Title Tension stiffening in concrete beams.

Part 1: FE analysis

Author(s) Ng, PL; Lam, JYK; Kwan, AKH

Proceedings Of The Institution Of Civil Engineers: Structures

Citation And Buildings, 2010, v. 163 n. 1, p. 19-28

Issued Date 2010

URL http://hdl.handle.net/10722/127261

This work is licensed under a Creative Commons Attribution-

NonCommercial-NoDerivatives 4.0 International License.;
Rights Permission is granted by ICE Publishing to print one copy for
personal use. Any other use of these PDF files is subject to
reprint fees
Proceedings of the Institution of
Civil Engineers
Structures and Buildings 163
February 2010 Issue SB1
Pages 19–28
doi: 10.1680/stbu.2009.163.1.19 Pui L. Ng Jeffrey Y. K. Lam Albert K. H. Kwan
Honorary Research Former Research Student, Professor and Head,
Paper 700065 Associate, Department of Department of Civil Department of Civil
Received 21/11/2007 Civil Engineering, The Engineering, The University Engineering, The University
Accepted 12/08/2009 University of Hong Kong of Hong Kong of Hong Kong
Keywords: beams & girders/
concrete structures/mathematical
modelling

Tension stiffening in concrete beams. Part 1: FE analysis

P. L. Ng PhD, J. Y. K. Lam PhD and A. K. H. Kwan PhD, MICE, CEng

Although after cracking, concrete has negligible tension action of the curvature of the flexural member. In other words,
capacity, the intact concrete between cracks within the there are two mechanisms that induce tensile stresses in the
tension zone of a reinforced concrete beam can still concrete of a cracked flexural member. The first mechanism, by
develop significant tensile stresses to contribute to the which the stress transfer through the bond induces tensile
flexural stiffness of the concrete beam. Such a tension stresses, is similar to that in a cracked axial member, as
stiffening effect in a flexural member is not quite the depicted in Figure 1(a). The second mechanism, by which the
same as that in an axial member because the tensile shearing action of the curvature induces tensile stresses, is like
stresses in a cracked flexural member are induced not that of a short length of unreinforced concrete beam between
only by the steel reinforcement–concrete bond but also two adjacent cracks subjected to curvature, as depicted in
by the curvature of the flexural member. In this study, Figure 1(b) which shows that near the cracks, the plane
the tensile stresses developed in cracked concrete beams sections no longer remain plane after cracking and shear
are analysed using a finite-element (FE) model that takes stresses are developed to pull the uncracked concrete in the
into account the non-linear biaxial behaviour of the tension zones thereby inducing significant tensile stresses in
concrete and the non-linear bond stress–slip behaviour the uncracked concrete. Hence, tension stiffening in flexural
of the steel reinforcement–concrete interface. Based on members is not the same as that in axial members.
the numerical results so obtained, a tensile stress block
is proposed for section analysis of the moment– Since the shearing action of curvature mentioned above is not
curvature curves of reinforced concrete beams at both yet common knowledge, further explanation is given in this
the uncracked and cracked states. It will be shown in paper. Basically, at locations near the cracks, vertical fibres (i.e.
part 2 of this paper that the tensile stress block may also fibres that are initially vertical) in the tension zones are no
be used for member analysis of the load–deflection longer perpendicular to horizontal fibres (fibres that are
curves of concrete beams without resorting to FE initially horizontal). Significant shear strains and stresses are
analysis. developed at the centroidal axis near the cracks. The shear
stresses pull the tension zones near the cracks so that at a
1. INTRODUCTION certain distance away from the cracks, the tension zones
In a reinforced concrete member, the stiffness of the concrete develop tensile stresses as in uncracked sections. This is like the
decreases after cracking but does not drop to zero because the shear lag near cracked concrete, as was pointed out by Hughes
uncracked concrete between adjacent cracks is still able to (see Figure 2 of Hughes (2008)). Alternatively, the action may
carry some tensile stresses to contribute to the overall stiffness be visualised as the development of shear stresses to maintain
of the member. This phenomenon is called tension stiffening. It equilibrium between the uncracked and cracked sections. At
occurs in both axial and flexural members. In axial members, the uncracked sections, tensile stresses develop due to
such as struts subjected to uniaxial loads and panels subjected curvature of the member, while only negligible tensile stresses
to biaxial loads, tensile stresses are induced in the concrete develop at the cracked sections. So, between an uncracked
between cracks mainly by the stress transfer through the steel section and a cracked section, there must be horizontal shear
reinforcement–concrete bond. Extensive studies on tension stresses developed along the centroidal axis to balance the
stiffening in axial members have been carried out, including difference in horizontal tensile forces. Without curvature, as in
experimental investigations (Jiang et al., 1984; Wollrab et al., an axial member, no such tensile stresses due to curvature
1996), theoretical modelling (Floegl and Mang, 1982; Gupta would be developed. That is why the authors have chosen to
and Maestrini, 1989; Wu et al., 1991; Choi and Cheung, 1996) call this ‘the shearing action of curvature’.
and the development of tensile stress blocks (Link et al., 1989;
Gupta and Maestrini, 1990; Massicotte et al., 1990; Fields and There have been relatively few reported studies on the tension
Bischoff, 2004). stiffening in flexural members compared with that in axial
members. Two alternative methods of accounting for the
In flexural members, tensile stresses are induced in the tension stiffening in flexural members have been developed.
concrete between cracks not only by the stress transfer through The first method aims at a direct evaluation of moment–
the steel reinforcement–concrete bond but also by the shearing curvature curves, while the second attempts the development

Structures and Buildings 163 Issue SB1 Tension stiffening in concrete beams. Part 1: FE analysis Ng et al. 19
 itself. The tensile stress block has been given in such a form
solely for the purpose that it may be applied at all stress states.
Centroidal In a cracked concrete beam, the tensile stress induced in the
Concrete axis
concrete between two adjacent cracks varies not only with the
depth but also along the beam axis. At a cracked section, the
Crack Crack
Steel reinforcing bar tensile stress is practically zero while at an uncracked section,
the tensile stress increases with the distance from the nearest
(a) cracked section. Since the cracks are to be smeared in the
structural analysis such that each beam section may be assumed
Centroidal to have a flexural stiffness changing smoothly with the mean
Concrete axis
curvature regardless of the actual positions and spacing of the
cracks, the tensile stress block should be taken as the
Crack Crack distribution of the mean tensile stress within the beam depth.
Only horizontal Plane sections
stresses shown; no longer
length of arrow However, due to the difficulties involved, the tension stress
remain plane
indicates magnitude fields in cracked concrete beams have never been thoroughly
(b) measured; consequently, there is still insufficient experimental
data for evaluating the distribution of the mean tensile stress
Figure 1. Tension stiffening in a flexural member: (a) stress within the beam depth. To overcome this difficulty, the present
transfer through bond; (b) shearing action of curvature study attempts to evaluate the tension stress fields in cracked
concrete beams by FE analysis. Typical reinforced concrete
beams with varying structural parameters have been analysed
of tensile stress blocks. For the first method, Creazza and Di and the tension stress fields under different conditions
Marco (1993) and Polak and Blackwell (1998) developed obtained. This is probably the first time that the actual
theoretical models to evaluate moment–curvature curves. distribution of mean tensile stress within a beam depth has
Piyasena et al. (2002) carried out finite-element (FE) analysis of been directly determined.
the variation of curvature between adjacent cracks, from which
the moment–mean curvature curve could be derived. However, 2. FINITE-ELEMENT ANALYSIS
despite the simplifying assumptions made, these theoretical
models are far too complicated for practical applications. For 2.1. Secant stiffness formulation and direct iteration
this reason, design codes CEB-FIP Model Code 1990 (CEB, In order to extend the analysis into the post-crack and post-
1993) and ACI building code (ACI, 2008) instead give empirical peak ranges within which the tangent stiffness can become
formulas for a quick and approximate evaluation of the undefined or negative, secant stiffness is used in the
moment–curvature curves. formulation of the stiffness matrices. For the non-linear
analysis, an iterative procedure with the loads applied in small
For the second method, several tensile stress blocks have been increments is used. At each load increment step, direct iteration
developed. As full-field measurement of the tensile stresses in a using the secant stiffness of the structure is employed. With
cracked concrete beam is extremely difficult, tensile stress this method, the loads can be applied either directly in the
blocks are generally developed by proposing a certain tensile form of prescribed forces or indirectly in the form of prescribed
stress block with unknown parameters and then determining displacements at the loading points.
the unknown parameters by curve-fitting of the theoretical
moment–curvature or load–deflection curves with The above formulation and numerical procedures for
experimental results. Gilbert and Warner (1978) considered developing FE codes have been used since the 1980s (Liauw
three alternative tensile stress blocks with different descending and Kwan, 1982). In this particular study, the FE code used is a
branches. Prakhya and Morley (1990) adopted a tensile stress tailor-made and refined version for reinforced concrete
block comprising a linear ascending branch and a non-linear members (Ng, 2007). It has been carefully validated before use.
descending branch, while Kaklauskas and Ghaboussi (2001)
adopted a tensile stress block composed of a linear ascending 2.2. Modelling of concrete
branch and a linear descending branch. Scott (1983) and Beeby The concrete is modelled by plane stress elements, which are
et al. (2005) proposed tensile stress blocks each comprising simply three-noded triangular constant strain elements. Its
multi-linear ascending and descending branches. Recently, non-linear biaxial stress–strain behaviour is accounted for in
Torres et al. (2004) adopted the strategy of determining the terms of the equivalent uniaxial strains defined by
unknown parameters by curve-fitting with empirical moment–
curvature curves given in a design code. The tensile stress 1
1a åe1 ¼ ðå1 þ v2 å2 Þ
block given in BS 8110 (BSI, 1985) is in the form of a linear 1  v1 v2
ascending curve with no descending branch at all. It is clear
then that all these tensile stress blocks differ widely and it is
difficult to judge which is more appropriate. 1
1b åe2 ¼ ðå2 þ v1 å1 Þ
1  v1 v2
It should be noted that although the tensile stress block is
normally given in the form of a stress–strain curve, it is not the where åe1 and åe2 are the equivalent uniaxial strains, å1 and å2
same as the tensile stress–strain curve of the concrete material are the principal strains, v1 and v2 are Poisson’s ratios and

20 Structures and Buildings 163 Issue SB1 Tension stiffening in concrete beams. Part 1: FE analysis Ng et al.
subscripts 1 and 2 denote quantities in the respective principal respectively. From these secant stiffness values, the constitutive
directions. Each of the principal stresses ó1 and ó2 is assumed matrix [D9c ] of the concrete in the local coordinate system,
to be a single variable function of the corresponding whose coordinate axes are the same as the principal directions,
equivalent uniaxial strain. In effect, therefore, the biaxial is derived as
stress–strain relation is decomposed into two independent
uniaxial stress–strain relations. 2 3
Ec1 v2 Ec1
6 07
6 1  v1 v2 1  v1 v2 7
6 7
½D9c  ¼ 6 7
Due to biaxial effects, the tensile and compressive strengths in
4 6 v1 Ec2 Ec2
07
the principal directions are not the same as the uniaxial tensile 6 7
4 1  v1 v2 1  v1 v2 5
and compressive strengths. They are determined using the
0 0 G
biaxial strength envelope developed by Kupfer and Gerstle
(1973), which consists of four distinct zones, namely the
tension–tension, tension–compression, compression– in which G is the shear modulus. Before cracking, the shear
compression and compression–tension zones. To evaluate the modulus is just taken as the initial elastic shear modulus Go .
tensile and compressive strengths in the principal directions, it After cracking, since the principal directions are the directions
is necessary first to find out within which zone the biaxial perpendicular and parallel to the cracks formed, the principal
stress state falls and then locate the point on the strength directions become the cracking directions and are then fixed
envelope that will be intercepted when the biaxial stresses keep (i.e. no longer allowed to rotate). As the crack may still be able
on increasing. The strength values at the intercepting point are to transmit a small amount of shear stress by aggregate
the tensile and compressive strengths in the principal interlock, the shear modulus after cracking is taken as ªGo , in
directions, denoted f tp and f cp respectively. which ª is a dimensionless shear retention factor depending on
the tensile strain perpendicular to the crack, as given by He
For any principal direction under tension, the stress–strain and Kwan (2001). From the above constitutive matrix, the
curve follows that of Guo and Zhang (1987), which is given by constitutive matrix [Dc ] of the concrete in the global
   6 coordinate system is obtained simply by the usual coordinate
ó å å
¼ 1:2  0:2 transformation.
2a f tp åtp åtp
2.3. Modelling of steel reinforcement
for the ascending branch
In order to allow for bond slip between the longitudinal
reinforcing bars and the surrounding concrete, the longitudinal
 ,"  17  # reinforcement is modelled by discrete bar elements connected
ó å å å to the concrete through bond elements. The bar elements are
¼ a 1 þ
2b f tp åtp åtp åtp one-dimensional elements possessing only axial stiffness. For
modelling the elastic, plastic and strain hardening behaviour of
for the descending branch
the steel, the constitutive model proposed by Mander (1984) is
adopted. Based on this constitutive model, the relation between
in which åtp is the strain at peak tensile stress and a is a the steel stress ós and steel strain ås is given by
dimensionless coefficient equal to 0.312 f 2t ( f t being the
uniaxial tensile strength in MPa). Since the crack tips in
5a ós ¼ E so ås for ås < f y =E so
reinforced concrete beams are usually located near the neutral
axis where the tensile stresses are small, the fracture energy of
the crack tips has been neglected in the analysis.
5b ós ¼ f y for f y =E so , ås < åsh
For any principal direction under compression, the stress–
strain curve follows that of Saenz (1964), which is given by
"   n#
  åu  ås
  , ós ¼ f y þ f u  f y 1
ó E co å åu  åsh
¼ 5c
f cp E cp åcp
" for åsh , ås < åu
   2 #
3

Eco å å
1þ 2 þ
Ecp åcp åcp
where Eso is the initial elastic modulus, f y is the yield strength,
f u is the ultimate tensile strength, åsh is the strain at start of
in which Eco is the initial elastic modulus, and Ecp and åcp are strain hardening, åu is the ultimate strain and n is a
the secant modulus and strain at peak compressive stress, dimensionless parameter depending on the strain-hardening
respectively. property of the steel. From the steel stress ós and steel strain
ås , the secant stiffness Es of the steel is calculated as ós /ås .
For each principal direction, the principal stress ó is obtained With this secant stiffness, the stiffness matrix of the bar
by substituting the corresponding equivalent uniaxial strain as element can be obtained in the usual way (ASCE, 1982). On the
å into Equation 2 or 3, whichever is applicable. Having other hand, the transverse reinforcement is assumed to be
evaluated the principal stresses ó1 and ó2 , the secant perfectly bonded to the concrete and modelled as smeared
stiffnesses Ec1 and Ec2 are calculated as ó1 /åe1 and ó2 /åe2 , reinforcement (ASCE, 1982).

Structures and Buildings 163 Issue SB1 Tension stiffening in concrete beams. Part 1: FE analysis Ng et al. 21
 2.4. Modelling of bond effective depth, that is the depth from the top of beam section
The bond between the longitudinal reinforcement and concrete to the centreline of the tension reinforcement, is 550 mm. In
is modelled using a four-noded bond element similar to the each beam, the tension reinforcement is provided in the form
interface element developed by Goodman et al. (1968). Each of two main bars. Figure 2(a) shows the general layout of the
bond element is assumed to have an infinitesimally small beams. In order to study the effects of the amount of tension
thickness. It has two pairs of duplicated nodes. The two nodes reinforcement, the tension reinforcement ratio rt was varied
in each pair of duplicated nodes have the same coordinates but (0.5, 1.0, 1.5 and 2.0%). The shear reinforcement ratio was
different degrees of freedom. Between them, one is connected fixed at 0.4%. In total, eight reinforced concrete beams were
to the steel reinforcement while the other is connected to the analysed.
concrete. The difference in displacement of the duplicated
nodes in the direction of the steel–concrete interface is taken The properties of the beam materials are presented in Table 1.
as the bond slip. For the steel reinforcement–concrete bond, the material
parameters pertinent to deformed bars embedded in unconfined
In this study, the bond stress–slip relation recommended by concrete recommended by CEB-FIP Model Code 1990 were
CEB-FIP Model Code 1990 (CEB, 1993) is employed. It is given adopted. Following the model code, the peak bond stress ôp
: :
by and residual bond stress ôf are taken as 2.0(f c )0 5 and 0.3(f c )0 5 ,
respectively.
 04
sb
6a ôb ¼ ôp for sb < s1
s1 The beams were discretised into 12 layers of FEs, as depicted
by the mesh shown in Figure 2(b). During FE analysis, loading
was applied in small increments to each beam as prescribed
6b ôb ¼ ôp for s1 , sb < s2 displacement when the beam was subjected to a single point
load at mid-span or as prescribed force when the beam was

 
sb  s2
ôb ¼ ðôf  ôp Þ þ ôp 600 550
s3  s2
6c
for s2 , sb < s3
6000

(a)

6d ôb ¼ ôf for s3 , sb

in which ôb is the bond stress, ôp is the peak bond stress, ôf is

the residual bond stress, sb is the bond slip, and s1 , s2 and s3 (b)
are the slip at the start of peak bond stress, slip at the end of
peak bond stress and slip at the start of residual bond stress, Figure 2. (a) General layout (dimensions in millimetres) and
respectively. (b) finite-element mesh

Initially, before bond slip occurs, the secant bond stiffness kb is

Concrete
taken as 200 N/mm3 , as recommended by CEB-FIP Model Code
1990. After bond slip has occurred, the bond stress ôb is
Uniaxial tensile strength f t : MPa 3.0
determined by substituting the bond slip sb evaluated from the
Uniaxial compressive strength f c : MPa 30.0
nodal displacements of the bond element into the above Initial elastic modulus Eco : GPa 30.0
equations and the secant bond stiffness kb is calculated as Poisson’s ratio  0.2
ôb /sb . Having obtained the secant bond stiffness, the stiffness
matrix of the bond element in the local coordinate system may Steel reinforcement
then be derived following the procedures developed by
Goodman et al. (1968) with the area of the interface taken as Yield strength f y : MPa 460
the length of the bond element multiplied by the total Ultimate tensile strength f u : MPa 552
perimeter of the steel reinforcing bars. Finally, the stiffness Initial elastic modulus Eso : GPa 200
matrix of the bond element in the global coordinate system is Tensile strain at start of strain hardening åsh : % 1
obtained by the usual coordinate transformation. Ultimate tensile strain åu : % 10

3. ANALYSIS OF REINFORCED CONCRETE BEAMS Steel reinforcement–concrete bond

3.1. Beams analysed Peak bond stress ôp : MPa 11.0

A series of typical simply supported reinforced concrete beams Residual bond stress ôf : MPa 1.6
Slip at start of peak bond stress s1 : mm 0.6
having different amounts of tension reinforcement and
Slip at end of peak bond stress s2 : mm 0.6
subjected to either a single point load at mid-span or a Slip at start of residual bond stress s3 : mm 1.0
uniformly distributed load over the entire span were analysed.
All the beams analysed have a span of 6000 mm and a uniform Table 1. Material properties
cross-section of 300 mm breadth by 600 mm depth. The

22 Structures and Buildings 163 Issue SB1 Tension stiffening in concrete beams. Part 1: FE analysis Ng et al.
subjected to a uniformly distributed load. At each load exhibits sharp changes at cracks. At the two opposite sides of a
increment step, direct iteration was employed. The direct crack, the bond stresses are in opposite directions. At the left-
iteration at each load increment step was repeated until the hand side of a crack, the bond slip of the reinforcement and
change in secant modulus in every FE was less than 2.0%. In the bond stress acting on the concrete are to the right, but at
the analysis of each beam, about 100 load increment steps the right-hand side, the bond slip of the reinforcement and the
were applied until the tension reinforcement yielded and the bond stress acting on the concrete are to the left. Consequently,
maximum tensile strain reached 2500 ìå. the bond between the reinforcement and the uncracked
concrete is always transferring tensile force to the concrete,
3.2. Analysis results leading to a gradual increase in the tensile stress of the
The crack pattern and stress distributions in a typical concrete with distance from the nearest crack.
reinforced concrete beam (rt ¼ 0.5% and subjected to a point
load at mid-span) obtained by the FE analysis are presented in The tensile force transferred from the reinforcement through
Figure 3. The first crack forms at mid-span where the bending the bond to the concrete (calculated by adding the bond forces
moment is largest. As the applied load increases, further cracks together) is illustrated and compared with the actual tensile
appear at regular spacings from previous cracks formed, as force in the tension zone of the concrete for the beam with
shown in Figure 3(a) for the crack pattern when the applied rt ¼ 0.5% and subjected to either a point load or a uniformly
load is equal to 45% of the peak load. distributed load in Figure 4. The tensile force transferred from
the reinforcement through the bond to the concrete accounts
From Figures 3(b) and 3(c), which show the stresses induced in for about 45–55% of the actual tensile force induced in the
the concrete and the tension reinforcement, it is evident that tension zone of the concrete. Hence, as explained earlier and
the concrete and steel stresses vary in the longitudinal illustrated in Figure 1, curvature of the beam should induce a
direction between two adjacent cracks. Basically, at a cracked significant tensile force in the tension zone of the concrete.
section, the tensile stress in the concrete first increases with
distance from the neutral axis and then decreases to a The load–deflection curves of the beams analysed are plotted
negligible value; at an uncracked section, the tensile stress in in Figure 5. For each beam, at an applied load smaller than
the concrete increases with distance from the neutral axis and
then remains at a certain significant value. On the other hand,
A B C
the tensile stress in the tension reinforcement always reaches a
Tensile force
local maximum value at a cracked section and decreases to a induced in concrete
195 kN
local minimum value at about halfway between two adjacent
cracks.

Figure 3(d), which shows the bond stress between the tension Tensile force
87 kN transferred by bond
reinforcement and the concrete, indicates that the bond stress

Applied load

A B C 0

A B C

Crack spacing (⬇0·8 times beam depth)

A B C
(a) (a)

A B C

Tensile force
A–A B–B C–C 153 kN induced in concrete

(b)
Tensile force
Tensile 79 kN transferred
stress by bond

Crack locations
0
(c)

Bond slip
Bond stress A B C
to the right

Crack spacing (⬇0·6 times beam depth)

Bond slip
to the left Crack locations (b)
(d)
Figure 4. Tensile force transferred through the bond and
Figure 3. (a) Crack pattern (rt ¼ 0.5%, applied load ¼ 45% of tensile force induced in concrete: (a) beam subjected to point
peak load); (b) stress distribution in concrete; (c) tensile load at mid-span (rt ¼ 0.5%); (b) beam subjected to uniformly
stress in reinforcement; (d) bond stress distributed load (rt ¼ 0.5%)

Structures and Buildings 163 Issue SB1 Tension stiffening in concrete beams. Part 1: FE analysis Ng et al. 23
 about 20% of the peak load, the load–deflection curve is simply as z, where  is the mean curvature and z is the depth
linear. As the applied load increases and when the beam from the neutral axis. However, after cracking, the curvature
eventually cracks, the gradient of the load–deflection curve fluctuates between adjacent cracks. To evaluate the mean
decreases. However, upon further increase in applied load, curvature, the local fluctuation of the curvature has to be
when no more new cracks are formed, the gradient of the curve smoothed. Here, the mean curvature at a beam section is
becomes constant until the beam starts to fail because of evaluated from FE analysis results by considering a short
yielding of the tension reinforcement and crushing of concrete. length of the beam starting and ending at one beam depth to
At the post-peak state, all the beams exhibit ductile behaviour each side of the beam section. The deflection of this short
because they are under-reinforced. length of beam is fitted with a cubic polynomial curve and
then the polynomial curve is differentiated twice with respect
to the length to obtain the smoothened curvature value as the
4. DERIVATION OF TENSILE STRESS BLOCK mean curvature at the beam section. Having obtained the mean
curvature at each beam section, the tensile stress in the
4.1. Tension stress field concrete may then be plotted against the theoretical tensile
The general shapes of the tension stress fields within cracked strain. As expected, the variation of tensile stress with
and uncracked sections were shown in Figure 3. For simply a theoretical tensile strain still changes with the stress state, but
presentation of numerical results, the tensile stress in the to a much lesser extent.
concrete obtained directly by FE analysis can be plotted
against depth. However, the variation of the tensile stress in Also as expected, the variation of tensile stress with theoretical
concrete with depth changes dramatically with the stress state. tensile strain is quite different at different sections, depending
To avoid dealing with many different tensile stress–depth on whether the section is cracked or uncracked and the
curves at different stress states, it is better to plot the tensile distance from the nearest cracked section. As the cracks are to
stress against the ‘theoretical tensile strain’. The theoretical be smeared in the structural analysis, it is the ‘mean tensile
tensile strain is not the true tensile strain of the concrete, but is stress’ that is of greater interest. The mean tensile stress is the
the tensile strain value evaluated during structural analysis smoothed tensile stress in the concrete with local fluctuation in
(more specifically, section analysis for deriving the moment– the longitudinal direction removed. Here, the variation of the
curvature curve) with the cracks smeared so that the flexural mean tensile stress with theoretical tensile strain at a beam
stiffness of the beam section changes smoothly with the mean section is obtained from FE analysis by considering a short
curvature. segment of the cracked beam starting and ending at a half
crack spacing to each side of the beam section. The short
In theory, the theoretical tensile strain may be evaluated segment has the same length as the crack spacing and is
divided into as many sections as possible. First, the variation of
600 tensile stress with theoretical tensile strain is plotted for each
rt ⫽ 2·0% section. Then the mean tensile stress at each theoretical tensile
500 rt ⫽ 1·5%
rt ⫽ 1·0% strain level is evaluated as the root-mean-square of the tensile
400 rt ⫽ 0·5% stress values of the sections at the theoretical tensile strain
Load: kN

level being considered. Root-mean-square is adopted as the

300
mean because this will result in the same strain energy after
200 stress smoothing.

100 The variation of the mean tensile stress at the mid-span section
0
with the theoretical tensile strain so obtained by the above
0 3 6 9 12 15 18 21 24 27 method for beams with rt ¼ 0.5% is plotted in Figure 6. The
Mid-span deflection: mm
figure shows that, before cracking (applied load ¼ 15% of peak
(a)
load), the mean tensile stress–theoretical tensile strain curve of
1200 each beam is basically a straight line ascending to a certain
rt ⫽ 2·0% peak value not higher than the tensile strength of the concrete.
1000 rt ⫽ 1·5%
rt ⫽ 1·0% When the beam has cracked slightly (applied load ¼ 20% of
800 rt ⫽ 0·5% peak load), the linear ascending branch of the curve reaches a
Load: kN

slightly lower peak value than before and a short descending

600 branch emerges. When the beam has cracked more extensively
400 (applied load > 30% of peak load), the ascending branch
becomes non-linear and reaches a peak value approximately
200 equal to half the tensile strength of concrete while the
descending branch extends to form a tail ending at a fairly
0
0 3 6 9 12 15 18 21 24 27 30 large theoretical tensile strain.
Mid-span deflection: mm
(b)
In general, tension stiffening is most significant when the beam
is under the servicing condition, that is when the applied load
Figure 5. Load–deflection curves: (a) beams subjected to
is equal to about 30–60% of the peak load. Within this range
point load at mid-span; (b) beams subjected to uniformly
distributed load of applied load, the mean tensile stress–theoretical tensile
strain curve changes slightly with the stress state. Nevertheless,

24 Structures and Buildings 163 Issue SB1 Tension stiffening in concrete beams. Part 1: FE analysis Ng et al.
 3·0 2·0
15% of peak load rt ⫽ 2·0%
2·5

Mean tensile stress: MPa

Mean tensile stress: MPa
20% of peak load rt ⫽ 1·5%
30% of peak load 1·5 rt ⫽ 1·0%
2·0 40% of peak load rt ⫽ 0·5%
50% of peak load
1·5 60% of peak load 1·0

1·0
0·5
0·5

0 0
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500

Theoretical tensile strain: με Theoretical tensile strain: με

(a) (a)

2·0
3·0
15% of peak load rt ⫽ 2·0%

Mean tensile stress: MPa

2·5 rt ⫽ 1·5%
Mean tensile stress: MPa

20% of peak load

1·5 rt ⫽ 1·0%
30% of peak load
2·0 40% of peak load rt ⫽ 0·5%
50% of peak load
60% of peak load 1·0
1·5

1·0
0·5
0·5

0
0 0 500 1000 1500 2000 2500
0 500 1000 1500 2000 2500
Theoretical tensile strain: με
Theoretical tensile strain: με (b)
(b)

Figure 7. Averaged curves of mean tensile stress plotted

Figure 6. Curves of mean tensile stress plotted against
against theoretical tensile strain: (a) beams subjected to point
theoretical tensile strain: (a) beams subjected to point load at
load at mid-span; (b) beams subjected to uniformly distributed
mid-span (rt ¼ 0.5%); (b) beams subjected to uniformly
load
distributed load (rt ¼ 0.5%)

needed because the mean tensile stress–theoretical tensile

the change is not large and, for simplicity, the four curves for strain curve changes as the concrete cracks or, in other words,
applied loads equal to 30, 40, 50 or 60% of the peak load may the curve obtained after cracking is not quite the same as that
be combined into a single curve for the post-crack state. In this before cracking. To deal with this phenomenon, the following
study, the single curve for the post-crack state was obtained by strategy is adopted.
arithmetic averaging of the aforementioned four curves. Such
averaging was repeated for the other concrete beams studied Figure 6 shows that before cracking, the mean tensile stress–
and the curves so obtained are presented in Figure 7. theoretical tensile strain curve is just an ascending straight line
reaching the concrete tensile strength. After cracking, the curve
It is evident from Figure 7 that among the beams with different changes to one with a non-linear ascending branch reaching
rt values under the same type of loading (point load or about half the tensile strength of concrete and a descending
uniformly distributed load), the respective curves for the post- branch with a long tail. The two distinct forms of the curve
crack state are fairly close to each other. Hence, the tension before and after cracking are shown schematically in Figure
reinforcement ratio rt does not appear to have a significant 8(a). During cracking, the form of the curve changes abruptly
effect on the variation of mean tensile stress with theoretical without a smooth transition. Hence, if tensile stress blocks are
tensile strain in the post-crack state. On the other hand, the to be developed for accurate analysis, there must be one for the
type of loading does seem to have a certain effect: the curves pre-crack state and another for the post-crack state. However,
for beams under a point load have lower peaks and longer tails traditional practice has been to adopt one tensile stress block
and the curves for beams subjected to a uniformly distributed for both the pre-crack and post-crack states. Furthermore, from
load have higher peaks and shorter tails. the practical application point of view, it should be more
convenient to use just one tensile stress block in the analysis
4.2. Proposed tensile stress block regardless of the stress state. Therefore, if possible, it is
From the mean tensile stress–theoretical tensile strain curves preferable to combine the distinct tensile stress blocks into one
obtained by FE analysis, a tensile stress block may be derived. that can be applied at both the pre-crack and post-crack states
The numerical procedures for obtaining the mean tensile stress without causing significant errors.
and theoretical tensile strain (explained in the previous section)
are completely general and should also be applicable to fibre- It is proposed to combine the two distinct tensile stress blocks
reinforced concrete beams and beams subjected to combined shown in Figure 8(a) into the one shown in Figure 8(b). The
axial load and bending. However, further manipulation is proposed combined tensile stress block is in the form of a

Structures and Buildings 163 Issue SB1 Tension stiffening in concrete beams. Part 1: FE analysis Ng et al. 25
 Design tensile stress
ft Compression
zone
Stress block before cracking
Tensile stress
Stress block after cracking Neutral axis

f ⬘t
Tension zone

Eco

εct εtu
Tensile strain Beam section Maximum εct ⬍ maximum Maximum
(a) theoretical theoretical theoretical
tensile strain tensile strain tensile strain
⭐ εct ⭐ α2εct ⬎ α2εct
ft
Figure 9. Variation of tension stress field with theoretical
Tensile stress

tensile strain

f ⬘t

Eco
fibre, as shown in Figure 9. If the maximum theoretical tensile
εct εtu strain is smaller than åct , then the tensile stress block
Tensile strain
representing the tension stress field in the tension zone is a
(b)
triangular block. If the maximum theoretical tensile strain is
Figure 8. (a) Distinct stress blocks before and after cracking. larger than åct but smaller than Æ2 åct , then the tensile stress
(b) Proposed combined tensile stress block block representing the tension stress field consists of a
triangular block for the pre-crack tension and a trapezoidal
block for the post-crack tension. Lastly, if the maximum
theoretical tensile strain is larger than Æ2 åct , then the tensile
multi-linear tensile stress–strain curve. It has a linear stress block is a triangular block for the pre-crack tension, a
ascending branch and a linear descending branch. In the triangular block for the post-crack tension and a zero stress
ascending branch, the tensile stress increases linearly from zero block for concrete no longer capable of carrying any tension.
to the tensile strength of the concrete f t at a gradient equal to
the initial elastic modulus of the concrete Eco . In the The values of Æ1 and Æ2 may be determined by referring back
descending branch, the tensile stress decreases linearly from a to Figure 7. From the curves presented in the figure, it has been
certain tensile stress of f t9 to zero at an ultimate tensile strain calculated that for beams subjected to a point load, Æ1 ranges
of åtu . The values of f t9 and åtu are expressed in terms of two from 0.37 to 0.47 with a mean of 0.41 and Æ2 is 17.2–19.7
dimensionless coefficients, Æ1 and Æ2 , as with a mean of 18.2; for beams subjected to a uniformly
distributed load, Æ1 is 0.45–0.55 with a mean of 0.51 and Æ2 is
7a f t9 ¼ Æ1 f t 13.0–15.8 with a mean of 14.2. For simplicity, the minor
effects of the reinforcement ratio may be neglected. Hence, for
beams subjected to a point load, Æ1 and Æ2 may be taken as 0.4
7b åtu ¼ Æ2 åct and 18, respectively, and for beams subjected to a uniformly
distributed load, Æ1 and Æ2 may be taken as 0.5 and 14,
respectively.
where Æ1 is a coefficient smaller than 1, Æ2 is a coefficient
greater than 1 and åct is the tensile strain at which the peak The use of just one tensile stress block for both the pre-crack
tensile stress occurs (note that åct ¼ f t /Eco ). and post-crack states would inevitably introduce errors in the
structural analysis, but it is considered that the errors so
The equations for the above proposed tensile stress block are introduced should be small. At the pre-crack state, the tensile
stress block is exactly the same as the mean tensile stress–
8a ó ¼ E co å for å < åct theoretical tensile strain curve before cracking and thus there
should be no errors introduced. At the post-crack state, the
tensile stress block would overestimate the tensile stresses at
Æ1 f t ðÆ2 åct  åÞ the locations where the theoretical tensile strain is smaller than
8b ó¼ for åct , å < Æ2 åct
ðÆ2 åct  åct Þ åct but these locations should be quite close to the neutral axis
and thus the errors in the bending moment so introduced
should be small, as will be shown in part 2 of this paper (Lam
8c ó ¼ 0 for Æ2 åct , å et al., 2009).

5. CONCLUSIONS
The actual tensile stresses induced in the beam section depend The tension stress fields in the concrete of reinforced concrete
on the theoretical tensile strains in the beam section, especially beams at both the pre-crack and post-crack states have been
the maximum theoretical tensile strain at the extreme tension analysed using an FE method that takes into account:

26 Structures and Buildings 163 Issue SB1 Tension stiffening in concrete beams. Part 1: FE analysis Ng et al.
(a) the cracking and non-linear biaxial behaviour of the Creazza G and Di Marco R (1993) Bending moment–mean
concrete curvature relationship with constant axial load in the
(b) the elastic, plastic and strain-hardening behaviour of the presence of tension stiffening. Materials and Structures
steel reinforcement 26(4): 196–206.
(c) the non-linear bond stress–slip behaviour of the steel Fields K and Bischoff PH (2004) Tension stiffening and
reinforcement–concrete bond. cracking of high-strength reinforced concrete tension
members. ACI Structural Journal 101(4): 447–456.
The results of the FE analysis indicate that, after cracking, the Floegl H and Mang HA (1982) Tension stiffening concept based
tensile stress in concrete varies not only with depth but also in on bond slip. ASCE Journal of Structural Engineering
the longitudinal direction between cracks. On the other hand, Division 108(12): 2681–2701.
the bond between the reinforcement and concrete is always Gilbert RI and Warner RF (1978) Tension stiffening in
transferring tensile stresses to the uncracked concrete, leading reinforced concrete slabs. ASCE Journal of Structural
to a gradual increase in the tensile stress in the concrete with Engineering Division 104(12): 1885–1900.
distance from the nearest crack. However, the tensile force Goodman RE, Taylor RL and Brekke TL (1968) A model for the
induced in the concrete is generally larger than the tensile mechanics of jointed rock. ASCE Journal of Soil Mechanics
force transferred through the bond, indicating that, along with and Foundations Division 94(3): 637–659.
the stress transfer through the bond, the curvature of the beam Guo ZH and Zhang XQ (1987) Investigation of complete
should also induce significant tensile forces in the tension stress–deformation curves of concrete in tension. ACI
zones of the uncracked sections. Materials Journal 84(4): 278–285.
Gupta AK and Maestrini SR (1989) Post-cracking behavior of
For the beams analysed, it was found that, in general, before membrane reinforced concrete elements including tension-
cracking, the mean tensile stress–theoretical tensile strain stiffening. ASCE Journal of Structural Engineering 115(4):
curve is a straight line ascending to the tensile strength of 957–993.
concrete. After cracking, the curve comprises a non-linear Gupta AK and Maestrini SR (1990) Tension-stiffness model for
ascending branch reaching about half the tensile strength and reinforced concrete bars. ASCE Journal of Structural
a descending branch with a long tail. Both the applied load Engineering 116(3): 769–790.
level and the tension reinforcement ratio have little effect on He XG and Kwan AKH (2001) Modeling dowel action of
the mean tensile stress–theoretical tensile strain curve. reinforcement bars for finite element analysis of concrete
However, the type of loading (point load or uniformly structures. Computer and Structures 79(6): 595–604.
distributed load) does have an influence. Although in theory, Hughes BP (2008) Early-age concrete crack control – is EC2
two distinct tensile stress blocks – one for the pre-crack state right or wrong? Structural Engineer 86(15): 32–37.
and another for the post-crack state – should be adopted, it is Jiang DH, Shah SP and Andonian AT (1984) Study of the
proposed for convenience in practical applications to combine transfer of tensile forces by bond. ACI Journal 81(3): 251–
the two tensile stress blocks into one that can be applied at all 529.
crack states. Based on the numerical results obtained, a Kaklauskas G and Ghaboussi J (2001) Stress–strain relations
combined tensile stress block defined in terms of two for cracked tensile concrete from RC beam tests. ASCE
parameters, Æ1 and Æ2 , which are dependent on the type of Journal of Structural Engineering 127(1): 64–73.
loading, was derived. Its validity and applicability to section Kupfer HB and Gerstle KH (1973) Behaviour of concrete under
and member analysis will be demonstrated in part 2 of this biaxial stresses. ASCE Journal of Engineering Mechanics
paper (Lam et al., 2009). Division 99(4): 853–866.
Lam JYK, Ng PL and Kwan AKH (2009) Tension stiffening in
REFERENCES reinforced concrete eams. Part 2: section and member
ACI (American Concrete Institute) (2008) Building Code analysis. Proceedings of the Institution of Civil Engineers,
Requirements for Structural Concrete and Commentary, Structures and Buildings 163(1): 29–39
Manual of Concrete Practice. ACI, Farmington Hills, MI, ACI Liauw TC and Kwan KH (1982) Nonlinear analysis of multi-
318M-08. storey infilled frames. Proceedings of the Institution of Civil
ASCE (American Society of Civil Engineers) (1982) State-of- Engineers, Part 2 73: 441–456.
the-Art Report on Finite Element Analysis of Reinforced Link RA, Elwi AE and Scanlon A (1989) Biaxial tension
Concrete Structures. ASCE, New York. stiffening due to generally oriented reinforcing layers. ASCE
Beeby AW, Scott RH and Jones AEK (2005) Revised code Journal of Engineering Mechanics 115(8): 1647–1662.
provisions for long-term deflection calculations. Proceedings Mander JB (1984) Seismic design of bridge piers. PhD thesis,
of the Institution of Civil Engineers, Structures and Buildings University of Canterbury, New Zealand.
158(1): 71–75. Massicotte B, Elwi AE and MacGregor JG (1990) Tension-
BSI (British Standards Institution) (1985) Structural Use of stiffening model for planar reinforced concrete members.
Concrete, Part 2: Code of Practice for Special Circumstances. ASCE Journal of Structural Engineering 116(11): 3039–
BSI, London, BS 8110. 3058.
Choi CK and Cheung SH (1996) Tension stiffening model for Ng PL (2007) Constitutive modelling and finite element analysis
planar reinforced concrete members. Computers and of reinforced concrete structures. PhD thesis, University of
Structures 59(1): 179–190. Hong Kong.
CEB (Comité Euro-International du Béton) (1993) CEB-FIP Piyasena R, Loo YC and Fragomeni S (2002) Tension stiffening
Model Code 1990: Model Code for Concrete Structures. effects in deflection of reinforced concrete beams. In
Thomas Telford, London. Advances in Mechanics of Structures and Materials (Loo Y-C,

Structures and Buildings 163 Issue SB1 Tension stiffening in concrete beams. Part 1: FE analysis Ng et al. 27
 Chowdry SH and Fragomeni, eds). Swets & Zeitlinger, Lisse, Scott RH (1983) The short-term moment–curvature relationship
pp. 205–210. for reinforced concrete beams. Proceedings of the Institution of
Polak MA and Blackwell KG (1998) Modeling tension in Civil Engineers, Part 2 75: 725–734.
reinforced concrete members subjected to bending and axial Torres Ll, Lopez-Almansa F and Bozzo LM (2004) Tension-
load. ASCE Journal of Structural Engineering 124(9): 1018– stiffening model for cracked flexural concrete members.
1024. ASCE Journal of Structural Engineering 130(8): 1242–1251.
Prakhya GKV and Morley CT (1990) Tension-stiffening and Wollrab E, Kulkarni SM, Ouyang C and Shah SP (1996)
moment–curvature relations of reinforced concrete elements. Response of reinforced concrete panels under uniaxial
ACI Structural Journal 87(5): 597–605. tension. ACI Structural Journal 93(6): 648–657.
Saenz LP (1964) Discussion of ‘Equation for the stress–strain Wu Z, Yoshikawa H and Tanabe T (1991) Tension stiffness
curve of concrete’ by Desayi and Krishnan. ACI Journal model for cracked reinforced concrete. ASCE Journal of
61(9): 1229–1235. Structural Engineering 117(3): 715–732.

What do you think?

To discuss this paper, please email up to 500 words to the editor at journals@ice.org.uk. Your contribution will be forwarded to the
author(s) for a reply and, if considered appropriate by the editorial panel, will be published as discussion in a future issue of the
journal.
Proceedings journals rely entirely on contributions sent in by civil engineering professionals, academics and students. Papers should be
2000–5000 words long (briefing papers should be 1000–2000 words long), with adequate illustrations and references. You can
submit your paper online via www.ice.org.uk/journals, where you will also find detailed author guidelines.

28 Structures and Buildings 163 Issue SB1 Tension stiffening in concrete beams. Part 1: FE analysis Ng et al.