Computers and Concrete, Vol. 8, No.

1 (2011) 1-22 1
DOI: http://dx.doi.org/10.12989/cac.2011.8.1.001
Technical Note

A numerical tension-stiffening model for ultra high

strength fiber-reinforced concrete beams
Chaekuk Na and Hyo-Gyoung Kwak *
Department of Civil and Environmental Engineering, Korea Advanced Institute of Science and
Technology, 335 Gwahak-ro, Yuseong-gu, Daejeon, Korea, 305-701
(Received July 30, 2009, Accepted December 3, 2009)

Abstract. A numerical model that can simulate the nonlinear behavior of ultra high strength fiber-
reinforced concrete (UHSFRC) structures subject to monotonic loadings is introduced. Since engineering
material properties of UHSFRC are remarkably different from those of normal strength concrete and
engineered cementitious composite, classification of the mechanical characteristics related to the biaxial
behavior of UHSFRC, from the designation of the basic material properties such as the uniaxial stress-
strain relationship of UHSFRC to consideration of the bond stress-slip between the reinforcement and
surrounding concrete with fiber, is conducted in this paper in order to make possible accurate simulation
of the cracking behavior in UHSFRC structures. Based on the concept of the equivalent uniaxial strain,
constitutive relationships of UHSFRC are presented in the axes of orthotropy which coincide with the
principal axes of the total strain and rotate according to the loading history. This paper introduces a
criterion to simulate the tension-stiffening effect on the basis of the force equilibriums, compatibility
conditions, and bond stress-slip relationship in an idealized axial member and its efficiency is validated by
comparison with available experimental data. Finally, the applicability of the proposed numerical model is
established through correlation studies between analytical and experimental results for idealized UHSFRC
beams.
Keywords: ultra high performance concrete (UHPC); steel fiber-reinforced concrete (SFRC); tension-
stiffening model; tensile properties; finite element analysis.

1. Introduction
Rapid increase in the number of construction projects involving long-span bridges and high-rise
buildings has necessitated the development of construction materials possessing increased strength.
Concrete, which has become one of the most important construction materials and is widely used in
many types of engineering structures, is not an exception. Following the introduction of normal
strength concrete, its strength has been continuously increased; however, the brittleness of concrete
is proportion to the increase in compressive strength. To overcome the brittleness of concrete,
engineered cementitious composite (ECC) has been developed by adding various types of fibers at
different volume fractions to the concrete matrix (Li 1998). ECC’s tensile strain hardening behavior
has a capacity in the range of 3% to 7%; however, the compressive strength has not been increased
compared to normal concrete. In an effort to improve the strength and toughness of concrete, ultra
high strength concrete with steel fibers added in the concrete matrix, namely, ultra high strength
* Corresponding author, Professor, E-mail: khg@kaist.ac.kr
2 Chaekuk Na and Hyo-Gyoung Kwak

fiber-reinforced concrete (UHSFRC), has been developed by many researchers.

In order to use UHSFRC as a construction material, however, the structural behavior of UHSFRC
members as well as the material properties of UHSFRC itself must be verified. In this regard,
numerous relevant experimental studies have been conducted (Kölle et al. 2004, Mansur et al.
1999). Within the framework of developing advanced design and analysis methods for UHSFRC
structures, experimental study is required, as experiments provide a firm basis for design equations
and also supply basic information for numerical analyses, such as material properties. In addition,
the results of numerical analyses should be evaluated through comparison with results of experiment
involving full-scale models of the structural sub-assemblages or the entire structures.
The development of reliable analytical models can reduce the number of required test specimens
for the solution of a given problem (Kwak and Kim 2001, Kwak and Na 2007). This is of notable
importance given that tests are time-consuming and costly and often do not simulate exactly the
loading and support conditions of the actual structure. Nevertheless, very little work has been
carried out on the structural behavior of UHSFRC systems on the basis of finite element analyses
(Foster et al. 2006), because of the computational effort involved and insufficient knowledge of the
material behavior of UHSFRC under biaxial stress state. With the recognition that many of the
material models for biaxial loading have yet to be fully verified so far (Demeke and Tegos 1994,
Hussein and Marzouk 2000, Kölle et al. 2004), one of the aims of this paper is to address some
model selection issues in the numerical analyses of UHSFRC structures, in particular, with regard to
the strength of the reinforcing steel and the tension-stiffening effect in concrete.
This paper introduces an improved numerical tension-stiffening model of UHSFRC members on
the basis of the force equilibriums, compatibility conditions, and simplified bond stress-slip
relationship. The introduced model is idealized with four boundary values corresponding to the
stabilized crack strain, the steel and fiber yielding strains, and the pull-out strain, respectively. The
use of the tension-stiffening model makes it possible to accurately simulate the post-cracking
behavior of UHSFRC members dominantly affected by the tension-stiffening effect rather than
normal strength concrete members. Finally, the introduced numerical model is validated by
comparison with test results for idealized UHSFRC beams, and additional parametric studies are
conducted to review the structural behavior of UHSFRC beams according to the tension-stiffening
effect and the change in material properties.

2. Material properties
2.1 Concrete
The uniaxial stress-strain curves of normal strength concrete have been proposed on the basis of
experimental studies of a numerous idealized relationships. Likewise, in the cases of steel fiber-
reinforced concrete, experimental studies have been conducted to describe the stress-strain relationship of
corresponding concrete (Hussein and Marzouk 2000, Kölle et al. 2004, Mansur et al. 1999). A
typical normalized compressive stress-strain relationship of SFRC is shown in Fig. 1. An increase of
the compressive strength accompanies a rapid decrease of ductility in the strain softening region;
however, this brittleness has been overcome by adding fibers to the concrete, as this increases both
the ductility of concrete and its fatigue strength (Ezeldin and Balaguru 1992). In general,
improvement of the ductility of concrete has made it possible to develop higher strength concrete
 A numerical tension-stiffening model for ultra high strength fiber-reinforced concrete beams 3

Fig. 1 Normalized compressive stress-strain relationship of UHSFRC

such as UHSFRC and has facilitated practical application of UHSFRC structures.

In describing the uniaxial compressive stress-strain behavior of UHSFRC, more attention must be
given to the strain softening region. Many empirical equations of NSC were proposed to define the
stress-strain relationship, and each general relationship was agreed well with the experimental test
(Hognestad 1951); since the principal variable was the compressive strength of concrete specimen.
However, unlike NSC, the compressive behavior of UHSFRC, which depends to fiber content and
specimen dimension, is difficult to choose a specific general stress-strain relationship. Upon this
background, because test data are not sufficient to suggest a unique compressive stress-strain
relationship, this paper introduces the stress-strain relationship given in Eq. (1), wherein the
ascending branch represents a similar equation to that popularly used for NSC (Hognestad 1951)
and the descending branch is a regression equation determined on the basis of experimental data
(Mansur et al. 1999). In Fig. 1, the stress-strain relationship of NSC with fiber (Ezeldin and
Balaguru 1992) and HSC without fiber (Attard and Setunge 1996) are too ductile and brittle,
respectively, compare to that of ultra high strength concrete with fiber (Mansur et al. 1999).
k1
fc ′ { 1 – ( 1 – ε ⁄ ε c ) 0 ≤ ε ≤ εc
⎧
⎪ }

f =⎨ (1)
⎪
– ε ⁄ εc – k2
⎪
⎧e
fc′⎨ ------------------
- + k3 ⎫⎬ ε ≥ εc
⎪
⎩
k2 ⎭
⎩

where fc′ is the compressive strength of UHSFRC, εc is the peak strain corresponding to fc′ , and
Ec = 3.840 fc′ represents the initial modulus of elasticity introduced by Graybeal (2007) for all
range of concrete strength. In general, the shape function k1 = ln (1 – η) ⁄ ln{1 – η ⁄ (Ecεc ⁄ fc′)} can be
determined by assuming the secant ratio η = f ⁄ fc ′ , and the material parameters of k2 = 0.65 ,
k3 = 0.15 , required to define the descending branch, can be determined through the correlation
between the regressive curve and the experimental data of Mansur et al. (1999).
With a uniaxial stress-strain relationship of concrete, the material behavior of concrete under
4 Chaekuk Na and Hyo-Gyoung Kwak

Fig. 2 Biaxial strength failure envelope of UHSFRC

biaxial loading needs to be defined, because the strength characteristic and stress-strain behavior of
concrete are somewhat different from those of concrete under uniaxial loading due to the effects of
Poisson’s ratio and micro crack confinement (Kwak and Na 2007). To simulate the change of
material properties according to the stress state, it is thus necessary to define the biaxial strength
envelope. As conducted for normal strength concrete, corresponding experimental studies have also
been performed for UHSFRC (Demeke and Tegos 1994, Hussein and Marzouk 2000, Kölle et al.
2004). Fig. 2 shows the biaxial strength failure envelope of UHSFRC introduced in this paper along
with that of NSC for comparison.
The accompanying equation for the failure envelope has been designed on the basis of Kölle's
experimental data for the compression-compression region and Demeke's and Hussein's experimental data
for the compression-tension and the tension-tension region, respectively. Some experimental results
show that the compressive-compressive behavior of UHSFRC always depend on the fiber content;
however, some experimental results show the opposite behavior. Differently from both equations
defined in the tension-tension region and the compression-tension region, which are similar to those
introduced by Kupfer's experimental data for NSC (Kupfer et al. 1969), the equation for the failure
envelope in the compression-compression region is expressed by
f2p = (0.25ξ3 – 1.25ξ2 + 1.25ξ + 1 )fc′, f1p = ξf2p (2)
where ξ = f1 ⁄ f2 is the principal stress ratio, and f p and f p are the maximum equivalent principal
1 2

stresses corresponding to the current principal stresses f and f , respectively.

1 2

After determination of the equivalent concrete compressive strength of f p and f p from the biaxial
1 2

failure surface of UHSFRC, the equivalent uniaxial stress-strain relationship in the compression-
compression region, corresponding to the current loading history, is constructed by replacing the
compressive strength fc′ in Eq. (1) with the equivalent compressive strength fip. In the compression-
tension and the tension-tension regions, however, the following assumptions are adopted in this
paper, because the response of a typical UHSFRC member is considerably more affected by the
tensile than the compressive behavior of concrete: (1) failure takes place by cracking when the
principal tensile strain exceeds the limit strain; therefore, the tensile behavior of the concrete
 A numerical tension-stiffening model for ultra high strength fiber-reinforced concrete beams 5

Fig. 3 Stress-strain relationship of steel

dominates the response; (2) the uniaxial tensile strength of concrete ft is reduced to the value feq, as
shown in Fig. 2, to account for the effect of the compressive stress under a biaxial state; and (3) the
concrete stress-strain relationship in compression is the same as that under uniaxial loading and
does not change with an increase of the principal tensile stress.
2.2 Steel
The stress-strain curves for steel are generally assumed to be identical in tension and compression.
For simplicity in the calculations, it is necessary to idealize a one-dimensional stress-strain curve for
the steel element. Normal strength steel is usually assumed to be a linear elastic, linear strain
hardening material whose yield stress is fy (bared steel bar in Fig. 3). As noted in previous studies,
however, normal strength steel embedded in a concrete matrix presents different behavior from
bared steel bar, because of the bond interaction along the steel bar between adjacent cracks. This
means that the averaged yield stress fn, which is significantly less than fy, must be used to avoid
overestimation of the post-yielding behavior of the reinforced concrete structures in the case of
taking the tension-stiffening effect into consideration in the stress-strain relationship of concrete.
More details related to the calculation of fn can be found elsewhere (Kwak and Kim 2004).

3. Tension-stiffening model for UHSFRC

When a symmetrical uncracked RC member is loaded in tension, the tensile force is distributed
between the reinforcing steel and the concrete in proportion to their respective stiffness, and cracks
in the concrete occur when the stress reaches a value corresponding to the tensile strength of
concrete. In a cracked cross-section, all tensile forces are balanced by the steel encased in the
concrete matrix only. However, between adjacent cracks, tensile forces are transmitted from the
steel to the surrounding concrete by bond forces. This phenomenon is defined as the tension-
stiffening effect, and the same response also appears in UHSFRC members.
This tension-stiffening effect can be adequately taken into account by the increased average
stiffness of an element. An increase of tensile stiffness of concrete can be accomplished by using
6 Chaekuk Na and Hyo-Gyoung Kwak

Fig. 4 Descriptions for a cracked UHSFRC axial member

a stress-strain relationship that includes a descending branch in the tension region. In this paper,
based on the force equilibriums, compatibility conditions, and bond stress-slip relationship
between the reinforcement and the surrounding concrete in an axial tension member, a
descending branch to define the post-cracking stress-strain relationship of concrete is proposed.
Additionally, the bonding resistance between the steel fibers and the concrete matrix in UHSFRC
is also taken into consideration, because steel fiber-reinforced concrete can exhibit significant
post-cracking tensile resistance at cracks, depending on the type and dosage of the steel fiber
used.
3.1 Force equilibrium
As was verified from experiments (Bischoff 2003, Kölle et al. 2004), the use of steel fibers can
improve the bond and reduce crack spacing. A cracked axial UHSFRC member subject to a direct
tensile force T is shown in Fig. 4(a). A part of the member bounded by adjacent cracks with a crack
spacing of 2a can be taken as the free body diagram.
Since the applied direct tensile force T is carried partly by the concrete matrix (Fc) and partly by
the reinforcing steel (Fs), the following force equilibrium equation can be obtained (see Fig. 4(b))
T = Fc + Fs (3)
The two force components carried by the concrete matrix and the reinforcing steel can be expressed by
du
Fc = AcEcεc = AcEc -------c (4a)
dz
du
Fs = AsEsεs = AsEs -------s (4b)
dz
 A numerical tension-stiffening model for ultra high strength fiber-reinforced concrete beams 7

where E, ε, u and A are the elastic modulus, the strain, the deformation, and the sectional area
corresponding to each material of the concrete matrix (subscript c) and the reinforcing steel
(subscript s), respectively.
Reinforcing bars transfer tensile stresses to the concrete matrix through the bond stresses along the
surface between the reinforcements and the surrounding concrete. Therefore, an infinitesimal element
of the length dz is taken out from the intact concrete between adjacent cracks to obtain the equilibrium
equations for the concrete matrix and the reinforcing steel. Fig. 4(c) presents the free body diagram at
the steel and concrete interface. The following equilibrium equations of force deviation for the steel
and concrete, which are expressed in terms of the bond parameters, can be obtained
dFc
-------
- = –pmfb (5a)
dz
dFs
-------
- = +pmfb (5b)
dz
where p is the perimeter of the reinforcing bar, m is the number of reinforcing bars placed, and fb is
the bond stress at the steel-concrete interface.
3.2 Bond-slip behavior
Since the bond slip ∆ at the steel-concrete interface is defined by the relative displacement
between the reinforcing steel and concrete matrix ( ∆ = us – uc ), substitution of Eqs. (4) and (5) into
the second order differential equation of the bond slip leads to Eq. (6), if the linear bond stress-slip
relationship given by fb = Eb∆ is assumed
d--------
2
∆
– k2 ∆ = 0 (6)
dz 2

where k = (pmEb/AsEs) · (1+nρ), Eb= the slip modulus, n = Es ⁄ Ec , and ρ = As ⁄ Ac . In particular, nρ

2

means the area parameter and k− represents the characteristic length (Gupta and Maestrini 1990).
1

The general solution to Eq. (6) given by ∆ = C1 sinhkz can be solved from the boundary
conditions: (1) the slip should be zero at the center (z = 0) between crack faces, and (2) the slips at
both crack faces must be the same because of the symmetry ( ∆(–z) = –∆(z) ). Integration of Eq. (5)
after substituting the obtained general solution leads to the following expression for the steel force FS
pmEbC1
Fs = ------------------
-coshkz + C3 (7)
k
where the constant of integration C3 = (T – Ff ) – (pmEbC1 ⁄ k) is obtained from the boundary
condition at the crack surface ( Fs = T – Ff at z = a ), because the steel fiber of the concrete matrix
resists the tensile force (Ff) at the crack face, as shown in Fig. 5. The general solutions for the
concrete and steel forces can be written as
pmEbC1
Fc = Ff + ------------------
- (coshka – coshkz) (8a)
k
pmEbC1
Fs = ( T – Ff) – ------------------
-(coshka – coshkz) (8b)
k
8 Chaekuk Na and Hyo-Gyoung Kwak

Fig. 5 Cracks and forces distribution between adjacent cracks

where Ff denotes the tensile force carried by the steel fiber in UHSFRC at the crack face.
If the resisting force by the steel fiber at the crack face (Ff) is assumed to be proportional to the
concrete force Fc at the center (z = 0), that is, Ff = αFc(0) (Bischoff 2003), then the following
relationship for Ff can be obtained, where α is a coefficient related to the steel fiber and concrete
properties and can be determined from the force equilibrium. Noted that α must be larger than or
equal to zero ( α ≥ 0 ).
α pmEb C1
Ff = ----------
- ⋅ -------------------(coshka – 1) (9)
1–α k
In advance, the displacements of concrete and steel along the reinforcement can be calculated
from Eqs. (4), (8) and (9) through integration with respect to the principal direction, z.
c1 ⎧ kz
uc = 1 + nρ ⋅ -------------
- -----------(coshka – α) – sinhkz ⎬ (10a)
⎫

1 + nρ ⎨⎩ 1 – α ⎭
 A numerical tension-stiffening model for ultra high strength fiber-reinforced concrete beams 9

c1 ⎧ kz
Tz - – -------------
us = --------- -⎨ -----------(coshka – α) – sinhkz ⎬ (10b)
⎫

As Es 1 + nρ ⎩ 1 – α ⎭

in which C1 = (T ⁄ AsEs )(1 – α) ⁄ {k(coshka–α )} is uniquely determined from the two relationships of
∆ = u s – u c and ∆ = C 1 sinh kz .

3.3 Tension-stiffening model

On the basis of the obtained equations for the displacements, the descending branch in the tension
region of the concrete stress-strain relationship can be determined. To take into account the tension-
stiffening behavior, the average behavior of UHSFRC needs to be defined with the effective tensile
stress (σcm) and the corresponding average strain ( εsm = εcm ) in the concrete matrix. First, the
equilibrium equation in Eq. (3) can be rearranged as
T = Acσcm + AsEsεsm (11)
These two values of σcm and εsm(= εcm) can be obtained from Eq. (10).
us ( a ) T 1 ⎧ ( 1 – α ) sinh ka ⎫
εsm = ------------ = ---------- ⋅ -------------- ⎨ nρ + ---------------------------------- ⎬ (12a)
a As Es 1 + n ρ ⎩ ka( coshka–α) ⎭

T 1 1 – α)sinhka ⎫
σcm = ----- ⋅ -------------- ⎨ 1 – ---------------------------------- ⎬
⎧ (
(12b)
As 1 + nρ ⎩ ka(coshka–α ) ⎭
The maximum tensile force in the concrete matrix occurs at z = 0, and the corresponding concrete
stress is σc max = Fc max ⁄ Ac ; that is, the maximum tensile stress in the concrete is directly
, ,

proportional to the applied principal tensile force T. Accordingly, σc max converges to the tensile ,

strength of the concrete (ft) as the applied tensile force T increases. At that point, a new crack will
be formed at z = 0, and the corresponding crack strain will be εcrack = σc max ⁄ Ec . ,

T 1 1–α
σc max = ----- ⋅ -------------- ⎛ 1 – -------------------------⎞ (13a)
As 1 + nρ coshka – α⎠
,
⎝

T 1 ⎧ 1–α ⎫
εcrack = ----------- ⋅ -------------- ⎨ 1 – ------------------------ ⎬ (13b)
A c Ec 1 + n ρ ⎩ cosh ka–α ⎭
After eliminates T from Eqs. (12) and (13), they are rewritten in a non-dimensional form.
( 1 – α ) sinh ka
1 – --------------------------------- -
σcm ka ( cosh ka – α )
------------- = -----------------------------------------
- (14a)
σc max ,
1 – α
1 – -------------------------
coshka – α
( 1 – α ) sinh ka
1 + ----------------------------------------- -
εcm n ρ ka ( cosh ka – α )
----------- = -------------------------------------------------- (14b)
εcrack 1–α -
1 – ------------------------
coshka – α
10 Chaekuk Na and Hyo-Gyoung Kwak

In Eq. (14), the crack spacing 2a is the same as the specimen length (L), and becomes
progressively shorter, L/2, L/4, ... and so on. Finally, σcm/σc,max converges to the value of 2/3 as the
parameter ka related to the crack spacing approaches zero. However, the actual crack spacing is not
narrowed any further but remains constant after reaching a certain value. The experimental study
indicated that the number of cracks is stabilized when the average strain is about 0.001 in an axial
normal strength reinforced concrete (NSRC) member (Rizkalla and Hwang 1984). An experimental
study for a UHSFRC member was also conducted and it was reported that the average crack width
(wm) and the average crack spacing (sm) are roughly half those obtained in NSRC members (Lorrain
et al. 1998). This means that the stabilized strain of the UHSFRC member converges to 0.001, as is
the case for the NSRC member, as given by the relationship wm = sm × εm , representing the average
crack width = the average crack spacing × the average strain. Accordingly, with the assumption that
the linear bond stress-slip relationship holds, Eq. (14) can be available up to εcm = 0.001 (point A
in Fig. 6).
Further deformation leads to yielding of the reinforcing steel, followed by an increase of the slip
while maintaining a plateau fb = τb . For continued increase of the slip, the bond stress decreases
linearly to the value of the ultimate frictional bond resistance. In the case of constant bond stress
and yielding of the reinforcing steel, the tensile force carried by the concrete matrix and reinforcing
steel can be calculated from Eq. (5) with the appropriate boundary conditions of the steel fiber force
Ff at the crack face, given in Fig. 5(c).
Fc = Ff + pmτb( a – z) (15a)
Fc = (T – Ff ) – pmτb( a – z) (15b)
The assumed force component for the steel fiber ( Ff = αFc(0) ) makes it possible to calculate the
concrete and the steel forces. From Eqs. (4) and (15), the displacements of the concrete and the
steel can be expressed as follows
pmτbz ⎛ ----------
a - – 1--- z⎞
uc = nρ -------------- (16a)
As Es 1 – α 2 ⎠
⎝

pmτbz ⎛ ----------
Tz - – -------------- a - – 1---z⎞
us = --------- (16b)
As Es As Es ⎝ 1 – α 2 ⎠
Furthermore, the average strain in the reinforcing steel εsm can be obtained by differentiating us
with respect to z at the crack face. Accordingly, the effective tensile stress of concrete can be
calculated from Eq. (11) with the average strain in the reinforcing steel εsm .
us(a ) T ⎧ pmτba 1 + α ⎫
εsm = ------------ = ---------- ⎨ T – --------------
- ⋅ ----------- (17a)
a As Es ⎩ 2 1 – α ⎬⎭
1 pmτba- ⋅ 1----------
σcm = --- ⋅ --------------
+ α- (17b)
A 2 1–α
The maximum tensile force in the concrete matrix occurs at z = 0 (Fc,max) and can be obtained
from Eq. (15) as:
 A numerical tension-stiffening model for ultra high strength fiber-reinforced concrete beams 11

pmτba 2
Fc max = --------------
- ⋅ ----------- (18)
,
2 1–α
The non-dimensional parametric equations, as applied in the case of linear bond stress-slip
relationship, are determined as follows (point B in Fig. 6)
- = 1----------
+ α-
σcm
------------ (19a)
σc, max 2
εcm
----------- = ---------- T - – 1--- ⋅ 1----------
1 – α-⎛ -------------- + α-⎞ (19b)
εcrack nρ ⎝ pm τ b a 2 1 – α⎠
where the corresponding concrete stress and strain are σc max = Fc max ⁄ Ac and εcrack = σc max ⁄ Ec ,
, , ,

respectively, and the principal tensile force T is given by T = As fy + αAc ft at the crack face. The coefficient
α = 0 corresponding to normal strength concrete without steel fibers indicates σc ⁄ σc, max = 1 ⁄ 2 , which
agrees with the previous studies (Gupta and Maestrini 1990, Kwak and Kim 2004).
The coefficient α is determined from Eq. (20), representing the force equilibrium in an axial
member composed of the ultra high strength concrete matrix and reinforcing steel at two typical
sections located at the center of the specimen ( z = 0 ) and at the crack face ( z = a ), respectively.
T = Fs ( 0 ) + Fc ( 0 ) = F s ( a ) + Fc ( a ) (20)
where Fc(0) = ftAc , Fs(0) = T – pmτba ⁄ (1 – α ) are the concrete and steel forces at the center,
respectively, and Fc(a) = αftAc is the steel fiber force at the crack face, which is proportional to the
concrete force at the center, Fs(a) = T – pmτba ⋅ α ⁄ (1 – α ) is the steel force at the crack face when
the reinforcing steel yields. Hence, the coefficient α can be expressed in terms of the concrete and
bond parameters.
pmτba
α = 1 – --------------
-≥0 (21)
ft Ac

Fig. 6 Effective concrete tensile stress-strain relationship

12 Chaekuk Na and Hyo-Gyoung Kwak

Accordingly, the tensile force carried by the steel fiber at the crack face, which is assumed to be
proportional to the concrete force at center ( Ff = αFc(0) ), remains constant even after yielding of
the reinforcing steel, while the average steel strain is rapidly enlarged. An increase in the average
steel strain at the post-yielding stage will continue until the elongation in the steel fiber
( ∆lf = (fy f ⁄ Ef) ⋅ lf ) equates with the average crack width (wm), where lf and fy,f represent the length
,

and its yielding strength of the steel fiber, respectively. Hence, from the relationship wm = sm × εm
(Lorrain et al. 1998), the average tensile strain for this critical condition, in which the normalized
concrete stress σcm ⁄ σc max = (1 + α) ⁄ 2 maintains a uniform value without any change from that
,

defined by point B in Fig. 6, can be determined (point C in Fig. 6).

Finally, the limit average strain εlim in the reinforcing steel needs to be defined. SETRA (2002)
indicates that the effective tensile stress of concrete disappears when the crack width ( wm = εlim ⋅ lc )
reaches 1/4 of the length of the steel fiber. Therefore, the limit average strain (εlim) in the reinforcing
steel can be expressed by the length of steel fiber (lf) and the characteristic length (lc), where
lf = 2h ⁄ 3 for a rectangular section with h = height of the section (point D in Fig. 6).
εlim = lf ⁄ 4 lc (22)

4. Solution procedure
Based on the material models of concrete including the tension-stiffening effect and reinforcing steel
defined previously to present the material nonlinearity, a finite element formulation was conducted.
The distributed steel model was adopted, because the reinforcement is uniformly distributed over
the concrete matrix with a particular orientation angle in an element. The steel fiber effect to
enhance the stiffness at post-cracking loading stage is indirectly taken into consideration through the
proposed tension-stiffening model. Accordingly, two-dimensional plane element is used for all
elements, since it can simulate the biaxial cracking behavior more effectively compare to one-
dimensional layered beam element.
To simulate the stress state of the concrete under biaxial loading, the orthotropic model was
adopted in this paper for its simplicity and computational efficiency. With reference to the principal
axes of orthotropy, the incremental constitutive relationship can be expressed by
⎧ dσ1 ⎫ ⎧ dε1 ⎫

⎨ d σ 2 ⎬ = [ D ] LO ⎨ d ε 2 ⎬ (23)
⎪ ⎪ ⎪ ⎪

⎩ dσ3 ⎭ ⎩ dε3 ⎭
⎪ ⎪ ⎪ ⎪

E1 v E1 E2 0
1-
D ]LO = ----------- (24)
[
1 – v2 v E 1 E 2 E2 0
0 0 ( 1 – v2 )G
where (1 – v2 )G = 0.25(E1 + E2 – 2v E1E2 ) , E , and E are the secant moduli of the elasticity in the
1 2

direction of the axes of orthotropy, which are oriented perpendicular and parallel to the crack
direction. Additionally, G is the shear modulus of the elasticity and v is Poisson’s ratio.
As cracks progress, changes in the crack direction are simulated using the rotating crack model
 A numerical tension-stiffening model for ultra high strength fiber-reinforced concrete beams 13

with the smeared crack model, it is assumed that a crack forms in a direction perpendicular to the
principal strain when the principal tensile strain exceeds the cracking strain εo. Since the material
matrix is defined with reference to the principal strain direction, it must be transformed to the global
coordinate system before all element stiffness matrixes can be assembled. This is accomplished by
the following transformation
T
[ D ] GL = [ T ] [ D ] LO [ T ] (25)
where tan 2θ = γxy ⁄ (εx – εy) , θ is the angle between the direction normal to the crack and the global
x-direction, and [T] is a transformation matrix.
The reinforcing bars embedded in the concrete element are replaced by an equivalent steel
element. Since the equivalent steel element has uniaxial properties in the direction parallel to the
axis of the reinforcing bars, the constitutive material matrix takes the simple form:

Fig. 7 Flowchart for solution procedure

14 Chaekuk Na and Hyo-Gyoung Kwak

0 ρi Esi 0
[ Ds ]LO = 0 0
0 (26)
0 0 0
The stiffness matrix of the composite reinforced concrete arrived at by the superposition of the
concrete and reinforcing steel stiffness matrix can be expressed as Eq. (27), where n is the number
of steel elements embedded in the concrete element.
n n
[ K ]el = [ Kc ]el + ∑K
i=1
[ s ]i, el = ∫ [B]T⎨[Dc ]GL +
V
⎧

⎩
∑D
i=1
[
⎫
s ]i, GL ⎬[ B ] dV
⎭
(27)

The arc-length method (Crisfield 1991) has recently been adopted as a solution scheme for the
material nonlinear analysis of UHSFRC structures displaying strength degradation after yielding of
steel. All of the remaining procedures, from the construction of the element stiffness matrix to the
convergence check, are identical to those used in a classical nonlinear analysis of RC structures. A
summary of the nonlinear solution algorithm is presented in Fig. 7, and more details of arc-length
method can be found elsewhere (Crisfield 1991).

5. Verification of tension-stiffening model

To verify the efficiency of the introduced tension-stiffening model, the cracking behaviors of
tension members subject to direct tensile force were analyzed. Two different tension members,
which are composed of high strength concrete and ultra high strength fiber-reinforced concrete
tested by Bischoff (2003) and Jungwirth and Muttoni (2004), respectively, were selected. These two
tension members have been the subject of analytical correlation studies (Fields and Bischoff 2004,
Redaelli 2006).

Fig. 8 Configuration of specimen 1

Table 1 Material properties used in specimen 1

Concrete properties Steel properties
Specimen
fc′ (MPa) ft(MPa) Ec (GPa)
*
fy(MPa) Es(GPa) ρ(%) ρfiber(%)
15M 420.5 202 2.0 0.78
62.4 4.8 33
20M 441.7 207 3.0 0.78
*not noted in Bischoff (2003), computed from ACI318 (Ec=3,320 ( fc′ ) +6,900)
½
 A numerical tension-stiffening model for ultra high strength fiber-reinforced concrete beams 15

Fig. 9 Normalized stress-strain relationship

Fig. 10 Member response with tension-stiffening effect

The first specimen (specimen 1), as shown in Fig. 8, is a specimen with a square cross section
dimension of 100 mm × 100 mm and is reinforced with either a single 15M or 20M bar, corresponding to
a steel ratio of 2% and 3%, respectively. The material properties of the test specimen are
summarized in Table 1, and more details related to the experimental study can be found elsewhere
(Bischoff 2003).
Figs. 9 and 10 show the cracking responses for the bonded members. Since this study did not take
into account the member shortening caused by shrinkage, the initial offset in the strain distribution
induced from shrinkage deformation of the concrete matrix has been subtracted from the original
experimental data. Fig. 9 shows the relationship between the normalized stress and the normalized
strain, and Fig. 10 represents the corresponding relationship between the applied axial load and the
average axial strain.
The comparisons of the experimental and analytical results show that the introduced tension-
stiffening model presents improved cracking behavior through all the loading stages. CEB model
(1993) recommends a constant value of 0.4 for the normalized stress at the post-cracking stage
regardless of the magnitude of the corresponding strain; therefore, the cracking load seems lower
16 Chaekuk Na and Hyo-Gyoung Kwak

Fig. 11 Configuration of specimen 2

Table 2 Material properties used in specimen 2

Concrete properties Steel properties
Specimen
fc ′ (MPa) ft(MPa)
*
Ec(GPa) fy(MPa) Es(GPa) ρ(%) ρfiber(%)
2.5% 3.14 2.5
146 8.9 63 556 200
4.1% 4.94 4.1
*not noted in Jungwirth (2004), conducted from KICT test result (2005)

than that of experimental and analytical model. In addition, during the yielding stage of steel rebar,
CEB model cannot fit the experimental results, because it does not take into account the
contribution of fibers. Another relationship of Collins and Mitchell (1991), where the normalized
stress = 1 ⁄ (1 + 500εm ) and εm equals the average member strain, slightly underestimates the
cracking behavior at the initial post-cracking stage. This is attributed to absence of consideration of
the bond effect in Collin’s model between the concrete matrix and included steel fiber. In advance,
this underestimation is expected to be enlarged as the amount of steel fiber is increased.
In order compare the differences in the post-cracking behavior according to changes in the amount
of steel fiber in the concrete matrix and the compressive strength of concrete, another specimen
(specimen 2) tested by Jungwirth and Muttoni (2004) was also selected. Its configuration and the
material properties used in the experiment are shown in Fig. 11 and Table 2, respectively. Figs. 12
and 13 show results corresponding closely with those obtained in the previous specimen (specimen
1) have been obtained in spite of the relatively high compressive strength of concrete. An uncertain
phenomenon is observed in Fig. 12; that is, the stiffness of member increases after steel yielding at

Fig. 12 Normalized stress-strain relationship

 A numerical tension-stiffening model for ultra high strength fiber-reinforced concrete beams 17

Fig. 13 Member response with tension-stiffening effect

point B in Fig. 6. This phenomenon can be explained as the bridging effect of fiber after steel
yielding at crack face. In spite of this phenomenon, the analytical results given by the introduced
tension-stiffening model provide good agreement with the experimental results.
Finally, from the obtained results, it can be inferred that the introduced model can effectively
simulate the tension-stiffening effect in a UHSFRC member regardless of changes in the compressive
strength of concrete and the amount of steel fibers. Meanwhile, the direct use of the conventional
relationship of normal strength reinforced concrete member leads to underestimation of the tension-
stiffening effect even in the case of a UHSFRC member.

6. Numerical analysis
The experimental results from several UHSFRC beams tested at KICT (2005) were used to
investigate the validity of the proposed analytical model. Because of a limited number of
experimental data for UHSFRC members, in this paper the correlation studies between the
analytical results and the experimental values have not been extended to various structural members
beyond UHSFRC beams.
Four types of simply supported UHSFRC beams have been investigated as shown in Fig. 14.
Each beam has a rectangular cross section size of b × h = 125 mm × 250 mm and is reinforced with
two mild steel bars. Two equal point loads are applied and their position is the major difference in each
specimen (see Table 3). In advance, the same material properties of concrete and steel as those in the
experimental study are used as follows: the yield strength of steel is fy = 538 MPa , the elastic modulus of

Fig. 14 Configuration of specimens

18 Chaekuk Na and Hyo-Gyoung Kwak

Table 3 Steel amounts and loading points

Specimen D10L16 D10L20 D10L24 D10L28 D13L24
As 2D10 2D10 2D10 2D10 2D13
la (mm) 396 550 770 990 770
span ratio (la/d) 1.8 2.5 3.5 4.5 3.5

Fig. 15 Load-deflection relationships according to span ratio

Fig. 16 Load-deflection relationships according to steel ratio

steel is Es = 200, 000 MPa , the compressive and the tensile strength of UHSFRC are fc′ = 146 MPa and
ft = 13.9 MPa , respectively, and the elastic modulus of concrete is Ec = 49, 000 MPa . Uniform steel
fiber of 2% is adopted for all specimens.
Fig. 15 represents a typical relationship between the applied lateral load and the corresponding
vertical deflection at the mid-span according to the span ratio. The proposed numerical model not
only gives accurate predictions for the ultimate load but also effectively simulates the nonlinear
behavior of UHSFRC beams as the lateral load increases from zero to its ultimate value. All the
 A numerical tension-stiffening model for ultra high strength fiber-reinforced concrete beams 19

specimens represented the bending failure, and Fig. 15 shows that an increase of the shear span
ratio of la/d accompanies a decrease of the ultimate resisting capacity with an increase of the lateral
deflection.
As the steel ratio increases, the ultimate capacity increases beyond steel yielding stress without
major changes of the neutral axis of the section. The experimental data in Fig. 16 reflects more
ductile behavior in specimen D13L24 in spite of having a larger steel ratio than specimen D10L24.
This appears to be induced from ultimate resisting capacities and the post-yielding behavior in these
specimens being governed by the concrete matrix in the tensile region rather than the yielding of the
reinforcing bars.
In order to investigate the contribution of the tension-stiffening effect to the structural response,

Fig. 17 Load-deflection relationships of specimen D10L24

20 Chaekuk Na and Hyo-Gyoung Kwak

Fig. 18 Load-deflection relationships of specimen D13L24

analytical results with and without tension-stiffening are compared for two representative specimens,
D10L24 and D13L24, in conjunction with changes in the compressive strength of concrete and the
yield strength of steel. These two specimens present typical brittle and ductile behavior, respectively.
For parametric studies, different strengths of materials from those used in the experiment are
considered, and attention is given to normal strength concrete (NSC) with fc′ = 40 MPa and
ft = 2.1 MPa , ultra high strength concrete (UHSC) with fc′ = 146 MPa and ft = 13.9 MPa , normal
strength steel (NSS) with fy = 538 MPa and high strength steel (HSS) with fy = 1, 538 MPa .
Figs. 17 and 18, which show the obtained results, lead to the following conclusions: (1) the
tension-stiffening effect in the NSC beam (NSRC), representing the bending behavior, is not large
enough to change the resisting capacity of the member (see Figs. 17(c) and 18(c)); (2) the entire
structural behavior of the UHSFRC beam (UHSFRC), from the initial cracking to reach the ultimate
 A numerical tension-stiffening model for ultra high strength fiber-reinforced concrete beams 21

resisting capacity, is dominantly governed by the tension-stiffening effect regardless of the change in
steel strength. This large contribution of the tension-stiffening effect is induced from the bond effect
between the steel fibers and the concrete matrix; (3) the structural response of specimen D10L24 is
changed to ductile behavior when the reinforcing steel is changed from normal strength steel (NSS)
to high strength steel (HSS). This means that the use of NSS is inappropriate with respect to
ensuring ductility in UHSFRC beams; accordingly, (4) HSS must be used in the case of UHSFRC
members for more rational design; (5) as shown in Figs. 17(a) and 18(a), when the tension-
stiffening effect is not taken into account in the numerical simulation, the ultimate loads of
UHSFRC case for both specimens D10L24 and D13L24 are almost the same as those of NSRC
case, because the magnitude of the ultimate resisting capacity is directly proportional to the amount
of reinforcing steel, and the equality between the developed moment by Mu = P ⋅ a ⁄ 2 and the
sectional resisting capacity gives similar ultimate loads for both specimens. This phenomenon
appears even when different materials are used (see Figs. 17(b) and 18(b) for the use of HSS and
Figs. 17(c) and 18(c) for the use of NSC, respectively); and finally, (6) in order to exactly evaluate
the ultimate resisting capacity of UHSFRC beams, the tension-stiffening effect derived on the basis
of the mechanical behavior of UHSFRC must be considered in the numerical simulation.

7. Conclusions
In this paper, a numerical tension-stiffening model that can simulate the post-cracking behavior of
UHSFRC members is proposed on the basis of the force equilibriums, compatibility conditions, and
bond stress-slip relationship between the reinforcing steel and the surrounding concrete. In advance,
the bond characteristics between the steel fibers and the concrete matrix are also taken into
consideration, and the efficiency of the proposed model in finite element analyses is verified by
comparison with reliable experimental results for UHSFRC beams.
Based on the results of limited correlation studies among the analytical results, the test values, and
associated parametric studies, the following conclusions are obtained: (1) in order to exactly evaluate the
ultimate resisting capacity of UHSFRC beams, the tension-stiffening effect must be considered; (2) in
defining the tension-stiffening effect for UHSFRC structures, the bond characteristics between the
concrete matrix and the embedded steel fibers must be incorporated; (3) the use of high strength steel
are recommended in UHSFRC members in order to effectively use the enlarged resisting capacity of
UHSFRC for tensile stress; and (4) the introduced tension-stiffening model can effectively be used in
evaluating the ultimate resisting capacity of UHSFRC beam structures.

Acknowledgments
The work presented in this paper was funded by Center for Concrete Corea (05-CCT-D11), supported by
Korea Institute of Construction and Transportation Technology Evaluation and Planning (KICTTEP)
under the Ministry of Land, Transport and Maritime Affairs.
22 Chaekuk Na and Hyo-Gyoung Kwak

References
Attard, M.M. and Setunge, S. (1996), “Stress-strain relationship of confined and unconfined concrete”, ACI
Mater. J., 93(5), 1-11.
Bischoff, P.H. (2003), “Tension-stiffening and cracking of steel fiber-reinforced concrete”, J. Mater. Civil. Eng. -
ASCE, 15(2), 174-182.
CEB-FIP (1993), CEB-FIP model code 1990: design code, CEB Bulletin d'Information No.213/214, Thomas
Telford, London.
Collins, M.P. and Mitchell, D. (1991), Prestressed concrete structures, Prentice-Hall, Englewood Cliffs, NJ.
Crisfield, M.A. (1991), Non-linear finite element analysis of solids and structures, John Wiley and Sons, London.
Demeke, A. and Tegos, I.A. (1994), “Steel fiber reinforced concrete in biaxial stress tension-compression
conditions”, ACI Struct. J., 91(5), 579-584.
Ezeldin, A.S. and Balaguru. P.N. (1992), “Normal- and high-strength fiber reinforced concrete under compression”, J.
J. Mater. Civil. Eng. - ASCE, 4(4), 415-429.
Fields, K. and Bischoff, P.H. (2004), “Tension-stiffening and cracking of high-strength reinforced concrete
tension members”, ACI Struct. J., 101(4), 447-456.
Foster, S.J., Voo, Y.L. and Chong, K.T. (2006), “FE analysis of steel fiber reinforced concrete beams failing in
shear: variable engagement model”, ACI Special Public., 237, 55-70.
Graybeal, B.A. (2007), “Compressive behavior of ultra-high-performance fiber-reinforced concrete”, ACI Mater.
J., 104(2), 146-152.
Gupta, A.K. and Maestrini, S.R. (1990), “Tension-stiffness model for reinforced concrete bars”, J. Struct. Eng. -
ASCE, 116(3), 769-790.
Hognestad, E. (1951), A study on combined bending and axial load in reinforced concrete members, Bulletin
Series No. 399, Bulletin No. 1., Engineering Experiment Station, University of Illinois, IL.
Hussein, A. and Marzouk, H. (2000), “Behavior of high-strength concrete under biaxial stresses”, ACI Mater. J.,
97(1), 27-36.
Jungwirth, J. and Muttoni, A. (2004), “Structural behavior of tension members in UHPC”, Proceeding of
International Symposium on UHPC.
Kölle, B., Phillips, D.V., Zhang, B., Bhatt, B. and Pearce, C.J. (2004), “Experimental investigation of the biaxial
properties of high performance steel fibre reinforced concrete”, Proceeding of 6th RILEM Symposium on
Fibre-Reinforced Concrete (FRC): RILEM, Varenna, Italy.
Korea Institute of Construction Technology (2005), Development of girder using ultra-high performance
cementitious composite for bridge, KICTEP, Korea.
Kupfer, H., Hilsdorf, H.K. and Rusch, H. (1969), “Behavior of concrete under biaxial stresses”, ACI J., 66(8),
656-666.
Kwak, H.G. and Kim, D.Y. (2001), “Nonlinear analysis of RC shear walls considering tension-stiffening effect”,
Comp. Struct., 79(5), 499-517.
Kwak, H.G. and Kim, D.Y. (2004), “Material nonlinear analysis of RC shear walls subject to monotonic
loading”, Eng. Struct., 26(11), 1517-1533.
Kwak, H.G. and Na, C.K. (2007), “Nonlinear analysis of RC structures considering bond-slip effect”, Korean
Soc. Civil Eng., 27(3), 401-412.
Li, V.C. (1998), Engineered cementitious composites-tailored composites through micromechanical modeling,
Canadian Society for Civil Engineering, Montreal.
Lorrain, M., Maurel, O. and Seffo, M. (1998), “Cracking behavior of reinforced high-strength concrete tension
ties”, ACI Struct. J., 95(5), 626-635.
Mansur, M.A., Chin, M.S. and Wee, T.H. (1999), “Stress-strain relationship of high-strength fiber concrete in
compression”, J. Mater. Civil. Eng. - ASCE, 11(1), 21-29.
Redaelli, D. (2006), “Testing of reinforced high performance fibre concrete members in tension”, Proceeding of
6th International Ph.D. Symposium in Civil Engineering.
Rizkalla, S.H. and Hwang, L.S. (1984), “Crack prediction for members in uniaxial tension”, ACI J., 81(6), 572-579.
SETRA (2002), Béntons Fibrés á Ultra-Hautes Peformances, Recommandations Provisoires, Documents
Scientifiques et Techniques, SETRA-AFGC, France.