Strength of Circular Concrete-Filled Tubes

with and without Internal Reinforcement

under Combined Loading
Jiho Moon 1; Dawn E. Lehman, M.ASCE 2; Charles W. Roeder, M.ASCE 3;
and Hak-Eun Lee 4
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.

Abstract: Concrete-filled tubes (CFTs) have been used in civil engineering practice as piles, caissons, columns, and bridge piers. Relative to
conventional structural steel and reinforced concrete components, CFTs have several advantages. The steel tube serves as both reinforcement
and formwork, eliminating the need for both, and provides large tensile and compressive capacities; the concrete fill restrains buckling of the
steel tube, which increases the strength, stiffness, and deformability of the section. In some cases, internal reinforcement is used to enhance
the strength and facilitate connection to adjacent members. Although these properties are well accepted, the use of CFTs in practice is
awkward because design provisions among codes vary significantly and previous research has not considered internal reinforcement.
An analytical research study was undertaken to evaluate and improve design provisions for CFTs with and without internal reinforcement
under combined axial load and bending. A continuum model was developed to simulate prior test results subjected to combined loading
and the validated model was used to investigate the strength and inelastic performance of CFTs under combined loading. Current design
provisions for CFTs were evaluated using the results of these finite-element analysis and previous test results. The comparisons indicate
that current design approach provides good prediction of CFT capacity subjected only to bending or axial demands, but current provisions
provide conservative values for the CFTs under general combined loading. An alternative P-M interaction curve for CFTs was proposed.
DOI: 10.1061/(ASCE)ST.1943-541X.0000788. © 2013 American Society of Civil Engineers.
Author keywords: Concrete-filled tubes (CFTs); Composite behavior; P-M interaction curve; Structural design; Buckling; Nonlinear
finite-element analysis; Metal and composite structures.

Introduction concrete fill is not necessary, however, internal reinforcement is

Concrete-filled tubes (CFTs) are composite members that consist of
a steel tube and concrete infill. In Asia, they have been used as
building columns and bridge piers as an alternative to conventional
reinforced concrete construction. A primary benefit of CFTs is
that the concrete fill provides large compressive load capacity
and restrains buckling of the steel tube, which results in signifi-
cant resistance and inelastic deformation capacity (Marson and
Bruneau 2004; Roeder et al. 2010). They also provide economy
and rapid construction because the steel tube replaces the form-
work and reinforcement. The use of self-consolidating concrete
without vibration of concrete further accelerates construction
(Roeder et al. 2010). In general, internal reinforcement in the
1
Research Professor, Civil, Environmental & Architectural Engineer-
ing, Korea Univ., 335 Engineering Building, Seoul 136-701, South Korea.
E-mail: jmoon1979@gmail.com
2
Associate Professor, Dept. of Civil & Environmental Engineering,
Univ. of Washington, 214B More Hall, Seattle, WA 98195. E-mail:
delehman@u.washington.edu
3
Professor, Dept. of Civil & Environmental Engineering, Univ. of
Washington, 233B More Hall, Seattle, WA 98195 (corresponding author).
E-mail: croeder@u.washington.edu
4
Professor, Civil, Environmental & Architectural Engineering, Korea
Univ., 305 Engineering Building, Seoul 136-701, South Korea. E-mail:
helee@korea.ac.kr
Note. This manuscript was submitted on August 18, 2011; approved on
December 7, 2012; published online on December 10, 2012. Discussion
period open until February 9, 2014; separate discussions must be submitted
for individual papers. This paper is part of the Journal of Structural En-
gineering, © ASCE, ISSN 0733-9445/04013012(12)/$25.00.
occasionally used in piles, caissons, and bridge piers in order to
enhance the strength and facilitate connection to other elements.
CFTs have had limited use in the United States. In part, this
results from U.S. construction practices and inconsistencies in
the design approach. Currently, AISC (2010), American Concrete
Institute (ACI) (2008), and AASHTO (2009) provide design pro-
visions for CFTs. AISC permits two methods for predicting the
resistance of CFT members: (1) a simplified plastic stress distribu-
tion approach, or (2) a more complex strain compatibility method.
Similarly, ACI uses a general strain compatibility method that
can be simplified to a plastic stress distribution similar to AISC.
The three major differences in the simplified approaches are
that (1) AISC approximates the stress in the uniform block to
be as 0.95fc0 for CFTs while ACI approximates the uniform
strength as 0.85fc0 , (2) AISC uses the entire height of the stress
block while ACI recommends that it be reduced by a factor β
(β < 0.85 depending on concrete strength), and (3) ACI limits the
axial strength to 80% of the squash load with an additional factor to
account for eccentricity where AISC treats the instability explicitly.
AASHTO also addresses the design of circular CFTs, but these pro-
visions are less mature than the AISC and ACI provisions
AISC and ACI also differ on column design in their handling
of both length effects and effective flexural stiffness with respect
to long-term effects for gravity loads. Several researchers (Varma
et al. 2002; Bruneau and Marson 2004; Choi et al. 2006; Roeder
et al. 2010) have investigated the design method of CFTs, yet
there is not consensus on the correct approach among the different
design provisions.

© ASCE 04013012-1 J. Struct. Eng.

J. Struct. Eng. 2013.139.

 In addition, CFT sections may have internal reinforcement and
there are no test data or rigorous analyses to support the design
methods for a reinforced CFT component. This research project
was undertaken to develop a uniform and consistent approach to
designing CFT beam-column elements (i.e., elements subjected
to combined loading) with or without internal reinforcement. Veri-
fied nonlinear continuum models were used where the verification
included both CFT components subjected to axial load only or
combined loading. The nonlinear continuum model directly simu-
lated confinement effects, local buckling, bond stress, and slip be-
tween steel tube and concrete infill, thereby permitting a robust
parametric study.
The verified model was used to conduct a parametric study to
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.
but concrete stresses of 0.95fc0 and 1.0fc0 are used in AISC
and Eurocode 4, respectively. The coefficient of 0.95 was used
in this paper.
In Fig. 1, t is wall thickness of the tube, y is the distance from the
center of the section to the neutral axis for the specific load combi-
nation, rm is the distance from the origin to the center of the steel
tube, ri is the radius from the origin to the inside of the steel tube,
c is the half-width of the compression area, fy and fyb are yield
stress of the steel tube and reinforcement, respectively, and fc0 is
the compressive strength of the concrete. The variable P is the
applied axial force and P is positive when it acts in compression.
For simplicity, the internal reinforcement was replaced by an inter-
nal steel ring with an equivalent total area and an equivalent thick-

investigate the impact of diameter to thickness (D=t), internal ness, teq , and radius, rbm .
reinforcement (ρi ), and axial load (P=Po ) ratios. The results were Based on the equilibrium condition in Fig. 1, the moment, M,
combined with past experimental results to evaluate internationally and axial force, P, can be determined from an assumed neutral axis
accepted design provisions including AISC (2010) and Eurocode 4 depth, as provided in Eqs. (1a) and (1b)
[European Committee for Sandardization (CEN) 2004] provisions.
Based on the results, an alternative P-M interaction curve for θ
P ¼ −2fyb NAb b − 4fy trm θ
CFT was proposed. The proposed curve is based on the plastic- π

 
stress distribution method (PSDM) combined with stability expres- 0.95fc0 2θ
sions to estimate the flexural strength of low and moderate axial þ ðπ − 2θÞr2i − 2cy − NAb 1 − b ð1aÞ
2 π
load ratios.

2 r2
Design Approach for Circular CFTs with and without M¼ fyb NAb rbm cos θb þ 4fy tc m
π ri
Internal Reinforcing
 
0 2c 2 2 rbm cos θb
þ 0.95f c ðr − y Þ − NAb ð1bÞ
To assess the flexural strength of CFTs under combined loading, 3 i π
either the plastic-stress distribution (AISC 2010; CEN 2004) or
train-compatibility methods (AISC 2010; ACI 2008) are used. where
The plastic stress distribution method in AISC (2010) was adapted y y
by Leon et al. (2007) from the Eurocode (Roik and Bergmann sin θb ¼ ; sin θ ¼ ;
rbm rm
1992). Roeder et al. (2010) reported that the plastic-stress distribu- ð1cÞ
tion method is the method of choice for its simplicity and accuracy. NAb
c ¼ ri cos θ; and teq ¼
Thus, only the plastic-stress distribution method was considered in 2πrbm
this study.
The plastic-stress distribution method is an equilibrium-based In Eq. (1), N and Ab are the number of internal reinforcing bars
method used to determine the strength of a CFT member where and the individual bar area, respectively. Eq. (1) is valid when the
instability effects due to global buckling are not considered. Fig. 1 value of θb is between −π=2 and π=2.
shows a cross section of a CFT component with internal reinforce- Fig. 2(a) shows implementation of the PSDM (dashed line) for a
ment. The figure also shows a typical stress distribution in the tube, CFT cross section without internal reinforcement and a D=t ratio of
reinforcement, and concrete fill. The length of each of the stress 60 (therefore, ρi ¼ 0). The axial load and moment are normalized
components is determined by satisfying equilibrium for a given bending only and axial load only cases, where Po;AISC and M o;AISC
externally applied axial load and moment. The AISC (2010) and are the squash load (M ¼ 0) and flexural strength corresponding to
Eurocode 4 (CEN 2004) plastic distribution methods are similar, no axial load (P ¼ 0) from AISC provisions, respectively.

Asc: compression area of steel tube, Acc: compression area of concrete infill
Abc: compression area of reinforcement, Ast: tension area of steel tube
Abt: tension area of reinforcement
2C
Steel tube Concrete infill
fy fyb 0.95fc fyAsc
0.95fc Acc
N.A.
fybAbc
rbm rm y
t b
teq o
P
ri fyAst
fybAbt
fy fyb
Internal Steel Rebar Concrete Equilibrium
reinforcement stress stress stress condition

Fig. 1. Geometry of CFTs with internal reinforcement, stress distribution, and equilibrium condition

© ASCE 04013012-2 J. Struct. Eng.

J. Struct. Eng. 2013.139.

 1.2 1.2 D/t=60, L/D=8, i=0% 1.2 D/t=60, L/D=8,
D/t=60, L/D=8, i=0% i=0%
A A
1.0 1.0 1.0 A
Interaction curve Interaction curve
from PSDM A' A"
0.8 A' 0.8 0.8
A' from PSDM
P
P/Po,AISC

P/Po,EC4
k P

P/Po,EC4
0.6 C H 0.6 0.6 H
AISC bilinear curve C'
0.4 0.4 0.4
without resistance factor d
D Interaction curve
0.2 D' 0.2 0.2 including global buckling
E
B B B
0.0 0.0 0.0
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5
M/Mo,AISC M/Mo,EC4 M/Mo,EC4
(a) (b)
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Example of P-M interaction curves: (a) AISC; (b) Eurocode 4

This interaction curve does not account for buckling. To con- factor on the gross cross-sectional stiffness of the concrete to ac-
sider length slenderness effect, the critical buckling load, Pcr , count for material nonlinearity (e.g., cracking).
must be determined. Based on the AISC provisions, Pcr can be Three different values of C3 were investigated in this study [EIeff
determined as from AISC (2010), Eurocode 4 (2004), and Roeder et al. (2010)]
2
and are given by
Pcr =Po;AISC ¼ 0.658λ when λ ≤ 1.5 ð2aÞ  
As
C3 ¼ 0.6 þ 2 ≤ 0.9 by AISC ð2010Þ ð5aÞ
Pcr =Po;AISC ¼ 0.877=λ2 when λ > 1.5 ð2bÞ As þ Ac

where C3 ¼ 0.6 by Eurocode 4 ðCEN 2004Þ ð5bÞ

sffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Po;AISC πEIeff  
λ¼ ; Pecr ¼ ; and C3 ¼ 0.15 þ
P
þ
As
≤ 0.9 by Roeder et al: ð2010Þ ð5cÞ
Pecr ðkLÞ2 Po As þ Ac
Po;AISC ¼ 0.95f c0 Ac þ f y As þ f yb Ar ð2cÞ

The variables used in Eq. (2) are defined as follows: k is the
effective buckling length factor; EIeff is the effective stiffness;
Ac and As are the area of the concrete infill and steel tube, respec-
tively; and Ar is the total area of the internal reinforcement.
Similarly, Eurocode 4 (CEN 2004) also provides the equation
for Pcr as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pcr =Po;EC4 ¼ 1=ðϕ þ ϕ2 − λ2 Þ ≤ 1 ð3aÞ
ϕ ¼ 0.5½1 þ αðλ − 0.2Þ þ λ2
;
Po;EC4 ¼ fc0 Ac þ f y As þ fyb Ar
where α = imperfection factor, which accounts for initial imperfec-
tion and residual stress of the section. The values of α are 0.21 and
0.34 for buckling curves a and b, respectively, which depend on the
section properties of CFTs. Eurocode 4 (CEN 2004) accounts for
the confinement effect of the CFT in predicting the squash load
when λ < 0.5 and el =D < 0.1, where el is the eccentricity of the
load. Thus, Po;EC4 could be larger than that from Eq. (3b). The de-
tailed equations are shown in Eurocode 4 (CEN 2004).
The effective stiffness of the CFT is an important section prop-
erty that is needed to determine the buckling capacity. Generally,
the effective stiffness of the CFT is expressed in the form
EIeff ¼ Es I s þ Eb I b þ C3 Ec I c
where Es , Eb , and Ec = modulus of elasticity values of the steel
tube, internal reinforcement, and concrete infill, respectively; I s ,
I b , and I c = second moment of area of the steel tube, internal
reinforcement, and concrete infill, respectively; and C3 = reduction
After the P-M interaction curve from the plastic-stress distribu-
tion method and Pcr are determined, the P-M interaction curve
considering the effect of global buckling can be obtained. Both
AISC (2010) and Eurocode 4 (2004) provide methods.
AISC provides three different methods to construct the P-M
interaction curve, and Method 2 was adopted in this study because
this method is based on the plastic-stress distribution method.
Fig. 2(a) highlights four points used to establish the P-M interac-
tion surface based on Method 2 in AISC provisions (AISC 2010):
point A is the full squash load (M ¼ 0), point B corresponds to
and flexural strength without axial load (M o;AISC ), point C has the same
flexural strength as point B but with axial load, and point D cor-
ð3bÞ responds to an axial load of one-half of that determined for point C.
The AISC method reduces points A, C, and D to points A 0 , C 0 , and
D 0 to account for global buckling effect by multiplying by a reduc-
tion factor of the critical load to the squash load Pcr =Po;AISC .
However, with this approach, the reduced flexural strength may
not be conservative at point D 0 because it may result in larger flexu-
ral strength than predicted by the plastic-stress distribution method.
This potential problem can be eliminated by (1) estimation of D 0 on
the PSDM curve or (2) removal of point D from the interaction
surface; the latter approach is adopted by AISC (2010). The result-
ing simple design interaction curve for CFTs is then constructed by
connecting points A 0 , C 0 , and B, as shown in Fig. 2(a). This bilinear
interaction curve is used through this paper for comparison with
ð4Þ results of analysis and tests.
Fig. 2(b) shows the construction of the P-M interaction curve
including the global buckling effect based on Eurocode 4 (CEN
2004). Points A and B [Fig. 2(b)] are the same as points A and
B in AISC (2010). Similarly to AISC, the full squash load is re-
duced to point A 0 with reduction factor Pcr =Po;EC4. The moment

© ASCE 04013012-3 J. Struct. Eng.

J. Struct. Eng. 2013.139.

 capacity at point A 0 , μk , is then removed from the interaction curve
and this removal moment is assumed to decrease linearly to
the point E. The detail of determination of point E is shown in
Eurocode 4 (CEN 2004).
Analytical Model Development and Verification
Model Overview
A total of 197 nonlinear finite-element models were analyzed with
the ABAQUS nonlinear finite-element analysis program to investi-
gate the strength of the CFT (36 models for CFTs under bending,
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.
48 models for CFTs under axial load, and 113 models of CFTs
under combined loading). Fig. 3 shows the configuration of the
finite-element model used to model a cantilever columns subjected
to combined loading. The model must accurately simulate the full
response to axial and bending loads, including any relative move-
ment between the steel shell and concrete fill. To enable this, four
different types of elements were used to model the concrete fill,
steel shell, internal reinforcement, and interface between the steel
shell and concrete fill, as shown in Fig. 3(a). The eight-node solid
element (C3D8R) was used to model the concrete, the four-node
shell element (S4R) was used to model the steel shell, and the two-
node truss element (T3D2) was used for the internal reinforcement.
The GAP element in ABAQUS was used to simulate the inter-
face between concrete infill and steel tube. The GAP element has
infinite stiffness for the compression and no stiffness in tension,
which permitted simulation of slip by separation of two nodes.
(Penetration of one node into an adjacent one was prevented.)
The normal stresses in the GAP element results in confinement
of the concrete, thus it permits explicit modeling of the confining
effect of the tube. Shear stress transfer between the steel tube and
the concrete infill is accomplished through friction, which was in-
troduced to the model through a friction coefficient assigned to the
GAP element. Thus, shear stress at the interface is generated with
pressure acting through the GAP element with a specified value
of friction coefficient. A friction coefficient of 0.47 was used based
on validation from a prior study (Moon et al. 2012). In contrast,
Concrete infill: Steel tube:
8-node solid element 4-node shell element
Interface between
Internal rebar:
steel tube & concrete infill:
2-node truss element
GAP element
(a)
Fig. 3. Finite-element model of cantilever CFT specimen
© ASCE 04013012-4
J. Struct. Eng. 2013.139.
relative movement (slip) between the internal reinforcement and
the concrete was not specifically modeled; the truss elements sim-
ulating the reinforcement were assumed to be perfectly bonded to
the concrete and were modeled using the EMBEDED option in
ABAQUS. Similar modeling approaches have been proposed by
previous researchers (Elremaily and Azizinamini 2001; Leon and
Hu 2008).
Boundary and loading conditions used are shown in Fig. 3(b).
The nodes at the base of the CFT column model were fully re-
strained to model a fixed-base column. Distributed axial loads were
applied uniformly at the top of the model to both the concrete and
steel section. A monotonic lateral displacement history was applied
at the top of the model. To ensure convergence, a mesh refinement
study was conducted. Based on this study, a total of 20 elements is
recommended to model the steel tube. This mesh was adopted for
analysis as shown in Fig. 3(a).
The concrete was modeled using the available damaged plastic-
ity model incorporated in ABAQUS (Lubliner et al. 1989; Lee and
Fenves 1998). This constitutive model simulates triaxiality depen-
dent plastic hardening. The uniaxial unconfined stress-strain rela-
tionship proposed by Saenz (1964) was employed in which it was
assumed that the compressive stress-strain relationship is linear up
to a stress of 0.5fc0 and the maximum compressive strength, f c0 , is
achieved when compressive strain is 0.003. The modulus of elas-
ticity of the concrete Ec was approximated as 4,700ðf c0 Þ0.5 MPa
(ACI 2011).
The tensile response of the concrete was modeled as follows:
(1) the tensile stress increases linearly up to maximum tensile
strength of the concrete, which is approximate as 0.09fc0 , and
(2) after this peak, the tensile stresses decrease linearly to zero
at a strain equal to 10 times the strain at maximum tensile strength
of the concrete, εct .
The dilation angle, ψ, of the concrete is an important model
parameter, which in part determines the plastic hardening of the
concrete. It was approximated as 20° based on the results of prior
research and parametric studies (Moon et al. 2012). For the steel
tube and internal reinforcement, a trilinear stress-strain relationship
was used with an isotropic-hardening plasticity rule. Young’s
modulus, Es , was approximated as 200,000 MPa and Poisson’s
Fixed end: U1, 2, & 3 of concrete
infill & steel tube are restrained.
CFT with internal rebar
Step 2: U2 is
applied on the
steel tube. U2
U3 U1
Step 1: Axial load (Direction 3) is applied
to both concrete infill & steel tube
(b)
J. Struct. Eng.
 ratio, υs , was approximated as 0.3. The plastic plateau terminated from 8 to 30 for Matsui et al. (1995). These test data were selected
when strain of the steel, εs , was set equal to 10 times of yield strain for the completeness of the available information.
of the steel (10εsy ) and stress increased up to ultimate strength of The test specimens were simulated to verify the finite-element
the steel, fu , which is achieved when the strain of the steel, εsu , is model used in this study. A half-sine wave initial imperfection was
0.1. The measured stress-strain curve from the CFT tests was used introduced to initiate global buckling. Two different imperfections
to verify the analytical models when available. were examined, including maximum amplitudes, e, of L=2,000 or
L=5,000. Figs. 4(a and b) show the comparison of simulated axial
load-transverse displacement to the test results for Specimen 18-0
Model Verification: CFT under Bending
from Matsui et al. (1995) and Specimen SC154-4 from Han
The bending behavior of CFTs was verified in a prior study (Moon and Yan (2000). The simulated results show good agreement with
et al. 2012). In this paper, the verification was extended to include test results.
slender CFTs under axial load and CFTs under axial-bending com- Analysis results with an eccentricity, e, of L=2,000 provided
bined loading. The brief summary of verification of bending behav- better simulation of tests, as shown in Table 1. The differences be-
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.

ior is as follows. tween the simulated and measured test results are likely due to sev-
Three-point bending test results conducted by Thody (2006) eral factors. First, it is almost impossible to achieve ideal pin-ended
were used to verify the finite-element model for the CFTs under support in experiments due to the friction generated between the
bending. The tests were simply supported components with a test specimen and the test setup. Second, the exact shape and mag-
5.49-m span and a single actuator located at midspan to apply nitude of initial imperfection are unknown and cannot be applied to
the cyclic lateral load. The steel tubes used for the test specimens the finite-element models. These factors indicate the finite-element
were 508 and 6.35 mm in diameter and thickness, respectively. The models used in this study provide a suitable, conservative estimate
measured compressive strength of the concrete, f c0 , and yield stress of the buckling strength of CFTs.
of the steel tube, f y , were 84.1 and 520.9 MPa, respectively.
The simulated moment-drift and slip-drift relationships agree
Model Verification: CFT under Combined Loading
well with those from the test. On average, the analysis estimated
a bending capacity approximately 6% smaller than that achieved The combined axial and bending test results of Marson and
in experiments. The computed maximum slip was slightly larger Bruneau (2004) were used to verify the finite-element analysis
than the experimentally measured slip. The local buckling shape models for CFT under general combined loading. The tests were
and cracking pattern that was observed by removing the steel tube conducted using a footing that provided near full fixity; the
after the test were also compared with simulation results, and it was CFT beam columns were connected to a pair of wide flange sec-
found that analysis results agree well with the test results. tions and embedded in the footing. As a result, the footing was not
modeled, but a fixed-base column model was employed. Table 2
provides geometry and material properties for the four specimens
Model Verification: CFT under Axial Load
simulated. The D=t ratio varied from 43.2 to 73.9 and axial load
Han and Yan (2000) and Matsui et al. (1995) conducted experi- ratio (P=Po;AISC ) ranged from 0.13 to 0.32.
ments on the buckling strength of a pin-ended CFT column. A total Fig. 5 compares the measured and simulated moment-drift re-
of 15 test specimens were analyzed. The profiles of their specimens lationship where the drift is defined as lateral displacement at the
and comparison results are shown in Table 1. The D=t ratios were top of a specimen divided by the column length. The models pro-
24 and 39.6 for the specimens of Han and Yan (2000) and Matsui vide good simulation of the measured stiffness, yield, and maxi-
et al. (1995), respectively. The ratios of the length to the diameter, mum strengths. The average difference in the predicted and
L=D, varied between 32.5 and 38.5 for Han and Yan (2000) and measured strengths was 7.3%. In the test program, cyclic loading

Table 1. Specimen Properties and Predicted Results for CFT under Axial Load
Average Average error Number of
Researcher D (mm) D=t L=D f y (MPa) fc0 (MPa) error (%) (e ¼ L=2,000) (%) (e ¼ L=5,000) specimens
Matsui et al. (1995) 165.2 39.6 12–30 358.7 40.9 2.8 9.5 4
Han and Yan (2000) 108 24 32.5–38.5 348.1 25.4–37.4 13 12.5 11

1600 18-0, Matsui et al. (1995), Test 400 SC154-4, Han & Yan (2000), Test
FEM, e=L/2,000
Applied axial load P

FEM, e=L/2,000
Applied axial load P

1200 P 300 P

800 200
L L
400 100

0 0
0 20 40 60 0 50 100 150 200
(a) Displacement at midspan (b) Displacement at midspan

Fig. 4. Verification of the finite-elemeny model for CFTs under axial load: (a) 18-0 specimen (data from Matsui et al. 1995); (b) SC154-4 specimen
(data from Han and Yan 2000)

© ASCE 04013012-5 J. Struct. Eng.

J. Struct. Eng. 2013.139.

 Table 2. Specimen Properties and Predicted Results for CFT under The CFT component was modeled as a cantilever to easily evaluate
Combined Loading combined loading. The ratio of the length to the diameter of the
fy f c0 Mu;Test = CFT L=D was 8 for the bending and reference models. This
Model D (mm) D=t L=D (MPa) (MPa) P=Po;AISC M u;FEM L=D ratio is equivalent to 16 for simply supported CFTs under cen-
CFT34 323.9 43.2 6.8 415 40 0.32 1.07
tral concentrated loads. According to a test program conducted by
CFT42 406.4 42.8 5.4 505 35 0.19 0.96 previous researchers (Elchalakani et al. 2001; Han et al. 2006;
CFT51 323.9 58.9 6.8 405 35 0.33 1.15 Thody 2006) for which L=D ratio of CFTs under bending varied
CFT64 406.4 73.9 5.4 449 37 0.13 0.98 from 6 to 15, an L=D of 16 minimized the effect of shear stress on
the bending behavior of CFTs. The diameter of the CFT was
1,524 mm (60 in.), which represents the typical diameter of the
bridge pier. Yield stress of the steel, f y , and compressive strength
was applied and test specimens lost lateral strength after tearing of of the concrete, fc0 , were 344.5 MPa (50 ksi) and 34.5 MPa (5 ksi),
steel tube followed local buckling; the tearing of a steel tube was respectively, for all models. The yield stress of internal reinforce-
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.

not modeled and therefore the impact on the postpeak response is ment, fyb , is 344.5 MPa (50 ksi).
not simulated. Thus, postpeak behavior at high drift ratio obtained The parameters were studied as follows. The D=t ratios studied
from the analysis could be different from that of the test. were 40, 60, 80, and 100. The internal reinforcement ratio, ρi , was
Fig. 6 shows the comparison of local buckling shape, stress dis- varied and included 0.0, 0.5, 1.0, 2.0, and 3.0%. Two different con-
tribution, and cracking pattern of the CFT64 specimen. The ob- crete cover depths, dc , of 50.8 mm (2 in.) and 152.4 mm (6 in.)
served location and shape of the local buckling is similar to the were used resulting in radii ratios, rbm =rm , of 0.933 and 0.8, respec-
simulated response, as shown in Fig. 6. The steel tube fully yielded tively (Fig. 1). The resulting values of the buckling parameter λ are
in tension and compression region and the maximum compression summarized in Table. 3.
was observed at the tip of the compression part. Further, at 7% drift,
cracking in both tension and compression side was observed, which
was also well simulated. Bending with No Axial Load
The flexural strength of the CFT is needed to establish P-M inter-
action curves, and 36 CFT members were analyzed as a fixed can-
Parametric Study of CFT under Combined Loading tilever column and the lateral displacement was applied at the top
of the CFT. Fig. 7 summarizes these analytical results. Fig. 7(a)
The verified models were used to study the salient parameters shows the variation of moment capacity normalized by the moment
including amount of internal reinforcement, D=t ratio, axial load capacity of the CFT without internal reinforcement as a function of
ratio, and relative radius of the reinforcement and the steel tube. ρi =ρtotal , where ρtotal is the total reinforcement ratio including steel
Finite-element analyses were performed on CFT models with tube and internal reinforcement.
and without internal reinforcement to establish P-M interaction The moment capacity increased with increasing ρi =ρtotal . The
curves, which include the effects of global and local buckling. flexural strength of the CFT increased by approximately 50% when

600 Test results, CFT34 1200 Test results, CFT42

FEM results, CFT34 FEM results, CFT42
Base Moment (kN. m)

Base Moment (kN. m)

500 1000

400 800
P P
300 H 600 H

200 400
L L
100 200

0 0
0 2 4 6 8 10 0 2 4 6 8 10
(a) Drift (%) (b) Drift (%)

500 Test results, CFT51 800 Test results, CFT64

FEM results, CFT51 FEM results, CFT64
Base Moment (kN. m)

Base Moment (kN. m)

400
600

300
P P
H 400 H
200
L 200 L
100

0 0
0 2 4 6 8 10 0 2 4 6 8 10
(c) Drift (%) (d) Drift (%)

Fig. 5. Verification of the finite-element model for CFTs under combined loading: (a) CFT34; (b) CFT42; (c) CFT 51; (d) CFT64 specimen

© ASCE 04013012-6 J. Struct. Eng.

J. Struct. Eng. 2013.139.

 * Von-Mises stress in steel tube

Yielding
* Min. Principle stress (Compression) in concrete infill
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.

* Comparison of local buckling

shape (at 7% drift)

* Cracking pattern of concrete infill

Max.
compression

Load direction

Fig. 6. Comparison of local buckling shape, stress distribution, and cracking pattern of CFT64 specimen

Table 3. Range of D=t Ratio and λ Fig. 7(d) shows the comparison of the predicted strength from
λ [based on Number current design codes with previous test results for the CFT without
Loading case D=t ratio AISC (2010)] of models internal reinforcement (Prion and Boehme 1994; Elchalakani et al.
2001; Han et al. 2006; Thody 2006). Again, AISC (2010) and
Flexure
Eurocode 4 (2004) provisions provide the conservative strength
FEM 40–100 — 36
Test data 40.2–109 — 33
estimates of CFTs under bending; the average discrepancy be-
Pure axial load tween test and predicted strength was 23.2 and 22.9% for AISC
FEM 40–100 0.70–1.9 48 and Eurocode 4 provisions, respectively. The scatter is compara-
Test data 18.4–68.7 0.34–2.17 113 tively large for the measured-to-code predicted response, as shown
Combined loading in Fig. 7(d). This is mainly attributed to the uncertainty of material
FEM 40–100 0.70–1.14 113 properties of test specimen or test procedure. In the case of the
Test data 34.5–111.9 0.31–0.65 34 analysis, the f u =f y ratio is the same for all models and scatter is
quite uniform. On the other hand, for the actual test, the fu =f y ratio
is not constant. Furthermore, it can be seen that nearly identical test
specimens in the same test program achieved considerably different
ρi =ρtotal is increased to 0.45%, but this is an 83% increase in the strength for some tests. This also supports potential anomalies in
quantity of steel. This result shows that adding internal reinforce- the test setup or procedure.
ment is a less efficient mechanism for increasing flexural resistance
than normal CFT members. To some extent this is compensated for
by the radius of the internal reinforcement. Increasing the rbm =rm Axial Load with No Lateral Loading
ratio increases the moment capacity because of the relatively larger A series of 48 models were analyzed under pure axial load. Again,
moment arm for the reinforcing steel. the predicted resistance values were compared to expressions avail-
Figs. 7(b and c) compare the moment capacity from finite- able in current design codes (AISC 2010; CEN 2004). Four differ-
element analysis to moment capacity computed by the AISC ent L=D ratios were considered as follows: 8, 12, 16, and 20. An
(2010) and Eurocode 4 (2004) provisions as a function of the initial imperfection with maximum amplitude of L=1,500, which is
ρi =ρtotal and D=t ratios, respectively. These analytical results show the value used to construct the AISC buckling curve, was used.
that current design codes provide reasonably conservative values The maximum load obtained from the analyses defines the
for moment capacity regardless of D=t ratio and ρi =ρtotal . On aver- bucking load. Fig. 8 shows the normalized axial load lateral drift
age, the AISC and Eurocode 4 provisions underestimate the resis- relationship curves for four models; the axial load was normal-
tance predicted by finite-element analysis by 9.1 and 8.5%, ized to the squash load, Po . For slender column models (for exam-
respectively. ple, L=D ¼ 16 and 20 as shown in Fig. 8), the axial strength

© ASCE 04013012-7 J. Struct. Eng.

J. Struct. Eng. 2013.139.

 Mo,FEM / Mo,FEM without internal rebar
1.75 1.75 Mo,FEM /Mo,AISC
rbm/rm=0.8
rbm/rm=0.933 Mo,FEM /Mo,EC4
1.50

Mo,FEM /Mo,Predicted
1.50

1.25 1.25

1.00 1.00

0.75 0.75
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5
/ (b) /
(a) i total
i total
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.

1.75 Mo,FEM /Mo,AISC 1.75 Mo,Test /Mo,AISC

Mo,FEM /Mo,EC4 Mo,Test /Mo,EC4
1.50 1.50
Mo,FEM /Mo,Predicted

Mo,Test /Mo,Predicted
1.25 1.25

1.00 1.00
Test results (CFT without rebar)
0.75 0.75
40 60 80 100 120 40 60 80 100 120
(c) D/t ratio (d) D/t ratio

Fig. 7. Results of parametric study of CFTs under bending: (a) variation of moment capacity with ρi =ρtotal ; (b) M o;FEM =Mo;predicted versus ρi =ρtotal ;
(c) Mo;FEM =M o;predicted versus D=t ratio; (d) M o;Test =M o;predicted versus D=t ratio (test results)

1.2 and Roeder et al. (2010)] result in different values of λ, but the
D/t=60, and i=2% impact on the predicted buckling capacity is small. The AISC
1.0 buckling curve provides good agreement with buckling strength
L/D=8 predicted by finite-element analysis for all D=t ratios and ρi for
0.8 both EIeff models (AISC 2010; Roeder et al. 2010). With the
P/Po,AISC

L/D=12 Eurocode 4 (CEN 2004) buckling curves, an imperfection factor,

0.6
0.4
0.2
0.0
0.0 0.3
Drift (%)
Fig. 8. Normalized axial load drift relationship for various L=D ratios
continuously increased after the buckling without strength degra-
dation. For these more slender columns, the buckling load was
obtained by using the Southwell plot procedure (Mandal and
Calladine 2002). Considerable prebuckling deformation was ob-
served for all models as shown in Fig. 8 and the magnitude of the
deformation increased with increasing L=D ratio. These prebuck-
ling deformations result in a secondary bending moment due to
P-Δ effect.
Fig. 9 compares the normalized buckling load Pcr;FEM =
Po;AISC or EC4 , where Pcr;FEM is the buckling strength obtained from
finite-element analysis as a function of the buckling parameter λ.
Different EIeff stiffness models as given by Eqs. (4) and (5) were
used to evaluate the AISC buckling provisions. Figs. 9(a and b)
show that the two different EIeff expressions [from AISC (2010)
α, of 0.21 provides a better estimation of buckling strengths for the
L/D=16 CFT models than buckling curve b, as shown in Fig. 9(c). Local
buckling of the steel tube did not occur in these analyses, and there
L/D=20 were no strength reductions due to local buckling. Bradford et al.
(2002) and Hu et al. (2003) noted that local buckling of circular
CFTs under axial load is unlikely to occur because the concrete
0.6 0.9 1.2 core and steel tube are in contact around the perimeter and the con-
fining pressure is applied to the concrete in all the radial directions.
The simulated and code-predicted buckling capacities are com-
pared in Fig. 10, which shows the normalized buckling load ratio
Pcr;FEM =Pcr;AISC or EC4 , where Pcr;AISC and Pcr;EC4 are the buckling
capacities from AISC (2010) and Eurocode 4 (CEN 2004), as a
function of the buckling parameter λ. The comparison suggests that
the design equations provide a conservative estimate of the buck-
ling capacity, and the conservatism is generally larger with increas-
ing λ for all design methods. The AISC buckling curve with EIeff
from the AISC provision shows the best estimation of the buck-
ling strength. The average discrepancies were 9.5, 13, and 16.9%
for AISC provisions when computed with EIeff expressions from
AISC (2010), Roeder et al. (2010), and Eurocode 4 (CEN 2004)
provisions, respectively.
Fig. 11 compares the predicted buckling strength with previous
CFT test results without internal reinforcement. A total of 98 mea-
sured buckling data points were used; all data are available on the
Goode CFT database (Goode and Lam 2008). The test specimens
were selected to be of reasonable scale (diameter greater than

© ASCE 04013012-8 J. Struct. Eng.

J. Struct. Eng. 2013.139.

 FEM, D/t=60, i=0%, FEM, D/t=100, i=0%
FEM, D/t=60 i=2%, rbm/rm=0.8, FEM, D/t=100 i=2%, rbm/rm=0.8

Eurocode 4 buckling curve a

1.0 1.0 1.0

0.8 0.8 0.8

Pcr,FEM /Po,AISC
Pcr,FEM /Po,AISC

Pcr,FEM /Po,EC4
0.6 0.6 0.6
Eurocode 4
0.4 0.4 0.4 buckling curve b
AISC buckling
AISC buckling curve
curve with EIeff
0.2 0.2 0.2
from Roeder et al. (2010)
0.0 0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5
(a) (b) (c) Buckling parameter
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.

Buckling parameter Buckling parameter

Fig. 9. Variation of buckling load with λ: (a) Pcr =Po versus λ (AISC buckling curve); (b) Pcr =Po versus λ [AISC buckling curve with EIeff from
Roeder et al. (2010)]; (c) Pcr =Po versus λ (buckling curve from Eurocode 4)

1.50 1.50 1.50

1.25 1.25 1.25
Pcr,FEM /Pcr,AISC
Pcr,FEM /Pcr,AISC

Pcr,FEM /Pcr,EC4
1.00 1.00 1.00
0.75 0.75 0.75
0.50 0.50 AISC buckling curve with 0.50
AISC buckling curve Buckling curve a from
0.25 with EIeff from AISC 0.25 EIeff from Roeder et al. (2010) 0.25 Eurocode 4
0.00 0.00 0.00
0.5 1.0 1.5 2.0 0.5 1.0 1.5 2.0 0.5 1.0 1.5 2.0
(a) Buckling parameter (b) Buckling parameter (c) Buckling parameter

Fig. 10. Comparison of predicted buckling load: (a) comparison of analysis results with AISC buckling curve; (b) comparison of analysis results with
AISC buckling curve [EIeff from Roeder et al. (2010)]; (c) comparison of analysis results with buckling curve a from Eurocode 4 (CEN 2004)

2.5 diameters with small D=t ratios. Thus, their buckling behavior is
likely controlled by the steel section, and it is likely that the end
loading of the columns was imperfect, resulting in a coupled
2.0
global-local instability mode. Furthermore, nearly identical test
Pcr,Test /Pcr,Predicted

specimens in the same test program achieved significantly larger

1.5 (an average of 20%) compressive resistance. This suggests poten-
tial anomalies in the test setup or procedure for these three tests.
1.0

CFTs under Combined Loading

AISC buckling curve with EIeff from AISC
0.5
AISC buckling curve with EIeff from Roeder et al. (2010)
Buckling curve a from Eurocode 4
0.0
0.0 0.5 1.0 1.5
Buckling parameter
Fig. 11. Comparison of test results with predicted buckling load
100 mm) and a resulting buckling parameter λ larger than 0.4.
Fig. 11 shows that the buckling strength predicted by current design
provisions is generally conservative compared with the test results
and the predictions are more conservative with increasing λ. The
average discrepancies were 23.5, 24.2, and 26.7% for AISC pro-
visions, AISC buckling curve with EIeff from Roeder et al. (2010),
and Eurocode 4 (CEN 2004) provisions, respectively. A limited
number of tests fall below the design provisions for columns with
λ values in the range between 0.5 and 1.0 (three test data points
out of a total of 98 points were below 95% of predicted buckling
load). These three test specimens in question have relatively small
A series of parametric analyses (a total of 113 models) were per-
formed to evaluate the strength of the CFT under general combined
loading. The analytical results were compared with the P-M inter-
2.0 2.5
action curve from AISC (2010) and Eurocode 4 (CEN 2004).
Fig. 12 shows a P-M interaction curve derived from the plastic
stress distribution method and the analytical prediction of the
strength of CFT with a D=t of 60, L=D of 8, and ρi ¼ 0%. The
triangles denote the moment capacity when P-Δ moments are
included and the circles indicate the nominal moment capacity ex-
cluding P-Δ moments. When P-Δ moments are considered, the
P-M interaction curve from the plastic stress distribution method
is the same as the curve that includes instability up to the buckling
load (Pcr;AISC =Po;AISC ¼ 0.8 for the model in Fig. 12). Conversely,
when the axial load exceeds buckling capacity, the plastic stress
distribution method overestimates the moment capacity of the
CFT. In this study, P-Δ effects were included in the moment
calculations for all comparisons.
Fig. 13 compares P-M interaction curves based on AISC (2010)
and Eurocode 4 (CEN 2004) provisions with finite-element analy-
sis results for a CFT column having a D=t ratio of 60 and rbm =rm

© ASCE 04013012-9 J. Struct. Eng.

J. Struct. Eng. 2013.139.

 PSDM
1.2 desirable, it has negative consequences with seismic design. Seis-
FEM results without P- effect
FEM results with P- effect mic design requires consideration of the expected capacity of
1.0 ductile elements because seismic performance does not depend
on the capacity of a single structural element. For example, if
0.8 Pcr,AISC=0.8Po,AISC
the conservatism of strength of a CFT column is large, the connec-
P/Po,AISC

0.6 tions and adjacent member such as footing or cap beam should
be conservatively designed to avoid premature failure of these
0.4 components in an extreme loading case. Thus, it is necessary to
accurately predict the strength of the CFT column. Furthermore,
0.2 most practical columns have relatively small axial load ratios
D/t=60, L/D=8, i=0%
0.0
and the AISC and Eurocode 4 interaction curves considerably
0.0 0.5 1.0 1.5 underestimate the moment capacity of CFTs under these condi-
M/Mo,AISC tions. This results in increased uncertainty as to the expected mo-
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.

ment capacity of the member and requires greater conservatism in

Fig. 12. P-Δ effect on the strength of CFTs under combined loading
of 0.8. Both reinforced and unreinforced CFT sections were con-
sidered. The figure shows that the plastic-stress distribution method
provides accurate and conservative predictions of the strength
of the CFT when the axial load is low, but as noted previously,
this method significantly overestimates the strength with high
axial load and slenderness ratio, because global buckling is not
considered in the P-M interaction curves from the plastic-stress
distribution method.
If the axial load is limited to the buckling resistance, the plastic-
stress distribution method may slightly overestimate the moment
capacity as the compressive load approaches the buckling, but
the overestimate is modest. The interaction curve suggested by
AISC (2010) and Eurocode 4 (CEN 2004) interaction curves were
similar to each other and give quite conservative estimation of the
strength of the CFT for all analyzed cases. While conservatism is
the seismic design. Taken as a whole, these results suggest a need to
improve the P-M interaction curve to provide a more economical
and rational design.
Alternative P-M Interaction Curve for Circular CFTs
An alternative P-M interaction curve is proposed to improve the
accuracy of the predicted resistance of CFT members under com-
bined load. The proposed curve is shown in Fig. 14 and is specially
designed to take the benefits from both the plastic-stress distribu-
tion method and current design codes. The curve is defined as
follows:
• The P-M interaction curve from the PSDM forms the basis of
the curve.
• Points A and B are the pure axial and pure flexural capacity cor-
responding to the PSDM P-M interaction curve (no considera-
tion of global buckling effects).

1.2 D/t=60, L/D=8, i=0% 1.2 D/t=60, L/D=12, i=0%

PSDM
PSDM 1.0
1.0 AISC
AISC
Eurocode 4
Eurocode 4 0.8
0.8 Proposed
Proposed
FEM
P/Po,AISC

P/Po,AISC

FEM
0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5
(a) M/Mo,AISC (b) M/Mo,AISC

1.2 D/t=60, L/D=8, =2%, r /r =0.8 1.2 D/t=60, L/D=12, =2%, r /r =0.8
i bm m i bm m

1.0 PSDM
1.0 PSDM
AISC AISC
0.8 0.8 eurocode
P/Po,AISC

Eurocode 4
P/Po,AISC

Proposed proposed
0.6 FEM 0.6 FEM

0.4 0.4

0.2 0.2
0.0 0.0
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5
(c) M/Mo,AISC (d) M/Mo,AISC

Fig. 13. Comparison of P-M interaction curves with analysis results: (a) D=t ¼ 60, L=D ¼ 8, ρi ¼ 0%; (b) D=t ¼ 60, L=D ¼ 12, ρi ¼ 0%;
(c) D=t ¼ 60, L=D ¼ 8, ρi ¼ 2%; (d) D=t ¼ 60, L=D ¼ 12, ρi ¼ 2%

© ASCE 04013012-10 J. Struct. Eng.

J. Struct. Eng. 2013.139.

 1.2 D/t=60, L/D=8, i=0% 3.0
A
1.0 2.5
A' A'' Interaction curve from PSDM
0.8

Mu,Test /Mu,Predicted
P/Po,AISC 2.0
C
0.6
1.5
0.4 Proposed interaction curve
C'
including global buckling effect 1.0
D AISC (AVG.=1.59, STD.=0.39)
0.2
0.5 Eurocode 4 (AVG.=1.58, STD.=0.39)
B Proposed curve (AVG.=1.29, STD.=0.17)
0.0
0.0 0.5 1.0 1.5 0.0
M/Mo,AISC 0.0 0.1 0.2 0.3 0.4 0.5
P/Pcr,AISC
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 14. Schematic view of alternative P-M interaction curve

Fig. 16. Comparison of test results with predicted strength

• Point C lies on the PSDM P-M interaction curve with the same
moment capacity as point B. The different between points B and Fig. 16 compares the predications using the proposed curve with
C is that point B does not have an axial load and point C does. previous test results (Elremaily and Azizinamini 2001; Bishop
• Points A 0 and C 0 are obtained by from points A and C by multi- 2009; Zhang et al. 2009; Park et al. 1982). A total of 34 test results
plying the axial load associated with points A and C by the ratio were compared. The D=t ratio of test specimens varied from 34.5 to
Pcr =Po;AISC . (This is similar to the AISC approach.) These 111.9, and axial load ratio P=Pcr had a range from 0.06 to 0.44.
points do not fall on the PSDM P-M interaction curve. [Only the tests of Park et al. (1982) had internal reinforcement.]
• Point D is located on the PSDM interaction curve. The point The proposed design curve provides the conservative predictions
is determined as PD ¼ 0.5PC 0 , where PD is the axial load at of all of the specimens. On average, the results provided moment
point D. capacities that were 29, 59, and 58% larger than the moment
• Finally, the maximum axial load, PA 0 , is used to determine the capacity predicted by the proposed model and AISC (2010) and
upper limit of the curve. The intersection of the PSDM P-M Eurocode 4 (CEN 2004) provisions, respectively. This increased
interaction curve and the horizontal line at PA 0 is used. capacity is expected; work by Roeder et al. (2010) shows that
• The alternative P-M interaction curve can be constructed by the PSDM method results in larger capacities than predicted,
connecting points A 0 , A 0 0 , D, and B. 24% on average. Taken as a whole, this thorough comparison in-
The proposed curve is also shown in Fig. 13. The proposed dicates that the proposal is accurate, conservative, and appropriate
curve compares well with the simulated results. Fig. 15 compares for design.
the ratio of moment predicted by the finite-element analysis (total
of 113 models) to the moment capacity predicted using AISC
(2010), Eurocode 4 provisions (CEN 2004), and proposed P-M in- Summary and Conclusions
teraction curve. The ratio is plotted as a function of the applied axial
load. The proposed interaction curve provides more accurate esti- In this study, the strength and behavior of CFTs with and without
mate and less scatter than the provisions, as indicated by the mean internal reinforcement under general combined loading was stud-
and standard deviation values presented in the figure. The scatter ied. The research approach was as follows. A finite-element model
increases as the axial load P approaches the critical load Pcr . Larger capable of simulating generalized conditions, including the impact
values of P beyond balance results in smaller moment capacities, of confinement, slip, and axial load, was developed. The model
therefore inaccuracies may be amplified. The proposed curve pro- was verified using previous tests for CFTs under various loading
vides more accurate flexural strength predictions at all levels of ax- conditions and the accuracy of the finite-element models was veri-
ial loads. Fig. 15 also shows a magnified portion of the curve at fied. Parametric studies were conducted using the verified model.
more common levels of axial load ratios; the proposed curve is both The study parameters included amount of internal reinforcement,
accurate and conservative. ratio of the tube to the internal reinforcement, axial load ratio,

AISC (AVG.=1.76 , STD.=1.00) 1.75 AISC (AVG.=1.18 , STD.=0.07)

6
EC4 (AVG.=2.10 , STD.=2.07) Eurocode 4 (AVG.=1.20 , STD.=0.09)
Proposed curve Proposed curve
5 (AVG.=1.07 , STD.=0.05)
Mu,FEM/Mu,Predicted

1.50
Mu,FEM/Mu,Predicted

(AVG.=1.06 , STD.=0.06)
4

3 1.25

2
1.00
1

0 0.75
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6
P/Pcr,AISC P/Pcr,AISC

Fig. 15. Comparison of analysis results with predicted strength

© ASCE 04013012-11 J. Struct. Eng.

J. Struct. Eng. 2013.139.

 slenderness ratio, and D=t ratio. These parameters were evaluated
for all combined loading conditions.
The results were used to both evaluate existing P-M interaction
curve design models and a proposed model. The AISC (2010) and
Eurocode 4 (CEN 2004) provisions provide P-M interaction curves
including consideration of buckling effects, but these curves con-
siderably underestimate the bending resistance of CFTs for the low
axial loads. This is the most practical range for column design, and
an alternate P-M interaction curve was proposed.
Under pure bending, the analysis showed that the AISC (2010)
and Eurocode 4 (CEN 2004) provisions provide reasonable esti-
mates of flexural capacity of the CFTs for all values of D=t and
internal reinforcing bar ratio. Under axial compression, the AISC
Downloaded from ascelibrary.org by East Carolina University on 06/15/14. Copyright ASCE. For personal use only; all rights reserved.
buckling curve agrees well with the results of simulated and test
results. Under general combined loading, the P-M interaction curve
from the plastic-stress distribution method considerably overesti-
mates the strength when the slenderness and axial load ratios
are high. This is expected because the plastic-stress distribution
method neglects the effect of global buckling.
The alternative P-M interaction curve reduced the significant
conservatism noted in current AISC (2010) and Eurocode 4
(CEN 2004) methods and provided reduced scatter. The proposed
curve employs aspects of both the plastic-stress distribution method
and the current design codes. The proposed curve was successfully
verified by comparing with results of finite-element analysis and
those of previous tests, and it was shown to provide more accurate
predictions of combined resistance than current design methods.
Acknowledgments
This paper was completed partially with funding provided by the
Washington State Department of Transportation (WSDOT) through
the project entitled “Design of Bridge Foundations with Steel Cas-
ings.” Mr. Bijan Khaleghi is the WSDOT Bridge Design Engineer
and the coordinator of this project. The advice and financial support
of the WSDOT is gratefully acknowledged.
References
AASHTO. (2009). AASHTO LRFD bridge design specification, 4th Ed.,
Washington, DC.
ABAQUS [Computer software]. Dassault Systemes Simulia Corp.,
Providence, RI.
American Concrete Institute (ACI). (2008). Building code requirements
for structural concrete and commentary, Farmington Hills, MI.
American Institute of Steel Construction (AISC). (2010). Specifications
for structural steel buildings, Chicago.
Bishop, E. S. (2009). “Evaluation of the flexural resistance and stiffness
models for circular concrete-filled steel tube members subjected to com-
bined axial-flexural loading.” M.S. thesis, Univ. of Washington, Seattle.
Bradford, M. A., Loh, H. Y., and Uy, B. (2002). “Slenderness limits for
filled circular steel tubes.” J. Constr. Steel Res., 58(2), 243–252.
Bruneau, M., and Marson, J. (2004). “Seismic design of concrete-filled
circular steel bridge piers.” J. Bridge Eng., 9(1), 24–34.
Choi, Y.-H., Foutch, D. A., and LaFave, J. M. (2006). “New approach
to AISC P-M interaction curve for square concrete filled tube (CFT)
beam-columns.” Eng. Struct., 28(11), 1586–1598.
Elchalakani, M., Zhao, X. L., and Grzebieta, R. H. (2001). “Concrete-filled
circular steel tubes subjected to pure bending.” J. Constr. Steel Res.,
57(11), 1141–1168.
© ASCE 04013012-12
J. Struct. Eng. 2013.139.
Elremaily, A., and Azizinamini, A. (2001). “Design provisions for connec-
tions between steel beams and concrete filled tube columns.” J. Constr.
Steel Res., 57(9), 971–995.
European Committee for Standardization (CEN). (2004). “Design of
composite steel and concrete structures, Part 1.1 General rules and rules
for buildings.” EN 1994-1-1, Eurocode 4, Brussels, Belgium.
Goode, C. D., and Lam, D. (2008). “Concrete—filled tube columns—tests
compared with Eurocode 4.” Proc., Engineering Foundation Conf. on
Composite Construction in Steel and Concrete, ASCE, Reston, VA,
317–325.
Han, L. H., Lu, H., Yao, G. H., and Liao, F. Y. (2006). “Further study on the
flexural behaviour of concrete-filled steel tubes.” J. Constr. Steel Res.,
62(6), 554–565.
Han, L.-H., and Yan, S.-Z. (2000). “Experimental studies on the strength
with high slenderness ratio concrete filled steel tubular columns.” Proc.,
6th ASCCS Conf., ASCCS, West Yorkshire, UK, 419–425.
Hu, H. T., Huang, C. S., Wu, M. H., and Wu, Y. M. (2003). “Nonlinear
analysis of axial loaded concrete-filled tube columns with confinement
effect.” J. Struct. Eng., 129(10), 1322–1329.
Lee, J., and Fenves, G. L. (1998). “Plastic-damage model for cyclic loading
of concrete structures.” J. Eng. Mech., 124(8), 892–900.
Leon, R. T., and Hu, J. W. (2008). “Design of innovative SMA PR con-
nections between steel beams and composite columns.” Proc., 6th
Int. Workshop on Connections in Steel Structures, Connections VI,
AISC, Chicago, IL, 513–524.
Leon, R. T., Kim, D. K., and Hajjar, J. F. (2007). “Limit state response of
composite columns and beam-columns part 1: Formulation of design
provisions for the 2005 AISC specification.” Eng. J., 44(4), 341–358.
Lubliner, J., Oliver, J., Oller, S., and Oñate, E. (1989). “A plastic-damage
model for concrete.” Int. J. Solids Struct., 25(3), 299–326.
Mandal, P., and Calladine, C. R. (2002). “Lateral-torsional buckling of
beams and the Southwell plot.” Int. J. Mech. Sci., 44(12), 2557–2571.
Marson, J., and Bruneau, M. (2004). “Cyclic testing of concrete-filled cir-
cular steel bridge piers having encased fixed-base detail.” J. Bridge
Eng., 9(1), 14–23.
Matsui, C., Tsuda, K., and Ishibashi, Y. (1995). “Slender concrete filled
steel tubular columns under combined compression and bending.”
Proc., 4th Pacific Structural Steel Conf., Steel-Concrete Composite
Structures, Vol. 3, Pergamon, Oxford, UK, 29–36.
Moon, J., Lehman, D. E., Roeder, C. W., and Lee, H.-E. (2012). “Analytical
modeling of bending of circular concrete-filled steel tubes.” Eng.
Struct., 42, 349–361.
Park, R. J. T., Priestley, M. J. N., and Walpole, W. R. (1982). “The seismic
performance of steel encased reinforced concrete bridge piles.”
Research Rep. 82-12, Civil Engineering Dept., Univ. of Canterbury,
Christchurch, New Zealand.
Prion, H. G. L., and Boehme, J. (1994). “Beam-column behaviour of
steel tubes filled with high strength concrete.” Can. J. Civ. Eng.,
21(2), 207–218.
Roeder, C. W., Lehman, D. E., and Bishop, E. (2010). “Strength and
stiffness of circular concrete filled tubes.” J. Struct. Eng., 136(12),
1545–1553.
Roik, K., and Bergmann, R. (1992). “Composite column.” Constructional
steel design: An international guide, P. J. Dowling, J. E. Harding, and
R. Bjorhovde, eds., Elsevier Applied Science, London.
Saenz, L. P. (1964). “Discussion of ‘Equation for the stress-strain curve
of concrete’ by P. Desayi, and S. Krishnan.” J. Am. Concr. Inst., 61,
1229–1235.
Thody, R. (2006). “Experimental investigation of the flexural properties
of high-strength concrete-filled steel tubes.” M.S. thesis, Univ. of
Washington, Seattle.
Varma, A. H., Ricles, J. M., Sause, R., and Lu, L. W. (2002). “Experimental
behavior of high strength square concrete-filled steel tube beam-
columns.” J. Struct. Eng., 128(3), 309–318.
Zhang, G. W., Xiao, Y., and Kunnath, S. (2009). “Low-cycle fatigue dam-
age of circular concrete-filled tube columns.” ACI Struct. J., 106(2),
151–159.
J. Struct. Eng.