all evaluated using wave equation
and elasticity, t the numerical prediction only when the results are
evaluated in the same manner, i.e. assuming elastic behavior of
concrete.
Traditionally, the evaluation of experimental results for concrete
and other quasi-brittle materials exposed to high loading rates
based on elasticity is widely accepted [21,25,60]. However, the
presented numerical results indicate that the true tensile strength
approximately follows the rate dependent constitutive law, while
the inertial forces are responsible for the apparent sudden and
progressive rise in the tensile strength beyond a certain strain rate.
This suggests that the progressive increase of tensile strength at
higher strain rates, suggested by many researches and design codes,
might need a more careful and better interpretation that can
consider the inuence of inertial forces as well as the damage in the
specimen. It is understandable that such an evaluation is not as
simple for experimental results as it for numerical ones. Nevertheless, it is highly important, since, the evaluation based on

Table 3
Numerically predicted concrete properties in the un-cracked part of the specimen.
Case

A
B
C

Max. stress
[MPa]

Max.
strain [%]

Secant modulus,
Esec [MPa]

Damage
(1  Esec/E0)

Center

Surface

Center

Surface

Center

Surface

Center

Surface

4.31
4.45
4.02

3.60
4.16
3.61

0.114
0.176
0.109

0.046
0.076
0.057

3781
2528
3688

7826
5474
6333

0.87
0.92
0.88

0.74
0.82
0.79

65

assumed elastic material behavior totally ignores relatively high

level of concrete damage before cracking.
Consequent to such evaluation and interpretation of the
experimental results, in many commercial nite element codes the
progressive increase of material strength is a part of the constitutive law. This is not very convincing from the physical point of view.
As discussed by O

zbolt et al. [18] and O

zbolt and Sharma [57], the
progressive increase of structural response observed at the macroscale (e.g. compact tension test) is a consequence of inertia forces
and not a consequence of rate dependent strength of the material.
Therefore, the progressive increase in the tensile strength shall not
be a part of the constitutive law for concrete, rather, it should be
predicted automatically from the dynamic analysis.
As discussed by O

zbolt et al. [17,18] and as demonstrated in
simple example in previous section, low quality concrete exhibit
more damage before localization of cracks. Consequently, for low
quality concrete the discrepancy between the true strength (rate
dependent constitutive law) and the strength predicted employing
elasticity wave theory (apparent strength) increases. On the contrary, for brittle material the apparent and true strength should be
the same since for brittle material there is almost no damage before
localization of crack. To prove this, the above numerical analysis for
MSHB is further carried out assuming elasticebrittle material (for
case A). The material properties are the same as in the above
analysis except that the fracture energy is set to GF 2.0 J/m2. The
results are evaluated in the same manner as before and are also
shown in Table 4. As can be seen, the predicted tensile strength and
strength calculated using wave equation and elasticity give
approximately the strength as predicted by the rate dependent
constitutive law. This conrms that wave elasticity theory can be
used only if before cracking there is no signicant damage. Moreover, recent experimental results for glass (brittle material) also
conrm that there is no progressive increase of DIF for tensile
strength at high strain rates [64].
As already pointed out, before localization of discrete cracks
tensile maximum principal stress in entire concrete specimen is
close to tensile strength. Because of this fact it is important to note
that the constitutive law for concrete should account for the concrete nonlinearity in the pre-peak tensile response.
5.2. Rate dependent fracture energy
Similar to the rate dependent tensile strength, it is still not clear
how the strain rate inuences fracture energy. As discussed in
Section 3.2, in the used constitutive law (microplane model) rate
sensitivity is accounted for based on the rate process theory (Eq.
(7)). With increasing strain rate tensile strength and stiffness increase linear in semi-log scale (see Fig. 11 and Table 1). In the model
there is no explicit relation between fracture energy and strain rate,
however, similar to tensile strength, the resulting fracture energy
(area under the stressestrain curves from Fig. 11 multiplied by the
element size) increases approximately linear in semi-log scale with
strain rate. Correct estimation of fracture energy of concrete from
experiments with high strain rates is extremely difcult [37,63].
The main problem is that in such experiments multiple cracks form
and many of them are not visible since the crack pattern over the
specimen depth is not exactly the same as at the surface of the
specimen. Moreover, concrete before localization of one or more
macro-cracks is highly damaged. Therefore it is not surprising that
with this respect there are still contradictions in the literature. The
evaluation of experimental results of Schuler et al. [41] (see Table 2)
suggested that for higher loading rates there is a progressive increase of fracture energy.
The evaluation of fracture energy from numerical results can be
performed either based on the response of a single cracked

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

66
Table 4
Numerically predicted data and their evaluation.
Case

Pull-back vel. [m/s]

A
2.81
B
2.81
C
2.85
Brittle material:
A
0.80

Strain rate [1/s]

ft const. law [MPa]

ft nite elements [MPa]

ft from (3), damaged [MPa]

Surface

Surface

Center

Surface

Center

Surface

34
52
83

4.56
4.64
4.67

3.80
4.27
3.59

4.36
4.46
4.30

4.23
3.46
4.24

6.09
5.09
5.56

11.87
11.87
12.04

34

4.56

4.63

4.84

3.4

4.45

3.38

element, as done here, or evaluating average fracture energy over

the surface of cracked elements. From measuring fracture energies
on several cracked elements, it was concluded that the average
fracture energy is not much different than the energy measured
element wise. Therefore, the fracture energy for three loading cases
(A, B and C) is evaluated for two nite elements of the macro-crack,
one element in the center of the cracked surface (center of the
specimen) and one on the surface of the specimen. Fracture energy
is calculated as the integral under the stressestrain curve (Fig. 15)
and multiplied by the crack band (element size), following the
concept of smeared cracks and crack band theory. In Table 5, the
results for the cracked section, at the center and on the specimen
surface are shown. Similar to tensile strength, fracture energy is not
uniformly distributed over the cracked cross-section. In Fig. 18,
experimental results and results of numerical analysis are
compared. The numerical results show that for the investigated
range of strain rate, DIF of fracture energy is slightly larger than one.
It is interesting to note that it is even lower than DIF predicted by
constitutive law. The reason for such seemingly anomalous
behavior cannot be given with full certainty, but it seems that
possibly there is some inuence of the 3D rate dependent structural
response such as lateral stresses. Due to these, in spite of higher
value for the fracture energy at the constitutive level (uni-axial), a
lower value of fracture energy is obtained at the structural level.
This means that with increasing strain rate the material response
becomes more brittle since strength increase is stronger than
fracture energy.

ft from (3),
elastic [MPa]

The presented values for fracture energy multiplied by the

cross-section of the specimen and number of cracks gives
approximately the total fracture energy consumed by the open
macro-cracks. However, total fracture energy consumed by the
specimen is larger because not only macro-cracks consume energy,
but also a number of smaller macro- and micro-cracks, which after
localization of macro-cracks get closed. This applies not only to unnotched specimens but also to specimens with notch. If one would
account for this part of consumed energy and calculate the average
specic energy per macro-crack than would this apparent specic
fracture energy exhibit progressive increase with strain rate,
similar as show the experimental results. However, the same as in
the case of tensile strength, this would not be the true specic
fracture energy.
The evaluation of most experimental results shows that up to
the strain rate of 1/s, DIF approximately increases linearly in semilog scale and for higher strain rates there is a progressive increase.
The reason for this can probably be attributed to the fact that the
total crack surface at high strain rates cannot be correctly measured
due to the rather complex crack pattern, i.e. the total crack surface
is most probably much larger then assumed in the evaluation of
experimental results. As discussed above, the un-cracked part of
the specimen also exhibits signicant damage (micro-cracking),
which also contributes to the overall fracture energy. Moreover, in
the crack zone there are local inertia effects that contribute to the
estimated fracture energy. Filtering out of these effects would
probably give similar results as predicted numerically. Again, such

Fig. 17. Different evaluations of DIF for tensile strength (experimental and numerical results) compared with experimental data from the literature evaluated assuming elasticity.

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

Table 5
Numerically evaluated fracture energy.
Case

Quasi-static
A
B
C

Strain
rate [1/s]

GF [J/m2]
Center

Surface

Center

Surface

e
34
52
83

107.00
84.10
124.12
154.97

107
116.95
110.32
104.43

1.0
0.78
1.16
1.44

1.0
1.09
1.03
0.98

DIF (GF)

an evaluation using the experiments, may not be possible with the

current state of the art and may be better instrumentation may
provide with the missing information in future.

3.

6. Summary and conclusions

The tensile behavior of concrete under high loading rates is
numerically investigated and the results are discussed. The analysis
is performed by employing 3D FE code for multi body dynamics
with fragmentation capability. As a constitutive law the rate sensitive microplane model is used. Based on the numerical results and
their comparison with experimental results the following can be
concluded.
1. It is shown that for quasi-brittle materials, such as concrete,
loaded by relatively high loading rates the resistance (apparent
strength) consists of two contributions. The rst contribution is
the true material strength, which is approximately a linear
function of strain rate (semi-log scale) and is controlled by the
rate dependent constitutive law. The second contribution is a
consequence of damage of concrete. Up to the strain rate of
approximately 10/s the apparent strength is nearly the same as
the true material strength. However, for higher strain rates there
is a progressive increase of apparent strength. Depending on the
test set-up and measurement technique, the progressive increase can be explained either by inertia invoked by material
softening or by the fact that in the evaluation of numerical or
experimentally measured data, damage of concrete is ignored.
2. The evaluation of numerical results of MSHB test shows that
before localization of cracks concrete specimen exhibit signicant level of damage. The general assumption that the concrete
before cracking is linear elastic, made in the evaluation of

4.

5.

6.

7.

67

experimental results, is not very convincing physically. This

seems to be the main reason that the experimentally estimated
apparent strength is much larger than the true material
strength. Due to a large database of tests evaluated based on this
assumption, the apparent strength is in the current engineering
practice and design codes considered as the true material
strength. This needs to be reviewed, since the sudden progressive increase in apparent strength is due to the inuence of
inertia and consequently dependent on the size and type of
structure on hand.
Quasi-brittle materials with relatively large FPZ (e.g. low quality
concrete) exhibit higher apparent strength than materials with
smaller FPZ (high strength concrete). Moreover, larger the
specimen size larger is the apparent strength. It is shown that
for brittle material, with small size of the FPZ (e.g. glass), the
apparent and true strength are nearly the same.
In the MSHB test, the number of tensile cracks depends on the
loading rate. For relatively low loading rate only one tensile
crack is generated, however, with increase of loading rate
number of tensile cracks increases and damage is distributed
over the large volume of the specimen.
Similar to the true tensile strength, numerical results show that
concrete fracture energy increases approximately linearly
(semi-log scale) with the increase of strain rate. It approximately
follows the fracture energy that is the outcome of the constitutive law. However, evaluation of experimental results show
that for higher loading rates the fracture energy increases progressively with the increase of strain rate, similar to the
apparent strength. The reason for this is probably due to the fact
that in experiments there are a number of discrete macro- and
micro-cracks whose total surface area is impossible to accurately measure. Inevitably, the total surface area is underestimated with the consequence that the evaluated fracture
energy is overestimated.
Numerical results suggest that the results of any indirect tension
test, such as split Hopkinson bar test, need a more careful
interpretation, i.e. due to the fact that in the concrete specimen
is damaged, and not elastic, the results of measurement need to
be corrected.
True material strength should be accounted by the rate dependent constitutive law and the inuence of inertia should automatically come from dynamic structural analysis. Progressive
increase of apparent strength is always related to the change of
the failure mode or multiple cracking and branching phenomena. Therefore, the apparent strength should not be considered
as a part of the constitutive law.

Acknowledgements
The authors would like to express their sincere gratitude towards Dr. Jaap Weerheijm, Delft University of Technology for
providing the experimental data used in Fig. 9 of this paper.
References

Fig. 18. Different evaluations of DIF for fracture energy (experimental and numerical
results) compared with experimental data from the literature evaluated assuming
elasticity.

[1] Reinhardt HW. Concrete under impact loading, tensile strength and bond.
Heron 1982;27(3).
[2] Banthia NP, Mindess S, Bentur A. Impact behaviour of concrete beams. Mater
Struct Matriaux et Constr 1987;20:293e302.
[3] Brandon DG. Dynamic loading and fracture. In: Blazynski TZ, editor. Materials
at high strain rates. London: Elsevier; 1987. pp. 187e218.
[4] Curbach M. Festigkeitssteigerung von Beton bei hohen Belastungs-geschwindigkeiten. Germany: Karlsruhe University; 1987 [Ph.D. thesis], 154 pages.
[5] Bischoff P, Perry S. Compressive behaviour of concrete at high strain rates.
Mater Struct Matriaux et Constr 1991;24:425e50.
[6] Reinhardt HW, Krmeling HA, Zielinski AJ. The split Hopkinson bar, a versatile
tool for the impact testing of concrete. Mater Struct 1985;19(109):55e63.

68

J. O
zbolt et al. / International Journal of Impact Engineering 69 (2014) 55e68

[7] Reinhardt HW, Weerheijm J. Tensile fracture of concrete at high loading rates
taking account of inertia and crack velocity effects. Int J Fract 1991;51:31.
[8] Schuler H. Experimentelle und numerische Untersuchungen zur Schdigung
von stobeanspruchtem Beton [Dissertation]. Germany: Universitt der
Bundeswehr Mnchen; 2004. 184 Pages.
[9] Mechteherine V, Millon O, Butler M, Thoma K. Mechanical behavior of strain
hardening cement-based composite under impact loading. Cem Conc Compos
2011;22:1e11.
[10] Freund LB. Crack propagation in an elastic solid subjected to general loading-I.
Constant rate of extension. J Mech Phys Solids 1972;20:129e40.
[11] Freund LB. Crack propagation in an elastic solid subjected to general loadingII. Non-uniform rate of extension. J Mech Phys Solids 1972;20:141e52.
[12] Kipp ME, Grady DE, Chen EP. Strain-rate dependent fracture initiation intern.
J Fract 1980;16:471e8.
[13] Mihashi H, Wittmann FH. Stochastic approach to study the inuence of rate of
loading on strength of concrete. Heron 1980;25(3).
[14] Holmquist T, Johnson G, Cook W. A computational constitutive model for
concrete subjected to large strain, high strain rates and high pressures. In:
14th Int. sym on ballistics, Quebec. Canada: American defense preparedness
association; 1993. pp. 591e600.
[15] Lu Y, Xu K. Modelling of dynamic behaviour of concrete materials under blast
loading. Int J Solid Struct 2004;41:131e43.
[16] O

zbolt J, Reinhardt HW. Rate dependent fracture of notched plain concrete
beams. In: Pijaudier-Cabot, Gerard, Acker, editors. Proceedings of the 7th international conference CONCREEP-7. Wiley, France: Ecole Centrale de Nantes;
2005. pp. 57e62.
[17] O

zbolt J, Rah KK, Mestrovi
c D. Inuence of loading rate on concrete cone
failure. Int J Fract 2006;139:239e52.
[18] O

zbolt J, Sharma A, Reinhardt HW. Dynamic fracture of concrete e compact
tension specimen. Int J Solid Struct 2011;48:1534e43.
[19] Ortlepp S. Zur Beurteilung der Festigkeitssteigerung von hochfestem Beton
unter hohen Dehnungsgeschwindigkeiten [Dissertation]. Technische Universitt Dresden; 2006. 182 pages.
[20] Pedersen RR, Simone A, Sluys LJ. Continousediscontinous modelling of dynamic failure of concrete using a viscoelastic viscoplastic damage model. In:
Motasoares CA, Martins JAC, Rodrigues HC, Ambrsio JAC, Pina CAB,
Motasoares CM, et al., editors. III European conference on computational
mechanics. Lisbon, Portugal: Springer; 2006. p. 370.
[21] Pedersen RR. Computational modelling of dynamic failure of cementitious
materials [Dissertation]. the Netherlands: TU Delft; 2009. 164 Pages.
[22] Larcher M. Development of discrete cracks in concrete loaded by shock waves.
Int J Impact Eng 2009;36:700e10.
[23] Reinhardt HW, O

zbolt J, Trava
s V. Response of concrete members to impact
loading, dynamic fracture of concrete e compact tension specimen. In: Oh BH,
Choi OC, Chung L, editors. FraMCoS-7, Jeju, Korea, May 23e28. Korea Concrete
Institute; 2010. pp. 1701e14.
[24] Malvar LJ, Ross CA. Review of strain rate effects for concrete in tension. ACI
Mater J 1998;95(6):735e9.
[25] Weerheijm J. Concrete under impact tensile loading and lateral compression
[Dissertation]. the Netherlands: TU Delft; 1992. 157 pages.
[26] Wu H, Zhang Q, Huang F, Jin Q. Experimental and numerical investigation on
the dynamic tensile strength of concrete. Int J Impact Eng 2005;32:605e17.
[27] Li QM, Meng H. About the dynamic strength enhancement of concrete-like
materials in a split Hopkinson pressure bar test. Int J Solids Struct 2003;40:
343e60.
[28] Kim DJ, Sirijaroonchai K, El-Tawil SE, Naaman A. Numerical simulation of the
split Hopkinson pressure bar test technique for concrete under compression.
Int J Impact Eng 2010;37:141e9.
[29] Li QM, Lu YB, Meng H. Further investigation on the dynamic compressive
strength enhancement of concrete-like materials based on split Hopkinson
pressure bar tests, part II: numerical simulations. Int J Impact Eng 2009;36:
1335e45.
[30] Zhang M, Wu HJ, Li QM, Huang FL. Further investigation on the dynamic
compressive strength enhancement of concrete-like materials based on split
Hopkinson pressure bar tests part I: experiments. Int J Impact Eng 2009;36:
1327e34.
[31] Zhou XQ, Hao H. Modelling of compressive behaviour of concrete-like materials at high strain rate. Int J Solids Struct 2008;45:4648e61.
[32] Cotsovos DM, Pavlovic MN. Numerical investigation of concrete subjected
to compressive impact loading. Part 1: a fundamental explanation for the
apparent strength gain at high loading rates. Comput Struct 2008;86:
145e63.
[33] Georgin JF, Reynouard JM. Modeling of structures subjected to impact: concrete behavior under high strain rate. Cem Conc Compos 2003;25:131e43.

[34] Grote DL, Park SW, Zhou M. Dynamic behavior of concrete at high strain rates
and pressures: I. experimental characterization. Int J Impact Eng 2001;25:
869e86.
[35] Rossi P. Strain rate effects in concrete structures: the LCPC experience. Mater
Struct 1997;00:54e62.
[36] Cusatis G. Strain-rate effects on concrete behavior. Int J Impact Eng 2011;38:
162e70.
[37] Weerheijm J, Van Doormaal JCAM. Tensile failure of concrete at high loading
rates: new test data on strength and fracture energy from instrumented
spalling tests. Int J Impact Eng 2007;34:609e26.
[38] Cotsovos DM, Pavlovic MN. Numerical investigation of concrete subjected to
high rates of uniaxial tensile loading. Int J Impact Eng 2008;35:319e35.
[39] Lu YB, Li QM. About the dynamic uniaxial tensile strength of concrete-like
materials. Int J Impact Eng 2011;38:171e80.
[40] O

zbolt J, Bo
snjak J, Sola E. Dynamic fracture of concrete compact tension
specimen: experimental and numerical study. Int J Solids Struct 2013;50:
4270e8.
[41] Schuler H, Mayrhofer C, Thoma K. Spall experiments for the measurement of
the tensile strength and fracture energy of concrete at high strain rates. Int J
Impact Eng 2006;32:1635e50.
[42] Hopkinson B. A method of measuring the pressure produced in the detonation
of high explosives or by the impact of bullets, Phil. Trans Roy Soc Lond Ser A
1914;213(10):437e56.
[43] Kolsky H. Stress waves in solids. Oxford: Clarendon Press; 1953. 212 pages.
[44] Belytschko T, Liu WK, Moran M. Nonlinear nite elements for continua and
structures. John Wiley & Sons Ltd; 2001. 672 pages.
[45] Ba

zant ZP, Oh BH. Crack band theory for fracture of concrete. RILEM
1983;93(16):155e77.
[46] Mohr O. Welche Umstnde bedingen die Elastizittsgrenze und den
Bruch eines Materiales. Z des Vereins Dtsch Ingenieure 1900;46:1524e30.
1572e1577.
[47] Taylor GI. Plastic strain in metals. J Inst Met 1938;62:307e24.
[48] O

zbolt J, Li Y, Ko

zar I. Microplane model for concrete with relaxed kinematic
constraint. Int J Solid Struct 2001;38:2683e711.
[49] Ba

zant ZP, Prat PC. Microplane model for brittle-plastic material e parts I and
II. J Eng Mech ASCE 1988;114:1672e702.
[50] O

zbolt J, Reinhardt HW. Numerical study of mixed mode fracture in concrete.
Int J Fract 2002;118:145e61.
[51] Pivonka P, O

zbolt J, Lackner R, Mang HA. Comparative studies of 3D-constitutive models for concrete: application to mixed-mode fracture. Int J Numer
Methods Eng 2004;60:549e70.
[52] Ba

zant ZP, Oh BH. Efcient numerical integration on the surface of a sphere.
Z angew Math Mech (ZAMM, Berlin) 1986;66(1):37e49.
[53] Ba

zant ZP, Adley MD, Carol I, Jirasek M, Akers SA, Rohani B, et al. Large-strain
generalization of microplane model for concrete and application. J Eng Mech
ASCE 2000;126(9):971e80.
[54] Ba

zant ZP, Caner FC, Adley MD, Akers SA. Fracturing rate effect and creep in
microplane model for dynamics. J Eng Mech ASCE 2000b;126(9):962e70.
[55] Krausz AS, Krausz K. Fracture kinetics of crack growth. Dordrecht, The
Netherlands: Kluwer; 1988. 196 pages.
[56] Dilger WH, Koch R, Kowalczyk R. Ductility of plain and conned concrete
under different strain rates. Detroit, Michigen, USA: American Concrete
Institute (ACI Journal); 1984. pp. 73e81.
[57] O

zbolt J, Sharma A. Numerical simulation of reinforced concrete beams with
different shear reinforcements under dynamic impact loads. Int J Impact Eng
2011;38:940e50.
[58] O

zbolt J, Sharma A. Numerical simulation of dynamic fracture of concrete
through uniaxial tension and L-specimen. Eng Fract Mech 2012;85:88e102.
[59] CEB. Concrete structures under impact and impulsive loading, synthesis
report, bulletin dInformation no. 187. Laussane, Switzerland: b; 1988.
184 pages.
[60] FIB. Model code for concrete structures. Germany: b, Wilhelm Ernst & Sohn;
2010. 434 pages.
[61] Private communication with Dr. Jaap Weerheijm, Delft University of
Technology.
[62] Trava
s V, O

zbolt J, Ko

zar I. Failure of plain concrete beam at impact load: 3D
nite element analysis. Int J Fract 2009;160:31e41.
[63] Vegt I, Weerheijm J, Breugel K. The rate dependency of concrete under tensile
impact loading e fracture energy and fracture characteristics. In: Proceedings
of 13th Int. Sym on Interaction of the Effects of Munitions with Structures
ISIEMS; 2009. Brhl, Germany, CD ROM, 11 pages.
[64] Peroni M, Solomos G, Pizzinato V, Larcher M. Experimental investigation of
high strain-rate behaviour of glass. Appl Mech Mater 2011;82:63e8.