International Journal of Impact Engineering 45 (2012) 28e38

Contents lists available at SciVerse ScienceDirect

International Journal of Impact Engineering

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

Tensile behaviour of high performance bre-reinforced cementitious composites

at high strain rates
Alessio Caverzan a, *, Ezio Cadoni b, Marco di Prisco a
a
Department of Structural Engineering (DIS), Politecnico di Milano, Milan, Italy
b
Department of Environment, Constructions and Design, University of Applied Sciences of Southern Switzerland, Lugano, Switzerland

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

Article history: The promise of bre-reinforced cementitious composites for dynamic loading application stems from
Received 25 October 2010 their observed good response under static loading. An experimental research aimed at contributing to
Received in revised form the understanding of the behaviour of advanced bre-reinforced cementitious composites subjected to
29 November 2011
low and high strain rates was carried out. The material behaviour was investigated at four strain rates
Accepted 18 January 2012
Available online 2 February 2012
(0.1, 1, 150 and 300 s
1) and the tests results were compared with their static behaviour. Tests at
intermediate strain rates (0.1e1 s
1) were carried out by means of a hydro-pneumatic machine (HPM).
High strain rates (150e300 s
1) were investigated by exploiting a Modied Hopkinson bar (MHB).
Keywords:
High performance bre-reinforced
Comparison between static and dynamic tests highlighted several relevant aspects. First, with the change
cementitious composites in the strain rate, the Dynamic Increase Factor (DIF) of the material appears well predicted by some
HPFRCC models proposed in the literature up to a value of 0.1 s
1, while at higher strain rates it increases less than
High strain rates expected from models. Moreover, the post-peak behaviour showed a stress plateau inuenced by the
Dynamic behaviour bres and dependent on the strain rate.
Hopkinson bar 2012 Elsevier Ltd. All rights reserved.

1. Introduction
The mechanical behaviour of bre-reinforced cementitious
composites when subjected to impact or blast still has many
aspects open to investigation [1], with specic reference to large
and socially-sensitive structures, as sheltering structures, high-rise
buildings, bridges, off-shore platforms, pipelines, gasication
reactors, secondary containment shells for nuclear power plants,
and tunnels. For such reasons, the mechanical response of concrete
structures exposed to blast and impact loading can only be
predicteddand controlleddby formulating proper materials
models for cementitious composites, including strain rate effects.
As a matter of fact, the scanty information provided so far by spe-
cialised equipment such as the Hopkinson bar for very high strain
rates shows a signicant increase in mechanical properties, but this
increase still needs to be related to the main engineering parame-
ters. The current understanding of the dynamic response and
impact resistance of cementitious composites, and especially of
high strength concrete, is very limited. There are also contradictory
results in the literature. With reference to compressive behaviour,
* Corresponding author. Dept. of Structural Eng. (DIS), Politecnico di Milano, P.zza
Leonardo da Vinci, 32, 20133 Milano, Italy. Tel.: 39 02 2399 4251.
E-mail address: caverzan@stru.polimi.it (A. Caverzan).
Banthia [2], Banthia et al. [3], Bentur et al. [4] and Ross [5] reported
a reduced value of the ratio between dynamic and static strength
with a growing of static strength of concrete, while in the results
reported by Bischoff and Perry [6] that ratio is independent of static
strength.
High performance bre cementitious composites are charac-
terised by high toughness and a hardening behaviour in bending.
Fibres enhance the ductility of brittle materials like concrete, and
this improvement is strictly related to the process by which the load
is transferred from the matrix to the bres and the bridging effect of
bres across the cracks. Fibre pull-out is the principal mechanism
contributing to the high toughness of the material and it is the
preferred failure mode, rather than the fracture failure mechanism,
because of its ability to redistribute stresses. Gokoz and Naaman [7]
carried out tests under static condition (with a velocity n equal to
4.2  10
5 m/s) and high displacement rate (n 3 m/s) for three
types of bres (smooth steel, glass and polypropylene). They
concluded that, while polypropylene bres were very sensitive to
the imposed displacement rate, smooth steel bres were insensitive
to it. They also reported that the post-peak behaviour of steel
smooth bres, whose pull-out behaviour is essentially based on
friction, was almost insensitive to displacement rate. Banthia and
Trottier [8] investigated the pull-out resistance of deformed bres
(hooked end, crimped and I-shaped bres) embedded in a cement-

0734-743X/$ e see front matter 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijimpeng.2012.01.006
 A. Caverzan et al. / International Journal of Impact Engineering 45 (2012) 28e38
based matrix. They performed a series of static (n 8.46  10
6 m/s)
pull-out tests using a hydraulic testing machine and a series of
dynamic (n 1.5 m/s) tests carried out by means a pendulum impact
device. The study highlighted that deformed steel bres embedded
in cementitious matrixes generally sustain a higher load under
impact than under static pull-out and that the pull-out energy is
also greater under impact as long as the bre pulls out and does not
fail. Similar results were found by Kim et al. [9]. The authors studied
the pull-out sensitivity to loading rates for two types of high
strength steel bres (hooked and twisted) by applying four
displacement rates (n was equal to 0.018, 0.18, 1.8 and 18 mm/s).
They concluded that the pull-out response of high strength steel
hooked bres showed no appreciable rate sensitivity, and this result
was attributed to the fact that microcracking, from which the rate
effects are thought to arise, is localised in a small region in the
vicinity of the hooks, and therefore should not affect signicantly
the rate sensitivity. On the other hand, high strength steel twisted
bres showed rate sensitive pull-out behaviour and they attributed
this to the radial and longitudinal interface cracking that takes place
along the entire embedded bre length. Since of bre behaviour,
cement matrix and the bond between them (or the pull-out
mechanism) are likely dependent on the strain rate, it is expected
that the response of bre-reinforced cementitious composites is
also rate dependent: however, the investigation of the rate depen-
dency of the pull-out mechanism cannot conclusively identify the
complete rate sensitivity of the composite.
Beginning in the 1980s, various researchers [10e14] studied
strain rate effects on the tensile properties of strain-softening bre-
reinforced concrete. They observed a signicant increase in tensile
strength, strain at maximum stress, and fracture energy due to high
strain rates. Moreover, these studies have pointed out that the
mechanical properties and rate sensitivity of bre-reinforced
cementitious composites are dependent upon the bre type, bre
volume fraction and matrix composition, which inuences the
bond between the bre and the matrix. Very few studies have
investigated on the effect of the strain rate on the tensile behaviour
of strain-hardening high performance bre-reinforced cementi-
tious composites. Kim et al. [15], following the work presented on
the bre pull-out rate sensitivity [9], have investigated the tensile
response of High Performance Fibre-Reinforced Cementitious
Composites (HPFRCC) to strain rates. The main parameters inves-
tigated were the type of high strength steel bres (hooked and
twisted), the bre volume fraction (Vf 1% and Vf 2%), and the
matrix strength using a specimen with a bell-shaped ends geom-
etry. They have investigated four different strain rates from

4
1
3_ 10 s up to 3_ 10
1 s
1 and between these extremes, the
strain rates of 3_ 10
3 s
1 and 3_ 10
2 s
1. They reported that, in
general, the HPFRCC specimens with twisted bres are generally
sensitive to strain rate, whereas their counterparts with hooked
bres are generally not. However, it is important to highlight that the
strain rates investigated by Kim et al. [15] are considered as a seismic
rate, because they are usually reached during an earthquake, while
blast and impact scenarios are characterised by strain rates thousand
times greater than seismic ones. Cadoni et al. [16] studied the
behaviour of HPFRCC (two different bres were studied: polyvinyl-
alcohol and steel bres) when subjected to high strain rates
(_3 50 s
1 and higher) by means of a modied Hopkinson bar
(MHB). They observed, after dynamic failure, that steel bres were
pulled-out from the matrix, while the majority of polyvinyl-alcohol
bres experienced tensile failure. The difference in failure could
justify the variation in terms of fracture energy as reported by the
authors. On the other hand, the tensile strength of the steel bre
composite increases with strain rate while the tensile strength of
PVA composites seems to be less sensitive to strain rates, even
though it is remarkably increased compared to the static values.
29
2. Mix design and manufacturing
This research studies a high performance cementitious
composite optimised with steel bres. The mix design of the
HPFRCC material is specied in Table 1. The steel bres were high
carbon straight bres, 13 mm long with a 0.16 mm diameter (aspect
ratio lf/df equal to 80); their proportion in the mix was equal to
100 kg/m3 (volume fraction Vf equal to 1.25%), while the density of
the composite material was equal to 2380 kg/m3. The sand used in
the material was sieved up to 2 mm and it can be classied as mixed
quartz sand. The mix procedure was composed of several phases.
First cement and slag were mixed in dry condition for 2 min, then
water and super plasticiser were added and the cement paste was
mixed for 5 min. Then the sand was introduced into the cement
paste and mixed for 5 min. Finally, bres were added while the
mortar material was mixing. Before casting, its workability was
measured by means of a ow test carried out with Abrams cone.
A 1.6 m  0.60 m in plane slab, 30 mm thick, was manufactured. The
slab was cast by applying a unidirectional ow as shown in Fig. 1. In
order to guarantee a certain bre orientation, the properties of the
self compacting material were used, taking advantage of the ow
direction. Three prismatic beam samples, 40 mm wide and 600 mm
long, were sawed from the slab and tested in bending to perform
a proper mechanical characterisation of the material according to
Italian Guidelines. The high bre content and the favourable
orientation imposed by the casting ow control allow us to guar-
antee a small dispersion of the response before and after single-
crack localisation and a hardening behaviour in uniaxial tension
[17e20]. From the bent specimens, several small cylinders 20 mm
long, the object of the present work, were cored in the direction of
tensile stresses. Their diameter was nominally equal to 20 mm
(Fig. 2(a)). Each cylinder cored from each prismatic specimen was
notched (notch depth 1.5 mm), to be tested in uniaxial tension at
different loading rates. The specimen identication used in the
paper consists in two parts: the rst one identies the strain rate
investigated (S0 for the quasi-static tests, S1 for a strain rate equal
to 0.1 s
1, S2 for a strain rate equal to 1 s
1, etc.), while the second
part is referred to the specimen number. The tests reported are part
of a large experimental campaign investigated in Caverzans PhD
thesis. In order to simplify the identication scheme the rst part
(T20/U-20) was deleted; a reader interested to collect all the data
coming from this experimental campaign performed on HPFRCC
material and to have a database focused on the HPFRCC mechanical
behaviour characterized by different tests type should consider the
extended notation (i.e. T20/U-20-S3-1).
3. Experimental procedures
An experimental programme was carried out to properly iden-
tify the mechanical properties of high performance steel bre-
reinforced cementitious composites under static and dynamic
conditions. In particular, two different mechanical testing machines
were used: a hydro-pneumatic machine (HPM) was employed to
investigate the strain rate range between 0.1 s
1 and 1 s
1, and
Table 1
Mix design.
Constituent Dosage [kg/m3]
Cement type I 52.5 600
Slag 500
Water 200
Sand 0e2 mm 983
Fibres (lf 13 mm; df 0.16 mm) 100
Super plasticizer 33 l/m3
30 A. Caverzan et al. / International Journal of Impact Engineering 45 (2012) 28e38
Fig. 1. Slab cast applying a unidirectional ow.
a modied Hopkinson bar (MHB) [21] was exploited to investigate
high strain rates 150e300 s
1.
3.1. Quasi-static tests
Uniaxial tensile tests on notched cylindrical specimens (S0
series) were performed with an electro-mechanical testing
machine INSTRON 5867 at the laboratory of the Politecnico di
MilanodPolo Regionale di Lecco. The machine has a maximum
loading capacity of 30 kN and it is characterised by a precision for
loading measures of 0.4% of the read value up to 1/100 of the
maximum load cell capacity, and of 0.5% of the read value up to 1/
500 of the maximum load cell capacity. The precision of the stroke
displacement measures is equal to 0.05% of the read value. The
samples were glued to the machine platens by means of an epoxy
resin. Two aluminium cylinders connected to the press by
a knuckled joint (Fig. 2(b)) were used as press platens. In both
cylinders, a 5 mm deep cylindrical cavity with a 22 mm diameter
was made in order to increase the glued sample surface. Stroke was
considered as feedback parameter during the tests. The displace-
ment rate imposed during the tests was equal to 5  10
5 mm/s up
to 1.5 mm and it was progressively increased to 10
3 mm/s.
3.2. Dynamic tests at medium strain rates
The HPM functioning scheme is presented in Fig. 3. At the
beginning of the test, a sealed piston divides the cylindrical tank
into two chambers, one being lled with gas at high pressure
(viz., 150 bars), and the other with water. An equal pressure is
initially established in the water and gas chambers so that the
forces acting on the two faces of the piston are in equilibrium. The
L*
a b
Fig. 2. Test set-up, press platens and glued specimen: press device (a); press platens and glued specimen (b).
test starts when the second chamber discharges the water through
a calibrated orice that is activated by a fast electro-valve. The
piston then starts to move, expelling the water. The specimen
(S1 and S2 series) is connected on one side to the piston shaft, and
on the other side, to the end of an elastic bar which is rigidly xed
to a supporting structure. When the piston shaft moves, the spec-
imen is pulled at a xed strain rate that depends on the velocity of
expulsion of the water from the chamber. It can be demonstrated
that the movement of the piston at a constant velocity assures
a constant strain rate throughout the test. To allow for better
control of constant velocity conditions, it was chosen to perform
small decreases in the gas pressure and thus small changes in the
piston force. The elastic bar is instrumented with a strain-gauge
that provides, through the elastic properties of the bar, the force
acting on the specimen during the test. Two targets are attached at
both ends of the specimen and their movement is measured by two
contact-less displacement transducers.
Finally, the load P resisted by the specimen is measured by the
dynamometric elastic bar, whereas the specimen elongation DL is
measured by the displacement transducers, sensing the displace-
ment of the plate target xed to both specimen ends. This data
acquisition system allows of deducing the stress versus strain
relationship for the investigated strain rate.
3.3. Dynamic tests at high strain rates
The Hopkinson Pressure Bar (HPB) is an experimental device
widely used for intermediate and high strain rate testing
(101e104 s
1). Several versions of this equipment were developed
to derive the dynamic properties of a material under different stress
conditions (tensile, compression, exural, etc.) [22e24].
In the present research programme, a Modied Hopkinson Bar
(MHB), placed in the DynaMat laboratory of the University of
Applied Sciences of Southern Switzerland (SUPSI) of Lugano,
Switzerland, was used to carry out the dynamic tests at high strain
rates (Fig. 4). The MHB consists of two circular aluminium bars,
called input and output bars (with a diameter of 20 mm and having
lengths of 3 and 6 m, respectively) between which the HPFRCC
specimen (S3 and S4 series) is glued using a bi-component epoxy
resin. The input bar is connected to a high strength steel pretension
bar (having 6 m length and 12 mm diameter), used as pulse
generator. The test with the MHB is performed as follows:
a) rst, a hydraulic actuator (of maximum loading capacity of
600 kN) pulls the high strength steel bar; the pretension stored
in this bar is assured by the blocking device (number 3 in
Fig. 4);
Section A-A
Cavity
A A
Press platen
 A. Caverzan et al. / International Journal of Impact Engineering 45 (2012) 28e38
Fig. 3. Hydro-pneumatic machine (HPM).
b) the second operation is the rupture of the fragile bolt in the
blocking device, which gives rise to a tensile mechanical pulse
of duration 2.4 ms and with a linear loading rate during the rise
time. The pulse then propagates along the input and output
bars, leading the specimen to failure.
As with the classical Hopkinson bar apparatus, the pulse prop-
agates along the input bar with the velocity of the elastic wave
equal to C0, and constant shape. When the incident pulse (3 I) rea-
ches the HPFRCC sample, it is partly reected by it (3 R). The portion
of pulse that is not reected, passes through the specimen (3 T) and
propagates into the output bar, as shown in Fig. 5. The relative
amplitudes of the incident, reected and transmitted pulses
depend on the mechanical properties of the specimen. Strain-
gauges glued on the input and output bars (800 mm from the
specimen) are used to measure the elastic deformation over time
generated by the incident, reected and transmitted pulses. The
variable measures were registered by a 6 channels AlazarTech data
acquisition system (ATS330). The system is designed to provide
high precision (12 bit) waveform acquisition with a maximum
50 MS/s real-time sampling rate, 25 MHz full power bandwidth and
acquisition memory of 128K samples per channel. Each individual
board has its own independent trigger circuitry and time base, so
that each channel can be recorded with different time windows.
The data acquisition system receives the signals from the Pacic
Instrument 6120 programmable transducer ampliers. This is an
Fig. 4. Modied Hopkinson bar (MHB).
31
expandable signal conditioning system which allows the user to
congure each individual signal conditioner from any host
computer with IEEE-488 or RS-232 capability. Individual channels
may be addressed, or an all channel command will simultaneously
address all channels, for fast initial setup.
In Fig. 5, the raw signals measured on the input and output bars
are shown. The clean resolution of the incident, reected and
transmitted pulses, the sharp rise time of the incident pulse (of the
order of 30 ms), and the almost constant amplitude of the incident
pulse can be observed. In the rst phase of the tests, when the
mechanical wave reaches the samples the equation 3 I 3 R 3 T,
hence one hypothesis of the Hopkinson bar theory is not respected.
The duration of this phase depends on the wave velocity of the
material and on the length of the sample: the shorter is the spec-
imen, the lower time needed to reach the equilibrium is. In the tests
reported, samples length was equal to 20 mm this means that the
unbalanced phase was limited to 100 ms or less. While the signals
(3 I 3 R) and 3 T are equal after the initial phase and the specimen is
in equilibrium throughout the fracture process. However, it should
be notice that the unbalanced phase is a peculiar aspect of Hop-
kinson technique and can be observed in tests performed on
different types of material: brittle materials like rocks, concretes
and ductile materials like metals and alloys. Nevertheless, the
Hopkinson bar is considered as a protable technique to charac-
terize the material behaviour at high strain rate and a large amount
of papers was published in order to proof it. Thereafter, the stress
and strain in the sample can be derived from Equations (1)e(3)
where L is the specimen length (20 mm):
A0
st E0 3 t (1)
A T
2C
3_ t
0 3 R t (2)
L
Zt
2C
3 t
0 3 R tdt (3)
L
0
On the other hand, the history of the crack opening displace-
ment (COD) can be calculated:
Zt
CODt
2C0 3 R tdt (4)
0
In the case of high performance bre-reinforced cementitious
composite specimens, the strain rate can be calculated as for metallic
32 A. Caverzan et al. / International Journal of Impact Engineering 45 (2012) 28e38

strain in input and output bars [m/m]

0.0005 30 300
input stress strain-rate
output
25 250
0.0004

stress [MPa]
20 200

strain rate [s ]
0.0003 reflected pulse
15 150
0.0002 incident
pulse 10 100

0.0001 5 50
raw signals
transmitted pulse
0 0 0
0 0.0005 0.001 0.0015 0.002 0 0.0002 0.0004 0.0006 0.0008
time [s] time [s]

Fig. 5. Signals measured on the input and output bar vs time curves of HPFRCC and stress and strain rate vs time curves.

and ductile materials, in which the plastic zone is well dened. It has where samples are glued, Fig. 2(b)) is approximately considered as
to be highlighted that the displacement rate between the two a rigid zone. It is interesting to observe that steel bres are able to
sample ends is the principal data and the strain rate is then obtained favour stable propagation at room temperature up to a strain of
as an average dividing the displacement rate for the sample length. about 4%, computed according to Equation (5). The pre-peak
In order to emphasize this concept the values of displacement rates behaviour is well described by a parabolaerectangular model,
in the paragraph concerning the dynamic tests results were also where the plateau is very close to a value of 2% (Table 2).
reported. Considering the peak strength computed in the paper, the The low values of the elastic stiffness obtained in these tests
transmitted wave was used to compute nominal stresses in the could be caused by the visco-elastic behaviour of the glue. In order
samples (see Equation (1)). When a region close to the peak is to account for the contribution of the visco-elastic behaviour of the
considered, it could be observed that the transmitted wave is always glue, the results were compared with some preliminary tests aimed
a little lower than the sum of incident and reected wave, hence the at evaluating the elastic modulus of the material. These tests
transmitted wave leads to a little lower values of nominal stresses. showed an upper limit for the elastic modulus equal to 40 GPa. The
Stresses computed in this way give values of the peak strength useful elastic deformation for a stress level equal to the peak strength
for engineering applications from a safety point of view. computed with an elastic modulus of 40 GPa was compared to that
computed with the elastic modulus obtained in the tensile tests. It
4. Test results was then possible to estimate the maximum value of the error
affecting the strain measures. The difference between the two
4.1. Quasi-static tests strain values discussed above (D3 peak) is equal to about 1%. The peak
strains, for the undamaged material, could be thereafter over-
Test results are shown in Fig. 6(a,b) in terms of nominal stress estimated at a maximum of 1%.
(sN) versus crack opening displacement (COD) and fracture energy According to the Italian Guidelines, the material can be dened
(Gf) versus crack opening displacement, respectively. Moreover, in as a hardening material. The rst cracking strength is close to
Fig. 6(a) a detail of the peak zones is plotted to highlight the rst 3 MPa, while the peak strength is about 10 MPa. This means that,
linear elastic and the post-peak behaviour, close to the peak. The even if a notch depth ratio equal to 0.15 is used, the cracking
nominal stress is dened as the ratio between the applied load P process is distributed over the entire length. Again, the geometrical
and the area of the cross section A0. defect introduced by the notch and the gluing procedure prevent
Peak strengths and the corresponding crack opening displace- with a correct evaluation of the actual ductility measured in terms
ments are listed in Table 2; the values of the peak strains (3 peak) can of the strain of the peak plateau.
be calculated as follow: On the basis of these results, the total energy absorbed by the
specimen during the test up to the reaching of a negligible stress
wpeak transmitted by bres (Gf; Table 2) can be computed. The energy
3 peak (5)
L* absorbed was always computed up to a crack opening displacement
where L* is the clear span equal to 10 mm (Fig. 2(b)), and repre- equal to 5 mm, if the tests were not nished yet, a vertical cut-off
senting the equivalent specimen length. According to this was applied. Gf represents the fracture energy computed as the
assumption, the cavity depth of the platens (5 mm for each side, subtended area in the stresseCOD plain (see Equation (6)).

a 12.5 S0-1
b 25
S0-2
S0-3 20
10 average
12.5
10
G f [kJ/m2 ]

15
N [MPa]

7.5

5 10
4

2.5 5 S0-1
S0-2
S0-3
average
0 0
0 1 2 3 4 5 0 1 2 3 4 5
COD [mm] COD [mm]

Fig. 6. Quasi-static test results: stress vs COD (a); fracture energy vs COD (b).
 A. Caverzan et al. / International Journal of Impact Engineering 45 (2012) 28e38 33

Table 2
Peak strengths, crack opening at the peak and fracture energy.

Sample code fpeak [MPa] fpeak,av [MPa] (std) wpeak [mm] wpeak,av [mm] (std) Gf [kJ/m2] Gf,av [kJ/m2] (std) Fibres n
S0-1 9.71 8.62 0.17 0.20 21.23 19.51 75
S0-2 7.83 (0.97) 0.19 (0.04) 18.11 (1.58) 77
S0-3 8.33 0.25 19.19 73

a 20 b 20
S1-1 S2-1
S1-2 S2-2
S1-3 S2-3
average S2-4
15 15 average
20
15
N [MPa]

15

N [MPa]
10
10 10 10

5 5

0
5 0 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0 0.1 0.2 0.3 0.4

0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 8 9 10
COD [mm] COD [mm]

Fig. 7. Stress vs COD curves. Dynamic tests performed at: 3_ 0:1 s
1 (a); 3_ 1 s
1 (b).

Z
COD
Gf sN wdw
0 computed up to a crack opening displacement equal to 5 mm. If
The fracture energy, Gf, computed according to Equation (6),
takes into account three different contributions: the energy
required to create a new fracture surface in the matrix, the energy
involved in the pull-out process, and the energy eventually needed
for the failure of the steel bres. These contributions could have
different COD activation thresholds and different relative impor-
tances. The evolution of the fracture energy with respect to the
crack opening displacement (COD) is shown in Fig. 6(b). During
tensile tests, the generation of a new fracture surface in the matrix
and the bre pull-out are the main mechanisms involved, while the
fracture energy associated with the bre failure is negligible.
4.2. Dynamic tests at medium strain rates
For a better understanding of the dynamic behaviour of the
material, a series of tests at strain rates equal to 0.1 s
1 (specimens
named S1) and 1 s
1 (specimens named S2) were carried out
exploiting the hydro-pneumatic device. Fig. 7 shows the results in
terms of nominal stresses versus COD. To highlight the rst linear
elastic and the post-peak behaviour, details of the peak zones are
also plotted in Fig. 7. The material behaviour is well described by an
elasticeplastic model. In fact, after the peak, a stress plateau, higher
than that observed in the quasi-static test (Fig. 6(a)), up to a COD of
0.4 mm, is shown. Finally, the good bre distribution can be
appreciated by the very low scatter obtained in the tests. Peak
strength and crack opening displacement are summarised in
Table 3. The values of peak strains can be derived from Equation (5)
using L, the specimen length (20 mm), instead of L*.
(6) The total fracture energy absorbed during the tests was always
a higher opening was measured at the end of the tests, a vertical
cut-off was performed. When the test control was lost due to bre
failure, the energy up to such limit value was computed. The plot of
the fracture energy versus the crack opening displacement is
shown in Fig. 8, while the results in terms of total energy are listed
in Table 3. As an average, the fracture energy showed an increase at
0.1 s
1 with respect to that derived for quasi-static tests.
4.3. Dynamic tests at high strain rates
The displacement rates reached in the tests here reported were
equal to about 3 m s
1 (specimen series S3) and 3 m s
1 which
correspond to an equivalent strain rates of about 150 and 300 s
1
respectively. The high strain rate results obtained on HPFRCC are
reported in Table 4 in terms of peak strength, crack opening
displacement at the peak and fracture energy. In Fig. 9(a,b), the
stress versus COD curves of the HPFRCC specimens tested at high
strain rates are shown. Details of the peak zones are plotted in
Fig. 9(a,b) in order to highlight the rst linear elastic branch and
post-peak behaviour close to the peak. It can be observed that after
a remarkable peak strength, which is greater than the peak
observed in the static condition and for a strain rate equal to 1 s
1,
the specimens show a small drop down to a post-peak stress
plateau up to a COD of 0.4 mm, as highlighted in the quasi-static
tests. The stress plateau reached a stress level equal to 15 MPa, as
observed in the dynamic tests performed at 3_ 1 s
1. This quite

Table 3
Peak strengths, crack opening at the peak and fracture energy at 3_ 0:1 s
1 and 3_ 1 s
1 .

Sample code fpeak [MPa] fpeak,av [MPa] (std) wpeak [mm] wpeak,av [mm] (std) Gf [kJ/m2] Gf,av [kJ/m2] (std) Fibres n
S1-1 13.07 12.36 0.38 0.49 34.61 32.45 71
S1-2 10.30 (1.81) 0.57 (0.10) 28.07 (3.80) 76
S1-3 13.71 0.53 34.68 72

S2-1 14.29 15.25 0.37 0.61 28.03 31.40 76

S2-2 14.09 (1.65) 0.46 (0.23) 24.90 (7.75) 71
S2-3 17.67 0.81 42.57 73
S2-4 14.96 0.82 30.11 77
34 A. Caverzan et al. / International Journal of Impact Engineering 45 (2012) 28e38

a 45 b 45
40 40
35 35
30 30

G f [kJ/m2 ]

G f [kJ/m2 ]
25 25
20 20
15 15
10 S2-1
S1-1 10 S2-2
S1-2 S2-3
5 S1-3 5 S2-4
average average
0 0
0 1 2 3 4 5 0 1 2 3 4 5
COD[mm] COD[mm]

Fig. 8. Fracture energy vs COD curves. Dynamic tests performed at: (a) 3_ 0:1 s
1 ; (b) 3_ 1 s
1 .

Table 4
Peak strengths, crack opening and fracture energy at 3_ 150 s
1 and 3_ 300 s
1 .

Sample code fpeak [MPa] fpeak,av [MPa] (std) wpeak [mm] wpeak,av [mm] (std) Gf [kJ/m2] Gf,av [kJ/m2] (std) Fibres n
S3-1 20.46 22.36 0.038 0.038 34.59 29.46 79
S3-2 22.28 (1.89) 0.032 (0.01) 31.43 (7.75) 90
S3-3 24.96 0.024 24.75 79
S3-4 21.74 0.056 27.06 e

S4-1 21.88 24.25 0.017 0.018 27.32 29.46 91

S4-2 22.49 (3.60) 0.015 (0.01) 15.89 (7.75) 67
S4-3 28.39 0.021 28.66 90

constant stress plateau seems to be a dynamic material property 5. Discussion

linked to the debonding phase and partially to the rst stage of the
pull-out mechanism. High strain rates mainly affect the peak In order to have an overview of the global material behaviour,
strength rather than the bre pull-out strength. The peak strength the average of the stress versus crack opening displacement curves
in static conditions and for medium strain rates is highly inu- and fracture energy versus crack opening displacement is plotted.
enced by the bres, which contributes to developing multiple Fig. 13(a,b) summarises the experimental results in terms of the
cracks. average curves for the HPFRCC. Each curves represents the material
In Fig. 10(a,b), the evolution of the fracture energy with respect response for a different strain rate. Besides, in Fig. 14(b) a detail of
to the crack opening displacement (COD) for each strain rate is the peak zone is plotted in order to highlight the rst elastic branch
shown. As in the quasi-static tests, the fracture energy is and the region close to the peak. It is worth observing the
computed by using Equation (6). The same criteria mentioned increasing initial stiffness with the increase of strain rates. This
above were followed: when the test curves were stopped due to increase is much more evident when the strain rate is equal to 150
the sudden drop induced by bre failure, a vertical jump or 300 s
1.
assumption was used and the energy absorbed was always The material tested at 0.1 s
1 has shown a small increment in
computed up to a crack opening displacement equal to 5 mm, by the initial stiffness. A small increment was also expected for the
applying a vertical cut-off. The fracture energy computed shows material tested at 1 s
1. On the contrary, the material showed in the
a comparable value with that of the tests carried out at lower pre-peak region a behaviour well described by a parabo-
strain rates. Figs. 11 and 12 show clearly that a single macro-crack laerectangular model with an initial stiffness close to the static
was triggered in the notched section of samples tested at high one (Fig. 13(a)). The behaviour described above could be attributed
strain rates, no other fracture plane can be detected. Some to a pre-cracking problem; nevertheless this explanation seems
external micro- and nano-fracture planes could be present but unlikely because all four samples tested at 1 s
1 showed a similar
their contribution to the total energy is negligible if it is compared pre-peak behaviour. Analysing the tests results, several interesting
to the energy involved in the pull-out process which is several considerations can be highlighted. When the strain rate grows, not
orders of magnitude higher even if could play a key role in the only does the initial stiffness increase, but so do the peak strength
pre-peak behaviour. and the post-peak plateaus observed for all the strain rates

a 30 S3-1
b 30 S4-1
S3-2 S4-2
25 S3-3 25 S4-3
S3-4 average
average 30
20 30 20 25
N [MPa]

N [MPa]

25
20
20
15 15 15
15
10
10
5
10 5 10
0
0 0 0.1 0.2 0.3 0.4
0 0.1 0.2 0.3 0.4
5 5

0 0
0 1 2 3 4 5 0 1 2 3 4
COD[mm] COD[mm]

Fig. 9. Stress vs COD curves. Dynamic tests performed at: 3_ 150 s
1 (a); 3_ 300 s
1 (b).
 A. Caverzan et al. / International Journal of Impact Engineering 45 (2012) 28e38
a 35
30
25
G f [kJ/m2 ]
20
15
10 S3-1
S3-2
5 S3-3
S3-4
average
0 0
0 1 2 3 4 5
COD[mm]
Fig. 10. Fracture energy vs COD curves. Dynamic tests performed at: 3_ 150 s
1 (a); 3_ 300 s
1 (b).
Fig. 11. Photos of the samples tested at displacement rate equal to 3 m s
1 (_3 150 s
1).
investigated. While the peak strength grows when the strain rate
increases, the post-peak plateau grows from 8.5 MPa in static
condition to about 15 MPa for a strain rate equal to 1; in between,
a value of about 11 MPa was reached for a strain rate equal to 0.1 s
1.
For strain rates greater than 1 s
1, the post-peak stress plateau was
always equal to about 15 MPa. This stress level seems to be
a dynamic threshold linked to the pull-out mechanism. During the
fracture process in a pull-out test, it is possible to divide the bre
contribution into two principal parts: the rst part is that due to the
debonding phase and the second part is that due to the pull-out.
The pull-out mechanism for a straight bre is a mechanism
partially based on friction and it is completely activated for a rela-
tive large crack opening displacement. The plateau stress zone
(for a COD between 0.2 and about 0.8 mm) is surely dominated by
the pull-out mechanism. Fig. 14(a) shows the pull-out load (P)
versus slip response for a straight bre close to the peak region.
Besides, Fig. 14(a) shows also a schematic representation of the
previous curve by a polyline. When the strain rate increases, the
level of the stress plateau gradually jumps to a limit stress value of
about 15 MPa for strain rates grater then 1 s
1. This limit seems to
be a dynamic property of the material. This increase in the plateau
could be explained by an increase of the point (I) and the slope (KII)
in Fig. 14(a) with the strain rates.
The rising of the point (I) could be linked to the delamination
process along the bres in the debonding phase. This delamination
consists in a propagation of a crack in the matrix surrounding the
bre along the bre length. The delamination process could sta-
bilise with the increase of the strain rates. Instead, the increase of
KII with the growth of strain rate could be related to the deforma-
tion capability of the bres and the surrounding zones during the
Fig. 12. Photos of the samples tested at displacement rate equal to 6 m s
1 (_3 300 s
1).
35
b 35
30
25
G f [kJ/m2 ]
20
15
10
S4-1
5 S4-2
S4-3
average
0 1 2 3 4 5
COD[mm]
pull-out process. For low strain rates, the bres and the adjacent
area have time to deform themselves and/or develop plastic
deformation in order to reduce the dynamic frictional force. When
the strain rate increases, the process described above cannot take
place: there is not enough time to develop such a plastic defor-
mation and hence the dynamic frictional force increases. Looking to
the matrix behaviour at high strain rates (_3 150e300 s
1),
a remarkable increase of the peak strength can be pointed out
(see Fig. 13(a)). This strength enhancement can be essentially
associated to the matrix. When the peak is reached, a localised
crack appears, causing a drop in the stress. The stress drops down to
a value of about 11 MPa for a strain rate equal to 150 s
1 and to
about 14 MPa for a strain rate equal to 300 s
1. These break points
and the following stress recovery could be associated with a stabi-
lisation of the delamination process in the dynamic eld.
Looking to the softening behaviour of the pull-out mechanism
for a for a COD between 1 and 2.5 mm is inuenced by the slope KIII
(see Fig. 13(a)). In that COD range an embrittlement of the material
response with the increase of strain rate can be highlighted. The
embrittlement can be related to a decrease of KIII with the increase
of strain rate. Analysing Fig. 13(a) at higher crack opening
displacements (from COD equal to 2.5 mme5 mm), an alignment of
the material response can be pointed out for all the strain rates
investigated.
In Fig. 15(a) the dynamic increase factor (DIF) versus strain rate
is plotted. The dynamic increase factor is computed as the ratio
between the dynamic strength of the material (fpeak,dynamic) and the
static strength of the material (fpeak,static). The DIF parameter can be
used to quantify the rate sensitivity of this material. The DIF curves
are compared with the curves proposed by Model Code 90 (MC90),
36 A. Caverzan et al. / International Journal of Impact Engineering 45 (2012) 28e38

a b 35
25 S0
S1 30
S2
20 S3 25
S4

G f [MPa]
N [MPa]
15 20

15
10
10 S00
S
S1
5 S2
5
S3
S4
0 0
0 1 2 3 4 5 6 7 0 1 2 3 4 5
COD[mm] COD[mm]

Fig. 13. Tests results: average curves. Stress vs COD (a) and fracture energy vs COD (b) at different strain rates.

Model Code 2010 (MC10), Malvar and Ross [25] and Birkimer [26]
for plain concrete (the tests performed by Birkimer do not
provide any information for the strain rates lower than 2, hence
a linear trend was supposed in this range). Different from what is
proposed by the codes, Malvar and Ross [25] the HPFRC investi-
gated here has a change in its DIF slope at 0.1 s
1. A transition zone
between 0.1 and 150 s
1, with a slope lower than the slope
proposed by all the formulations taken into account, could char-
acterise the DIF trend of the material. Up to 150 s
1 a zone with
higher slope than the previous one could be dened. More data for
strain rates higher than 300 s
1 are needed in order to conrm the
different slope for strain rates higher than 150 s
1. However the DIF
trend for high strain rates has a slope lower than that suggested by
the codes. Finally, even if the transition zone is neglected and a bi-
linear curve is used to describe the DIF trend, the slope would
change for a strain rate equal to about 10 s
1. To the best knowledge
of the authors, no studies have been reported on the effect of strain
rates on fracture energy in bre-reinforced cementitious compos-
ites. As mentioned previously, many experiments have been per-
formed to study the tensile strength over a wide range of strain
rates, but little data is available for the fracture energy in plain
concrete [27e30]. Weerheijm [27] measured in direct tension tests
with a Hopkinson bar over the complete stressecrack opening
relation. The author reported the fracture energy at strain rates
between 10
1 and 100 1/s: from those results an increase in the
fracture energy with the strain rate or crack opening velocity
cannot be seen. On the other hand, Schuler et al. [28,29] focussed
their studies on the fracture energy at strain rates between 101 and
102 by exploiting a split Hopkinson bar. In spall experiments, it is
not possible to measure stress over strain or crack opening as can
be done in a quasi-static tension test, and for this reason, Schuler at
al. computed the fracture energy from the reduction of the frag-
ment velocities. From the experimental results they observed
a dynamic increase factor for the fracture energy between 2 and 3.
Also Brara and Klepaczko [30] performed several dynamic tests on
micro concrete by exploiting a split Hopkinson bar. The micro
concrete tested in their experiments was characterised by
a maximum aggregate size, as used in the HPFRCC object of the
present paper, equal to 2 mm but with a lower strength. Analysing
the experimental results the authors pointed out a drastic increase
in the total fracture energy: the value computed at 120 s
1 is three
times higher than the static one. In Fig. 15(b) the results in terms of
the ratio between dynamic and static fracture energy versus the
strain rate is plotted. It can be highlighted that the material had
always a dynamic fracture energy higher than the static one, as
reported by Schuler and Hansson [29]. The dynamic increase factor
for the fracture energy (DIFGf) does not appear to be rate dependent
as shown for the plain high performance concrete by Schuler and
Hansson [29], but it is close to 1.6 up to a strain rate equal to 150 s
1
and it decreases to 1.2 for a strain rate equal to 300 s
1. However, it
is difcult to compare directly the results in terms of fracture
energy for plain and bre concrete where the presence of bres
plays a signicant role. The mechanism of failure for plain concrete
at high loading rates is characterised by the propagation of many
micro-cracks at the same time. This mechanism could explain the
increase in fracture energy. At high strain rates the micro-cracks
could not have enough time to follow the minimum energy path
as in the quasi-static tests where the micro-cracks nucleate in
a macro-crack which can propagate with a certain velocity. At high
strain rates the micro-cracks are forced to propagate along the
shortest path characterised by a higher resistance [25,30,31]. This
phenomenon can explain the drastic increase in fracture energy
observed in plain concrete at high strain rates (>1 s
1). On the
other hand, bre-reinforced cement-based materials are charac-
terised by high fracture energy due to the pull-out mechanism.
The energy involved during the bre pull-out represents the prin-
cipal energetic contribution in the failure process. The pull-out
mechanism and its rate dependancy are completely different
from the failure behaviour of plain concrete as Figs. 13 and 15(b)
show. Finally, it should be noted that the good bre orientation

a P b 25
P S0
S1
20 S2
S3
S4
15
N [MPa]

10

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0 COD[mm]

Fig. 14. Expected pull-out behaviour for straight bres (a); stress vs COD for different strain rates: detail of the peak zone (b).
 A. Caverzan et al. / International Journal of Impact Engineering 45 (2012) 28e38
a 3
2.5 HPFRCC
MC1990
MC2010
Malvar
Birkimer
DIF
2 1
1.5
1 0
1e-06 1e-05 0.0001 0.001 0.01 0.1
.
Fig. 15. Static and dynamic results: comparisons. DIF vs strain rate (a) and fracture energy ratio vs strain rate (b).
obtained in the samples produced a fracture energy stabilisation in
the whole strain rate range investigated.
6. Conclusions
The behaviour of an advanced high performance bre-
reinforced cementitious composite when subjected to different
strain rates was investigated. An extensive experimental investi-
gation was carried out by means of uniaxial tension tests in the
static eld. Furthermore, a dynamic characterisation was per-
formed in uniaxial tension by exploiting a hydro-pneumatic
machine and a modied Hopkinson bar.
From the experimental results presented in this research work
some important conclusions can be drawn:
e With reference to the material self compacting properties,
a good bre alignment was obtained by imposing a unidirec-
tional casting ow. This result is conrmed by the low scatter
observed in the material response. The low scatter was
observed in the overall test series carried out, and at all the
strain rates investigated. The material so cast, and thus char-
acterised by a good bre alignment, showed a very high
performance compared with other cementitious composites at
comparable cost.
e On the basis of the results coming from the static and dynamic
tests carried out, it is possible to consider a cylindrical sample
20 mm high, with a diameter equal to 20 mm, as a represen-
tative volume for this material characterised by a good bre
orientation.
e The comparison between static and variable strain rate tests,
carried out by means of three different mechanical devices,
exhibits high values of the dynamic increase factor. For strain
rates up to 0.1 s
1, the test results seem to be well predicted by
the trend proposed in the Model Code 2010 for plain concrete.
Nonetheless, the DIF values for HPFRCC start to increase at
a lower value of the strain rate (between 0.1 and 1 s
1) than that
suggested by different models (1 s
1 for Malvar and the Model
Code 2010, 30 s
1 for the Model Code 1990). Between 1 and
150 s
1 a transition zone with a lower slope compared to that
expected for plain concrete was observed. By increasing the
strain rate to 300 s
1, the DIF increases, but the models consid-
ered overestimate the slope also in this range of strain rates.
e The stress plateau shown in the quasi-static tests was obtained
also at different strain rates. Increasing the strain rate from
10
6 to 1 s
1 the stress plateau grows from 8.5 to 15 MPa. For
higher strain rates (from 3_ 150 s
1 to 3_ 300 s
1) the peak in
strength considerably increases, because it is related to the
matrix properties and to the bre debonding phase. However,
in the post-peak region, after a drop, the stresses reach the
stress plateau at 15 MPa up to a crack opening displacement of
37
b 2
1.5
DIF Gf
0.5
1 10 100 1000 1e-06 1e-05 0.0001 0.001 0.01 0.1 1 10 100 1000
0.8 mm. This value of the stress plateau (15 MPa) appears to be
an upper bound limit and a dynamic property of the investi-
gated material.
e The high strain rates seem to modify the strain localisation
process. In fact, for low strain rates the stable crack propagation
seems to be associated to the growth of a process zone char-
acterised by a stable multi-localisation. On the contrary, at high
strain rates a strain localisation takes place for very small crack
opening favouring, the uncoupling of unstable mode I crack
propagation from the pull-out delamination.
e The dissipated energy up to a large crack opening (5 mm)
remains almost constant for growing strain rates up to a value
of 150 s
1 and scantly decreases at 300 s
1.
Acknowledgements
The Authors are grateful to Matteo Dotta and Daniele Forni of
the University of Applied Sciences of Southern Switzerland, Matteo
Colombo and Giulio Zani of the Politecnico di Milano for their
precious collaboration in the execution of the laboratory tests. The
research was partially nanced by the project ACCIDENT ID
7629770 in the frame of INTERREG IT/CH 2006-2013 program.
References
[1] Bindiganavile V, Banthia N, Aarup B. Impact response of ultra-high-strength
ber-reinforced cement composite. ACI Materials Journal 2002;99:543e8.
[2] Banthia N. Impact resistance of concrete. Ph.D. thesis, University of British
Columbia, Vancouver, British Columbia, Canada; 1987.
[3] Banthia N, Mindess S, Bentur A, Pigeon M. Impact testing of concrete using
a drop-weight impact machine. Experimental Mechanics 1989;29:63e9.
[4] Bentur S, Mindess N, Banthia. Behavior of reinforced concrete under impact:
the effect of concrete strength. Journal of Society of Experimental Mechanics;
1987:449e58.
[5] Ross. Review of strain rate effects in materials. In: Structures under extreme
loading conditions, ASME pressure vessels and piping conference, pp.
255e262.
[6] Bischoff PH, Perry SH. Impact behavior of plain concrete loaded in uniaxial
compression. Journal of Engineering Mechanics 1995;121:685e93.
[7] Gokoz U, Naaman AE. Effect of strain-rate on the pull-out behaviour of bres
in mortar. International Journal of Cement Composites 1981;3:187e202.
[8] Banthia N, Trottier J-F. Deformed steel berecementitious matrix bond under
impact. Cement and Concrete Research 1991;21:158e68.
[9] Kim D, El-Tawil S, Naaman A. Loading rate effect on pullout behavior of
deformed steel bers. ACI Materials Journal 2008;105:576e84.
[10] Nammur G, Naaman AE. Strain rate effects on tensile properties of ber
reinforced concrete. In: Mindess S, editor. In: MRS symposium on
cement based composites: strain rate effects on fracture, vol. 64, Pitts-
burgh. pp. 97e118.
[11] Kormeling H, Reinhardt H. Strain rate effects on tensile properties of ber
concrete in uniaxial tension. International Journal of Cement Composites and
Lightweight Concrete 1987;9:197e204.
[12] Gopalaratnam VS, Shah SP. Properties of steel ber reinforced concrete sub-
jected to impact loading. ACI Materials Journal 1986;83:117e26.
[13] Suaris W, Shah SP. Strain-rate effects in bre-reinforced concrete subjected to
impact and impulsive loading. Composites 1982;13:153e9.
38 A. Caverzan et al. / International Journal of Impact Engineering 45 (2012) 28e38
[14] Lok TS, Zhao PJ. Impact response of steel ber-reinforced concrete using a split
Hopkinson pressure bar. Journal of Materials in Civil Engineering 2004;16:
54e9.
[15] Kim DJ, El-Tawil S, Naaman A. Rate-dependent tensile behavior of high
performance ber reinforced cementitious composites. Materials and Struc-
tures 2009;42:399e414.
[16] Cadoni E, Meda A, Plizzari G. Tensile behaviour of FRC under high strain-rate.
Materials and Structures 2009;42:1283e94.
[17] Caverzan A, Colombo M, di Prisco M, Lapolla S. On high-temperature behav-
iour of HPFRC thin plates. In: Uomoto T, Banthia N, editors. Proceedings of
ConMat09. Japan: Nagoya; 2009. p. 444e9.
[18] Ferrara L, Faifer M, Muhaxheri M, Toscani S, Ottoboni R. Connecting non-
destructive ber dispersion measurements with tensile HPFRCC behaviour. In:
Parra-montesinos GJ, Reinhardt HW, Naaman AE, editors. Proceedings of
HPFRCC 6, RILEM, Ann Arbor. USA: Springer; 2011. p. 43e50.
[19] Ferrara L, di Prisco M, Khurana R. Tailoring optimum performance for struc-
tural use of self consolidating SFRC. In: Gettu R, editor. Proceedings of Beb
2008, RILEM, Chennai, India, pp. 739e750.
[20] Faifer M, Ottoboni R, Toscani S, Ferrara L. Nondestructive testing of steel-ber-
reinforced concrete using a magnetic approach. IEEE Transactions on Instru-
mentation and Measurement 2011;60:1709e17.
[21] Cadoni E, Albertini C, Solomos G. Analysis of the concrete behaviour in tension
at high strain-rate by a modied Hopkinson bar in support of impact resistant
structural design. Journal of Physics: IV (France) 2006;134:647e52.
[22] Ross C. Split-Hopkinson pressure bar tests. Technical Report. FL: Air Force
Engineering and Services Center Tyndall AFB; 1989. pp. 1e80.
[23] Reinhardt H, Kormeling H, Zielinski A. The split Hopkinson bar, a versatile
tool for the impact testing of concrete. Materials and Structures 1986;19:
55e63.
[24] Cadoni E, Solomos G, Albertini C. Mechanical characterisation of concrete in
tension and compression at high strain rate using a modied Hopkinson bar.
Magazine of Concrete Research 2009;61:221e30.
[25] Malvar LJ, Ross C. Review of strain rate effects for concrete in tension. ACI
Materials Journal 1998;95:735e9.
[26] Birkimer D. Critical normal fracture strain of Portland cement concrete. Ph.D.
thesis, University of Cincinnati; 1968.
[27] Weerheijm J. Concrete under impact tensile loading and lateral compression.
Ph.D. thesis, Technische Universiteit Delft; 1992.
[28] Schuler H, Mayrhofer C, Thoma K. Spall experiments for the measurement of
the tensile strength and fracture energy of concrete at high strain rates.
International Journal of Impact Engineering 2006;32:1635e50.
[29] Schuler H, Hansson H. Fracture behaviour of high performance concrete (hpc)
investigated with a Hopkinson-bar. Journal of Physics: IV (France) 2006;134:
1145e51.
[30] Brara JR, Klepaczko. Fracture energy of concrete at high loading rates in
tension. International Journal of Impact Engineering 2007;34:424e35.
[31] Zielinski, Fracture of concrete and mortar under uniaxial impact tensile
loading. Ph.D. thesis, Technische Universiteit Delft; 1982.