You are on page 1of 10

Engineering Structures 92 (2015) 186–195

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Effect of tension stiffening on the behaviour of reinforced concrete


circular columns under torsion
Tarutal Ghosh Mondal, S. Suriya Prakash ⇑
Department of Civil Engineering, Indian Institute of Technology, Hyderabad, India

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

Article history: Combination of torsion and axial compressive loading frequently occurs in bridge columns during
Received 28 May 2014 earthquake vibrations. However, very few investigations in the past have explored the behaviour of
Revised 4 March 2015 Reinforced Concrete (RC) members under such load combinations. Softened Truss Model (STM) can be
Accepted 5 March 2015
used to predict the torsional response of RC members such as beams and columns. In the present study,
Available online 1 April 2015
STM is used to predict the behaviour of RC circular columns under combined action of torsion and axial
compression and the analytical prediction has been compared with experimental observation. It has been
Keywords:
found that the overall torque–twist curve and in particular the post-peak behaviour of the tested columns
Circular column
Tension stiffening
is not accurately predicted by the original STM. The present study focuses on development of an
Softened truss model improved analytical model for RC circular columns under torsional load conditions. The tensile strength
Torque of concrete being very small, its contribution to torsional strength of RC members was considered
Twist insignificant and was neglected in the original STM. The present investigation revealed that neglecting
Axial compression tension stiffening is the main reason behind the less accurate predictions of STM in the past compared
to test data. The original STM has been extended in this study to include the effect of tension stiffening
(TS) to get an improved prediction. Three different tension stiffening models have been considered for
this purpose. The efficiency of each model in predicting the overall torsional response of RC columns
under combined torsion and axial compression has been validated by comparing the analytical predic-
tions of TS-STM with experimentally observed values. It is observed that exponential tension stiffening
model can best represent the torsional behaviour of circular RC columns.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction load effects. This combination of seismic loading and structural


constraints can result in combined torsion and axial compression
The presence of torsional loading significantly alters the in columns.
behaviour of reinforced concrete (RC) members. Although, a num- Considering the brittle nature of torsional failure of RC
ber of studies have addressed pure flexure, pure shear and pure members and high safety requirement for important buildings
torsion independently, investigations on such loads in combination and bridges, it is essential to accurately predict the torsional beha-
are scarce. Among them, the combination of torsion and axial load viour of such members for the purpose of efficient design of these
is the least studied until now in spite of its occurrence in bridge members. Though STM was developed [1–3] for this purpose, their
columns during earthquakes. The presence of torsion loading is prediction for circular RC columns has not been validated enough
more likely in skewed or horizontally curved bridges, bridges with due to paucity of experimental data. It is found that the inaccurate
unequal spans or column heights, and bridges with outrigger predictions of STM were found to be due to the total disregard of
bents. Construction of bridges with these configurations is often the tensile strength of concrete. Hence, the objective of this study
unavoidable due to site constraints. In addition, multidirectional is to develop an improved analytical model which can efficiently
earthquake motions (including significant vertical motions), predict the torsional behaviour of circular RC columns. The original
structural constraints due to a stiff decking, movement of joints, STM is modified to include the effect of tensile stiffening in order to
abutment restraints, and soil conditions may also lead to combined obtain an improved prediction of torsional behaviour of RC
members. Three different tension stiffening models have been con-
sidered in the present study. The experimental results obtained
⇑ Corresponding author.
from two large scale circular columns (H/D(3)-T/Mð1Þ-1:32%
E-mail addresses: ce13m1023@iith.ac.in (T.G. Mondal), suriyap@iith.ac.in
(S.S. Prakash).
and H/D(6)-T/M(1)-0.73%) tested in University of Missouri [4–6]

http://dx.doi.org/10.1016/j.engstruct.2015.03.018
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
T.G. Mondal, S.S. Prakash / Engineering Structures 92 (2015) 186–195 187

under torsional loading in presence of axial compression are used warping stresses. Circular RC columns investigated in this study
for validation of the developed model. have different failure behaviour than the non-circular sections
investigated by Karayannis and Chalioris [10]. Three empirical
2. Tension stiffening effect models were proposed by Vecchio and Collins [11], Belarbi and
Hsu [12] and Stevens et al. [13] for tension stiffening behaviour
Shear flow created in an RC section due to the applied torsional of concrete. However, they were empirical expressions developed
moment can be resolved into a diagonal compression and a diago- using test data from 2-dimensional shear panels and do not closely
nal tension. When this diagonal tensile stress in an RC member model the 3-dimensional torsional response of RC members [14].
reaches the tensile strength of concrete, cracks are developed. Hence, a modified set of equations, comprising a linear, a parabolic
When a crack is developed, redistribution of stress takes place and an exponential relationship, were proposed by Greene [14]
(Fig. 1). Load carried by concrete at the crack location is transferred based on regression analysis of torsional test data which is used
to the reinforcement crossing the crack. The reinforcement in turn in this investigation. The exponential model was earlier used by
transfers the load to the intact concrete between the adjacent You and Belarbi [15] under torsional load condition but was not
cracks through bond stress at steel concrete interface. Therefore, verified with experimental data. The investigation by You and
the concrete in between the cracks continues to carry tensile load Belarbi [15] focused on back-calculating the shear flow thickness
and offers stiffness. This phenomenon is termed as tension based on the given experimental data. In the present work, the
stiffening. A realistic STM model that takes into account this objective is to accurately predict the entire torque–twist behaviour
tension stiffening effect needs to be formulated for better predic- by including tensions stiffening relationships. All the three tension
tions. If this effect is disregarded, steel stresses are overestimated stiffening models including linear, parabolic and exponential mod-
and as a result, structural safety is underestimated considerably. els are verified with experimental data. It has been observed that
As load increases, crack formation continues resulting in a drop the exponential model is most reliable in predicting the behaviour
of concrete stress in between the cracks. This process continues of circular RC columns under torsional loading with axial
until the cracks are so close that effective length of intact concrete compression.
between the cracks reduces to such a level that they are incapable
of taking any more tensile stress perpendicular to the crack direc-
tions and the concrete is then considered to be cracked totally. 3. Tension-stiffened softened truss model
Subsequently, the effect of tension stiffening becomes insignificant
and inflicts no influence on the torsional behaviour of RC members. 3.1. Background
The present study shows the significance of tension stiffening
through validation of STM predictions with experimental data on This section describes the development of the existing Softened
circular RC columns tested under combined torsion and axial Truss Model (STM) [16] by considering the effect of concrete
compression. tension stiffening (TS) for improved predictions. The inclusion of
Several investigations [7–9] in the past studied the influence of tension stiffening in the model is important because it allows an
tension stiffening effect on the behaviour of RC members and dif- improved prediction of the service-level twist. A stress–strain
ferent tension stiffening models were proposed. However, none relationship for concrete in tension was taken from literature
of them were verified and validated with test data under torsional [14]. The proposed TS-STM adopts the equilibrium and compatibil-
loading condition. Karayannis and Chalioris [10] presented an ity equations developed for an RC panel under a membrane stress
analytical model for prestressed concrete members under torsional field. The torsional moment and thickness of shear flow are given
loading which takes into account the tension stiffening effect. It is by Bredt’s [17] thin-tube theory. The analysis of any member (hol-
worth mentioning that in cases of higher axial compression, the low or solid) under torsion can be modelled as a thin-tube as
torsional behaviour of RC columns is similar to that of prestressed shown in Fig. 2. The concrete core of a solid member is assumed
concrete elements. However, applicability of the model proposed to not contribute to the torsional resistance. These theories
by Karayannis and Chalioris [10] was limited to plain prestressed commonly assume reinforced concrete member as assemblies of
concrete members or prestressed concrete members with light two-dimensional membrane elements, also called panels,
transverse reinforcement. In addition, shape of the cross section subjected to in-plane shear and normal stresses. Therefore, the
plays a major role on the damage of RC members under torsion behaviour of a reinforced concrete member under torsion can be
due to change in the shear flow characteristics and presence of predicted via the behaviour of membrane elements.

Fig. 1. Distribution of concrete stresses after cracking [7].


188 T.G. Mondal, S.S. Prakash / Engineering Structures 92 (2015) 186–195

t
D
1 q
q
td A
C
a B
q q
1
ll
l

Fig. 2. Torsion in circular column (a) thin tube assumption (b) shear element from thin tube.

3.2. Shear flow zone reinforced concrete membrane elements subjected to torsion with-
out considering contribution from steel reinforcement. However, if
An external torque applied to a reinforced concrete prismatic steel reinforcement is also taken into consideration, which is
member is resisted by an internal torque formed by a circulatory aligned in the l (longitudinal) and t (transverse) directions, the
shear along the periphery of the cross section (Fig. 2a). In general, in-plane stress condition developed in the membrane element is
the shear stress values are more at the periphery of the section. as shown in Fig. 3 and its formulations are explained in detail in
Owing to such variation of shear stress, it can be stated that shear the following section.
stress resisted by core of the section is small compared to that
resisted by the outer part of the section. Thus, the shear flow ‘q’ 3.3. Shear element in shear flow zone
effectively occupies a zone, known as the shear flow zone towards
the periphery of the section, and has a thickness denoted as td Fig. 3 shows stress conditions in reinforced concrete membrane
(Fig. 2a). This thickness td is a variable determined from the equi- element subjected to in-plane stress which made a foundation of
librium and compatibility conditions. It is not the same as the wall basic governing equations for shear. The directions of the longitu-
thickness of a hollow member. Fig. 2b shows stress condition in dinal and transverse steel bars are designated as the l and t axes,

Fig. 3. RC membrane element subjected to in-plane stresses [1,3].


T.G. Mondal, S.S. Prakash / Engineering Structures 92 (2015) 186–195 189

respectively, constituting the l–t coordinate system. Accordingly, to some extent. However, it is observed in the past that ignoring
the normal stresses are rl and rt ; and the shear stress is slt After Poisson’s effect in the formulation of STM does not impair the
the development of diagonal cracks, the concrete struts are efficiency of the model significantly [15]. Hence, the original STM
subjected to compression and the steel bars act as tension links, proposed by Hsu [1] is used in this study which ignores the con-
thus forming an apparent truss action. The compression struts finement effect on overall torsional response of RC members.
are oriented in the d-axis, which is inclined at an angle a to the
longitudinal steel bars. This direction is also assumed to be the 3.5.1. Equilibrium equations
direction of the principal compressive stress and strain. Taking The two dimensional equilibrium equations relate the average
the direction perpendicular to the d-axis as the r-axis, we have a internal stresses developed in the concrete (rd and rr ) and in the
d–r coordinate system in the direction of the principal stresses reinforcement (f l and f t ) to the average applied stresses (rl , rt
and strains. The normal principal stresses in the d and r directions and slt ) with respect to the angle of inclination of d-axis to the
are rd and rr respectively. l-axis, a. The torsional moment induced by internal shear stress
can be expressed as Eq. (6).
3.4. Estimation of thickness of shear flow zone
rl ¼ rd cos2 a þ rr sin2 a þ ql f l ð2Þ
An important issue in extending the behaviour of the mem-
brane elements to a 3D member under pure torsion and combined rt ¼ rd sin2 a þ rr cos2 a þ qt f t ð3Þ
loading is in the accurate estimation of the thickness of shear flow
zone, td. It can be back calculated from the observed torque and slt ¼ ðrd þ rr Þ sin a cos a ð4Þ
twist values as shown by You and Belarbi [15]. However, in the
present study a different approach has been adopted. The objective Al
ql ¼ ð5aÞ
of this study is to accurately predict the entire torque–twist beha- po t d
viour by including tensions stiffening relationships. In the original
STM, the shear flow zone is assumed to extend into the member up At
qt ¼ ð5bÞ
to the neutral axis. The concrete inside the neutral axis is acting in std
tension and considered ineffective. Such a member is assumed to
be fully cracked, with the concrete and reinforcement acting as a T ¼ slt ð2A0 td Þ ð6Þ
truss. Although research has shown this to be an effective way to where Al ; At are total cross-sectional areas of longitudinal and trans-
model a fully cracked RC member [16], it does not reflect the beha- verse mild steels and p0 , t d are perimeter of the shear flow zone and
viour of an uncracked member. However, in the present study, the the thickness of the shear flow zone respectively.
developed formulations are valid for uncracked elements as well.
Since the pre-cracking torsional behaviour of RC members under
3.5.2. Compatibility equations
torsion is known to be linear, therefore it is not difficult to plot
The two dimensional compatibility equations relate the average
the variation of torque with twist before cracking from the crack-
strains in l—t co-ordinate system (el , et and clt ) to that (ed and er ) in
ing torsional moment and cracking twist values calculated from
d—r co-ordinate system (principal axes). The angle of twist (h) can
the expressions given by Collins and Mitchell [18]. The expression
be related by geometry to the angle of inclination (aÞ, the thickness
for td0 is given by Eq. (1) [19] and it represents the effective thick-
of shear flow zone (td ) and the compressive strain of the concrete
ness of a thin tube at cracking. In Eq. (1), pc is the perimeter of the
strut (ed ) as shown below.
section and Acp is the area of concrete bounded by pc .
el ¼ ed cos2 a þ er sin2 a ð7Þ
3 Acp
t d0 ¼ ð1Þ
4 pc
et ¼ ed sin2 a þ er cos2 a ð8Þ
The thickness of shear flow zone increases after cracking
according to the increase in torsional moment till the peak point. clt
¼ ðed þ er Þ sin a cos a ð9Þ
2
3.5. Governing equations in TS-STM
h ¼ ðpo =2Ao Þclt ð10Þ
The proposed TS-STM adopts the equilibrium equations devel-
oped for an RC panel under a membrane stress field which are 3.5.3. Constitutive laws
shown in Eqs. (2)–(6). The torque, T, resisted by the section is given 3.5.3.1. Concrete stress strain curves under compression. In-plane
by Eq. (6) for a shear stress flowing around the perimeter of the shear stress developed in a reinforced concrete membrane element
section as described by Bredt’s thin-tube theory [17]. The compati- due to torsional loading leads to diagonal tension and diagonal
bility relationships given by Eqs. (7)–(10), which describe the compression (principal stresses). Concrete being weak in tension,
relationship between the unit twist h, and the shear strain, cLT is cracks form in the direction normal to the diagonal tension direc-
adopted without modification. Finally, constitutive relationships tion. It is observed that peak compressive stress and strain in
are used for the concrete and reinforcement as shown in diagonal compressive direction are reduced by the presence of
Eqs. (11)–(18). cracking along the diagonal compression. This phenomenon of
Softened truss model assumes that RC members behave as thin reinforced concrete under shear and torsion is known as softening
tubes under torsional loading. An RC member under torsional of concrete. Previous studies have indicated that not considering
loading is considered to be an assembly of membrane elements softening effect leads to the over-estimation of concrete capacity
which are subjected to bi-dimensional state of stress (shear plus under shear and torsion. Vecchio and Collins [11] incorporated this
normal) and stress in the third direction (inward to the member) softening effect and proposed softened concrete stress–strain
is neglected in the formulation. Any stress acting normal to the curves under compression. The uniaxial constitutive relationship
plane of membrane element (e.g. confinement pressure) may influ- of non-softened concrete is assumed to follow a parabolic model.
ence the in-plane behaviour of the element due to Poisson’s effect The ascending and descending branches of the softened stress
190 T.G. Mondal, S.S. Prakash / Engineering Structures 92 (2015) 186–195

strain relationship are derived from this parabolic model are 1.4
shown by Eqs. (11a) and (11b). Linear
"     #   1.2 Parabolic
0 ed ed 2 ed
rd ¼ ff c 2  61 ð11aÞ 1.0
Exponential
fe0 fe0 fe0

"  2 #   0.8
0 ed =fe0  1 ed
rd ¼ ff c 1 61 ð11bÞ 0.6
4=f  1 fe0
0.4
where rd and ed are concrete compressive stress and strain
0
respectively. f c and e0 are peak values of stress and strain (assumed 0.2
to be 0.002 [11]) of concrete under compression. f is peak soften-
ing coefficient which is quantitatively a measure of diagonal tensile 0.0
cracking of concrete. Hsu [2] proposed the following expression for 0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007
f in terms of diagonal tensile strain er . Principal Tensile Strain, εr (mm/mm)

0:9 Fig. 4. Normalized tensile stress versus tensile strain data and tension stiffening
f ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð12Þ
1 þ 600er relationships.

3.5.3.2. Concrete stress strain curves under tension. Models of ten- Table 2
sion stiffening are typically linear before cracking as shown in Parameters used in evaluation of cracking torque and twist.
Eq. (13). Parameter Expression

rr ¼ Es er ð13Þ tcr 3=4Ac =pc


GK G 4A
2

po t
o

After cracking, the tensile stress decreases rapidly with further qffiffiffiffiffiffiffiffiffiffiffiffiffi
T cr Ac2 f
increase in tensile strain. The present study considers three differ- p f cr 1 þ f pc
c cr

ent tension stiffening equations proposed by Greene [14] based hcr T cr =GK
torsional test data which are shown in Eqs. (14a), (14b) and
(14c) respectively. The term br is a function of er and is given by
Eq. (16). The kt term in Eq. (14c) is a constant that controls the rate
at which the function decays. The term ecro in Eq. (16) is the strain Longitudinal steel:
at which the tensile stress intersects the strain axis. Also, the f l ¼ Es el el < ely ð17aÞ
parabolic function is tangent to the stress axis at ecro . The selected
values of the constants are summarised in Table 1. Normalized f l ¼ f ly el P ely ð17bÞ
stress–strain data generated from the proposed equations are plot-
Transverse steel:
ted and shown in Fig. 4. The parameters used for estimation of
cracking torque and twist at cracking are provided in Table 2. f t ¼ Es et et < ety ð18aÞ
rr ¼ f cr ð1  br Þ ð14aÞ
f t ¼ f ty et P ety ð18bÞ
rr ¼ f cr ð1  2br þ b2r Þ ð14bÞ Young’s modulus of steel, Es was assumed to be 200 GPa [17]. The
yield strain of steel (ely and ety) was calculated dividing yield
rr ¼ f cr ekt ðer ecr Þ ð14cÞ strength by Es .
qffiffiffiffi
Ag 0 3.5.4. Ao and po
f cr ¼ C fc ð15Þ
Acp Area (A0 ) enclosed by centre line of shear flow zone and the
perimeter (p0 ) of the enclosed area can be expressed as functions
er  ecr of td as shown below
br ¼ ð16Þ
ecr0  ecr pc td
Ao ¼ Ac  þ t2d ð19Þ
where ecr and f cr are cracking strain and cracking stress 2
respectively. po ¼ pc  4t d ð20Þ

3.5.3.3. Stress–strain relationship in mild steel. An elastic–perfectly


3.5.5. Expressions for el and et
plastic material model identical under tension and compression
The angle of inclination of diagonal struts to the member long-
is assumed for steel reinforcements. The stress (f l and f t ) in steel
itudinal axis of the member, a, is explicitly included in the equilib-
can be related to strain (el and et ) before and after yielding as
rium and compatibility equations. However, use of a can cause
shown below.
numerical instability in the iterative solution process. So, it was
eliminated from the governing equations and the following rela-
Table 1 tions were obtained.
Constants used in concrete stress strain curves under tension.
Ao ðed Þðrd þ rr Þ
C ecro kt el ¼ ed þ ð21Þ
po td ðrl  rr Þ þ ðAl f l Þ
Linear 0.500 0.00450 –
Parabolic 0.500 0.00700 – Ao sðed Þðrd þ rr Þ
Exponential 0.500 – 350 et ¼ ed þ ð22Þ
po ½st d ðrt  rr Þ þ ðAt f t Þ
T.G. Mondal, S.S. Prakash / Engineering Structures 92 (2015) 186–195 191

The above equations can be solved simultaneously with the 4.2. Solution procedure
stress–strain relationships for mild steel.
The following procedure was used to solve the proposed
3.5.6. Verification for er Circular TS-STM. The algorithm is summarised as a flow chart in
From Eqs. (7) and (8), sin(a) and cos(a) can be estimated in Fig. 5.
terms of the strain components. Then using sin2(a) + cos2(a) = 1,
inter-relationship among the strain components can be obtained 1. Select a value of ed .
as shown in Eq. (23). 2. Assume a value of er .
3. Assume a value of td .
er ¼ el þ et  ed ð23Þ 4. Calculate f, rd , A0 and p0 from Eqs. (12), (11), (19) and (20)
respectively. Before cracking, calculate rr using Eq. (13).
After cracking, use the tension stiffening models in Eq. (14)
3.5.7. The a can be calculated from the equation
to calculate rr .
From the estimated values of sin(a) and cos(a), angle of crack
5. Calculate el using Eq. (21).
inclination (a) can be expressed in terms of strain components as
shown in Eq. (24).
el  ed
tan2 a ¼ ð24Þ
et  ed

3.5.8. Young’s modulus (Ec ) and shear modulus (G) of concrete


Young modulus of concrete (Ec ) can be calculated from cube
compressive strength (f ck ) of concrete using the relation given in
Eq. (25). The shear modulus (G) is related with Young modulus
(Ec ) and Poisson’s ratio (c) as shown in Eq. (26).
p
Ec ¼ 5000 f ck MPa
ð25Þ
f ck ¼ Compressive strength of concrete in MPa:

G ¼ Ec =ð1 þ 2cÞ
ð26Þ
c ¼ Poisson’s ratio

3.5.9. Parameters at cracking


GK is the torsional stiffness. t cr ; T cr and hcr are the depth of shear
flow zone, values of torque and twist at cracking respectively. Ac is
the area enclosed by the outer perimeter of the concrete cross-
section and pc is the outside perimeter of concrete cross-section.
f pc is applied axial stress in the member. The above parameters
are listed in Table 2.

4. Solution procedure

The proposed Circular TS-STM uses the basic equilibrium and


compatibility equations developed for the original STM. Also, the
material relationships for concrete compression and tension have
been modified as described in this section. The additional equa-
tions needed for calculation is presented here. An efficient solution
procedure for solving the system of equations is also described
here in. Given the dimensions of the cross section, the reinforce-
ment, and the material properties, a ‘displacement controlled’
solution to the equations can be invoked by first selecting ed and
then assuming trial values for er and td . Next, an iterative proce-
dure is used to find values of the assumed variables by solving
the equilibrium equations, compatibility equations, and stress–
strain relationships. This will find one point on torque–twist curve.
Subsequent points are found by varying the selected values of ed
from a near zero to a maximum value that causes compression
failure of concrete. The indeterminacy in the equations is settled
by putting rl ¼ constant;rt ¼ 0 and er and td as variables.

4.1. Initial calculations

The following initial calculations were made for variables that


are constant during the solution process. Calculations of Ac , pc ,
t cr , f cr and Ec were done from the respective equations. Fig. 5. Solution algorithm for circular TS-STM [6].
192 T.G. Mondal, S.S. Prakash / Engineering Structures 92 (2015) 186–195

6. Calculate et using Eq. (22). Fig. 6b. Pure torsion was created by applying equal but opposite
7. Calculate er from Eq. (23). If the difference between the forces with the two actuators. The axial compressive load was
assumed and calculated value of er is not within a tolerable applied by a hydraulic jack on top of the load stubs; this jack
limit, then repeat Steps 2–6 until the convergence is transferred the axial load to the column via seven un-bonded
achieved by assuming different values of td and er . high-strength pre-stressed steel strands. A number of instruments
8. Calculate a, slt , T, clt and h corresponding to one value of ed were used to measure the applied loads, deformations, and internal
using Eqs. (24), (4), (6), (9) and (10) respectively. strains. The axial load was measured by a load cell between the
9. Repeat the process for different values of ed . hydraulic jack and the top of the load stub. Electrical strain gages
10. Stop the procedure when the strains start decreasing in the were attached to the surface of the longitudinal and transverse
longitudinal and transverse reinforcement. reinforcement to measure strains permitting study of the deforma-
tion of reinforcement under different loading conditions. It is
5. Experimental corroboration of proposed model worth mentioning that only two specimens tested under pure tor-
sion were used for the validation of the model presented in this
5.1. Specimen details study.

Two half-scale test specimens (H/D(3)-T/M(1)-1.32% and H/ 5.3. Overall torsional response
D(6)-T/M(1)-0.73%) used in this study are representative of typical
bridge columns. The proposed Circular TS-STM was validated by The torque–twist behaviour of the two columns predicted by
comparing the predicted response to the response observed during Circular TS-STM is shown and compared with experimental data
experiment. The two specimens considered in this study had in Fig. 7. The observed behaviour is similar to the torsional
aspect ratios of 3 and 6 respectively. Aspect ratio of bridge columns response of plain concrete element [21] as well as reinforced nor-
mainly depends on bridge geometry and column diameter and mal and high strength concrete elements [22,23] as observed in
typically varies between 1 and 10 [20]. For typical bridge columns, previous studies. It is evident that the response predicted by the
the transverse reinforcement ratio generally varies between 0.5% three tension stiffening models used in this study is identical until
and 1.5%. The test specimens used for this study had transverse near the peak torque. This is because the three tension stiffening
reinforcement ratios of 0.73% and 1.32% which are well within models are identical before cracking. For large principal tensile
the typical range. Specimen H/D(6)-T/Mð1Þ-0:73% had normal cir- strains, the average tensile stresses predicted by the models are
cular rings, whereas, specimen H/D(3)-T/Mð1Þ-1:32% had spiral as markedly different and therefore, the predicted behaviours are
transverse reinforcement. The longitudinal reinforcement ratio for no longer identical. Though the magnitude of difference is small
the test specimens was 2.1% which also falls within the typical because the effect of tension stiffening is small near the peak. It
range of 2–4% [21] for generic bridge columns. The axial load due is observed that behaviour predicted by exponential model is
to the superstructure dead load to bridge columns generally varies closest to experimental values. Prediction of parabolic model is
between 5% and 10% of the capacity of the columns [20]. Hence, an more or less similar to that of experimental model. However, both
axial load equivalent to 7% of the concrete capacity of the columns the models are less stiff than the observed experimental behaviour.
was applied during testing. The details of the test specimens are They underestimate the torque values before the peak and over-
summarised in Table 3 and the cross section details and reinforce- estimate the same after the peak. The linear model yields identical
ment layouts are shown in Fig. 6. Separate predictions were made prediction as the earlier two up to the peak torque but thereafter it
using the linear, parabolic, and exponential models for tension stif- becomes excessively stiff and underestimate the torque values
fening. The cracking and ultimate torque and twist values obtained considerably. So this model is unacceptable after the peak. The
from the analytical and experimental results, on comparison, behaviours predicted by ignoring the effect of tension stiffening
yielded an appreciable correlation. have also been shown in the same graph to quantify the effect of
tension stiffening. It is found that the torque values are highly
under predicted and the behaviour is unacceptably under stiff in
5.2. Test setup and instrumentation the region between the cracking point and the peak. But after the
peak, the prediction is not significantly different from that of
A brief description of test-setup for torsion tests used for the exponential and parabolic models as the effect of tension stiffening
validation of the developed model is presented here. Cyclic is minimal in this range. It is worth mentioning that the model
torsional loading was generated by controlling two horizontal neglects dowel action and assumes perfect bond between the con-
servo-controlled hydraulic actuators shown schematically in crete and reinforcement and considering bond–slip behaviour is a
scope for further work.
Table 3
Specimen details. 5.4. Peak response
Specimen ID/parameters H/D(3)-T/ H/D(6)-T/
Mð1Þ-1:32% M(1)-0.73% A comparison of the predicted and experimental values of ulti-
mate torque and twist is shown in Table 4. It is observed that all
Section shape Circular Circular
Diameter (mm) 610 610 the four (including the zero TS model) tension stiffening models
Clear cover (mm) 25 25 used in this study predicts the ultimate torque values with fair
Total column height (m) 2.74 4.55 accuracy. This is due to the fact that effect of tension stiffening is
Effective column height (m) 1.83 3.65 minimum at this stage of loading and so the ultimate strength is
Cylinder strength of concrete (MPa) 27.97 37.90
not improved significantly. However, the ultimate twist values pre-
Longitudinal reinforcement 12–25.4 mm 12–25.4 mm
Transverse reinforcement 12.5 mm @ 9.5 mm @ 76 dicted by all the three models are slightly overestimated mostly.
70 mm o.c. o.c.
Longitudinal steel yield strength (MPa) 462 462 5.5. Strains in longitudinal and transverse reinforcements
Transverse steel yield strength (MPa) 457 457
Axial force (kN) 600.51 600.51
Young’s modulus of steel (MPa) 200,000 200,000
Fig. 8 shows the variation in strain in the longitudinal and the
transverse reinforcement with applied torque for the two test
T.G. Mondal, S.S. Prakash / Engineering Structures 92 (2015) 186–195 193

Strong Wall Load Cell


Hydralic Jack
Load Stub

Two Hydraulic
Actuators
Steel Strands
(Inside Column)

Test Column

X
Support Blocks Anchors
Y
Strong Floor Z (out)

(a) Cross Sectional Details (b) Test Setup


Fig. 6. Cross sectional details and test setup.

350
400
350 300

300 250
Torque (kN-m)

Torque (kN-m)

250
200
200
150 Experimental
Experimental
150 Exponential
Exponential
100 Parabolic
100 Parabolic
Linear
Linear 50
50 Zero TS
Zero TS
0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12 14 16 18
Twist (Degree) Twist (Degree)
(a) H/D(3)-T/M(∞)-1.32% (b) H/D(6)-T/M(∞)-0.73%

Fig. 7. Comparison of efficiency of different TS model in predicting torque–twist behaviour.

Table 4
Comparison of predictions at ultimate torque and twist.

Specimen Parameter Experimental Linear model Parabolic model Exponential model Zero tension stiffening case
H/D(3)-T/Mð1Þ-1.32% Tu (kN m) 327.54 334.93 340.3 347.98 346.16
hu (deg) 3.23 4.66 4.77 5.046 6.36
Tu,calc/Tu,test NA 1.02 1.04 1.06 1.06
hu,calc/hu,test NA 1.44 1.48 1.56 1.97
H/D(6)-T/Mð1Þ-0.73% Tu (kN m) 281.20 262.36 311.10 298.62 294.56
hu (deg) 6.11 3.87 9.04 7.26 7.73
Tu,calc/Tu,test NA 0.93 1.11 1.06 1.05
hu,calc/hu,test NA 0.63 1.48 1.19 1.27

specimens. The prediction of all the three tension stiffening models The strain in both kinds of reinforcement is tensile bringing
closely follow the experimental measurements. The behaviour about an apparent truss action. It can be inferred from the graphs
shows an initial steep slope as torque increases without significant that any of the three tension stiffening models can be used inter-
increase in stress in the reinforcements. This is due to the shear changeably to represent the state of stress in the reinforcements.
flow in concrete resisting most of the applied torque. The response Though the model developed by ignoring the effect of tension stif-
is effectively linear before cracking and subsequently it flattens fening overestimates the level of strain in the reinforcements. It is
out. A similar behaviour was observed by Valipour and Foster further observed that transverse steel yields prior to longitudinal
[24] in force based formulation of reinforced concrete framed steel which can be attributed to lower percentage of transverse
element under torsion. steel in the specimens compared to longitudinal steel.
194 T.G. Mondal, S.S. Prakash / Engineering Structures 92 (2015) 186–195

400 350
350 300

Torque (kN-m)

Torque (kN-m)
300
250
250
200 Experimental
200 Experimental
150 Exponential
150 Exponential
Parabolic
100 Parabolic 100
Linear
Linear 50
50 Zero TS
Zero TS
0 0
0.0000 0.0005 0.0010 0.0015 0.0000 0.0005 0.0010 0.0015
Average Strain in Longitudinal Reinforcement Average Strain in Longitudinal Reinforcement

400 350
350 300
Torque (kN-m)

Torque (kN-m)
300
250
250
Experimental 200 Experimental
200
Exponential 150 Exponential
150
Parabolic 100 Parabolic
100 Linear Linear
50 Zero TS 50 Zero TS
0 0
0.000 0.002 0.004 0.006 0.000 0.004 0.008 0.012
Average Strain in Transverse Reinforcement Average Strain in Transverse Reinforcement
H/D(3)-T/M(∞)-1.32% H/D(6)-T/M(∞)-0.73%

Fig. 8. Efficiency of different TS model in predicting average strain in reinforcements.

6. Summary and conclusions However, circular bridge columns are subjected to shear and
flexure during earthquake excitations. Experiments in the past
The improved model discussed in this study includes the effect [25,26] has shown that flexural strength decreases more rapidly
of concrete tension stiffening to provide a continuous prediction of with torsional loading. Also, from a flexural design perspective,
torsional behaviour both before and after cracking. In this model, the transverse reinforcement requirements for confinement of con-
the alignment of cracks rotate to remain normal to the principal crete may not be adequate in the presence of torsional loading.
tensile stress and the contribution of cracked concrete in shear Thus, a significant understanding is still required for rational
through aggregate interlocking is neglected. The model has been model development owing to large number of variables involved
validated by comparing the predicted and experimental behaviour in the behaviour of RC circular columns under combined loading.
of circular columns tested under torsion with axial compression. This is a scope for future work.
The following conclusions can be derived based on the compar-
ison of predictions of Circular TS-STM and experimental data: Acknowledgements

 Validation of analytical predictions with experimental data Experimental data used in this study was carried out as a part of
shows that tension stiffening has a significant influence in the a project funded by NEES-NSF-NEESR, USA, and the National
torque–twist response of an RC circular member under com- University Transportation Centre, and the Intelligent Systems
bined torsion and axial compression. Centre of Missouri S&T, USA. Their financial support during the
 The exponential tension stiffening model gives the best predic- PhD work of second author is gratefully acknowledged. The
tion of torsional behaviour of circular RC columns. Parabolic analytical work is as part of the project funded by SERB,
model can also be used without significant loss in accuracy. Department of Science and Technology, India. Their financial
However, the linear tension stiffening model failed to predict support is also gratefully acknowledged.
the torsional behaviour particularly in the post peak region.
 The tension stiffening models considered in this study are found References
to be efficient in predicting the peak torque values precisely but
somewhat lacking in predicting the peak twist values with same [1] Hsu TTC. Softened truss model theory for shear and torsion. ACI Struct J
1988;85-S56:624–35.
degree of accuracy and hence they need improvements.
[2] Hsu TTC. Nonlinear analysis of concrete torsional members. ACI Struct J
 Both the specimen used for the validation of the models had the 1991;88-S69.
following failure progression: shear cracking, transverse [3] Hsu TTC, Mo YL. Unified theory of reinforced concrete. Library of congress
reinforcement yielding, longitudinal bar yielding and diagonal cataloguing. CRC Press; 1993. p. 257–87.
[4] Suriya Prakash S, Belarbi Z, Ayoub A. Behaviour of RC bridge columns under
compression failure of concrete. This failure progression was combined cyclic bending and torsion with moderate shear. In: Proceedings of
adequately captured by the developed circular TS-STM. the 14th world conference on earthquake engineering, October 2008, Beijing,
 Comparison of strain measurement has shown that all the three China. CD-ROM Version, Paper ID 05-02-0083.
[5] Belarbi A, Ayoub A, Greene G, Suriya Prakash S, Mullapudi R. Seismic
tension stiffening models used in this study are capable of pre- performance of RC bridge columns subjected to combined loadings including
dicting reinforcement strains with sufficient accuracy. torsion. In: Proceedings of ASCE structures congress, May 2007, Long Beach,
California, USA.
[6] Prakash S. Seismic behavior of RC circular columns under combined loading
The scope of the work was limited to torsional behaviour of cir- including torsion. PhD thesis, Department of Civil Engineering, Missouri
cular bridge columns in combination with axial compression. University of Science and Technology, Missouri, USA; 2009.
T.G. Mondal, S.S. Prakash / Engineering Structures 92 (2015) 186–195 195

[7] Choi C-K, Cheung S-H. Tension stiffening model for planar reinforced concrete [16] Hsu TTC, Mo YL. Softening of concrete in torsional members – theory and tests.
members, vol. 59. Great Britain: Computers & Structures Elsevier Science Ltd.; J Am Concr Inst Proc 1985;82(3):290–303.
1996. p. 179–90. [17] Bredt R. Kritische Bemerkungen zur Drehungselastizitat. Zeitschrift des
[8] Elenas A, Vasilidis L, Pouliou E, Emmanouilidou N. Influence of tension Vereines Deutscher Ingenieure 1986; Band 40, No. 28. p. 785–90, No. 29. p.
stiffening effect on design and behaviour of reinforced concrete structures. 813–17.
In: Measuring, monitoring and modelling concrete properties, an [18] Collins MP, Mitchell D. Prestressed concrete structures. Response Publications
international symposium dedicated to professor SP. Shah, Springer; 2006. Canada; 1991. 766 pp.
p. 215–20. [19] ACI 318-08. Building code requirements for reinforced concrete and
[9] Stramandinoli RSB, Rovere HLL. An efficient tension stiffening model for commentary. American Concrete Institute, Detroit, MI, USA; 2008. 430 pp.
nonlinear analysis of reinforced concrete members. Eng Struct 2008;30: [20] Lehman DE, Moehle JP. Performance based seismic design of reinforced
2069–80. concrete bridge columns. In: 12th World earthquake engineering conference,
[10] Karayannis CG, Chalioris CE. Strength of prestressed concrete beams in torsion. January 2000, New Zealand. p. 215–23.
J Struct Eng Mech 2000;10(2):163–80. [21] Karayannis CG, Chalioris CE. Experimental validation of smeared analysis of
[11] Vecchio FJ, Collins M. The modified compression-field theory for plain concrete in torsion. J Struct Eng ASCE 2000;126(6):646–53.
reinforced concrete elements subjected to shear. ACI Struct J 1986; [22] Jeng CH, Hsu TTC. A softened membrane model for torsion in reinforced
83(2):219–31. concrete members. Eng Struct 2009;31:1944–54.
[12] Belarbi A, Hsu TTC. Constitutive laws of concrete in tension and reinforcing [23] Bernardo LFA, Andrade JMA, Lopes SMR. Softened truss model for reinforced
bars stiffened by concrete. ACI Struct J 1994;91(4):465–74. NSC and HSC beams under torsion: a comparative study. Eng Struct
[13] Stevens NJ, Uzumeri SM, Collins MP. Analytical modelling of reinforced 2012;42:278–96.
concrete subjected to monotonic and reversed loadings, vol. [24] Valipour HR, Foster SJ. Nonlinear reinforced concrete frame element with
87(1). Department of Civil Engineering, University of Toronto; 1987. torsion. Eng Struct 2010;32(4):988–1002.
[14] Greene GG. Behaviour of reinforced concrete girders under cyclic torsion and [25] Prakash S, Belarbi A, You Y-M. Seismic performance of circular RC columns
torsion combined with shear: experimental investigation and analytical subjected to axial force, bending and torsion with low and moderate shear.
models. PhD thesis, Department of Civil, Architectural, and Environmental Eng Struct 2010;32:46–59.
Engineering, University of Missouri, Rolla; 2006. 227 pp. [26] Tirasit P, Kawashima K. Seismic performance of square reinforced concrete
[15] You YM, Belarbi A. Thickness of shear flow zone in a circular column under columns under combined cyclic flexural and torsional loadings. J Earthquake
pure torsion. Eng Struct 2011;33:2435–47. Eng 2007;11:425–52.

You might also like