Numerical Analysis of Composite Steel-Concrete Columns

of Arbitrary Cross Section

João Batista Marques de Sousa Jr.1 and Rodrigo Barreto Caldas2
Downloaded from ascelibrary.org by Indian Institute of Tech Gandhinagar on 12/22/15. Copyright ASCE. For personal use only; all rights reserved.

Abstract: This paper presents a numerical formulation for the nonlinear analysis of slender steel-concrete composite columns of generic
cross-sectional shape, subjected to axial force and biaxial bending. The cross section is defined in terms of a number of closed polygonal
loops of a specific material, each one with its own stress-strain relation, with reinforcement bars embedded in the polygons. The material
and geometrically nonlinear equlibrium problem is solved by the finite element method, with displacement-based stress resultant beam-
column elements. The proposed scheme turns possible, with a unified treatment, to perform analyses of concrete-filled steel tubes, fully
or partially encased steel profiles, or less usual cross sections present on composite construction. The robustness and accuracy of the
formulation is verified against numerical and experimental results available in the literature.
DOI: 10.1061/共ASCE兲0733-9445共2005兲131:11共1721兲
CE Database subject headings: Composite columns; Steel; Concrete; Finite element method; Numerical analysis; Cross sections.

Introduction design codes 共Oehlers and Bradford 1995兲. Significant differences

Composite structures, also called mixed or hybrid structures,
combine steel and reinforced concrete to benefit from each mate-
rial characteristic. Composite construction takes advantage of the
speed of construction, light weight and strength of steel, and the
higher mass, stiffness, damping properties, and economy of rein-
forced concrete. One of the most suitable structural elements for
this combination is the composite column, which in recent years
has received much attention by researchers and practicing engi-
neers. Steel-concrete composite columns have been employed in
high-rise buildings, bridges, piers, piles, and earthquake-resistant
structures.
The most usual types of composite columns are the concrete
filled steel tubes and the partially or fully encased steel profiles.
Encased steel elements have the additional advantage of protect-
ing from fire damage and local buckling of steel plates, while
concrete-filled tubes benefit from the increase in strength due to
concrete confinement. In general, reinforcement bars may be
present, as well as shear connectors to prevent slip.
Most practice codes, such as the widely used AISC-LRFD
共1999兲; ACI-318 共2002兲; Eurocode 4 共1994兲; and BS-5400 共BSI
1979兲 have incorporated simplified methods for analysis and de-
sign of composite columns. These provisions are generally ex-
trapolated from either reinforced concrete column or steel column
1
Associate Professor, Civil Engineering Dept., Escola de Minas, Univ.
Federal de Ouro Preto, 35400-000, Ouro Preto MG, Brazil
共corresponding author兲. E-mail: joao@em.ufop.br
2
Graduate Student, Civil Engineering Dept., Escola de Minas, Univ.
Federal de Ouro Preto, 35400-000, Ouro Preto MG, Brazil. E-mail:
engarte@bol.com.br
Note. Associate Editor: Donald W. White. Discussion open until April
1, 2006. Separate discussions must be submitted for individual 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 was submitted
for review and possible publication on April 13, 2004; approved on
January 31, 2005. This paper is part of the Journal of Structural Engi-
neering, Vol. 131, No. 11, November 1, 2005. ©ASCE, ISSN 0733-9445/
2005/11-1721–1730/$25.00.
between prescriptions of these codes 共El-Tawil et al. 1995; Zhang
and Shahrooz 1999; Saw and Richard Liew 2000; Weng and Yen
2002兲 indicate that further numerical and experimental research
must be undertaken to gain understanding about the structural
behavior of these elements.
The finite element method 共FEM兲 is the most popular tool for
simulation of structural response. Recent numerical work on com-
posite columns has been focused on either three-dimensional or
beam-column elements. Three-dimensional element formulations
are able to predict the complex behavior of the composite struc-
tures, using sophisticated material constitutive laws and interac-
tion models between the materials. Beam-column elements, on
the other hand, are simpler to develop and implement and less
expensive on computational terms.
Practical design situations often employ composite columns
under simultaneous axial load and biaxial bending. However, not
many numerical and experimental results are available for com-
bined axial force–biaxial bending loading, and design rules based
on linearization of the interaction curves may be too conservative.
The purpose of this paper is to develop and implement a nu-
merical procedure for the static analysis of slender steel-concrete
composite columns subjected to axial force and biaxial bending.
The main advantage of the formulation is its suitability for a
generic shape of cross section under different loading conditions,
providing a valuable tool for the analysis of composite columns.
At the end of the paper, selected examples prove the ability of
the proposed scheme to provide accurate and reliable results
compared to experimental data as well as other numerical
formulations.
Cross-Section Analysis
Composite cross-section analysis remains an active field of re-
search. In composite construction, the formability and robustness
of concrete, associated to the high resistance of steel profiles
allow the columns to be designed with various geometries. Ana-
lytical 共Roik and Bergmann 1990兲 and numerical procedures for

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2005 / 1721

J. Struct. Eng., 2005, 131(11): 1721-1730


Downloaded from ascelibrary.org by Indian Institute of Tech Gandhinagar on 12/22/15. Copyright ASCE. For personal use only; all rights reserved.
Fig. 1. Geometry of composite section
the analysis and design of reinforced concrete, steel and compos-
ite sections have been developed for over 30 years and related
recent developments have been published 共Rodriguez and
Aristizabal-Ochoa 1999; Chen et al. 2001; Sfakianakis 2002;
Bonet et al. 2004兲. Practice codes are often restricted to symmet-
ric sections and simple mechanical models, such as the rigid-
plastic assumption.
Cross-sectional analysis is a fundamental step on the formula-
tion and implementation of a successful numerical analysis of
composite columns by FEM. For stocky columns or in simplified
models which use approximations for geometric nonlinearity, it
may provide the solution of the column problem itself. The basic
numerical procedures involved are the determination of resultant
forces under predefined deformation variables and the evaluation
of derivatives of resultant forces with respect to the same
variables 共generalized stiffnesses兲. According to Spacone and
El-Tawil 共2004兲, two different approaches are generally used,
namely 共1兲 resultant model and 共2兲 fiber section models. Fiber
section model involves the subdivision of the section into small
elements which are assumed to be in a state of uniaxial stress.
This simple scheme allows the representation of any shape of
steel-concrete composite section. Effects such as concrete con-
finement, cracking, strain localization, and steel local buckling
may be simulated. Resultant section models define section re-
sponses in terms of values of moment-curvature relationships and
works with combined axial force-bending moment elastic bound-
ing surfaces. In both cases, plane sections are assumed to remain
plane even into the inelastic range.
In the present paper, cross-sectional response is defined in
terms of resultant axial force and bending moments, Nx, M y, and
M z 共Fig. 1兲, which are in turn functions of the section deformation
state, described by generalized strains, namely axial strain ␧0 and
curvatures ky and kz
Nx = 冕冕
Am
␴x共␧x兲dA
1722 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2005
J. Struct. Eng., 2005, 131(11): 1721-1730
My = 冕冕
Am
␴x共␧x兲zdA
Mz = − 冕冕
Am
␴x共␧x兲ydA 共1兲
The generalized strains are measured relative to a local system of
axes whose origin is conventionally adopted at the plastic cen-
troid of the cross section. Strain at any point 共y , z兲 may be evalu-
ated by ␧ = ␧0 + kyz − kzy. In usual situations 共doubly symmetric
sections兲 this point coincides with the geometric centroid. The
resultant forces and derivatives are obtained from analytical inte-
gration of the stress-strain relationship for each material 共Caldas
2004兲. This formulation has close similarities with the fiber
model.
The geometry of the cross section is described by a number of
closed counterclockwise loops, with internal openings described
clockwise, Fig. 1. Each loop encloses a specific material with a
piecewise-defined stress-strain relation. Curved boundaries are
approximated by an equivalent set of rectilinear segments. The
resultant forces of Eqs. 共1兲 may be split into a sum of integrals
over each material polygon, while reinforcement bar contribu-
tions are evaluated pointwise. If the uniaxial stress-strain relations
are composed by polynomials 共or adequately approximated by a
set of polynomials兲, Eqs. 共1兲 are integral expressions of integer
powers of y and z, and cross-sectional resultant forces and deriva-
tives with respect to the deformation variables may be evaluated
analytically. This well-known procedure is based on the transfor-
mation of the surface integrals into line integrals along the closed
boundaries of each region and appears to have been first sug-
gested for reinforced concrete section analysis by Werner 共1974兲
and later employed, for instance, by Rotter 共1985兲. Although quite
convenient, this scheme seems not to have been fully explored for
analysis of composite cross sections.
With piecewise polynomial stress-strain laws, it is necessary to
subdivide each material polygon into subpolygons, correspondent
to each subrange of the constitutive law. This can be accom-
plished in an elegant way by means of a well-known surface
contouring algorithm from computer graphics 共Martha et al.
1997兲. This method of analysis enables one to consider, for in-
stance, different levels of confinement for distinct concrete re-
gions, different residual stresses and effective widths for steel
profile plates, or even modified stress-strain relationships for the
case of structures subjected to fire action. Compared to the fiber
section model, it is unnecessary to subdivide the cross section into
small square regions, the results are exact for polynomial stress-
strain laws, there are no restrictions on the geometry of the sec-
tion and the integrations are much faster.
In the examples presented in this paper the stress-strain rela-
tions for steel 共and reinforcement兲 and concrete were assumed as
shown in Figs. 2 and 3. The concrete peak stress is dependent on
the confinement level assumed to each region of interest, and the
limit value of concrete strain ␧cu may vary according to its cylin-
der strength f cyl. If f cyl is under 50 MPa, ␧cu is taken equal to
−0.0035, and as −0.175/ f cyl but less than −0.0022 otherwise.
Finite Element Formulation
Several works have dealt with numerical analysis of composite
steel-concrete structures. Composite beams have received consid-
erable attention. Less research, however, has been devoted to

Fig. 2. Stress-strain relation for steel
Downloaded from ascelibrary.org by Indian Institute of Tech Gandhinagar on 12/22/15. Copyright ASCE. For personal use only; all rights reserved.
composite columns, and for biaxial bending of slender composite
columns there is not much numerical and experimental research.
Muñoz and Hsu 共1997a, b兲 performed experimental and numeri-
cal analysis of concrete-encased composite colums under biaxial
bending and proposed a simplified design procedure. More re-
cently, Lakshmi and Shanmugam 共2002兲 developed a semianalyti-
cal procedure for numerical analyses of in-filled composite col-
umns under biaxial bending which showed good agreement with a
large set of experimental data.
The numerical procedures may be roughly divided in two main
approaches.
1. Three-dimensional analysis with solid elements: This ap-
proach offers the widest range of possibilities for simulating
structural behavior. Sophisticated constitutive relations may
be employed, including plasticity, damage, and time depen-
dent effects. Shear connection may be modeled precisely as
well as bond-slip between the elements. Johansson and Gyll-
toft 共2002兲 employed this kind of strategy via commercial
code to analyze the influence of forms of load application;
and
2. Bar or beam-column elements: This approach employs one-
dimensional elements to model the beam-column axis. It
requires accurate evaluation of cross-sectional properties as
the physical nonlinearities are taken into account at section
level and carried to element level by means of numerical
integration.
It has been recognized that the employment of solid elements,
despite its accuracy and reliability, may be excessively time-
consuming for practical purposes and, in general, beam-column
elements offer a better balance between response precision and
computational cost. For beams, specialized formulations have
been developed to consider the effect of bond slip in the steel-
concrete interface 共Salari and Spacone 2001; Faella et al. 2002兲.
This effect is generally neglected when dealing with columns, due
to its not significant influence in the column’s ultimate capacity.
Recent developments on beam-column elements for reinforced
concrete, steel, and composite structures have employed classical
displacement-based as well as mixed and force-based formula-
tions 共Ayoub and Filippou 2000; El-Tawil and Deierlein 2001a,b;
Fig. 3. Stress-strain relation for concrete
JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2005 / 1723
J. Struct. Eng., 2005, 131(11): 1721-1730
Salari and Spacone 2001兲, and a lot of effort is still being put on
research with the objective of finding better formulations.
The FE model employed in this paper is based on cubic inter-
polation of transverse displacements and quadratic interpolation
for axial displacements. The main advantage of this rather classi-
cal scheme is its simplicity of formulation and implementation,
with easy incorporation into existing FE codes. Its drawback lies
on the fact that for inelastic response the solution will not in
general satisfy equilibrium and several elements may be neces-
sary to correctly simulate the behavior of a single bar. This has
been a major concern 共Chan 2001兲, leading to the development of
alternative procedures for beam-column analysis, such as mixed
or flexibility-based elements 共Ayoub and Filippou 2000; El-Tawil
and Deierlein 2001a,b兲 and the quartic interpolation technique
with constant axial force criterion 共Izzudin et al. 2002兲, among
many others. Notwithstanding, Teh 共2001兲 argued that this con-
cern about cubic elements is unjustified and stated that the cubic
formulation provides very good results for steel frame analyses
with just a few elements per member. Although this statement
cannot be directly extrapolated for reinforced concrete or com-
posite elements, the writers obtained accurate results with not
more than four elements per column, which in a personal com-
puter does not imply prohibitive computational cost.
The maximum concrete compressive stress value f c depends
on the confinement level, see, for instance, the discussion by El-
Tawil and Deierlein 共1999兲. In this work, for in-filled circular
steel tubes, concrete peak stress is considered to be equal to its
maximum cylinder compressive stress f cyl, otherwise it is as-
sumed to be equal to 0.85 f cyl. These values are based on the
Eurocode 4 共1994兲 specifications.
The element axis is supposed to coincide with the plastic cen-
troid of the cross section. Previous work 共Roik and Bergmann
1990; Chen et al. 2001兲 has showed that this assumption leads to
better results when applied to nonsymmetrical cross sections
under biaxial bending as it provides a smoother mapping between
the resultant moment and the neutral axis direction, which are in
general not parallel.
Element Formulation
The point of departure for the FE formulation is the principle of
virtual work stated as a condition of equilibrium for the structure
␦Wint − ␦Wext = 0 共2兲
It is assumed that each fiber of the column is in a uniaxial state of
strain. The expression for the internal virtual work for an element
of volume V, when only uniaxial stresses and strains are taken
into account, is
␦Wint = 冕冕 冕
V
␴x␦␧xdV 共3兲
Integration is performed relative to the undeformed element con-
figuration 共total Lagrangian approach兲. The end nodal degrees of
freedom are two transverse displacements, two rotations, and one
axial displacement, Fig. 4. As mentioned before, cubic interpola-
tion 共Hermitian basis functions兲 for the transverse displacements
v and w and quadratic interpolation for the axial displacement u
are employed at the element level
 The numerical analyses undertaken were stopped when the maxi-
mum concrete strain reached the conventional limit ␧cu.

Examples

Several researchers have carried out experimental analyses of

composite columns, and some have performed numerical investi-
gations as well. Shanmugam and Lakshmi 共2001兲 presented a
comprehensive compilation of experimental results dating back
from 1969. In this section, results from experimental and numeri-
Downloaded from ascelibrary.org by Indian Institute of Tech Gandhinagar on 12/22/15. Copyright ASCE. For personal use only; all rights reserved.

cal analyses from other sources are compared with the ones by the
present formulation. The examples try to encompass the largest
number of possibilities to illustrate the versatility of the formula-
tion. In all the examples Young modulus for the steel and rein-
forcement was taken as 210,000 MPa.

Slender Concrete Encased I-Sections

Mirza and coworkers 共1996兲 carried out an experimental and nu-
Fig. 4. Finite element for composite column merical study on the behavior of slender concrete-encased com-
posite beam-columns, with the presence of reinforcement, com-
prising a total number of 16 specimens. The beam-columns were
v = NTv qv, w = NTwqw, and u = NTu qu 共4兲 tested under a combination of axial force and transverse loads,
with high influence of second-order effects. Steel ribs were
At the structural level, the central node axial displacement is stati- welded along the steel flanges of I-sections of the specimens in
cally condensed out from the final equations. Upon substitution of order to study the influence of the mechanical connection between
the strain-displacement relation the materials. In the original work, the FE analysis employed
1 three-node, three-dimensional beam elements from a commercial
␧x = u⬘ + 关共v⬘兲2 + 共w⬘兲2兴 − y v⬙ − zw⬙ 共5兲 program 共Abaqus v. 4.8兲. I-beam elements were used to model the
2
steel section, with 15 integration points. For concrete 15⫻ 15 in-
共the prime indicates differentiation with respect to x兲, on Eq. 共5兲, tegration points were used and for reinforcement one integration
replacement of interpolation of the nodal variables and integration point per bar. The steel ribs were modeled using 10 integration
on the cross section, the element internal force vector may be points and eccentric elements. A modified version of the Desayi–
obtained Krishnan curve was employed for concrete 共Saenz 1964兲. More

冕冤 冥
details, including the geometries of the cross sections and the
NxN⬘u
material properties for concrete, steel, and reinforcement of each
fm = Nxv⬘N⬘v + M zN⬙v dx 共6兲 specimen, may be found in the original paper.
ᐉm
Nxw⬘Nw⬘ − M yNw⬙ The proposed model employed four elements with four inte-
gration points each, disposed so that the nodes coincided with the
Following standard procedure, derivation of the internal force points of application of the transverse loads. The comparison of
vector with respect to the nodal displacements leads to the ele- the numerical and experimental results obtained by Mirza et al.
ment tangent stiffness matrix: 共1996兲 are plotted against the present formulation in Table 1 and

再 冎
冤 冥
⳵Nx T Figs. 5–7. It can be seen that the numerical results display a very
N⬘u good agreement.
⳵q

kT = 冕 冉 N⬘v v⬘
⳵q
再 冎
⳵Nx T
冊
+ Nx
0u N⬘v 0w + N⬙v
⳵M z
⳵q
再 冎 T
dx Slender Circular Concrete Filled Tubes

冉 再 冎 冊 再 冎
ᐉm
⳵Nx T
⳵M y
Nw⬘ w⬘ + Nx
0u 0v Nw⬘  − Nw⬙
⳵q ⳵q
The element resultant forces Nx, M y, M z and their derivatives with
respect to the generalized displacements q are obtained analyti-
cally by the technique presented in the previous section.
The solution of the nonlinear equilibrium problem is obtained
using the Newton–Raphson technique with either load or dis-
placement control. The second choice is recommended to traverse
load limit points which is the general case for isolated composite
columns. Although a descending branch after peak load is usually
present, the examples analyzed herein generally do not show this
behavior strongly, mainly due to the constitutive relation adopted.
T
Results of nine experiments on slender concrete filled tubes sub-
jected to eccentric axial load were obtained from the work of
Rangan and Joyce 共1992兲. The cross-section specimens consisted
共7兲
of a steel tube of 1.6 mm thickness and external diameter
101.6 mm. The concrete cubic strength is 67.4 MPa and the steel
yield strength 218 MPa.
The numerical analysis using the proposed formulation em-
ployed four elements with four integration points each. Loading
was applied at both ends of the column resulting in single curva-
ture. The maximum allowable value for concrete strain was re-
laxed. This consideration was adopted because previous works
共Varma et al. 2002; Han 2004兲 have concluded that high-strength
concrete, when confined by a circular steel tube, displays a
strongly ductile behavior, with little decrease in the maximum

1724 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2005

J. Struct. Eng., 2005, 131(11): 1721-1730

 Table 1. Comparison of Results with Tests by Mirza et al. 共1996兲
Test Analysis Test Analysis
Specimen deflection deflection strength strength Test
designation 共mm兲 共mm兲 共kN兲 共kN兲 analysis
RHB-1 27.8 26.3 950.0 917.2 1.04
RHB-2 49.1 36.3 550.0 674.3 0.82
RHB-3 31.5 41.5 570.0 480.0 1.19
RHB-4 59.3 46.2 307.5 275.2 1.12
RHB-4A 56.8 48.7 154.3 155.9 0.99
RHB-5 56.4 50.7 95.0 90.1 1.05
Downloaded from ascelibrary.org by Indian Institute of Tech Gandhinagar on 12/22/15. Copyright ASCE. For personal use only; all rights reserved.

RNHB-1 39.8 33.2 925.0 867.4 1.07

RNHB-2 40.5 34.9 775.0 790.9 0.98
RNHB-3 46.5 41.3 540.0 511.4 1.06
RNHB-4 56.9 43.6 352.5 282.5 1.25
RNHB-5 53.7 51.4 107.5 96.0 1.12
RHNB-1 35.9 29.5 927.0 953.9 0.97
RHNB-2 37.9 35.2 720.0 778.9 0.92
RHNB-3 55.3 40.5 540.0 531.2 1.02
RHNB-4 64.1 45.8 296.0 295.4 1.00
RHNB-5 67.2 50.9 100.0 100.7 0.99
Average: 1.04
Standard deviation: 0.10
Note: RHB⫽specimens with perforated plate shear connectors; and RHNB⫽specimens without shear connection.

Fig. 5. Results for RHB series 关adapted from Mirza et al. 共1996兲兴

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2005 / 1725

J. Struct. Eng., 2005, 131(11): 1721-1730

Downloaded from ascelibrary.org by Indian Institute of Tech Gandhinagar on 12/22/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Results for specimens with non-perforated plate shear connectors 共RNHB兲 series 关共adapted from Mirza et al. 1996兲兴

Fig. 7. Results for specimens without shear connection 共RHNB兲 series 关共adapted from Mirza et al. 1996兲兴

1726 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2005

J. Struct. Eng., 2005, 131(11): 1721-1730

 Table 2. Comparison with Tests by Rangan and Joyce 共1992兲
Test Analysis
Specimen Length Eccentricity strength strength Test
designation 共mm兲 共mm兲 共kN兲 共kN兲 analysis
1 807.5 10 430 402 1.07
2 807.5 30 235 243 0.97
3 1,312.5 10 350 342 1.02
4 1,312.5 30 190 191 0.99
5 1,565 10 315 320 0.98
6 1,817.5 10 280 295 0.95
Downloaded from ascelibrary.org by Indian Institute of Tech Gandhinagar on 12/22/15. Copyright ASCE. For personal use only; all rights reserved.

7 1,817.5 30 140 161 0.87

8 2,322.5 10 220 246 0.89
9 2,322.5 30 126 134 0.94
Average: 0.97
Standard deviation: 0.06

stress beyond its peak value. Table 2 shows the results of the
analysis and comparison with values obtained by Rangan and
Joyce 共1992兲.
Slender Rectangular Concrete-Filled Tubes
under Biaxial Bending
From the work by Lakshmi and Shanmugam 共2002兲 the writers
reproduced the results obtained by Matsui et al. 共1995兲, in which
16 square concrete-filled tubes were tested. The tubes had sides
with 149.8 mm 共external兲 and thickness of 4.27 mm. The steel
tube had a yield stress of 412 MPa, and the concrete had cylinder
strength of 31.9 MPa 共average values兲. Columns from
0.6 to 4.5 m high were tested, with eccentricity varying from zero
to 125 mm along the two principal axes of the cross section.
Four identical elements with four integration points were em-
ployed in the numerical analysis. Table 3 displays the results and
comparisons with the experimental data, displaying very good
agreement. The greater difference observed for very slender col-
umns is probably due to the presence of geometrical imperfec-
tions which are more likely to be present in these cases. Lakshmi
and Shanmugam 共2002兲 took this into account by employing an
initial eccentricity equal to 0.001 times the column length. Al-
though it would be simple to incorporate this eccentricity into the
present formulation, without a reliable measure of imperfection
from the test program any adopted value would be arbitrary. For
design purposes, the imperfection values established by the de-
sign codes should be used.
Rectangular Concrete-Filled Tube with High Strength
Materials
Liu et al. 共2003兲 published results from tests of 21 short rectan-
gular concrete-filled steel tubes, assembled with high strength ma-
terials. The wall thickness was constant and equal to 4.18 mm, the
sides of the tube varied from 80.1 to 200.2 mm, and the length of
the specimens varied from 300 to 600 mm. The average yield
stress for steel was 550 MPa and the cylinder resistances for con-

Table 3. Comparison with Tests by Matsui et al. 共1995兲

Test Analysis
Specimen Length Eccentricity strength strength Test
designation 共mm兲 共mm兲 共kN兲 共kN兲 analysis
1 600 25 1,184 1,161 1.02
2 75 734 692 1.06
3 125 514 488 1.05
4 1,200 25 1,133 1,118 1.01
5 75 665 667 1.00
6 125 484 473 1.02
7 1,800 25 1,025 1,056 0.97
8 75 631 631 1.00
9 125 445 449 0.99
10 2,700 75 553 564 0.98
11 3,600 25 705 802 0.88
12 75 441 490 0.90
13 125 326 359 0.91
14 4,500 25 588 666 0.88
15 75 373 420 0.89
16 125 277 315 0.88
Average: 0.97
Standarddeviation: 0.07

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2005 / 1727

J. Struct. Eng., 2005, 131(11): 1721-1730

 Table 4. Comparison with Test by Liu et al. 共2003兲
Test Analysis
Specimen Width Height Thickness Length strength strength Test
designation 共mm兲 共mm兲 共mm兲 共mm兲 共kN兲 共kN兲 analysis
C1-1 100.3 98.2 4.18 300 1,490 1,376 1.083
C1-2 101.5 100.6 4.18 300 1,535 1,413 1.086
C2-1 101.2 101.1 4.18 300 1,740 1,447 1.202
C2-2 100.7 100.4 4.18 300 1,775 1,434 1.238
C3 182.8 181.2 4.18 540 3,590 3,468 1.035
Downloaded from ascelibrary.org by Indian Institute of Tech Gandhinagar on 12/22/15. Copyright ASCE. For personal use only; all rights reserved.

C4 181.8 180.4 4.18 540 4,210 3,659 1.151

C5-1 120.7 80.1 4.18 360 1,450 1,375 1.055
C5-2 119.3 80.6 4.18 360 1,425 1,368 1.042
C6-1 119.6 80.6 4.18 360 1,560 1,397 1.117
C6-2 120.5 80.6 4.18 360 1,700 1,405 1,210
C7-1 179.7 121.5 4.18 540 2,530 2,525 1.002
C8-1 180.4 119.8 4.18 540 2,970 2,626 1.131
C8-2 179.2 121.3 4.18 540 2,590 2,636 0.983
C9-1 160.2 81.4 4.18 480 1,710 1,747 0.979
C9-2 160.7 80.5 4.18 480 1,820 1,739 1.047
C10-1 160.1 81.0 4.18 480 1,880 1,786 1.053
C10-2 160.6 80.1 4.18 480 2,100 1,777 1.182
C11-1 199.8 101.2 4.18 600 2,350 2,426 0.969
C11-2 200.2 98.9 4.18 600 2,380 2,393 0.995
C12-1 199.2 102.1 4.18 600 2,900 2,538 1.143
C12-2 199.8 99.6 4.18 600 2,800 2,500 1.120
Average: 1.087
Standard deviation: 0.081

Table 5. Comparison with Test Results by Wang 共1999兲

Major axis moment Minor axis moment
Test Analysis
Specimen Eccentricity Moment Eccentricity Moment strength strength Test
designation 共mm兲 ratio 共mm兲 ratio 共kN兲 共kN兲 analysis
RHS-1 0 — 55 −1 368 428 0.86
RHS-2 0 — 55 0 246 236 1.04
RHS-3 55 0 110 0 172 172 1.00
RHS-4 55 0 110 −1 238 231 1.03
RHS-5 55 −1 110 −1 251 236 1.06
RHS-6 55 0 55 0 234 223 1.05
RHS-7 55 −1 0 — 520 555 0.94
RHS-8 55 0 0 — 480 385 1.25
RSJ-1 55 0 0 — 960 957 1.00
RSJ-2 0 — 55 −1 816 868 0.94
RSJ-3 0 — 55 0 570 626 0.91
RSJ-4 55 0 55 0 427 350 1.22
RSJ-5 55 0 110 −1 380 374 1.02
RSJ-6 55 −1 110 −1 330 258 1.28
RSJ-7 55 0 110 0 257 231 1.11
Average: 1.05
Standard deviation: 0.12
Note: Negative ratio between end extremities moments indicates reverse curvature. RHS⫽concrete filled rectangular hollow section; and RSJ⫽concrete
encased I section.
1728 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2005

J. Struct. Eng., 2005, 131(11): 1721-1730

 crete were 60.8 and 72.1 MPa. Four elements with four integra- f cylinder , f cyl ⫽ concrete cylinder strength;
tion points were used in the numerical analysis. Table 4 displays fy ⫽ steel yield strength;
the comparison between the tests and the numerical results. Once f ys ⫽ reinforcement steel yield strength;
again there is excellent agreement between the experimental and fm ⫽ element internal force;
numerical results. kT ⫽ element tangent stiffness matrix;
ky ⫽ curvature about y;
kz ⫽ curvature about z;
Slender Composite Columns under Biaxial Bending,
My ⫽ resultant bending moment about y;
Single and Reverse Curvature
Mz ⫽ resultant bending moment about z;
Wang 共1999兲 performed experiments with eight rectangular hol- Nx ⫽ resultant axial force in x direction;
low concrete-filled and seven I-section concrete-encased columns Nu ⫽ interpolation functions for displacement along
Downloaded from ascelibrary.org by Indian Institute of Tech Gandhinagar on 12/22/15. Copyright ASCE. For personal use only; all rights reserved.

under static loding. The specimens were subjected to eccentric x;

axial loading in both ends resulting in uniaxial and biaxial bend- Nv ⫽ interpolation functions for displacement along
ing with either single or reverse curvature. The cylinder strength y;
for concrete ranged from 25 to 55 MPa and yield stress from Nw ⫽ interpolation functions for displacement along
310 to 370 Mpa. Four elements with two Gauss points were em- z;
ployed. Table 5 summarizes the data for the experiments and the qu ⫽ displacements in x direction;
comparison of the ultimate loads obtained by the experimental qv ⫽ displacements in y direction;
program with those from the present formulation. In terms of qw ⫽ displacements in z direction;
displacements, some discrepancies are once more unavoidable as u ⫽ displacement along x;
in general geometrical imperfections are present in experimental V ⫽ volume;
results and these were not taken into account here. For instance, v ⫽ displacement along y;
the original work measured nonzero displacements at the mid- w ⫽ displacement along z;
height of reverse curvature specimens that the numerical analysis Wint ⫽ internal work;
is not able to detect. Wext ⫽ external work;
␧x ⫽ strain along x;
␧0 ⫽ axial strain;
Summary and Conclusions ␴x ⫽ normal stress along x;
␦ ⫽ indicates variation;
A FE formulation for the numerical analysis of composite steel- ␪ ⫽ rotation;
concrete slender columns of arbitrary cross section was devel- 0u ⫽
0 0 0T, null vector; and
oped, implemented, and tested. The section formulation permitted 0v , 0w ⫽
0 0 0 0T, null vector.
the simulation of behavior for all common types of composite
columns, differently from most of the published work that usually
focuses on a particular type of section. The results obtained by the References
numerical procedure agree well with other numerical simulations
and with experimental results. The model proved to be accurate American Concrete Institute 共ACI兲. 共2002兲. “Building code requirements
and reliable in terms of ultimate load capacity as well as displace- for structural concrete and commentary.” ACI-318, Detroit.
ment evaluation. The results obtained so far encourage one to American Institute of Steel Construction 共AISC-LRFD兲. 共1999兲. Load
develop critical analysis of simplified procedures from various and resistance factor design specification for steel buildings, with
code practices, especially with respect to the effects of slender- errata incorporated as of September 4, 2001.
ness on the displacement pattern and second order effects on col- Ayoub, A., and Filippou, F. C. 共2000兲. “Mixed formulation of nonlinear
umns with various loading and support conditions. The presented steel-concrete composite beam element.” J. Struct. Eng., 126共3兲,
formulation is restricted to isolated members with well-defined 371–381.
boundary conditions. An issue to be addressed further is whether Bonet, J. L., Romero, M. L., Miguel, P. F., and Fernandez, M. A. 共2004兲.
it is suitable for numerical analyses of composite framed struc- “A fast stress integration algorithm for reinforced concrete sections
with axial loads and biaxial bending.” Comput. Struct. 82, 213–225.
tures. It is the writers’ belief that, once issues such as nonlinear
British Standards Institution 共BSI兲. 共1979兲. “Steel, concrete and compos-
boundary conditions 共e.g., semirigid composite connections兲 have ite bridges: Part 5: Code of practice for design of composite bridges.”
been properly taken into account, the formulation will be able to BS 5400, London.
model composite framed systems properly. This work is currently Caldas, R. B. 共2004兲. “Numerical analysis of steel-concrete composite
in progress. columns.” MSc dissertation 共in Portuguese兲, Dept. of Civil Engineer-
ing, Escola de Minas, Universidade Federal de Ouro Preto, Ouro
Preto 共MG兲, Brazil.
Acknowledgment Chan, S. L. 共2001兲. “Review: Non-linear behaviour and design of steel
structures.” J. Constr. Steel Res., 57, 1217–1231.
The writers wish to thank USIMINAS S. A. for financial support Chen, S. F., Teng, J. G., and Chan, S. L. 共2001兲. “Design of biaxially
in the form of a grant for the second writer. loaded short composite columns of arbitrary section.” J. Struct. Eng.,
127共6兲, 678–685.
El-Tawil, S., and Deierlein, G. G. 共1999兲. “Strength and ductility of con-
crete encased composite columns.” J. Struct. Eng., 125共9兲, 1009–
Notation 1019.
El-Tawil, S., and Deierlein, G. G. 共2001a兲. “Nonlinear analysis of mixed
The following symbols are used in this paper: steel-concrete frames. I: Element formulation.” J. Struct. Eng.,
E ⫽ steel elasticity modulus; 127共6兲, 647–655.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2005 / 1729

J. Struct. Eng., 2005, 131(11): 1721-1730

 El-Tawil, S., and Deierlein, G. G. 共2001b兲. “Nonlinear analysis of mixed
steel-concrete frames. II: Implementation and verification.” J. Struct.
Eng., 127共6兲, 656–665.
El-Tawil, S., Sanz-Picon, C. F., and Deierlin, G. G. 共1995兲. “Evaluation
of ACI-318 and AISC 共LRFD兲 strength provisions for composite
columns.” J. Constr. Steel Res., 34共1兲, 103–126.
Eurocode 4-European Committee for Standardization. 共1994兲. “Design of
composite steel and concrete structures: Part 1.1: General rules and
rules for buildings.” DRAFT prEN 1994-1-1:2001.
Faella, C., Martinelli, E., and Nigro, E. 共2002兲. “Steel and concrete com-
posite beams with flexible shear connection: ‘Exact’ analytical expres-
sion of the stiffness matrix and applications.” Comput. Struct. 80,
Downloaded from ascelibrary.org by Indian Institute of Tech Gandhinagar on 12/22/15. Copyright ASCE. For personal use only; all rights reserved.
1001–1009.
Han, L.-H. 共2004兲. “Flexural behaviour of concrete-filled steel tubes.”
J. Constr. Steel Res., 60, 313–337.
Izzuddin, B. A., Siyam, A. A. F. M., Smith, D. L. 共2002兲. “An efficient
beam-column formulation for 3D reinforced concrete frames.”
Comput. Struct., 80共7兲, 659–676
Johansson, M., and Gylltoft, K. 共2002兲. “Mechanical behavior of circular
steel-concrete composite stub columns.” J. Struct. Eng., 128共8兲,
1073–1081.
Lakshmi, B., and Shanmugam, N. E. 共2002兲. “Nonlinear analysis of in-
filled steel-concrete composite columns.” J. Struct. Eng., 128共7兲,
922–933.
Liu, D., Gho, W. M., and Yuan, J. 共2003兲. “Ultimate capacity of
high-strength rectangular concrete-filled steel hollow section stub
columns.” J. Constr. Steel Res., 59, 1499–1515.
Martha, L. F., Carvalho, M. T. M., and Seixas, R. B. 共1997兲. “Volume
contouring of generic unstructured meshes.” J. Braz. Comput. Soc.,
3共3兲, 43–51.
Matsui, C., Tsuda, K., and Ishibashi, Y. 共1995兲. “Slender concrete-filled
steel tubular columns under combined compression and bending.”
Proc., 4th Pacific Structural Steel Conf., Vol. 3, 29–36.
Mirza, S. A., Hyttinen, V., and Hyttinen, E. 共1996兲. “Physical tests and
analyses of composite steel-concrete beam-columns.” J. Struct. Eng.,
122共11兲, 1317–1326.
Muñoz, P. R., and Hsu, C. T. T. 共1997a兲. “Behavior of biaxially loaded
concrete-encased composite columns.” J. Struct. Eng., 123共9兲,
1163–1171.
Muñoz, P. R., and Hsu, C. T. T. 共1997b兲. “Biaxially loaded concrete-
encased composite columns: Design equation.” J. Struct. Eng.,
123共12兲, 1576–1585.
Oehlers, D. J., and Bradford, M. 共1995兲. Composite steel and concrete
structural members: Fundamental behaviour, Pergamon, Elsevier,
New York.
1730 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / NOVEMBER 2005
J. Struct. Eng., 2005, 131(11): 1721-1730
Rangan, B. V., and Joyce, M. 共1992兲. “Strength of eccentrically loaded
slender steel tubular columns filled with high strength concrete.” ACI
Struct. J., 89共6兲, 676–681.
Rodriguez, J. A., and Aristizabal-Ochoa, J. D. 共1999兲. “Biaxial interac-
tion diagrams for short RC columns of any cross-section.” J. Struct.
Eng., 125共6兲, 672–683.
Roik, K., and Bergmann, R. 共1990兲. “Design method for composite col-
umns with unsymmetrical cross-sections.” J. Constr. Steel Res. 33,
153–172.
Rotter, J. M. 共1985兲. “Rapid exact inelastic biaxial bending analysis.”
J. Struct. Eng., 111共12兲, 2659–2674.
Saenz, L. P. 共1964兲. “Discussion of ‘Equation for the stress-strain curve
of concrete,’ by P. Desayi and S. Krishnan.” ACI J., 61共9兲,
1229–1235.
Salari, M. R., and Spacone, E. 共2001兲. “Finite element formulations of
one-dimensional elements with bond-slip.” Eng. Struct., 23, 815–826.
Saw, H. S., and Richard Liew, J. Y. 共2000兲. “Assessment of current meth-
ods for the design of composite columns in buildings.” J. Constr. Steel
Res., 53, 121–147.
Sfakianakis, M. G. 共2002兲. “Biaxial bending with axial force of rein-
forced, composite and repaired concrete sections of arbitrary shape by
fiber model and computer graphics.” Adv. Eng. Software, 33,
227–242.
Shanmugam, N. E., and Lakshmi, B. 共2001兲. “State of the art report on
steel-concrete composite columns.” J. Constr. Steel Res., 57,
1041–1080.
Spacone, E., and El-Tawil, S. 共2004兲. “Nonlinear analysis of steel-
concrete composite structures: State of the art.” J. Struct. Eng.,
130共2兲, 159–168.
Teh, L. H. 共2001兲. “Cubic beam elements in practical analysis and design
of steel frames.” Eng. Struct., 23, 1243–1255.
Varma, A. H., Ricles, J. M., Sause, R., and Lu, L.-W. 共2002兲. “Experi-
mental behavior of high strength square concrete-filled steel tube
beam-columns.” J. Struct. Eng., 128共3兲, 309–318.
Wang, Y. C. 共1999兲. “Tests on slender composite columns.” J. Constr.
Steel Res., 49, 25–41.
Weng, C. C., and Yen, S. I. 共2002兲. “Comparisons of concrete-encased
composite column strength provisions of ACI code and AISC speci-
fication.” Eng. Struct., 24, 59–72.
Werner, H. 共1974兲. “Schiefe biegung polygonal umrandeter Stahlbeton-
Querschnitte.” Beton-und Stahlbetonbau 69共4兲, 92–97.
Zhang, W., and Shahrooz, B. M. 共1999兲. “Comparison between ACI and
AISC for concrete-filled tubular columns.” J. Struct. Eng., 125共11兲,
1213–1223.