Académique Documents
Professionnel Documents
Culture Documents
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
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 The behavior of CFST columns under short-term loading has been
investigated in depth and well addressed by existing design codes.
During the past few decades, concrete-lled steel tubular (CFST) The related studies include Furlong [1], Neogi et al. [2], Knowles
columns have been increasingly used in a wide variety of structural and Park [3], Ghosh [4], Rangan and Joyce [5], Ge and Usami [6],
engineering applications, particularly in bridges and high-rise Fujinaga et al. [7], Schneider [8], Han et al. [9], Huang et al. [10],
buildings. The popularity of CFST columns can be attributed to Sakino et al. [11], Elremaily and Azizinamini [12], and Wheeler
their superior structural properties (e.g. high strength and ductil- and Bridge [13], to name just a few.
ity, good re resistance, and large energy absorption capacity), In permanent structures and facilities, CFST columns are usually
resulting from the combined action of the composite constituent subjected to long-term service loads before ultimate failure is
materials (steel and concrete). The steel tube serves as longitudinal experienced. Compared with the short-term deformation of CFST
and transverse reinforcement offering conning pressure to the columns, the creep and shrinkage of concrete core and its interac-
concrete inll. On the other hand, the concrete core can stabilize tion with the surrounding steel tube produce great uncertainties
and stiffen the surrounding steel tube in the sense that the local on the ultimate response of CFST columns with long-term service
buckling of the steel tube is prevented or at least delayed. loading applied. A few investigations, including both theoretical
Furthermore, this type of structural member has proven to reduce and experimental ones, were carried out in order to study the creep
the labor and material cost during the construction process since and shrinkage behavior of CFST columns, e.g. Morino et al. [14], Uy
external steel tubes can act as integral and permanent formworks. [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-
Corresponding author. Tel.: +1 212 851 0408; fax: +1 212 854 6267.
ture egress in a sealed environment [19,20]. In addition, the stress
E-mail address: waisman@civil.columbia.edu (H. Waisman).
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 309
redistribution between the concrete core and steel tube could simplied design method without repeating the complicated iter-
reduce the creep strain of concrete to some extent [21]. ative algorithm for each loading increment of the whole duration.
Compared with the research on the time-dependent creep and In other words, only one iterative process is needed for predicting
shrinkage behavior of CFST columns, models that describe the ulti- the nal status of creep. Another novel contribution of our model is
mate mechanical performance of CFST columns after long-term that a monolithic iterative scheme is developed instead of the stag-
service load remain relatively undeveloped. According to the gered one used in [16] to achieve better convergence of creep
pioneering experimental study conducted by Han et al. [9] and iterations.
Wang et al. [22], the circular and square CFST column specimens
under sustained load showed increase of up to 20% in 2.1. Creep model for conned concrete
load-bearing capacities when compared with their companion
load-free specimens. However, an opposite trend, i.e. reduction in The mechanical behavior of conned concrete is considerably
load-carrying capacity, has also been reported by other research- different from that of plain concrete, due to the sealed wrapping,
ers, e.g. Wang et al. [23]. It is still an open question [24] to explain variable stress history, and multi-axial stress status caused by
such divergent changes in structural strength. Aside from the the steel tube connement. Without loss of generality, we adopt
unclear mechanism, the lack of understanding of the potential the B3 model, developed by Bazant and Baweja [25], in the creep
inuence of long-term service loading on structural deformation analysis of conned concrete, considering that the B3 model is sim-
capacity (i.e. ductility) also underlines the necessity to conduct fur- ple yet well justied. Any other creep model, such as the CEB
ther research on this important, but generally neglected topic. model [26], the ACI 209 model [27] or the B4 model [28,29] intro-
This paper is devoted to exploring the mechanism underlying duced very recently, can equally be employed in our framework.
the experimentally observed strength and ductility change of Although full details of the B3 model can be found in Bazant and
CFST columns after long-term service loading. The remainder of Baweja [25], a brief description with the relevant equations is pro-
the paper is organized as follows. In Section 2, we establish a vided for clarity.
framework for efcient evaluation of creep strain of CFST columns Since the moisture migration of the concrete core is prevented
at any designated target time, imperative for the long-term failure by the surrounding steel tube [30], we neglect the drying creep
analysis of CFST columns. In this framework, the efciency of creep and shrinkage strain of the concrete inll in our study.
analysis is considerably enhanced by a monolithic iterative Furthermore, the concrete creep together with the deformation
scheme. In Section 3, we show that the post-creep mechanical compatibility can change the stress distribution between the con-
behavior of CFST columns depends on the combined action of crete core and the steel tube of CFST columns even though a xed
enhanced compressive strength of plain concrete and residual load is applied. Thus the varying stress history in concrete should
creep deformation. A uniaxial constitutive model, incorporating be taken into account.
the aforementioned effects brought about by long-term service The total strain et of a concrete member subjected to a uniax-
load, is proposed for steel-conned concrete core with both circu- ial stress rt can be calculated as follows:
lar and square cross-sections. Section 4 reports a nite element Z t
based numerical study on the effect of creep deformation on the et Jt; t 0 drt0 1
conning strength, which serves to validate the mechanical analy- 0
sis presented in Section 3. In Section 5, we further apply the pro- with the creep compliance function Jt; t 0 representing the total
posed stressstrain relation to ber element analysis where the strain at target time t due to a unit constant stress applied at age
nonlinear behavior of CFST columns under compression is investi- t0 . According to the B3 model, the compliance function Jt; t 0 can
gated. By comparison with the available experimental data, our be expressed as
model is found to perform well for both the circular and square
CFST members. In Section 6, the validated post-creep constitutive Jt; t 0 q1 C 0 t; t0 C d t; t 0 ; td 2
model is used for pushover analysis of CFST columns with where q1 is the instantaneous elastic strain due to the unit stress
long-term service load applied. A parametric study is conducted and C 0 t; t 0 the basic creep compliance. C d t; t 0 ; t d denotes the
to examine the inuence of critical design parameters on the additional compliance function for drying creep which started at
post-creep ultimate strength and ductility of CFST column mem- age td and has been neglected as stated above. Then the compliance
bers subjected to combined compression and bending loads. Final function can be further expanded as
concluding remarks are given in Section 7.
t
Jt; t 0 q1 q2 Q t; t 0 q3 ln 1 t t 0 n q4 ln 3
t0
2. Creep analysis of CFST columns
where q2 ; q3 , and q4 are empirical constitutive parameters; Q t; t0
can be easily obtained either from an approximate explicit formula
The creep deformation analysis of CFST columns includes two
or by numerical integration, such as those given in [25]; n is an
main ingredients, i.e. the initial short-term elastic analysis and
empirical parameter whose value can be taken as 0.1 for all normal
the subsequent long-term creep analysis. The focus in this paper
concrete types.
is on perfectly bonded composite columns, in which the concrete
In practice, the total integrated strain given in Eq. (1) can be
core and steel tube sustain the axial load together. The combined
approximated by a sum of strains caused by small stress incre-
effects of concrete creep, lateral connement provided by the steel
ments applied at all time intervals, that is
tube, and stress redistribution induced by geometric compatibility
(see Fig. 1) make the deformation of CFST columns under sustained X
n
1
load a complex time-dependent process. Following the framework etn Jtn ; t i Jt n ; ti1 Drt i 4
i1
2
proposed by Naguib and Mirmiran [16], we establish hereinafter a
predictive model with all aforementioned effects considered for where the subscripts n and i denote total and individual time steps,
the efcient assessment of creep deformation of CFST columns at respectively.
any designated target time. However, and in contrast to the previ- In Eqs. (1) and (4), the salient feature of conned concrete,
ously proposed framework [16], the creep strain of CFST columns namely the status of triaxial stresses, is not reected. According
at any target time can be conveniently obtained through a 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
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
Fig. 2. Schematic representation of geometric and loading information for steel tube with various cross-sections: (a) circular section; (b) square section.
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
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
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
Table 1
Material, geometric, and loading parameters for the creep test conducted by Terrey
et al. [19].
600 80
Naguib's method
70 Simplified method
500
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.
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
are dened as the master and slave surfaces, respectively. Hard assumed to be 0.2. Considering the strong correlation between
contact is specied in the normal direction of the interface to pre- the Youngs modulus Ec of concrete and its compressive strength,
vent penetration whereas the Rough option in ABAQUS is used the empirical formula recommended by ACI Committee 318 [34]
in the tangential direction to reproduce the experimental observa- is employed to calculate the Youngs modulus Ec
tion that there is little or no slip between the steel tube and con- q
crete inll. In order to obtain a better convergence rate in Ec 4700 f c0 MPa 51
contact analysis, the meshes on the steelconcrete contact surface
are matched. where the compressive strength f c0 is in MPa.
Fig. 8 shows an example of the mesh geometry of a typical cir- A linear DruckerPrager yield criterion with a non-associated
cular specimen with its boundary conditions. The column is axially ow rule model available in ABAQUS is employed to model the
loaded at the top with the stiff plate of the test machine. For the concrete failure [48] in this study. The adopted plastic model
failure analysis of CFST columns, displacement control is recom- accounts for the pressure-sensitive yielding exhibited by
mended to capture the potential softening branch of the axial load quasi-brittle and frictional materials [49] such as concrete and
(N)- axial strain (e) curves. Apart from the failure analysis, the rock. Several parameters are required for using the linear
time-dependent creep behavior of CFST columns before ultimate DruckerPrager model, these being, the friction angle (b), the ratio
failure needs to be investigated. In this case, the axial sustained of ow stress in triaxial tension to that in compression (K), the dila-
loading should be kept constant. To this end, a reference point at tion angle (w), and the ow potential eccentricity (). They are cal-
the center of the top end is dened and all nodes located at the ibrated to be 50 , 1.0, 30 , and 0.1 [50], respectively. In addition,
top end are kinematically coupled with the reference point by the strain hardening/softening curve, dened as a function of the
using multi-point constrains provided by ABAQUS. In the analysis equivalent plastic strain, can be determined from the short-term
step of creep, a concentrated force in the axial direction is applied uniaxial stressstrain relation (31) of concrete, either with or with-
at the reference point whose translational degrees of freedom at out connement. For a better comparison between the models, we
the other two directions are restrained. In the subsequent failure use the short-term constitutive relation, and so the potential inu-
analysis, the axial loading condition is replaced by a displacement ence of the concrete strength enhancement caused by the sus-
boundary condition specied along the column length. For the col- tained load can be eliminated from all numerical tests. Hence the
umn base, all three translational degrees of freedom are con- conning strength change due to sustained loading is only attribu-
strained in all analysis steps. ted to the existence of residual creep deformation and the correla-
tion between them can be more easily investigated in the
4.3. Material models numerical study.
In order to model the time-dependent creep deformation of
4.3.1. Steel tube concrete, a creep constitutive law should be dened. Considering
The Youngs modulus Es and Poissons ratio ls of steel are taken the variable stress history during the analysis, we employ the
to be 200 GPa and 0.3, respectively, if their values were not strain hardening power law available in ABAQUS, which is dened
reported in the test. Various constitutive models have been used in the following rate form
in the literature to model the steel tube, for example, the 1
elastic-perfectly plastic model used in [44] and the elasticplastic cr n m 1 ecr m m1
e_ cr A r 52
model with multi-linear hardening used in [45]. As stated in Tao
where r cr and ecr denote the uniaxial equivalent creep stress and
et al. [46], the steel tube exhibits insignicant strain hardening at
the strain levels of general structural interest (normally less than strain, e_ cr the uniaxial creep strain rate, A; n, and m the
5%), especially in cases of square CFST columns. Therefore, the user-specied creep material parameters. The creep law can be
elastic-perfectly plastic model is adopted in this study for the mod- integrated with respect to time either analytically or numerically.
eling of steel tube. The stressstrain relation of the steel material is The integrated equation is then compared to the results of creep
controlled by the Von Mises yield criterion with associated plastic or relaxation tests to calibrate the creep material parameters A; n,
ow rule using the PLASTIC option in ABAQUS. A schematic plot and m. In order to represent the creep test data [19], A; n, and m
0:8
of the uniaxial stressstrain curve of steel is depicted in Fig. 9 for are chosen to be 7 1012 MPa5 d , 5, and 0.2, respectively.
clarity. Note that steel creep is not considered here since it is
assumed to be negligible compared to concrete creep. 4.4. Model verication
4.3.2. Conned concrete In order to examine and verify the accuracy of the nite element
The Poissons ratio lc of concrete normally ranges from 0.15 to model, the short-term failure analysis is carried out rst on a total
0.22 [47]. When experimental values are not available, lc is of 24 CFST columns. The short-term analysis can be treated as the
Y
Z
X
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.)
318 H. Liu et al. / Engineering Structures 100 (2015) 308325
Table 2
Short-term test data and simulation results for circular CFST columns.
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.
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)
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
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
Steel tube
Concrete core
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)
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).
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.
1000 1200
1000
800
800
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 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)
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)
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- [16] Naguib W, Mirmiran A. Creep modeling for concrete-lled steel tubes. J Constr
Steel Res 2003;59(11):132744.
tion, we propose a uniaxial stressstrain relation aimed to charac-
[17] Ma YS, Wang YF. Creep of high strength concrete lled steel tube columns.
terize the post-creep failure behavior of steel-conned concrete, Thin-Wall Struct 2012;53:918.
which is then validated against extensive FE simulations and [18] Gunnarsson AK, Lehman D, Roeder C, Kuder K. Long-term and seismic
experimental database. The constitutive model is based on a sim- performance of concrete-lled steel tube columns with conventional and
high-volume scm concrete, Tech. Rep. TNW2012-18, University of
plied analytical method solved iteratively to efciently predict Washington; 2012.
long-term deformation of CFST composite columns at any desig- [19] Terrey P, Bradford M, Gilbert R. Creep and shrinkage of concrete in concrete-
nated target time. The proposed constitutive model is then imple- lled circular steel tubes. In: Proc of 6th inter symposium on tubular
structures, Melbourne, Australia; 1994. p. 2938.
mented in a ber-type element to study the response of CFST [20] Ichinose L, Watanabe E, Nakai H. An experimental study on creep of concrete
columns under axial compression. The predicted post-creep failure lled steel pipes. J Constr Steel Res 2001;57(4):45366.
behavior agrees well with limited experimental results for CFST [21] Uy B, Das S. Time effects in concrete-lled steel box columns in tall buildings.
Struct Des Tall Build 1997;6(1):122.
columns with circular and square cross-sections. [22] Wang Y, Geng Y, Ranzi G, Zhang S. Time-dependent behaviour of expansive
Furthermore, the effects of critical parameters on the long-term concrete-lled steel tubular columns. J Constr Steel Res 2011;67(3):47183.
failure behavior of CFST columns are investigated through a para- [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,
metric analysis. Pushover analysis, consisting of combined axial shrinkage and durability mechanics of concrete and concrete structures; 2008.
compression and lateral bending is studied, where it is shown that p. 595600.
the strength enhancement ratio kc0 , as well as the steelconcrete [24] Ranzi G, Leoni G, Zandonini R. State of the art on the time-dependent
behaviour of composite steelconcrete structures. J Constr Steel Res
area ratio a, plays an important role in the long-term response
2013;80:25263.
(strength and ductility) of CFST columns. The time-dependent [25] Bazant ZP, Baweja S. Creep and shrinkage prediction model for analysis and
effects of the sustained loading on the strength and ductility of 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,
CFST columns become evident for large strength enhancement
no. 213-214, FIB-Fd. Int. du Bton; 1993.
ratio kc0 and small steelconcrete area ratio a. [27] A.C. 209. Prediction of creep, shrinkage, and temperature effects in concrete
While the model proposed is simple and efcient, more exper- structures. ACI Spec Pub 1982;76:193300.
imental results are required to further improve and better calibrate [28] Rilem draft recommendation: Tc-242-mdc multi-decade creep and shrinkage
of concrete: material model and structural analysis. Mater Struct 2015; 48(4):
its parameters when analyzing the long-term failure behavior of 75370.
CFST columns. [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.
Acknowledgements [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
The study in this paper was conducted during the rst authors of concrete with reference to multiaxial compression. In: ACI journal
visit to Columbia University, which was supported by the China proceedings, vol. 67. ACI; 1970.
Scholarship Council (201306210079), the National Science [32] Liu H. Load-bearing performance of prestressed conned concrete considering
creep [Ph.D. thesis]. Tsinghua University; 2014.
Foundation of China (51038006, 51378292, and U1134110), and [33] Richart FE, Brandtzaeg A, Brown RL. A study of the failure of concrete under
Beijing Natural Science Foundation (8144051). The nancial sup- combined compressive stresses, Tech. rep., Engineering Experiment Station,
port is greatly appreciated. University of Illinois, Bull. No. 185; 1928.
[34] ACI. Building code requirements for structural concrete and
commentary. American Concrete Institute; 1999.
References [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.
[1] Furlong RW. Strength of steel-encased concrete beam columns. J Struct Div Am
[36] Susantha K, Ge H, Usami T. Uniaxial stress-strain relationship of concrete
Soc Civil Eng 1967;93:113.
conned by various shaped steel tubes. Eng Struct 2001;23(10):133147.
[2] Neogi P, Sen H, Chapman J. Concrete-lled tubular steel columns under
[37] Han L-H, Yao G-H, Zhao X-L. Tests and calculations for hollow structural steel
eccentric loading. Struct Eng 1969;47(5):18795.
(HSS) stub columns lled with self-consolidating concrete (SCC). J Constr Steel
[3] Knowles RB, Park R. Axial load design for concrete lled steel tubes. J Struct Div
Res 2005;61(9):124169.
1970;96(10):212553.
[38] Cusson D, Paultre P. Stress-strain model for conned high-strength concrete. J
[4] Ghosh RS. Strengthening of slender hollow steel columns by lling with
Struct Eng 1995;121(3):46877.
concrete. Can J Civ Eng 1977;4(2):12733.
[39] Hughes B, Ash J. Some factors inuencing the long term strength of concrete.
[5] Rangan BV, Joyce M. Strength of eccentrically loaded slender steel tubular
Matriaux et Construction 1970;3(2):814.
columns lled with high-strength concrete. ACI Struct J 1992;89(6):67681.
[40] Dhir R, Sangha C. A study of the relationships between time, strength,
[6] Ge H, Usami T. Strength analysis of concrete-lled thin-walled steel box
deformation and fracture of plain concrete. Mag Concr Res
columns. J Constr Steel Res 1994;30(3):25981.
1972;24(81):197208.
[7] Fujinaga T, Matsui C, Tsuda K, Yamaji Y. Limiting axial compressive force and
[41] Coutinho AS. A contribution to the mechanism of concrete creep. Matriaux et
structural performance of concrete lled steel circular tubular beam-columns.
Construction 1977;10(1):316.
In: Proceeding of the fth pacic structural steel conference, Seoul, Korea;
[42] Hellesland J, Aas-Jakobsen IA, Green R. A stress and time dependent strength
1998. p. 97984.
law for concrete. Cem Concr Res 1972;2(3):26175.
[8] Schneider SP. Axially loaded concrete-lled steel tubes. J Struct Eng
[43] Hibbitt D, Karlsson B, Sorensen P. ABAQUS users manual: version 6.12,
1998;124(10):112538.
Pawtucket, RI: ABAQUS; 2012.
[9] Han L-H, Tao Z, Liu W. Effects of sustained load on concrete-lled hollow
[44] Hu H, Huang C, Wu M, Wu Y. Nonlinear analysis of axially loaded concrete-
structural steel columns. J Struct Eng 2004;130(9):1392404.
lled tube columns with connement effect. J Struct Eng
[10] Huang C, Yeh Y-K, Liu G-Y, Hu H-T, Tsai K, Weng Y, et al. Axial load behavior of
2003;129(10):13229.
stiffened concrete-lled steel columns. J Struct Eng 2002;128(9):122230.
[45] Han L-H, Yao G-H, Tao Z. Performance of concrete-lled thin-walled steel
[11] Sakino K, Tomii M, Watanabe K. Sustaining load capacity of plain concrete stub
tubes under pure torsion. Thin-Wall Struct 2007;45(1):2436.
columns by circular steel tubes. In: Proc international special conf on concrete-
[46] Tao Z, Wang Z-B, Yu Q. Finite element modelling of concrete-lled steel stub
lled steel tubular structure; 1985. p. 1128.
columns under axial compression. J Constr Steel Res 2013;89:12131.
[12] Elremaily A, Azizinamini A. Behavior and strength of circular concrete-lled
[47] Luccioni B, Rougier V. A plastic damage approach for conned concrete.
tube columns. J Constr Steel Res 2002;58(12):156791.
Comput Struct 2005;83(27):223856.
[13] Wheeler A, Bridge R. The behaviour of circular concrete-lled thin-walled steel
[48] Yu T, Teng J, Wong Y, Dong S. Finite element modeling of conned concrete-i:
tubes in exure. In: Proceedings of the 5th international conference on
DruckerPrager type plasticity model. Eng Struct 2010;32(3):66579.
composite construction in steel and concrete V. Reston, Virginia: American
[49] Bigoni D, Piccolroaz A. Yield criteria for quasibrittle and frictional materials. Int
Society of Civil Engineers; 2006. p. 41323.
J Solids Struct 2004;41(11-12):285578.
[14] Morino S, Kawaguchi J, Cao Z-S. Creep behavior of concrete-lled steel tubular
[50] Lan Y-M. Finite element study of concrete columns with ber composite
members. In: Proc of an engineering foundation cotffer on steel-concrete
jackets, Ph.D. thesis, Purdue University; 1998.
composite structure. Irsee: ASCE; 1996. p. 51425.
[51] Sakino K, Nakahara H, Morino S, Nishiyama I. Behavior of centrally loaded
[15] Uy B. Static long-term effects in short concrete-lled steel box columns under
concrete-lled steel-tube short columns. J Struct Eng 2004;130(2):1808.
sustained loading. ACI Struct J 2001;98(1):96104.
H. Liu et al. / Engineering Structures 100 (2015) 308325 325
[52] Schneider S. Axially loaded concrete-lled steel tubes. J Struct Eng [55] McKenna F, Fenves G, Scott M, Jeremic B. Open system for earthquake
1998;124(10):112538. engineering simulation (opensees), Pacic Earthquake Engineering Research
[53] Huang C, Yeh Y-K, Liu G-Y, Hu H-T, Tsai K, Weng Y, Wang S, Wu M-H. Axial Center, University of California, Berkeley, CA; 2000.
load behavior of stiffened concrete-lled steel columns. J Struct Eng [56] Cook D, Chindaprasirt P. Inuence of loading history upon the compressive
2002;128(9):122230. properties of concrete. Mag Concr Res 1980;32(111):89100.
[54] Tao Z, Uy B, Liao F-Y, Han L-H. Nonlinear analysis of concrete-lled square [57] Han L-H. Tests on stub columns of concrete-lled RHS sections. J Constr Steel
stainless steel stub columns under axial compression. J Constr Steel Res Res 2002;58(3):35372.
2011;67(11):171932.