Engineering Structures 100 (2015) 308325

Contents lists available at ScienceDirect

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

Strength and ductility performance of concrete-lled steel tubular

columns after long-term service loading
Hui Liu a,b, Yongxiang Wang b, Minghua He a, Yongjiu Shi a, Haim Waisman b,
a
Department of Civil Engineering, Tsinghua University, Beijing 100084, PR China
b
Department of Civil Engineering and Engineering Mechanics, Columbia University, New York, NY 10027, USA

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

Article history: Concrete-lled steel tubular (CFST) columns are widely used in infrastructure applications and thus usu-
Received 6 February 2015 ally are subject to long-term service loading. However, understanding the inuence of sustained loading
Revised 23 April 2015 on the ultimate performance of these structural members is still lacking. The objective of this work is to
Accepted 12 June 2015
develop a constitutive model to account for strength and ductility change of CFST columns under sus-
Available online 25 June 2015
tained loading, validated by experimental data reported in the literature.
In this framework, a simplied analytical method equipped with a monolithic iterative scheme is
Keywords:
developed to efciently estimate the creep deformation of these composite columns at any designated
Concrete-lled steel tubular columns
Long-term service loading
target time. Based on the calculated creep status, an analytical stressstrain curve is proposed to charac-
Creep terize the post-creep mechanical behavior of steel-conned concrete. This stressstrain behavior incor-
Stressstrain relationship porates the combined effects of enhanced compressive strength of plain concrete and reduced
Strength conning strength provided by steel tube, both of which are caused by sustained load. Finite element
Ductility based numerical study together with the available test database are used to validate the mechanical anal-
ysis and to assess the performance of the proposed constitutive model. The predicted post-creep response
is found to be in good agreement with the experimental results for CFST columns with circular and square
cross-sections. Finally, an extensive parametric study based on a pushover analysis is conducted to exam-
ine the inuence of individual critical design parameters on structural ultimate strength and ductility due
to long-term service loading.
2015 Published by Elsevier Ltd.

1. Introduction
During the past few decades, concrete-lled steel tubular (CFST)
columns have been increasingly used in a wide variety of structural
engineering applications, particularly in bridges and high-rise
buildings. The popularity of CFST columns can be attributed to
their superior structural properties (e.g. high strength and ductil-
ity, good re resistance, and large energy absorption capacity),
resulting from the combined action of the composite constituent
materials (steel and concrete). The steel tube serves as longitudinal
and transverse reinforcement offering conning pressure to the
concrete inll. On the other hand, the concrete core can stabilize
and stiffen the surrounding steel tube in the sense that the local
buckling of the steel tube is prevented or at least delayed.
Furthermore, this type of structural member has proven to reduce
the labor and material cost during the construction process since
external steel tubes can act as integral and permanent formworks.
Corresponding author. Tel.: +1 212 851 0408; fax: +1 212 854 6267.
E-mail address: waisman@civil.columbia.edu (H. Waisman).
The behavior of CFST columns under short-term loading has been
investigated in depth and well addressed by existing design codes.
The related studies include Furlong [1], Neogi et al. [2], Knowles
and Park [3], Ghosh [4], Rangan and Joyce [5], Ge and Usami [6],
Fujinaga et al. [7], Schneider [8], Han et al. [9], Huang et al. [10],
Sakino et al. [11], Elremaily and Azizinamini [12], and Wheeler
and Bridge [13], to name just a few.
In permanent structures and facilities, CFST columns are usually
subjected to long-term service loads before ultimate failure is
experienced. Compared with the short-term deformation of CFST
columns, the creep and shrinkage of concrete core and its interac-
tion with the surrounding steel tube produce great uncertainties
on the ultimate response of CFST columns with long-term service
loading applied. A few investigations, including both theoretical
and experimental ones, were carried out in order to study the creep
and shrinkage behavior of CFST columns, e.g. Morino et al. [14], Uy
[15], Naguib and Mirmiran [16], Ma and Wang [17], Gunnarsson
et al. [18]. It was found that shrinkage strain in CFST columns
was very small or totally eliminated due to the prevention of mois-
ture egress in a sealed environment [19,20]. In addition, the stress

http://dx.doi.org/10.1016/j.engstruct.2015.06.024
0141-0296/ 2015 Published by Elsevier Ltd.
 H. Liu et al. / Engineering Structures 100 (2015) 308325
redistribution between the concrete core and steel tube could
reduce the creep strain of concrete to some extent [21].
Compared with the research on the time-dependent creep and
shrinkage behavior of CFST columns, models that describe the ulti-
mate mechanical performance of CFST columns after long-term
service load remain relatively undeveloped. According to the
pioneering experimental study conducted by Han et al. [9] and
Wang et al. [22], the circular and square CFST column specimens
under sustained load showed increase of up to 20% in
load-bearing capacities when compared with their companion
load-free specimens. However, an opposite trend, i.e. reduction in
load-carrying capacity, has also been reported by other research-
ers, e.g. Wang et al. [23]. It is still an open question [24] to explain
such divergent changes in structural strength. Aside from the
unclear mechanism, the lack of understanding of the potential
inuence of long-term service loading on structural deformation
capacity (i.e. ductility) also underlines the necessity to conduct fur-
ther research on this important, but generally neglected topic.
This paper is devoted to exploring the mechanism underlying
the experimentally observed strength and ductility change of
CFST columns after long-term service loading. The remainder of
the paper is organized as follows. In Section 2, we establish a
framework for efcient evaluation of creep strain of CFST columns
at any designated target time, imperative for the long-term failure
analysis of CFST columns. In this framework, the efciency of creep
analysis is considerably enhanced by a monolithic iterative
scheme. In Section 3, we show that the post-creep mechanical
behavior of CFST columns depends on the combined action of
enhanced compressive strength of plain concrete and residual
creep deformation. A uniaxial constitutive model, incorporating
the aforementioned effects brought about by long-term service
load, is proposed for steel-conned concrete core with both circu-
lar and square cross-sections. Section 4 reports a nite element
based numerical study on the effect of creep deformation on the
conning strength, which serves to validate the mechanical analy-
sis presented in Section 3. In Section 5, we further apply the pro-
posed stressstrain relation to ber element analysis where the
nonlinear behavior of CFST columns under compression is investi-
gated. By comparison with the available experimental data, our
model is found to perform well for both the circular and square
CFST members. In Section 6, the validated post-creep constitutive
model is used for pushover analysis of CFST columns with
long-term service load applied. A parametric study is conducted
to examine the inuence of critical design parameters on the
post-creep ultimate strength and ductility of CFST column mem-
bers subjected to combined compression and bending loads. Final
concluding remarks are given in Section 7.
2. Creep analysis of CFST columns
The creep deformation analysis of CFST columns includes two
main ingredients, i.e. the initial short-term elastic analysis and
the subsequent long-term creep analysis. The focus in this paper
is on perfectly bonded composite columns, in which the concrete
core and steel tube sustain the axial load together. The combined
effects of concrete creep, lateral connement provided by the steel
tube, and stress redistribution induced by geometric compatibility
(see Fig. 1) make the deformation of CFST columns under sustained
load a complex time-dependent process. Following the framework
proposed by Naguib and Mirmiran [16], we establish hereinafter a
predictive model with all aforementioned effects considered for
the efcient assessment of creep deformation of CFST columns at
any designated target time. However, and in contrast to the previ-
ously proposed framework [16], the creep strain of CFST columns
at any target time can be conveniently obtained through a
309
simplied design method without repeating the complicated iter-
ative algorithm for each loading increment of the whole duration.
In other words, only one iterative process is needed for predicting
the nal status of creep. Another novel contribution of our model is
that a monolithic iterative scheme is developed instead of the stag-
gered one used in [16] to achieve better convergence of creep
iterations.
2.1. Creep model for conned concrete
The mechanical behavior of conned concrete is considerably
different from that of plain concrete, due to the sealed wrapping,
variable stress history, and multi-axial stress status caused by
the steel tube connement. Without loss of generality, we adopt
the B3 model, developed by Bazant and Baweja [25], in the creep
analysis of conned concrete, considering that the B3 model is sim-
ple yet well justied. Any other creep model, such as the CEB
model [26], the ACI 209 model [27] or the B4 model [28,29] intro-
duced very recently, can equally be employed in our framework.
Although full details of the B3 model can be found in Bazant and
Baweja [25], a brief description with the relevant equations is pro-
vided for clarity.
Since the moisture migration of the concrete core is prevented
by the surrounding steel tube [30], we neglect the drying creep
and shrinkage strain of the concrete inll in our study.
Furthermore, the concrete creep together with the deformation
compatibility can change the stress distribution between the con-
crete core and the steel tube of CFST columns even though a xed
load is applied. Thus the varying stress history in concrete should
be taken into account.
The total strain et of a concrete member subjected to a uniax-
ial stress rt can be calculated as follows:
Z t
et Jt; t 0 drt0 1
0
with the creep compliance function Jt; t 0 representing the total
strain at target time t due to a unit constant stress applied at age
t0 . According to the B3 model, the compliance function Jt; t 0 can
be expressed as
Jt; t 0 q1 C 0 t; t0 C d t; t 0 ; td 2
where q1 is the instantaneous elastic strain due to the unit stress
and C 0 t; t 0 the basic creep compliance. C d t; t 0 ; t d denotes the
additional compliance function for drying creep which started at
age td and has been neglected as stated above. Then the compliance
function can be further expanded as
 
  t
Jt; t 0 q1 q2 Q t; t 0 q3 ln 1 t  t 0 n q4 ln 3
t0
where q2 ; q3 , and q4 are empirical constitutive parameters; Q t; t0
can be easily obtained either from an approximate explicit formula
or by numerical integration, such as those given in [25]; n is an
empirical parameter whose value can be taken as 0.1 for all normal
concrete types.
In practice, the total integrated strain given in Eq. (1) can be
approximated by a sum of strains caused by small stress incre-
ments applied at all time intervals, that is
X
n
1
etn  Jtn ; t i Jt n ; ti1 Drt i 4
i1
2
where the subscripts n and i denote total and individual time steps,
respectively.
In Eqs. (1) and (4), the salient feature of conned concrete,
namely the status of triaxial stresses, is not reected. According
to the study of Gopalakrishnan et al. [31], creep strains of concrete
310 H. Liu et al. / Engineering Structures 100 (2015) 308325

Steel tubes
Concrete
Concrete

Reference Deformed configuration

configuration after sustained load
Fig. 1. Schematic illustration of the stress redistribution when a CFST column is subjected to long-term service loading. The amplitude of axial stresses are represented by the
length of arrows.

under triaxial stresses conform to the Boltzmann principle of with the subscript sa denoting the axial direction of the steel tube.
superposition in spite of the nonlinear nature of creep. That is to
say, the creep strain in the axial direction can be obtained by add- 2.3. Short-term elastic analysis of CFST columns
ing corresponding strain contributions caused by stress compo-
nents acting in all directions. Accordingly, the total axial and In this section, we elaborate on the algorithm used for
radial strains of conned concrete can be calculated as follows. short-term elastic analysis of CFST columns through which the
Z tn Z tn instantaneous stress status of concrete core at loading age t 0 is
eca tn Jtn ; t 0 drca t 0  2lc Jt n ; t0 drcr t 0 determined. These short-term stresses will serve as the basis for
0 0
the subsequent creep analysis. The initial guess of load distribution
X
n
1
 Jt n ; ti Jtn ; t i1 Drca ti  2lc Drcr ti  5 factor ac is taken as
i1
2
Ec Ac
ac 11
Ec Ac Es As
ecr tn lepr eca tn 6
where E is the Youngs modulus in the axial direction and A is the
with the subscripts ca and cr denoting the axial and radial direc-
cross-sectional area. The subscripts c and s are used to distinguish
tions of the concrete core, respectively. lc is the static Poissons
the two constituent materials: concrete and steel, respectively.
ratio of concrete whereas lepr stands for the effective Poissons ratio
Then the axial stresses in the concrete core and steel tube are esti-
as dened in [31]. mated as

2.2. Equilibrium and geometric compatibility conditions Nc ac N L Ns 1  ac NL

rca t0 ; rsa t0 12
Ac Ac As As
For the mechanical analysis of CFST columns, the static equilib- Given the axial stress rca t 0 in concrete, the axial strain eca t0
rium and geometric compatibility should be satised simultane- can be obtained from the short-term stressstrain relationship of
ously. In the axial and lateral directions, the static equilibrium the conned concrete, which will be detailed in Section 3.1.
conditions read Applying Eqs. (6) and (9), we then get the radial stain ecr t 0 of
NL Nc Ns 7 the concrete core and the hoop strain esh t 0 of the steel tube. In
line with the framework presented by [16], the steel tube is
2t s assumed to be in a plane stress state and moreover to behave elas-
rcr  rsh 8
tically under long-term service load. Thus its constitutive relations
D  2ts
are given by
where NL is the total axial load sustained by CSFT columns; N c and
N s are the load-carrying capacities provided by the concrete core rsa t0 rsh t0
esa t0  ls 13
and steel tube, respectively; the subscript sh represents the hoop Es Es
direction of the steel tube; ts and D are the thickness and external
dimension of the steel tube, respectively, as shown in Fig. 2. Note rsh t0 rsa t0
esh t0  ls 14
that for CFST columns with square cross-sections, the conning Es Es
stress rcr is assumed to distribute uniformly along the tube edges.
Substitution of the estimated esh t 0 and rsa t0 into Eq. (14)
Given the assumption that the concrete core and steel tube are
yields the hoop stress rsh t0 of the steel tube. Then the axial strain
perfectly bonded during the whole loading process, the following
compatibility conditions of deformation can be obtained
esa t0 of the steel tube and the radial conning stress rcr t0 can be
estimated according to Eqs. (13) and (8), respectively.
straightforwardly:
Keeping in mind the geometric compatibility condition (10) in
ecr esh 9 the axial direction, we need to assess the difference between the
estimated axial strains eca t0 and esa t0 of the concrete inll and
eca esa 10 steel tube. If the difference between them exceeds a
 H. Liu et al. / Engineering Structures 100 (2015) 308325 311

Fig. 2. Schematic representation of geometric and loading information for steel tube with various cross-sections: (a) circular section; (b) square section.

user-specied tolerance ns , we update the value of the concrete

strain eca t0 to be esa t0 , followed by the calculation of new axial
stress rca t0 and load distribution factor ac . The aforementioned
process is repeated until the tolerance limit is satised. The entire
solution procedure of the initial short-term elastic analysis is illus-
trated in Fig. 3.

2.4. Long-term creep analysis of CFST columns

Once the instantaneous stresses of the concrete core were

determined, the long-term creep analysis of CFST columns can sub-
sequently be performed. Compared with the creep deformation of
concrete, the steel creep under sustained load is assumed to be
negligible. It will be shown in Section 3 that the post-creep behav-
ior of CFST columns relies on the creep status at the target time
when ultimate failure is experienced, rather than the history of
creep deformation. Motivated by this feature, we propose a simpli-
ed model for the creep analysis of CFST columns. The proposed
method enables an efcient estimation of creep strains at any des-
ignated time without performing the time-consuming iterative
process for each loading increment of the whole duration. The for-
mat is simple and thus convenient to use. Although the nal creep
strains may be slightly overestimated, the accuracy is found to be
quite satisfactory for practical engineering design. The details of
this simplied model are elaborated below.
First, the total axial strain of the conned concrete given in Eq.
(5) can be further expanded considering that long-term service
loads are applied instantaneously at the initial time t 0

1X n
eca tn Jtn ; t0 rca t0 Jt n ; t i1 Jt n ; t i rca t i  rca t i1 
2 i1
(
1X n
 2lc  Jtn ; t 0 rcr t0 Jt n ; t i1 Jt n ; ti rcr ti
2 i1
o
rcr t i1  15

After some manipulations, the following expression can be obtained

1
eca tn fJtn ;t0 rca t0 rca t1  Jtn ; tn rca tn  rca tn1 g
2 Fig. 3. Flow chart of the initial short-term elastic analysis of CFST columns.
 lc  fJt n ; t0 rcr t0 rcr t1  Jt n ;t n rcr t n  rcr t n1 g
1Xn1 X
n1
Along with the development of concrete creep, the axial force
Jtn ; ti rca ti1  rca t i1   lc Jtn ;t i rcr ti1 sustained by the concrete core will be gradually transferred to
2 i1 i1
the external steel tube. Taking the stress and strain to be positive
 rcr t i1  16
in compression, we have
312 H. Liu et al. / Engineering Structures 100 (2015) 308325

rca ti1  rca ti1 < 0 17

The constitutive relations (13) and (14) of the steel tube are
reorganized here as
Es
rsa esa ls esh 18
1  l2s

Es
rsh esh ls esa 19
1  l2s
Substitution of Eq. (19) into the equilibrium condition (8) in the lat-
eral direction yields the conning stress as a function of the steel
strain:
Fig. 4. Schematic illustration of the simplied model for concrete creep.
2ts Es
rcr  esh ls esa 20
D  2t s 1  l2s
D  2t s
With the use of geometric compatibility conditions and Eq. (6), we rsh t  x 24
2ts
have
2t s Es Substitution of the initial guess into our simplied model (23)
rcr ti1  rcr ti1 l  ls eca ti1  eca ti1  yields the axial strain eca t of concrete as a function of x
D  2t s 1  l2s epr
21 eca t Ax B 25
Due to the presence of lateral connement (i.e. lepr > ls ) and where A and B are given by
concrete creep (i.e. eca t i1 > eca t i1 ), the positiveness of the A lc Jt; t 0 Jt; t 26
above Eq. (21) can be ensured. That is to say, the conning pressure
keeps increasing during the process of long-term loading. and
Combining Eqs. (16), (17), and (21), one can easily get the upper r ca t rca t0 r ca t  rca t0
bound of the axial strain eca t n B Jt; t 0  Jt; t 
2 2
1  lc rcr t0 Jt; t0  Jt; t 27
eca tn < fJtn ; t0 rca t0 rca t1  Jtn ; tn rca tn  rca tn1 g
2
From the geometric compatibility conditions (9), (10) and the for-
 lc  fJt n ;t 0 rcr t 0 rcr t1  Jt n ; tn rcr tn  rcr t n1 g 22
mula for ecr , we know that
In the present model, the axial strain eca at target time t is
esa t Ax B; esh t lepr Ax B 28
approximated by this upper bound by taking n 1 and t t n
Substituting the estimated rsh t; esa t, and esh t into the constitu-
rca t rca t0 rca t  rca t0
eca t Jt; t0  Jt; t  tive relation (19) of the steel tube, we obtain a linear equation from
2 2
which x (i.e. rcr t) can be easily solved:
rcr t rcr t0
 2lc Jt; t0   2lc Jt; t
2 1  l2s D  2t s
x Es lepr  ls Ax B 29
rcr t  rcr t0 2t s
 23
2 Subsequently, esa t and esh t can also be obtained. Given their
Essentially, the convex curve of the actual stress history [9] is value, we can obtain the axial stress rsa t of the steel tube from
replaced by a linearly varied one in the simplied model. As can Eq. (18). According to the equilibrium condition in the axial direc-
be seen in Fig. 4, the simplied method will give a relatively higher tion (7), a new value of rca t can be calculated as
prediction of the creep strain compared with the real value.
NL  As rsa t
However, the proposed method will considerably reduce the com- rca t 30
Ac
putational cost.
In Eq. (23), the instantaneous stress status at the initial loading The difference between r  ca t and rca t should be less than a preset
age t0 has been determined by the framework given in Section 2.3. tolerance nl . Otherwise, one needs to adjust the value of r  ca t to be
To evaluate the creep strain eca at target time t, one needs to know rca t and repeat the aforementioned process until convergence is
the values of the axial and radial stresses of the conned concrete achieved. The entire solution procedure of the long-term creep
at t. For this purpose, an initial guess should be chosen for rca t analysis is illustrated in Fig. 5.
and rcr t, which is followed by an iterative procedure aiming to It bears emphasis that the proposed creep analysis method is
ensure the static equilibrium and geometric compatibility condi- applicable to cases in which the bond between the steel tube and
tions. In the work of Naguib and Mirmiran [16], the authors the concrete inll is assumed to be perfect. In these situations
employed a staggered iterative scheme where rca t and rcr t the concrete inll undergoes triaxial stress states. If the level of
are solved for separately. In contrast, we propose a monolithic sustained loading is low, the unstable mortar crack propagation,
solution technique by taking advantage of the correlation between which is the cause of concrete dilation, does not develop in the
the axial and radial stresses in the concrete component. In this concrete inll and thus the Poisson ratio of the steel tube, ls , will
framework, both the rca t and rcr t can be determined simulta- be larger than that of the concrete inll, lc . That is to say, the lat-
neously within the same iteration loop, leading to better conver- eral expansion of the steel tube is larger than that of the concrete
gence performance and less computational cost. The monolithic core. A small gap, namely, interface debonding, may occur if the
iteration scheme is detailed as follows. bond strength of the steelconcrete interface is insufcient. In such
At the beginning of iteration, the initial guess of rca t is taken cases both the steel tube and the concrete inll are uniaxially
to be r ca t rca t 0 whereas the value of rcr t is represented by x. loaded and the elastic and creep analyses can be remarkably sim-
Applying Eq. (8), we get the hoop stress rsh t of the steel tube as plied. The corresponding analysis method is beyond the scope of
 H. Liu et al. / Engineering Structures 100 (2015) 308325 313

the paper. The interested reader is referred to Wang et al. [22] for 3. Ultimate response of CFST columns after long-term service
more details. load

In the literature, various closed-form stressstrain models have

2.5. Case study of accuracy and convergence
been proposed for steel-conned concrete without considering the
effects of long-term service loading. These models are based
The accuracy and convergence performance of the proposed
directly on the interpretation and regression analysis of axial com-
creep analysis method is validated in this section. We choose the
pression test results. Since direct failure tests have been conducted
creep test data of Terrey et al. [19] for a circular CFST specimen
on CFST columns by many researchers for several decades, the
as the benchmark, with the corresponding geometric, material,
available test database used in the model development is large
and loading parameters listed in Table 1. The total axial strain eca
and reliable, which ensures the successful application of these
of the concrete core is evaluated by the simplied method for dif-
models in predicting the short-term inelastic response of CFST col-
ferent time instants t 18, 28, 50, 118, and 145 d. For comparison,
umns. However, it is not the case for the failure tests after sus-
the method proposed by Naguib and Mirmiran [16] is also adopted.
tained loading. The lack of the corresponding experiments can be
The predicted axial strains are plotted against the experimental
partially attributed to their expensive and time-consuming nature.
observation in Fig. 6a. It appears that the total axial strain esti-
Without enough experimental samples, it becomes a challenging
mated by the proposed method is slightly higher than that of
task to develop new constitutive models accounting for the effects
Naguibs method, as we expected in Section 2.4. Even so, the sim-
of sustained loading.
plied method is still accurate enough for most engineering appli-
It is well recognized that the behavior of CFST columns is inu-
cations. Fig. 6b compares the numbers of creep iterations for the
enced by many geometrical and material parameters, particularly
simplied method and Naguibs method. Apparently, the simpli-
the diameter-to-thickness ratio D=t s , concrete compressive
ed method equipped with the monolithic iterative scheme
strength f c0 , and steel yield strength f y . These parameters deter-
improves the convergence performance considerably.
mine the conning effect provided by the steel tube, which leads
to the strength and ductility enhancement of the concrete core.
As the departure point for our study, we carefully analyze the indi-
vidual inuence of sustained loading on these critical parameters.
Finite element based numerical study as well as published experi-
mental results are resorted to validate our analysis. By adopting
these new parameters with the effects of sustained load consid-
ered, the short-term constitutive models, which are
well-established from the short-term test database covering a wide
parametric range, can be fully utilized to predict the ultimate
response of CFST columns with sustained loading previously
applied.
In this section, the mechanism underlying the experimentally
observed strength change of CFST columns after long-term service
load is explored. It is found that the post-creep response of CFST
columns is controlled by the competition between the residual
creep strain, which reduces the conning level provided by the
steel tube and hence the compressive strength and ductility, and
the strength enhancement of plain concrete, which is caused by
the sustained load. Based on the short-term constitutive model
previously proposed by the authors [32] for steel-conned con-
crete, we propose a modied uniaxial stressstrain relationship
for conned concrete, which incorporates the aforementioned
inuence of sustained load. This model is used later in ber ele-
ment analysis to predict the ultimate performance of CFST columns
after sustained load.

Table 1
Material, geometric, and loading parameters for the creep test conducted by Terrey
et al. [19].

Property name Symbol Value Unit

Length of specimen L 600 mm
Outer diameter of specimen D 199 mm
Thickness of steel tube ts 1.5 mm
Compressive strength of concrete f c0 40 MPa
Yield stress of steel fy 345 MPa
Youngs modulus of concrete Ec 30 GPa
Youngs modulus of steel Es 200 GPa
Poisson ratio of concrete lc 0.2
Poisson ratio of steel ls 0.3
Long-term service load NL 350 kN
Initial loading time t0 18 d
Target time t 145 d
Fig. 5. Flow chart of the long-term creep analysis of CFST columns.
314 H. Liu et al. / Engineering Structures 100 (2015) 308325

600 80
Naguib's method
70 Simplified method
500

Total axial strain (x10 )

-6
60
400
50

Iteration #
300 40

30
200
20
Experiment
100
Naguib's method 10
Simplifid method
0 0
0 20 40 60 80 100 120 140 160 28 2
50 3
118 4
145
Time (d) Time (d)
(a) (b)

Fig. 6. Accuracy and convergence performance for different creep analysis methods: (a) time history of total axial strain; (b) iteration number.

3.1. Short-term constitutive model of steel-conned concrete f cc f c0 k1 f L 32

 
The exponential constitutive model proposed in [32] is adopted fL
ecc ec0 1 k2 33
in this study to determine the short-term deformation behavior of f c0
steel-conned concrete. Its full range stressstrain curve is
schematically depicted in Fig. 7. As can be seen from Fig. 7, this where f c0 and ec0 denote the compressive strength and the corre-
constitutive relation with a concise format can provide a smooth sponding strain of unconned concrete, respectively. In this study,
softening branch. Its envelope curve can be expressed as ec0 is taken as 0.003 as suggested by the ACI Specication [34]. k1
( and k2 are empirical coefcients which can be obtained from exper-
f cc n2  n if n eeccc 2 0; 1 imental data, and are set as 4.1 and 20.5. f L is the lateral conning
rc
f cc  f Res expKec  ecc f Res if n eeccc 2 1; 1 strength provided by steel tube, which is expected to be dependent
on the material properties and geometric characteristic of CFST col-
31
umns. Thus different formulas are needed to determine the conn-
where rc and ec are the axial stress and strain of conned concrete, ing strength in circular and square CFST columns. For circular
respectively; f cc and ecc are the axial peak stress and the corre- members, the lateral conning strength can be calculated as [35]:
sponding strain of conned concrete, respectively; K is the softening
2t s
parameter used to control the slope of the descending phase of con- fL b f 34
D  2ts y
stitutive curve. f Res is the residual strength of conned concrete,
whose value is taken to be a percentage of the peak stress f cc . !2
Here the ratio k3 of f Res over f cc is chosen to be 0.2. f f f
b 0:3582me  0:1524 c0 4:843me c0  9:169 c0  0:2688
It is well known that the compressive strength and the corre- fy fy fy
sponding strain can be considerably enhanced in conned concrete 35
due to the lateral restraining action when compared with those of
unconned concrete. Such an enhancement can be characterized
me 0:881  106 D=ts 3  2:58  104 D=ts 2 1:953
by the linear equation proposed by Richart et al. [33].
 102 D=t s 0:4011 36

with f y the yield stress of steel tube. In the case of square CFST col-
umns, the conning strength is given by the following equations
[36]:
1:46
f c0 1:03
f L 6:5R 0:12f c0 37
fy

with
s
D 3f y 1  l2s
R 38
ts Es p2

It is clear that the aforementioned formulas (34)(38) characterize

the dependence of the lateral conning strength f L on the
diameter-to-thickness ratio D=t s and the yield strength f y of the
steel tube. In addition, the conning strength f L is also a function
of the compressive strength f c0 of the concrete core. This is a rea-
sonable assumption considering that the f c0 =f y ratio plays a vital
role in the dilation of the concrete core. When the f c0 =f y ratio is
small, the micro cracks in the concrete core will develop at the early
Fig. 7. Comparison of short-term and post-creep uniaxial stressstrain relations for loading stage and will lead to dilation of the concrete, resulting in a
steel-conned concrete. high conning strength.
 H. Liu et al. / Engineering Structures 100 (2015) 308325
In order to determine the softening slope K, one can introduce a
neutral point ec50 (see Fig. 7) located at the falling branch with the
corresponding stress equal to 0:5f cc f Res . Applying the constitu-
tive relation Eq. (31) yields
0:5f cc f Res f cc  f Res expKec50  ecc f Res 39
Then the softening parameter K can be derived from Eq. (31)
ln 0:5
K 40
ec50  ecc
The softening parameter K is an important index indicating the level
of lateral restraint. In other words, a higher value of K represents a
gentler degradation slope and thus a higher ductility of conned
concrete, and vice versa.
The neutral point of the conned concrete can be calculated
according to the regression of axial loading experiment data [37],
which reads
ec50 ec50 0:21f L =f c0 1:16 41
with ec50 the strain of the neutral point of plain concrete. The exist-
ing empirical study suggested a value of ec50 =0.04 [38].
3.2. Effects of sustained loading on CFST columns
The short-term constitutive relation discussed above should be
adapted to characterize the ultimate behavior of conned concrete
after long-term service loading. The effects of sustained load are
mainly related to residual creep strain and strength increase of
plain concrete, which are investigated in detail as below.
1. The creep recovery of concrete is assumed to be negligible in
this study. This assumption is consistent with the common sce-
narios encountered in practice: failure loading usually takes
place right after the duration of the service load. Thus the time
interval between them is short enough to neglect creep
recovery. As discussed in Section 2, creep deformation of the
concrete core can develop during the service life of CFST col-
umns. Without the creep recovery, a residual hoop strain ^esh
in the steel tube should exist due to the compatibility of defor-
mation. The residual hoop strain ^esh t at target time t can be
expressed as
^esh t ^ecr t lepr ^eca t
lepr eca t  eca rca t 42
where ^eca and ^ecr denote the residual creep strains of the con-
crete core in the axial and radial directions, respectively. The
axial stress rca t and strain eca t of the concrete core at
target time t are calculated by the simplied creep analysis
method proposed in Section 2. eca rca t denotes the instanta-
neous elastic strain under the stress level of rca t, which
can be obtained from the short-term constitutive model
discussed above.
The residual hoop strain developed in the steel tube will reduce
its elastic deformation capacity in the radial direction. Its effect
can be represented by a weakened yielding strength of the steel
tube, which is referred to as the effective yielding strength ~f y
hereinafter. Due to the positive correlation between the conn-
ing strength and the steel yielding strength shown in Eqs. (34)
(38), the residual hoop strain caused by the concrete creep will
exert a negative effect on the conning action between the con-
crete inll and steel. In accordance with Eq. (32), the reduced
conning action will further decrease the compressive strength
of the concrete core. We propose a linear formula to characterize
the effective yielding strength of the steel tube after sustained
load, given by
315
~f y ey  j^esh j f 43
y
ey
with ey the yielding strain of the steel tube.
2. Aside from the negative effect of residual creep strain on the
compressive strength of the concrete core, conned concrete
can benet from the strength enhancement brought about by
long-term service load. Such an enhancement of concrete
strength has been observed in numerous plain concrete exper-
iments. Hughes and Ash [39] found that the strength enhance-
ment of concrete is positively correlated to cement content,
aggregate size, and stress level, and negatively correlated to
concrete age at loading. These correlation relations also exist
in creep deformation. Dhir and Sangha [40] presented a com-
prehensive review on this problem where the strength
enhancement was suggested to be around 11% for 28 days tests.
This phenomenon could be explained by increased consolida-
tion of concrete [41] and improved healing of cracks perpendic-
ular to the load [42]. A similar statement is also made in the
study of Wang et al. [22] on the time-dependent behavior of
CFST columns. Based on these experimental observations, we
introduce an enhancement coefcient kc0 , not less than 1, to
characterize the compressive strength ~f c0 of the plain concrete
after long-term service loading
~f c0 kc0 f with kc0 P 1 44
c0
In the axial compression tests of CFST columns, it is difcult to
directly measure the conning pressure provided by the steel tube.
Furthermore, as we stated before, the conning strength depends
both on the yielding strength of the steel tube and on the compres-
sive strength of the concrete core. Therefore, the inuence of the
residual hoop strain on the conning strength is combined with
that of the enhanced concrete strength and is therefore not easily
extracted even if the experimental results on conning pressure
are available. For these two reasons, the nite element based
numerical study is resorted to justify our rst argument, which will
be detailed in Section 4. The main benets of the FE validation
include the following: (1) substantial savings in the cost and time
compared with the long-term creep and ultimate failure tests of
CFST columns; (2) allows to see the conning pressure at any loca-
tion and loading level; (3) allows to vary any parameter of interest
to evaluate its inuence on CFST columns. In this context, the effect
of the residual creep strain on the conning strength can be quan-
titatively investigated by removing the strength change of the con-
crete core caused by sustained loading. Regarding the second
argument we made, it has been supported by numerous experi-
mental observations [3942] and its effectiveness can therefore
be ensured.
3.3. Post-creep constitutive model for steel-conned concrete
According to the statement given in Section 3.2, f y and f c0
should be updated to the effective yielding strength ~f y of the steel
tube and the post-creep compressive strength ~f c0 of the plain con-
crete, respectively, if sustained loading is applied. Following the
formulas (34)(38) for composite columns under short-term load-
ing, the post-creep lateral conning strength ~f L for circular and
square CFST columns can be obtained by adopting the updated
parameters ~f y and ~f c0 . Note that the superscript ~ is employed
to distinguish quantities after sustained loading with their
short-term counterparts. By checking the positiveness of the
derivatives of f L with respect to f c0 and f y , the positive correlation
between f L and f y , as well as the negative correlation between f L
316 H. Liu et al. / Engineering Structures 100 (2015) 308325
and f c0 , can straightforwardly be derived for both circular and
square CFST columns. Considering that ~f y 6 f y and ~f c0 P f c0 , one
can get
 
~f L ~f L ~f c0 ; ~f y ~f L kc0 f ; ey  ^esh f 6 fL
c0
ey y The post-creep constitutive model for conned concrete is com-
Namely, the post-creep conning strength ~f L is predicted to be
smaller than the short-term one. Accordingly, the post-creep peak
stress ~f cc and the corresponding effective strain ~ecc can be calculated
as follows
~f cc ~f c0 k1 ~f L
! strain on the conning action. In the simulation, both the viscous
~f L
~ecc ec0 1 k2
~f c0
with
~ecc etotal  ^eca
cc
In Eq. (48), the total axial strain etotal
cc of concrete corresponding to
the post-creep peek stress ~f cc is additively decomposed into the
residual creep strain ^eca and the effective peak strain ~ecc . This
decomposition aims to take into account the inuence of irre-
versible creep strain developed in the concrete core.
According to Eq. (46), the post-creep compressive strength of
conned concrete may be higher than its short-term peak strength,
as observed in Han et al. [9] and Wang et al. [22]. On the other
hand, the competing effects of increase in plain concrete strength
and decrease in conning strength may also give rise to a reduction
in the compressive strength of conned concrete. Thus, our model
is capable of explaining and characterizing the seemingly divergent
observations in the literature on the strength change of the
steel-conned concrete. The specic value of strength change
depends on the material and geometric properties of CFST columns
as well as the level and duration of the sustained load.
In contrast to the uncertain inuence of sustained load on the
ultimate strength, the ductility performance of steel-conned con-
crete, namely, the capability to experience large plastic deforma-
tion without signicant strength reduction, will be lowered by
long-term service loading. This trend can be clearly seen from
Eqs. (40) and (41), where both the reduced conning strength ~f L
and the enhanced plain concrete strength ~f c0 increase the degrad-
ing rate of the post-yield softening branch. This prediction is rea-
sonable considering the following experimental evidence: (1)
concrete with a higher compressive strength behaves in a more
brittle manner; (2) weakened conning action between the steel
tube and concrete core indicates a more brittle post-peak descend-
ing branch. Therefore, one should be cautious of the negative effect
on ductility brought about by sustained loading in designing CFST
columns for permanent facilities.
Herein we adapt the short-term stressstrain relation (31) by
replacing the short-term peak stress f cc and strain ecc with the
newly estimated post-creep peak stress ~f cc and strain ~ecc ^eca ,
through which the aforementioned effects of long-term service
load are accounted for. In addition, the post-creep residual strength
should also vary with the change of the compressive strength,
namely ~f Res k3 ~f cc . The proposed post-creep constitutive model
is expressed by
8 circular and D=20 for square CFST columns, respectively. In the
>
> 0 if n ec~ecc^eca 2  ^e~ecacc ; 0
< axial direction, the element size is selected as twice the mesh size
rc ~f cc n2  n if n ec~ecc^eca 2 0; 1
>
>
: ~
f cc  ~f Res expK
~ ec  ^eca  ~ecc ~f Res if n ec ^eca 2 1; 1
~ecc
with
ln 0:5
~
K 50
1:16
e 
c50 0:21~f L =~f c0  ~ecc
45
pared with the short-term uniaxial stressstrain relation in Fig. 7,
from which the effects of sustained load can be clearly seen.
4. Finite element modeling and validation
The nite element modeling presented in this section aims to
46 validate the aforementioned analysis of the effect of residual creep
and plastic deformations of CFST columns are taken into account.
47
Initially, the structural members are subjected to a rapidly applied
axial service loading, held constant for a long period of time. At this
stage, creep deformation plays a dominant role. At the end of the
48 long-term simulation, the CFST specimens are rapidly loaded to
failure with a plastic process dominating the structural deforma-
tion. To the best knowledge of the authors, the present contribu-
tion is the rst numerical study on the long-term failure
behavior of CFST columns. Besides the coupled long-term simula-
tion, a short-term failure analysis without creep is also performed
and compared with the experimental results in order to verify the
accuracy of the adopted plastic constitutive models for steel and
concrete.
4.1. Software
The general-purpose commercial FE software ABAQUS is used in
the present study to simulate the long-term creep response as well
as the failure of CFST columns under axial compression. In the sim-
ulation, three major components should be properly modeled to
reproduce the realistic behavior of CFST columns. Those are the
concrete inll, the surrounding steel tube, and the interaction
between the concrete inll and steel tube. The initial application
of the axial service loading is assumed to occur quickly. It involves
purely elastic response, which is obtained by using the static anal-
ysis procedure. Creep response and plastic failure are then devel-
oped in the second and third steps, in which the quasi-static
analysis procedure provided by ABAQUS is done. Due to the nonlin-
ear nature of the studied problem, an incremental-iterative solu-
tion scheme is adopted and the NewtonRaphson method is
invoked for the solution of each incremental step.
4.2. Model geometry and boundary conditions
Three-dimensional 8-node solid elements (C3D8R) with
reduced integration are used to discretize the steel tube and con-
crete inll. Since the reduced integration scheme may lead to a
zero-energy mode (i.e. element distortion does not generate any
strain energy), hourglass control with an articial stiffness intro-
duced in [43] is used in this work to alleviate the zero-energy
deformation mode. In order to determine an optimal mesh size
providing relatively accurate prediction at the expense of reason-
able computational time, a mesh convergence study is conducted.
The mesh is rened until the variation of the axial loadstrain
curve tends to be minor. The results of the mesh sensitivity analy-
sis suggest the element size along the cross section to be D=18 for
along the radial direction.
The interaction action between the steel tube and concrete inll
is modeled using surface-to-surface contact where the inner sur-
49
face of the steel tube and the outer surface of the concrete inll
 H. Liu et al. / Engineering Structures 100 (2015) 308325
are dened as the master and slave surfaces, respectively. Hard
contact is specied in the normal direction of the interface to pre-
vent penetration whereas the Rough option in ABAQUS is used
in the tangential direction to reproduce the experimental observa-
tion that there is little or no slip between the steel tube and con-
crete inll. In order to obtain a better convergence rate in
contact analysis, the meshes on the steelconcrete contact surface
are matched.
Fig. 8 shows an example of the mesh geometry of a typical cir-
cular specimen with its boundary conditions. The column is axially
loaded at the top with the stiff plate of the test machine. For the
failure analysis of CFST columns, displacement control is recom-
mended to capture the potential softening branch of the axial load
(N)- axial strain (e) curves. Apart from the failure analysis, the
time-dependent creep behavior of CFST columns before ultimate
failure needs to be investigated. In this case, the axial sustained
loading should be kept constant. To this end, a reference point at
the center of the top end is dened and all nodes located at the
top end are kinematically coupled with the reference point by
using multi-point constrains provided by ABAQUS. In the analysis
step of creep, a concentrated force in the axial direction is applied
at the reference point whose translational degrees of freedom at
the other two directions are restrained. In the subsequent failure
analysis, the axial loading condition is replaced by a displacement
boundary condition specied along the column length. For the col-
umn base, all three translational degrees of freedom are con-
strained in all analysis steps.
4.3. Material models
4.3.1. Steel tube
The Youngs modulus Es and Poissons ratio ls of steel are taken
to be 200 GPa and 0.3, respectively, if their values were not
reported in the test. Various constitutive models have been used
in the literature to model the steel tube, for example, the
elastic-perfectly plastic model used in [44] and the elasticplastic
model with multi-linear hardening used in [45]. As stated in Tao
et al. [46], the steel tube exhibits insignicant strain hardening at
the strain levels of general structural interest (normally less than
5%), especially in cases of square CFST columns. Therefore, the
elastic-perfectly plastic model is adopted in this study for the mod-
eling of steel tube. The stressstrain relation of the steel material is
controlled by the Von Mises yield criterion with associated plastic
ow rule using the PLASTIC option in ABAQUS. A schematic plot
of the uniaxial stressstrain curve of steel is depicted in Fig. 9 for
clarity. Note that steel creep is not considered here since it is
assumed to be negligible compared to concrete creep.
4.3.2. Conned concrete
The Poissons ratio lc of concrete normally ranges from 0.15 to
0.22 [47]. When experimental values are not available, lc is
Constrained in X and Y directions
Fig. 8. Finite element discretization and boundary conditions of a typical circular CFST columns. The steel tube is colored in red whereas the concrete core is represented in
blue. (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)
317
assumed to be 0.2. Considering the strong correlation between
the Youngs modulus Ec of concrete and its compressive strength,
the empirical formula recommended by ACI Committee 318 [34]
is employed to calculate the Youngs modulus Ec
q
Ec 4700 f c0 MPa 51
where the compressive strength f c0 is in MPa.
A linear DruckerPrager yield criterion with a non-associated
ow rule model available in ABAQUS is employed to model the
concrete failure [48] in this study. The adopted plastic model
accounts for the pressure-sensitive yielding exhibited by
quasi-brittle and frictional materials [49] such as concrete and
rock. Several parameters are required for using the linear
DruckerPrager model, these being, the friction angle (b), the ratio
of ow stress in triaxial tension to that in compression (K), the dila-
tion angle (w), and the ow potential eccentricity (). They are cal-
ibrated to be 50 , 1.0, 30 , and 0.1 [50], respectively. In addition,
the strain hardening/softening curve, dened as a function of the
equivalent plastic strain, can be determined from the short-term
uniaxial stressstrain relation (31) of concrete, either with or with-
out connement. For a better comparison between the models, we
use the short-term constitutive relation, and so the potential inu-
ence of the concrete strength enhancement caused by the sus-
tained load can be eliminated from all numerical tests. Hence the
conning strength change due to sustained loading is only attribu-
ted to the existence of residual creep deformation and the correla-
tion between them can be more easily investigated in the
numerical study.
In order to model the time-dependent creep deformation of
concrete, a creep constitutive law should be dened. Considering
the variable stress history during the analysis, we employ the
strain hardening power law available in ABAQUS, which is dened
in the following rate form
  1
 cr n  m 1  ecr m m1
e_ cr A  r 52
where r  cr and ecr denote the uniaxial equivalent creep stress and
strain, e_ cr the uniaxial creep strain rate, A; n, and m the
user-specied creep material parameters. The creep law can be
integrated with respect to time either analytically or numerically.
The integrated equation is then compared to the results of creep
or relaxation tests to calibrate the creep material parameters A; n,
and m. In order to represent the creep test data [19], A; n, and m
0:8
are chosen to be 7  1012 MPa5 d , 5, and 0.2, respectively.
4.4. Model verication
In order to examine and verify the accuracy of the nite element
model, the short-term failure analysis is carried out rst on a total
of 24 CFST columns. The short-term analysis can be treated as the
Y
Z
X
Constrained in all directions
318 H. Liu et al. / Engineering Structures 100 (2015) 308325
Stress
0 dicted change of conning strength using the analytic formula
Strain
Fig. 9. Uniaxial elastic-perfectly plastic stressstrain curve for steel. ES denotes the
Youngs Modulus of steel in the linear elastic region and ry is the onset of yielding.
limited case of long-term failure analysis with creep deformation
approaching to zero. The ultimate axial strengths as well as the full
range axial load (N)-strain (e) curves predicted by the FE model are
compared with the corresponding experimental results collected
from the literature [5153]. The selected test data cover a wide
range of parameters including the diameter-to-thickness ratio,
the concrete compressive strength, and the steel yield strength.
The details of the test data are summarized in Tables 2 and 3 for cir-
cular and square CFST columns, respectively. Note that the speci-
men names in Tables 2 and 3 are consistent with the names given
in the literature [5153].
Fig. 10a depicts comparisons between predicted and measured
axial loadstrain curves for three representative circular specimens
CU-022, CC4-A-4-1, and CU-100. The diameter-to-thickness ratio
D=t s of these representative specimens ranges broadly from 22 to
100. Similarly, the predicted axial loadstrain curves are plotted
against experimental results in Fig. 10b for three square represen-
tative specimens (SU-022, CR4-C-4-1, and SU-070) with D=ts rang-
ing from 22 to 70. As can be seen from Fig. 10, the nite element
simulation gives excellent prediction of stiffness, ultimate strength,
and post-peak behavior of CFST columns under axial loading. In
addition, both the test data and numerical results show that the
axial loadstrain curves of specimens with signicant conning
effects (i.e. circular specimens with low diameter-to-thickness ratio
D=t s ) exhibit no descending branch, whilst other specimens show
apparent strain softening, especially in square-type CFST columns.
Thus it is necessary to dene a consistent ultimate strength for
comparison purpose. Following Tao et al. [54], the ultimate
strength is taken as the rst peak load in the full range N-e curve
if the corresponding strain is less than %1. If there is no softening
branch or the strain at the rst peak load is greater than %1, the
ultimate strength is selected to be the load at %1 strain limit. In
accordance with the denition, the ultimate strengths obtained
from the FE modeling N uP and experimental results N uE are summa-
rized in Tables 2 and 3 for all specimens. For circular CFST columns,
the mean value and standard deviation (STD) of N uP =N uE is 0.991
and 0.057, whereas a mean ratio N uP =N uE of 1.023 with a standard
deviation of 0.049 is obtained for square CFST columns.
4.5. Long-term failure analysis of CFST columns
In this section, we analyze the long-term failure behavior of
four representative specimens: CU-022, CC4-A-4-1, SU-022, and
SU-040. As stated before, the numerical tests serve as the
benchmark solution, to validate the analytical model presented
in Sections 3.2 and 3.3 with regards to the effect of residual creep
strain on conning strength.
In consistence with the denition of ultimate strength, the con-
ning strength is taken as the average pressure at the steelcon-
crete contact surface when the ultimate strength is reached. Both
the outputs of resultant norm force (CNORMF) at the interface
and total area in contact (CAREA) are supported by ABAQUS.
Then, the average pressure is equal to the resultant contact force
per unit contact area. In the long-term failure analysis, the loading
period before the ultimate test is varied to obtain different levels of
axial creep strain ^eca . The variations of conning strength ~f L ,
obtained from the numerical tests and analytic formula (45), are
plotted as a function of axial creep strain ^eca in Figs. 11a and 12a
for circular and square specimens, respectively. Overall, the pre-
shows a reasonable match to the numerical FE results. A gradually
decreasing conning strength is clearly observed with the increase
of axial creep strain.
Apart from the values of conning strength, the relative reduc-
tion in conning strength obtained by the numerical test and ana-
lytic approach are also compared in Figs. 11b and 12b for these
specimens. As can be seen, the relative reduction in conning
strength caused by residual creep is minor at a low creep strain
level (less than 500 le). However, sometimes, the relative reduc-
tion may become manifest in extreme loading cases. For example,
creep strain can be very high when such composite structures are
exposed to re. In such situations, the creep effect should be fully
considered.
5. Performance assessment
In this section, the performance of the proposed constitutive
model for steel-conned concrete after long-term service load is
assessed against limited data available from two sets of experi-
ments: square CFST columns presented by Han et al. [9] and circu-
lar CFST columns reported by Wang et al. [22]. Considering that the
major objective of the axial stressstrain relation is its application
in ber analysis, the ber element method implemented in
OpenSees [55] is adopted to investigate the nonlinear behavior of
CFST columns under compression. As a simple yet very effective
numerical technique, the ber element method achieves a reason-
able balance between the efciency of a 1D beam-column element
and the accuracy of a 3D solid element. The following assumptions
are made in the ber element formulation:
Original plane cross-sections remain plane and perpendicular to
the centroidal axis during the whole analysis process.
Relative slip does not occur between the materials in
cross-sections.
Material components within the cross-sections are subjected to
uniaxial stress states.
In the ber element method, a circular or square column section
can be discretized into a cluster of ber elements as depicted in
Fig. 13 and such a discretization can be easily adapted to other dif-
ferent applications. Each ber element represents a ber of material
deforming longitudinally along the member and can be assigned
specic material properties. Then ber stresses are calculated from
ber strains using the assigned uniaxial stressstrain relations
incorporating the conning effects. In this study, the exponential
strainstress relationship proposed in Section 3 and the
elasto-plastic uniaxial model Steel01 provided by the material
library of OpenSees are adopted for the concrete inll and steel
tube, respectively. Following the experimental results reported by
 H. Liu et al. / Engineering Structures 100 (2015) 308325 319

Table 2
Short-term test data and simulation results for circular CFST columns.

No. Specimen D  ts (mm) D f y (Mpa) f c0 (Mpa) N uE (KN) N uP (KN) N uP Source

ts N uE

1 CC4-A-4-1 149  2:96 50.3 308 40.5 1060 1025 0.97 [51]
2 CC4-C-4-1 300  2:96 101.4 279 41.1 3277 3506 1.07
3 CC4-D-4-1 450  2:96 152 279 41.1 6870 6389 0.93
4 CC8-A-8 108  6:47 16.7 853 77 2713 2957 1.09
5 CC8-C-8 222  6:47 34.4 843 77 7108 7321 1.03
6 CC8-D-8 337  6:47 52 823 85.1 13776 13500 0.98
7 CU-022 140  6:5 21.5 313 23.8 1340 1410 1.05 [52]
8 CU-047 140  3:0 46.7 285 28.18 832 782 0.94
9 CU-100 300  3:0 100 232 27.23 2810 2670 0.95
10 CU-040 200  5:0 40 265.8 27.15 1989 2009 1.01 [53]
11 CU-070 280  4:0 70 272.6 31.15 3025 2874 0.95
12 CU-150 300  2:0 150 341.7 27.23 2608 2425 0.93

Mean 0.991
STD 0.057

Table 3
Short-term test data and simulation results for square CFST columns.

No. Specimen D  ts (mm) D f y (Mpa) f c0 (Mpa) N uE (KN) N uP (KN) N uP Source

ts N uE

1 CR4-A-4-1 148  4:38 33.8 262 40.5 1414 1442 1.02 [51]
2 CR4-C-4-1 215  4:38 49.1 262 41.1 2424 2327 0.96
3 CR4-D-4-1 323  4:38 73.7 262 41.1 4950 5297 1.07
4 CR8-A-8 119  6:47 18.4 835 77 3318 3451 1.04
5 CR8-C-8 175  6:47 27 835 77 5366 5259 0.98
6 CR8-D-8 265  6:47 40.9 835 80.3 8990 8451 0.94
7 SU-017 127  7:47 17 347 23.8 1892 1949 1.03 [52]
8 SU-022 127  5:67 22.4 312 23.8 1242 1317 1.06
9 SU-029 127  4:34 29.3 357 26 1106 1161 1.05
10 SU-040 200  5:0 40 265.8 27.15 2312 2243 0.97 [53]
11 SU-070 280  4:0 70 272.6 31.15 3401 3605 1.06
12 SU-150 300  2:0 150 341.7 27.27 3062 3338 1.09

Mean 1.023
STD 0.049

3500 4000
CC4-A-4-1 FE simulation CR4-C-4-1 FE simulation
CC4-A-4-1 Experiment 3500 CR4-C-4-1 Experiment
3000
CU-022 FE simulation SU-022 FE simulation
CU-022 Experiment SU-022 Experiment
3000
2500 CU-100 FE simulation SU-070 FE simulation
CU-100 Experiment SU-070 Experiment
2500
Axial load N (KN)

Axial load N (KN)

2000
2000
1500
1500

1000
1000

500 500

0 0
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.00 0.01 0.02 0.03 0.04
Axial strain Axial strain
(a) (b)
Fig. 10. Comparison between predicted and measured axial loadstrain curves for (a) circular CFST specimens; (b) square CFST specimens.

Cook and Chindaprasirt [56] on plain concrete strength after sus- usually characterized by n N L =N u , namely, the ratio of the sus-
tained load, the value of the strength enhancement coefcient kc0 tained loading N L over the cross-sectional ultimate capacity N u .
is selected to be 1.2 for the long-term failure analysis of CFST According to Han [57], the cross-sectional ultimate capacity N u
columns. can be estimated as follows
Among the test results for circular [22] and square [9] CFST col-
umns, only selected cases, in which both the long-term creep and Nu Asc 1:212 Bn Cn2 f ck 53
ultimate failure of specimens were tested, are utilized for the per-
formance assessment. The level of the axial sustained loading N L is B 0:138f y =235 0:7646 54
320 H. Liu et al. / Engineering Structures 100 (2015) 308325

8 0

Relative reduction in confining strength (%)

7
-2

6
Confining strength (MPa)

-4

5
-6
4
-8
3

-10
2
CC4-A-4-1 FE test CC4-A-4-1 FE test
CC4-A-4-1 Analytical prediction CC4-A-4-1 Analytical prediction
1 -12
CU-022 FE test CU-022 FE test
CU-022 Analytical prediction CU-022 Analytical prediction
0 -14
0 300 600 900 1200 1500 0 300 600 900 1200 1500
Axial creep strain () Axial creep strain ()

(a) (b)
Fig. 11. Comparison of conning strength obtained from analytic formula and numerical test for circular specimens: (a) conning strength versus axial creep strain; (b)
relative reduction in conning strength versus axial creep strain.

3 0
Relative reduction in confining strength (%)

-4
Confining strength (MPa)

2
-8

-12
1

SU-040 FE test CC4-A-4-1 FE test

-16
SU-040 Analytical prediction CC4-A-4-1 Analytical prediction
SU-022 FE test SU-022 FE test
SU-022 Analytical prediction SU-022 Analytical prediction
0 -20
0 200 400 600 800 1000 0 200 400 600 800 1000
Axial creep strain () Axial creep strain ()
(a) (b)
Fig. 12. Comparison of conning strength obtained from analytic formula and numerical test for square specimens: (a) conning strength versus axial creep strain; (b)
relative reduction in conning strength versus axial creep strain.

Steel tube

Concrete core

Circular section Square section

Fig. 13. Fiber grids for circular and square CFST columns.
 H. Liu et al. / Engineering Structures 100 (2015) 308325 321

C 0:0727f ck =20 0:0216 55 followed by ultimate failure experiment. Their companion coun-
terparts which were maintained unloaded during the creep test
where f ck is the characteristic strength of concrete, equaling to are labeled as S-S-1 and S-S-2, respectively, with a main difference
0:67f cu for normal strength concrete with f cu the corresponding in their steelconcrete area ratios a As =Ac . Note that two batches
cube strength. Asc As Ac is the total cross-sectional area for the were tested for specimens S-L-1 and S-L-2.
specimens. n As f y =Ac f ck denotes the connement effect Fig. 15 presents a comparison of the experimentally measured
coefcient. loadstrain curves between specimens (S-L-1 and S-L-2) with sus-
The basic information for the tested specimens is summarized tained load previously applied and their loading-free counterparts
in Table 4. For convenience, we use the following nomenclature (S-S-1 and S-S-2). It is observed that the ultimate strength of the
consisting of two letters followed by a number to characterize each square columns subjected to sustained load is about 4.5% higher
specimen. The rst letter, C or S, denotes the cross-sectional than that of their loading-free counterparts. The strength enhance-
shape of a specimen, circular or square, respectively. The second ment is relatively small compared with that of circular specimens.
letter, S or L, represents a specimen subjected to short-term In addition the ductility change is not obvious after sustained load-
or long-term loading, respectively. Finally, the last number follow- ing. It may be explained by the non-uniform distribution of conn-
ing these letters is used to distinguish specimens subjected to dif- ing stress in these square members with a high D=t s ratio, which
ferent levels of sustained loading or with different geometric leads to little conning effect. Thus the effect of residual creep
dimensions. strain is minimized in these specimens.
For circular CFST columns, long-term tests were conducted rst The axial loadstrain responses obtained from the ber element
on specimens C-L-1 and C-L-2 for around 120 days, while their analysis are also plotted for all square specimens in Fig. 15 against
companion specimen C-S-1 with the same concrete mix was kept the experimentally measured N-e curves. As can be seen from
unloaded for the whole duration. Then all the specimens were fur- Fig. 15, the predicted ultimate strengths provide a good match to
ther loaded to failure. The major difference between specimens the experimental data. For the measured short-term loadstrain
C-L-1 and C-L-2 is that the latter specimen is loaded with a higher responses, three parts can be clearly identied. In the elastic stage
sustained load ratio (n 0:58), nally leading to a 40% higher of loading, an ascending branch can be observed. After reaching the
creep strain than that of the former one with a lower loading ratio peak point, the axial loading decreases drastically with the increase
(n 0:33). Note that two batches of specimen C-L-2 were tested. of axial strain, which is followed by a more gentle strain-softening
A comparison of the experimentally measured loadstrain behavior. Overall, the predicted responses from the ber element
curves between specimens (C-L-1 and C-L-2) subjected to sus- analysis capture the curve slope except for the second branch.
tained load and their loading-free counterpart C-S-1 is plotted in Regarding the long-term loadstrain curves, the ber element
Fig. 14. It is found that the ultimate strength of the circular speci- modeling provides a more accurate prediction compared with
mens after sustained load is about 12.5% higher than that of the the short-term one. In addition, the predicted ultimate strengths
loading-free counterpart. However, such a strength enhancement N uP are compared with the experimental results N uE in Table 5. A
is obtained at the expense of reduced ductility. The reason for this mean strength ratio N uP =N uE of 0.967 with a standard deviation
phenomenon has been explained by the mechanical analysis pre- of 0.020 is obtained.
sented in Section 3. For performance assessment, the axial load
strain curves predicted by the ber element analysis are compared
against the experimental results in Fig. 14. As can be seen, a gener- 6. Pushover analysis of CFST columns
ally good agreement is obtained between the predicted and tested
results, except that the slope of strain-softening branch is slightly In order to examine the effects of different parameters on the
overestimated by the ber element analysis. In consistence with long-term failure behavior of CFST columns, a parametric study
the test results, the predicted N-e curves for specimens C-L-1 and is conducted in this section using the above ber element analysis
C-L-2 also show higher ultimate strengths and steeper softening procedure. Given the similar behavior of circular and square CFST
branches than those of specimen C-S-1. Overall, the predictions columns, only the CFST columns with circular cross-sections are
from the ber element analysis are conservative but with reason- investigated in this parametric study. We investigate the perfor-
able accuracy. In addition, a summary of the predicted ultimate mance of CFST columns subject to a combined axial compression
strengths N uP and those ultimate strengths N uE from experimental and lateral bending, so called pushover analysis, as depicted in
tests are compared in Table 5. For specimens with two batches Fig. 16. The specimens are loaded with a sustained axial force N L
tested, the average of the measured ultimate strengths is used at t0 28 days after concrete casting. For the long-term failure
for comparison. A mean ratio N uP =N uE of 0.961 is obtained with a analysis, the specimens are laterally loaded (N P in Fig. 16) to failure
standard deviation of 0.049. at t = 2000 days. In other words, the axial sustained load is main-
In the case of square CFST columns, specimens S-L-1 and S-L-2 tained for 1972 days during which concrete creep deformation
underwent the same test procedure as experienced by the circular develops. For comparison, a short-term failure analysis is also stud-
members S-L-1 and S-L-2. That is to say, long-term creep test ied, where the lateral bending force is applied at the same time as
the axial compression (i.e. when t t 0 28 days). Note that only
short columns are considered in these studies and potential buck-
ling effects due to geometric nonlinearities are excluded.
Table 4
Basic information of selected test members for performance assessment [22,9]. Apart from the test setup, the geometric and material
parameters of the circular CFST columns as well as the axial
Specimen NL n t  t0 f c0 fy Es ts D
loading condition are summarized in Table 6. Based on that,
(KN) (d) (MPa) (MPa) (GPa) (mm) (mm)
the following coefcients are calculated: steelconcrete area
C-S-1 0 0 0 42.7 315 179 2.65 140 ratio a As =Ac 0:11 and long-term service load ratio
C-L-1 290 0.33 123 42.7 315 179 2.62 140
C-L-2 515 0.58 121 42.7 315 179 2.6 140
n N L =N u 0:4. We analyze the sensitivity of long term failure
S-S-1 0 0 0 34.3 293.5 202 2.93 100 of CFST columns to two critical parameters as listed in Table 7.
S-L-1 360 0.62 120 34.3 293.5 202 2.93 100 In order to characterize the ductility of CFST columns under
S-S-2 0 0 0 34.3 293.5 202 2.93 120 combined axial loading and bending, a ductility index g is intro-
S-L-2 470 0.62 120 34.3 293.5 202 2.93 120
duced in the study, dened as
322 H. Liu et al. / Engineering Structures 100 (2015) 308325

1400 1400

1200 1200

1000 1000
Axial load (kN)

Axial load (kN)

800 800

600 600
C-S-1 Experiment
400 C-S-1 Experiment 400 C-L-2 Experiment 1
C-L-1 Experiment C-L-2 Experiment 2
200 C-S-1 Prediction 200 C-S-1 Prediction
C-L-1 Prediction C-L-2 Prediction
0 0
0 5000 10000 15000 20000 0 5000 10000 15000 20000
Axal strain () Axial strain ()

(a) (b)
Fig. 14. Comparison of the predicted and experimental axial loadstrain curves for circular CFST columns: (a) C-S-1 and C-L-1 with n 0:33; (b) C-S-1 and C-L-2 with
n 0:58.

Table 5
Comparison of predicted and experimentally measured ultimate strengths (N uP stands
for predicted values and N uE experimental values).

Specimen C-S-1 C-L-1 C-L-2 S-S-1 S-L-1 S-S-2 S-L-2

N uP (KN) 1196.2 1298.5 1250.5 763.6 823.4 996.6 1085.3
N uE (KN) 1200.2 1352.3 1348.5 802 826.5 1054 1116
N uP =N uE 0.997 0.960 0.927 0.952 0.996 0.946 0.972

ueu
g 56
uey
where uey is the end deection of CFST columns at the lateral peak
strength. ueu denotes the end deection when the bending load
drops to 90% of the lateral peak strength.

6.1. Effect of strength enhancement coefcient kc0

The strength enhancement coefcient kc0 of plain concrete is
one of the most important factors that inuence the long-term fail-
ure behavior of CFST columns. It is usually estimated from the
plain concrete experiments with analogous concrete mix and sus-
tained loading duration. The enhancement coefcient kc0 adopted
in the parametric study ranges from 1.00 to 1.20. Fig. 17 shows
the effect of different strength enhancement coefcients kc0 on
the long-term lateral loading-deection curves for circular CFST
columns. In addition, the short-term lateral loaddeection behav-
ior of circular CFST columns is also plotted in Fig. 17 for
Fig. 16. Setup of pushover test of CFST columns.
comparison. The short-term and long-term pushover tests are
denoted hereinafter by the letter S and L following the param-
eter value, respectively. Note that the scale of the vertical coordi-
nate is selected to illustrate the difference between each case.
As shown in the gure, for kc0 0, the lateral peak strength of
the long-term pushover test is lower than that of the short-term
one. This is attributed to the presence of residual creep deforma-
tion in the long-term test. With the increase in kc0 , the lateral peak
strength of CFST columns is observed to gradually increase.

1000 1200

1000
800

800
Axial load (kN)

Axial load (kN)

600

600

400
S-S-1 Experiment S-S-2 Experiment
400
S-L-1 Experiment 1 S-L-2 Experiment 1
200
S-L-1 Experiment 2 S-L-2 Experiment 2
S-S-1 Prediction 200
S-S-2 Prediction
S-L-1 Prediction S-L-2 Prediction
0 0
0 2000 4000 6000 8000 10000 12000 14000 0 2000 4000 6000 8000 10000 12000 14000
Axial strain () Axial strain ()
(a) (b)
Fig. 15. Comparison of the predicted and experimental axial loadstrain curves for square CFST columns: (a) S-S-1 and S-L-1; (b) S-S-2 and S-L-2.
 H. Liu et al. / Engineering Structures 100 (2015) 308325 323

Table 6 However, such an increase in the lateral ultimate strength is

Geometric, material, and loading parameters for pushover analysis. accompanied by a signicant decrease in the ductility index, which
L (mm) D (mm) ts (mm) kc0 f y (MPa) f c0 (MPa) N L (kN) is listed in Table 8.
300 100 2.50 1.10 345 40 196
6.2. Effect of steelconcrete area ratio a

Table 7 For CFST composite structures, the concrete core is more sensi-
Different values assigned to each parameter involved in the parametric study. tive to the time-dependent effects caused by long-term loading
Parameters Case 1 Case 2 Case 3 Case 4 Case 5 than the steel tube. Thus the steelconcrete area ratio a, which
determines the weight ratio of each constituent, is expected to
kc0 1.00 1.05 1.10 1.15 1.20
a As =Ac 0.05 0.11 0.15 0.20 have a signicant effect on the time-dependent behavior of CFST
columns. In the parametric study, the steelconcrete area ratio a
is varied from 0.05 to 0.20 by changing the outer diameter D while
maintaining the same steel tube thickness t s 2:5 mm. The corre-
60 sponding lateral loaddeection curves of CFST columns are pre-
sented in Fig. 18 for both the short and long-term pushover tests.
The weight ratio of concrete section will decrease with an
58
increasing a, and accordingly its role in CFST composite columns
will be weakened. That is to say, with the increase of a, the CFST
56 columns will behave in a more ductile manner like steel and the
time-dependent effects brought about by sustained load will
Lateral load NP (KN)

54 become indistinctive. These analyses agree with the observation

from Fig. 18. As can be seen in Fig. 18, the lateral strength of
CFST columns is enhanced whereas the ductility is reduced by
52
kc0=1.00-S the sustained load when a =0.05. However, the difference between
kc0=1.00-L short-term and long-term lateral loaddeection curves is mar-
50
kc0=1.05-L ginal when a = 0.20.
kc0=1.10-L
48 kc0=1.15-L
7. Summary and conclusions
kc0=1.20-L
46
0 2 4 6 8 10 12 14 16 In this paper, we focus on the investigation of the mechanism
Normalized deflection u e /L (%) underlying the experimentally observed strength and ductility
change of CFST columns after sustained loading. The effect of the
Fig. 17. Effects of strength enhancement coefcient kc0 on the lateral load sustained loading is analyzed in detail. It is found that the
deection behavior of CFST columns. post-creep mechanical behavior of CFST columns depends on the
combined action of enhanced compressive strength of plain

Table 8
Ductility index with respect to kc0 .

Case kc0 = 1.00-S kc0 = 1.00-L kc0 = 1.05-L kc0 = 1.10-L kc0 = 1.15-L kc0 = 1.20-L
ueu =L (%) 13.75 12.9 12.15 11.4 10.7 10.25
uey =L (%) 3.5 3.4 3.25 3.15 3 2.9
g 3.93 3.79 3.74 3.62 3.57 3.53

300 60

=0.05-S =0.11-S
280 50
=0.11-L
=0.05-L
=0.15-S
Lateral load NP (KN)

Lateral load NP (KN)

260 40 =0.15-L
=0.20-S
240
=0.20-L
30

220 20

200 10

180 0
0 2 4 6 8 10 0 2 4 6 8 10
Normalized deflection ue /L (%) Normalized deflection ue /L (%)

(a) (b)

Fig. 18. Effects of steelconcrete area ratio on the lateral loaddeection behavior of CFST columns: (a) a 0:05; (b) a 0:11, 0.15, and 0.20.
324 H. Liu et al. / Engineering Structures 100 (2015) 308325
concrete and residual creep deformation. Inspired by this observa-
tion, we propose a uniaxial stressstrain relation aimed to charac-
terize the post-creep failure behavior of steel-conned concrete,
which is then validated against extensive FE simulations and
experimental database. The constitutive model is based on a sim-
plied analytical method solved iteratively to efciently predict
long-term deformation of CFST composite columns at any desig-
nated target time. The proposed constitutive model is then imple-
mented in a ber-type element to study the response of CFST
columns under axial compression. The predicted post-creep failure
behavior agrees well with limited experimental results for CFST
columns with circular and square cross-sections.
Furthermore, the effects of critical parameters on the long-term
failure behavior of CFST columns are investigated through a para-
metric analysis. Pushover analysis, consisting of combined axial
compression and lateral bending is studied, where it is shown that
the strength enhancement ratio kc0 , as well as the steelconcrete
area ratio a, plays an important role in the long-term response
(strength and ductility) of CFST columns. The time-dependent
effects of the sustained loading on the strength and ductility of
CFST columns become evident for large strength enhancement
ratio kc0 and small steelconcrete area ratio a.
While the model proposed is simple and efcient, more exper-
imental results are required to further improve and better calibrate
its parameters when analyzing the long-term failure behavior of
CFST columns.
Acknowledgements
The study in this paper was conducted during the rst authors
visit to Columbia University, which was supported by the China
Scholarship Council (201306210079), the National Science
Foundation of China (51038006, 51378292, and U1134110), and
Beijing Natural Science Foundation (8144051). The nancial sup-
port is greatly appreciated.
References
[1] Furlong RW. Strength of steel-encased concrete beam columns. J Struct Div Am
Soc Civil Eng 1967;93:113.
[2] Neogi P, Sen H, Chapman J. Concrete-lled tubular steel columns under
eccentric loading. Struct Eng 1969;47(5):18795.
[3] Knowles RB, Park R. Axial load design for concrete lled steel tubes. J Struct Div
1970;96(10):212553.
[4] Ghosh RS. Strengthening of slender hollow steel columns by lling with
concrete. Can J Civ Eng 1977;4(2):12733.
[5] Rangan BV, Joyce M. Strength of eccentrically loaded slender steel tubular
columns lled with high-strength concrete. ACI Struct J 1992;89(6):67681.
[6] Ge H, Usami T. Strength analysis of concrete-lled thin-walled steel box
columns. J Constr Steel Res 1994;30(3):25981.
[7] Fujinaga T, Matsui C, Tsuda K, Yamaji Y. Limiting axial compressive force and
structural performance of concrete lled steel circular tubular beam-columns.
In: Proceeding of the fth pacic structural steel conference, Seoul, Korea;
1998. p. 97984.
[8] Schneider SP. Axially loaded concrete-lled steel tubes. J Struct Eng
1998;124(10):112538.
[9] Han L-H, Tao Z, Liu W. Effects of sustained load on concrete-lled hollow
structural steel columns. J Struct Eng 2004;130(9):1392404.
[10] Huang C, Yeh Y-K, Liu G-Y, Hu H-T, Tsai K, Weng Y, et al. Axial load behavior of
stiffened concrete-lled steel columns. J Struct Eng 2002;128(9):122230.
[11] Sakino K, Tomii M, Watanabe K. Sustaining load capacity of plain concrete stub
columns by circular steel tubes. In: Proc international special conf on concrete-
lled steel tubular structure; 1985. p. 1128.
[12] Elremaily A, Azizinamini A. Behavior and strength of circular concrete-lled
tube columns. J Constr Steel Res 2002;58(12):156791.
[13] Wheeler A, Bridge R. The behaviour of circular concrete-lled thin-walled steel
tubes in exure. In: Proceedings of the 5th international conference on
composite construction in steel and concrete V. Reston, Virginia: American
Society of Civil Engineers; 2006. p. 41323.
[14] Morino S, Kawaguchi J, Cao Z-S. Creep behavior of concrete-lled steel tubular
members. In: Proc of an engineering foundation cotffer on steel-concrete
composite structure. Irsee: ASCE; 1996. p. 51425.
[15] Uy B. Static long-term effects in short concrete-lled steel box columns under
sustained loading. ACI Struct J 2001;98(1):96104.
[16] Naguib W, Mirmiran A. Creep modeling for concrete-lled steel tubes. J Constr
Steel Res 2003;59(11):132744.
[17] Ma YS, Wang YF. Creep of high strength concrete lled steel tube columns.
Thin-Wall Struct 2012;53:918.
[18] Gunnarsson AK, Lehman D, Roeder C, Kuder K. Long-term and seismic
performance of concrete-lled steel tube columns with conventional and
high-volume scm concrete, Tech. Rep. TNW2012-18, University of
Washington; 2012.
[19] Terrey P, Bradford M, Gilbert R. Creep and shrinkage of concrete in concrete-
lled circular steel tubes. In: Proc of 6th inter symposium on tubular
structures, Melbourne, Australia; 1994. p. 2938.
[20] Ichinose L, Watanabe E, Nakai H. An experimental study on creep of concrete
lled steel pipes. J Constr Steel Res 2001;57(4):45366.
[21] Uy B, Das S. Time effects in concrete-lled steel box columns in tall buildings.
Struct Des Tall Build 1997;6(1):122.
[22] Wang Y, Geng Y, Ranzi G, Zhang S. Time-dependent behaviour of expansive
concrete-lled steel tubular columns. J Constr Steel Res 2011;67(3):47183.
[23] Wang Y, Han B, Zhang D. Advances in creep of concrete lled steel tube
members and structures. In: Proc 8th international conference on creep,
shrinkage and durability mechanics of concrete and concrete structures; 2008.
p. 595600.
[24] Ranzi G, Leoni G, Zandonini R. State of the art on the time-dependent
behaviour of composite steelconcrete structures. J Constr Steel Res
2013;80:25263.
[25] Bazant ZP, Baweja S. Creep and shrinkage prediction model for analysis and
design of concrete structures: Model b3. ACI Spec Pub 2000;194:184.
[26] Comit euro-international du Bton. CEB-FIP model code 1990: design code,
no. 213-214, FIB-Fd. Int. du Bton; 1993.
[27] A.C. 209. Prediction of creep, shrinkage, and temperature effects in concrete
structures. ACI Spec Pub 1982;76:193300.
[28] Rilem draft recommendation: Tc-242-mdc multi-decade creep and shrinkage
of concrete: material model and structural analysis. Mater Struct 2015; 48(4):
75370.
[29] Wendner R, Hubler M, Bazant Z. Optimization method, choice of form and
uncertainty quantication of model b4 using laboratory and multi-decade
bridge databases. Mater Struct 2015;48(4):77196.
[30] Neville AM. Creep of concrete: plain, reinforced, and
prestressed. Amsterdam: Elsevier Science; 1971.
[31] Gopalakrishnan BK, Neville AM, Ghali A. A hypothesis on mechanism of creep
of concrete with reference to multiaxial compression. In: ACI journal
proceedings, vol. 67. ACI; 1970.
[32] Liu H. Load-bearing performance of prestressed conned concrete considering
creep [Ph.D. thesis]. Tsinghua University; 2014.
[33] Richart FE, Brandtzaeg A, Brown RL. A study of the failure of concrete under
combined compressive stresses, Tech. rep., Engineering Experiment Station,
University of Illinois, Bull. No. 185; 1928.
[34] ACI. Building code requirements for structural concrete and
commentary. American Concrete Institute; 1999.
[35] Tang J, Hino S, Kuroda I, Ohta T. Modeling of stressstrain relationships for
steel and concrete in concrete lled circular steel tubular columns. Steel
Constr Eng 1996;3(11):3546.
[36] Susantha K, Ge H, Usami T. Uniaxial stress-strain relationship of concrete
conned by various shaped steel tubes. Eng Struct 2001;23(10):133147.
[37] Han L-H, Yao G-H, Zhao X-L. Tests and calculations for hollow structural steel
(HSS) stub columns lled with self-consolidating concrete (SCC). J Constr Steel
Res 2005;61(9):124169.
[38] Cusson D, Paultre P. Stress-strain model for conned high-strength concrete. J
Struct Eng 1995;121(3):46877.
[39] Hughes B, Ash J. Some factors inuencing the long term strength of concrete.
Matriaux et Construction 1970;3(2):814.
[40] Dhir R, Sangha C. A study of the relationships between time, strength,
deformation and fracture of plain concrete. Mag Concr Res
1972;24(81):197208.
[41] Coutinho AS. A contribution to the mechanism of concrete creep. Matriaux et
Construction 1977;10(1):316.
[42] Hellesland J, Aas-Jakobsen IA, Green R. A stress and time dependent strength
law for concrete. Cem Concr Res 1972;2(3):26175.
[43] Hibbitt D, Karlsson B, Sorensen P. ABAQUS users manual: version 6.12,
Pawtucket, RI: ABAQUS; 2012.
[44] Hu H, Huang C, Wu M, Wu Y. Nonlinear analysis of axially loaded concrete-
lled tube columns with connement effect. J Struct Eng
2003;129(10):13229.
[45] Han L-H, Yao G-H, Tao Z. Performance of concrete-lled thin-walled steel
tubes under pure torsion. Thin-Wall Struct 2007;45(1):2436.
[46] Tao Z, Wang Z-B, Yu Q. Finite element modelling of concrete-lled steel stub
columns under axial compression. J Constr Steel Res 2013;89:12131.
[47] Luccioni B, Rougier V. A plastic damage approach for conned concrete.
Comput Struct 2005;83(27):223856.
[48] Yu T, Teng J, Wong Y, Dong S. Finite element modeling of conned concrete-i:
DruckerPrager type plasticity model. Eng Struct 2010;32(3):66579.
[49] Bigoni D, Piccolroaz A. Yield criteria for quasibrittle and frictional materials. Int
J Solids Struct 2004;41(11-12):285578.
[50] Lan Y-M. Finite element study of concrete columns with ber composite
jackets, Ph.D. thesis, Purdue University; 1998.
[51] Sakino K, Nakahara H, Morino S, Nishiyama I. Behavior of centrally loaded
concrete-lled steel-tube short columns. J Struct Eng 2004;130(2):1808.
 H. Liu et al. / Engineering Structures 100 (2015) 308325 325

[52] Schneider S. Axially loaded concrete-lled steel tubes. J Struct Eng
1998;124(10):112538.
[53] Huang C, Yeh Y-K, Liu G-Y, Hu H-T, Tsai K, Weng Y, Wang S, Wu M-H. Axial
load behavior of stiffened concrete-lled steel columns. J Struct Eng
2002;128(9):122230.
[54] Tao Z, Uy B, Liao F-Y, Han L-H. Nonlinear analysis of concrete-lled square
stainless steel stub columns under axial compression. J Constr Steel Res
2011;67(11):171932.
[55] McKenna F, Fenves G, Scott M, Jeremic B. Open system for earthquake
engineering simulation (opensees), Pacic Earthquake Engineering Research
Center, University of California, Berkeley, CA; 2000.
[56] Cook D, Chindaprasirt P. Inuence of loading history upon the compressive
properties of concrete. Mag Concr Res 1980;32(111):89100.
[57] Han L-H. Tests on stub columns of concrete-lled RHS sections. J Constr Steel
Res 2002;58(3):35372.