Journal of Constructional Steel Research 81 (2013) 3543

Contents lists available at SciVerse ScienceDirect

Journal of Constructional Steel Research

Load-reversed push-out tests on rectangular CFST columns

*
Xiushu Qu a, Zhihua Chen a, , David A. Nethercot b, Leroy Gardner b, Marios Theofanous b
a
School of Civil Engineering, Tianjin University, Tianjin, 300072, PR China
b
Department of Civil and Environmental Engineering, South Kensington Campus, Imperial College London, SW7 2AZ, UK

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

Article history: Load-reversed push-out tests have been carried out on 6 rectangular concrete-lled steel tubular (CFST) col-
Received 4 February 2012 umns with the aim of investigating the nature of the bond between the concrete inll and the steel tube, the
Accepted 3 November 2012 contribution of each bond stress component (i.e. chemical adhesion, microlocking and macrolocking) and the
Available online 29 November 2012
development of macrolocking within four half-cycles of loading. The contribution of microlocking to the total
bond strength was obtained from the comparison between the ultimate strength of normal specimens and
Keywords:
Concrete-lled steel tubes (CFST)
lubricated specimens, which also revealed the detrimental effect of lubrication on the bond strength. The
Interface bond strength macrolocking contribution was obtained from the comparison between the ultimate strength achieved in
Interface condition the rst half-cycle of loading u1 and the ultimate strength achieved in the third half-cycle of loading u3 of
Load-reversal the non-lubricated specimens. The developed bond mechanisms were explained and details of the interface
Macrolocking bond stress distribution were obtained from the recorded axial strain gradients in the steel tube. Finally, the
Microlocking concept of a critical shear force transfer length was introduced, and its implications on practical design
Push-out test discussed.
Crown Copyright 2012 Published by Elsevier Ltd. All rights reserved.

1. Introduction
Concrete-lled steel tube (CFST) compression members have at-
tracted considerable attention from both practising engineers and re-
searchers owing to their favourable structural properties and signicant
nancial advantages over their structural steel or reinforced concrete
counterparts that make them well-suited for structural applications.
The steel tube connes the concrete inll thus enhancing its strength
and ductility, whilst the concrete core delays the onset of local buckling
of the tube wall, particularly for RHS and SHS. Hence the resulting com-
posite columns possess high strength, stiffness and ductility thereby lead-
ing to smaller column sizes and to larger lettable areas particularly in the
lower storeys of multi-storey buildings [1]. Moreover the steel tubes act
as permanent formwork for the concrete and can support signicant con-
struction loads before the hardening of the concrete thereby speeding up
construction times and lowering the cost. The slenderness of the steel
tubes and the strength of the concrete inll utilised in practical applica-
tions vary from one country to another [2], with the majority of the
research being directed towards relatively thick-walled tubes lled with
normal strength concrete, whilst thin-walled tubes lled with high-
strength concrete have received less attention [3]. Traditionally CFST
have employed square, rectangular and circular hollow sections (SHS,
RHS and CHS respectively) or even octagonal hollow sections [4,5],
* Corresponding author. Tel./fax: +86 22 27403736.
E-mail addresses: quxiushu@bucea.edu.cn (X. Qu), zhchen@tju.edu.cn (Z. Chen),
d.nethercot@imperial.ac.uk (D.A. Nethercot), leroy.gardner@imperial.ac.uk
(L. Gardner).
whilst, more recently, concrete lled elliptical hollow sections (EHS)
have been introduced [6,7].
Key to the satisfactory structural response of CFST is the effective
shear stress transfer from one material to the other (i.e. composite
action) particularly in regions of geometric discontinuity of the struc-
tural members, where bond stress demand is the greatest [2]. Com-
posite action may be achieved via the natural bond between steel
and concrete, similarly to the bond between steel reinforcing bars
and surrounding concrete, or with the aid of shear connectors of var-
ious forms, including structural bolts [810], Hilti nails [9], threaded
bars [10], self-taping screws [11] and tab stiffeners [12].
Early research on the bond strength in circular CFST was con-
ducted by Virdi and Dowling [13] who performed push-out tests on
stocky CHS lled with concrete of various grades. They examined
the effect of various parameters on the bond strength, such as the
age, strength, compaction and curing conditions of concrete, the in-
terface length, the tube size and various surface treatments and
concluded that the most signicant factor contributing to the bond
strength is the mechanical keying between steel and concrete.
Morishita et al. [4,14] tested CHS, SHS and octagonal hollow sections
lled with concrete to investigate the effects of cross-section shape
and concrete grade on bond strength. They reported that concrete-
lled CHS display increased bond strength compared to SHS and
octagonal cross-sections and that high strength concrete results in
decreased bond strength compared to normal strength concrete for
CHS due to increased shrinkage. Similar observations regarding the
effect of the concrete grade on bond strength were also made else-
where [8,13], and have also been attributed to the higher shrinkage

0143-974X/$ see front matter. Crown Copyright 2012 Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jcsr.2012.11.003
36 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 3543
associated with increased concrete strength. The bond strength
achieved by different cross-section geometries was also investigated
by Shakir-Khalil [8] and the superior bond strength achieved by CHS
compared to RHS was attributed to the more effective connement
pressure exerted by the CHS as evidenced by the friction marks
which were distributed uniformly around the CHS, whereas friction
marks due to connement pressure were only observed in the corner
regions of the RHS specimens. Moreover, the diminished bond
strength observed for the larger tested specimens compared to the
smaller ones, was attributed to the more pronounced effect of con-
crete shrinkage for large cross-sections. A detailed study on the
effects of shrinkage and tube dimensions was reported by Roeder et
al. [2], who also highlighted the effect of pressure of the wet concrete
and internal surface irregularity of the tube on the bond strength at
the concretesteel interface for lled CHS.
The inuence of tube interface on bond strength was studied by
Tomii et al. [5,15], who tested CFST with checkered internal walls,
Shakir-Khalil [8], who compared the bond strength achieved by
lubricated and non-lubricated concrete-lled RHS and CHS specimens
and Kilpatrick and Rangan [11], who investigated various interface con-
ditions. Other investigations have studied the inuence of cyclic shear
force [16], cyclic push-out force [17], concrete compaction [18] and
utilisation of expansive cement [19], whilst the effect of elevated tem-
peratures on bond strength has also been investigated [20]. A detailed
summary of relevant past experimental programmes is given in [21].
Based on the ndings of the aforementioned research, the main
parameters affecting the bond strength are the cross-section shape
[2,4,8,9,14,20], interface roughness [2,5,8,11,1315], variations in in-
ternal tube dimensions [13,20], concrete age [13,17,20], compaction
[13,18] and shrinkage [2,4,8,13,14,19] and cross-section dimensions
[2,3,8,20]. There seems to be a lack of consensus regarding the effect
of interface length on bond strength, since bond strength seems to in-
crease with interface length according to [20], whereas changing the
interface length did not have a direct effect on the maximum bond
strength of specimens according to [8,13]. Similarly, discrepancies
regarding the effect of concrete strength can be found in the literature
with some researchers nding that higher concrete grade leads to
increased bond strengths [19,22], whilst others observed the opposite
trend, due to the increased shrinkage associated with higher concrete
strength [3,4,8,13,14], particularly for CHS, where the reduction in
conning pressure due to shrinkage is more pronounced.
In addition to studies of ultimate interface bond strength, the mech-
anisms contributing to it have also been assessed. Virdi and Dowling
[13] identied two distinct components contributing to the bond
strength apart from chemical adhesion, namely microlocking and
macrolocking. Microlocking is dened as the keying between the con-
crete and the roughness of the steel surface, whilst macrolocking refers
to the resistance to movement of the concrete core along the tube
concrete interface due to the manufacturing tolerances associated
with the internal tube dimensions. The same bond mechanisms have
also been identied for RHS [8,23].
The response of a CFST subjected to push-out forces is qualitatively
depicted in Fig. 1, where the contribution of the three components of
bond strength at the various stages of loading is shown in terms of an
idealized forceslip curve. The bond due to chemical adhesion and
microlocking has to be broken for slip to occur, whereupon macrol-
ocking is activated. Chemical adhesion and microlocking govern the ini-
tial linear part of the curve and contribute mainly to the attainment of
the maximum bond stress, whereas macrolocking determines the resid-
ual bond stress that remains at the end of the bond stressslip curve.
The stressslip curve can assume three different shapes as discussed
in [20], depending on the relative contribution of macrolocking upon
the loss of bond. It may display a maximum followed by a falling branch,
a maximum followed by a falling branch, which raises again at large
slips, or display no maximum at all, as was the case for all tests reported
in [13].
Fig. 1. Idealized response of push-out specimens.
Recent research to investigate the contribution of microlocking to
the interface bond stress has been conducted by Chen et al. [24], who
assessed the contribution of microlocking by comparing the ultimate
average bond stress achieved by normal and lubricated specimens
similarly to previous investigations [8,13]. Based on the test results
reported in [13,24], the ratios of microlocking to bond stress are
32%75% for circular CFST specimens, 10%20% for square CFST spec-
imens, and 10%50% for rectangular CFST specimens, respectively. In
[8] it was observed that lubricated (i.e. with reduced effect from
microlocking) specimens attained approximately half the ultimate
bond stress of the non-lubricated ones. However, studies into the
contribution of macrolocking are rather limited and warrant further
investigation.
The main objective of this paper is to assess the contribution of the
various components of the bond strength (i.e. chemical adhesion,
microlocking, macrolocking) in the case of concrete lled RHS with
the focus being on the contribution of macrolocking. Six rectangular
CFST columns have been tested using the load-reversed push-out test
method with one specimen lubricated at the interface. Details of the in-
terface bond stress distribution through four half-cycles of loading are
given. The microlocking contribution is derived from the relative ratio
of bond stress between the normal specimens and the lubricated spec-
imen, recorded at the rst loading stage. The macrolocking contribution
is obtained by comparing the ultimate bond stresses from the rst and
third half-cycles of loading. Based on the test data, the effect of interface
length and concrete strength on the interface bond strength of rectan-
gular CFST columns was also investigated, supplementing previous re-
search in this area. Finally, the critical shear force transfer length was
studied to provide guidance for practising engineers.
2. Experimental study
2.1. General
A total of six rectangular CFST specimens were prepared and tested
under load reversal. Table 1 provides a summary of the test specimens.
Specimens CP1CP5 had essentially similar properties. Specimen CP6
was slightly longer and was lled with concrete of slightly higher
strength. The specimen length was selected such that the specimens
behave essentially as stub columns with no inuence from member
slenderness, whilst still including a representative distribution of initial
geometric imperfections along the length, thus allowing macrolocking
to develop. The interface length was selected as the nominal specimen
length less approximately 60 to 80 mm to allow an adequate space for
the concrete core to displace inside the specimens. The interface length
of specimens CP1 to CP5 was maintained at approximately 700
750 mm, whilst the interface length for specimen CP6 was 843 mm.
For CP5, lubrication was used in the form of butter at the concrete
steel interface to study the loss of microlocking. The section labelling
convention is shown in Fig. 2.
 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 3543 37

Table 1
Specimen details for the load-reversed push-out tests.

Specimen Depth D (mm) Width B (mm) Thickness t (mm) Length L (mm) Interface length Li (mm) Concrete grade Steel grade Interface treatment

CP1 300 200 5.67 800 715 C40 Q235 No lubrication

CP2 300 200 5.67 800 733 C40 Q235 No lubrication
CP3 300 200 5.67 800 726 C40 Q235 No lubrication
CP4 300 200 5.67 800 718 C40 Q235 No lubrication
CP5 300 200 5.67 800 713 C40 Q235 Lubrication
CP6 300 200 5.67 900 843 C50 Q235 No lubrication

2.2. Preparation of specimens
The interface bond strength was measured by means of load-
reversed push out tests. The ends of the steel tubes were cut and ma-
chined to the required length before delivery, ensuring that the two
ends were parallel to each other and normal to the sides. Any deposits
of dust or oil on the inside of the steel tubes were removed. The inter-
face of specimen CP5 was lubricated with butter prior to the casting of
the concrete core. The concrete was cast within the steel tubes, leav-
ing approximately 50100 mm air gap at one end of each specimen.
All the specimens were compacted by hand.
2.3. Material properties
Two different grades of commercial concrete were used in the
tests. For each batch, six 100 mm concrete cubes were cast and
cured in standard laboratory conditions for 28 days [25]. After that
period, experiments to measure the concrete elastic modulus and
concrete compressive strength were conducted. The average values
of the material parameters are shown in Table 2.
The material properties of the steel tubes were obtained by means
of tensile coupon tests. Four coupons (see Fig. 2) were tested in accor-
dance with [26]. The key average measured properties are shown in
Table 3.
2.4. Experimental set-up
All specimens were tested at around 40 days of age. The load-
reversed push-out tests were conducted using a 500 kN capacity hy-
draulic jack as depicted in Fig. 3. The specimens were set up in the
testing machine in the vertical position with the air gap at the bottom.
A layer of sand was rst spread on the top surface of the specimens.
Then, a steel block, which had a slightly smaller cross-section than
that of the concrete core, was placed on the top of the specimen.
This ensured that the load was applied only on the concrete core
and allowed the concrete core to be pushed downward when testing.
The load was measured by a load cell placed on the hydraulic jack.
The movement of the concrete core with respect to the steel tube at
z represent the interface bond strength. This is determined from:
Flat coupon
the top end was measured by means of two linearly varying displace-
ment transducers (LVDTs) placed at the two sides of the specimen.
The deection of the concrete core was then taken as the average of
the two transducer values. In order to study the strain distribution
in the steel hollow section, strain gauges were xed on the narrow
and wide faces of the steel tubes along the longitudinal length of
each specimen. The spacing between two adjacent strain gauges
was 100 mm. Since the strain may vary signicantly near to the load-
ed end, two strain gauges, placed 50 mm from the loaded end, were
added to the two faces of the steel tube. Details of the strain gauges
for specimen CP1 are shown in Fig. 4.
During testing, load was applied at the top of the specimen to the
concrete core and was resisted at the base by the steel section alone.
Initially, load was increased at the rate of 10 kN/min and the deec-
tions were recorded at every 2 kN step. Once the specimen started
to show a marked change in the slip between the steel tube and con-
crete core, as indicated by the LVDTs, the recordings were taken after
every 2 mm of movement of the concrete core. Loading continued
until the recorded load showed little change for large increases of
slip. At this point, the rst load application was complete. After that,
the specimen was inverted and reversed load was applied to the spec-
imens following the above-mentioned procedure until the recorded
load showed little change for large increases of slip. Then, the speci-
men was reversed again to apply a further cycle. Each specimen was
subjected to 4 half-cycles of loading.
2.5. Experimental results
For each test, the slip between the steel and the concrete core was
initially linearly related to the increments in loading. At a certain
stage, slip developed more rapidly and the interface carrying capacity
was observed to vary with the loading direction and the load reversal.
Generally, for loading in the same direction, the interface carrying ca-
pacity decreased as the load cycles increased. After the test, the state
of the concrete contact surface was examined. A general view of the
top and bottom contact surface of the specimens observed after test-
ing is shown in Fig. 5. It may be seen that the concrete had been
slightly crushed in some local contact surface regions.
For design purposes, the average interface bond stress is used to
N=CLi 1

t
in which N is the push-out load value, C is the perimeter of the con-
y y
D

crete section in contact with the steel tube and Li is the interface
length.

Table 2
Measured concrete properties.

z Weld Concrete strength Young's modulus EC Compressive strength (cube) fcu

grade (MPa) (MPa)
B C40 29,380 39
C50 38,070 49
Fig. 2. Section labelling convention and location of at tensile coupons.
38 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 3543
Table 3
Key material properties from tensile coupons tests.
Specimen Young's modulus Yield stress Ultimate tensile
ES (MPa) fy (MPa) strength fu (MPa)
RHS 300 200 5.67 (Q235b) 224,400 257.6 512.3
Values of , calculated using Eq. (1), were used to determine the
interface bond stress () versus slip (S) curves of Fig. 6. As expected,
the peak bond stress under the rst application of loading is typically
higher than under subsequent loading. Additionally, two somewhat
different forms of curve may be observed: for the rst, second and
fourth half-load cycles, the curves of all the specimens have the
same feature, with an initial linear portion followed by a transitional
portion. After reaching the peak strength, a rapidly declining portion
appears before a relatively stable residual strength is attained. For the
third load reversal, the curves of all the specimens are characterized
by the lack of any falling branch. The reason for the behaviour in
the third half-load cycle being different from the others is explained
in Section 3.4.
Fluctuation may be observed in some of the bond stressslip
curves, especially for specimen CP5. A possible explanation for the
higher uctuations in the case of CP5 is that lubrication on the inside
of the tube may not have been entirely uniform. For all other speci-
mens, uctuations may be attributed to the nonhomogenous conn-
ing pressure exerted by the tubes on the concrete core, due to local
irregularities of the corner surface of the steel tubes, causing extra
microlocking in these regions.
The ultimate average bond strength and its corresponding slip for all
specimens at every load reversal are shown in Table 4. For each speci-
men, un (n = 1, 2, 3 or 4) is the ultimate average bond strength at the
nth (n = 1, 2, 3 or 4) load reversal, as derived from Eq. (1), where Nun
is the peak load in the nth curve, un = Nun / CLi. For all specimens, the
slip at the peak load in the nth half-cycle of loading is Sun (n = 1, 2, 3
or 4). For specimen CP5, although the peak interface bond strength of
0.112 MPa can be observed after approximately 10 mm of slip, at the
uctuation of the rst curve, the rst peak point at the bond strength
of 0.106 MPa and lower slip is believed to be a more relevant value. It
should be noted that all S curves displayed a lack of a maximum in
the third half-cycle, similarly to some results reported in [8,13,20].
Hence Su3 is the displacement which corresponds to the maximum
recorded average stress during the third half-cycle of loading. Although
this is not indicative of the ultimate bond stress of the specimens, it still
provides some useful information about the behaviour of the speci-
mens. In this case it is the shape of the loadslip curve rather than the
actual value of the u3 that is signicant.
Reaction frame
Load cell
LVDTs Hydraulic jack
Air gap Specimen
Rigid base
Fig. 3. Set-up for the test specimens.
(a) CP1 (b) CP1 1-1 (c) CP1
cross-section cross-section 2-2 cross-section
2
1
300
1
200
2
200 300
Fig. 4. Specimen details: CP1 (dimensions in mm).
3. Analysis of test results
3.1. Bond mechanisms
As discussed in Section 1, the interface bearing capacity arises
from three distinct mechanisms: chemical adhesion, microlocking
and macrolocking. Chemical adhesion is generated by capillary action,
which arises during the cement hydration process. According to
previous research [22] chemical adhesion makes only a minor contri-
bution to the bond strength and is inuenced by many factors, such as
the cement content and watercement ratio. Microlocking, is caused
by the roughness of the bond interface from the micro-perspective
[2,13,8,20]. Macrolocking, also referred to as interface friction, arises
due to the manufacturing tolerances associated with variations of
the internal dimensions of the tube [2,13]. A graphical representation
of each component is shown in Fig. 7.
During load-reversed push-out tests, the contribution of each of the
aforementioned factors to the bond strength is different for different
load reversals and load levels. Generally, at the beginning of the rst
loading cycle, there is no slip between the steel tube and the concrete
core and the interface bond strength is due entirely to chemical adhe-
sion. As the push-out load increases, slip between the steel tube and con-
crete core occurs over part of the end regions, where the chemical
adhesion is damaged and microlocking starts to engage. As the load is
further increased, the interface slippage grows and begins to move to-
wards the middle part of the interface length, the chemical adhesion
only exists in the areas without slippage, whilst the interface bond
strength in the areas with slip develops mainly from microlocking.
Once relative slip occurs along the whole interface length, chemical ad-
hesion ceases to contribute and microlocking is also broken in the
areas of greater slip. At this stage, the push-out load equals the interface
carrying capacity and the interface bond strength results mainly from
microlocking and macrolocking. As slip increases further, the effect of
the roughness of the bond interface reduces and the microlocking contri-
bution begins to decline. Eventually, the interface bond strength is due
solely to macrolocking. After the rst reversal, there is no chemical adhe-
sion and microlocking is very small and may be assumed to be zero.
Therefore, the interface bond strength may be assumed to be provided
almost entirely from macrolocking under all subsequent load reversals.
3.2. Bond stress distribution
To understand the bond mechanisms, it is necessary to check the
bond stress distribution along the length of the tubes. The bond stresses
can be obtained from the axial strain gradients in the steel tube; these
are related by statics to the interface bond stress distribution [2]. If Nx
 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 3543 39

(a) CP2 top (b) CP3 bottom (c) CP5 bottom

Fig. 5. Typical contact surfaces of specimens after testing.

is the axial load in the steel tube at location x, then the bond stress, (x), occur at the bottom of the steelconcrete interface and may be
can be obtained from: inuenced by the pinching effect, which causes the steel tube to
deform outwards at this location.
dN x
x : 2
2B  2t 2D  2t dx 3.3. Inuence of different parameters

Figs. 8 and 9 illustrate the bond stress distributions on the narrow
and wide faces of the steel tube for specimen CP1 at different load
levels. The bond stress distribution is not uniform along the tube
length, but the value of local bond stress increases as the load
increases until the ultimate load is reached. Some local calculated
bond stress values were zero in Figs. 8 and 9. These conditions usually
As discussed above, the interface bond carrying capacity included
all three bond mechanisms only in the rst load cycle. Therefore, the
study of the effect of different factors on bond strength in this section
refers to their inuence on the ultimate bond strength under the rst
load application. As shown in Fig. 6, the slip initially varies linearly
with load. To illustrate this more clearly, the bond stress versus slip

(a) CP1 (b) CP2

0.4 0.4
Bond stress (MPa)

Bond stress (MPa)

0.3 First half-cycle 0.3 First half-cycle

0.2
0.2
0.1
Third half -cycle 0.1
0.0 Third half -cycle
0 10 20 30 40 50 60 70 0.0
-0.1 0 10 20 30 40 50 60 70
-0.2 Fourth half-cycle -0.1
Fourth half-cycle
Second half-cycle Second half-cycle
-0.3 -0.2
Slip (mm) Slip (mm)
(c) CP3 (d) CP4
0.4 0.4
Bond stress (MPa)

Bond stress (MPa)

0.3 First half-cycle 0.3 First half-cycle

0.2
0.2
0.1
0.1 Third half-cycle
Third half -cycle 0.0
0.0 0 10 20 30 40 50 60 70 80
0 10 20 30 40 50 60 70 -0.1
-0.1 -0.2 Second half-cycle
Second half-cycle Fourth half-cycle
-0.2 Fourth half-cycle -0.3
Slip (mm) Slip (mm)

(e) CP5 (f) CP6

0.15 0.4
First half-cycle
Bond stress (MPa)

Bond stress (MPa)

0.10 0.3 First half-cycle

0.05 0.2
Third half-cycle Third half-cycle
0.00 0.1
0 20 40 60 80 100
-0.05 0.0
0 10 20 30 40 50 60
-0.10 Fourth half-cycle -0.1
Fourth half-cycle
Second half-cycle Second half-cycle
-0.15 -0.2
Slip (mm) Slip (mm)
Fig. 6. Bond stress versus slip S curves from load-reversed push-out tests for each specimen.
40 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 3543

Table 4 to smoothing of the interface associated with deterioration of friction,

Results from load-reversed push-out tests. though this reduction seems to be relatively small.
Specimen Su1 u1 Su2 u2 Su3 u3 Su4 u4 The third half-cycle exhibits a notably different response on the
reference (mm) (MPa) (mm) (MPa) (mm) (MPa) (mm) (MPa) rest loading half-cycles and displays a lack of maximum average
CP1 0.997 0.305 0.805 -0.192 52.9 0.166 0.38 -0.203 bond strength. Hence it is both qualitatively and quantitatively differ-
CP2 0.973 0.271 1.09 -0.144 57.9 0.140 1.09 -0.130 ent to the second and the fourth half-cycles, despite the fact that in all
CP3 0.881 0.282 0.436 -0.181 62.1 0.152 1.01 -0.131 three of them macrolocking governs the response. This may be attrib-
CP4 1.15 0.259 0.938 -0.188 42.2 0.143 2.42 -0.129
uted to the constraining force alterations in response to change in the
CP5 3.56 0.106 0.503 -0.126 69.2 0.053 0.300 -0.103
CP6 0.841 0.319 5.84 -0.135 40.1 0.11 1.75 -0.103 load direction, due to a trend of increasing or decreasing diameter
with length, as shown, in an exaggerated fashion, in Fig. 11. For the
case illustrated, in the rst half-cycle, the constraining force will
reduce with increasing load and the concrete will slip more easily as
relationship for the rst load application to specimens CP1 to CP6 is a result. However, when the load direction is reversed, the concrete
shown in Fig. 10 with a 10 mm range on horizontal axis. An initial lin- would be wedged into a narrowing funnel-like shape and the cons-
ear portion followed by a transitional ascending portion appears for training force would, as a consequence, increase.
all curves. It should be noted that the transitional ascending portion
of CP5, for which the interface is lubricated, is 2.99 mm, which is 3.5. Relative contribution of each bond mechanism
larger than that of the other specimens. This tendency is consistent
with the results observed by Shakir-Khalil [8] and Virdi and Dowling Based on the composition and variation in the interface bearing
[13], who also observed the initiation of nonlinear response and the capacity, it can be assumed that the ultimate average bond stress in
subsequent reduction in stiffness at lower load levels for lubricated half-cycle 1 u1 for the lubricated specimen CP5 is derived, for the
specimens. most part, from microlocking and that the ultimate average bond stress
In Fig. 10, it can be seen that the average ultimate bond stress of in half-cycle 3 u3 for normal specimens is largely derived from
CP6, which had a higher concrete compressive strength of C50, is macrolocking. Since in the third half-cycle the load is in the same direc-
slightly higher than that of the other specimens. It may hence be tion and orientation as in the rst half-cycle, the macrolocking contribu-
concluded that for the specimens considered the concrete strength tion in the rst load application may, therefore, be approximated by u3
does have a degree of inuence on the bond strength. However as for normal specimens.
discussed in Section 1, there is a lack of consensus regarding the effect Specimens CP1CP5 have the same cross-section, concrete grade
of concrete grade on bond strength, due to the fact that the benecial and steel grade characteristics, and although the interface length varies
effect of increased concrete strength is usually accompanied by the slightly, this is unlikely to inuence the interface bond stress for these
detrimental effect of increased shrinkage. specimens. As a result, their recorded test data can be used to analyse
the contribution of each bond stress component without any variation
in other parameters. Comparing u1 for specimen CP5 (with lubrication)
3.4. Development of macrolocking with that of the u1 for normal specimens (CP1CP4), the microlocking
bond strength was found to be about 40% of the average ultimate bond
After the rst half-cycle of loading, the chemical adhesion is strength. Comparing u3 and u1 for specimens CP1 to CP4, the average
destroyed and the microlocking is now at an almost negligible level. contribution of macrolocking was found to be approximately 55%.
As a consequence, after the rst load half-cycle, it can be assumed that Hence it can be concluded that the bond strength exhibited by the
the interface bearing capacity is provided only by macrolocking. concrete-lled RHS is mainly attributed to macrolocking and micro-
Macrolocking is generated by changes in the internal diameter of the locking, whilst the effect of chemical adhesion appears to be very
tube and refers not only to the constraining force but also to the friction limited.
between the steel tube and the concrete core. Xue and Cai [22]
suggested that the constraining force varies as a function of the 3.6. Critical shear force transfer length
cross-sectional diameter and the slip between the steel and concrete,
whilst, at the same time, the friction coefcient gradually decreases In practice, CFST columns may be loaded through the concrete
with increasing load cycles. core alone, by connection to the steel tube only or by a combination
With reference to Fig. 6, it may be observed that the second and thereof, for example with through-plate connections. Shear connec-
the fourth half-cycle S curves of all specimens resemble each tors may also be present if required.
other in shape. This veries the aforementioned assumption, that The ultimate bond capacity Nu may be given by:
the interface bond strength established after the rst half-cycle can
be entirely attributed to the macrolocking effect. Furthermore, the ul- Nu Ls 2B  2t D  2u 3
timate interface bond strengths observed in the second half-cycle are
slightly larger than those observed in the fourth half-cycle. This indi- where u is the average ultimate bond strength, taken herein as u1 and Ls
cates that the friction coefcient reduces with increasing cycles due is the shear force transfer length. Assuming that the capacity of the two

(a) Chemical adhesion (b) Microlocking (c) Macrolocking

Fig. 7. Interface bond stress components.

 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 3543 41

(a) First half-cycle (b) Second half-cycle

0.4

Bond stress of the steel narrow face (MPa)

Bond stress of the steel narrow face (MPa)

1.5
N=0.383Nu N=0.276Nu
N=0.525Nu N=0.525Nu
N=0.755Nu N=0.756Nu
0.3
N=Nu N=Nu
1.0

0.2

0.5
0.1

0.0 0.0
0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800
Distance away from the loaded end (mm) Distance away from the loaded end (mm)

(c) Third half-cycle (d) Fourth half-cycle

0.3

Bond stress of the steel narrow face (MPa)

Bond stress of the steel narrow face (MPa)

4
N=0.292Nu
N=0.307Nu
N=0.523Nu
N=0.525Nu
N=0.705Nu
3 N=0.661Nu
N=Nu
N=0.789Nu 0.2
N=Nu

0.1
1

0 0.0
0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800
Distance away from the loaded end (mm) Distance away from the loaded end(mm)

Fig. 8. Bond stress distribution in the narrow face of the steel tube along specimen CP1 length at different load levels.

components of the CFST column is given by their respective yield or length and B the outer SHS width), whilst the opposite trend was
crushing resistances, the minimum failure load Nf is: observed for circular tubes. Hence it can be concluded that there is
n o no consensus regarding the effect of interface length on the bond
Nf min Ac f c; As f s 4 strength for concrete-lled RHS, which indicates that the interface
length cannot be considered a dominant factor affecting bond stress.
where Ac is the cross-sectional area of concrete core, As is the cross-section For specimens CP14 and 6, which have a D/t ratio of approxi-
area of steel tube, fc is the compressive cylinder strength of concrete, mately 50, the experimental average value of Ls,cr is 5445 mm. If the
taken as fc =0.79fcu ( fcu being the cube strength) and fy is the yield full bond between the steel tube and the concrete core is to be relied
strength of the steel. upon in design, then the member length adopted in practice should
Hence, the critical length Ls,cr required for full shear transfer exceed Ls,cr or suitable measures to ensure effective load transfer,
(i.e. Nu = Nf ) can be obtained as: such as through-plate connections or shear connections, should be
adopted.
n o
min Ac f c; As f s
Ls;cr : 5
2B  2t D  2t u 4. Conclusions

It should be noted that the aforementioned derivation is based on Six load-reversed push-out tests have been conducted to investi-
the assumption that interface length has a negligible effect on bond gate the bond mechanisms in rectangular CFST columns, and the rel-
strength. As mentioned in Section 1, numerous researchers [8,13,20] ative contribution of each component. Each specimen was subjected
have studied the effect of interface length on bond strength without to four half-cycles of loading. Bond stress versus slip curves were
being able to verify a direct link. In [13] it was concluded that for obtained for each specimen. The axial bond strain distribution in the
circular CFST the contact length has no appreciable effect on bond narrow and wide faces of the steel tube, and along the entire length
strength, whilst Shakir-Khalil [8] pointed out that the interface length of the specimen at the point of the interface reaching its ultimate
does not seem to have a direct and proportional effect on the load load-carrying capacity, and, in the case of specimen CP1, at a range
carrying capacity of the specimens. Tao et al. [20] reported that the of different load levels, have been successfully recorded. Based on
bond strength increased with increasing Li / B (Li being the interface the experimental data, bond stress distributions in the narrow and
42 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 3543

(a) First half-cycle (b) Second half-cycle

0.6 0.6

Bond stress of the steel wide face (MPa)

Bond stress of the steel wide face (MPa)

N=0.383Nu N=0.383Nu
N=0.525Nu 0.5 N=0.525Nu
N=0.755Nu N=0.755Nu
0.4 N=Nu 0.4 N=Nu

0.3

0.2 0.2

0.1

0.0 0.0
0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800
Distance away from the loaded end (mm) Distance away from the loaded end (mm)

(c) Third half-cycle (d) Fourth half-cycle

3 1.5

Bond stress of the steel wide face (MPa)

Bond stress of the steel wide face (MPa)

N=0.307Nu N=0.292Nu
N=0.552Nu N=0.523Nu
1.2
N=0.661Nu N=0.705Nu
2 N=0.789Nu N=Nu
N=Nu 0.9

0.6
1

0.3

0 0.0
0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800
Distance away from the loaded end (mm) Distance away from the loaded end (mm)

Fig. 9. Bond stress distribution in the wide face of the steel tube along specimen CP1 length at different load levels.

wide faces of specimen CP1 were determined. The bond stress distri- full shear transfer was studied, and design recommendations were
bution was found to be non-uniform along the length of the tube. By made.
analysing the average ultimate bond stress for each of the four half
cycles of loading (u1, u2, u3 and u4), the relative contributions of Symbols
the three bond mechanisms were assessed. Macrolocking was found
to be the dominant mechanism contributing to bond strength,
followed by microlocking, whereas the effect of chemical adhesion
appeared to be limited. Finally, the critical interface length to ensure Ac cross-sectional area of concrete core
As cross-section area of steel tube
B width of the rectangular steel tube
0.35
C perimeter of the concrete section in contact with the steel
tube
0.30
D depth of the rectangular steel tube
Ec modulus of elasticity of concrete
0.25
Bond stress (MPa)

Es modulus of elasticity of rectangular steel tube

fc compressive cylinder strength of concrete
0.20
fcu compressive cube strength of concrete
fy yield strength of steel
0.15 L length of the rectangular steel tube
Li length of the steelconcrete interface
0.10 Ls shear force transfer length
CP1 CP4
Ls,cr critical shear force transfer length
0.05 CP2 CP5
N push-out load value
CP3 CP6 Nf minimum failure load of steel tube or concrete core
0.00 Nu ultimate interface bearing capacity
0 2 4 6 8 10
N(x) axial load in the steel tube at location x
Slip (mm)
S slip
Fig. 10. Bond stress versus slip S curves for each specimen during the rst half cycle of Su1 slip corresponding to ultimate push-out load in the rst
loading. half-cycle of loading
 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 3543 43

N N N

Fig. 11. Macrolocking action in CFST columns.

Su2 slip corresponding to ultimate push-out load in the second [7] Espinos A, Gardner L, Romero ML, Hospitaler A. Fire behaviour of concrete lled
elliptical steel columns. Thin-Walled Struct 2011;49(2):239-55.
half-cycle of loading [8] Shakir-Khalil H. Push-out strength of concrete-lled steel tubes to push out
Su3 slip corresponding to ultimate push-out load in the third forces. Struct Eng 1993;71(13):230-3.
half-cycle of loading [9] Shakir-Khalil H. Resistance of concrete-lled steel tubes to push out forces. Struct
Eng 1993;71(13):234-43.
Su4 slip corresponding to ultimate push-out load in the fourth [10] Shakir-Khalil H, Hassan NKA. Push-out resistance of concrete-lled tubes.
half-cycle of loading Proceedings of the sixth international symposium on tubular structures, Melbourne,
t wall thickness of the steel tube Australia, 1416 December 1994; 1994. p. 285-91.
[11] Kilpatrick AE, Rangan BV. Inuence of interfacial shear transfer on behavior of
interface bond stress
concrete-lled steel tubular columns. ACI Struct J 1999;96(4):642-8.
u1 average ultimate bond strength in the rst half-cycle of [12] Petrus C, Hamid HA, Ibrahim A, Nyuin JD. Bond strength in concrete lled built-up
loading steel tube columns with tab stiffeners. Can J Civ Eng 2011;38(6):627-37.
[13] Virdi KS, Dowling PJ. Bond strength in concrete lled circular steel tubes. Compos-
u2 average ultimate bond strength in the second half-cycle of
ite columns. CESLIC Report. CC11. London: Engineering Structures Laboratories,
loading Civil Engineering Department, Imperial College, London; 1975.
u3 average ultimate bond strength in the rst half-cycle of [14] Morishita Y, Tomii M, Yoshimura K. Experimental studies on bond strength in
loading concrete lled circular steel tubular columns subjected to axial loads. Transac-
tions of Japan Concrete Institute; 1979. p. 351-8.
u4 average ultimate bond strength in the fourth half-cycle of [15] Tomii M, Yoshimura K, Morishita Y. A method of improving bond strength in
loading between steel tube and concrete core cast in circular steel tubular columns. Trans-
(x) bond stress at location x actions of Japan Concrete Institute, 2; 1980. p. 319-26.
[16] Morishita Y, Tomii M. Experimental studies on bond strength between square
steel tube and encased concrete core under cyclic shearing force and constant
axial force. Transactions of Japan Concrete Institute, 4; 1982. p. 363-70.
[17] Aly T, Elchalakani M, Thayalan P, Patnaikuni I. Incremental collapse threshold for
Acknowledgements
push-out resistance of circular concrete lled steel tubular columns. J Constr Steel
Res 2010;66(1):11-8.
The authors are grateful to the China Standard Management Group [18] Han L-H, Yang Y-F. Inuence of concrete compaction on the behavior of concrete
for the BG50017-2003 Structural Steel Design Code for their special com- lled steel tubes with rectangular sections. Adv Struct Eng 2001;4(2):93100.
[19] Xu C, Chengkui H, Decheng J, Yuancheng S. Push-out test of pre-stressing concrete
posite structure research funding (GB 5000172010-04), and would like lled circular steel tube columns by means of expansive cement. Construct Build
to thank the Taian Kenuo prole steel stock company limited for the Mater 2009;23(1):491-7.
supply of test specimens and Feng Qilei, Xu Xiu, Sun Guojun and other [20] Tao Z, Han LH, Uy B, Chen X. Post-re bond between the steel tube and concrete in
concrete-lled steel tubular columns. J Constr Steel Res 2011;67(3):484-96.
students in the steel research group of Tianjin University for their assis- [21] Gourley BC, Tort C, Denavit MD, Schiller PH, Hajjar JF. A synopsis of studies of the
tance with the laboratory work. monotonic and cyclic behavior of concrete-lled steel tube members, connections
and frames. Report No. NSEL-008. Urbana-Champaign: Newmark Structural Engi-
neering Laboratory, Department of Civil and Environmental Engineering, University
References of Illinois; 2008.
[22] Xue LH, Cai SH. Bond strength at the interface of concrete-lled steel tubular
[1] Giakoumelis G, Lam D. Axial capacity of concrete-lled tube columns. J Constr
columns: part I. Build Sci 1996;12(3):22-8 [in Chinese].
Steel Res 2004;60(7):1049-68.
[23] Parsley MA, Yura JA, Jirsa JO. Push-out behavior of rectangular concrete-lled steel
[2] Roeder CW, Cameron B, Brown CB. Composite action in concrete lled tubes.
tubes. Composite and hybrid systems, ACI SP-196. Farmington Hills, Michigan: ACI;
J Struct Eng ASCE 1999;125(5):477-84.
2000.
[3] O'Shea MD, Bridge RQ. Design of circular concrete lled thin-walled steel tubes.
[24] Chen ZH, Qu XS, Wang XD, Sun RR, Li LM. Experimental study on the interface bear-
J Struct Eng ASCE 2000;126(11):1295-303.
ing capacity on concrete-lled square steel tube. J Harbin Inst Tech 2009;41(2):
[4] Morishita Y, Tomii M, Yoshimura K. Experimental studies on bond strength in
27-32.
concrete lled square and octagonal steel tubular columns subjected to axial
[25] GB50152-92. Standard methods for testing of concrete structures. Chinese Standard;
loads. Transactions of Japan Concrete Institute; 1979. p. 359-66.
1992.
[5] Tomii M, Yoshimura K, Morishita Y. A method of improving bond strength in
[26] GB/T 228-2002. Metallic materialstensile testing at ambient temperature. Chinese
between steel tube and concrete core cast in square and octagonal steel tubular
standard; 2002.
columns. Transactions of Japan Concrete Institute, 2; 1980. p. 327-34.
[6] Yang H, Lam D, Gardner L. Testing and analysis of concrete-lled elliptical hollow
sections. Eng Struct 2008;30(12):3771-81.