Meso Scale Model for Fiber-Reinforced-Concrete:

Microplane Based Approach

Željko Smolčić 1 and Joško Ožbolt 1,2
1
Faculty of Civil Engineering of University of Rijeka, Radmile Matejčić 3, Rijeka, Croatia
2
Institut für Werkstoffe im Bauwesen, Universität Stuttgart, Pfaffenwaldring 4, Stuttgart, Germany

(Received keep as blank , Revised keep as blank , Accepted keep as blank )

Abstract. In the present paper experimental and numerical analysis of hook-ended steel fiber reinforced concrete is
carried out. The experimental tests are performed on notched beams loaded in 3-point bending using fiber volume
fractions up to 1.5 %. The numerical analysis of fiber reinforced concrete beams is performed at meso scale. The
concrete is discretized with 3D solid finite elements and microplane model is used as a constitutive law. The fibers
are modelled by randomly generated 1D truss finite elements, which are connected with concrete matrix by discrete
bond-slip relationship. It is demonstrated that the presented approach, which is based on the modelling of concrete
matrix using microplane model, able to realistically replicate experimental results. In all investigated cases failure is
due to the pull-out of fibers. It is shown that with increase of volume content of fibers the effective bond strength and
slip capacity of fibers decreases.

Keywords: concrete; steel fibers; 3D finite element analysis; meso-scale; microplane model; bond-slip.

1. Introduction

Fiber reinforced concrete is a composite material, to which, in addition to the usual concrete
ingredients, small discontinuous fibers of high tensile strength are added during the mixing
process. The fibers used in fiber reinforced concrete are classified according to the material they
are made of: steel fibers, glass fibers, synthetic fibers and natural fibers (ACI Committee 544
2001). Compared to the same concrete composition without fibers, fiber reinforced concrete has
significantly higher fracture energy but also compressive and tensile strengths are higher.
However, for some fiber types or for excessively large volume fractions of fibers, compressive or
tensile strengths might be reduced. The main advantages of fiber reinforced concrete are visible at
the post peak response, when the fibers bridge the cracks and contribute to the resistance and
ductility (ACI Committee 544 2001).
Due to its composition concrete is an extremely heterogeneous quasi-brittle material, not
simple to model realistically. From the numerical point of view this makes the modelling even
more complex. In the numerical modelling macro scale models are mostly used in engineering
practice, however, numerically more demanding meso scale models provide better insight into the
actual concrete behaviour. Numerical studies are essential to improve the properties of the

Corresponding author, Ph.D., E-mail: [email protected]
material, such as concrete or fiber reinforced concrete. Having a realistic numerical model,
numerical parametric studies together with experiments can serve as important tool to improve
material and structural properties. An overview of the models for modelling of fiber reinforced
concrete is given in Kunieda et al. (2011), according to which the majority of fiber reinforced
concrete models are based on: (1) Models formulated in the framework of continuum mechanics
using stress-strain (σ-ε) relationship and smeared crack approach (Han et al. 2003, Suwada and
Fukuyama 2006); (2) Discrete models based on the stress-crack opening law (σ-w, discrete cracks)
(Fischer et al. 2007, Maalej 2001) and (3) The combination of (1) and (2) (Bolander and Saito
1997).
The numerical analysis of fiber reinforced concrete can be carried out at the macro or meso
scale. For the modelling at the macro scale material properties have to be be homogenised and,
depending whether the cracking is modelled using smeared or discrete approach, stress-strain
softening law or stress-crack opening law is employed. These models are suitable for analysis of
structures (e.g. beams, frames, slabs, etc.). As an example for the modelling of the fiber reinforced
concrete within the concept of smeared crack approach at the macro scale, where the effect of
fibers is implicitly accounted for in the constitutive law, is the recently proposed microplane model
M7f (Caner et al. 2013). However, when modelling at the meso scale, the fiber reinforced
concrete is usually considered as a three - phase material consisting of: cement matrix (concrete),
fibers and interface between cement matrix and fibers. Every single fiber has to be modelled
separately so that the fiber position and its orientation, as random variables, have to be specified.
In such detailed models where the concrete, fibers and its connection (bond) are explicitly
discretized (e.g. using finite elements) and characterised by the specific constitutive law, the
macroscopic response is the result of the meso scale analysis. These models are useful for detailed
parametric studies, however, they are computationally too expensive to be employed in structural
analysis. Typical examples of such modelling approach are based on the Lattice models and the
Rigid-Body-Spring Model (RBSM) (Kunieda et al. 2011, Ogura et al. 2013, Bolander and Saito
1997, Jun and Mechtcherine 2010, Bolander and Sukumar 2005, Bolander et al. 2008, Kang et al.
2014, Schauffert and Cusatis 2012, Schauffert et al. 2012).
In the present paper experimental and numerical results of 3-point bending pre-notched beams
made of fiber reinforced concrete are presented and discussed. In order to provide better insight
into the actual behaviour of fiber reinforced concrete, the numerical analysis is performed at the
meso scale. The aim is to investigate whether the meso scale model based on the microplane
model for matrix (concrete) is able to replicate the macroscopic response of concrete and to
investigate the effective bond-slip relationship between fibers and concrete for different volume
content of fibers. In the analysis fiber reinforced concrete is considered as a three-phase composite
consisting of concrete, fibers and the interface between them. Concrete is discretized by 3D four
nodes constant strain finite elements with the microplane model as a constitutive law (Ožbolt et
al., 2001). The fibers are modelled as a simple truss finite elements that are randomly distributed
over the concrete and represented by the uniaxial stress-strain relationship for steel. The
connection between concrete and fibers is modelled with discrete bond-slip relationship. The
calibration of the parameters for concrete and for the bond-slip relationship is obtained by fitting
experimental results of concrete beams without and with different fiber content. Finally, numerical
parametric study is carried out to find out the relationship between volume content of fibers and
effective bond-slip relationship. The results indicate that effective bond strength and slip capacity
degrade with an increase in volume content of fibers, beyond a threshold value.
2. Experimental investigations

Three-point bending tests are performed on pre-notched fiber reinforced concrete beams. In the
experiments the concrete quality is kept constant and only the volume content of steel fibers is
varied in the range from 0 to 1.5 %. The experimental results are then used to calibrate numerical
model in order to investigate the influence of volume content of fibers on the effective bond-slip
constitutive law for fibers.

2.1 Material properties

The compressive strength of concrete is measured according to EN 12390-3 standard (EN

12390-3:2001 2001) on three concrete cubes dimensions 150x150x150 mm. Tensile strength is
measured by employing tensile splitting tests according to EN-12390-6 standard (EN 12390-
6:2000 2000) on 150x150x150 mm concrete cubes using three specimens. The mean values of
compressive and tensile splitting strength are fc=74.47 MPa and fts=4.38 MPa, respectively.
In the fiber reinforced concrete the hooked-end steel fibers Dramix RC 65/35 BN are used. The
fiber length being lf = 35 mm, the diameter df = 0.55 mm (Fig. 1) and the aspect ratio lf /df = 65.
The fiber tensile strength is fs = 1.345 GPa and Young’s modulus Es = 210 GPa. To obtain bond-
slip relation of a single fiber, three pull-out tests are carried out on 40x40x160 mm concrete prisms
(Fig. 2). The steel fibers were embedded in the middle of the concrete prism, with the embedment
length equal to one fourth of the fiber length (le = lf/4 = 35/4 ≈ 9 mm). The axis of the fiber is
perpendicular to the surface of the concrete specimen. The pull-out tests are carried out by
displacement control, the constant displacement rate being 0.005 mm/s. The experiment is
conducted until the fiber is completely pulled out from the specimen (total displacement of 9 mm).
The measured pull-out load vs. fiber displacement curves and the corresponding mean value curve
are shown in Fig. 3. Note the presence of the post-peak hump, which is characteristic of the actions
of the hook-ended of the fiber during pull-out. Note that due to the averaging of test data (n =3) the
hump is not clearly seen in the average response curve. The post peak hump is a consequence of
fully straightened end of hooked fiber.

2.2 Three-point bending tests on pre-notched beams

The bending tests are carried out on square cross-section beams dimensions bxh = 150x150
mm of the total length L = 550 mm. The beam is simply supported with the span of l = 500 mm
(see Fig. 4), according to RILEM recommendations for fiber reinforced concrete structures TC
162-TDF (Vandewalle 2002). The notch at the mid-length of the beam was cut 28 days after
casting using the wet sawing method. The beam was turned on its side at 90° against the casting
surface and the notch was sawn over the entire width. The notch width is 5 mm and 25 mm in
depth (see Fig. 4).
The beam is loaded by displacement control with displacement rate at the mid-span of 0.01
mm/min for plain concrete beams and 0.2 mm/min for the fiber reinforced beams, respectively.
For fiber reinforced beams the displacement controlled test is carried out up to the total
displacement of approximately 30 mm. During the test the load and displacement at the bottom
surface of the beam were continuously measured (see Fig. 5). The tests are performed for four
different fiber volume fractions (Vf = 0%, 0.5%, 1.0% and 1.5%) and for each fiber fraction three
specimens are tested, i.e. in total 12 specimens.
Experimentally obtained load-deflection curves for the beams, with different fiber volume
fractions (Vf=0%, 0.5%, 1.0% and 1.5%) are plotted in Fig. 6. Each figure shows the load-
displacement relationship for all tested specimens as well as the mean value (in black). For plain
concrete beams (Vf=0 %) the displacement up to 0.5 mm is shown, while for fiber reinforced
concrete (Vf=0.5%, 1.0% and 1.5%) the displacement up to 5.0 mm is plotted. It can be seen that
up to the formation of the first crack in concrete the load-displacement curves are almost identical
for all beams, and only after the first crack appearance large scatter of measured data can be
observed. As expected, with addition of steel fibers the beam response becomes more ductile. It is
obvious that with increasing the fiber volume fraction, the resistance and ductility of the beams
increase. In all cases the failure is due to the mode-I fracture with formation of single discrete
crack.

3. Meso scale modelling approach

In the numerical 3D finite element analysis at the meso scale, fiber reinforced concrete is
considered as a three-phase composite consisting of concrete, fibers and the discrete interface
between them. Concrete is discretized by 3D finite elements with the microplane model as a
constitutive law. The fibers are modelled as simple 1D truss finite elements that are randomly
distributed over the concrete and the uniaxial stress-strain relationship for steel is used as a
constitutive law. The connection between concrete and fibers is simulated with discrete bond-slip
relationship.

3.1 Concrete

The numerical analysis is carried out using three-dimensional finite element (3D FE) code
MASA developed at the University of Stuttgart (Ožbolt 1998). Through a number of numerical
studies it has been demonstrated that the code is able to realistically predict behaviour of concrete
structures (Ožbolt et al. 2001, Ožbolt et al. 2002, Ožbolt and Reinhardt 2002, Ožbolt et al. 2000).
In the analysis the concrete is simulated by the microplane model in which damage and cracking
phenomena are modelled within the concept of smeared cracks. To assure mesh objective results
relatively simple crack band approach is used (Bažant and Oh 1983).
It is important to note that for the realistic modelling of concrete the constitutive law plays the
most important role. In the microplane model, the material response is calculated based on the
monitoring of stresses and strains in different predefined directions. Integrating microplane
stresses in a thermodynamically consistent way, from a known macroscopic strain tensor it is
possible to calculate macroscopic stress tensor. The constitutive framework is similar to discrete
type of the models (e.g. random particle model) with the difference that the model is formulated in
the framework of continuum. The physical concept behind the microplane model was already
discussed at the beginning of last century by Mohr (1900) and Taylor (1938). In the model, the
material is characterized by uniaxial relations between stress and strain components on planes of
various orientations. These planes may be imagined to represent damage planes or weak planes in
the microstructure, such as those that exist at the contact between aggregate and the cement
matrix. Unlike phenomenological models for concrete (e.g. plasticity or damage based models),
which are based on tensor invariants, in the microplane model the tensorial invariance restrictions
need not be directly enforced. Superimposing, in a suitable manner, the responses from all the
microplanes automatically satisfies them.
The used microplane model (Ožbolt et al. 2001) is based on the so-called relaxed kinematic
constraint concept. It is a modification of the M2 microplane model proposed by Bažant and Prat
(1988). Each microplane is defined by its unit normal vector components ni (see Fig. 7).
Microplane strains are assumed to be the projections of macroscopic strain tensor ij (kinematic
constraint). On the microplane are considered normal ( N, N) and two shear stress-strain
components (K, M, K, M).
To realistically model concrete, the normal microplane stress and strain components have to be
decomposed into volumetric and deviatoric parts ( N = V+D, N = V+D). Unlike most
microplane formulations for concrete, which are based on the kinematic constrain approach, to
prevent unrealistic model response for dominant tensile load and to prevent stress locking
phenomena, in the present model kinematic constrain is relaxed by additional discontinuity
function that is introduced at the macro level. Based on the micro-macro work conjugancy of
volumetric-deviatoric split and using pre-defined microplane stress-strain constitutive laws, the
macroscopic stress tensor is calculated as an integral over all possible, pre-defined, microplane
orientations:

(1)

where S denotes the surface of the unit radius sphere and ij denotes Kronecker delta and ki and mi
are directions of shear microplane components. The integration is performed by numerical
integration using 21 integration points (symmetric part of the sphere, see Fig. 7). For more detail
see (Ožbolt et al. 2001).

3.2 Steel fibers

The constitutive law for fibers is defined by the uniaxial stress-strain relationship. The three-
linear stress-strain law shown in Fig. 8 is used, which is defined by five parameters: initial Young’s
modulus E0, hardening modulus Eh, yield stress fy and tensile (compressive) strength fs and limit
strain f. Once the strain reaches the limit value it is assumed that the stress immediately drops to
zero. However, as will be shown later, in all here investigated cases the failure of steel was not
relevant since the fibers were pulled-out before the limit strain was reached. The constitutive law
is assumed to be the same for tension and compression.

3.3 Connection between fibers and concrete: discrete bond-slip relationship

The discrete bond model for the modelling of bond between concrete and deformed
reinforcement bars is here used to simulate the connection between fibers and concrete (Ožbolt et
al. 2002). In the finite element model the bond model simulates the connection between concrete
3D finite elements and the fibers that are represented by 1D truss finite elements. The connection
perpendicular to the fiber orientation is assumed to be perfect and the connection in direction of
fiber axis is defined by discrete bond-slip law. The fiber slip s represents a relative displacement
between “the same“ concrete and the fiber node in direction of fiber axis. It is modelled using zero
length nonlinear spring element. The equation used for the discrete bond model is divided into four
parts (Ožbolt et al. 2002, see Fig. 9):

(2)

where the parameters for s (slip), τ (bond stress), k (stiffness) and R (radius of curvature of the
ascending branch) are defined as shown in Fig. 9.

3.4 Generation of fiber finite elements

For the defined 3D space of the beam the fibers are generated randomly. For the given length of
fibers (35 mm) the position of their centre of gravity and orientation are chosen as a random
variables. The random generator fulfils two constrains: (1) the generation of fibers is terminated
when in the advanced specified volume content of fibers is reached and (2) the crossing between
fibers is not allowed, i.e. if two fibers are crossing one of them is removed. Once the fibers are
generated they are discretized by four truss finite elements as a straight bars, i.e. the hooks are not
modelled (see Fig. 10). Note that their mechanical effect is taken into account indirectly through
the effective bond-slip relationship. Truss finite elements are used as a constraint for the generation
of four node concrete solid finite elements. At the common points two nodes are introduced and
connected with zero length nonlinear spring elements, which represent bond-slip constitutive law,
as defined above.
In order to save computational time, in the 3-point bending finite element analysis of the beam,
only central part of the beam (150x150x150 mm) is modelled as a three-phase material. The rest of
the beam, which is out of the zone of interest, is modelled using standard 3D solid finite elements
representing fiber reinforced concrete at macro scale. Furthermore, in the analysis simplified
version of the bond-slip relationship from Fig. 9 is used in most cases with the assumption R = 
(see Fig. 10). The essential parameters for the calibration of the bond - slip relationship (τ-s) are:
k1= k2= ksecant, τmax=τm+τf and s3. The contribution of the hooks is smeared out over the length of the
fibers and it is represented by the mechanical bond (τm). The friction between deformed fiber and
concrete is accounted for through τf. Both contributions are assumed to be the same in all five
nodes of the fiber finite elements. This simplification is introduced in order to avoid modelling of
complex processes related to the deformation of hooks when a fiber is pulled out from the
concrete. In such modelling approach the fiber has to be discretized with several beam finite
elements in order to simulate pull-out from the concrete realistically. In the present simulations
four finite elements are used.
4. Numerical analysis

4.1 Finite element model

The numerical analysis is carried out for notched beams with different fiber volume fractions
(Vf=0%, 0.5%, 1.0% and 1.5%). For the finite element discretization of the beam with fiber content
Vf=0.5 % and Vf=1 %, in total 1874 and 3784 fibers, respectively, are generated in the middle zone
of the beam. For the fiber volume fraction of Vf=1.5 % the number of concrete 3D finite elements
becomes extremely large and the analysis very demanding. Therefore, the finite element mesh for
this case (Vf=1.5 %) is the same as for the case Vf=1.0 %, however, the cross-section area and
diameter of the fibers are adopted accordingly, i.e. an increase of 1.5 times in the fiber cross
section area and fiber circumference are accounted for. The models for plain concrete (Vf=0 %) are
generated for concrete meshes that correspond to Vf=0.5 % and Vf=1.0 %. These two models are
marked with Vf=0.0 % (0.5 %) and Vf=1.0 % (1.0 %), respectively. Note that the discretizations
without bond elements and fibers need to be verified with respect to the mesh objectivity, i.e. two
different meshes for plain concrete beams should result to the same response. The typical finite
element discretizations for beams with 0.5 % and 1.0 % of volume fiber fractions together with
random distribution of fibers are shown in Fig. 11.

4.2 Calibration of concrete parameters

The model parameters are obtained by fitting of the test results for plain concrete beams loaded
in 3-point bending under displacement control. The experimentally obtained load-displacement
curve for mean value is fitted by the numerical simulation. In Fig. 12 are shown stress-strain
curves for uniaxial compression and tension, respectively. The curves are plotted for the crack
band width (element size) h = 5 mm. The resulting macroscopic properties of concrete are:
Young’s modulus Ec = 38652 MPa, Poisson’s ratio c = 0.18, uniaxial compressive strength
fc = 75.0 MPa, uniaxial tensile strength ft = 5.24 MPa and fracture energy GF = 92 J/m2.
In Fig. 13 are plotted experimentally and numerically obtained curves for two different meshes.
As can be seen the agreement between experimental and numerical results for plain concrete is
reasonably good for both discretizations.

4.3 Effective bond - slip relationship of single fiber

The mechanical properties of steel fiber are given by the producer and are as follows: Young’s
modulus Es = 210 GPa, Poisson’s ratio νs = 0.33, yield stress fy = 1.211 GPa, strength fs = 1.345
GPa, hardening modulus Eh = 14.56 GPa and limit strain s = 0.05. The bond-slip relationship (τ-s)
for single fiber is obtained from the single fiber pull-out test (see Fig. 3) with bond stress:

(3)

where P is the pull-out load, df is the fiber diameter (0.55 mm) and le is the embedment length (9
mm).
The model parameters of the effective bond-slip curve (see Fig. 9) for single fiber loaded in
axial direction are obtained by fitting experimentally obtained bond-slip relationship (mean value,
Fig. 3). Based on the fit, the discrete bond model parameters are obtained as: τm=10.17 MPa,
τf=1.83 MPa, ksecant=17.129 MPa/mm, k1=41.808 MPa/mm, k2=2.707 MPa/mm, s2*=0.2 mm, s3= 4.8
mm and R=3.039. The comparison between the experimental bond-slip curve and the constitutive
law is shown in Fig. 14.
The influence of the bond-slip relationship on the response of the beam is first studied for the
concrete mix with volume fiber fraction of 0.5%. The analysis is carried out using bond-slip
constitutive law for pull-out of single fiber (case BS-P, see Table 1). The comparison with the
experimental results (see Fig. 15) shows that the peak resistance is well estimated, however, the
pre-peak stiffness is too low and the post peak response too brittle. There are two reasons for this,
first, the bond-slip relationship for a single fiber does not account for the interaction between
fibers (damage of concrete between fibers) and, second, in the critical section of the beam fibers
are loaded (pulled out) in directions that do not coincide with the their axis, i.e. there is also dowel
action, which leads to the increase of the stiffness of the effective bond-slip relationship. The
analysis clearly shows that the initial stiffness of the bond-slip relationship significantly influences
the response. Therefore, to more realistically estimate the effective bond-slip relationship, single
fiber tests, where the inclination of fibers is varied, should be carried out. Moreover, additional
numerical parametric study would be useful to investigate the effect in more detail.
To obtain better fit of the experimental results a parametric study is carried out. In the study the
simplified version of the bond-slip relationship (see Fig. 10a) is used in which, compared to the
single bond-slip relationship, stiffness k1, k2 and ksecant are increased and all equal to 1200 MPa/mm,
keeping all other parameters the same (case BS-1, see Table 1). With these parameters for bond-
slip relationship the analysis shows very good agreement with the experiments almost in the entire
displacement range, up to 5 mm. To further improve the response for large displacements, in the
next step slip parameter s3 is slightly decreased, from 4.8 mm to 4.0 mm (case BS-2). The resulting
curve obtained from the meso-scale model exhibit now nice agreement with the experimental
curve, not only for the pre-peak response and resistance but also for the entire post-peak response
in the range up to displacement of 5 mm. Finally, the effect of bond strength on the response of the
beam is investigated by varying peak resistance (12.0 MPa, case BS-1) in the range of 20 %
(cases BS-3 and BS-4). From Fig. 15 can be seen that this leads to the positive and negative shift,
respectively, of the post-peak response. Based on the comparison of the numerical and
experimental data it can be concluded that the best fit for Vf = 0.5 % is obtained for the bond-slip
relationship BS-2.
If one employs the bond-slip relationship obtained from the calibration of the model with
Vf = 0.5 % for the beams with higher content of fibers then the resulting resistance is overestimated
and the post-peak response is too ductile. This is illustrated in Fig. 16 which shows the comparison
between the corresponding experimental data and numerical simulation using bond slip
relationship for the case BS-2. The reason for this is obvious, namely, due to the higher content of
fibers the interaction between fibers over the surrounding concrete leads to the reduction of the
pull-out capacity of fibers. The concrete of the fracture process zone between fibers becomes more
damaged and consequently the pull-out capacity of fibers is lower.
The evaluation of numerical results shows that the effective bond-slip relationship should be
adapted to the volume content of fibers. This is obtained through the parametric study, similar as
carried out for the fiber volume content of 0.5 %. In the study the starting constitutive bond-slip
relationship is the case BS-2 (see Table 1). The peak resistance m and the limit slip s3 are varied
such that the analysis fit the mean load-displacement curve for the fiber volume content of 1.0 %
and 1.5 %, respectively. The optimal parameters of the bond-slip constitutive law for all three
volume contents are summarised in Table 2 and the corresponding bond-slip relationships are
plotted in Fig. 17.
The resulting load-displacement curves for the beams with volume fraction of fibers varied
from 0 % to 1.5 % are shown in Fig. 18. As can be seen for the entire range of the post peak
response the agreement between numerical and experimental curves is very good. Note that for the
case with 1% of fiber content the analysis was stopped at displacement of approximately 4 mm
due to numerical difficulties. It can be seen that for relative low volume fraction of fibers after the
peak resistance is reached the load-displacement curve exhibit a sudden drop and subsequent
recovery. This is due to the relative low volume content of fibers but is also dependent on the beam
size. In case of large beams this drop would probably be even more pronounced.
In all simulations the failure is due to the bending Mode-I failure type. For the lower content of
fibers damage tends to be more localized around the final discrete crack whereas for the cases with
higher content of fibers damage tends to be more distributed. Typical crack patters observed in the
experiments and analyses for Vf = 1.0 % and 1.5 %, respectively, are shown in Fig. 19. It can be
seen that concrete in the fracture process zone is more damaged if the content of fibers is higher
(compare Figs. 19c and 19d). This explains the reason for decrease of effective bond resistance of
fibers with increase of volume fractions of fibers.
In Table 3 are summarized the maximum and minimum stresses in fibers. It is evident that in no
case the stress in the fibers reaches yield stress fy=1210.5 MPa. Moreover, the maximum tensile
stresses in fibers descreases with increase of volume fraction of fibers. In the contrary, the
compressive stresses in fibers increases with increase of volume fraction of fibers. Obviously,
failure is due to the pull-out of fibers and not to the failure of steel. This suggests that for the
correct prediction of the failure of beams the pull-out constitutive law is relevant.
The importance of the bond-slip constitutive law is demonstrated in Fig. 20, which shows load-
displacement response of the beam (Vf = 0.5 %) assuming three different bond-slip relationships:
(1) bond-slip law the same as in the case BS-2 (see Table 1), (2) perfect bond (no slip) and no limit
strain of steel and (3) perfect bond and limit strain failure of steel equal to 5 %. From Fig. 20 can
be seen that perfect bond leads to high peak resistance and high ductility only if no failure of steel
fibers takes place. With accounting for the limit on the strain in steel (5 %) peak load also
increases, however, once the strength of steel fibers is reached there is relative brittle failure of the
beam.

4.4 Proposed bond-slip relationship

Based on the results of the numerical study the correlation between the maximum shear stress
τmax=τm+τf [MPa] and the fiber volume fraction Vf [%] is proposed as:

(4)

where τmax,s is the bond strength of a single fiber and A and B are constants obtained from the above
discussed fitting procedure. For the fibers investigated in the present study τmax,s = 12 MPa,
A = 4.05 [MPa] and B = 3.97 [MPa] with correlation coefficient equal 0.99.
The correlation between the slip s3 [mm] at which τm becomes zero and the fiber volume
fraction Vf [%] is proposed as follows:

(5)

in which s3,max is the maximum value of s3 (4.0 mm), which approximately corresponds to single
pull-out constitutive law, C (1.0 mm) and D (1.5 mm) are a constants obtained from the fitting
procedure with correlation coefficient 1.00. The relationships from (4) and (5) are plotted in
Fig. 21. They can be used for the estimation of the parameters τmax and s3 for any fiber volume
fraction value between 0.0 % and 1.5 %. Of course, these functions are valid only for the fiber type
used in the present study. Moreover, they could also be dependent on the concrete type.
The above equations principally cover the the range of the volume content of fibers from 0.5%
up to 1.5%, since for this range the study was carried out. However, for relatively low amount of
fibers (Vf < 0.5%) it is assumed that the distance between the fibers is such that the effective bond-
slip resistance is not affected, i.e. the maximum bond resistance is reached for Vf = 0.5%.

5. Conclusions

Based on the results of the study, the following conclusions can be drown out.

 The present numerical results show that meso scale approach based on the microplane
model for concrete and discrete bond-slip relationship for steel fibers is able to realistically
replicate experimental tests. As expected, with increase of the volume fraction of fibers the
resistance and especially ductility increases.
 The effective bond-slip relationship is calibrated based on the fit of the experimental results
for 3-point bending test data by the numerical results using bond-slip relation of a single fiber
as the starting relationship.
 It is shown that the effective bond-slip relationship cannot be obtained only from a single
fiber pull-out experiments, where the fibers were aligned with the loading direction. Single
fiber tests, where the inclination of fibers is varied, can lead to a better representation of
effective bond properties. From such a test only bond strength and maximum slip can be
realistically evaluated. However, the initial stiffness is underestimated, with a consequence that
the resulting pre-peak resistance is underestimated and the post-peak response is too brittle.
 With the increase of the volume content of fibers the bond strength and maximum slip
decrease. The likely reason is the interaction among the fibers when they come too close to
each other, i.e. the local damage of concrete around fibers leads to degradation of effective
bond capability.
 The evaluation of the numerical results shows that for the investigated beams the failure is
due to the failure of bond between concrete and fibers, i.e. fibers are pulled-out from the
concrete. In all cases the yield stress in steel is not reached.
  All beams fail in mode-I bending failure type. With increase of the volume content of fibers
local damage around discrete bending crack increases, which also leads to the increase of
ductility.
 The proposed meso scale model is shown to be powerful numerical tool able to realistically
predict behaviour of fiber reinforced concrete. In combination with experimental results it can
be effectively employed in design of new steel fiber reinforced concretes and structural
elements.

References

ACI Committee 544. (2001), "State-of-the Art Report on Fiber Reinforced Concrete", ACI Manual of
Concrete Practice, ACI 544.1R-96.
Bažant, Z.P. and Oh, B.H. (1983), "Crack band theory for fracture of concrete", RILEM, 93(16), 155-177.
Bažant, Z.P. and Prat, P.C. (1988), "Microplane model for brittle-plasic material-parts I and II", Journal of
Engineering Mechanics, ASCE, 114, 1672-1702.
Bolander, J.E., Choi, S. and Duddukuri, S.R. (2008), "Fracture of fiber-reinforced cement composites:
Effects of fiber distribution", International Journal of Fracture, 154, 73-86.
Bolander, J.E. and Saito, S. (1997), "Discrete modelling of short-fiber reinforcement in cementitious
composites", Advanced Cement Based Material, 6, 76-86.
Bolander, J.E. and Sukumar, N. (2005), "Irregular lattice model for quasistatic crack propagation", Physical
Review B, 71, 1-12.
EN 12390-3:2001 (2001), Testing hardened concrete - Part 3, Compressive strength of test specimens,
Bruxelles.
EN 12390-6:2000 (2000), Testing hardened concrete - Part 6, Tensile splitting strength of test specimens,
Bruxelles.
Caner, F.C., Bažant, Z.P. and Wendner, R. (2013), "Microplane model M7f for fiber reinforced concrete",
Engineering Fracture Mechanics, 105, 41-57.
Fischer, G., Stang, H. and Dick-Nielsen, L. (2007), "Initiation and development of cracking in ECC
materials: experimental observations and modelling", Fracture mechanics of concrete and concrete
structures FRAMCOS6, 1517-1522.
Han, T.-S., Feenstra, P.-H. and Billington, S.-L. (2003), "Simulation of highly ductile fiber-reinforced
cement-based composite components under cyclic loading", ACI Structural Journal, 100(6), 749-757.
Jun, P. and Mechtcherine, V. (2010), "Behaviour of strain-hardening cement-based composites (SHCC)
under monotonic and cyclic tensile loading: Part 2 - Modelling", Cement & Concrete Composites, 32,
810-818.
Kang, J., Kim, K., Lim, Y.M. and Bolander, J.E. (2014), "Modeling of fiber-reinforced cement composites:
Discrete representation of fiber pullout", International Journal of Solids and Structures, 51, 1970-1979.
Kunieda, M., Ogura, H., Ueda, N. and Nakamura, H. (2011), "Tensile fracture process of Strain Hardening
Cementitious Composites by means of three-dimensional meso-scale analysis", Cement & Concrete
Composites, 33, 956-965.
Maalej, M. (2001), "Tensile properties of short fiber composites with fiber strength distribution", Journal of
Materials Science, 36, 2203-2212.
Mohr, O. (1900), "Welche Umstände bedingen die Elastizitätsgrenze und den Bruch eines Materiales?",
Zeitschrift des Vereins Deutscher Ingenieure, 46, 1524-1530, 1572-1577.
Ogura, H., Kunieda, M. and Nakamura, H. (2013), "Meso-Scale Analysis Considering Effect of Fiber
Inclination in Fiber Reinforced Cementitious Composites", RILEM Bookseries, 8, 261-272.
Ožbolt, J. (1998), "MASA – MAcroscopic Space Analysis, Internal Report", Institut für Werkstoffe im
Bauwesen, Universität Stuttgart.
Ožbolt, J., Meštrović, D., Li, Y.J. and Eligehausen, R. (2000), "Compression failure –beams made of
different concrete types and sizes", Journal of Structural Engeneering, ASCE, 126, 200-209.
Ožbolt, J., Li, Y.J. and Kožar, I. (2001), "Microplane Model for Concrete with Relaxed Kinematic
Constraint", International Journal of Solids and Structures, 38, 2683-2711.
Ožbolt, J., Lettow, S. and Kožar, I. (2002) "Discrete Bond Element for 3D Finite Element Analysis of
Reinforced Concrete Structures". In Balázs-Bartos-Cairns-Borosnyói (eds), Proceedings of the 3 rd
International Symposium: Bond in Concrete - from research to standards, Budapest, University of
Technology and Economics.
Ožbolt, J. and Reinhardt, H.W. (2002), "Numerical study of mixed mode fracture in concrete", International
Journal of Fracture, 118, 145-161.
Schauffert, E.A. and Cusatis, G. (2012), "Lattice Discrete Particle Model for Fiber-Reinforced Concrete. I:
Theory", Journal of Engineering Mechanics, 138, 826-833.
Schauffert, E.A., Cusatis, G., Pelessone, D., O’Daniel, J.L. and Baylot J.T. (2012), "Lattice Discrete Particle
Model for Fiber-Reinforced Concrete. II: Tensile Fracture and Multiaxial Loading Behavior", Journal of
Engineering Mechanics, 138, 834-841.
Suwada, H. and Fukuyama, H. (2006), "Nonlinear finite element analysis on shear failure of structural
elements using high performance fiber reinforced cement composite", Journal of Advanced Concrete
Technology, 4(1), 45-57.
Taylor, G.I. (1938), "Plastic strains in metals", Journal of the Institute of Metals, 62, 307-324.
Vandewalle, L. (2002), "RILEM TC162-TDF Test and desig methods for steel fiber reinforced concrete:
Bending test final Recommendation", Materials and Structures, 35, 579-582.

Figures and Tables

Fig. 1 Hook-ended steel fibers Dramix RC 65/35 BN

Fig. 2 Single fiber pull-out test

 200 1
2
Pullout force [N]

150 3
Mean (n=3)
100

50

0
0 1 2 3 4 5 6 7 8 9 10 11 12
Displacement [mm]
Fig. 3 Single fiber pull-out force vs. slip curves

Fig. 4 Pre-notched beam geometry and 3-point bending test set-up (Vandewalle 2002)
Fig. 5 Displacement controlled flexural 3-point bending test set-up (Vf=1.5 %)
 40 40
0.0_1 0.5_1
0.0_2 0.5_2
30 30 0.5_3
0.0_3
Force [kN]

0.5 (n=3)

Force [kN]
0.0 (n=3) 0.0 (n=3)
20 20

10 10

0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.0 1.0 2.0 3.0 4.0 5.0
Displacement [mm] Displacement [mm]
(a) Vf=0 % (b) Vf=0.5 %
40 40
1.0_1
1.0_2
30 1.0_3 30
1.0 (n=3)
Force [kN]

Force [kN]

0.0 (n=3)
20 20 1.5_1
1.5_2
10 10 1.5_3
1.5 (n=3)
0.0 (n=3)
0 0
0.0 1.0 2.0 3.0 4.0 5.0 0.0 1.0 2.0 3.0 4.0 5.0
Displacement [mm] Displacement [mm]
(c) Vf=1.0 % (d) Vf=1.5 %

Fig. 6 Experimentally measured load - displacement curves for beans with different fiber volume fractions
 z
z 
microplane microplane
integration point
T D y
K
n
V
M

x y
microplane
FE integration
point x
Fig. 7 Decomposition of the macroscopic strain vector into microplane strain components – normal
(volumetric and deviatoric) and shear

Fig. 8 Constitutive law of steel fiber

 Fig. 9 Discrete bond-slip relationship

(a) τ

τm
τmax

k1 τf
s3 s
(b)
s3
35 mm

(c)
4 mm

Fig. 10 Finite element idealization of fibers: (a) bond-slip relationship, (b) four truss finite elements and (c)
actual fiber shape
 (a) (b)

(c) (d)

(e)

Fig. 11 Finite element discretization of concrete and fibers for fiber volume content of 0.5 % and 1.0 %: (a)
Vf=0.5 %, entire beam view, (b) Vf=1.0 %, entire beam view, (c) Vf=0.5 %, left half of the beam (d)
Vf=0.5 %, isometric view of the left half of the beam and (e) detail of fiber-concrete of connection
 Strain [‰] 6
-50 -40 -30 -20 -10 0 5

Stress [MPa]
0 4

-20 3

Stress [MPa]
2
-40 1

-60 0
0 1 2 3 4 5 6 7 8 9 10
-80 Strain [‰]

(a) (b)
Fig. 12 Stress-strain curves for concrete: (a) uniaxial compression and (b) uniaxial tension

30
Experiment

20
Numerical Vf=0.0 % ( 0.5 %)
Force [kN]

Numerical Vf=0.0 % (1.0 %)

10

0
0.0 0.1 0.2 0.3 0.4 0.5
Displacement [mm]

Fig. 13 Load-displacement curves for plain concrete beam, numerical and experimental results
 14

12 Experiment
10
τ [MPa]

8 Analysis

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

Fig. 14 Experimental single fiber bond - slip relationship (τ-s) and fit obtained from numerical analysis

40
Experiment BS-P

BS-1 BS-2
30
BS-3 BS-4
Force [kN]

20

10

0
0.0 1.0 2.0 3.0 4.0 5.0
Displacement [mm]

Fig. 15 Comparison between experimentally and numerically obtained load – mid span displacement curves
(Vf=0.5%) for different bond-slip relationships
 40

30
Force [kN]

20

Experiment
10
BS-2

0
0.0 1.0 2.0 3.0 4.0 5.0
Displacement [mm]

Fig. 16 Comparison between experimentally and numerically obtained load – mid span displacement curves
(Vf=1.0%)

14
0.5 %
12

10 1.0 %
τ [MPa]

8
1.5 %
6

0
0 1 2 3 4 5 6 7
s [mm]

Fig. 17 Fiber bond - slip relationship (τ-s) for fiber reinforced concrete beams with different fiber volume
fractions
 40 40
Experiment Experiment
30 Numerical Vf=0.0 % ( 0.5 %) 30 Numerical

Force [kN]
Force [kN]

Numerical Vf=0.0 % (1.0 %)

20 20

10 10

0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.0 1.0 2.0 3.0 4.0 5.0
Displacement [mm] Displacement [mm]
(a) Vf=0 % (b) Vf=0.5 %
40 40
Experiment Experiment
30 Numerical 30 Numerical
Force [kN]

Force [kN]

20 20

10 10

0 0
0.0 1.0 2.0 3.0 4.0 5.0 0.0 1.0 2.0 3.0 4.0 5.0
Displacement [mm] Displacement [mm]
(c) Vf=1.0 % (d) Vf=1.5 %

Fig. 18 Comparison between experimentally and numerically obtained load – mid span deflection curves
 (a) (b)

(c) (d)

(e) (f)

Fig. 19 Cracks (maximum principal strain) at peak load: (a) Vf=1.0% and (b) Vf=1.5% and in the last
increment: (c) Vf=1.0% and (d) Vf=1.5%; Experimental failure mode for: (e) Vf=1.0% and (f) Vf=1.5%
 40
BS-2
Perfect bond - steel with no limit strain
30 Perfect bond - steel with limit strain
Force [kN]

20

10

0
0.0 1.0 2.0 3.0 4.0 5.0
Displacement [mm]

Fig. 20 Load-displacement response of the beam (Vf = 0.5%) assuming three different bond-slip
relationships

15 5
τmax [MPa]

4
s3 [mm]

10 3
5 2
1
0 0
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5
Vf [%] Vf [%]
(a) (b)

Fig. 21 The correlation between (a) maximum bond strength (τmax) and volume fraction of fibers (Vf) and (b)
limit slip (s3) and volume fraction of fibers (Vf)
Table 1 Discrete bond model parameters for Vf = 0.5%
Case Model parameters
τm τf ksecant k1 k2 s2* s3 R
[MPa] [MPa] [MPa/mm] [MPa/mm] [MPa/mm] [mm] [mm]
BS-P 10.17 1.83 17.13 41.81 2.71 0.20 4.80 3.04
BS-1 10.17 1.83 1200 1200 1200 0.20 4.80 -
BS-2 10.17 1.83 1200 1200 1200 0.20 4.00 -
BS-3 12.55 1.83 1200 1200 1200 0.20 4.80 -
BS-4 7.76 1.83 1200 1200 1200 0.20 4.80 -

Table 2 Discrete bond model parameters for all models

Volume Model parameters
content of τm τf ksecant k1 s2* s3
fibers [MPa] [MPa] [MPa/mm] [MPa/mm] [mm] [mm]
0.5 10.17 1.83 1200 1200 0.20 4.00
1.0 6.17 1.83 1200 1200 0.20 2.50
1.5 4.89 1.83 1200 1200 0.20 2.00

Table 3 Maximum and minimum stresses in fibers

Model maximum stress in fibers [MPa] minimum stress in fibers [MPa]
Vf [%]
0.5 1154 -359
1.0 771 -741
1.5 648 -632