Engineering Structures 100 (2015) 410–424

Contents lists available at ScienceDirect

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

Behaviour and design of square concrete-filled double skin tubular

columns with inner circular tubes
M.F. Hassanein a,⇑, O.F. Kharoob a, L. Gardner b
a
Department of Structural Engineering, Faculty of Engineering, Tanta University, Tanta, Egypt
b
Department of Civil and Environmental Engineering, Imperial College London, London SW7 2AZ, United Kingdom

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

Article history: Concrete-filled double skin tubular (CFDST) columns consist of two concentric steel tubes, the void
Received 18 February 2015 between which is filled with concrete. For the current studied configuration, a square tube is used for
Revised 28 April 2015 the outer columns, which enables more straightforward fabrication and installation of
Accepted 11 June 2015
beam-to-column joints compared with the case of circular outer columns. To date there have been no sig-
nificant applications of square CFDST columns with inner circular tubes worldwide, partly due to the lack
of design provisions. This paper addresses the axial compressive performance of these columns.
Keywords:
Nonlinear finite element (FE) analyses are employed to conduct parametric studies, having first validated
Concrete-filled steel tubes
Finite element analysis
the models against available experimental data. The depth-to-thickness (D/t) ratios of the outer tubes
Ultimate axial strength were varied such that all cross-section classes were considered. Based on the generated numerical
Short columns results, a new design model is proposed to provide reliable predictions for the ultimate axial strength
Slender columns of short CFDST columns. The work is then extended to investigate the behaviour of slender CFDST col-
SHS outer and CHS inner umns. Comparisons of the generated FE ultimate loads for the slender columns are made with strength
predictions according to the provisions of Eurocode 4 (EC4) and the AISC 360 Specification for CFST col-
umns, but with allowance made for the inner tube. It is found that the AISC provisions generally give good
predictions of the strengths of the CFDST columns, while EC4 gives rather conservative predictions for the
columns with slender and very-slender cross-sections.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction lighter-weight composite columns has been increasing around the

world to reduce the foundation loads, particularly for high-rise
Columns are fundamental components of most structures and applications [3]. One solution was to use the concrete-filled double
hence, accurate prediction of their capacity is vital for overall skin steel tubular (CFDST) column. The application of CFDST col-
structural efficiency and reliability. Composite columns, in which umns, examples of which are shown in Fig. 1, was found to reduce
steel elements act compositely with the concrete to resist the com- the self-weight of the structure due to the hollow inner voids [3].
pressive force, are being increasingly used in modern construction. CFDST columns have been recognised to have a series of beneficial
There is a wide variety of types of composite column of varying properties, such as high strength and bending stiffness, good fire
cross-sections, but the most commonly studied and used are [4] and seismic performance [3,5–9]. It was also found that
encased I-sections and concrete-filled steel tubes. In contrast to CFDST columns have similar performance to traditional CFST col-
encased composite columns, concrete-filled steel tubular (CFST) umns of the same dimensions of outer steel tube and strength of
columns have the advantage that they do not need any formwork materials, but with less weight due to the void inside the inner
or reinforcement. CFST columns offer several advantages in terms tube [10].
of structural performance over bare steel, reinforced concrete or As can be seen in Fig. 1, CFDST columns consist of concentric
encased I-section columns. In addition, the use of CFST columns inner and outer steel tubes with infill concrete between them. By
in high-rise composite buildings leads to rapid construction, which using circular and square hollow sections (CHSs and SHSs, respec-
can bring significant economic benefits [1,2]. Recently, demand for tively), four column typologies can be generated. On the basis that
CHSs are less susceptible to local buckling than SHSs, it has been
recommended in a number of previous investigations [5–8] to
⇑ Corresponding author. Mobile: +20 1228898494; fax: +20 403315860.
use CHSs for both outer and inner tubes. On the other hand, the
E-mail addresses: [email protected], [email protected]
(M.F. Hassanein).
fabrication, installation and design [3,7] of beam-to-column joints

http://dx.doi.org/10.1016/j.engstruct.2015.06.022
0141-0296/Ó 2015 Elsevier Ltd. All rights reserved.
 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424 411

Nomenclature

Ac , Asc cross-sectional area of sandwiched concrete Pul;Prop proposed ultimate axial strengths for short square
Ac;nominal nominal cross-sectional area of concrete, given by CFDST columns with inner CHSs, either P ul;Prop;1 or
D2  Aso P ul;Prop;2
ADS cross-sectional area of CFDST column Ppl;Rd;mod modified plastic resistance of columns
Asi , Asi;g gross cross-sectional area of inner steel tubes ti thickness of inner tube of CFDST columns
Aso , Aso;g gross cross-sectional area of outer steel tubes to thickness of outer tube of CFDST columns
Aso;eff effective cross-sectional area of outer steel tubes um lateral deflection at mid-height section of CFDST col-
D diameter/depth of steel tube in circular/square CFDST umns
columns
Dc depth of concrete core Greek letters
d diameter of inner steel tube in CFDST columns a imperfection factor
Ec Young’s modulus of concrete bc factor reflecting confinement effect on concrete ductil-
Es Young’s modulus of steel ity
ðEIÞe effective elastic flexural stiffness of CFDST columns Bs larger depth of rectangular cross-sections
f ck characteristic concrete strength, given by 0:67f cu cc strength reduction factor [24]
f cu characteristic cube strength of concrete ea axial strain in CFDST column
0
fc cylindrical compressive strength of sandwiched con- ec longitudinal compressive concrete strain
crete e0cc 0
strain at f cc
0
f cc effective compressive strength of confined concrete el longitudinal strain
f syi yield strength of inner tubes e0c strain corresponding to peak stress of concrete (f c )
0

ft tensile strength of concrete k column slenderness ratio or local slenderness ratios fol-
f yo yield strength of outer tubes lowing the AISC Specification [14]
IDS moment of inertia of CFDST section kp and kr limiting local slenderness ratios according to the AISC
KL, Le effective buckling length Specification [14]
L physical column length k global column slenderness (relative slenderness)
P axial load v hollow ratio of CFDST column or buckling reduction fac-
PAISC Ultimate axial capacity according to AISC [14] tor in EC4 [13]
PAISC;mod modified AISC formula for ultimate axial capacity f confinement factor, given by (Aso f so =Ac;nominal f ck )
Pcr , Pe elastic critical buckling load qs ratio of cross-sectional area of steel tube to that of con-
PEC4 ultimate axial capacity according to EC4 [13] crete core
PEC4;mod modified EC4 formula for ultimate axial capacity rc longitudinal compressive concrete stress
Pi;u capacity of inner tube according to Han et al. [11] q reduction factor for local plate buckling according to
Posc;u compressive capacity of outer tube with sandwiched EC3 [28]
concrete
Ppl;Rd plastic resistance to axial compression as specified in Abbreviations
EC4 [13] CFDST concrete-filled double skin tubular column
Ppl;Rd;mod modified plastic resistance to axial compression
CFST concrete-filled steel tubular column
Pul;design predicted strengths from three design models (Pul;Han , CHS circular hollow section
P ul;Prop;1 and Pul;Prop;2 ) F fully-effective cross-section
Pul;Exp ultimate axial strength of column obtained from exper- HSC high strength concrete
iments
NSC normal strength concrete
Pul;FE ultimate axial strength of column obtained from FE S slender cross-section
analyses SHS square hollow section
Pul;Han strength of short square CFDST columns with inner
UHSC ultra-high strength concrete
CHSs, according to Han et al. [11] VS very slender cross-section

for square columns are much easier compared with those of circu- behaviour of slender square CFDST columns. Hence, the focus of
lar columns; this led to the examination of square CFDST columns the present paper is firstly, on the strength of short square CFDST
by Huang et al. [7] and Han et al. [11] to unveil their general struc- columns with inner CHSs, and then on the behaviour and strength
tural behaviour. To investigate the cross-sectional behaviour of of slender square CFDST columns. This study is performed by
square CFDST columns with inner CHSs, twelve tests were carried means of the general purpose finite element (FE) package
out by Han et al. [11]. Latter, Huang et al. [7] reported a finite ABAQUS [12]. A key aspect of the research is to assess the applica-
element (FE) study to support these experimental results. Test data bility of existing design rules for CFST columns to CFDST columns.
in the literature on slender square CFDST columns with inner CHSs Comparisons are made against the provisions of Eurocode 4 [13]
were found to be rather limited, with only four column test results and AISC [14], though other design codes for CFST columns are also
identified by the authors. Two of the column tests (scbc4-1 and available, such as CIDECT [15] and Chinese code [16].
scbc4-2) were reported by Han et al. [11], while the other two
(DSc-1 and DSc-2) were described by Han et al. [9]. Clearly, the 2. Strength of short CFDST columns
behaviour and strength of these slender columns has yet to be fully
explored. Given the limited existing data, there remain a number of 2.1. Existing design approach and test results
different areas still to be investigated. Among these are the effect
of the depth-to-thickness (D/t) ratios of the SHSs on the strength Proposals to predict the strength of short square CFDST columns
of short CFDST columns under applied axial compression and the (i.e. the cross-section resistance) with inner CHSs (Pul,Han) have
412 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424

Sandwiched Sandwiched
concrete concrete
to to

ti
d
ti

D D

(a) Circular CFDST: CHS (b) Circular CFDST: SHS

inner and CHS outer inner and CHS outer

Sandwiched to
concrete
ti
to d
ti

Sandwiched
d concrete

D D

(c) Square CFDST: SHS (d) Square CFDST: CHS

inner and SHS outer inner and SHS outer
Fig. 1. Types of CFDST column cross-sections.

been made by Han et al. [11]. The predicted strength, Pul,Han, is fyo is the yield strength of the outer SHS in MPa, f is the confinement
 
given as follows: factor A Aso f so f , and Ac,nominal is the nominal cross-sectional area
c;nominal ck

Pul;Han ¼ P osc;u þ Pi;u ð1Þ of the concrete, given by D2  Aso.

The accuracy of the above design model was assessed herein
where Posc,u is the compressive capacity of the outer tube with the
through comparison with the experimental results of the CFDST
sandwiched concrete and Pi,u is the capacity of the inner tube com-
columns tested by Han et al. [11]. Table 1 shows the details and
puted as (Asifsyi), where Asi and fsyi are the cross-sectional area and
material properties of these columns, while Table 2 provides the
the yield strength of the inner CHSs, respectively. To determine the
comparison between the design strengths of the columns
capacity Posc,u, the following equation was put forward:
(Pul,Han) and their ultimate experimental strengths (Pul,Exp). The
Posc;u ¼ f scy Asco with Asco ¼ Aso þ Asc ð2Þ outer SHSs were, firstly, classified (based on the maximum
depth-to-thickness ratios for the compression parts of the
in which Asc and Aso are the cross-sectional areas of the sandwiched cross-sections according to EC3 [17]) into fully-effective (F:
concrete and the outer steel tube, respectively. The strength fscy (D  2to)/to 6 42e) and slender (S: (D  2to)/to > 42e) sections;
defined, in MPa, was given as: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
e ¼ 235=f yo . The slenderness ((D  2to)/to) of some of the SHSs
     
f yo f tested by Han et al. [11] was found to exceed the maximum value
f scy ¼ 1:212 þ 0:138 þ 0:7646 f þ 0:0727 ck þ 0:0216 f2 f ck
235 20 of 52e specified in EC4 [13]; these cross-sections were considered
ð3Þ to be very slender (VS). From the comparative results of Table 2,
it can be seen that the strength predictions are good on average,
where fck is the characteristic concrete strength in MPa (0.67fcu),
but the predicted strengths are conservative in the case of CFDST
where fcu is the characteristic cube strength of concrete in MPa,

Table 1
Details of the tested short square CFDST columns [11].
0
Column Classification of the SHSs D  to (mm) d  ti (mm) L (mm) fyo (MPa) fyi (MPa) f c (MPa)

ssc2 F 120  3 32  3 360 275.9 422.3 37.4

ssc3 F 120  3 58  3 360 275.9 374.5 37.4
ssc4 F 120  3 88  3 360 275.9 370.2 37.4
ssc5 VS 180  3 88  3 540 275.9 370.2 37.4
ssc6 VS 240  3 114  3 720 275.9 294.5 37.4
ssc7 VS 300  3 165  3 900 275.9 320.5 37.4
 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424 413

Table 2 A quarter of the short square CFDST columns were modelled,

Comparison between experimental and strength predictions of short square CFDST exploiting the symmetry of geometry and loading, as shown in
columns [11].
Fig. 2, in which the end plates at the loaded end has been removed
Column Category of SHS Pul,Exp (kN) Pul,Han (kN) Pul;Han
P ul;Exp
to present the cross-section. Owing to the thin-walled nature of
both the inner and outer steel tubes, shell elements were employed
scc2-1 F 1054 972 0.92
scc2-2 F 1060 972 0.92
to discretise them. Element S3 [12] which is a three-noded triangu-
scc3-1 F 990 935 0.94 lar general-purpose shell element with finite membrane strains
scc3-2 F 1000 935 0.93 was utilised. For the sandwiched concrete and the two end plates,
scc4-1 F 870 821 0.94 three dimensional four-node linear tetrahedral solid elements,
scc4-2 F 996 821 0.82
denoted C3D4 [12], were employed.
scc5-1 VS 1725 1740 1.01
scc5-2 VS 1710 1740 1.02 To simulate the bond between the steel tubes and the sand-
scc6-1 VS 2580 2710 1.05 wiched concrete, a surface-based interaction with a contact
scc6-2 VS 2460 2710 1.10 pressure-overclosure model in the normal direction, and a
scc7-1 VS 3240 3794 1.17 Coulomb friction model with a coefficient of friction of 0.25 in
scc7-2 VS 3430 3794 1.11
the longitudinal direction was employed. Based on a previous con-
Mean 1.00
vergence study, a mesh size of approximately 25 mm was used for
Standard deviation 0.099 modelling the steel tubes and the sandwiched concrete. The CFDST
columns had fixed ends, with only longitudinal displacement at the
loaded end allowed. Additional surface-based interaction was
with fully-effective SHSs and on the unsafe side for CFDST with defined between each end plate and the sandwiched concrete,
very slender SHSs. using the ‘‘No adjustment’’ option in ABAQUS. Each end plate
was connected to both tubes using ‘‘Shell-to-Solid coupling’’ ensur-
2.2. Numerical modelling ing that the displacements and rotations of the connected elements
remained equal. A uniform distributed load was applied statically
In this sub-section, nonlinear numerical simulations, based on at the top of the upper end plate using displacement control. The
the FE method and using the software package ABAQUS/Standard load was applied incrementally using the modified Riks method
[12], are made to expand the available results on short square [12]. The nonlinear geometry parameter (⁄NLGEOM) was included
CFDST columns. The FE model and its validation are first described. to allow for changes in geometry under load. The steel material
Based on the FE results, a design model providing suitable capacity was assumed to behave in an elastic-perfectly plastic manner.
predictions for short square CFDST columns with inner CHSs, con-
sidering different SHS classes as well as normal, high and 2.2.2. Material model for sandwiched concrete
ultra-high strength concrete, is proposed in Section 3.3. Experiments indicate that the confinement effect provided by
square steel tube increases the ductility of the concrete core in a
2.2.1. General description of FE model short CFST column but not its ultimate strength for columns with
A summary of the FE model is presented in this section since full depth-to-thickness ratios greater than about 30 [20]. Fig. 3 shows
details were provided in previous publications by the authors, such the general stress–strain curve used in the present FE analyses to
as [8], in which the behaviour of short circular CFDST columns with simulate the material behaviour of the confined concrete in square
inner CHSs is examined. The key difference between the two mod- CFDST columns. The part OA of the stress-strain curve is repre-
els is the constitutive model of the sandwiched concrete; this topic sented using the equations suggested by Mander et al. [21] as:
will therefore be presented in more detail. It is worth noting that 0
f cc kðec =e0cc Þ
local initial imperfections were not considered in the current FE rc ¼ ð4Þ
modelling of the short CFDST columns. This is because the strength k  1:0 þ ðec =e0cc Þk
reduction of the thin-walled hollow tubes is not significant owing
where
to the delaying effect of concrete core on the tube buckling as pre-
viously discussed by Tao et al. [18] who concluded that initial Ec
k¼ 0 ð5Þ
imperfections decreases the strength of similar sections by about Ec  ðf cc =e0cc Þ
1% on average. Furthermore, local buckling is not precluded due
to the effective imperfection generated by Poisson expansion adja- rc is the longitudinal compressive concrete stress, f 0cc is the effective
cent to the restrained ends, as described by Theofanous et al. [19]. compressive strength of the confined concrete, ec is the longitudinal

End plate
Outer SHS
Sandwiched
Plane of symmetry concrete

Plane of symmetry
Inner CHS

Fig. 2. FE mesh for short square CFDST columns with inner CHS.
414 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424

σc accounts for the effect of the D/t ratio of the steel tube on the soft-
A B ening of the concrete. The concrete ultimate strain ecu is taken as
f cc′ 0.03.

2.2.3. Validation of the FE model

C D To assess the accuracy of the generated FE models, the tests
β c f cc′ conducted by Han et al. [11] were simulated herein. Details of
the tests were given previously in Table 1. Note that Han et al.
[11] conducted repeated tests for each of the examined columns,
denoted Nos. 1 and 2 specimens. The ultimate axial loads of the
CFDST columns obtained from the FE analyses are compared with
O ε' cc 0.005 0.015 0.03 εc
test data in Table 3. As can be seen, the FE model yields good pre-
Fig. 3. Stress–strain curve for confined concrete in square CFDST columns utilised dictions of the ultimate loads of the short CFDST columns. The
in the present study. mean value of Pul,FE/Pul,Exp is 0.99 with a standard deviation of
0.084 for No. 1 specimens. For the No. 2 specimens, Pul,FE/Pul,Exp is
0.97 with a standard deviation of 0.080. Fig. 4 shows a comparison
0
compressive concrete strain, e0cc is the strain at f cc given by Eq. (7) between the experimental and FE load–axial strain curves for the
and Ec is the Young’s modulus of concrete, which is given by ACI CFDST columns. It should be noted that the strain is the average
[22] as: axial strain, which is computed by dividing the end shortening
qffiffiffiffiffiffi by the column length. End shortening is inherently more variable
0
Ec ¼ 3320 f cc þ 6900 ðMPaÞ ð6Þ and hence harder to predict accurately than ultimate strength
due to the flat nature of the load-deformation response near the
8
> 0:002
0
for f cc 6 28 ðMPaÞ failure point. Overall, it can be observed that the FE model predicts
>
< 0
0 well the complete axial load-strain curves for these tested speci-
e0cc ¼ 0:002 þ f54;000
cc 28
for 28 < f cc 6 82 ðMPaÞ ð7Þ mens, except that of specimen ssc2. The unconservative results
>
>
: 0
for specimen ssc2 may be due to the compressive strength of sand-
0:003 for f cc > 82 ðMPaÞ
wiched concrete being lower than that given in Table 1, or may be
Parts AB, BC and CD of the stress–strain curve for confined con- due to voids that could have been present in the concrete. Fig. 5
crete depicted in Fig. 3 are based on the model given by Tomii and presents a typical FE deformed shape for the short CFDST columns,
Sakino [23] and are defined as: which is similar to that observed experimentally. As can be seen,
8 0
for : e0cc < ec 6 0:005 the failure mode of the outer tube involved outward local buckling,
< f cc
>
0 0 0 involving separation of the tube from the concrete core.
rc ¼ bc f cc þ 100ð0:015  ec Þðf cc  bc f cc Þ for : 0:005 < ec 6 0:015
>
: 0
bc f cc for : ec > 0:015
2.2.4. Parametric study and proposed design equation
ð8Þ The validated FE model was employed to generate parametric
where bc reflects the confinement effect on the concrete ductility results for short square CFDST columns. The length of the columns
and it depends on the width-to-thickness ratio (Bs/t) of the CFST col- (L) was taken as three times the external depth (D) of the outer
umn section, where Bs is taken as the larger depth of the rectangular SHSs to avoid global buckling effects. Table 4 shows the dimen-
cross-section. Based on the experimental results presented by Tomii sions and material properties considered in this study. SHSs with
and Sakino [23], bc is proposed as: a range of local slendernesses (i.e. fully-effective, slender and very
8 slender cross-sections) filled with normal, high and ultra-high
< 1:0
> for : Bts 6 24 strength concrete were simulated. The parameters considered in
0
1
bc ¼ 1:5  48 Bs
for : 24 < Bts 6 48 ð9Þ the analyses include the cylinder strength ðf c Þ of the concrete
> t
: Bs
0
and the ratio of the tube dimensions (d/D). The values of f c ranged
0:5 for : t
> 48
from 25 MPa to 120 MPa, covering normal strength (NSC), high
0
The effective compressive strength of concrete ðf cc Þ is influ- strength (HSC) and ultra-high strength concrete (UHSC) according
enced mainly by the column size, the quality of concrete and the to EC4 [13]. The current concrete model is considered to be appli-
rate of loading, as found previously by Liang [24]. Hence, the value cable to high strength concrete since it was initially proposed [24]
0 0 0 0
of f cc was determined as f cc ¼ cc f c , where f c is the compressive for normal and high strength concrete confined by either a normal
cylinder strength of the concrete and cc is the strength reduction or high strength steel tubes. The yield strength of the steel of both
factor [24], taken as: tubes was constant at 355 MPa. The columns were placed into

cc ¼ 1:85D0:135
c ð0:85 6 cc 6 1:0Þ ð10Þ
Table 3
where Dc is the depth of the concrete core. It is worth noting that FE and experimental ultimate loads for short square CFDST columns [11].
this strength reduction factor (cc) was proposed [24] to account Column Pul,FE (kN) Pul;FE
for the effects of the column size, the quality of concrete and the Pul;Exp

loading rate on the concrete compressive strength. No. 1 specimens No. 2 specimens
The Drucker–Prager yield criterion was adopted for the concrete ssc2 1189 1.13 1.12
to define the extent of the elastic response and the hardening ssc3 946 0.96 0.95
behaviour under a triaxial stress state. The material angle of fric- ssc4 927 1.07 0.93
ssc5 1614 0.94 0.94
tion (b) and the ratio of the flow stress in tension to that in com-
ssc6 2391 0.93 0.97
pression (K) were taken as 20° and 0.8, respectively, as suggested ssc7 3050 0.94 0.89
by Hu et al. [20]. The softening behaviour of the concrete in the
Mean 0.99 0.97
post-yield range is determined by the parameter bc and the con-
Standard deviation 0.084 0.080
crete ultimate strain ecu. The parameter bc, given in Eq. (9),
 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424 415

P [kN]

P [kN]
Strain [ με ] Strain [ με ]
(a) scc2 (b) scc3
P [kN]

P [kN]
Strain [ με ] Strain [ με ]
(c) scc4 (d) scc5
P [kN]

P [kN]

Strain [ με ] Strain [ με ]
(e) scc6 (f) scc7
Fig. 4. Comparisons of experimental and FE axial load–strain curves for short square CFDST columns with inner CHSs [11].

three groups (G1, G2 and G3) based on their d/D ratios, as shown in based on Eq. (11), considering the gross (Aso,g) and the effective
Table 4. The d/D ratio was 0.5 in Group 1 (G1), in Group 2 (0.2) and (Aso,eff) cross-sectional areas of the outer steel tubes, respectively.
0.8 in Group 3 (G3). In each group, different dimensions for the The effective cross-sectional areas of the outer steel tubes were
inner CHSs were employed to maintain the d/ti ratios. calculated following the rules provided in Clause 4.4 of EN
Design models for determining the ultimate strengths of short 1993-1-5 [28], where the effective area of the compression zone
CFDST columns have been proposed by Hassanein et al. [25] and of a plate is reduced by a reduction factor for plate buckling (q).
Hassanein and Kharoob [8]. Additional models for concrete-filled For the current square columns Aso,eff = qAso,g, where q is given
stainless steel tubular (CFSST) short columns and the by Eq. 4.2 of [28]. Fig. 6 shows the relationship between the rela-
0
concrete-filled stainless steel–carbon steel tubular (CFSCT) short tive design strengths for varying strengths ðf c Þ of the sandwiched
columns under axial compression have been developed by concrete.
Hassanein et al. [26,27], respectively. Based on these design models From Table 5 and Fig. 6, it can be observed that the strength pre-
[8,25–27], a new design model for calculating the ultimate axial dictions of Han et al. [11] (Pul,Han) are generally on the unsafe side
strengths of axially loaded short square CFDST columns with inner when compared to the FE results except for those columns with
CHSs is proposed as: d/D = 0.8. Using Eq. (11) and considering the gross cross-sectional
0 areas (Aso,g) of the outer tubes, the predicted strengths (Pul,Prop,1)
Pul;Prop ¼ f yo Aso þ cc f c Asc þ f yi Asi ð11Þ
of the columns are close to the FE results in the case of fully effec-
It should be noted that above design model is based on the plas- tive outer tubes. However, the predicted strengths for the other
tic resistance of square or rectangular composite columns (Ppl,Rd), cases (i.e. columns with S and VS outer tubes) are on the unsafe
as specified in the EC4 [13], but with a modified contribution from side. As can be seen in the table, the average value of
the sandwiched concrete. In Pul,Prop, the contribution of the sand- Pul,Prop,2/Pul,FE is 1.02 compared to 1.11 and 1.08 for Pul,Han/Pul,FE
wiched concrete is reduced by a strength reduction factor (cc) [11] and Pul,Prop,1/Pul,FE, respectively. Despite the mean value of
which is a function of the column size, as given previously in Eq. Pul,Prop,2/Pul,FE being slightly greater than unity, this is considered
(10). A comparison of the FE strengths generated in the parametric to be the best design model among those examined herein, with
study and predicted strengths (Pul,design) from three design models by far the lowest scatter.
are shown in Table 5. The first is that proposed by Han et al. [11], Eq. (11) was also assessed against the experimental results [11],
while the other two design models (Pul,Prop,1 and Pul,Prop,2) were as shown in Table 6. The average predictions of Pul,Han [11], Pul,Prop,1
416 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424

conditions were assigned to these reference points, while the load

Applied load was uniformly distributed over the upper end plate. To simulate
direction the pin-ended supports, the restraints shown in the same figure
were applied to the reference points. A mesh similar to that used
for the short columns was utilised for the slender columns, and
the bond between each steel tube and the sandwiched concrete
was also simulated in the same manner. Each end plate was con-
nected to both tubes using the ‘‘Tie’’ constraint [12] to ensure that
the displacements and rotations of the connected elements were
equal.
The principal difference in the modelling of the square CFDST
columns compared to those of the circular CFST, CFSST and
CFDST columns is the definition of the constitutive model of the
sandwiched concrete. In the latter case (i.e. circular outer col-
umns), the increase in the compressive strength of the infill con-
crete due to the confinement effects should be included.
However, for square outer columns, the confinement is substan-
Local buckling tially reduced, and so the concrete constitutive model given in
Section 2.2.2 was used for the slender square CFDST columns with
inner CHSs studied herein. The ‘‘Concrete Damaged Plasticity’’
model was applied to describe the behaviour of the sandwiched
concrete. To represent the inelastic behaviour of concrete, this
model uses the concept of isotropic damaged elasticity, in combi-
nation with isotropic tensile and compressive plasticity. The model
assumes that the uniaxial tensile and compressive responses of
concrete are characterized by damaged plasticity. The plasticity
parameters required by the ‘‘Concrete Damaged Plasticity’’ model
Fig. 5. Deformed shape, showing local buckling, of a typical short square CFDST
[12] are the dilation angle, the eccentricity, the ratio of the strength
column.
in the triaxial state to that in the uniaxial state and the K parame-
ter. They were taken, respectively, as 20°, 0.1, 1.16 and 0.667
and Pul,Prop,2 are 1.00, 0.99 and 0.93, respectively. Although the pre- [29,32]. The tensile strength of concrete was taken as
qffiffiffiffiffiffiffiffiffi
dictions of Han et al. [11] are the most accurate on average, the 0
f t ¼ 0:6 cc f c ; cc is a strength reduction factor given in Eq. (10)
proposed approaches have considerably lower scatter, and provide 0
and f c is the unconfined compressive concrete strength. The con-
generally safe side predictions. Strength predictions using Pul,Prop,2
crete tensile softening behaviour was additionally considered.
are therefore recommended.
The ultimate tensile strain of the concrete considering the tensile
softening model is 10 times the strain at cracking [33].
3. Behaviour and strength of slender CFDST columns
3.3. Validation of the FE model of slender square CFDST columns
In this section, the behaviour and strength of slender square
CFDST columns are investigated numerically, following which For the case of slender square CFDST columns with inner CHSs,
design recommendations are made. validation was made herein against specimens scbc4-1 and
scbc4-2 tested by Han et al. [11]. Specimens scbc4-1 and scbc4-2
3.1. Previous FE investigations on slender columns are identical, with dimensions D  to and d  ti (in mm) of
120  3 and 58  3, respectively. The values of fyo and fyi were
The FE method was successfully employed by the authors in a 275.9 and 374.5 MPa, respectively. The test and FE
previous study [29] to investigate the behaviour of slender circular load-mid-height deflection curves for these columns are presented
CFDST columns under axial compression, accurately capturing the in Fig. 9. From the figure, it can be seen that the predicted initial
composite action, concrete damage, as well as local and global stiffness and the post-peak softening behaviour of the columns
instabilities. The model was validated against experiments on slen- are, generally, in good agreement with the experimental results.
der CFST [30], CFSST (concrete-filled stainless steel tubular col- The difference between the FE and test strengths of the columns
umns) [31] and CFDST columns [5]. Fig. 7 shows comparisons is likely to be due to the uncertainty of the actual concrete strength
between the test and FE load-mid-height lateral deflection curves and stiffness. Fig. 10 presents the FE deformed shape for the col-
for three samples of slender circular CFST, CFSST and CFDST col- umn scbc4, which is similar to that observed experimentally.
umns; additional results are also given in Ref. [29]. From the figure, Hence, it can be concluded that this validation, in addition to that
it can be seen that the FE model simulates well the general beha- for the other slender column types, shows that the current FE
viour of the slender circular concrete-filled columns in terms of model is able to capture accurately the behaviour of slender square
the initial stiffness, strength and the post-peak softening behaviour. CFDST columns.

3.2. Description of FE model of slender square CFDST columns 3.4. Parametric study

Fig. 8 shows a typical FE model for the slender square CFDST Three basic cross-sections – C2, C8 and C14 from Table 4 – were
columns. The model contains upper and lower rigid end plates, used throughout the parametric study. This covered fully-effective,
though the lower end plate has been removed in Fig. 8 to aid visu- slender and very slender SHSs. Variation in column length was
alisation of the cross-section of the column. At the centre of each used to generate a range of member slendernesses. The initial geo-
end plate a reference point (RP) was defined. The boundary metric imperfections were of the form of the lowest global
 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424 417

Table 4
Details of FE models of short square CFDST columns.
0 ti
Group Column reference Outer tube Inner tube L (mm) f c (MPa) d
D te

D (mm) te (mm) D d (mm) ti (mm) d

te ti

G1 C1 300 10.00 30 150 5.00 30 900 25 0.5 0.50

C2 300 10.00 30 150 5.00 30 900 40 0.5 0.50
C3 300 10.00 30 150 5.00 30 900 60 0.5 0.50
C4 300 10.00 30 150 5.00 30 900 80 0.5 0.50
C5 300 10.00 30 150 5.00 30 900 100 0.5 0.50
C6 300 10.00 30 150 5.00 30 900 120 0.5 0.50
C7 400 10.00 40 200 6.66 30 1200 25 0.5 0.66
C8 400 10.00 40 200 6.66 30 1200 40 0.5 0.66
C9 400 10.00 40 200 6.66 30 1200 60 0.5 0.66
C10 400 10.00 40 200 6.66 30 1200 80 0.5 0.66
C11 400 10.00 40 200 6.66 30 1200 100 0.5 0.66
C12 400 10.00 40 200 6.66 30 1200 120 0.5 0.66
C13 500 10.00 50 250 8.33 30 1500 25 0.5 0.83
C14 500 10.00 50 250 8.33 30 1500 40 0.5 0.83
C15 500 10.00 50 250 8.33 30 1500 60 0.5 0.83
C16 500 10.00 50 250 8.33 30 1500 80 0.5 0.83
C17 500 10.00 50 250 8.33 30 1500 100 0.5 0.83
C18 500 10.00 50 250 8.33 30 1500 120 0.5 0.83
G2 C19 300 10.00 30 60 5.00 12 900 25 0.2 0.50
C20 300 10.00 30 60 5.00 12 900 40 0.2 0.50
C21 300 10.00 30 60 5.00 12 900 60 0.2 0.50
C22 300 10.00 30 60 5.00 12 900 80 0.2 0.50
C23 300 10.00 30 60 5.00 12 900 100 0.2 0.50
C24 300 10.00 30 60 5.00 12 900 120 0.2 0.50
C25 400 10.00 40 80 6.66 12 1200 25 0.2 0.66
C26 400 10.00 40 80 6.66 12 1200 40 0.2 0.66
C27 400 10.00 40 80 6.66 12 1200 60 0.2 0.66
C28 400 10.00 40 80 6.66 12 1200 80 0.2 0.66
C29 400 10.00 40 80 6.66 12 1200 100 0.2 0.66
C30 400 10.00 40 80 6.66 12 1200 120 0.2 0.66
C31 500 10.00 50 100 8.33 12 1500 25 0.2 0.83
C32 500 10.00 50 100 8.33 12 1500 40 0.2 0.83
C33 500 10.00 50 100 8.33 12 1500 60 0.2 0.83
C34 500 10.00 50 100 8.33 12 1500 80 0.2 0.83
C35 500 10.00 50 100 8.33 12 1500 100 0.2 0.83
C36 500 10.00 50 100 8.33 12 1500 120 0.2 0.83
G3 C37 300 10.00 30 240 5.00 48 900 25 0.2 0.50
C38 300 10.00 30 240 5.00 48 900 40 0.8 0.50
C39 300 10.00 30 240 5.00 48 900 60 0.8 0.50
C40 300 10.00 30 240 5.00 48 900 80 0.8 0.50
C41 300 10.00 30 240 5.00 48 900 100 0.8 0.50
C42 300 10.00 30 240 5.00 48 900 120 0.8 0.50
C43 400 10.00 40 320 6.66 48 1200 25 0.8 0.66
C44 400 10.00 40 320 6.66 48 1200 40 0.8 0.66
C45 400 10.00 40 320 6.66 48 1200 60 0.8 0.66
C46 400 10.00 40 320 6.66 48 1200 80 0.8 0.66
C47 400 10.00 40 320 6.66 48 1200 100 0.8 0.66
C48 400 10.00 40 320 6.66 48 1200 120 0.8 0.66
C49 500 10.00 50 400 8.33 48 1500 25 0.8 0.83
C50 500 10.00 50 400 8.33 48 1500 40 0.8 0.83
C51 500 10.00 50 400 8.33 48 1500 60 0.8 0.83
C52 500 10.00 50 400 8.33 48 1500 80 0.8 0.83
C53 500 10.00 50 400 8.33 48 1500 100 0.8 0.83
C54 500 10.00 50 400 8.33 48 1500 120 0.8 0.83

buckling mode and of amplitude L/1000, following the Chinese G6, respectively. A constant concrete strength of 40 MPa was
code GB50017-2003 [16] and the recommendation of Han et al. employed throughout this parametric study.
[34]. The slenderness ratios (k) of the CFDST columns, defined in
Eq. (12), varied from 6 to 168, as can be seen in Table 7. 3.5. Structural behaviour

Le Fig. 11 shows the ultimate axial FE strength (Pul,FE) against the

k ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð12Þ
IDS =ADS slenderness ratios ðkÞ for the three generated groups. Within each
group, the anticipated trend of reducing resistance with increasing
where Le is the effective buckling length of the column, IDS is the slenderness may be seen.
moment of inertia of the CFDST section and ADS is the Axial load (P) versus longitudinal strain (el) curves, recorded on
cross-sectional area of the CFDST column. The column lengths were the concave and convex sides of the external and the internal tubes
increased in 2000 mm increments until the slenderness parameter k at the mid-height of columns C58 and C62 are shown in Fig. 12.
defined in EC4 [13] reached approximately 2.0. The columns of These columns belong to the group of columns which are charac-
cross-section C2, C8 and C14 were assigned to groups G4, G5 and terised by fully-effective outer steel sections. In the figure, the
418 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424

Table 5
FE strengths compared to different design models.

Group Column reference Pu,FE (kN) Pul,Han (kN) P ul;Han Pul,Prop,1 (kN) Pul,Prop,2 (kN) Pul;Prop;1 Pul;Prop;2
Pul;FE Pul;FE Pul;FE

G1 C1 6219 5821 0.94 6239 6239 1.00 1.00

C2 7002 6845 0.98 7027 7027 1.00 1.00
C3 8038 8271 1.03 8077 8077 1.00 1.00
C4 9075 9718 1.07 9127 9127 1.01 1.01
C5 10,110 11,173 1.11 10,177 10,177 1.01 1.01
C6 11,133 12,633 1.13 11,227 11,227 1.01 1.01
C7 8989 9005 1.00 9375 8689 1.04 0.97
C8 10,334 10,890 1.05 10,816 10,129 1.05 0.98
C9 11,686 13,460 1.15 12,736 12,050 1.09 1.03
C10 12,849 16,050 1.25 14,657 13,971 1.14 1.09
C11 15,296 18,648 1.22 16,578 15,891 1.08 1.04
C12 16,950 21,250 1.25 18,498 17,812 1.09 1.05
C13 11,764 12,771 1.09 13,056 11,299 1.11 0.96
C14 13,517 15,759 1.17 15,368 13,610 1.14 1.01
C15 16,291 19,798 1.22 18,450 16,693 1.13 1.02
C16 19,406 23,857 1.23 21,532 19,775 1.11 1.02
C17 22,443 27,924 1.24 24,615 22,857 1.10 1.02
C18 25,408 31,995 1.26 27,697 25,940 1.09 1.02
G2 C19 6027 6348 1.05 6058 6058 1.01 1.01
C20 6994 7582 1.08 7038 7038 1.01 1.01
C21 8302 9300 1.12 8345 8345 1.01 1.01
C22 9608 11,044 1.15 9652 9652 1.00 1.00
C23 10,913 12,798 1.17 10,959 10,959 1.00 1.00
C24 12,219 14,558 1.19 12,265 12,265 1.00 1.00
C25 8675 9667 1.11 9044 8358 1.04 0.96
C26 10,139 11,939 1.18 10,821 10,135 1.07 1.00
C27 12,266 15,036 1.23 13,191 12,504 1.08 1.02
C28 14,087 18,158 1.29 15,560 14,874 1.10 1.06
C29 16,842 21,289 1.26 17,929 17,243 1.06 1.02
C30 19,142 24,425 1.28 20,299 19,612 1.06 1.02
C31 11,132 13,537 1.22 12,539 10,781 1.13 0.97
C32 13,622 17,139 1.26 15,376 13,619 1.13 1.00
C33 17,139 22,007 1.28 19,159 17,402 1.12 1.02
C34 21,106 26,899 1.27 22,943 21,185 1.09 1.00
C35 24,986 31,801 1.27 26,726 24,969 1.07 1.00
C36 28,680 36,707 1.28 30,509 28,752 1.06 1.00
G3 C37 6141 4413 0.72 6145 6145 1.00 1.00
C38 6552 5046 0.77 6575 6575 1.00 1.00
C39 7099 5928 0.84 7149 7149 1.01 1.01
C40 7645 6824 0.89 7722 7722 1.01 1.01
C41 8019 7725 0.96 8295 8295 1.03 1.03
C42 8728 8628 0.99 8869 8869 1.02 1.02
C43 8723 7012 0.80 9225 8538 1.06 0.98
C44 9209 8178 0.89 10,041 9354 1.09 1.02
C45 10,150 97,68 0.96 11,128 10,442 1.10 1.03
C46 10,852 11,371 1.05 12,216 11,529 1.13 1.06
C47 11,736 12,979 1.11 13,303 12,617 1.13 1.08
C48 12,576 14,589 1.16 14,391 13,704 1.14 1.09
C49 11,377 10,152 0.89 12,822 11,065 1.13 0.97
C50 12,124 12,002 0.99 14,158 12,400 1.17 1.02
C51 12,934 14,501 1.12 15,938 14,181 1.23 1.10
C52 13,596 17,013 1.25 17,719 15,961 1.30 1.17
C53 15,967 19,530 1.22 19,499 17,742 1.22 1.11
C54 17,352 22,050 1.27 21,280 19,522 1.23 1.13
Mean 1.11 1.08 1.02
Standard deviation 0.149 0.069 0.041

negative and positive strains represent compressive and tensile cross-sectional types (F, S and VS) are shown in
strains, respectively. As can be seen, the entire sections of both Fig. 13(a) and (b). Fig. 13(a) shows the load-mid-height deflection
the external and internal tubes of column C58 were under com- curves for columns C58, C67 and C79 with k values about 70, while
pression at the ultimate load Pul,FE, with some variation in strains Fig. 13(b) presents the curves for the columns C60, C70 and C83
across the sections. Beyond the maximum load, strain reversal with k values of about 120. The loads have been normalised by
may be observed on the concave side of the outer tube. For column the modified plastic resistance of the columns (Ppl,Rd,mod) as deter-
C62, which is a more slender column, the strains on the two sides mined by Eq. (11), to allow direct comparisons between the col-
of the tubes began to diverge from the early stages of loading due umns in the different groups (i.e. G4, G5 and G6). Generally, for a
to lateral deflection of the columns, and are tensile in nature on the given slenderness, the columns with the different outer steel tubes
convex side of the outer tube at Pul,FE. (F, S or VS) exhibit similar behaviour. For the intermediate length
Normalised axial load-mid-height lateral deflection curves for columns with k values of around 70, the curves show an initially
three CFDST columns with similar k values but with different linear region before a sharp change in stiffness at Pul,FE. The longer
 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424 419

d /D = 0.2 d /D = 0.5 d /D = 0.8

Pul,design Pul,design Pul,design

Pul,FE Pul,FE Pul,FE

f c′ [MPa] f c′ [MPa] f c′ [MPa]

(a) Fully effective SHSs

d /D = 0.2 d /D = 0.5 d /D = 0.8

Pul,design Pul,design Pul,design

Pul,FE Pul,FE Pul,FE

f c′ [MPa] f c′ [MPa] f c′ [MPa]

(b) Slender SHSs

d /D = 0.2 d /D = 0.5 d /D = 0.8

Pul,design Pul,design Pul,design

Pul,FE Pul,FE Pul,FE

f c′ [MPa] f c′ [MPa] f c′ [MPa]

(c) Very slender SHSs
Fig. 6. Comparison between the FE strengths and design predictions for CFDST columns with (a) fully effective, (b) slender and (c) very slender SHSs.

columns, with k values of around 120, show more rounded beha- strength predictions for CFST columns based on the provisions of
viour from the initiation of the loading, owing to the more domi- EC4 [13] and the AISC Specification [14]. The material partial safety
nant second order effects. factors were set to unity for comparison purposes. The plastic
resistances (Ppl,Rd) of the cross-sections in both code predictions
3.6. Design strengths were modified herein by adding the contribution of the internal
tubes (Asifyi). Based upon the comparisons, suitable design recom-
In the absence of specific compressive design strength formulae mendations are made. It should be noted that AISC 360 specifies a
pffiffiffiffiffiffiffiffiffiffi
for slender square CFDST columns with inner CHSs, the ultimate FE maximum permitted width-to-thickness ratio ð5:00 E=F y Þ which
loads from the parametric study Pul,FE were compared with is much higher than that given in EC4 (52e). Hence, the studied
420 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424

Table 6 p2 ðEIÞe
Comparison between experimental results [11] and the proposed design strengths for Pcr ¼ ð15Þ
short square CFDST columns. ðKLÞ2

Column Category of Pul,Exp Pul,Han Pul;Han Pul;Prop;1 P ul;Prop;2 where KL is the effective length of the member and (EI)e is the effec-
Pul;Exp P ul;Exp Pul;Exp
SHS (kN) (kN) tive elastic flexural stiffness.
scc2-1 F 1054 972 0.92 0.90 0.90 The reduction factor (v) is calculated using the European strut
scc2-2 F 1060 972 0.92 0.89 0.89 curves as:
scc3-1 F 990 935 0.94 0.97 0.97 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
scc3-2 F 1000 935 0.93 0.96 0.96 v ¼ 1=ðu þ u2  k2 Þ 6 1:0 but v 6 1:0 ð16Þ
scc4-1 F 870 821 0.94 1.08 1.08
scc4-2 F 996 821 0.82 0.94 0.94
scc5-1 VS 1725 1740 1.01 1.00 0.90 u ¼ 0:5ð1 þ aðk  0:2Þ þ k2 Þ ð17Þ
scc5-2 VS 1710 1740 1.02 1.00 0.91
scc6-1 VS 2580 2710 1.05 1.00 0.86 with a = 0.34 (buckling curve (b)) for 3% < qs 6 6% which is the
scc6-2 VS 2460 2710 1.10 1.04 0.91 case for the current models where qs is the ratio of the
scc7-1 VS 3240 3794 1.17 1.10 0.94
cross-sectional area of the steel tube to that of the concrete core.
scc7-2 VS 3430 3794 1.11 1.04 0.89
The ultimate FE resistance Pul,FE of the modelled CFDST columns
Mean 1.00 0.99 0.93
were also compared with the strength predictions of EC4, but using
Standard deviation 0.099 0.065 0.056 the plastic resistance to axial compression (Ppl,Rd,mod) as proposed
by Eq. (11). This modified EC4 strength formula (PEC4,mod) for a
CFDST column is therefore given as:
columns with very slender outer tubes are permitted by the AISC
PEC4;mod ¼ vPpl;Rd;mod ð18Þ
Specification.

3.6.1. Eurocode 4 3.6.2. AISC Specification

The ultimate FE loads of the CFDST columns are compared with According to the AISC Specification [14], the unfactored ulti-
the unfactored design strengths of CFST columns predicted by mate axial strengths (PAISC) of rectangular CFST columns are given
Eurocode 4 (EC4) [13]. In the calculations, the buckling lengths of as:
the pin-ended columns were taken as equal to their physical 8 h P
i
< Pno 0:658 Pnoe : PPnoe 6 2:25
lengths. The EC4 formula for the ultimate axial capacity (PEC4) of
PAISC ¼ ð19Þ
a CFST column is given as: : 0:877P : PPnoe > 2:25
e

PEC4 ¼ vPpl;Rd ð13Þ

where
where Ppl,Rd is the plastic resistance to axial compression given by 8 0
>
> P ¼ f yo Aso;g þ 0:85f c Ac þ f yi Asi;g For compact sections
Eq. (14), in which the effective areas of the steel tubes are employed < p
P p Py 2
in the case of slender cross-sections. Pno ¼ Pp  kr kp ðk  kp Þ For non-compact sections
>
>
: 0
0
Ppl;Rd ¼ f yo Aso þ f c Ac þ f yi Asi ð14Þ f cr Aso;g þ 0:7f c Ac þ f yi Asi;g For slender sections
ð20Þ
The critical buckling load is calculated from:

CFST [29]
Axial load [kN]
Axial load [kN]

CFSST [30]

Deflection at mid-height (u m ) [mm] Deflection at mid-height (u m ) [mm]

Circular CFDST [5]

Axial load [kN]

Deflection at mid-height (u m) [mm]

Fig. 7. Comparisons of numerical and experimental axial load-mid-height deflection curves for slender columns of different types.
 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424 421

y Upper end plate

z x

Fig. 8. A typical FE model for slender CFDST column and boundary conditions.

1000

800
Axial load [kN]

600

400 scbc4-1: FE
scbc4-1: Exp
200
scbc4-2: Exp
0
0 10 20 30 40 50
Deflection at mid-height (um ) [mm]

Fig. 9. Comparisons of numerical and experimental axial load-mid-height deflec-

tion curves for slender square CFDST columns with inner CHSs [11].

in which
0
Py ¼ f yo Aso;g þ 0:7f c Ac þ f yi Asi;g ð21Þ

and

9Es
f cr ¼ ð22Þ
ðD=tÞ2

where k, kp and kr are the local slenderness ratios determined from

Table II.1A of the AISC Specification [14] and Pe in Eq. (19) is the
elastic critical buckling load determined from Eq. (15). Note that
the AISC Specification [14] does not use the effective area approach
in designing composite columns with slender cross-sections, but
instead, through the definition of Pno, uses an effective stress Fig. 10. FE deformed shape of scbc4.

approach, based on the local slenderness k.

The ultimate FE capacities Pul,FE of the CFDST columns were also
compared with a modified version of the AISC strength predictions
(PAISC,mod) that used the plastic resistance to axial compression given in Eq. (11). The third design strength was that predicted by
Ppl,Rd,mod, as given by Eq. (11), in place of Eq. (20). the AISC Specification [14] (PAISC), which uses an effective stress
approach (Eq. (20)). The fourth design strength (PAISC,mod) utilises
the AISC column curves [14], but the cross-section strength is
3.6.3. Comparisons and discussion
based on the modified plastic resistance (Ppl,Rd,mod) of the CFDST
In this sub-section, predicted capacities from the above four
column (Eq. (11)), and hence follows the effective width approach
design methods are compared with the FE strengths (Pul,FE). The
[28] for slender sections. Comparisons between the FE results and
first design strength was calculated according to EC4 (PEC4), as
predicted design strengths are shown in Fig. 14. As can be seen
explained in Section 3.6.1, utilising the effective width approach
from the figure, the PEC4, PEC4,mod and PAISC,mod design expressions
for the slender cross-sections. The second design strength
provide strength predictions on the safe side of the FE results.
(PEC4,mod) represents the strength of EC4 [13] but by using the mod-
The mean values of PEC4, PEC4,mod and PAISC,mod relative to the
ified plastic resistance (Ppl,Rd,mod) of the CFDST cross-section, as
422 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424

Table 7 P [kN]
Details and FE strengths Pul,FE of slender square CFDST columns.

Group Basic cross- Column L (m) k k Pul,FE

section reference (kN)
G4 C2 C55 1.00 11 0.13 6474
C56 3.00 32 0.39 5423
C57 5.00 53 0.65 4848
C58 7.00 75 0.92 4401 Concave side-Internal
C59 9.00 96 1.18 3485 Convex side-Internal
C60 11.00 117 1.44 2858 Concave side-External
C61 13.00 139 1.70 2239 Convex side-External
C62 15.00 160 1.96 1791
G5 C8 C63 1.00 8 0.10 10,489 με l
C64 3.00 24 0.30 10,270
C65 5.00 40 0.50 10,366
C66 7.00 56 0.70 8982
(a) C58: column – L=7.00m
C67 9.00 72 0.90 8034
C68 11.00 88 1.10 7609 P [kN]
C69 13.00 104 1.30 6592
C70 15.00 120 1.50 5513
C71 17.00 136 1.70 4430
C72 19.00 152 1.90 3597
C73 21.00 168 2.10 2893
G6 C14 C74 1.00 6 0.08 14,523
C75 3.00 19 0.24 14,070
C76 5.00 32 0.41 13,646
Concave side-Internal
C77 7.00 45 0.57 13,010
Convex side-Internal
C78 9.00 58 0.73 12,363
C79 11.00 70 0.90 11,673 Concave side-External
C80 13.00 83 1.06 10,980
Convex side-External
C81 15.00 96 1.22 9674 με l
C82 17.00 109 1.38 8133
C83 19.00 122 1.55 6902
C84 21.00 134 1.71 5527 (b) C62: column – L=15.00m
C85 23.00 147 1.87 4816
C86 25.00 160 2.03 4019 Fig. 12. Load–longitudinal strain relationships for (a) C58 and (b) C62 CFDST
columns.

16000 λ = 70
G6
14000
G5 λ = 72
12000
G4 P λ = 75
10000
Ppl, Rd, mod
Pul,FE 8000
[ kN] 6000
4000 u m [mm]
2000
λ
(a) Intermediate length columns
0
0 45 90 135 180
Fig. 11. FE ultimate strengths against slenderness ratios k.

λ = 120
P
Pul,FE are 0.88, 0.85 and 0.96, respectively, with the latter having λ = 122
Ppl,Rd, mod
slightly lower scatter. The PAISC design formulae provide a mean λ = 117
ratio of PAISC/Pul,FE = 1.00, but with a number of predictions on
the unsafe side, particularly for groups G5 and G6. The slight
difference between the PEC4 and PEC4,mod is attributed to the small
u m [mm]
decrease in the plastic resistance due to adopting the strength
reduction factor cc in the latter approach. The lower results
(b) Long columns
obtained from the AISC Specification when utilising the Fig. 13. Normalised load-mid-height (um) relationships for (a) intermediate length
modified plastic resistance (Ppl,Rd,mod) of the CFDST column and (b) long columns CFDST columns.
(Eq. (11)) is attributed to the use of effective cross-sectional prop-
erties for the G5 and G6 specimens instead of the gross areas used
by Ref. [14]. behaviour observed from the FE models. The most accurate predic-
Overall, the three design approaches – EC4, modified EC4 and tions come from the AISC Specification [14], though there are a
the AISC Specification – provide reasonable predictions of the number of results on the unsafe side. This is largely remedied in
the PAISC,mod proposals.
 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424 423

(a) (b)
Mean PEC4 /Pul,FE = 0.88 Mean PEC4,mod /Pul,FE = 0.85
STDEV=0.117 STDEV=0.102

PEC4 PEC4, mod

[ kN] [ kN]

Pul,FE [kN] Pul,FE [kN]

(c) (d)
Mean PAISC / Pul , FE = 1.00 Mean PAISC,mod /Pul,FE = 0.96
STDEV=0.090 STDEV=0.096

PAISC PAISC,mod
[ kN] [ kN]

Pul,FE [kN] Pul,FE [kN]

Fig. 14. Comparison between FE and design strengths from (a) EC4, (b) modified EC4, (c) AISC and (d) modified AISC.

4. Summary and conclusions were also made to reduce the number of predictions on the unsafe
side. Further research into slenderness limits for SHS with concrete
The axial compressive behaviour of short (i.e. cross-section infill and the influence of the inner-to-outer tube dimension ratios
response) and slender (i.e. member buckling response) of square in CFDST columns is recommended.
concrete-filled double-skin tubular (CFDST) columns with circular
inner tubes (CHSs) has been studied. Finite element (FE) models
were first developed and validated against existing experimental References
data. Following validation of the models, a series of parametric
studies in which the influence of the local slenderness of the outer [1] Johansson M. Composite action and confinement effects in tubular steel–
concrete columns, Ph.D. thesis, Chalmers University of Technology, Goteborg,
square tube and the strength of the concrete, was investigated. The
Sweden; 2002.
generated results were analysed and compared with the strength [2] Chitawadagi MV, Narasimhan MC, Kulkarni SM. Axial strength of circular
predictions of Han et al. [11], which were found to be on the unsafe concrete-filled steel tube columns – DOE approach. J Constr Steel Res
side in the case of slender outer tubes. A revised design model was 2010;66:1248–60.
[3] Han L-H, Wei L, Bjorhovde R. Developments and advanced applications of
proposed, based on the provisions of EC4 [13] but with a reduced concrete-filled steel tubular (CFST) structures: members. J Constr Steel Res
contribution from the sandwiched concrete to account for different 2014;100:211–28.
column sizes, and shown to yield improved predictions of [4] Lu H, Zhao X-L, Han L-H. Testing of self-consolidating concrete-filled double
skin tubular stub columns exposed to fire. J Constr Steel Res 2010;66:1069–80.
cross-section strength. [5] Tao Z, Han L-H, Zhao X-L. Behaviour of concrete-filled double skin (CHS inner
Attention was then given to the behaviour of slender square and CHS outer) steel tubular stub columns and beam-columns. J Constr Steel
CFDST columns. Again, the FE models were first validated against Res 2004;60:1129–58.
[6] Uenaka K, Kitoh H, Sonoda K. Concrete filled double skin circular stub columns
a series of existing test results and shown to be capable of accu- under compression. Thin-Walled Struct 2010;48:19–24.
rately capturing the load-deformation response and failure modes [7] Huang H, Han L-H, Tao Z, Zhao X-L. Analytical behaviour of concrete-filled
observed in the experiments. Parametric studies were subse- double skin steel tubular (CFDST) stub columns. J Constr Steel Res
2010;66:542–55.
quently performed to examine the influence of local slenderness [8] Hassanein MF, Kharoob OF. Compressive strength of circular concrete-filled
of the outer square tube and global member slenderness. After double skin tubular short columns. Thin-Walled Struct 2014;77:165–73.
examining the structural behaviour of the CFDST columns, their [9] Han L-H, Li Y-J, Liao F-Y. Concrete-filled double skin steel tubular (CFDST)
columns subjected to long-term sustained loading. Thin-Walled Struct
capacities were compared with strength predictions based on the
2011;49:1534–43.
provisions of EC4 [13] and the AISC Specification [14] for CFST col- [10] Zhao XL, Han LH. Double skin composite construction. Prog Struct Mat Eng
umns, since no specific rules for CFDST columns currently exist. It 2006;8(3):93–102.
was found that both design approaches [13,14] provide reasonable [11] Han L-H, Tao Z, Huang H, Zhao X-L. Concrete-filled double-skin (SHS outer and
CHS inner) steel tubular beam-columns. Thin-Walled Struct
predictions compared to the FE strengths, with the AISC method 2004;42(9):1329–55.
[14] yielding the more accurate results. This indicated that the [12] ABAQUS Standard User’s Manual The Abaqus Software is a product of Dassault
behaviour and strength of the sandwiched concrete was generally Systèmes Simulia Corp., Providence, RI, USA Dassault Systèmes, Version 6.8,
USA; 2008.
similar to that of the concrete in conventional CFST columns. [13] Eurocode 4. Design of composite steel and concrete structures – Part 1-1:
Slightly modified proposals based on the AISC and EC4 provisions General rules and rules for buildings. British Standard Institution, CEN; 2004.
424 M.F. Hassanein et al. / Engineering Structures 100 (2015) 410–424

[14] AISC 360. Load and resistance factor design specification, for structural steel [25] Hassanein MF, Kharoob OF, Liang QQ. Circular concrete-filled double skin
buildings. American Institute of Steel Construction, Chicago; 2010. tubular short columns with external stainless steel tubes under axial
[15] Bergmann R, Matsui C, Meinsma C, Dutta D. Design guide for concrete filled compression. Thin-Walled Struct 2013;73:252–63.
hollow section columns under static and seismic loading. Verlag TUD [26] Hassanein MF, Kharoob OF, Liang QQ. Behaviour of circular concrete-filled lean
Rheinland, Koln: CIDECT; 1995. duplex stainless steel tubular short columns. Thin-Walled Struct
[16] GB50017-2003. Code for design of steel structures. China Planning Press, 2013;68:113–23.
Beijing; 2003 [in Chinese]. [27] Hassanein MF, Kharoob OF, Liang QQ. Behaviour of circular concrete-filled lean
[17] Eurocode 3. Design of steel structures – Part 1-1: General rules and rules for duplex stainless steel–carbon steel tubular short columns. Eng Struct
buildings. British Standard Institution, CEN; 2005. 2013;56:83–94.
[18] Tao Z, Uy B, Han LH, Wang ZB. Analysis and design of concrete-filled stiffened [28] Eurocode 3. Design of steel structures – Part 1-5: Plated structural elements.
thin-walled steel tubular columns under axial compression. Thin-Walled British Standard Institution, CEN; 2006.
Struct 2009;47(12):1544–56. [29] Hassanein MF, Kharoob OF. Analysis of circular concrete-filled double skin
[19] Theofanous M, Chan TM, Gardner L. Structural response of stainless steel oval tubular slender columns with external stainless steel tubes. Thin-Walled
hollow section compression members. Eng Struct 2009;31(4):922–34. Struct 2014;79:23–37.
[20] Hu HT, Huang CS, Wu MH, Wu YM. Nonlinear analysis of axially loaded [30] Han LH. Tests on concrete filled steel tubular columns with high slenderness
concrete-filled tube columns with confinement effect. J Struct Eng, ASCE ratio. Adv Struct Eng 2000;3(4):337–44.
2003;129(10):1322–9. [31] Uy B, Tao Z, Han L-H. Behaviour of short and slender concrete-filled stainless
[21] Mander JB, Priestley MJN, Park R. Theoretical stress–strain model for confined steel tubular columns. J Constr Steel Res 2011;67:360–78.
concrete. J Struct Eng ASCE 1988;114(8):1804–26. [32] Tao Z, Wang Z-B, Yu Q. Finite element modelling of concrete-filled steel stub
[22] ACI-318. Building code requirements for reinforced concrete. ACI, Detroit (MI); columns under axial compression. J Constr Steel Res 2013;89:121–31.
2002. [33] Liang QQ. High strength circular concrete-filled steel tubular slender beam-
[23] Tomii M, Sakino K. Elastic–plastic behavior of concrete filled square steel columns, Part I: Numerical analysis. J Constr Steel Res 2011;67:164–71.
tubular beam-columns. Trans Archit Inst Jpn 1979;280:111–20. [34] Han L-H, Zhao X-L, Tao Z. Tests and mechanics model of concrete-filled SHS
[24] Liang QQ. Performance-based analysis of concrete-filled steel tubular beam- stub columns, columns and beam-columns. Steel Compos Struct – Int J
columns, Part I: Theory and algorithms. J Constr Steel Res 2009;65:363–72. 2001;1(1):51–74.