Nonlinear Analysis of In-Filled Steel-Concrete Composite

Columns
B. Lakshmi1 and N. E. Shanmugam2

Abstract: A semianalytical method to predict the behavior of in-filled columns is presented in this paper. Moment-curvature-thrust
Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

relationships are generated for column cross sections by an iterative process. Nonlinear equilibrium equations resulting from geometric
and material nonlinearities are solved by an incremental-iterative numerical scheme based on the generalized displacement control
method. Square, rectangular, and circular cross sections of compact steel tubes filled with concrete are considered in the analysis. The
columns are pin-ended and subjected to uniaxial or biaxial loading. The accuracy of the proposed analytical method is established by
comparing the results with the corresponding experimental values.
DOI: 10.1061/共ASCE兲0733-9445共2002兲128:7共922兲
CE Database keywords: Nonlinear analysis; Composite columns; Steel; Concrete.

Introduction 1989; Shakir-Khalil and Mouli 1990; Rangan and Joyce 1992兲
provide extensive data which strengthen the knowledge on the
Researchers have suggested analytical methods and design proce- behavior of real columns. These experimental results are useful to
dures for composite columns and design codes have been formu- understand the elastic and ultimate load behavior and to develop
lated. Large discrepancies between various design codes exist in an appropriate analytical model. The behavior of thin-walled cir-
terms of geometric and strength parameters, even when the same cular steel tubes filled with high strength concrete for use in tall
design philosophy is adopted. The disagreement between the re- buildings formed the basis of an experimental investigation by Uy
sults indicates that more accurate design guidelines are required. and Das 共1997兲. An empirical reduction factor that accounts for
Further, with the development of high strength concrete and struc- the effect of in-filled concrete prism size and the concrete strength
tural steel, there is a need to develop a good understanding of the class was introduced to evaluate the compressive strength of con-
fundamental behavior of slender composite columns under crete 共Bradford 1996兲. Research has been directed to investigate
uniaxial and biaxial bending. the use of high strength concrete 共Kilpatrick 1996; Uy and Patil
A number of theoretical and experimental studies have been 1996兲. Finite element modeling using the software ABAQUS was
reported in the recent past. Neogi et al. 共1969兲 conducted a short- developed to examine the ultimate load-carrying capacity and
term test on 18 eccentrically loaded circular concrete-filled steel load-deflection relationship up to and beyond failure 共Shakir-
tubular columns bent into single curvature with lengths varying Khalil and Al-Rawdan 1996兲.
from 1.4 to 3.3 m. A simple design procedure suitable for manual Mirza and Skrabek 共1992兲 examined the effect of concrete and
calculation has been proposed recently 共Wang and Moore 1997兲. steel strengths, the cross-sectional dimensions, and the presence
In this method the properties of the bare steel section are replaced of reinforcing bars on strength of encased composite columns.
with those of the composite section, which is in accordance with Effects of stiffeners in preventing or minimizing local buckling in
the recommendations given in BS 5950 for a bare steel column. stiffened and unstiffened infilled columns were investigated 共Ge
Design formulas for composite columns, based on standards and Usami 1992, 1994兲. Studies 共Kitada 1998兲 have shown that
given by ISO for steel structures, with appropriate modifications ductility of the composite beam-columns under large axial com-
have been proposed by Kato 共1996兲. Experiments in which full- pression is small for a rectangular cross section compared to a
scale in-filled columns were tested to failure in the late 1980s and circular cross section.
early 1990s 共Prion and Boehme 1989; Shakir-Khalil and Zeghiche This paper is concerned with a semianalytical method to pre-
dict the elastic and ultimate load behavior of square, rectangular,
1
Graduate Student, Structural Steel Research Group 共SSRG兲, Dept. of and circular steel-concrete composite columns. The nonlinear
Civil Engineering, National Univ. of Singapore, 10 Kent Ridge Crescent, analysis technique by Yang and Kuo 共1994兲 is used in the method
Singapore 119260. and the incremental equilibrium equation of a slender composite
2
Professor, Dept. of Civil Engineering, National Univ. of Singapore, column is formed based on the updated Lagrangian formulation.
10 Kent Ridge Crescent, Singapore 119260. The generalized displacement control 共GDC兲 method is then ap-
Note. Associate Editor: Enrico Spacone. Discussion open until De- plied to solve the incremental equation. The strength analysis is
cember 1, 2002. Separate discussions must be submitted for individual
carried out for concentric and eccentric loads with initial imper-
papers. To extend the closing date by one month, a written request must
be filed with the ASCE Managing Editor. The manuscript for this paper fection but the effects of local buckling are neglected. The col-
was submitted for review and possible publication on April 9, 2001; umn is pin-ended and subjected to the combined action of axial
approved on October 25, 2001. This paper is part of the Journal of compression and biaxial bending. The accuracy of the proposed
Structural Engineering, Vol. 128, No. 7, July 1, 2002. ©ASCE, ISSN method is assessed by comparing the analytical values with the
0733-9445/2002/7-922–933/$8.00⫹$.50 per page. corresponding experimental results.

922 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng. 2002.128:922-933.

 Cross-Section Analysis
The term ‘‘Short column’’ refers to a compression member that
can attain its ultimate capacity, known as squash load, without
overall buckling. The column must be straight and subjected to
axial load. In the early stages of loading, the Poisson’s ratio for
concrete is lower than that for steel and the steel tube has no
restraining effect on the concrete core. As the longitudinal strain
increases, Poisson’s ratio of concrete, which is 0.15–0.2 in elastic
range, can remarkably increase in the equitriaxial compressive
regime. Therefore the lateral expansion of uncontained concrete
gradually becomes greater than that of steel. A radial pressure
Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

develops at the steel–concrete interface thereby restraining the

concrete core and setting up a hoop tension in the tube. At this
Fig. 1. Stress-strain relation for concrete stage, the concrete core is stressed triaxially and the steel tube
biaxially.
In an actual building framework, the columns are frequently
subjected to bending moments acting in two perpendicular direc-
tions in addition to axial load due to action in space of the entire
Theoretical Analysis framing system. Biaxial bending can be significant in some appli-
cations such as a corner column in a multistorey structure and in
Assumptions composite columns functioning as bridge piers. There are only a
few studies on composite columns subjected to biaxial bending
The following assumptions have been made in the analysis:
and detailed investigation of such loading effect is imperative.
1. Constitutive relationships for concrete and steel are known;
For sections under biaxial bending, the plane of the neutral
2. Adequate bond exists between steel and concrete elements;
axis is not necessarily perpendicular to the line joining the point
3. Cross sections remain plane during various stages of loading
of application of the external load to the centroid of the section.
resulting in linear distribution of strain;
For a slender column, the lateral deflections vary along the length
4. Effects of strain hardening of steel, shear strain, tensile
strength of concrete, and local buckling of component plates of the column. The total moment, i.e., the summation of primary
are neglected; and secondary moment at each section, varies along the length. If
5. The influence of creep and shrinkage is neglected; one of the principal axes is weaker, the tendency of the column is
6. Maximum amplitude for initial lack of straightness is 0.001 to deflect in a direction perpendicular to the weaker axis. This
times the column length; makes the neutral axis change its position and undergo an in-
7. No residual stresses are considered; and crease in ␣, the angle of inclination in the same direction. Due to
8. Failure occurs when the extreme fiber compressive strain in this phenomenon, the lateral deflection of a section under biaxial
concrete reaches a limiting value of 0.003. bending is accompanied by translation and rotation of the neutral
The ultimate load is related to design strength of concrete and axis in addition to change in bending moment and curvature.
steel by the relation Therefore the inclination of the neutral axis can also be expected
to vary along the length, thus yielding to a nonplanar deflection in
N pl.rd ⫽A a f yd ⫹0.85A c f cd (1) three dimensions.
in which A a and A c ⫽cross-sectional area of structural steel and If axial force P and bending moment M alone are considered,
concrete, respectively, in the axial direction and f yd and f cd it is a simple matter to construct an interaction curve relating P
⫽corresponding design strengths. The idealized stress-strain and M under the condition that the entire section will be fully
model for both concrete 共Hognestad parabola兲 and steel used in plastic. Such an approach is not valid for compression members
the current study is shown in Figs. 1 and 2, respectively. where the change in geometry affects equilibrium if normal force
P and bending moments M x and M y acting in two perpendicular
directions are considered. The basic concepts for integration of
stresses consistent with strains at the limit state of strength can be
applied to any cross section with any orientation of the neutral
axis. For a column subjected to biaxial loading as shown in Fig. 3,
moments and curvatures about principal axes are calculated as
follows:
The moment about x axis

M x⫽ 兺 共 A c ␴ c y c ⫹A s ␴ s y s 兲 (2)

where A c and A s ⫽elemental area of concrete and steel strips,

respectively; ␴ c and ␴ s ⫽their corresponding stresses, and y c and
y s ⫽distance between the centroid of elemental area and plastic
centroid of the whole cross section along the y direction. Simi-
larly the moment about y axis

Fig. 2. Stress-strain relation for steel M y⫽ 兺 共 A c ␴ c x c ⫹A s ␴ s x s 兲 (3)

JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002 / 923

J. Struct. Eng. 2002.128:922-933.

Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Typical strain and stress distribution

Fig. 4. M-␾-P relation for major axis bending

in which x c and x s ⫽distance between the centroid of the elemen-
tal area and the plastic centroid of the whole cross section along
the x direction. The curvature about the x axis The resultant stress distribution at each element is obtained by
using the material properties of the corresponding element. Given
K x ⫽␧ cc /Y and (4) the axial load and corner strain ␧ cc , M -␸ values in each principal
the curvature about y axis direction can be determined by iteration of the neutral axis posi-
tion to satisfy force equilibrium. Repeating the procedure for suc-
K y ⫽␧ cc /X (5) cessive increments of extreme fiber strain ␧ cc until it reaches the
where X and Y ⫽intercepts of neutral axis on the directions par- crushing strain of concrete ␧ u , a set of M -␸ relations is obtained.
allel to breadth and depth, respectively, of the cross section and In the present model, ␧ cc is assumed to vary from 0 to 0.003 until
␧ cc ⫽extreme concrete fiber strain 共Fig. 3兲. The two moments M x the moment value reaches a maximum. During this procedure
and M y are related to the resultant moment M as 共i.e., for each increment of strain兲, the angle ␣ of the neutral axis
should be kept constant. The M -␸ relations are computed if
M 2 ⫽M 2x ⫹M 2y (6) agreement with the applied load is within a predefined tolerance.
Similarly for curvature along both axes, resultant curvature is If not, the extreme strain is altered and the whole procedure is
given by repeated. Thus the complete moment-curvature relationship is
constructed.
K 2 ⫽K 2x ⫹K 2y using (7) Figs. 4 and 5 show sets of moment-curvature relationships for
various axial force ratios for a steel section 120⫻80⫻5 RHS
K x ⫽K sin ␣ and (8)
with yield strength 350 N/mm2 filled with concrete of strength 45
K y ⫽K cos ␣ (9) N/mm2 subjected to uniaxial bending about major and minor axis
共Lakshmi and Shanmugam 2000a兲. In all cases there exists both
where ␣⫽inclination of neutral axis with x axis.
an ascending and a descending part. It is obvious from the curves
that moment capacity decreases with increasing axial-force ratio
Moment-Curvature-Thrust „ M -␸- P … Relationship 共defined as axial force to the squash load N u 兲 for both major and
minor axis bending. Further, when the axial force is large, the rate
A finite approach is used to compute the moment-curvature-thrust
of reduction in flexural stiffness is large when the ultimate load
(M -␸- P) relationship numerically in which the concrete and
condition is reached. However, this is not true in the elastic range.
steel areas in the cross section are broken into small discrete
areas. Summation of the acting elemental forces thus allows the
axial force to be determined, while summation of moments of the
elemental forces is used to determine the internal moments. Ef-
fects such as strain hardening of steel and finite nonzero stresses
in the concrete below the neutral axis position are usually consid-
ered secondary and are therefore neglected. A typical section of
the composite column is shown in Fig. 3. The coordinate system
is chosen to pass through the centroid of the section. The dis-
placements of the column joints are assumed to be small, so that
the total curvature in the two major bending planes labeled as K x
and K y can be represented in the form of second order derivatives
of the displacements as follows:
K x ⫽⫺⳵ 2 u/⳵x 2 (10)
K y ⫽⫺⳵ v /⳵y
2 2
(11)
The strain at any position can be expressed as
Fig. 5. M-␾-P relation for minor axis bending
␧⫽␧ o ⫹K x y⫹K y x (12)

924 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng. 2002.128:922-933.

 3. By transforming the element stiffness equations from local
coordinates to a common global coordinate system for each
element, all the element equations are assembled to yield the
structure stiffness equations. This ensures the satisfaction of
compatibility and equilibrium conditions for the entire col-
umn at each nodal point.
4. By imposing appropriate geometric boundary conditions the
structure will achieve its kinematic stability, which is also
indicated by the positive definiteness of the stiffness matrix.
5. For a given set of applied loading, the nodal displacements
can then be solved from the structure stiffness equations.
The standard stiffness matrix for the space frame element having
Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

nodal forces and moments as given in Fig. 7 was used in the

analysis. Flexural and axial rigidity referred to as EI and EA in
the stiffness matrices are derived using the slope of the moment-
curvature and axial load-shortening curves, respectively. The
slope for each element is obtained by the method of bisections
from the actual curves generated by the cross-sectional analysis.
For a typical load-deflection curve, the phenomena such as soft-
Fig. 6. Segmentation of a long column ening, stiffening, loading, and unloading are typified by the oc-
currence of limit points and snap-back points.
Formulation of incremental stiffness equations for each ele-
Column Strength Analysis ment in the structure can be expressed in a general form as
The study of complete column response at all stages of loading up
关 K 兴兵U其⫽兵2P其⫺兵1P其 (13)
to ultimate load requires proper consideration of both material
performance and geometric nonlinearity because of large defor- 关 K 兴 ⫽structure stiffness matrix, 兵 U 其 ⫽structure displacement vec-
mations. A linear elastic analysis differs from nonlinear analysis tor, 兵 2 P 其 ⫽total loads to be applied on the structure, and 兵 1 P 其
in that the equations of equilibrium for the structure under con- ⫽external loads applied to the structure at the beginning of the
sideration are established at the initial undeformed configuration. incremental step.
Slenderness for columns can be expressed analytically as a mea-
sure of the member flexural stiffness EI/L. Direct application of
Generalized Displacement Control Method „GDC…
the modulus of elasticity and the cross-section moment of inertia
The generalized displacement control 共GDC兲 method originally
which is customary for the design of bare structural steel columns
proposed by Yang and Shieh 共1990兲 is used to solve the incre-
cannot be used for composite and reinforced concrete members.
mental equilibrium equation 共13兲 considering its numerical stabil-
Also, the conventional definition of radius of gyration cannot be
ity near all types of limit points. With the superscript ‘‘i’’ denoting
applied to composite cross sections.
the current load increment step and the subscript ‘‘j’’ denoting the
current iteration number, Eq. 共13兲 can be rewritten as
Procedure
Behavior of composite columns is highly nonlinear since the 关 K ij⫺1 兴 兵 ␦U ij 其 ⫽ 兵 P ij 其 ⫺ 兵 F ij⫺1 其 (14)
M -␸- P relationships vary from section to section along the
length of the member. It is therefore essential to adopt a numeri- where 关 K ij⫺1 兴 ⫽tangent stiffness matrix formed at the beginning
cal technique in order to simplify the analysis. The procedure/ of the jth iteration based on the known element details at the ( j
algorithm used for the analysis can be summarized as follows: ⫺1)th iteration, 兵 ␦U ij 其 ⫽iterative displacement vector obtained
1. The entire column is broken down at the nodal points or for the jth iteration, 兵 P ij 其 ⫽total external nodal loads applied on
nodes along the length into a number of line elements as the structure at jth iteration, and 兵 F ij⫺1 其 ⫽internal element forces
shown in Fig. 6 and each of the cross sections is analyzed summed at each node of the structure up to the ( j⫺1)th iteration
with the stress distribution shown in Fig. 3. during the ith load increment.
2. Stiffness equations are derived for each element in terms of A generalized stiffness parameter 共GSP兲 is introduced to com-
the nodal degrees of freedom 共Fig. 7兲. This step ensures the pute the iterative parameter until the convergence of incremental
equilibrium of individual elements are satisfied. equilibrium equation is achieved. In GSPi for any load increment,
‘‘i’’ is defined as the ratio of the norm of the first iterative dis-
placement vector of the first load increment step to those at the
current load increment step

兵 ␦Û 11 其 T 兵 ␦Û 11 其
GSP ⫽ i
(15)
兵 ␦Û i⫺1
1 其 兵 ␦Û 1 其
T i

where 兵 ␦Ū ij 其 ⫽iterative residual displacement vector at the jth

iteration of the ith load increment and 兵 ␦Û ij 其 ⫽iterative tangential
displacement vector at the jth iteration of the ith load increment.
For the first load increment, GSP1 value is equal to one. The first
Fig. 7. Nodal forces and moments
iterative load parameter of the first load increment, ␦␭ 11 , is equal

JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002 / 925

J. Struct. Eng. 2002.128:922-933.

 Table 1. Comparison with Short Column Test Results by Shakir-
Khalil and Mouli 共1990兲

P expt P calc Pexpt

Column fy f cu
section 共N/mm2兲 共N/mm3兲 共kN兲 共kN兲 Pcalc
120⫻100⫻5 357.5 41.6 850 881.0 0.97
341.0 42.6 900 854.0 1.05
341.0 46.2 920 875.0 1.05
362.5 42.4 950 895.5 1.06
362.5 40.8 955 881.8 1.08

150⫻100⫻5 346.7 46.0 1210 1206.0 1.00

Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

346.7 46.2 1200 1210.0 0.99 Fig. 9. Typical load-moment interaction diagram
340.0 46.6 1190 1194.0 0.99
340.0 47.2 1200 1203.0 0.99
The M -␸- P and load-axial strain relationship plots are the
Note: Mean⫽1.02 and Standard deviation⫽0.04.
final outcome of the cross-section analysis. For every load incre-
ment, the M -␸- P and load-axial strain relationship plots are thus
to the input value, and for the subsequent load increments, it can used to update the structure stiffness matrix to ensure the equilib-
be computed using the known GSPi value of that particular load rium. After the stiffness is modified, the solution of Eq. 共14兲 共by
increment as the GDC method兲 gives the final results for forces and lateral
deformations of the column along its length.
␦␭ i1 ⫽⫾␦␭ 11 兩 GSPi 兩 1/2 (16)
The iterative load parameter, ␦␭ ij
, for all the subsequent itera- Results and Discussion
tions ( j⬎1) is computed using the following expressions:
The proposed analytical method was applied to investigate the
behavior of pin-ended square, rectangular, and circular box col-
兵 ␦Û i⫺1
1 其 兵 ␦Ū j 其
T i
␦␭ ij ⫽⫺ (17) umns filled with concrete. Parameters such as concrete and steel
兵 ␦Û i⫺1
1 其 兵 ␦Û j 其
T i strengths, the cross-sectional dimensions of section, and eccen-
tricity of applied loading were considered. Several columns tested
As the GDC method is based on the bounded characteristics of by other researchers were analyzed 共Lakshmi and Shanmugam
load parameter and displacement increments, it is able to bypass 2000b兲 and the results are presented in the form of interaction
the limit points without causing any numerical instability. diagrams, load-deflection plots, and comparison in tables. The
analytical values are compared with the corresponding experi-
Convergence Criteria mental results in order to assess the accuracy of the proposed
For each load increment, the equilibrium equation is solved by method.
iterations until the unbalanced force vector 兵 R ij⫺1 其 becomes neg-
ligible. This is indirectly achieved by the following energy crite-
ria: Short Columns
Sections 120⫻100⫻5 rectangular hollow section 共RHS兲 and
兩 兵 ␦U ij 其 T 兵 R ij⫺1 其 兩 150⫻100⫻5 RHS were tested by Shakir-Khalil and Mouli
⬍␳ E (18) 共1990兲 as short columns of length 200 mm with very little vari-
兩 兵 ␦U i1 其 T 共 ␦␭ i1 兵 P̂ 其 兲 兩
ance in material strength of both concrete and steel. These speci-
in which ␳ E is a user-specified tolerance, usually chosen between mens were tested for the squash load capacity of slender columns
1⫻10⫺6 and 1⫻10⫺10. Alternatively to terminate the iteration 共listed in Table 3兲. For the 120⫻100⫻5 RHS a tube with a length
for nonconverging and slow-converging systems, a maximum of 200 mm was used to determine thesquash load, while for the
number of iterations per load step is also imposed. 150⫻100⫻5 RHS, tubes of 100 mm length and another of 200

Table 2. Comparison with Square Columns Tested by Bridge 共1976兲

Size t L Load e P expt P calc Pexpt

Column 共mm兲 共mm兲 共mm兲 incl. 共deg兲 共mm兲 共kN兲 共kN兲 Pcalc
SHC-1 200 9.96 2130 0 38 1956 1845 1.06
SHC-2 10.01 3050 0 0 2869 2562 1.12
SHC-3 10.03 2130 30 38 2180 1946 1.12
SHC-4 9.88 2130 45 38 2162 1948 1.11
SHC-5 10.01 3050 30 38 2037 1869 1.09
SHC-6 9.78 3050 45 64 1623 1503 1.08
SHC-7 150 6.48 3050 0 38 680 613 1.11
SHC-8 6.48 3050 0 64 513 466 1.10
Fig. 8. Comparison of interaction curves
Note: Mean⫽1.10 and Standard deviation⫽0.02.

926 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng. 2002.128:922-933.

 Table 3. Comparison with Slender Columns Tested by Shakir-Khalil and Zeghiche 共1989兲 and Shakir-Khalil and Mouli 共1990兲

P expt P calc Pexpt

ex ey f sd f cu
Authors Section 共mm兲 共mm兲 共N/mm2兲 共N/mm2兲 共kN兲 共kN兲 Pcalc Mean and standard deviation

Shakir-Khalil and Zeghiche 共1989兲 120⫻80⫻5 0 0 386.3 44.0 600.0 619.0 0.97 Mean⫽1.04
24 0 386.3 40.0 393.0 374.0 1.05 Standard deviation⫽0.07
60 0 384.7 40.0 232.0 215.0 1.08
0 16 384.7 44.0 260.0 263.0 0.99
0 40 343.3 43.0 210.0 191.0 1.10
24 16 343.3 45.0 268.0 237.0 1.13
60 40 357.5 44.0 160.0 165.0 0.97
Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

Shakir-Khalil and Mouli 共1990兲 120⫻100⫻5 0 8 357.5 41.6 407.2 384.2 1.06 Mean⫽1.01
12 8 341.0 42.6 348.0 322.2 1.08 Standard deviation⫽0.05
42 28 341.0 42.6 198.5 204.6 0.97
24 40 362.6 42.4 206.8 217.7 0.95
60 16 362.6 40.8 209.8 205.7 1.02
150⫻100⫻5 0 0 346.7 46.0 1003 1003 1.00
15 10 346.7 46.2 596.2 621.0 0.96
45 30 340.0 46.6 329.2 316.5 1.04
75 50 340.0 47.2 254.6 242.5 1.05

mm length were used. However, analytical results have been P calc is 1.02 and it can therefore be concluded that the proposed
computed for short columns of 200 mm length in all cases and the method provides an accurate solution.
results are compared in Table 1. The calculated values in all cases
are close to the corresponding experimental results. The mean
value of the ratio of experimentalload P expt to calculated load Interaction Diagrams
Short columns subjected to uniaxial bending tested by Shakir-
Khalil and Al-Rawdan 共1996兲 were analyzed using the proposed
Table 4. Comparison with Specimens Tested by Matsui et al. 共1995兲 method. The section 150⫻100⫻5 RHS of yield strength f y
⫽350 N/mm2 with in-filled concrete of strength f cu 45 N/mm2
e P expt P calc Pexpt was used in the study. The side walls in the tubular sections are
Length
Number 共mm兲 共mm兲 共kN兲 共kN兲 Pcalc assumed as plates, supported on four sides for local buckling
calculations. Uy and Bradford 共1994兲 recommended that these
1 0.0 1598 1493 1.07
plates can be taken as clamped along the longitudinal edges and
2 600 25.0 1184 1117 1.06
the buckling coefficient k be therefore assumed as 10.3. The sec-
3 75.0 734 680 1.08
tion considered, having a plate slenderness b/t of 30, can be
4 125.0 514 504 1.02 considered compact and thus no reduction in strength due to local
5 0.0 1586 1510 1.05 buckling is accounted for. The interaction diagram for the above
6 1200 25.0 1133 1089 1.04 section obtained by the proposed method is compared in Fig. 8
7 75.0 665 621 1.07 with those by Shakir-Khalil and Al-Rawdan. The results lie close
8 125.0 484 448 1.08 to each other almost throughout the whole load range except at
9 0.0 1573 1605 0.98 lower axial load. The present model is slightly more conservative
10 1800 25.0 1025 1079 0.95
11 75.0 631 631 1.00
12 125.0 445 408 1.09
Table 5. Comparison with Slender Columns Tested by Wang 共1999兲
13 0.0 1356 1267 1.07
14 2700 25.0 846 798 1.06 P expt P calc Pexpt
ex ey
15 75.0 553 595 0.93 Column 共mm兲 共mm兲 共kN兲 共kN兲 Pcalc
16 125.0 893 971 0.92
RHS1 0 55 368 338 1.09
17 0.0 1143 1089 1.05
RHS2 0 55 246 220 1.12
18 3600 25.0 705 691 1.02
19 75.0 441 408 1.08 RHS3 55 110 172 150 1.15
20 125.0 326 313 1.04 RHS4 55 110 238 225 1.06
21 0.0 909 850 1.07 RHS5 55 110 251 226 1.11
22 4500 25.0 588 565 1.04 RHS6 55 55 234 215 1.09
23 75.0 373 419 0.89 RHS7 55 0 520 481 1.08
24 125.0 277 280 0.99 RHS8 55 0 480 425 1.13
Note: Mean⫽1.03 and Standard deviation⫽0.06. Note: Mean⫽1.10 and Standard deviation⫽0.03.

JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002 / 927

J. Struct. Eng. 2002.128:922-933.

 Table 6. Properties of Square Column Analyzed under Uniaxial
Bending
Column Section detail e 共mm兲 e/D
LC1 200⫻200⫻10 0 0.00
LC2 of length 3.0 m and 10 0.05
LC3 L/D ratio⫽15. 20 0.10
LC4 30 0.15
LC5 40 0.20
LC6 50 0.25
Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

excess of its ultimate moment of resistance M u 共27.3 kN m兲 with-

out axial loading. This is similar to the effect that a prestressing
force has on a reinforced concrete section. However, if the axial
Fig. 10. Effect of eccentricity on load-deformation response force and bending moment are independently applied to the com-
posite section, it should be noted that the removal of the axial
load would destabilize the section and cause its failure. It is there-
compared to the other two methods at lower axial load. The dif- fore recommended in the design of composite columns to con-
ference can be attributed to the inclusion of tensile strength of sider the cutoff at M u to form an integral part of the interaction
concrete 共taken as 10% of its compressive strength in bending兲 in curve. The interaction curve for combined compression and bend-
the finite element analysis by Shakir-Khalil and Al-Rawdan, ing can be found by moving the neutral axis over the whole cross
which has been neglected in the present method. section and determining the internal action effects from the result-
Unlike in hollow steel tubes, buckling of the steel tube in ing stress blocks.
in-filled composite columns is prevented by the outward push of
the contained concrete against the tube wall. As a result of this
Slender Columns
interaction between steel and concrete, the axial load acting on a
short concrete filled hollow section may be increased beyond the Tests on slender composite columns were carried out by a number
load level at which the steel yields. The ultimate load of the of researchers and these columns were analyzed by the proposed
column is reached when the stress in the contained concrete method in order to prove its validity. Bridge 共1976兲 investigated
reaches the maximum strength of concrete. Typical interaction the behavior of pin-ended concrete-filled square steel tubes
diagrams obtained using the present method for sections 120 loaded eccentrically about any axis; parameters such as the thick-
⫻80⫻5 and 150⫻100⫻5 mm of yield strength 350 N/mm2 filled ness of the steel tube t, eccentricity e, length L, and angle at
with concrete of strength 45 N/mm2 and subjected to uniaxial which the line of load is inclined to the horizontal axis were
bending about the major axis are given in Fig. 9. It can be seen considered in the study. The geometric and material properties of
from the figure that when subjected to a small axial load, the the test specimens are given in Table 2. The columns were ana-
composite section is capable of supporting a bending moment in lyzed by the present method and the results are compared with the

Fig. 11. Effect of eccentricity on load-deflection

928 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng. 2002.128:922-933.

 yield strength 412 N/mm2 was filled with concrete of compressive
strength 31.9 N/mm2. The length of the columns varied from 600
mm to 4.5 m covering a wide range of column slenderness and
the eccentricities about both axes varied from zero to a maximum
of 125 mm. Theoretical and experimental strengths of 24 columns
are presented in Table 4. Close agreement between the experi-
mental and analytical results is observed over a wide range of
column slenderness 共Table 5兲.

Load-Deflection Response
Fig. 12. Load-deflection curves of columns in biaxial bending Fig. 10 shows the relation between axial load and midheight de-
Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

formation obtained by the proposed analytical method for a sec-

tion of 250⫻150⫻6.3 mm subjected to load applied at an eccen-
experimental values. The predictions are conservative, although tricity varying from 0 to 35 mm about the major axis. These
not very much so. Comparisons between measured values P expt columns of length 3 m pinned at the ends are subjected to single
and predicted values P calc of column capacities show that the curvature bending. A comprehensive parametric study shows that
proposed method is sufficiently accurate. the two main parameters that influenced the force-deflection re-
Shakir-Khalil and Zeghiche 共1989兲 tested 2.76 m long rectan- sponse are column slenderness and eccentricity of the applied
gular columns of section 120⫻80⫻5 RHS under biaxial load load. It can be seen from the figure that the load-deflection re-
applied at eccentricities about major (e x ) and minor (e y ) axes. sponse remains very stiff even up to the ultimate load in the case
Further tests on rectangular columns 共120⫻80⫻5 and 150 of columns subjected to axial load. As the eccentricity increases
⫻100⫻5 RHS兲 of 3 m length were carried out by Shakir-Khalil the load-carrying capacity drops significantly and the initially
and Mouli 共1990兲. The experimental and the corresponding ana- shown stiffness decreases.
lytical values of the failure load obtained by the proposed method Square columns of section 200⫻200⫻10 mm were considered
are listed in Table 3. The comparison between the two sets of to study the complete load-deflection response under eccentric
values show good agreement thus confirming the ability of the compression producing single curvature bending. The geometric
proposed method to predict the ultimate load accurately. properties of columns LC1 to LC6 with yield strength at 290
Slender columns tested by Matsui et al. 共1995兲 were also con- N/mm2 for steel tubes and with in-filled concrete of compressive
sidered in this study. The columns were of square cross sections strength of 30 N/mm2 are listed in Table 6. The eccentricity ratio
of 149.8 mm width and 4.27 mm thickness. The steel tube of e/D was varied from a minimum value of 0.0 to a maximum of

Fig. 13. Experimental and analytical strengths for columns of square and rectangular sections

JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002 / 929

J. Struct. Eng. 2002.128:922-933.

 Table 7. Comparison with Specimens Tested by Rangan and Joyce
共1992兲

P expt P calc Pexpt

Length
Number mm e mm kN kN Pcalc
1 807.5 10 430 434 0.99
2 807.5 30 235 224 1.05
3 1312.5 10 350 324 1.08
4 1312.5 30 190 179 1.06
5 1565.0 10 315 325 0.97
6 1817.5 10 280 275 1.02
7 1817.5 30 140 143 0.98
Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

8 2322.5 10 220 212 1.04

Fig. 14. Load-deflection plots for O’Brien and Rangan’s specimen
9 2322.5 30 126 117 1.08
Note: Mean⫽1.03 and Standard deviation⫽0.04.

corresponding experimental values of ultimate failure loads for a

wide range of 65 composite columns described above is summa-
0.25 for a constant slenderness ratio. The resulting load-deflection rized in Fig. 13. This comparison covers square and rectangular
responses are shown in Fig. 11. It is evident from the figure that in-filled columns tested by various researchers. It is observed
the increase in eccentricity of the force produces a more flexible from the figure that all points lie within ⫾10% lines thus con-
response in both ascending and softening regions with a concomi- firming the ability of the proposed method to predict the failure
tant reduction in column strength. Conversely, as force eccentric- accurately.
ity decreases, the overall response becomes steeper in the ascend- The factor 0.85 used in the determination of load capacity of
ing portion with a sharper decline once the column strength is the cross section, which represents the ratio between cylinder and
exceeded. cube strength, can vary from a minimum of 0.67 to a maximum of
Load-deflection responses for a column tested by Shakir- 0.9. This ratio depends on many factors and thus is not constant.
Khalil and Zeghiche 共1989兲 are shown in Fig. 12. Deflections The factor used in the analysis can be attributed to its conserva-
about the x as well as y axis are plotted in the figure and the tive nature instead of underestimating the experimental results in
corresponding curves predicted by the present method are super- spite of using an unconfined concrete model.
imposed for comparison. Variations of deflections in the x and y
directions with applied loading are given in Fig. 12. It is obvious
from the figure that the experimental and analytical load- Circular Composite Columns
deflection responses about both the x and y axes are close to each
other, in the elastic as well as in the inelastic regions. The proposed method was also employed to analyze columns of
Comparison between the predicted and measured values of circular cross sections. A number of columns tested by other re-
ultimate load is generally good. Assessment of analysis is depen- searchers have been analyzed and the results are compared with
dent primarily on the accuracy with which the theoretical the corresponding experimental failure loads.
moment-curvature-thrust relationships reflect the actual response Experimental results of nine eccentrically loaded slender steel
of the cross section. These relationships are in turn directly influ- tubular columns filled with high strength concrete 共Rangan and
enced by assumptions regarding the distribution of strain through Joyce 1992兲 are compared against the corresponding analytical
the cross section and the stress-strain relationships of both con- values obtained by the present method in Table 7. The steel tube
crete and steel. Comparison between the analytical results and the had an outer diameter of 101.6 mm with a wall thickness of 1.6
mm. Average values of steel and concrete strength, 218 and 67.4
MPa, respectively, reported by the researchers have been used in
the analysis.
The length of steel tubes was varied from 800 to 2,300 mm
Table 8. Comparison with Specimens Tested by O’Brien and Rangan and the columns were pinned at both ends with eccentric loading.
共1993兲 Parameters under consideration include effective length of col-
Wall Yield umn and eccentricity of the applied load. Mean value of the ratio
P expt P calc Pexpt between experimental to calculated failure load is 1.03 implying
Length thickness strength
Number 共mm兲 共mm兲 共N/mm2兲 共kN兲 共kN兲 Pcalc that the proposed analytical method is capable of predicting ac-
curately the ultimate load capacity of circular composite columns.
10 1755 2.6341 190 176 1.08
Similar results corresponding to further tests carried out by
11 1375 241 254 0.95
O’Brien and Rangan 共1993兲 are presented in Table 8 which shows
12 995 279 293 0.95
a good agreement between the experimental and analytical values.
13 615 327 363 0.90
Fig. 14 shows typical load-deflection plots for specimens 10, 11,
14 520 425 401 1.06
15, and 16 共listed in Table 7兲.
15 1755 3.2 332 185 189 0.98
Circular in-filled composite columns were tested under short-
16 1375 237 228 1.04 term loading by Neogi et al. 共1969兲 and Kilpatrick 共1996兲. All
17 995 306 333 0.92 columns were subjected to single curvature bending with applied
18 615 388 392 0.99 load at equal end eccentricities. Table 9 shows the parameters
Note: Mean⫽0.99 and Standard deviation⫽0.06. involved in the study for the test specimens and comparison be-

930 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng. 2002.128:922-933.

 Table 9. Comparison with Specimens Tested by Neogi et al. 共1969兲 and Kilpatrick 共1996兲

P expt P calc Pexpt

Authors Number e/D L/D 共kN兲 共kN兲 Pcalc Mean and standard deviation
Neogi et al. 共1969兲 M1 0.28 78.6 622 604 1.03 Mean⫽1.04
Standard deviation⫽0.05
M2 0.23 78.7 702 731 0.96
M3 0.28 78.7 600 561 1.07
M4 0.28 79.0 625 613 1.02
M5 0.28 78.7 653 599 1.09
M6 0.23 78.7 739 754 0.98
M7 0.28 78.2 758 729 1.04
Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

M8 0.23 94.9 548 522 1.05

M9 0.23 94.9 548 503 1.09
M10 0.23 94.4 417 453 0.92
C5 0.05 44.3 965 894 1.08
C6 0.05 44.4 1046 987 1.06
C7 0.05 54.0 852 789 1.08
C8 0.05 54.0 805 759 1.06
C9 0.13 63.9 354 328 1.08
C10 0.13 64.0 526 510 1.03
C11 0.18 64.0 344 355 0.97
C12 0.18 64.0 503 479 1.05

Kilpatrick 共1996兲 1 0.10 31.8 430 453 0.95 Mean⫽1.01

2 0.30 31.8 235 255 0.92 Standard deviation⫽0.06
3 0.10 51.7 350 330 1.06
4 0.30 51.7 190 183 1.04
5 0.10 61.6 315 292 1.08
6 0.10 71.6 280 292 0.96
7 0.30 71.6 140 141 0.99
8 0.10 91.4 220 213 1.03
9 0.30 91.4 126 120 1.05
10 0.13 92.2 190 178 1.07
11 0.13 72.3 241 254 0.95
12 0.13 52.3 279 303 0.92
13 0.13 32.3 327 306 1.07
14 0.13 27.3 425 401 1.06
15 0.13 92.2 185 221 0.89
16 0.13 72.3 237 221 1.07
17 0.13 52.3 306 297 1.03
18 0.13 32.3 388 392 0.99

tween the experimental and the corresponding theoretical failure composite columns of square, rectangular, and circular cross sec-
loads using the proposed method. tions has been proposed. This approach eliminates the limitation
Close agreement between the two values is observed in the of the conventional analysis in which a deflection shape or pattern
table. All specimens failed at the midheight as concrete attained is usually assumed. Comparison of ultimate strength, which was
its crushing strain in compression whereas the extreme fiber did obtained by using the proposed method with the corresponding
not reach the yield strain of steel. Again the comparison with experimental results for columns, that were tested by other re-
analytical results shows good agreement thus ensuring the accu- searchers, has proven the accuracy of the proposed method. Based
racy of the method. A summary of results for a total of 54 circular on the results, moment capacity of columns is found to decrease
composite columns is shown in Fig. 15. Comparison between the with the increase in applied axial load. For the eccentrically
experimental failure loads and the corresponding analytical re- loaded columns, load carrying capacity is found to drop signifi-
sults obtained by the proposed method is given in the figure. It is
cantly with an increase of eccentricity. Giving due consideration
clear from the figure that all points lie within ⫹10% and ⫺10%
to uncertainties concerning the properties of the materials and the
thus confirming the accuracy of the proposed method to predict
effects of inevitable experimental errors, it has been found that
the ultimate load capacity of circular composite columns.
the proposed analytical model could predict accurately the
strength of concrete filled circular, square, and rectangular steel
Conclusions tubular columns having a wide range of column slenderness. The
method can, therefore, be used for direct analysis of short and
An analytical method to predict the inelastic and ultimate load slender composite columns subjected to uniaxial or biaxial bend-
behavior and to compute the ultimate strength of steel-concrete ing.

JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002 / 931

J. Struct. Eng. 2002.128:922-933.

Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 15. Experimental and analytical strengths for columns of circular sections

Notation L ⫽ length of column;

M ⫽ bending moment;
The following symbols are used in this paper: Mu ⫽ ultimate moment of resistance;
A a , A c ⫽ cross-sectional area of structural steel and con- Mx , My ⫽ moment about x and y axes, respectively;
crete, respectively; N pl,rd ⫽ ultimate load of composite cross section;
A s , A c ⫽ elemental area of steel and concrete strips, Nu ⫽ squash load of composite cross section;
respectively; P ⫽ applied axial load;
b/t ⫽ breadth to thickness ratio; P expt , P test ⫽ experimental strength of column;
D ⫽ depth/diameter of cross section; P calc ⫽ calculated strength of column by using pro-
E s ⫽ elastic modulus of steel; posed analytical method;
EA ⫽ axial stiffness of cross section;
兵2P其 ⫽ total load to be applied;
EI ⫽ flexural stiffness of cross section;
兵1P其 ⫽ external load at beginning of incremental
e x ⫽ eccentricity about major axis;
step;
e y ⫽ eccentricity about minor axis;
F cs ⫽ total compressive force in steel tube;
兵 P ij 其 ⫽ vector of total external nodal loads at jth
iteration of ith load increment;
F cc ⫽ total compressive force in concrete;
F st ⫽ total tensile force in steel tube; 兵 R ij⫺1 其 ⫽ vector of unbalanced forces during ( j⫺1)th
兵 F ij⫺1 其 ⫽ vector of total internal elemental forces up iteration of ith load increment;
to ( j⫺1)th iteration of ith load increment; t ⫽ thickness of steel tube;
f cu ⫽ compressive strength of concrete; 兵U其 ⫽ structure displacement vector;
f cd , f yd ⫽ design strengths of structural steel and con- u, v ⫽ lateral deflection along x and y directions,
crete, respectively; respectively;
f sd ⫽ design strength of steel tube; X, Y ⫽ intercepts of neutral axis with x and y axes;
f y ⫽ yield strength of steel tube; xc , xs ⫽ distance from centroid of elemental area of
GSPi ⫽ Generalized stiffness parameter at ith load concrete and steel, respectively, to plastic
increment; centroid of the cross section along x direc-
K ⫽ resultant curvature; tion;
K x , K y ⫽ curvature about x and y axes, respectively; yc , ys ⫽ distance from centroid of elemental area of
关K兴 ⫽ structure stiffness matrix; concrete and steel, respectively, to plastic
关 K ij⫺1 兴 ⫽ tangent stiffness matrix formed at beginning centroid of the cross section along y direc-
of jth iteration of ith load increment; tion;
k ⫽ buckling coefficient; ␣ ⫽ inclination of neutral axis with x axis;

932 / JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002

J. Struct. Eng. 2002.128:922-933.

 兵 ␦U ij 其 ⫽ iterative displacement vector at jth iteration steel tubular columns under combined compression and bending.’’
of ith load increment; Proc., 4th Pacific Struct. St. Conf., Singapore, Pergamon, 3共10兲, 29–
36.
兵 ␦Ū ij 其 ⫽ iterative residual displacement vector at jth Mirza, S. A., and Skrabek, B. W. 共1992兲. ‘‘Statistical analysis of slender
iteration of ith load increment; composite beam-column strength.’’ J. Struct. Eng., 118共5兲, 1312–
兵 ␦Û ij 其 ⫽ iterative tangential displacement vector at 1332.
jth iteration of ith load increment; Neogi, P. K., Sen, H. K., and Chapman, J. C. 共1969兲. ‘‘Concrete-filled
␦␭ 11 ⫽ initial load parameter; tubular steel columns under eccentric loading.’’ Struct. Eng., 47共5兲,
␦␭ i1 ⫽ iterative load parameter for jth iteration of 187–195.
ith load increment; O’Brien, A. D., and Rangan, B. V. 共1993兲. ‘‘Tests on slender tubular steel
␧ ⫽ strain at any fiber of cross section; columns filled with high strength concrete.’’ Aust. Civil Eng. Trans.,
␧ cc ⫽ extreme concrete fiber strain; 35共4兲, 287–292.
␧ o ⫽ strain at plastic centroid of cross section; Prion, H. G. L., and Boehme, J. 共1989兲. ‘‘Beam-column behaviour of
Downloaded from ascelibrary.org by University of Illinois at Chicago on 05/15/13. Copyright ASCE. For personal use only; all rights reserved.

␧ u ⫽ crushing strain of concrete; steel tubes filled with high-strength concrete.’’ 4th Int. Colloquium,
SSRC, New York, 439– 450.
␳ E ⫽ specified tolerance;
Rangan, B. V., and Joyce, M. 共1992兲. ‘‘Strength of eccentrically loaded
␴ c , ␴ s ⫽ stress in concrete and steel, respectively;
slender steel tubular columns filled with high strength concrete.’’ ACI
and Struct. J., 89共6兲, 676 – 681.
␸ ⫽ curvature. Shakir-Khalil, H., and Al-Rawdan, A. 共1996兲. ‘‘Composite construction in
steel and concrete III.’’ Proc., Engineering Foundation Conf., Swa-
bian Conference Center, Germany, D. Buckner and B. M. Shahrooz,
References eds., ASCE, 222–235.
Shakir-Khalil, H., and Mouli, M. 共1990兲. ‘‘Further tests on concrete-filled
Bradford, M. A. 共1996兲. ‘‘Design strength of slender concrete-filled rect- rectangular hollow section columns.’’ Struct. Eng., 68共20兲, 405– 413.
angular steel tubes.’’ ACI Struct. J., 93共2兲, 229–235. Shakir-Khalil, H., and Zeghiche, J. 共1989兲. ‘‘Experimental behaviour of
Bridge, R. Q. 共1976兲. ‘‘Concrete-filled steel tubular columns.’’ Civil Eng. concrete-filled rectangular hollow-section columns.’’ Struct. Eng.,
Trans., Inst. Eng. Australia, 18, 127–133. 67共19兲, 346 –353.
Ge, H., and Usami, T. 共1992兲. ‘‘Strength of concrete-filled thin-walled Uy, B., and Bradford, M. A. 共1994兲. ‘‘Local buckling of concrete-filled
steel box columns: Experiment.’’ J. Struct. Eng., 118共11兲, 3036 –3054. high strength steel box columns for tall buildings: behaviour and de-
Ge, H., and Usami, T. 共1994兲. ‘‘Strength of concrete-filled thin walled sign.’’ Struct. Design Tall Buildings, 3, 75–93.
steel box columns: Experiment.’’ J. Constr. Steel Res., 37, 607– 612.
Uy, B., and Das, S. 共1997兲. ‘‘Behaviour and design of concrete filled
Kato, B. 共1996兲. ‘‘Column curves of steel-concrete composite members.’’
fabricated steel box columns.’’ Proc., 15th Australian Conf. on the
J. Constr. Steel Res., 39共2兲, 121–135.
Mechanics of Structure and Materials, Balkema, Rotterdam, 129–
Kilpatrick, A. E. 共1996兲. ‘‘The behaviour of high-strength composite con-
134.
crete columns.’’ PhD thesis, Curtin Univ. of Technology, Australia.
Kitada, T. 共1998兲. ‘‘Ultimate strength and ductility of state of art Uy, B., and Patil, S. B. 共1996兲. ‘‘Concrete-filled high strength steel box
concrete-filled steel bridges piers in Japan.’’ Eng. Struct., 20共4 – 6兲, columns for tall buildings: behaviour and design.’’ Struct. Design Tall
347–354. Buildings, 5, 75–94.
Lakshmi, B., and Shanmugam, N. E. 共2000a兲. ‘‘Behaviour of steel- Wang, Y. C. 共1999兲. ‘‘Tests on slender composite columns.’’ J. Constr.
concrete composite columns.’’ Proc., 6th ASCCS Int. Conf. on Steel Steel Res., 49, 25– 41.
and Concrete Composite Structures, Association for International Wang, Y. C., and Moore, D. B. 共1997兲. ‘‘A design method for concrete-
Cooperation and Research in Steel-Concrete Composite Structures filled hollow section composite columns.’’ Struct. Eng., 75共21兲, 368 –
共ASCCS兲, Los Angeles, 449– 456. 372.
Lakshmi, B., and Shanmugam, N. E. 共2000b兲. ‘‘RHS/SHS columns with Yang, Y. B., and Kuo, S. R. 共1994兲. Theory and analysis of nonlinear
concrete in-fill.’’ Proc., Structures Congress’2000, ASCE, Philadel- framed structures, Prentice Hall, Singapore.
phia. Yang, Y. B., and Shieh, M. S. 共1990兲. ‘‘Solution method for nonlinear
Matsui, C., Tsuda, K., and Ishibashi, Y. 共1995兲. ‘‘Slender concrete filled problems with multiple critical points.’’ AIAA J., 28共12兲, 2110–2116.

JOURNAL OF STRUCTURAL ENGINEERING / JULY 2002 / 933

J. Struct. Eng. 2002.128:922-933.