Journal of Constructional Steel Research 81 (2013) 35–43

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 Chinab Department of Civil and Environmental Engineering, South Kensington Campus, Imperial College London, SW7 2AZ, UK

* Corresponding author. Tel./fax: +86 22 27403736.E-mail addresses: quxiushu@bucea.edu.cn (X. Qu), z

d.nethercot@imperial.ac.uk (D.A. Nethercot), leroy.gard(L. Gardner).

0143-974X/$ – see front matter. Crown Copyright © 20http://dx.doi.org/10.1016/j.jcsr.2012.11.003

a b s t r a c t

a r t i c l e i n f o

Article history:Received 4 February 2012Accepted 3 November 2012Available online 29 November 2012

Keywords:Concrete-filled steel tubes (CFST)Interface bond strengthInterface conditionLoad-reversalMacrolockingMicrolockingPush-out test

Load-reversed push-out tests have been carried out on 6 rectangular concrete-filled steel tubular (CFST) col-umns with the aim of investigating the nature of the bond between the concrete infill and the steel tube, thecontribution of each bond stress component (i.e. chemical adhesion, microlocking and macrolocking) and thedevelopment of macrolocking within four half-cycles of loading. The contribution of microlocking to the totalbond strength was obtained from the comparison between the ultimate strength of normal specimens andlubricated specimens, which also revealed the detrimental effect of lubrication on the bond strength. Themacrolocking contribution was obtained from the comparison between the ultimate strength achieved inthe first half-cycle of loading τu1 and the ultimate strength achieved in the third half-cycle of loading τu3 ofthe non-lubricated specimens. The developed bond mechanisms were explained and details of the interfacebond stress distribution were obtained from the recorded axial strain gradients in the steel tube. Finally, theconcept of a critical shear force transfer length was introduced, and its implications on practical designdiscussed.

Crown Copyright © 2012 Published by Elsevier Ltd. All rights reserved.

1. Introduction

Concrete-filled steel tube (CFST) compression members have at-tracted considerable attention from both practising engineers and re-searchers owing to their favourable structural properties and significantfinancial advantages over their structural steel or reinforced concretecounterparts that make them well-suited for structural applications.The steel tube confines the concrete infill thus enhancing its strengthand ductility, whilst the concrete core delays the onset of local bucklingof the tube wall, particularly for RHS and SHS. Hence the resulting com-posite columnspossess high strength, stiffness and ductility thereby lead-ing to smaller column sizes and to larger lettable areas particularly in thelower storeys of multi-storey buildings [1]. Moreover the steel tubes actas permanent formwork for the concrete and can support significant con-struction loads before the hardening of the concrete thereby speeding upconstruction times and lowering the cost. The slenderness of the steeltubes and the strength of the concrete infill utilised in practical applica-tions vary from one country to another [2], with the majority of theresearch being directed towards relatively thick-walled tubes filled withnormal strength concrete, whilst thin-walled tubes filled with high-strength concrete have received less attention [3]. Traditionally CFSThave employed square, rectangular and circular hollow sections (SHS,RHS and CHS respectively) or even octagonal hollow sections [4,5],

hchen@tju.edu.cn (Z. Chen),ner@imperial.ac.uk

12 Published by Elsevier Ltd. All rig

whilst, more recently, concrete filled elliptical hollow sections (EHS)have been introduced [6,7].

Key to the satisfactory structural response of CFST is the effectiveshear stress transfer from one material to the other (i.e. compositeaction) 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 steeland concrete, similarly to the bond between steel reinforcing barsand surrounding concrete, or with the aid of shear connectors of var-ious forms, including structural bolts [8–10], Hilti nails [9], threadedbars [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 onstocky CHS filled with concrete of various grades. They examinedthe effect of various parameters on the bond strength, such as theage, strength, compaction and curing conditions of concrete, the in-terface length, the tube size and various surface treatments andconcluded that the most significant factor contributing to the bondstrength is the mechanical keying between steel and concrete.Morishita et al. [4,14] tested CHS, SHS and octagonal hollow sectionsfilled with concrete to investigate the effects of cross-section shapeand concrete grade on bond strength. They reported that concrete-filled CHS display increased bond strength compared to SHS andoctagonal cross-sections and that high strength concrete results indecreased bond strength compared to normal strength concrete forCHS due to increased shrinkage. Similar observations regarding theeffect of the concrete grade on bond strength were also made else-where [8,13], and have also been attributed to the higher shrinkage

hts reserved.

Fig. 1. Idealized response of push-out specimens.

36 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 35–43

associated with increased concrete strength. The bond strengthachieved by different cross-section geometries was also investigatedby Shakir-Khalil [8] and the superior bond strength achieved by CHScompared to RHS was attributed to the more effective confinementpressure exerted by the CHS as evidenced by the friction markswhich were distributed uniformly around the CHS, whereas frictionmarks due to confinement pressure were only observed in the cornerregions of the RHS specimens. Moreover, the diminished bondstrength observed for the larger tested specimens compared to thesmaller ones, was attributed to the more pronounced effect of con-crete shrinkage for large cross-sections. A detailed study on theeffects of shrinkage and tube dimensions was reported by Roeder etal. [2], who also highlighted the effect of pressure of the wet concreteand internal surface irregularity of the tube on the bond strength atthe concrete–steel interface for filled CHS.

The influence of tube interface on bond strength was studied byTomii et al. [5,15], who tested CFST with checkered internal walls,Shakir-Khalil [8], who compared the bond strength achieved bylubricated and non-lubricated concrete-filled RHS and CHS specimensand Kilpatrick and Rangan [11], who investigated various interface con-ditions. Other investigations have studied the influence of cyclic shearforce [16], cyclic push-out force [17], concrete compaction [18] andutilisation of expansive cement [19], whilst the effect of elevated tem-peratures on bond strength has also been investigated [20]. A detailedsummary of relevant past experimental programmes is given in [21].

Based on the findings of the aforementioned research, the mainparameters affecting the bond strength are the cross-section shape[2,4,8,9,14,20], interface roughness [2,5,8,11,13–15], 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 effectof interface length on bond strength, since bond strength seems to in-crease with interface length according to [20], whereas changing theinterface length did not have a direct effect on the maximum bondstrength of specimens according to [8,13]. Similarly, discrepanciesregarding the effect of concrete strength can be found in the literaturewith some researchers finding that higher concrete grade leads toincreased bond strengths [19,22], whilst others observed the oppositetrend, due to the increased shrinkage associated with higher concretestrength [3,4,8,13,14], particularly for CHS, where the reduction inconfining 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] identified two distinct components contributing to the bondstrength apart from chemical adhesion, namely microlocking andmacrolocking. Microlocking is defined as the keying between the con-crete and the roughness of the steel surface, whilst macrolocking refersto the resistance to movement of the concrete core along the tube–concrete interface due to the manufacturing tolerances associatedwith the internal tube dimensions. The same bond mechanisms havealso been identified for RHS [8,23].

The response of a CFST subjected to push-out forces is qualitativelydepicted in Fig. 1, where the contribution of the three components ofbond strength at the various stages of loading is shown in terms of anidealized force–slip curve. The bond due to chemical adhesion andmicrolocking has to be broken for slip to occur, whereupon macrol-ocking is activated. Chemical adhesion andmicrolocking govern the ini-tial linear part of the curve and contribute mainly to the attainment ofthemaximumbond stress,whereasmacrolocking determines the resid-ual bond stress that remains at the end of the bond stress–slip curve.The stress–slip curve can assume three different shapes as discussedin [20], depending on the relative contribution of macrolocking uponthe loss of bond. Itmay display amaximum followedby a falling branch,a maximum followed by a falling branch, which raises again at largeslips, or display nomaximum at all, as was the case for all tests reportedin [13].

Recent research to investigate the contribution of microlocking tothe interface bond stress has been conducted by Chen et al. [24], whoassessed the contribution of microlocking by comparing the ultimateaverage bond stress achieved by normal and lubricated specimenssimilarly to previous investigations [8,13]. Based on the test resultsreported in [13,24], the ratios of microlocking to bond stress are32%–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 frommicrolocking) specimens attained approximately half the ultimatebond stress of the non-lubricated ones. However, studies into thecontribution of macrolocking are rather limited and warrant furtherinvestigation.

The main objective of this paper is to assess the contribution of thevarious components of the bond strength (i.e. chemical adhesion,microlocking, macrolocking) in the case of concrete filled RHS withthe focus being on the contribution of macrolocking. Six rectangularCFST columns have been tested using the load-reversed push-out testmethod with one specimen lubricated at the interface. Details of the in-terface bond stress distribution through four half-cycles of loading aregiven. The microlocking contribution is derived from the relative ratioof bond stress between the normal specimens and the lubricated spec-imen, recorded at the first loading stage. Themacrolocking contributionis obtained by comparing the ultimate bond stresses from the first andthird half-cycles of loading. Based on the test data, the effect of interfacelength 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 wasstudied to provide guidance for practising engineers.

2. Experimental study

2.1. General

A total of six rectangular CFST specimens were prepared and testedunder load reversal. Table 1 provides a summary of the test specimens.Specimens CP1–CP5 had essentially similar properties. Specimen CP6was slightly longer and was filled with concrete of slightly higherstrength. The specimen length was selected such that the specimensbehave essentially as stub columns with no influence from memberslenderness, whilst still including a representative distribution of initialgeometric imperfections along the length, thus allowing macrolockingto develop. The interface length was selected as the nominal specimenlength less approximately 60 to 80 mm to allow an adequate space forthe concrete core to displace inside the specimens. The interface lengthof 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 labellingconvention is shown in Fig. 2.

Table 1Specimen 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 lubricationCP2 300 200 5.67 800 733 C40 Q235 No lubricationCP3 300 200 5.67 800 726 C40 Q235 No lubricationCP4 300 200 5.67 800 718 C40 Q235 No lubricationCP5 300 200 5.67 800 713 C40 Q235 LubricationCP6 300 200 5.67 900 843 C50 Q235 No lubrication

37X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 35–43

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 twoends were parallel to each other and normal to the sides. Any depositsof 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 ofthe concrete core. The concrete was cast within the steel tubes, leav-ing approximately 50–100 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 thetests. For each batch, six 100 mm concrete cubes were cast andcured in standard laboratory conditions for 28 days [25]. After thatperiod, experiments to measure the concrete elastic modulus andconcrete compressive strength were conducted. The average valuesof the material parameters are shown in Table 2.

The material properties of the steel tubes were obtained by meansof tensile coupon tests. Four coupons (see Fig. 2) were tested in accor-dance with [26]. The key average measured properties are shown inTable 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 thetesting machine in the vertical position with the air gap at the bottom.A layer of sand was first spread on the top surface of the specimens.Then, a steel block, which had a slightly smaller cross-section thanthat of the concrete core, was placed on the top of the specimen.This ensured that the load was applied only on the concrete coreand 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

y

Flat coupon

y

z

z

Weld

B

D

t

Fig. 2. Section labelling convention and location of flat tensile coupons.

the top end was measured by means of two linearly varying displace-ment transducers (LVDTs) placed at the two sides of the specimen.The deflection of the concrete core was then taken as the average ofthe two transducer values. In order to study the strain distributionin the steel hollow section, strain gauges were fixed on the narrowand wide faces of the steel tubes along the longitudinal length ofeach specimen. The spacing between two adjacent strain gaugeswas 100 mm. Since the strain may vary significantly near to the load-ed end, two strain gauges, placed 50 mm from the loaded end, wereadded to the two faces of the steel tube. Details of the strain gaugesfor specimen CP1 are shown in Fig. 4.

During testing, load was applied at the top of the specimen to theconcrete 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 deflec-tions were recorded at every 2 kN step. Once the specimen startedto show a marked change in the slip between the steel tube and con-crete core, as indicated by the LVDTs, the recordings were taken afterevery 2 mm of movement of the concrete core. Loading continueduntil the recorded load showed little change for large increases ofslip. At this point, the first 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 recordedload showed little change for large increases of slip. Then, the speci-men was reversed again to apply a further cycle. Each specimen wassubjected to 4 half-cycles of loading.

2.5. Experimental results

For each test, the slip between the steel and the concrete core wasinitially linearly related to the increments in loading. At a certainstage, slip developed more rapidly and the interface carrying capacitywas 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 stateof the concrete contact surface was examined. A general view of thetop and bottom contact surface of the specimens observed after test-ing is shown in Fig. 5. It may be seen that the concrete had beenslightly crushed in some local contact surface regions.

For design purposes, the average interface bond stress τ is used torepresent the interface bond strength. This is determined from:

τ ¼ N=CLi ð1Þ

in which N is the push-out load value, C is the perimeter of the con-crete section in contact with the steel tube and Li is the interfacelength.

Table 2Measured concrete properties.

Concrete strengthgrade

Young's modulus EC(MPa)

Compressive strength (cube) fcu(MPa)

C40 29,380 39C50 38,070 49

Table 3Key material properties from tensile coupons tests.

Specimen Young's modulusES (MPa)

Yield stressfy (MPa)

Ultimate tensilestrength fu (MPa)

RHS 300×200×5.67 (Q235b) 224,400 257.6 512.3

1

300

200

1

2

2

200 300

(a) CP1 cross-section

(b) CP1 1-1 cross-section

(c) CP1 2-2 cross-section

Fig. 4. Specimen details: CP1 (dimensions in mm).

38 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 35–43

Values of τ, calculated using Eq. (1), were used to determine theinterface bond stress (τ) versus slip (S) curves of Fig. 6. As expected,the peak bond stress under the first application of loading is typicallyhigher than under subsequent loading. Additionally, two somewhatdifferent forms of curve may be observed: for the first, second andfourth half-load cycles, the curves of all the specimens have thesame feature, with an initial linear portion followed by a transitionalportion. After reaching the peak strength, a rapidly declining portionappears before a relatively stable residual strength is attained. For thethird load reversal, the curves of all the specimens are characterizedby the lack of any falling branch. The reason for the behaviour inthe third half-load cycle being different from the others is explainedin Section 3.4.

Fluctuation may be observed in some of the bond stress–slipcurves, especially for specimen CP5. A possible explanation for thehigher fluctuations in the case of CP5 is that lubrication on the insideof the tube may not have been entirely uniform. For all other speci-mens, fluctuations may be attributed to the nonhomogenous confin-ing pressure exerted by the tubes on the concrete core, due to localirregularities of the corner surface of the steel tubes, causing extramicrolocking in these regions.

The ultimate average bond strength and its corresponding slip for allspecimens at every load reversal are shown in Table 4. For each speci-men, τun (n=1, 2, 3 or 4) is the ultimate average bond strength at thenth (n=1, 2, 3 or 4) load reversal, as derived from Eq. (1), where Nun

is the peak load in the nth curve, τun=Nun/CLi. For all specimens, theslip at the peak load in the nth half-cycle of loading is Sun (n=1, 2, 3or 4). For specimen CP5, although the peak interface bond strength of0.112 MPa can be observed after approximately 10 mm of slip, at thefluctuation of the first curve, the first peak point at the bond strengthof 0.106 MPa and lower slip is believed to be a more relevant value. Itshould be noted that all τ–S curves displayed a lack of a maximum inthe third half-cycle, similarly to some results reported in [8,13,20].Hence Su3 is the displacement which corresponds to the maximumrecorded average stress during the third half-cycle of loading. Althoughthis is not indicative of the ultimate bond stress of the specimens, it stillprovides some useful information about the behaviour of the speci-mens. In this case it is the shape of the load–slip curve rather than theactual value of the τu3 that is significant.

Reaction frame

Load cell

LVDTs Hydraulic jack

Rigid base

Air gap Specimen

Fig. 3. Set-up for the test specimens.

3. Analysis of test results

3.1. Bond mechanisms

As discussed in Section 1, the interface bearing capacity arisesfrom three distinct mechanisms: chemical adhesion, microlockingandmacrolocking. Chemical adhesion is generated by capillary action,which arises during the cement hydration process. According toprevious research [22] chemical adhesion makes only a minor contri-bution to the bond strength and is influenced by many factors, such asthe cement content and water–cement ratio. Microlocking, is causedby the roughness of the bond interface from the micro-perspective[2,13,8,20]. Macrolocking, also referred to as interface friction, arisesdue to the manufacturing tolerances associated with variations ofthe internal dimensions of the tube [2,13]. A graphical representationof each component is shown in Fig. 7.

During load-reversed push-out tests, the contribution of each of theaforementioned factors to the bond strength is different for differentload reversals and load levels. Generally, at the beginning of the firstloading cycle, there is no slip between the steel tube and the concretecore 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 chemicaladhesion is damaged and microlocking starts to engage. As the load isfurther increased, the interface slippage grows and begins to move to-wards the middle part of the interface length, the chemical adhesiononly exists in the areas without slippage, whilst the interface bondstrength 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 theareas of greater slip. At this stage, the push-out load equals the interfacecarrying capacity and the interface bond strength results mainly frommicrolocking and macrolocking. As slip increases further, the effect ofthe roughness of the bond interface reduces and themicrolocking contri-bution begins to decline. Eventually, the interface bond strength is duesolely tomacrolocking. After the first 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 providedalmost entirely from macrolocking under all subsequent load reversals.

3.2. Bond stress distribution

To understand the bond mechanisms, it is necessary to check thebond stress distribution along the length of the tubes. The bond stressescan be obtained from the axial strain gradients in the steel tube; theseare related by statics to the interface bond stress distribution [2]. If Nx

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

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

39X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 35–43

is the axial load in the steel tube at location x, then the bond stress, τ(x),can be obtained from:

τ xð Þ ¼ dNx

2 B� 2tð Þ þ 2 D� 2tð Þ½ �dx : ð2Þ

Figs. 8 and 9 illustrate the bond stress distributions on the narrowand wide faces of the steel tube for specimen CP1 at different loadlevels. The bond stress distribution is not uniform along the tubelength, but the value of local bond stress increases as the loadincreases until the ultimate load is reached. Some local calculatedbond stress values were zero in Figs. 8 and 9. These conditions usually

(a) CP1

(c) CP3

(e) CP5

0 10 20 30 40 50 60 70

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

First half-cycle

Third half -cycle

Second half-cycleFourth half-cycle

Bon

d st

ress

τ (M

Pa)

Slip (mm)

0 20 40 60 80 100

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15First half-cycle

Third half-cycle

Second half-cycleFourth half-cycle

0 10 20 30 40 50 60 70

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

First half-cycle

Third half -cycle

Fourth half-cycleSecond half-cycle

Slip (mm)

Slip (mm)

Bon

d st

ress

τ (M

Pa)

B

ond

stre

ss τ

(MP

a)

Fig. 6. Bond stress τ versus slip S curves from load

occur at the bottom of the steel–concrete interface and may beinfluenced by the pinching effect, which causes the steel tube todeform outwards at this location.

3.3. Influence of different parameters

As discussed above, the interface bond carrying capacity includedall three bond mechanisms only in the first load cycle. Therefore, thestudy of the effect of different factors on bond strength in this sectionrefers to their influence on the ultimate bond strength under the firstload application. As shown in Fig. 6, the slip initially varies linearlywith load. To illustrate this more clearly, the bond stress versus slip

(b) CP2

(d) CP4

(f) CP6

ττ

τ

0 10 20 30 40 50 60 70

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

First half-cycle

Third half -cycle

Second half-cycleFourth half-cycle

Slip (mm)

0 10 20 30 40 50 60 70 80

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

First half-cycle

Third half-cycle

Fourth half-cycleSecond half-cycle

Bon

d st

ress

(M

Pa)

Bon

d st

ress

(M

Pa)

0 10 20 30 40 50 60

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

First half-cycle

Third half-cycle

Second half-cycleFourth half-cycle

Bon

d st

ress

(M

Pa)

Slip (mm)

Slip (mm)

-reversed push-out tests for each specimen.

Table 4Results from load-reversed push-out tests.

Specimenreference

Su1(mm)

τu1(MPa)

Su2(mm)

τu2(MPa)

Su3(mm)

τu3(MPa)

Su4(mm)

τu4(MPa)

CP1 0.997 0.305 0.805 -0.192 52.9 0.166 0.38 -0.203CP2 0.973 0.271 1.09 -0.144 57.9 0.140 1.09 -0.130CP3 0.881 0.282 0.436 -0.181 62.1 0.152 1.01 -0.131CP4 1.15 0.259 0.938 -0.188 42.2 0.143 2.42 -0.129CP5 3.56 0.106 0.503 -0.126 69.2 0.053 0.300 -0.103CP6 0.841 0.319 5.84 -0.135 40.1 0.11 1.75 -0.103

40 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 35–43

relationship for the first load application to specimens CP1 to CP6 isshown in Fig. 10 with a 10 mm range on horizontal axis. An initial lin-ear portion followed by a transitional ascending portion appears forall curves. It should be noted that the transitional ascending portionof CP5, for which the interface is lubricated, is 2.99 mm, which islarger than that of the other specimens. This tendency is consistentwith the results observed by Shakir-Khalil [8] and Virdi and Dowling[13], who also observed the initiation of nonlinear response and thesubsequent reduction in stiffness at lower load levels for lubricatedspecimens.

In Fig. 10, it can be seen that the average ultimate bond stress ofCP6, which had a higher concrete compressive strength of C50, isslightly higher than that of the other specimens. It may hence beconcluded that for the specimens considered the concrete strengthdoes have a degree of influence on the bond strength. However asdiscussed in Section 1, there is a lack of consensus regarding the effectof concrete grade on bond strength, due to the fact that the beneficialeffect of increased concrete strength is usually accompanied by thedetrimental effect of increased shrinkage.

3.4. Development of macrolocking

After the first half-cycle of loading, the chemical adhesion isdestroyed and the microlocking is now at an almost negligible level.As a consequence, after the first load half-cycle, it can be assumed thatthe interface bearing capacity is provided only by macrolocking.Macrolocking is generated by changes in the internal diameter of thetube and refers not only to the constraining force but also to the frictionbetween the steel tube and the concrete core. Xue and Cai [22]suggested that the constraining force varies as a function of thecross-sectional diameter and the slip between the steel and concrete,whilst, at the same time, the friction coefficient gradually decreaseswith increasing load cycles.

With reference to Fig. 6, it may be observed that the second andthe fourth half-cycle τ–S curves of all specimens resemble eachother in shape. This verifies the aforementioned assumption, thatthe interface bond strength established after the first half-cycle canbe entirely attributed to the macrolocking effect. Furthermore, the ul-timate interface bond strengths observed in the second half-cycle areslightly larger than those observed in the fourth half-cycle. This indi-cates that the friction coefficient reduces with increasing cycles due

(a) Chemical adhesion (b) Microl

Fig. 7. Interface bond s

to smoothing of the interface associated with deterioration of friction,though this reduction seems to be relatively small.

The third half-cycle exhibits a notably different response on therest loading half-cycles and displays a lack of maximum averagebond strength. Hence it is both qualitatively and quantitatively differ-ent to the second and the fourth half-cycles, despite the fact that in allthree of themmacrolocking governs the response. This may be attrib-uted to the constraining force alterations in response to change in theload direction, due to a trend of increasing or decreasing diameterwith length, as shown, in an exaggerated fashion, in Fig. 11. For thecase illustrated, in the first half-cycle, the constraining force willreduce with increasing load and the concrete will slip more easily asa result. However, when the load direction is reversed, the concretewould be wedged into a narrowing funnel-like shape and the cons-training force would, as a consequence, increase.

3.5. Relative contribution of each bond mechanism

Based on the composition and variation in the interface bearingcapacity, it can be assumed that the ultimate average bond stress inhalf-cycle 1 τu1 for the lubricated specimen CP5 is derived, for themost part, frommicrolocking and that the ultimate average bond stressin half-cycle 3 τu3 for normal specimens is largely derived frommacrolocking. Since in the third half-cycle the load is in the same direc-tion and orientation as in thefirst half-cycle, themacrolocking contribu-tion in the first load application may, therefore, be approximated by τu3for normal specimens.

Specimens CP1–CP5 have the same cross-section, concrete gradeand steel grade characteristics, and although the interface length variesslightly, this is unlikely to influence the interface bond stress for thesespecimens. As a result, their recorded test data can be used to analysethe contribution of each bond stress component without any variationin other parameters. Comparing τu1 for specimen CP5 (with lubrication)with that of the τu1 for normal specimens (CP1–CP4), the microlockingbond strength was found to be about 40% of the average ultimate bondstrength. Comparing τu3 and τu1 for specimens CP1 to CP4, the averagecontribution of macrolocking was found to be approximately 55%.Hence it can be concluded that the bond strength exhibited by theconcrete-filled RHS is mainly attributed to macrolocking and micro-locking, whilst the effect of chemical adhesion appears to be verylimited.

3.6. Critical shear force transfer length

In practice, CFST columns may be loaded through the concretecore alone, by connection to the steel tube only or by a combinationthereof, for example with through-plate connections. Shear connec-tors may also be present if required.

The ultimate bond capacity Nu may be given by:

Nu ¼ Ls2 B� 2tð Þ þ D� 2τð Þð Þτu ð3Þ

where τu is the average ultimate bond strength, taken herein as τu1 and Lsis the shear force transfer length. Assuming that the capacity of the two

ocking (c) Macrolocking

tress components.

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

0 100 200 300 400 500 600 700 8000.0

0.5

1.0

1.5

Distance away from the loaded end (mm)

N=0.383Nu

N=0.525Nu

N=0.755Nu

N=NuB

ond

stre

ss o

f th

e st

eel n

arro

w f

ace

τ(M

Pa)

0 100 200 300 400 500 600 700 8000.0

0.1

0.2

0.3

0.4N=0.276Nu

N=0.525Nu

N=0.756Nu

N=Nu

Bon

d st

ress

of

the

stee

l nar

row

fac

e τ(

MP

a)

Distance away from the loaded end (mm)

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

0 100 200 300 400 500 600 700 8000.0

0.1

0.2

0.3N=0.292Nu

N=0.523Nu

N=0.705Nu

N=Nu

Bon

d st

ress

of

the

stee

l nar

row

fac

e τ(

MP

a)

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

0

1

2

3

4

N=0.307Nu

N=0.525Nu

N=0.661Nu

N=0.789Nu

N=Nu

Bon

d st

ress

of

the

stee

l nar

row

fac

e τ(

MP

a)

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.

41X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 35–43

components of the CFST column is given by their respective yield orcrushing resistances, the minimum failure load Nf is:

Nf ¼ min Ac f c;As f sn o

ð4Þ

whereAc is the cross-sectional area of concrete core,As is the cross-sectionarea of steel tube, fc is the compressive cylinder strength of concrete,taken as fc=0.79fcu ( fcu being the cube strength) and fy is the yieldstrength of the steel.

Hence, the critical length Ls,cr required for full shear transfer(i.e. Nu=Nf) can be obtained as:

Ls;cr ¼min Ac f c;As f s

n o

2 B� 2tð Þ þ D� 2tð Þð Þτu: ð5Þ

It should be noted that the aforementioned derivation is based onthe assumption that interface length has a negligible effect on bondstrength. As mentioned in Section 1, numerous researchers [8,13,20]have studied the effect of interface length on bond strength withoutbeing able to verify a direct link. In [13] it was concluded that forcircular CFST the contact length has no appreciable effect on bondstrength, whilst Shakir-Khalil [8] pointed out that the interface lengthdoes not seem to have a direct and proportional effect on the loadcarrying capacity of the specimens. Tao et al. [20] reported that thebond strength increased with increasing Li /B (Li being the interface

length and B the outer SHS width), whilst the opposite trend wasobserved for circular tubes. Hence it can be concluded that there isno consensus regarding the effect of interface length on the bondstrength for concrete-filled RHS, which indicates that the interfacelength cannot be considered a dominant factor affecting bond stress.

For specimens CP1–4 and 6, which have a D/t ratio of approxi-mately 50, the experimental average value of Ls,cr is 5445 mm. If thefull bond between the steel tube and the concrete core is to be reliedupon in design, then the member length adopted in practice shouldexceed Ls,cr or suitable measures to ensure effective load transfer,such as through-plate connections or shear connections, should beadopted.

4. Conclusions

Six load-reversed push-out tests have been conducted to investi-gate the bond mechanisms in rectangular CFST columns, and the rel-ative contribution of each component. Each specimen was subjectedto four half-cycles of loading. Bond stress versus slip curves wereobtained for each specimen. The axial bond strain distribution in thenarrow and wide faces of the steel tube, and along the entire lengthof the specimen at the point of the interface reaching its ultimateload-carrying capacity, and, in the case of specimen CP1, at a rangeof different load levels, have been successfully recorded. Based onthe experimental data, bond stress distributions in the narrow and

(a) First half-cycle

0 100 200 300 400 500 600 700 8000.0

0.2

0.4

0.6

N=0.383Nu

N=0.525Nu

N=0.755Nu

N=Nu

Bon

d st

ress

of

the

stee

l wid

e fa

ce τ

(MP

a)

Distance away from the loaded end (mm)

(b) Second half-cycle

0 100 200 300 400 500 600 700 8000.0

0.1

0.2

0.3

0.4

0.5

0.6

N=0.383Nu

N=0.525Nu

N=0.755Nu

N=Nu

Bon

d st

ress

of

the

stee

l wid

e fa

ce τ

(MP

a)

Distance away from the loaded end (mm)

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

0 100 200 300 400 500 600 700 8000

1

2

3

N=0.307Nu

N=0.552Nu

N=0.661Nu

N=0.789Nu

N=Nu

Bon

d st

ress

of

the

stee

l wid

e fa

ce τ

(MP

a)

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

0.0

0.3

0.6

0.9

1.2

1.5

N=0.292Nu

N=0.523Nu

N=0.705Nu

N=NuB

ond

stre

ss o

f th

e st

eel w

ide

face

τ(M

Pa)

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.

42 X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 35–43

wide faces of specimen CP1 were determined. The bond stress distri-bution was found to be non-uniform along the length of the tube. Byanalysing the average ultimate bond stress for each of the four halfcycles of loading (τu1, τu2, τu3 and τu4), the relative contributions ofthe three bond mechanisms were assessed. Macrolocking was foundto be the dominant mechanism contributing to bond strength,followed by microlocking, whereas the effect of chemical adhesionappeared to be limited. Finally, the critical interface length to ensure

0 2 4 6 8 100.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Slip (mm)

Bon

d st

ress

τ (M

Pa)

CP1

CP2

CP3

CP4

CP5

CP6

Fig. 10. Bond stress τ versus slip S curves for each specimen during the first half cycle ofloading.

full shear transfer was studied, and design recommendations weremade.

Symbols

Ac cross-sectional area of concrete coreAs cross-section area of steel tubeB width of the rectangular steel tubeC perimeter of the concrete section in contact with the steel

tubeD depth of the rectangular steel tubeEc modulus of elasticity of concreteEs modulus of elasticity of rectangular steel tubefc compressive cylinder strength of concretefcu compressive cube strength of concretefy yield strength of steelL length of the rectangular steel tubeLi length of the steel–concrete interfaceLs shear force transfer lengthLs,cr critical shear force transfer lengthN push-out load valueNf minimum failure load of steel tube or concrete coreNu ultimate interface bearing capacityN(x) axial load in the steel tube at location xS slipSu1 slip corresponding to ultimate push-out load in the first

half-cycle of loading

NN N

Fig. 11. Macrolocking action in CFST columns.

43X. Qu et al. / Journal of Constructional Steel Research 81 (2013) 35–43

Su2 slip corresponding to ultimate push-out load in the secondhalf-cycle of loading

Su3 slip corresponding to ultimate push-out load in the thirdhalf-cycle of loading

Su4 slip corresponding to ultimate push-out load in the fourthhalf-cycle of loading

t wall thickness of the steel tubeτ interface bond stressτu1 average ultimate bond strength in the first half-cycle of

loadingτu2 average ultimate bond strength in the second half-cycle of

loadingτu3 average ultimate bond strength in the first half-cycle of

loadingτu4 average ultimate bond strength in the fourth half-cycle of

loadingτ(x) bond stress at location x

Acknowledgements

The authors are grateful to the China Standard Management Groupfor the BG50017-2003 Structural Steel Design Code for their special com-posite structure research funding (GB 5000172010-04), and would liketo thank the Taian Kenuo profile steel stock company limited for thesupply of test specimens and Feng Qilei, Xu Xiu, Sun Guojun and otherstudents in the steel research group of Tianjin University for their assis-tance with the laboratory work.

References

[1] Giakoumelis G, Lam D. Axial capacity of concrete-filled tube columns. J ConstrSteel Res 2004;60(7):1049-68.

[2] Roeder CW, Cameron B, Brown CB. Composite action in concrete filled tubes.J Struct Eng ASCE 1999;125(5):477-84.

[3] O'Shea MD, Bridge RQ. Design of circular concrete filled thin-walled steel tubes.J Struct Eng ASCE 2000;126(11):1295-303.

[4] Morishita Y, Tomii M, Yoshimura K. Experimental studies on bond strength inconcrete filled square and octagonal steel tubular columns subjected to axialloads. Transactions of Japan Concrete Institute; 1979. p. 359-66.

[5] Tomii M, Yoshimura K, Morishita Y. A method of improving bond strength inbetween steel tube and concrete core cast in square and octagonal steel tubularcolumns. Transactions of Japan Concrete Institute, 2; 1980. p. 327-34.

[6] Yang H, Lam D, Gardner L. Testing and analysis of concrete-filled elliptical hollowsections. Eng Struct 2008;30(12):3771-81.

[7] Espinos A, Gardner L, Romero ML, Hospitaler A. Fire behaviour of concrete filledelliptical steel columns. Thin-Walled Struct 2011;49(2):239-55.

[8] Shakir-Khalil H. Push-out strength of concrete-filled steel tubes to push outforces. Struct Eng 1993;71(13):230-3.

[9] Shakir-Khalil H. Resistance of concrete-filled steel tubes to push out forces. StructEng 1993;71(13):234-43.

[10] Shakir-Khalil H, Hassan NKA. Push-out resistance of concrete-filled tubes.Proceedings of the sixth international symposium on tubular structures, Melbourne,Australia, 14–16 December 1994; 1994. p. 285-91.

[11] Kilpatrick AE, Rangan BV. Influence of interfacial shear transfer on behavior ofconcrete-filled steel tubular columns. ACI Struct J 1999;96(4):642-8.

[12] Petrus C, Hamid HA, Ibrahim A, Nyuin JD. Bond strength in concrete filled built-upsteel tube columns with tab stiffeners. Can J Civ Eng 2011;38(6):627-37.

[13] Virdi KS, Dowling PJ. Bond strength in concrete filled circular steel tubes. Compos-ite columns. CESLIC Report. CC11. London: Engineering Structures Laboratories,Civil Engineering Department, Imperial College, London; 1975.

[14] Morishita Y, Tomii M, Yoshimura K. Experimental studies on bond strength inconcrete filled circular steel tubular columns subjected to axial loads. Transac-tions of Japan Concrete Institute; 1979. p. 351-8.

[15] Tomii M, Yoshimura K, Morishita Y. A method of improving bond strength inbetween steel tube and concrete core cast in circular steel tubular columns. Trans-actions of Japan Concrete Institute, 2; 1980. p. 319-26.

[16] Morishita Y, Tomii M. Experimental studies on bond strength between squaresteel tube and encased concrete core under cyclic shearing force and constantaxial force. Transactions of Japan Concrete Institute, 4; 1982. p. 363-70.

[17] Aly T, Elchalakani M, Thayalan P, Patnaikuni I. Incremental collapse threshold forpush-out resistance of circular concrete filled steel tubular columns. J Constr SteelRes 2010;66(1):11-8.

[18] Han L-H, Yang Y-F. Influence of concrete compaction on the behavior of concretefilled steel tubes with rectangular sections. Adv Struct Eng 2001;4(2):93–100.

[19] Xu C, Chengkui H, Decheng J, Yuancheng S. Push-out test of pre-stressing concretefilled circular steel tube columns by means of expansive cement. Construct BuildMater 2009;23(1):491-7.

[20] Tao Z, Han LH, Uy B, Chen X. Post-fire bond between the steel tube and concrete inconcrete-filled steel tubular columns. J Constr Steel Res 2011;67(3):484-96.

[21] Gourley BC, Tort C, Denavit MD, Schiller PH, Hajjar JF. A synopsis of studies of themonotonic and cyclic behavior of concrete-filled steel tube members, connectionsand frames. Report No. NSEL-008. Urbana-Champaign: Newmark Structural Engi-neering Laboratory, Department of Civil and Environmental Engineering, Universityof Illinois; 2008.

[22] Xue LH, Cai SH. Bond strength at the interface of concrete-filled steel tubularcolumns: part I. Build Sci 1996;12(3):22-8 [in Chinese].

[23] Parsley MA, Yura JA, Jirsa JO. Push-out behavior of rectangular concrete-filled steeltubes. Composite and hybrid systems, ACI SP-196. Farmington Hills, Michigan: ACI;2000.

[24] Chen ZH, Qu XS, Wang XD, Sun RR, Li LM. Experimental study on the interface bear-ing capacity on concrete-filled square steel tube. J Harbin Inst Tech 2009;41(2):27-32.

[25] GB50152-92. Standardmethods for testing of concrete structures. Chinese Standard;1992.

[26] GB/T 228-2002. Metallic materials—tensile testing at ambient temperature. Chinesestandard; 2002.