Construction and Building Materials 126 (2016) 812–822

Contents lists available at ScienceDirect

Construction and Building Materials

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

Constitutive model for confined ultra-high strength concrete in steel

tube
Yan-Bo Wang a,b, J.Y. Richard Liew a,c,⇑
a
Department of Civil & Environmental Engineering, National University of Singapore, Singapore 117576, Singapore
b
College of Civil Engineering, Tongji University, Shanghai 200092, China
c
College of Civil Engineering, Nanjing University of Technology, Nanjing 211800, China

h i g h l i g h t s

A constitutive model for confined ultra-high strength concrete is proposed.

The dilation angle model is a function of lateral pressure and plastic deformation.

The hardening/softening rule is dependent on pressure and plastic deformation.

The interaction between the steel tube and the confined concrete can be predicted.

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

Article history: With the development of material science and production technology, ultra-high strength concrete
Received 15 February 2016 (UHSC) with uniaxial compressive strength up to 200 MPa has been made available commercially and
Received in revised form 13 September used for concrete filled steel tubular (CFST) columns in high-rise buildings. Finite element analysis is a
2016
necessary tool to analyze CFST columns, but its accuracy depends on the generic constitutive model
Accepted 21 September 2016
for the confined UHSC in the steel tube. This paper proposes a new constitutive model for confined
UHSC based on (1) a yield criterion which is a function of hydrostatic pressure and lode angle, (2) a
non-associated flow rule with a dilation angle that is a function of the confining pressure and the equiv-
Keywords:
Ultra-high strength concrete
alent plastic strain, and (3) a hardening/softening rule which is dependent on the confining pressure and
Constitutive model the equivalent plastic strain. The parameters of the proposed constitutive model are calibrated by a series
Confinement effect of uniaxial compression, biaxial compression and triaxial compression tests of UHSC specimens. The con-
Concrete filled steel tubular column stitutive model is then implemented in ABAQUS and verified by the test results of short CFST columns.
Numerical simulation Comparison of the test and predicted results in terms of compression load versus axial strain and lateral
strain curves demonstrates that the proposed model can predict accurately the maximum resistance of
the stub CFST columns as well as the interactive behavior between the steel tube and the confined con-
crete core.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction pressive strength and ductility of concrete are enhanced owing to

Concrete filled steel tubular (CFST) columns have been increas-
ingly used in high-rise buildings, offshore structures and large span
structures, as they can take full advantages of both steel and con-
crete, such as high strength, high stiffness and good ductility. Due
to the synergistic interaction between concrete core and steel tube,
the best usage of construction materials can be achieved. The com-
⇑ Corresponding author at: Department of Civil & Environmental Engineering,
National University of Singapore, Singapore 117576, Singapore.
E-mail addresses: 2ybwang@tongji.edu.cn (Y.-B. Wang), ceeljy@nus.edu.sg (J.Y.R.
Liew).
the confinement effect provided by steel tube, while the inward
local buckling of the steel tube is prevented by the concrete core.
With the recent advances in material science and production tech-
nology, the concrete of higher compressive strength over 120 MPa
and up to 200 MPa has been made available commercially. Charac-
terized by the extremely high uniaxial compressive strength, this
new type of material is denoted as ultra-high strength concrete
(UHSC) in the literatures [1,2]. UHSC has been employed mainly
in columns for high-rise building construction to improve the
structural efficiency in term of smaller strength to column size
ratio [1,3]. Finite element (FE) method has been widely used to
generate additional data of CFST columns and their accuracy was

http://dx.doi.org/10.1016/j.conbuildmat.2016.09.079
0950-0618/Ó 2016 Elsevier Ltd. All rights reserved.
 Y.-B. Wang, J.Y.R. Liew / Construction and Building Materials 126 (2016) 812–822
verified by comparing the predicted results with the test data [4–
10]. To simulate the interaction between the steel tube and con-
crete core interface and hence to obtain an accurate prediction of
the behavior of CFST columns, an accurate constitutive model for
the confined ultra high strength concrete is necessary. However,
an accurate constitutive model that can predict the failure of UHSC
under tri-axial stresses is currently unavailable. Most of the
research work on confined concrete model was derived from the
test data of stub CFST columns with normal strength concrete
and high strength concrete [4–10]. This type of model ignored
the fact that UHSC exhibited a more brittle post peak behavior than
that of normal and high strength concretes when it was under
compression [11]. Moreover, the previous numerical work focused
mainly on predicting the compression resistance of CFST columns
instead of investigating the confinement behavior and the failure
mechanism of the confined concrete core in the steel tube. Most
of the proposed constitutive models fail to capture the plastic dila-
tion of the concrete core in CFST columns.
Associated flow rule was adopted in literatures [4–6,8] and non-
associated flow rule with a constant value of dilation angle was
assumed in the literatures [7,9,10,12]. As pointed out by Yu and
Teng et al. [13,14], those assumptions on flow rule are not in line
with the experimental findings that the flow rule of concrete is
dependent on confining pressure and plastic deformation [15].
Therefore, the cross-sectional dilation of concrete cannot be accu-
rately described by the flow rule with a constant value, even for
non-associated flow rule. With those adopted flow rules, it is not
possible to expect a reasonable simulation of the interaction
between the concrete core and steel tube. On the other hand, in
order to simulate the confinement effect on the plastic behavior
of concrete core, uniaxial stress-strain relationship was modified
by matching the FE prediction with the result of the tested CFST
stub columns via trial and error process. This approach leads to
an empirical solution to a particular problem. In this way, the
researchers proposed various empirical stress-strain models for
the confined concrete in CFST columns [5–10]. However, the accu-
racy of such empirical stress-stain model is restricted to the given
specimen subset, since the trial and error method makes no
attempt to generalize the solution (stress-strain model) to other
specimens with varying parameters such as cross sectional shape,
diameter to thickness ratio, width to thickness ratio, steel grade,
etc., out of the prescribed experimental range. Moreover, due to
the deviation of the predicted cross-sectional dilation of concrete
core caused by using inappropriate flow rules, compensations were
implicitly made in the determination process of the uniaxial stress-
strain relationship, which indicated that the proposed stress-strain
model was not able to represent the real behavior of the confined
concrete in the CFST columns.
A review of the existing numerical models of CFST stub column
revealed the limits of the current constitutive model and high-
lighted the need to develop a comprehensive and generic constitu-
tive model for confined UHSC. The measurements of cross-
sectional dilation of the concrete core and the confining pressure
exerted on the interaction surface between concrete core and steel
tube are unfeasible due to the configuration of CFST specimens. To
this end, the knowledge of concrete under multiaxial stress states
established from uniaxial compression, biaxial compression and
triaxial compression tests can be used in the development of the
constitutive model for steel tube confined concrete. After the
pioneering study by Richart et al. [16], systematic investigations
have been carried out on the behavior of confined normal strength
concrete [17–20]. Several researchers have studied the behavior of
high strength concrete under multiaxial stresses [21–25]. A com-
prehensive review of unconfined and actively confined concrete
test results indicates the stress-strain behavior and lateral strain-
to-axial strain relationship are different from normal strength con-
813
crete to high strength concrete [15,26]. Thus, concerns have been
expressed in the performance of UHSC under uniaxial compressive
and tensile loadings [27,28] and biaxial compressive loading
[29,30].
The present work aims to develop a previously unavailable con-
stitutive model for steel tube confined UHSC, which is of good gen-
erality and accuracy. The paper starts with the discussion on key
characteristics of such constitutive model. Then, the concrete plas-
ticity model is described in term of yield criterion, flow rule, and
hardening/softening rule, in the following section. A type of UHSC
with uniaxial compressive strength over 200 MPa was tested under
triaxial compressive loading with various confinement pressures
from 25 to 400 MPa by the authors [11]. Based on these experi-
mental results the model parameters are calibrated. After that,
the developed constitutive model is implemented in general FE
software ABAQUS by the user-defined subroutine VUSDFLD and
finally the predicted results are verified against the test data of
steel tubes infilled with such UHSC.
2. Key characteristics of a constitutive model for steel tube
confined concrete
Concrete elements under laterally confining pressure can
undergo pronounced inelastic axial deformation prior to reaching
the failure load. A constitutive model based on plasticity theory
is appropriate to describe the material response of such concrete.
Due to the difficulties in measuring the lateral dilation of concrete
and the confining pressure on the interface between the steel tube
and concrete core, the previous constitutive models [5–10] were
developed by matching the predicted axial force-displacement
response of CSFT columns with those from tests via a trial and error
process. This semi-empirical approach, however, provides a good
agreement between the FE prediction and the test results, but it
is not a sufficient condition to establish a generalized constitutive
model.
Fig. 1 shows a CFST column subjected to compression. Under
axial compression, the concrete core shows cross-sectional dila-
tion. At the initial stage of elastic deformation, the inner surface
of steel tube may separate from the concrete because the Poisson’s
ratio of steel is higher than that of concrete. Thus, there is no con-
finement effect in the early stage of loading. With the increasing of
axial load, the formation and propagation of micro cracks in the
core concrete increase the rate of radial dilation. Consequently,
the core concrete gets in contact with the steel tube and pushes
the latter outwards causing radial deformation. This induces hoop
stress, rh , in the steel tube as shown in Fig. 2a. This hoop stress is
counter-balanced by the confining pressure, rcp , exerted on the
concrete core as shown in Fig. 2b. Owing to the confinement effect,
the strength and ductility of the confined concrete are enhanced.
Under axial Contact with
compression steel tube
Lateral Response
dilation of from steel
concrete core tube
Hoop
stress
Confining
Shape of yield pressure on )
surface concrete
surface Hardening
and softening
behavior
Fig. 1. Interaction between steel tube and concrete core.
814 Y.-B. Wang, J.Y.R. Liew / Construction and Building Materials 126 (2016) 812–822

cp

cp

t d t

h t h t
(a) Steel tube (b) Concrete core
Fig. 2. Confining action of steel tube to concrete core.

Because of this, the compression resistance of a CFST cross section hardening/softening rule. To calibrate the parameters included in
is generally higher than the sum of that of steel tube and concrete the proposed model, axial stress-strain relationships and lateral
core acting independently. strain-to-axial strain relationships of concrete under triaxial com-
A closer examination of the interaction behavior between steel pression with confining pressure ranging from low to high are nec-
tube and concrete core reveals that the lateral dilation of concrete essary. These relationships cannot be obtained directly from stub
is the key parameter affecting the axial force distribution between column tests. The authors carried out a series of tests on UHSC
the steel and concrete. The concrete confinement effect is affected cylinders with uniaxial compressive strength over 200 MPa,
by the enhanced hardening/softening behavior of concrete. In view including unconfined compression test, hydrostatic compression
of this, the flow rule and hardening/softening rule should be care- test, and triaxial compression test with confining pressure ranging
fully developed to describe the real behavior of the concrete con- from 25 to 400 MPa. The test results published in reference [11] are
fined by a steel tube. Extensive experimental results have shown used here for calibrating the parameters used in the proposed con-
that the influences of confining pressure on the lateral dilation of stitutive model.
concrete and the hardening/softening rule are notable and cannot
be ignored [15,26]. Meanwhile, the flow rule and hardening/soft- 3.1. Yield criterion
ening rule are a function of plastic deformation. Hence, a non-
associated flow rule with a variable value of dilation angle is nec- A yield criterion, which describes the shape of yield surface and
essary to describe the influences of the confining pressure and the the subsequent loading surface in stress space, is used to deter-
resulting plastic deformation. The maximum strength and ductility mine whether a material is undergoing plastic deformation. Vari-
of concrete can be enhanced in different extent as the presence of ous yield functions have been made to describe the pressure-
confining pressure ranging from low to high. The concrete core in a sensitive characteristics of concrete materials, such as the well
CFST stub column is not confined at the initial stage of loading fol- know Drucker-Prager criterion (2 parameters), Ottosen criterion
lowed by low to medium confining pressure in the subsequent load (4 parameters) and William-Warnke criterion (5 parameters).
stages when more and more micro cracks developed in the con- The Drucker-Prager criterion is commonly employed in the numer-
crete core. To this end, various strain hardening/softening defini- ical modelling of confined concrete due to its simplicity compared
tions are needed to fully cover the gradually increased confining with other yield functions. In the material library of the commer-
pressures associated with the entire loading history of CFST stub cial FE software ABAQUS, the Drucker-Prager yield criterion has
column. For the concrete yield criterion, it is understood that the been extended by introducing a new parameter related to J3 in
yield surface should be sensitive to the hydrostatic pressure (con- the yield function to offer different descriptions of tensile meridian
fining and axial pressures) and lode angle (the third invariant of and compressive meridian as the Mohr-Coulomb yield surface,
the deviatoric stress tensor). however, with a smoothed noncircular yield surface in the devia-
The required key characteristics of the proposed constitutive toric plane [31]. This 3-parameter yield function was adopted in
model for steel tube confined concrete can be summarized as the present study. The linear extended Drucker-Prager yield crite-
follows: rion has the form:
pffiffiffiffi I1
(1) A yield criterion which is a function of hydrostatic pressure F ¼ tðKÞ J 2  tan u  k ¼ 0 ð1Þ
(I1/3) and lode angle, F (I1, J2, J3); 3
(2) A non-associated flow rule with a dilation angle b, that is a where
function of the confining pressure and the equivalent plastic pffiffiffi "   pffiffiffi !#
strain, G – F and b – constant; and 3 1 1 3 3 J3
tðKÞ ¼ 1þ  1 ð2Þ
(3) A hardening softening rule which is dependent on the con- 2 K K 2 J 23=2
fining pressure and the equivalent plastic strain, kðrcp ; ^eÞ;
I1 ¼ r1 þ r2 þ r3 ð3Þ
F is the yield function, I1 is the first invariant of the stress tensor,
J2 and J3 are the second invariant and the third invariant of the 1 1h i
deviatoric stress tensor, G is the flow potential function, b is the J2 ¼ sij sji ¼ ðr1  r2 Þ2 þ ðr2  r3 Þ2 þ ðr3  r1 Þ2 ð4Þ
2 6
dilation angle, and k is the hardening/softening parameter.
1
J3 ¼ sij sjk ski ð5Þ
3. Proposed constitutive model for UHSC
3
where ri is the principle stresses, i = 1 represents the axial direction,
In this section, a constitutive model with the required key i = 2 or 3 represents the radial direction, and sij is the deviatoric
parameters is developed in terms of yield criterion, flow rule and stress tensor. The yield surface is defined by the three material
 Y.-B. Wang, J.Y.R. Liew / Construction and Building Materials 126 (2016) 812–822 815

parameters: u = friction angle, which represents the slope of the e

compressive meridian, k = softening/hardening parameter, which 900 fac
ur
is the interception of the compressive meridian to the hydrostatic l ds
800 yie
pressure axis, and K = flow stress ratio, which is the added parame- at e
ter in the extended Drucker-Prager criterion to control the shape of l tim e
700 U fac
the yield surface in the deviatoric plane with 0.778 6 K 6 1. When ur
ds
i el
K = 1, Eq. (2) reduces to the classic Drucker-Prager yield criterion 600
ϕ ly
tia
Ini

Sqrt (J2) (MPa)

with a circular cross-section in the deviatoric plane.
500
To calibrate the three parameters for the yield surface of
ϕ
UHSC, concrete cylinder tests under at least three different mul- 400
tiaxial stress states are needed, where two stress states located
on the compressive meridian and one stress state located on 300
the tensile meridian. For the compressive meridian, uniaxial
compression tests and triaxial compression tests under five dif- 200
ferent confining pressures were adopted here, as summarized Peak point
0 100 Yield point
in Table 1, where f c0 is the unconfined compressive strength k0=41.9
0
and f cc is the confined compressive strength. The compressive 0
meridian for the ultimate yield surface (failure surface) was 0 100 200 300 400 500 600 700
determined through least square fit to the test data, as shown -I1/3 (MPa)
in Fig. 3. The friction angle u is 45.7° with R2 = 0.98. For The ini-
tial yield surface, the same friction angle of 45.7° with a reduced Fig. 3. Compressive meridian of UHSC.
0
sized corresponding 0:3f c0 was assumed. Thus, the interception
of the initial yield surface to hydrostatic pressure axis is
k0 = 41.9. On the other hand, the biaxial compressive test data - 1
obtained by Curbach and Speck [29] were adopted in the deter-
mination of the dependence of the yield surface on the interme-
diate principal stress. It is found that the parameter K for UHSC
K=1
is 0.935. According to Kupfer et al. [17], the parameter K for nor- K=0.935
mal strength concrete is 0.862. Compared with normal strength K=0.862
concrete, the yield surface of UHSC in the deviatoric plane is
nearly a circular, as shown in Fig. 4 in comparison with the sur-
face with K = 1.0 and K = 0.862.

3.2. Flow rule

The radial strain versus axial strain curves of UHSC under

unconfined and confined compression were measured from the - 2 - 3
tests [11], as shown in Fig. 5. According to Fig. 2, the maximum
confining pressure exerted by steel tube can be roughly estimated
by Maxðrcp Þ ¼
2f y
d=t
. For commonly used S355 steel tubes with d/t
ratio from 90 to 20, Maxðrcp Þ is ranging from 11.9 MPa to Fig. 4. Yield surface on the deviatoric plane.

53.6 MPa. Considering the axial resistance contribution from steel

tube at the same time, the confining pressure in CFSTs would be
eij ¼ eeij þ epij ð6Þ
not more than 50 MPa. Thus, only the test results with confining
pressure up to 50 MPa are used in this model, as shown in Fig. 5.
These test results of UHSC cylinders were employed here for the
@G
calibration of the flow rule parameters. In the stage of elastic load- epij ¼ k ð7Þ
@ rij
ing, the elastic strain ee can be calculated from the axial stress,
Young’s modulus and Poisson’s ratio according to the Hooke’s
law. The mean values of the measured Young’s modulus Ec and where
Poisson’s ratio v of UHSC are 56.8 GPa and 0.25, respectively [28]. eij is the strain tensor rij is the stress tensor, k is a non-negative
In the stage of inelastic deformation, the total strain is the summa- scalar parameter, and G is the plastic potential function. As pointed
tion of elastic strain and plastic strain out in Section 2, a non-associated flow rule (G – F) was adopted to

Table 1
Peak stress and yield stress under triaxial compression (Unit: MPa).

rcp 0
f cc 0ðf c0 Þ Ultimate yield surface Initial yield surface
pffiffiffiffi pffiffiffiffi
I1 =3 J2 I1 =3 J2

0 212.0 70.7 212.0 21.2 63.6

25 345.0 131.7 320.0 59.1 102.4
50 418.4 172.8 368.4 97.1 141.2
100 546.9 248.1 444.3 172.9 218.7
200 860.5 420.2 660.5 324.6 373.9
400 1244.5 681.5 844.5 628.1 684.2
816 Y.-B. Wang, J.Y.R. Liew / Construction and Building Materials 126 (2016) 812–822

pffiffiffi !
0.0 3 2ðr3  r1 Þ tan b
de ¼ de
p
2
p
3 ¼ dk pffiffiffiffi þ ð14Þ
2 J2 3 3
σcp= 0 MPa
-0.1
σcp= 25 MPa By using Eqs. (13) and (14), the dilation angle can be expressed
in terms of plastic strains as
σcp= 50 MPa
pffiffiffi p
Radial strain ε2=ε3 (%)

-0.2 1 dep1 þ 2dep3 3 dI1

tan b ¼ ¼ qffiffiffiffi ð15Þ
2 dep1  dep3 2 J p2
-0.3
where

dI1 ¼ dep1 þ dep2 þ dep3

p
-0.4 ð16Þ

and
-0.5 1
½ðdep1  dep2 Þ þ ðdep2  dep3 Þ þ ðdep3  dep1 Þ 
p 2 2 2
dJ2 ¼ ð17Þ
6
0.0 0.5 1.0 1.5 2.0 2.5
Therefore, the dilation angle can be calibrated by the results of
Axial strain ε1 (%) unconfined and confined compression tests through Eq. (15).
Fig. 6 shows the dilation angle versus equivalent plastic strain
Fig. 5. Radial strain-to-axial strain relationships of unconfined and confined UHSC.
curves under different confining pressures. The positive value of
tan b represents volumetric dilation and the negative value of
determine the direction of plastic strain flow, which is in the form tan b represents volumetric contraction. It can be seen from
of Drucker-Prager yield function, Fig. 6 that the dilation angle b is a function of not only the equiv-
alent plastic strain, but also the confining pressure. For the uncon-
pffiffiffiffiffiffiffi I1
G¼ 3J 2  tan b  c ð8Þ fined compression test of UHSC, the initial value of dilation angle is
3 58.3°. With the increase in the equivalent plastic strain, the dila-
where b is the dilation angle and c is constant. As the flow potential tion angle increases to 66.1° at peak load and followed by a brittle
function presents in the form of the partial derivative of stress ten- crash. The UHSC cylinder under the confining pressure of 25 MPa
sor in Eq. (7), there is only one valid parameter, which is the dilation starts with a volumetric contraction with the dilation angle of
angle b. It controls the direction of plastic strain flow. Previous 22.0°. The dilation angle increases with the equivalent plastic
experiments revealed that a constant dilation angle is not adequate strain and reaches the maximum value of 13.1° (volumetric dila-
to describe the plastic dilation of concrete under various confining tion) near the peak load. Then, it turns into volumetric dilation
pressures [15,19,24]. The dilation angle b is a function of confining during softening branch and levels off with the dilation angle of
pressure and plastic deformation as 51.3°. The UHSC cylinder under the confining pressure of
50 MPa has the similar b  ^ep curve as that of rcp ¼ 25 MPa; how-
b ¼ bðrcp ; ^ep Þ ð9Þ ever, with the increased ductility in the softening branch.
In order to simplify the implementation of bðrcp ; ^ep Þ in numer-
where rcp is the confining pressure and ^ep is the equivalent plastic ical simulation, a trilinear dilation angle model based on the nor-
strain. The increment of the equivalent plastic strain is given by, malized axial plastic strain is proposed for monotonic loading, as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi shown in Fig. 7. The three controlling points are the initial point
d^ep ¼ depi depi ¼ dep1 dep1 þ dep2 dep2 þ dep3 dep3 ð10Þ at ep1 =epcc ¼ 0, the peak point at ep1 =epcc ¼ 1:15 and the start point of
the level off branch at ep1 =epcc ¼ 2. The dilation angle of the control-
In the case of uniformly confined concrete, epi can be derived
from Eq. (6) and the Hooke’s law. Noted with r2 ¼ r3 ¼ rcp , we
have

1
ep1 ¼ e1  ðr1  2v r3 Þ ð11Þ
Ec

And
1
ep2 ¼ ep3 ¼ e3  ½ð1  v Þr3  v r1  ð12Þ
Ec
Hence, the axial plastic strain, the radial plastic strain and the
corresponding equivalent plastic strain can be obtained from the
test results by applying Eq. (11), Eq. (12) and Eq. (10). To calibrate
the dilation angle brcp ; ^eÞ, the obtained radial plastic strain-to-axial
plastic strain relationships should be converted into the dilation
angle versus equivalent plastic strain curves under different con-
fining pressures. Substitution of Eq. (8) into Eq. (7) yields the
expression for plastic strain increment,
pffiffiffi !
3 2ðr1  r3 Þ tan b
de p
1 ¼ dk pffiffiffiffi þ ð13Þ
2 J2 3 3
Fig. 6. Dilation angle versus equivalent plastic strain curves of unconfined and
and confined UHSC.
 Y.-B. Wang, J.Y.R. Liew / Construction and Building Materials 126 (2016) 812–822 817

where tan u is the friction angle determined in the calibration

of yield surface in Section 3.1, and r1 and r3 shall be obtained from
the triaxial compression test of concrete cylinders, as shown in
Fig. 8. Based on the triaxial compression test results of UHSC
[11], the k  ^ep curves under various confining pressure up to
50 MPa can be obtained by using Eq. (20) and Eq. (10). In the
numerical simulation, the hardening/softening parameter is usu-
ally inputted as an equivalent uniaxial stress-plastic strain curve
in terms of tabular data. To this end, the hardening/softening
parameter k under different confining pressures is better to be con-
verted into the equivalent uniaxial stress r ^ 1 corresponding to the
case of unconfined compression. In this case, k is a known variable,
and r2 = r3 = 0. Substitution of k, r2 and r3 into Eq. (20) yields the
expression for the equivalent uniaxial stress,
k
r^ 1 ¼ ð21Þ
1  tan3 u

Consequently, the r ^ 1  ep1 curves can be calculated by using Eq.

Fig. 7. Trilinear dilation angle model for unconfined and confined UHSC.
ling points are summarized in Table 2, where rcp =f c0 is the confine-
0
ment ratio, bi , bp and bc are the dilation angles corresponding to
ep1 =epcc ¼ 0; ep1 =epcc ¼ 1:15 and ep1 =epcc ¼ 2, and epcc and epc0 are the strain
at peak stress point under confined and unconfined compression,
respectively. In the case of confining pressure between 0 MPa
and 50 MPa, linear interpolations of the values given in Table 2
should be performed.
3.3. Hardening/softening rule
Hardening/softening rule defines the evolution of the loading
surface due to the development of the plastic strain, including
the hardening part from the initial yield surface to the ultimate
yield surface (failure surface) and the followed softening part
finally up to the residual surface. For simulating the CFST stub col-
umn subjected to monotonic loading, isotopic hardening is
adopted. In the extended Drucker-Prager yield criterion as given
in Eq. (1), k is the hardening/softening function. Experiments on
normal strength concrete, high strength concrete and UHSC has
indicated that the development of the subsequent yielding surface
is sensitive to plastic deformation and confining pressure
[11,19,24,26]. Therefore, the hardening/softening parameter
k ¼ kðrcp ; ^ep Þ
should be calibrated under different confining pressures. Solv-
ing Eq. (1) for k yields
pffiffiffiffi I1
k ¼ tðKÞ J 2  tan u
3 concrete surface are obtained by the analysis of surface-to-
Noted that since r2 = r3 = rcp, Eq. (19) can be rewritten as
tan u
k ¼ ðr3  r1 Þ  ðr1 þ 2r3 Þ
3
(21). In accordance with the tabular inputs in ABAQUS, a piece-
wise linearized r^ 1  ep1 model is proposed, as shown in Fig. 9.
The values of the controlling points in Fig. 9 are given in Table 3
with the equivalent uniaxial stress normalized by the uniaxial
compressive strength. For hardening/softening parameter with
confining pressure between 0 MPa and 50 MPa, linear interpola-
tions of the values given in Table 3 should be performed.
4. Verification of the proposed constitutive model for UHSC
4.1. Implementation
In order to verify the accuracy of the developed constitutive
model for UHSC, the calibrated linear extended Drucker-Prager
yield surface (Eq. (1), u ¼ 45:7 , k0 ¼ 41:9 and K ¼ 0:935) and
the proposed plastic dilation model (Eq. (9) and Table 2) and hard-
ening/softening model (Eq. (18) and Table 3) are implemented in
the commercial FE software ABAQUS through the available user
subroutine interface VUSDFLD of ABAQUS/Explicit. Two field vari-
ables, confining pressure and equivalent plastic strain, are defined
and directly added at the integration points of elements. The FOR-
TRAN source file is programmed to obtain stress and strain status
from material points and pass values to the field variables of the
user subroutine. The real-time simulation of steel tube confined
UHSC can be obtained through incremental approach with suffi-
ciently small time increments, as follows:
(1) For a given time increment, the axial strain increment is
ð18Þ known. The corresponding radial strain is predicted by using
the proposed plastic dilation model (Eq. (9) and Table 2) and
current confining pressure as well as plastic strain (0 for ini-
tial time step).
ð19Þ (2) The hoop stress in steel tube and confining pressure on the
surface contact, based on force equilibrium and radial dis-
placement compatibility between steel tube and concrete
ð20Þ core.

Table 2
Dilation angles of controlling points.

rcp(MPa) rcp
0
f c0
Ec (GPa) v epcc ðepc0 Þ (%) tan bi tan bp tan bc

0 0 60.8 0.25 0.073 1.619 2.251 2.214

25 0.118 74.0 0.28 0.387 0.404 0.228 1.246
50 0.236 72.4 0.29 0.729 0.764 0.131 1.403
818 Y.-B. Wang, J.Y.R. Liew / Construction and Building Materials 126 (2016) 812–822

obtained from tensile coupon test are converted to true stress-

Fig. 8. Measured stress difference-equivalent plastic strain curves of unconfined
and confined UHSC.
strain relationships and employed in the material model for steel
tube. The concrete core of the CFST column is modelled by using
three-dimensional 8-node solid elements (C3D8R) with reduced
integration. The steel tube is modelled using 4-node shell elements
(S4R) with reduced integration. Mesh convergent analysis is per-
formed to find an appropriate mesh size, which could provide
accurate results with reasonable computational times. Stable time
increment based on the fixed time increment is determined by
applying element-by-element estimate. As mentioned in step (3),
the interaction between concrete core and steel tube is simulated
by surface-to-surface contact defined in ABAQUS with the finite
sliding formulation. Penalty contact method is adopted to model
the mechanical constraint between the two sets of surface. A sen-
sitivity study was carried out on the effect of coefficient of friction
and it was found that the maximum compression resistance is not
too sensitive (less than 0.5%) to the change of coefficient of friction
from 0.25 to 0.5. The coefficient of friction is taken as 0.25. The
internal surface of steel tube is allowed to separate from the flank
of the concrete core under tensile force. However, the two con-
tacted surfaces are not allowed to penetrate each other. Two rigid
plates are modelled to simulate the supports of the experiment, as
shown in Fig. 10. All the degrees of freedoms of the bottom and top
rigid plate are fixed except the axial displacement of the top rigid
plate is free, with a displacement-controlled loading is gradually
applied.

4.2. Verification

The proposed constitutive model for steel tube confined UHSC is

Fig. 9. Equivalent uniaxial stress-plastic strain curves of unconfined and confined
UHSC.
(3) With the axial plastic strain and confining pressure obtained
from step (1) and (2), the axial stress of concrete is identified
from the proposed hardening/softening model (Eq. (18) and
Table 3).
(4) The full loading process is completed by repeating step (1) to
(3).
Subsequently, the numerical model of stub CFST column is
developed in ABAQUS/Explicit. Mechanical properties of steel tube
verified by the test results of 8 stub CFST columns reported by
Xiong [3]. The methodology of the verification carried out in the
present paper is different from that the similar test data of the stub
CFST columns are used for both deriving and verifying the confined
concrete model. Since the proposed model is derived from the the-
oretical analysis and experimental calibration by triaxial compres-
sion test, the test results of stub CFST columns are considered
independent from the development of the material model. The
dimension details and material properties of the CFST columns
are summarized in Table 4, where L is the length of the specimen,
D and t are the external diameter and thickness of steel tube, fy is
the yield strength of steel, and d is the steel contribution ratio. Six
batches of UHSC were used in the eight specimens. The concrete
cylinder compressive strength ranges from 149 MPa to 193 MPa.
The steel tubes with nominal yield strengths of 235 MPa and
355 MPa were used in the experiment. The D/t ratio of the speci-
mens ranges from 45.1 to 22.6, which can be classified into class
1 and 3 in accordance with Eurocode 3 [32]. The comparison of
the maximum test load Ftest with the predicted maximum load
FFE by numerical simulation is shown in Table 4. The mean value
of Ftest/FFE is 1.006 with the standard deviation of 0.029. This

Table 3
Hardening/softening parameter for unconfined and confined UHSC.

rcp =f 0c0 ¼ 0 rcp =f 0c0 ¼ 0:118 rcp =f 0c0 ¼ 0:236

e (%)
p
1
^ 1 =f 0c0
r e (%)
p
1
^ 1 =f 0c0
r ep1 (%) r^ 1 =f 0c0
0 0.30 0 0.30 0 0.29
0.002 0.75 0.025 0.74 0.046 0.71
0.013 0.86 0.104 1.07 0.174 1.07
0.035 0.95 0.212 1.24 0.409 1.29
0.073 1.00 0.387 1.31 0.729 1.35
0.429 0.15 0.554 1.26 1.014 1.30
0.513 0.05 0.823 1.03 1.415 1.14
0.857 0.00 1.212 0.85 1.751 1.06
 Y.-B. Wang, J.Y.R. Liew / Construction and Building Materials 126 (2016) 812–822 819

and 25 MPa is adopted in the model for UHSC. To this end, the
residual strength of UHSC under such confining pressure ratio
may be overestimated by the linear interpolation, which is the
main reason for the difference between the predicted and the mea-
sured residual resistance of the specimens with relatively low steel
contribution ratio. This points to the necessary of further investiga-
tion on the behavior of UHSC under triaxial compression with low
confining pressure. In addition to the accurate prediction of the
resistance of the whole CFST specimen, the good agreement
between the test result and numerical prediction in load-lateral
strain curve demonstrates the capability of the proposed model
in providing an accurate prediction of the interaction behavior
between steel tube and concrete core.

4.3. Discussion

4.3.1. Interactions between steel and concrete

Fig. 10. Finite element model of stub CFST column.
indicates that the proposed model can accurately capture the con-
finement effect on the compressive strength of UHSC.
The predicted load-axial strain and load-lateral strain curves are
compared with three typical test specimens with different steel
tube configurations, as shown in Fig. 11. It can be observed that
the predicted initial stiffness, yielding and hardening regions
match well with the experimental curves. Due to the brittle failure
behavior of UHSC, the test load suddenly dropped and gradually
recovered to the residual resistance in the post-peak region, as
shown in Fig. 11. Thus, the softening branch cannot be captured
by the test. In the numerical simulation, the sharp drop after the
peak point is observed in the load-strain curves. It can be seen that
the predicted residual resistances of the specimens S2-2-3, S3-1-2
and S2-1-4 are higher than the test results. Correspondingly, at the
residual strength branch, the maximum confining pressure for the
specimens are 20.9 MPa, 18.6 MPa and 19.7 MPa, respectively. It
should be noted that, due to the lack of test data, linearly interpo-
lation of residual strength with confining pressure between 0 MPa
Numerical simulation revealed the interactions between steel
tube and concrete core. To illustrate the development of confining
pressure, Fig. 12 shows stress-strain curves at the mid-height of
steel tube and concrete in both axial and radial directions, where
horizontal axis points A, B, C and D are the strains corresponding
to the concrete core re-contacts with the steel tube, the initial
yielding of steel tube, the peak resistance of the tested CFST short
column, and the initial hardening of steel tube. In the initial stage
of elastic deformation, there is no confining pressure developed
because the Poisson’s ratio of steel is higher than that of UHSC.
In this stage, the Mises stress of steel is same as its axial stress
since the hoop stress in the steel tube is zero. Under a given axial
strain in inelastic deformation stage, UHSC dilates faster than steel
in the radial direction. Thus, the concrete core gets in contact with
the steel tube again at point A. From point A onward, the dilation of
UHSC will push steel tube outward and cause hoop stress. The
hoop stress will be counter-balanced by the confining pressure
exerted on the concrete core. Thus, the confining pressure
increases with axial deformation and radial dilation. The Mises
stress of steel is composed of axial stress and hoop stress and
reaches yield strength at point B. Due to the forced radial dilation
and the reduced axial stiffness of steel tube, the axial force is redis-
tributed between steel tube and concrete core with increasing axial
deformation. Concrete core achieves the maximum confined
strength at point C when the tested CFST column reaches the max-
imum resistance. In the post peak branch, UHSC shows a sudden
drop in strength while steel tube resists more axial force. Conse-
quently, hoop stress decreases during the yield plateau and turns
to increase at the hardening branch after point D, as well as confin-
ing pressure.
4.3.2. Comparison with existing models
As demonstrated in Section 3.2, the measured dilation angel of
UHSC is a function of confining pressure and plastic deformation.
However, a range of constant dilation angle from 20° to 40°was

Table 4
Configuration details and test results of stub CFST columns.

Specimen L (mm) D  t (mm) Class fy (MPa) 0

f c0 (MPa) Maxðrcp Þ (MPa) d F test (kN) F FE (kN) F test =F FE

S2-1-3 600 219  5 3 380 185.1 19.4 0.232 7837 7968 0.984
S2-1-4 600 219  5 3 380 193.3 19.7 0.224 8664 8197 1.057
S3-1-1 600 219  6.3 1 300 163.0 18.3 0.258 6915 7073 0.978
S3-1-2 600 219  6.3 1 300 175.4 18.6 0.244 7407 7400 1.001
S3-1-3 600 219  6.3 1 300 148.8 17.9 0.276 6838 6609 1.035
S3-1-4 600 219  6.3 1 300 174.5 18.6 0.245 7569 7354 1.029
S2-2-3 600 219  10 1 381 185.1 20.9 0.394 9085 9167 0.991
S2-2-4 600 219  10 1 381 193.3 21.1 0.384 9187 9427 0.975
Mean 1.006
Standard deviation 0.029
820 Y.-B. Wang, J.Y.R. Liew / Construction and Building Materials 126 (2016) 812–822

Fig. 12. Interaction behavior between steel tube and concrete of specimen S2-2-3.

model b ¼ bðrcp ; ^ep Þ, which is dependent on both the confining

pressure and plastic deformation, are used in the numerical simu-
lation, as shown in Fig. 13. The initial value of dilation angle bi at
each confinement ratio is adopted for the pressure sensitive dila-
tion angle model b ¼ bi ðrcp Þ. From the comparison with the pro-
posed model b ¼ bðrcp ; ^ep Þ; it can be seen that the volume
dilation is firstly underestimated by the model b ¼ bi ðrcp Þ up to
over peak plastic strain epc0 or epcc ; as highlighted by the shadow
in Fig. 13. Due to the volume dilation, concrete core is subjected
to the lateral confining by the steel tube. After the peak value at
1:2epcc ; the dilation angle turns to decrease with the increase in
plastic strain. Consequently, the volume dilation is overestimated
by the model b ¼ bi ðrcp Þ during the post-peak regime, as shown
in Fig. 13. This directly results in the inaccurate response from steel
tube (hoop stress) and indirectly deviates the performance of CFST
column via the value of confining pressure and the pressure sensi-
tive hardening/softening parameter. Figs. 14 and 15 show the sim-
ilar trends in the caused hoop stress and the resistance of whole
CFST column. Firstly underestimate and then overestimated, in
the predictions of the hoop stress of steel tube and the resistance
of the whole CSFT column. In the case of the constant dilation angle
model, the volume dilation is initially underestimated under low

Fig. 11. Comparison of predicted and experimental load-axial strain and load-
lateral strain curves.

assumed for steel tube confined concrete in most literatures [4–

10,12]. To investigate the influence of dilation angle on the interac-
tion between steel tube and confined concrete core, numerical
studies with different dilation angle models are carried out. In
the case study of the specimen S2-2-3, a constant dilation angle
with values of 20°, 30° and 40°, a dilation angle model
b ¼ bi ðrcp Þ only pressure sensitive and the proposed dilation angle Fig. 13. Comparison of different dilation angle models.
 Y.-B. Wang, J.Y.R. Liew / Construction and Building Materials 126 (2016) 812–822
Fig. 14. Comparison of the induced hoop stress by different dilation angle models.
Fig. 15. Comparison of the load-axial strain curve with different dilation angle
models.
confining pressure and then overestimated under increased confin-
ing pressure, as shown in Fig. 13. At the axial strain ecc correspond-
ing to peak load, the predicted volume dilation by using b ¼ 40 is
close to that by using the proposed model b ¼ bðrcp ; ^ep Þ. Thus, as
well as b ¼ bðrcp ; ^ep Þ, gives a reasonable prediction on the ultimate
strength of the specimen S2-2-3. In contrast, a constant dilation
angle of 20° and 30° fails to give an accurate prediction due to
the underestimated volume dilation at the peak point, as shown
in Figs. 14 and 15. However, any value of constant angle can only
give an accurate perdition of confining pressure and hoop stress
at an instant moment instead of the full loading process, since
the real variation of dilation angle cannot be simulated in a numer-
ical model by a constant value. It should be noted that the devia-
tion of the predicted resistance, which is caucused by adopting
an inappropriate dilation angle model, could be revised by the
modification of stress-strain relationship of concrete via matching
the test result by trial and error process. In this case, however, the
accurate prediction of the dilation of concrete core, the hoop stress
of steel tube, the confining pressure to concrete core and the stress
status of concrete cannot be expected. The combination of the
adopted dilation angle model and the modified concrete stress-
821
strain curve is limited to predict the resistance of CFST columns
with similar configurations as those columns used in the derivation
of the modified concrete stress-strain curve. Comparatively, the
proposed constitutive in this paper is derived from theoretical
analysis and a series of triaxial compression tests and verified by
the independent stub CFST column tests. It concerns the real inter-
action behavior between steel tube and concrete core. The general-
ity of the proposed constitutive model for steel tube confined UHSC
is highlighted.
5. Conclusion
A generic constitutive model for confined ultra-high strength
concrete (UHSC) with compressive strength greater than 150 MPa
is proposed based on the following three essential formulations:
(1) a yield criterion which is dependent on hydrostatic pressure
(I1/3) and lode angle, F (I1, J2, J3);
(2) a non-associated flow rule with a dilation angle b that is a
function of the confining pressure and the equivalent plastic
strain, G – F and b – constant;
(3) a hardening/softening rule which is dependent on the con-
fining pressure and the equivalent plastic strain, kðrcp ; ^eÞ;
The proposed constitutive model is calibrated against a series of
tests involving UHSC specimens subject to uniaxial compression,
biaxial compression and triaxial compression. The constitutive
model is implemented in the numerical simulation via a user-
defined subroutine VUSDFLD in ABAQUS. Test results of short
tubular steel columns infilled with UHSC are used to validate the
numerical model. Comparison of the test and predicted results in
terms of compression load versus axial strain and lateral strain
curves shows that the proposed model can predict accurately the
maximum resistance of the stub CFST columns as well as the inter-
active behavior between the steel tube and the concrete core.
Moreover, good agreement between the prediction and test results
among the specimens with different configurations and casted
from various batches of concrete shows the generality of the pro-
posed model. However, it should be noted that the linear interpo-
lation of proposed hardening/softening model may cause an
overestimation of the residual resistance of CFST column. Further
investigation on the behavior of UHSC under triaxial compression
with low confining pressure is needed.
Acknowledgements
The authors would like to acknowledge the funding support by
the National Research Foundation Singapore, Sembcorp Industries
Ltd. and National University of Singapore under the Sembcorp-NUS
Corporate Laboratory and by China Postdoctoral Science Founda-
tion under Grant No: 2014M561518.
References
[1] J.Y.R. Liew, D.X. Xiong, Ultra-high strength concrete filled composite columns
for multi-storey building construction, Adv. Struct. Eng. 15 (2012) 1487–1503.
[2] M.-X. Xiong, J.Y.R. Liew, Mechanical behaviour of ultra-high strength concrete
at elevated temperatures and fire resistance of ultra-high strength concrete
filled steel tubes, Mater. Des. 104 (2016) 414–427.
[3] D.X. Xiong, Structural Behaviour of Concrete Filled Steel Tubes With High
Strength Materials, National University of Singapore, Singapore, 2012.
[4] H.B. Ge, T. Usami, Strength analysis of concrete-filled thin-walled steel box
columns, J. Constr. Steel Res. 30 (1994) 259–281.
[5] H.-T. Hu, C.-S. Huang, M.-H. Wu, Y.-M. Wu, Nonlinear analysis of axially loaded
concrete-filled tube columns with confinement effect, J. Struct. Eng. 129 (2003)
1322–1329.
[6] E. Ellobody, B. Young, D. Lam, Behaviour of normal and high strength concrete-
filled compact steel tube circular stub columns, J. Constr. Steel Res. 62 (2006)
706–715.
822 Y.-B. Wang, J.Y.R. Liew / Construction and Building Materials 126 (2016) 812–822
[7] L.-H. Han, G.-H. Yao, Z. Tao, Performance of concrete-filled thin-walled steel
tubes under pure torsion, Thin-Walled Struct. 45 (2007) 24–36.
[8] X. Dai, D. Lam, Numerical modelling of the axial compressive behaviour of short
concrete-filled elliptical steel columns, J. Constr. Steel Res. 66 (2010) 931–942.
[9] Z. Tao, Z.-B. Wang, Q. Yu, Finite element modelling of concrete-filled steel stub
columns under axial compression, J. Constr. Steel Res. 89 (2013) 121–131.
[10] H.-T. Thai, B. Uy, M. Khan, Z. Tao, F. Mashiri, Numerical modelling of concrete-
filled steel box columns incorporating high strength materials, J. Constr. Steel
Res. 102 (2014) 256–265.
[11] Y.-B. Wang, J.Y.R. Liew, S.C. Lee, D.X. Xiong, Experimental study of ultra-high-
strength concrete under triaxial compression, Mater. J. 113 (2016) 105–112.
[12] F. Aslani, B. Uy, Z. Tao, F. Mashiri, Behaviour and design of composite columns
incorporating compact high-strength steel plates, J. Constr. Steel Res. 107
(2015) 94–110.
[13] T. Yu, J.G. Teng, Y.L. Wong, S.L. Dong, Finite element modeling of confined
concrete-I: Drucker-Prager type plasticity model, Eng. Struct. 32 (2010) 665–
679.
[14] T. Yu, J.G. Teng, Y.L. Wong, S.L. Dong, Finite element modeling of confined
concrete-II: plastic-damage model, Eng. Struct. 32 (2010) 680–691.
[15] J. Lim, T. Ozbakkaloglu, Lateral strain-to-axial strain relationship of confined
concrete, J. Struct. Eng. (2014) 04014141.
[16] F.E. Richart, A. Brandtzaeg, R.L. Brown, A Study of the Failure of Concrete under
Combined Compressive Stresses, University of Illinois, Urbana, IL, 1928.
Engineering Experimental Bulletin No. 185.
[17] H. Kupfer, H.K. Hilsdorf, H. Rusch, Behavior of concrete under biaxial stresses,
ACI J. Proc. 66 (1969) 656–666.
[18] J.-C. Chern, H.-J. Yang, H.-W. Chen, Behavior of steel fiber reinforced concrete
in multiaxial loading, ACI Mater. J. 89 (1993) 32–40.
[19] I. Imran, S.J. Pantazopoulou, Experimental study of plain concrete under
triaxial stress, ACI Mater. J. 93 (1996) 589–601.
[20] D. Sfer, I. Carol, R. Gettu, G. Etse, Study of the behavior of concrete under
triaxial compression, J. Eng. Mech. 128 (2002) 156–163.
[21] J. Xie, A.E. Elwi, J.G. MacGregor, Mechanical properties of three high-strength
concretes containing silica fume, ACI Mater. J. 92 (1995) 135–145.
[22] M.M. Attard, S. Setunge, Stress-strain relationship of confined and unconfined
concrete, ACI Mater. J. 93 (1996) 432–442.
[23] D. Candappa, J. Sanjayan, S. Setunge, Complete triaxial stress-strain curves of
high-strength concrete, J. Mater. Civ. Eng. 13 (2001) 209–215.
[24] F. Ansari, Q. Li, High-strength concrete subjected to triaxial compression, ACI
Mater. J. 95 (1998) 747–755.
[25] Q. Li, F. Ansari, High-strength concrete in triaxial compression by different
sizes of specimens, ACI Mater. J. 97 (2000) 684–689.
[26] J.C. Lim, T. Ozbakkaloglu, Stress–strain model for normal- and light-weight
concretes under uniaxial and triaxial compression, Constr. Build. Mater. 71
(2014) 492–509.
[27] B.A. Graybeal, Compressive behavior of ultra-high-performance fiber-
reinforced concrete, ACI Mater. J. 104 (2007) 146–152.
[28] B.A. Graybeal, F. Baby, Development of direct tension test method for ultra-
high-performance fiber-reinforced concrete, ACI Mater. J. 110 (2013) 177–186.
[29] M. Curbach, K. Spek, Ultra high performance concrete under biaxial
compression, in: E. Fehling, M. Schmidt, S. Sturwald (Eds.), Proceedings of
the Second International Symposium on Ultra High Performance Concrete,
Kassel University Press, Kassel, Germany, 2008, pp. 477–484. March 05–07.
[30] T. Leutbecher, E. Fehling, Structural behaviour of UHPC under tensile stress
and biaxial loading, in: M. Schmidt, E. Fehling, C. Geisenhansluke (Eds.),
Proceedings of the International Symposium on Ultra High Performance
Concrete, Kassel University Press, Kassel, Germany, 2004, pp. 435–446.
September 13–15.
[31] ABAQUS Theory Manual USA: Hibbitt, Karlsson & Sorensen Inc.
[32] CEN. Eurocode 3: Design of steel structures, Part 1–1: General rules and rules
for buildings, EN 1993-1-1. Brussels: European Committee for Standardization,
2005.