Engineering Structures 151 (2017) 554–567

Contents lists available at ScienceDirect

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

Size effect in circular concrete-filled steel tubes with different

diameter-to-thickness ratios under axial compression
Wenjing Wang, Hua Ma, Zhenbao Li, Zhenyun Tang ⇑
Beijing University of Technology, Key Laboratory of Urban Security and Disaster Engineering of Education Ministry, Beijing, China

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

Article history: Axial compression tests of circular concrete-filled steel tubes with different diameters (219 mm, 426 mm,
Received 25 February 2017 and 630 mm) and ratios of tube diameter to steel thickness (55 and 88) were conducted to investigate the
Revised 11 August 2017 effect of size on the bearing capacity. The experimental results indicated that the peak nominal stress
Accepted 12 August 2017
decreased as the size increased, and the decrease in the nominal stress due to the size effect increased
at higher ratios of diameter to thickness. At the peak load moment, an increased specimen diameter cor-
responded to a decreased hoop stress in the steel tube as well as a decreased concrete strength due to the
Keywords:
confinement effect of the steel tube. When the ratio of diameter to thickness increased, the extent of
Concrete-filled steel tube
Size effect
reduction of the hoop stress and the confining effect of the steel tube influenced by the increasing spec-
Ratio of diameter to thickness imen size increased. However, the vertical stress in the steel tube was increased at increased size, and
Hoop stress in steel tube stress increases in the ratio of diameter to thickness improved the increase degree of the vertical stress of steel
Confinement effect tube due to the enlargement of specimen size. Hence, the vertical bearing capacity of the steel tube was
Bearing capacity affected by both the specimen size and the ratio of diameter to thickness. Based on the size effect law
(SEL) proposed by Bazant, and taken the effect of the ratio of diameter to thickness into consideration,
a size-dependent formula to evaluate hoop stress in the steel tube was developed. A size-related model
considering situations with different ratios of diameter to thickness was established in order to estimate
the bearing capacity of large-size circular concrete-filled steel tubes. The model and experimental results
showed good agreement.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction large-size structures, particularly concerning the effect of speci-

Concrete-filled steel tubes (CFTs) possess the superior mechan-
ical properties of high bearing capacity, good plasticity, and tough-
ness, with the advantage of convenient construction suitable for
modern engineering technology. CFTs also satisfy the structural
requirements for high-rise buildings, large-spans bridges, heavy-
load structures, and construction in harsh environments. Thus,
CFTs are widely used in high-rise and super-high-rise buildings,
industrial plants, long-span bridges, and underground structures,
providing good economic benefits and construction effects. As
the required bearing capacity increases, the component sizes for
CFT structures also increase. For example, the diameters of the
CFT columns used in the outer frame of the Jin Tower in Tianjin,
China, is 1200–1600 mm, and the maximum diameter of the CFT
columns used in the bottom layer of the Union Square building
complex in Seattle, WA, USA, is 3050 mm. The safety of these
⇑ Corresponding author.
E-mail addresses: s200904014@126.com (W. Wang), mahua@bjut.edu.cn
(H. Ma), lizb@bjut.edu.cn (Z. Li), tzy@bjut.edu.cn (Z. Tang).
men diameter on the confinement effect of the steel tube, has
become a very concerning issue. So far, studies on the bearing
capacity of CFTs have used specimens with diameters below 450
mm, with many focusing on the range of 100–200 mm [1], without
showing the existence of a size effect in CFTs. However, the exis-
tence of size effects is increasingly accepted in theory [2–7] and
has been confirmed by many experiments on plain concrete sam-
ples and concrete components, such as reinforced concrete (RC)
beams, RC columns, RC beam-column joints, and concrete confined
by fibre-reinforced plastic (FRP) [8–15].
The size effect on the bearing capacity of concrete columns has
been studied from two perspectives, analysing the effect of mate-
rial size on the behaviour of plain concrete under axial compres-
sion and the effect of component size on the confinement effect
of enclosing reinforcements.
With increases in specimen size, concrete develops many more
internal defects, increased heterogeneity, and quasi-brittleness;
the size effect clearly exists in plain concrete. The study of statisti-
cal size effect considers the random distribution of internal defects
of concrete; therefore, the corresponding study describes size

http://dx.doi.org/10.1016/j.engstruct.2017.08.022
0141-0296/Ó 2017 Elsevier Ltd. All rights reserved.
 W. Wang et al. / Engineering Structures 151 (2017) 554–567
effect based on statistical theory (1951). A deterministic size effect
is observed in the uniaxial compression of short concrete speci-
mens with non-symmetric stress fields (wall friction), attributed
to a global energy release during the buckling of inclined or trans-
verse bands of microslabs in axial splitting failure [19]. The effect
of size on the uniaxial compression of concrete has been exten-
sively investigated. Gonnerman [16] found and verified experi-
mentally the size effect on the strength of concrete. Axial
compression tests on concrete cylinders with diameters of 152–
914 mm, conducted by Blanks et al. [17], showed that increases
in diameter decreased the compressive strength of the cylinders,
with decreases in strength reaching 16%. Weibull [2] believed that
the various defects causing strength reduction in concrete were
randomly distributed and therefore proposed a statistical theory
for the size-effect distribution law. Sabnis et al. [18] considered
that the effect of size on concrete strength was influenced by many
factors, and performed a regression analysis on compressive con-
crete strengths as measured by 12 researchers. They proposed a
formula for calculating the compressive strength of concrete as a
function of size, considering the influences of concrete strength,
age, and curing conditions. Bazant et al. [19] demonstrated that
energy dissipation after peak loading was related to specimen size
because of the localization of strain softening damage, and thereby
proposed a size-dependent model of the compressive strength of
concrete based on fracture mechanics. Kim et al. [20] applied
Bazant’s size effect law (SEL) for tensile strength [21] to the com-
pressive strength of concrete and then derived a size-effect for-
mula for the compressive strength of concrete cylinders, called
the modified size effect law (MSEL) of Bazant, based on a statistical
analysis of 678 experimental data including 20 specimens from
Smadi et al. [22], 172 from Gonnerman [16], 26 from Blanks
et al. [17], 337 from Kesler [23], and 123 from Murdock et al.
[24]. Noguchi et al. [25] experimentally studied the effect of size
on the cylinder compressive strength of high-strength concrete
and provided two formulas for calculating the compressive
strength of concrete cylinders, considering the size effect based
on the statistical theory of Weibull [2] and the model of Kim
et al. [26], respectively. Sakino et al. [27] determined the size effect
reduction factor for the compressive strength of concrete cylinders
by a regression analysis of 26 test data from Blanks et al. [17]. Yi
et al. [28] experimentally investigated the compressive strengths
of concrete samples with different sizes, shapes, and pouring direc-
tions, obtaining new parameters for the MSEL formula by fitting
the test results. The existing results indicate that the uniaxial com-
pressive strength of concrete is decreased as the specimen size
increases; the decrease can be estimated by different theoretical
methods.
Because of the interaction between concrete and confining
materials such as steel and FRP, the size effect in confined concrete
is more complicated than that in plain concrete materials. For con-
crete confined by circular spiral lateral steel subjected to axial
compression, Kim et al. [20] analysed the existing test data and
concluded that decreasing the volumetric ratio of lateral steel
increased the effect of size on the compressive strength of the con-
fined concrete, and then proposed a formula for estimating the
compressive strength of confined concrete that considered the
interaction between the specimen size and the volumetric ratio
of lateral steel, based on the MSEL model. After conducting an axial
compression experiment on concrete specimens confined by circu-
lar and square lateral steel with the maximum specimen size of
800 mm, Li et al. [29,30] found that the peak stress of the confined
concrete decreased as the size increased, and degree of decrease
was enlarged as the volumetric ratio of lateral steel lessened. A
size-dependent stress–strain constitutive model of confined con-
crete under varied volumetric ratios of lateral steel was established
based on the SEL model [21]. An experimental study on concrete
555
confined by carbon FRP (CFRP), conducted by Akogbe et al. [31],
showed that the compressive stress of the confined concrete
decreased when the diameter was increased from 100 mm to
300 mm. Through axial compression tests on concrete confined
by aramid FRP (AFRP) of different sizes, Wang et al. [15] concluded
that increases in size caused decreases in the strength of AFRP-
confined concrete at the peak bearing capacity of the specimen;
the degree of decrease was increased as the confinement ratio of
AFRP decreased. A formula to estimate the strength of AFRP-
confined concrete considering the influence of both size and con-
finement ratio was established based on the SEL [21]. Elkadi
et al. [32] tested concrete cylinders under triaxial compression
and found that the triaxial compressive strength of concrete was
decreased for increased specimen size, and then fitted the size
effect formula for estimating the triaxial compressive strength
based on the SEL [19]. Previous research has shown that the peak
stress of confined concrete under axial compression is decreased
with increasing size, and degree of decrease is increased with
decreases in the confinement factor (such as the volumetric ratio
of lateral steel or the confining ratio of FRP). Formulas for the peak
stresses of various types of confined concrete under axial compres-
sion relating to the interaction of specimen size and confinement
factor have been obtained based on the basic SEL [19,21].
For the specific confined concrete type of CFT under axial com-
pression, research on the size effect is limited. Luksha et al. [33]
performed axial compression tests on 10 short CFT columns with
diameters ranging from 159 mm to 1020 mm, all with different
concrete strengths and ratios of diameter to thickness (D/t). The
results showed that increased column sizes corresponded to
changes in the failure mode of the columns from plastic to shear.
Experimental studies conducted by Yamamoto et al. [34] and Chen
et al. [35] found that the peak stress of circular CFT subjected to
axial compression was insensitive to specimen size increases for
D/t values of less than 50. Based on the analysis of a micro-plane
numerical model of CFTs with different D/t subjected to axial com-
pression, Caner et al. [36] speculated that the steel ratio of CFTs
which are affected by size ranges from 4% to 8%, or D/t of 50 to
100. Wang et al. [37] tested short cylindrical CFT columns with
the constant D/t of 88 and diameters ranging from 219 to
820 mm under axial compression. The nominal stress of the spec-
imen was decreased with increases in the specimen size, while the
tested hoop stress of the steel tube was decreased as the size
increased; therefore, the confinement effect on concrete provided
by the steel tube was subject to size effects. A size-related model
was proposed to predict the bearing capacity of CFTs. Existing
research has implied that, for large values of D/t, the peak stresses
of cylindrical CFTs subjected to axial compression experienced size
effects; however, when D/t is varied, the influence of interaction
between size and D/t remains unclear.
Actually, the columns used in the practice engineering are
always subjected to eccentric compression but not to uniaxial
compression. While the axial compression performance is the basic
performance of the column, the calculation of axial bearing capac-
ity is also the basic index of engineering design. The research on
the size effect of the compressive performance of the concrete-
filled steel tube (CFT) column is also helpful for the mechanism
analysis of the size effect on its performance under eccentric com-
pression, bending moment or even seismic action. Currently, the
studies of the size effect on the mechanical properties of CFT col-
umns are lack of strong scientific basis. Therefore, this paper
started to investigate the size effect on CFT columns from the basic
performance, namely axial compression performance. In this study,
the axial bearing capacities of cylindrical CFTs with different diam-
eters and D/t values were investigated experimentally. The diame-
ters of the specimens were 219, 426, and 630 mm; the D/t values
were 55 and 88. Under different D/t values, the size effects on
556 W. Wang et al. / Engineering Structures 151 (2017) 554–567
the stress of the steel tube and the confinement effect of core con-
crete provided by the steel tube were analysed. Finally, a model for
predicting the bearing capacity of cylindrical CFTs, considering
interactions of size and D/t, was proposed.
2. Experimental program
2.1. Material properties
The mechanical properties of the concrete used in the speci-
mens are shown in Table 1. Aggregates in the concrete included
medium-grain sand and gravel with the maximum aggregate size
of 25 mm. A superplasticizer was used and the mass density of
the concrete was 2406.5 kg/m3 while the water–cement ratio of
the concrete was 0.451. All specimens were cast from the same
batch of concrete, meaning that the aggregate gradation in all spec-
imens was equivalent. This is because the study is focused on the
size effect at the component level, rather than at the material level;
variations in aggregate gradation cause difficulties in separating
the component size effect from other factors, and impede analysis
of the experimental results. According to the Chinese standards for
the mechanical properties of concrete (GB/T 50081-2002) [38], the
average compressive stress of standard cubic concrete samples
measuring 150
150
150 mm is 59.78 MPa, and the average
compressive strength of standard prismatic concrete samples mea-
suring 150
150
300 mm was 49.64 MPa.
The mechanical properties of the steel tubes used in the test are
listed in Table 2. The steel grade used for the specimens was Q235.
According to the outer diameters and thicknesses of the steel
tubes, they were divided into six types as follows: 219
4.0 mm,
219
2.5 mm, 426
8.0 mm, 426
5.0 mm, 630
11.0 mm, and
630
7.0 mm. Regarding the production of the steel tubes, the
219-mm-diameter tubes were straight-seam tubulars, while the
others were spiral-weld tubulars. According to the Chinese stan-
dard for the mechanical properties of steel (GB/T 228.1-2010)
[39], three vertical samples were taken from each type of tube to
determine the mechanical properties, including the yield strength,
ultimate tensile strength, and elongation after fracture, as shown
in Table 2.
2.2. Specimen details
In the axial compression tests, six groups of 12 short cylindrical
CFT columns with three different diameters were cast from the
same batch of concrete to reduce dispersion in the test results.
Pumped commercial concrete was used in the specimens. After
steel plate was welded to the bottoms of the steel tubes, the con-
crete was cast vertically into the tubes by layers. Samples were
reserved during casting to obtain the mechanical properties of
the plain concrete. The specimens were steam-cured after casting.
After curing, the top surface of each column was plastered and then
welded to another piece of steel plate. The surface of each steel
tube was painted white to prevent rusting and permit convenient
observation of test phenomena.
The detailed design parameters of the 12 cylindrical CFT col-
umns in the test are listed in Table 3. The ratio of height to diam-
eter for all specimens is 3. The variable factors in the test were the
Table 1
Mechanical properties and mix proportions of concrete utilized in the CFT columns.
Prism compressive strength Cube compressive strength Cement
of concrete fc [MPa] of concrete fcu [MPa] density
[kgm3]
49.64 59.78 366
specimen diameter (219, 426, and 630 mm) and D/t (55 and 88).
The values listed in Table 3 are as measured; the meaning of the
specimen designation is as follows: the number 2, 4, or 6 before
D represents the specimen diameter of 219, 426, or 630 mm,
respectively; D/t55 and D/t88 means that the D/t of the specimen
is 55 and 88 respectively; the final -1 or -2 denotes the first or sec-
ond specimen in a group. The steel plates at the top and bottom of
the column were placed in order to keep the top and bottom sur-
faces of the specimen perpendicular to its vertical axis, thereby
ensuring that the steel tube and concrete were subjected to axial
compression loading at the same time. The steel plates also provide
confinement at the ends of the specimens. The thickness and side
length of the steel plate was increased with the diameter of the
specimen.
2.3. Loading and measuring
Size effect experiments on CFT columns under axial compres-
sion were conducted in a multifunctional electrohydraulic servo
loading system with the ultimate load of 100 MN at the Key Labo-
ratory of Urban Security and Disaster Engineering of the Education
Ministry at the Beijing University of Technology, China. The assem-
bling and loading of specimen is displayed in Fig. 1(a). The loading
device is diagrammed in Fig. 1(b). The two ends of the CFT column
are supported by spherical hinges to eliminate eccentricity in load-
ing. A force sensor is located between the upper spherical hinge
and the top of the specimen to measure the exerted axial load.
Continuous and monotonic axial compressive loading was applied
to the specimens. The load was controlled by force or displace-
ment. Before 90% of the bearing capacity was reached, the axial
load was controlled by uniform increases in force; after 90%, the
control method switched to uniform increases in displacement to
obtain accurate peak and decline phases. When the total axial
deformation of the specimen reached 6% of the specimen height,
the test was completed.
The arrangements of measuring points on the specimens are
illustrated in Fig. 1(b). An IMP static data acquisition system was
used for data collection in the testing. Measured items included
the axial load, axial deformation, and strain of the steel tube of
the specimen. The axial load was measured by force sensors. The
axial deformation of the specimen was measured by four linear
variable differential transformer (LVDT) displacement meters
arranged on the column surface with a gauge length comprising
the central 2/3 of the height of the specimen. The range of the
LVDTs used for specimens with diameters of 219 mm was
50 mm; it was 100 and 200 mm for specimens with diameters of
426 and 630 mm, respectively. 12 pairs of steel strain gauges were
located in the gauge length of the middle 1/3 height of the column
to measure the strain in the steel tube. Each pair included one hoop
strain gauge and one vertical strain gauge.
3. Experimental results and discussion
3.1. Observed behaviour
The failure process of each specimen was consistent with main
stages illustrated in Fig. 2, using specimen 6D/t88-2 as an example.
Water density Sand density Gravel Admixture Fly ash
[kgm3] [kgm3] density density [kgm3] density
[kgm3] [kgm3]
165 746 1031 6.5 92
 W. Wang et al. / Engineering Structures 151 (2017) 554–567 557

Table 2
Mechanical properties of steel tubes used in the CFT columns.

Specimen No. Outer diameter
thickness [mm] Yield strength of steel fy [MPa] Ultimate tensile strength fu [MPa] Elongation after fracture A [%]
2D/t55 219
4.0 352.1 450.0 21.4
4D/t55 426
8.0 276.2 386.0 29.5
6D/t55 630
11.0 269.1 393.8 29.9
2D/t88 219
2.5 590.4 619.5 3.7
4D/t88 426
5.0 259.8 418.4 24.4
6D/t88 630
7.0 276.0 437.9 29.1

Notes: Compressive stresses are positive values; tensile stresses have negative values.

Table 3
Geometric parameters of the CFT columns.

Specimen No. Outer diameter of specimen D [mm] Height of specimen H [mm] Thickness of steel tube t [mm] D/t
2D/t55-1 219.4 657 4.3 51.4
2D/t55-2 220.7 657 4.3 51.3
4D/t55-1 425.2 1278 7.9 53.8
4D/t55-2 426.1 1278 8.2 51.8
6D/t55-1 626.3 1890 11.2 56.1
6D/t55-2 626.0 1890 11.2 55.9
2D/t88-1 215.9 657 2.6 84.1
2D/t88-2 216.9 657 2.6 82.4
4D/t88-1 426.8 1278 5.1 84.2
4D/t88-2 425.7 1278 5.1 83.5
6D/t88-1 625.6 1890 7.2 86.9
6D/t88-2 632.1 1890 7.0 89.9

Notes: In the specimen number, the number of 2, 4, or 6 before D denotes that the diameter of specimen is 219, 426, or 630 mm respectively; D/t55 or D/t88 represents the D/t
value of 55 or 88; the last -1 and -2 denotes the first or second specimen in the same group with the same nominal geometric parameters.

Fig. 1. Loading device and measuring arrangement.

Stage 1 is the initial loading phase of the specimen, when the sur-
face of the specimen experiences no significant change. With
increased load, the peak load is reached at Stage 2, but the surface
of the specimen is similar to that at Stage 1, indicating that this
phase entails slight deformation of the specimen concentrated in
axial direction, but no formation of oblique cracks. The develop-
ment of axial deformation accelerates after entering the decline
phase. During Stages 3 and 4, the white paint begins to peel off,
and oblique cracks gradually form on the surface of the specimen.
The oblique fracture gradually penetrates the column, and local
buckling occurs on the surface of the steel tube during Stage 5.
Finally, the specimen loses its bearing capacity and fails at Stage
6. At the end of loading, the state of the specimen is as displayed
as Stage 6(a) in Fig. 2. After testing ends, the state of the specimen
after removing the peeled white paint on the specimen surface is
shown in Stage 6(b) in Fig. 2; the shape of the shear oblique frac-
ture zone in the specimen is clearly visible. The inclined macroc-
rack band formed in the post-peak stage of the specimen, shown
in Fig. 2, indicates the formation of an inclined band of microslabs
at the peak load, which causes a global energy release accompa-
nied by size effect [19].
Fig. 3 shows the ultimate failure mode of each specimen after
the test; all specimens experience shear failure, which is one of
the quasi-brittle failure modes. This indicates that, during the test,
the confinement provided by the steel tube is not enough to con-
vert the brittleness of the concrete completely into plasticity;
558 W. Wang et al. / Engineering Structures 151 (2017) 554–567

Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6(a) Stage 6(b)

Fig. 2. Failure evolution of specimen 6D/t88-2.

2D/t55-1 2D/t55-2 4D/t55-1 4D/t55-2 6D/t55-1 6D/t55-2

2D/t88-1 2D/t88-2 4D/t88-1 4D/t88-2 6D/t88-1 6D/t88-2

Fig. 3. Final failure patterns of all specimens.

therefore, the size effect remains present in the CFT columns D/t values, the hoop expansion is smaller, the shear fracture band
because of the quasi-brittle nature of concrete. Fig. 3 illustrates is narrow, and the main shear fracture is obvious. For specimen
that, for small D/t values, the hoop expansion at ultimate failure group 2D/t88, because the fracture elongation of the steel tube is
is large and the shear fracture band is wide and full. For larger small, cracking occurs in the steel tube at Stage 5. All cracks in
 W. Wang et al. / Engineering Structures 151 (2017) 554–567
group 2D/t88 appear in the shear oblique fracture zone but not in
the welding seam of the tube. This phenomenon does not occur in
the other groups of specimens, as shown in Fig. 3.
3.2. Main test results
Fig. 4 depicts curves of the axial load versus axial deformation
for all specimens. The axial load is measured by the force sensor,
while the axial deformation is the average of displacements mea-
sured by the four LVDTs. Fig. 5 shows the nominal stress–strain
curves for all specimens. The nominal stress is the ratio of the axial
load to the cross-sectional area of the specimen. The nominal strain
is processed in two parts. In Fig. 5, nominal strain before the peak
nominal stress is defined as the average of vertical strains of the
steel tube as measured by the 12 gauge points arranged on the
specimen; after the peak nominal stress, the nominal strain is
the average of the strains obtained by the four LVDTs, where each
strain measured by the LVDTs is the ratio of the measured dis-
placement to the corresponding actual gauge length. The shift in
calculation methods is used because the deformation before the
peak nominal stress is small, so the strain value obtained from
the strain gauges of the steel tube is more accurate than that
obtained by the LVDTs. After entering the decline section, the local
softening of the specimen is such that the average strain of LVDTs
reflects the state of the specimen more accurately. From Fig. 5, it
can be seen that the initial tangent modulus of specimens with
the same D/t value is essentially constant; with increased size,
the peak nominal stress decreases, while the peak nominal strain
varies little. Because the yield strength of the steel tube of group
2D/t88 is large, its peak nominal stress is near that of group 2D/
t55, while its peak nominal strain is significantly greater than that
of the other specimens with the same D/t.
The primary test results of all specimens at the moment of peak
bearing capacity are listed in Table 4. The detailed process to
obtain the stress of confined concrete (fcc_exp) is given in Section 3.4.
Nu is the bearing capacity calculated by the model proposed in Sec-
tion 4. Fig. 6 shows the relation between the peak nominal stresses
of all specimens and specimen size. Fig. 6 indicates that, for a con-
stant D/t value, the peak nominal stress of the specimen is
decreased as the size increases; when the D/t value of the specimen
increases, the degree of decrease in the peak nominal stress of the
specimens by the size effect is increased.
3.3. Size effect on the stress of steel tube
The bearing capacity of an empty steel tube is determined by its
stability. For larger D/t, the empty steel tube is more likely to
40
Axial compression load (MN)
30
20
10
0 0
0 10 20 30
Displacement in the length of 2/3H (mm)
(a)
Fig. 4. Axial load–displacement curves of all specimens.
559
collapse, namely because the effects of the initial imperfections
of the steel are greater. In a CFT, because of the support by the core
concrete, local buckling of the steel tube does not occur before the
bearing capacity of the CFT is reached (as shown in Fig. 2), so the
axial compression strength of the CFT is determined by the ulti-
mate strength of the material. In addition, the steel is homoge-
neous while concrete is anisotropic. The size effect of the steel
occurs at the microscale, but that of the concrete is in the mesos-
cale. In components composited from steel and concrete, the size
effect of steel imperfections is small enough to be neglected gener-
ally [12,29,34]. Therefore, the initial imperfections of the empty
steel tubes have little effect on the bearing capacities of CFTs with
different diameters, and tests on the axial compression perfor-
mance of empty steel tubes of different sizes for comparison were
not performed.
For the situation in which the concrete and steel tube in the CFT
are simultaneously subjected to uniaxial compression, after the
steel tube and concrete deform together, the steel tube of the
CFT is in the biaxial stress state of vertical compression and hoop
tension. The axial load bearing capacity of CFT is distributed over
the steel tube and core concrete. In order to separate the bearing
capacity of the core concrete from the total bearing capacity of
the CFT, the biaxial stress state of the steel tube must be acquired.
The vertical strain and hoop strain of the steel tube were measured
by the strain gauges arranged on the steel tube during the test.
Then the measured strain components of the steel tube were trans-
formed into the vertical compressive stress and hoop tensile stress
of the steel tube, using the method of Gu [40]. The method [40]
obeys the Von Mises yield criterion with details found in the liter-
ature [37].
In the method given by Gu [40] to calculate the component
stress in the steel tube, the equivalent stress-strain curve includes
the elastic stage, the plastic yield platform stage, and the plastic
strengthening stage. The stress and the strain value at each key
point in the equivalent stress-strain curve is taken from the
mechanical behaviour of the steel tube in the test, and the details
of the extraction and parameters can be easily found in the litera-
ture [37]. From known values of the equivalent stress-strain rela-
tionship and the measured data of vertical and hoop component
strain in the steel tube, the vertical and hoop stress in the steel
tube can be obtained by the following process: in the elastic stage
of equivalent stress-strain curve, the component stress in the steel
tube can be obtained by the generalized form of Hooke’s law; at the
initial point in the plastic yield platform stage of the equivalent
stress-strain curve, the component stress in the steel tube is deter-
mined by total deformation theory of the small elastic-plastic
deformation of the Ilyushin, the equation of the Von Mises
40
2D/t88-1
Axial compression load (MN)
2D/t55-1
2D/t55-2 2D/t88-2
4D/t55-1 4D/t88-1
4D/t55-2 30 4D/t88-2
6D/t55-1 6D/t88-1
6D/t55-2 6D/t88-2
20
10
40 0 10 20 30 40
Displacement in the length of 2/3H (mm)
(b)
560 W. Wang et al. / Engineering Structures 151 (2017) 554–567

120 120
2D/t55-1 2D/t88-1
2D/t55-2 2D/t88-2
100 100

Nominal stress (MPa)

Nominal stress (MPa)
4D/t55-1 4D/t88-1
4D/t55-2 4D/t88-2
80 6D/t55-1 80 6D/t88-1
6D/t55-2 6D/t88-2
60 60

40 40

20 20

0 0
0.00 0.01 0.02 0.03 0.04 0.00 0.01 0.02 0.03 0.04
Nominal strain (ε) Nominal strain ( ε)
(a) (b)
Fig. 5. Nominal stress–strain curves of all specimens.

Table 4
Primary test results at peak bearing capacity.

Specimen No. Peak load Nexp [kN] Peak nominal stress fcs [MPa] Stress of confined concrete fcc_exp [MPa] Nexp/Nu Average of Nexp/Nu
2D/t55-1 4234 112.03 109.69 0.96 0.97
2D/t55-2 4245 111.00 96.50 0.97
4D/t55-1 14469 101.90 94.60 0.96 0.98
4D/t55-2 15077 105.75 99.79 0.99
6D/t55-1 29463 95.64 86.43 0.98 0.98
6D/t55-2 29294 95.17 88.45 0.98
2D/t88-1 4030 110.08 103.06 1.06 1.04
2D/t88-2 4244 114.86 104.14 1.02
4D/t88-1 12360 86.38 80.83 1.03 1.00
4D/t88-2 13092 91.98 87.59 0.97
6D/t88-1 23663 76.98 71.64 1.13 1.09
6D/t88-2 26002 82.86 77.10 1.04

Notes: The peak load (Nexp) is measured from the present experiment. Peak nominal stress (fcs) is defined as fcs = Nexp/Acs, where Acs is the cross section area of specimen. The
process to obtain the stress of confined concrete (fcc_exp) from the test results refers to Section 3.4, as calculated by Eq. (1). Nu is the bearing capacity calculated by the model
proposed in Section 4.

120 120
Peak nominal stress fcs (MPa)
Peak nominal stress fcs (MPa)

110 110

100 100

90 90

80 80
Test value of group D/t55
Test value of group D/t88
Average test value of group D/t55
Average test value of group D/t88
70 70
0 200 400 600 800 0 200 400 600 800
Specimen diameter D (mm) Specimen diameter D (mm)
(a) (b)
Fig. 6. Peak nominal stress of all specimens with different sizes.

criterion and the numerical iterative method; in the plastic yield corresponding to each stage of the equivalent stress-strain curve
platform stage of the equivalent stress-strain curve, Von Mises cri- of the steel tube can be obtained by the method given by Gu [40].
terion and incremental theory of Prandtl-Lester flow rule are Fig. 2 shows that local softening has not occurred in the speci-
solved simultaneously to establish the solution of the component men at the peak bearing capacity (Stage 2). Therefore, the average
stress in the steel tube; in the plastic strengthening stage, the com- of the stress components of the steel tube from all gauge points
ponent stress in the steel tube is acquired through the equivalent uniformly distributed on the specimen can be taken to represent
stress value and the incremental theory of Prandtl-Lester flow rule. the biaxial stress state of the steel tube throughout the specimen,
Summing up the above, the component stress in the steel tube with hoop stress fsh_exp and vertical stress fsz_exp. Using specimen
 W. Wang et al. / Engineering Structures 151 (2017) 554–567 561

2D/t55-1 as an example, the hoop and vertical stresses at all gauge
points when the specimen reaches peak bearing capacity are listed
in Table 5, and the corresponding fsh_exp and fsz_exp acquired by aver-
aging are -229.28 and 140.36 MPa, respectively, where the com-
pressive value is positive and the tensile value is negative. In
Table 5, the position of gauge points arranged on the steel tube
is shown in Fig. 7. The distribution of hoop stress in the steel tube
of gauge points is shown in Fig. 8. The distribution of vertical stress
in the steel tube of gauge points is shown in Fig. 9. Table 5 implies
that the equivalent stress of all gauge points at the moment of ulti-
mate bearing capacity is mostly in the stage of plastic yielding,
indicating that the steel tube is in the yield state at this point.
In order to eliminate the influence of different yield strengths of
the steel tubes, the coefficient of the stress component of the steel
tube is defined as the ratio of the stress component of the steel
tube to the yield strength of the tube. Therefore, as the CFT reaches
ultimate bearing capacity, the test value of the hoop stress coeffi-
cient of the steel tube is defined as Gh_exp = fsh_exp/fy, and the test
value of the vertical stress coefficient of the steel tube is Gz_exp = -
fsz_exp/fy. The stress component coefficient of the steel tube in
each specimen at the ultimate bearing capacity acquired in the test
is listed in Table 6. Fig. 10 illustrates the relation between the spec-
imen size and the stress component coefficient of the steel tube at
the ultimate bearing capacity. Fig. 10 implies that, when the ulti-
mate bearing capacity is reached, specimens with equal D/t values
show decreased hoop stress coefficients of the steel tube (Gh_exp)
and increased vertical stress coefficients (Gz_exp) as the specimen
sizes increase. For increasing D/t values, the degrees of decrease
and increase for Gh_exp and Gz_exp, respectively, related to the influ-
ence of diameter are intensified.
3.4. Size effect on the confined concrete
Since the test values including the axial bearing capacity of the
specimen (Nexp), vertical stress of the steel tube (fsz_exp), cross-
sectional area of the steel tube (As, mm2), and cross-sectional area
of the core concrete (Ac, mm2) are known, the stress of the confined
concrete (fcc_exp) at the moment of the ultimate bearing capacity
can be determined by Eq. (1), with test results shown in Table 4.
Nexp  As f sz exp
f cc exp ¼ ð1Þ
Ac
The relation between fcc_exp and specimen size is plotted in
Fig. 11. The figure shows that, at the peak bearing capacity, the
stress of the confined concrete decreases with the increase of spec-
imen diameter; the extent of decrease increases with the increase
in D/t of the specimen. This implies that the interaction between
the size of the CFT and the ratio of confinement (D/t) is similar to
that of other confined concretes, such as concrete confined by lat-
eral steels or AFRP.
The stress of confined concrete consists of two parts: the mate-
rial strength of the concrete and the confinement strength of the
concrete provided by the steel tube. In order to compare the influ-
ence of size and D/t on these two parts, it is necessary to separate
the two parts from the stress of the confined concrete.
Firstly, the concrete material strength is essentially the cylindri-
cal compressive strength of the concrete (fcd). Section 1 showed
that research on the size effect for this part is abundant. In recent
studies on CFTs, a reduction factor is typically introduced and mul-
tiplied by the cylindrical compressive strength of concrete in order
to account for the effect of size on the strength of the concrete. For
example, Liang et al. [41] revised the size-effect reduction factor
proposed by Sakino et al. [27] and utilized the revised factor in cal-
culating the bearing capacity. However, the amount of test data
that Sakino et al. [27] used to propose the size-effect reduction fac-
tor was quite small and unlikely to accurately reflect the effect of
size on the strength of concrete. Comparing typical models, includ-
ing those of Weibull [2], Noguchi et al. [25], Bazant’s SEL [21], MSEL
of Kim et al. [20], and Sakino et al. [27], the MSEL model proposed
by Kim et al. [20] was based on the size effect law of Bazant [21]
and the relevant parameters were determined by the regression
analysis of much previously collected experimental data. There-
fore, the model has a comprehensive consideration of influencing
factors and high credibility. Hence, the MSEL of Kim et al. [20]
was utilized to consider the effect of size on the cylindrical com-
pressive strength of concrete in this experiment, and tests on the
axial compression performance of concrete cylinders of different
sizes were not conducted.
In addition, when the compressive strength of the concrete is
transferred from that of a prism measuring 150
150
300 mm
to that of a cylinder measuring 150
300 mm, a shape conversion
factor should be considered. In this scenario, the conversion factor
is 1.073 according to the recommendation of the American Con-
crete Institute in the literature [42]. Therefore, fcd can be obtained
by Eq. (2), and fcd of each specimen is illustrated in Fig. 11. In Eq.
(2), the number of 1.073 is a shape conversion coefficient of con-
crete compressive strength, and the other part except 1.073 in
Eq. (2) is the model to evaluate the compressive strength of cylin-
drical concrete considering size effect proposed by Kim et al. [20].
2 3
6 0:4 7
f cd ¼ 1:073
f c
40:8 þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5 ð2Þ
1 þ HD 50

Table 5
Stress components of steel tube of specimen 2D/t55-1 at ultimate bearing capacity.

Gauge points Hoop stress of steel tube Vertical stress of steel tube Equivalent stress of steel tube State of equivalent stress
No. [MPa] [MPa] [MPa]
G-1, 2 306.94 77.28 352 Stage of yield platform
G-3, 4 226.35 179.20 352 Stage of yield platform
G-5, 6 225.77 179.82 352 Stage of yield platform
G-7, 8 174.21 230.92 352 Stage of yield platform
G-9, 10 311.49 70.38 352 Stage of yield platform
G-11, 12 342.39 18.49 352 Stage of yield platform
G-13, 14 92.85 260.03 316.83 Up to 90% of yield strength of steel tube
G-15, 16 39.93 370.26 352 Stage of yield platform
G-17, 18 359.60 15.72 352 Stage of yield platform
G-19, 20 117.17 278.48 352 Stage of yield platform
G-21, 22 399.41 134.46 352 Stage of yield platform
G-23, 24 235.09 169.60 352 Stage of yield platform
Average -229.28 140.36 – –

Notes: Taking specimen 2D/t55-1 as an example, the stress components of the steel tube for the whole specimen (hoop stress fsh_exp, vertical stress fsz_exp) acquired by
averaging is fsh_exp = 229.28 MPa, fsz_exp = 140.36 MPa respectively, in which the value in compression is positive and the value in tension is negative.
562 W. Wang et al. / Engineering Structures 151 (2017) 554–567

Fig. 7. Distribution of strain gauges points of the steel tube in Table 5.

Fig. 8. Distribution of the hoop stress in the steel tube in Table 5.

Fig. 9. Distribution of the vertical stress in the steel tube in Table 5.

Table 6
Stress component coefficient of steel tube of each specimen at ultimate bearing capacity.

Specimen No. Gh_exp Average of Gh_exp Gz_exp Average of Gz_exp

2D/t55-1 65.1% 42.8% 39.9% 60.6%
2D/t55-2 20.5% 81.3%
4D/t55-1 40.0% 43.2% 70.5% 67.6%
4D/t55-2 46.3% 64.6%
6D/t55-1 28.2% 35.0% 81.0% 74.7%
6D/t55-2 41.8% 68.4%
2D/t88-1 72.8% 66.8% 42.8% 49.2%
2D/t88-2 60.8% 55.5%
4D/t88-1 31.6% 36.0% 76.6% 73.0%
4D/t88-2 40.3% 69.4%
6D/t88-1 43.1% 38.2% 68.5% 71.9%
6D/t88-2 33.4% 75.4%

Notes: At the moment of ultimate bearing capacity, the test value of hoop stress coefficient of steel tube is defined as Gh_exp = fsh_exp/fy, and the test value of vertical stress
coefficient of steel tube is Gz_exp = fsz_exp/fy, where fy is the yield strength of steel tube under axial tension listed in Table 2.

Then, the stress of confinement by the steel tube at the moment The obtained proportion of the components of the core concrete
when the bearing capacity is reached can be determined by sub- stress at the ultimate bearing capacity are illustrated in Fig. 12,
tracting the size-related cylindrical compressive strength of the con- which uses the average values of the two specimens per group.
crete from the stress of the confined concrete, namely fcc_exp  fcd. Fig. 12 shows that, at the moment the ultimate bearing capacity is
 W. Wang et al. / Engineering Structures 151 (2017) 554–567 563

1.0 1.0
Test value of D/t55
0.9
Average test value of D/t55
0.8 Average test value of D/t88 0.8
Test value of D/t88
0.7
0.6 0.6
Gθ

Gz
0.5
0.4 0.4
0.3 Test value of D/t55
0.2 0.2 Average test value of D/t55
Average test value of D/t88
0.1 Test value of D/t88
0.0 0.0
0 200 400 600 800 0 200 400 600 800
Specimen Diameter D (mm) Specimen Diameter D (mm)

(a) Coefficient of hoop stress of steel tube (b) Coefficient of vertical stress of steel tube
Fig. 10. Stress component coefficients of steel tube at ultimate bearing capacity versus specimen size.

120 120

100 100
fcc_exp (MPa)

fcc_exp (MPa)
80 80

60 60

40 40
Test value fcc_exp of D/t55 Test value fcc_exp of D/t88
20 Average of fcc_exp of D/t55 20 Average of fcc_exp of D/t88
fcD of D/t55 fcD of D/t88
0 0
100 200 300 400 500 600 700 100 200 300 400 500 600 700
Specimen diameter D (mm) Specimen diameter D (mm)
(a) (b)
Fig. 11. Stress of confined concrete of specimens at ultimate bearing capacity versus specimen diameter.

reached, the size effect on the stress of confinement is much greater
than that on the strength of the concrete; when D/t increases, the
size effect on the stress of confinement of the steel tube is increased.
This may be attributed to the hoop deformation of the concrete,
which decreases with increased specimen size; the confinement of
the core concrete provided by the steel tube is generated by the
deformation of the steel tube with that of the core concrete. There-
fore, the confinement effect of the steel tube is decreased with the
decreased hoop expansion of the core concrete. When the hoop con-
fining ratio is decreased, namely when the D/t increases, the con-
crete becomes much more brittle, so that the degree of reduction
in confinement becomes greater because of the size effect.

100 100
100% 100% 96.3% 94.6% 100% 100% 96.4% 94.6%
92.4%
Strength distribution ratio
of confined concrete (%)

Strength distribution ratio

of confined concrete (%)

80 80
75.8%
67.5%
60 60
50.9%
40 40

20 20

D/t55 D/t88
0 0
0 200 400 600 800 0 200 400 600 800
Specimen diameter D (mm) Specimen diameter D (mm)
Confinement part Concrete material Concrete material Confinement part

(a) (b)
Fig. 12. Component proportions of stress of confined concrete at peak bearing capacity.
564 W. Wang et al. / Engineering Structures 151 (2017) 554–567
3.5. Size-dependent model of bearing capacity
During the initial loading of cylindrical CFT columns under axial
loading, Poisson’s ratio of concrete is less than that of steel;
thereby, the core concrete and steel tube experience axial com-
pression separately. When the vertical deformation is increased
under increased loading, Poisson’s ratio of concrete exceeds that
of the steel tube until a certain load. Then the steel tube generates
hoop tensile stress when it starts to deform with the hoop expan-
sion of the core concrete, hence providing triaxial compressive
confinement to the core concrete. Meanwhile, the vertical stress
of the steel tube is decreased because of the biaxial stress state
of the steel. The interaction between the steel tube and core con-
crete causes a significant improvement in the axial compressive
bearing capacity of the CFT, but also significantly complicates the
calculation of the bearing capacity.
The axial compressive bearing capacity of the cylindrical CFT is
composed of two parts, including the vertical bearing capacity of
the confined concrete and that of the steel tube. The above discus-
sion of the test results implies that the test factors of the size and
D/t of the specimen have significant effects on both parts of the
bearing capacity of the CFT. In order to consider the mutual influ-
ences of size and D/t in the calculation of the axial compressive
bearing capacity of the CFT, the two parts of bearing capacity are
modelled separately.
First, the confined concrete and its corresponding stress (fcc) is
modelled as two components of the cylindrical compressive
strength of concrete (fcd) and the strength of confinement (Kfr) at
the peak bearing capacity, as shown in Eq. (3):
f cc ¼ f cd þ Kfr ð3Þ
in which fcc is the stress of the confined concrete at the ultimate
bearing capacity (MPa), fr is the confining stress on the core con-
crete provided by the steel tube at the ultimate bearing capacity
(MPa), and K is the coefficient of confinement under triaxial com-
pression. The model considering the effect of size on the strength
of the concrete material (fcd) as a part of the stress of the confined
concrete was discussed in Section 3.4, as expressed in Eq. (2). The
analysis of the influence of size and D/t on the strength of confine-
ment in the modelling proceeds as follows.
The confining stress of the steel tube (fr) is given as follows:
2t
fr ¼  f ð4Þ
D  2t sh
f sh ¼ Gh f y ð5Þ
f sz ¼ Gz f y ð6Þ
where fsh is the hoop stress of the steel tube at the ultimate bearing
capacity (MPa), fsz is the vertical stress of the steel tube at the ulti-
mate bearing capacity (MPa), Gh is the coefficient of hoop stress in
the steel tube at the peak bearing capacity, and Gz is the coefficient
of vertical stress in the steel tube at the peak bearing capacity. It can
be seen from Eq. (4)–(6), as well as Section 3.3, that the influence of
size and D/t on the confining stress (fr) is mainly reflected in the
hoop stress of the steel tube, and more precisely on the coefficient
of hoop stress in the steel tube. Hence, it is necessary to establish
the relationship between the coefficient of the stress component
in the steel tube and factors including the specimen size and D/t.
When establishing a model for predicting the stress of confined
concrete considering the influence of specimen size and confining
ratio, models for other types of confined concrete generally
reflected the impact of the confining ratio based on the SEL model
proposed by Bazant [21], as with the models proposed by Kim et al.
[20], Song et al. [30], and Wang et al. [15]. Therefore, considering
the influence of D/t based on the size effect model of Bazant [21],
the relationship between the coefficient of hoop stress in the steel
tube (Gh) and factors including specimen size and D/t can be
obtained by a regression analysis of the test data in Fig. 10(a), as
given by Eq. (7). In addition, performing regression analysis on
the test data in Fig. 10(b) allows the acquisition of the relationship
between the coefficient of vertical stress in the steel tube (Gz) and
factors including specimen size and D/t, as given by Eq. (8). The
model of Gh and Gz is plotted in Fig. 13.
3:18  146 Dt
Gh ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi ð7Þ
1 þ 17:1 D
1  52:9 Dt
 2
780 t
Gz ¼ 0:75  1:5
þ 140 ð8Þ
D D
Although the coefficient of confinement under triaxial compres-
sion (K) is not directly related to the size and D/t of the specimen, it
is necessary to confirm the variation in K under different lateral
pressures. The test values of K (Kexp) can be obtained by substitut-
ing the known test parameters into Eqs. (2)–(6). Comparing the
commonly used models for calculating K, including the model of
Richart et al. [43], Zeng et al. [44] and Cai et al. [45] with the test
value Kexp, the results are displayed in Fig. 14. Among the models,
that provided by Richart et al. [43] views the relation between K
and the lateral pressure ratio (fr/fcd) as linear; Zeng et al. [44] show
a logarithmic nonlinear relationship between K and fr/fcd based on
their test data; Cai et al. [45] propose a nonlinear relation of K and
fr/fcd based on a large number of test data under tri-axial compres-
sion. The results from this experiment show that Kexp is nonlinear
with fr/fcd; the model provided by Cai et al. [45] is similar to the
trend of the experimental data, but the other models are very dif-
ferent. In order to obtain the nonlinear trend of the test data of K,
regression analysis is performed to yield the expression of K as
follows:
 0:62
fr
K ¼ 2:28 ð9Þ
f cd
Eqs. (6) and (8) imply that the effect of factors including the size
and D/t of the specimen on the vertical bearing capacity of the steel
tube is present in the coefficient of the vertical stress in steel tube
at the ultimate bearing capacity (Gz). According to the axial com-
pression mechanism of CFTs, the effect of size and D/t of the spec-
imen on the vertical stress in the steel tube is associated with the
effect on the hoop stress in the steel tube, but the effect trend in
the vertical stress is opposite that in the hoop stress.
According to the above analysis, substituting Eqs. (2)–(9) into
Eq. (10) yields the size-dependent model for predicting the axial
compressive bearing capacity of cylindrical CFTs for different val-
ues of D/t:
Nu ¼ As f sz þ Ac f cc ð10Þ
where Nu is the calculated axial compressive bearing capacity of
the cylindrical CFT (kN). This model is applicable to the circum-
stance of CFT with the concrete grade of C60, the steel tube grade
of Q235, and the D/t of 55–88. Eq. (10) concentrates on the bearing
capacity of short CFT columns, therefore a buckling coefficient for
the situation of long CFT column is not utilized here. Table 4 lists
the ratio of the experimental results (Nexp) to the model calculation
results (Nu). Fig. 15 illustrates the comparison between the exper-
imental and calculation results. Table 4 and Fig. 15 both indicate
that the values calculated by the model agree well with the exper-
imental values.
A comparison of the test data from similar specimens in refer-
ences [27] and [33] (listed in Table 7) with those of the model pro-
posed here is shown in Fig. 15. The calculated values of the
 W. Wang et al. / Engineering Structures 151 (2017) 554–567 565

(a) Coefficient of hoop stress in steel tube (b) Coefficient of vertical stress in steel tube
Fig. 13. The model of the coefficients of the stress components of the steel tube at ultimate bearing capacity.

30
Test value
specimens with concrete strengths of 15–28.9 MPa in the previous
Model of Cai [45] studies decreases obviously, supporting this point. According to
25
Model of Richart [43] the study of Cai [45], the proportion of the increase in strength from
Model of Zeng [44] confinement in this experiment remains within a reasonable range.
20 Eq. (9) Therefore, the adaptability of the model in this study when using
other test data still requires further study. Without regarding the
15
K

influence of the differences in concrete strength percentages

between this experiment and others in the literature, the compar-
10
ison ratios of specimens with different sizes in other studies are at
5 the same level, indicating that the model can evaluate the size effect
on the bearing capacities of specimens with different sizes.
0 The theoretical model proposed in this paper was established
0.00 0.05 0.10 0.15 0.20 0.25 by dividing each part of the bearing capacity based on the mecha-
f r /f cd nism of CFTs under axial compression, and then fitting the exper-
imental data to obtain the evaluation equation for each
Fig. 14. The coefficient of confinement under triaxial compression (K) as a function parameter in the model. The semi-empirical, semi-theoretical
of the lateral pressure ratio (fr/fcD) at ultimate bearing capacity. model is very simple, but it effectively validates and evaluates
the size effect on the axial compression behaviour of CFTs.
specimens are somewhat larger than the test results in other stud- The test results in Figs. 5 and 6 showed that the increase in the
ies. This is because the percentages of the concrete strength con- D/t increased the reduction degree of the peak nominal stresses of
tributing to the axial compressive strength of the specimens in the the CFT due to the increasing specimen size. This was because
referenced works are greater than that of the specimen in this paper. when the D/t increased, the reduction degree of the hoop stress
In Fig. 15, the ratio of the calculated to the experimental value of the in the steel tube due to the increase of specimen size at peak load
increased, thereby decreasing the confinement provided by the
steel tube to the concrete. The decrease of hoop stress in the steel
60
tube was the result of the decrease of the hoop deformation of the
concrete with the increase of the specimen size. Therefore, the size
50 effect on CFT is still the result of the concrete material instead of
the steel tube.
40 According to research results based on small-sized specimens, it
Nu (MN)

is debatable whether the strength contribution of confinement

30 should be considered when estimating the axial compressive bear-
2D/t55 4D/t55
6D/t55 2D/t88 ing capacity of CFTs. However, this study shows that, because of
20 4D/t88 6D/t88 the size effect on the concrete material, the confinement of the
[33] fcp>30MPa core concrete provided by the steel tube in large-sized CFTs is
[33] fcp<30MPa weakened. Moreover, the degree of weakening in the confinement
10
[27] fcp>30MPa is increased as the D/t increases, and the extent of weakening of the
[27] fcp=25.4MPa confinement far exceeds that of the strength of the concrete mate-
0
0 10 20 30 40 50 60 rial due to the size effect. Therefore, the design formulas based on
Nexp (MN) research of small-sized specimens do not accurately or safely eval-
uate the axial compressive bearing capacity of large-sized CFTs,
Fig. 15. Comparison between experimental and calculated results of axial com- and the interaction effect of factors including the size and D/t of
pressive bearing capacity. specimens should be considered in designing CFTs.
566 W. Wang et al. / Engineering Structures 151 (2017) 554–567

Table 7
Detail of specimens for comparison.

Specimen No. [27] D/mm D/t fcp/MPa fy/MPa Specimen No. [33] D/mm D/t fcp/MPa fy/MPa
CC4-A-2 149.0 50.3 25.4 308 SB2 630.0 90.0 36.0 291.4
CC4-A-4-1 40.5 SB6 82.8 35.0 349.5
CC4-A-4-2 SB7 74.6 34.5 350.0
CC6-C-2 239.0 52.6 25.4 507 SB3 61.7 38.4 323.3
CC6-C-4-1 238.0 52.4 40.5 SB4 54.3 46.0 347.2
CC6-C-4-2 SB5 820.0 91.8 45.0 331.0
CC6-D-2 361.0 79.5 25.4 525 SB9 1020.0 105.8 16.9 336.0
CC6-D-4-1 361.0 79.5 41.1 SB10 77.0 28.9 368.7
CC6-D-4-2 360.0 79.3 SB8 720.0 86.8 15.0 312.0
CC8-D-2 337.0 52.1 25.4 823
CC8-D-4-1 41.1
CC8-D-4-2

Notes: fcp is the cylinder strength of concrete. fy is the tensile yield strength of steel.

4. Conclusions Consequently, the vertical bearing capacity of the steel tube

In order to investigate the effect of size on the axial compressive
bearing capacity of cylindrical CFTs, axial compressive experiments
were conducted in which the ratios of diameter to thickness of
specimens (D/t) were 55 and 88, and the range of specimen diam-
eters was 219–630 mm. The results verified the existence of the
size effect in determining the axial compressive bearing capacity
of CFTs with different D/t, and the following conclusions are
reached.
(1) All specimens experienced shear failure at the end of the
loading tests. This indicates that, for D/t greater than 50,
although the confinement of the core concrete provided by
the steel tube still exists in the CFT, the core concrete
remains brittle, which is associated with the generation of
the size effect. When the size and D/t of the specimen are
larger, the size effect on the cylindrical CFT becomes stron-
ger; for example, the nominal stress of the CFT decreases
more. This is consistent with the results of theoretical anal-
ysis in previous reports [36].
(2) Based on the analysis of test stresses in the steel tube, the
bearing capacities of the steel tube and confined concrete
are separated from the bearing capacity of the CFT to study
the size effect on each component of the bearing capacity.
The results imply that, because of the size effect on the hoop
stress in the steel tube, with increases in the size and D/t of
the specimen, the degree of decrease in the confinement of
the core concrete provided by the steel tube is increased,
meaning a stronger size effect is observed. This is similar
to the size effect law of concrete confined by lateral steels
[29] and concrete confined by AFRP [15] under the influence
of different confining ratios.
(3) In addition, by comparing the compressive strength of
cylindrical concrete with the strength of confinement, the
effect of size and D/t of the specimen on the latter is signif-
icantly larger than that on the former. This implies that the
design theory of CFT components based on small-sized spec-
imens is likely to be unsafe when applied to large-sized
components. Therefore, it is suggested that the interacting
effects of the size and D/t of the specimen should be
considered in designing large-size components with D/t
exceeding 50.
(4) Unlike concrete confined by lateral steel or AFRP, the steel
tube in the CFT is subjected to hoop tension and vertical
compression stresses simultaneously. Therefore, the size
effect on the vertical stress in the steel tube opposes the size
effect on the hoop stress in the steel tube at the same time.
is influenced by both the size and D/t of the specimen. This
complicates the size effect law for large-sized CFTs.
(5) Based on the SEL model of Bazant and considering the effect
of D/t, the relationship between the hoop stress in the steel
tube and factors including the size and D/t of the specimen
was obtained to reflect the size effect on the confinement
of the core concrete provided by the steel tube. In this
way, a size-dependent model for evaluating the axial com-
pressive bearing capacity of cylindrical CFTs under the influ-
ence of different D/t values was established. The calculated
results agree well with the experimental data, providing
effective reference data for practical engineering.
Acknowledgments
This project was supported by the National Key Basic Research
and Development Program of China (2016YFC0701100) and the
Key Laboratory Open Fund of Urban Safety and Disaster Reduction
(2013ZD02).
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.engstruct.2017.
08.022.
References
[1] Gourley BC, Tort C, Denavit MD, Schiller PH, Hajjar JF. A synopsis of studies of
the monotonic and cyclic behavior of concrete-filled steel tube members,
connections, and frames. NSEL report series, No. NSEL-008; 2008.
[2] Weibull W. A statistical theory of the strength of materials. R Swedish Acad
Eng Sci Proc 1939;151:1–45.
[3] Carpinteri A, Chiaia B, Ferro G. A new explanation for size effects on the
flexural strength of concrete. Mag Concrete Res 1997;49(178):45–53.
[4] Duan K, Hu X, Wittmann FH. Scaling of quasi-brittle fracture: boundary and
size effect. Mech Mater 2006;38(1/2):128–41.
[5] Bazant ZP. Size effect. Int J Solids Struct 2000;37:69–80.
[6] Bazant ZP, Pang SD. Activation energy based extreme value statistics and size
effect in brittle and quasibrittle fracture. J Mech Phys Solids 2007;55:91–131.
[7] Bazant ZP, Yu Q. Universal size effect law and effect of crack depth on quasi-
brittle structure strength. J Eng Mech 2009;135(2):78–84.
[8] Turkel A, Ozkul MH. Size and wall effects on compressive strength of concretes.
ACI Mater J 2010;107(4):372–9.
[9] Caibal MRJ, Iwashita Y, Nakanishi M. Size effect on structural performance of
reinforced concrete beams under flexure. J Struct Constr Eng AIJ 2008;73
(626):625–32 [in Japanese].
[10] Belgin CM, Sener S. Size effect on failure of overreinforced concrete beams. Eng
Fract Mech 2008;2008(75):2308–19.
[11] Nemecek J, Bittnar Z. Experimental investigation and numerical simulation of
post-peak behavior and size effect of reinforced concrete columns. Mater
Struct 2004;37(3):161–9.
 W. Wang et al. / Engineering Structures 151 (2017) 554–567
[12] Jin L, Li D, Du XL. Mechanical behavior and size effect of moderate high-
strength RC columns under monotonic and cyclic axial compression. Eng
Struct 2016;124:269–85.
[13] Abrams DP. Scale relations for reinforced concrete beam-column joints. ACI
Struct J 1987;84(6):502–12.
[14] Barbhuiya S, Choudhury AM. A study on the size effect of RC beam–column
connections under cyclic loading. Eng Struct 2015;95:1–7.
[15] Wang YF, Wu HL. Size effect of concrete short columns confined with aramid
FRP jackets. J Compos Constr 2011;15(4):535–44.
[16] Gonnerman HF. Effect of size and shape of test specimen on compressive
strength of concrete. Proc ASTM 1925;25:237–50.
[17] Blanks RF, McNamara CC. Mass concrete tests in large cylinders. Proc ACI J
1935;31:280–303.
[18] Sabnis GM, Mirza SM. Size effects in model concrete. J Struct Div 1979;105
(6):1007–20.
[19] Bazant ZP, Xiang YY. Size effect in compression fracture: splitting crack band
propagation. J Eng Mech 1997;123(2):162–72.
[20] Kim JK, Yi ST, Park CK, Eo SH. Size effect on compressive strength of plain and
spirally reinforced concrete cylinders. ACI Struct J 1999;96(1):88–96.
[21] Bazant ZP. Size effect in blunt fracture: concrete, rock, metal. ASCE J Eng Mech
1984;110(4):518–35.
[22] Smadi MM, Slate FO. Micro-cracking of high and normal strength concrete
under short-and-long-term loadings. ACI Mater J 1989;86(2):117–27.
[23] Kesler CE. Effect of length to diameter ratio on compressive strength on ASTM
cooperative investigation. Proc ASTM 1959;59:1216–29.
[24] Murdock JW, Kesler CE. Effect of length to diameter ratio of specimen on the
apparent compressive strength of concrete. Proc ASTM 1957;221:68–73.
[25] Noguchi T, Tomosawa F. Effect of size and shape of specimen on mechanical
properties of high strength concrete in compression. J Struct Constr Eng, AIJ
1995;473:19–28 [in Japanese].
[26] Kim JK, Eo SH. Size effect in concrete specimens with dissimilar initial cracks.
Mag Concr Res 1990;42(153):233–8.
[27] Sakino K, Nakahara H, Morino S, Nishiyama I. Behavior of centrally loaded
concrete-filled steel-tube short columns. J Struct Eng 2004;130(2):180–8.
[28] Yi ST, Yang EI, Choi JC. Effect of specimen sizes, specimen shapes, and
placement directions on compressive strength of concrete. Nucl Eng Des
2006;2006(236):115–27.
[29] Li ZB, Song J, Du XL, Yang XG. Size effect of confined concrete subjected to axial
compression. J Cent South Univ 2014;21:1217–26.
[30] Song J, Li ZB, Wang YQ, Zuo ZY. Stress-strain constitutive model of concrete
confined by hoops with considering size effect. J Build Struct 2015;36
(8):99–107 [in Chinese].
567
[31] Akogbe RK, Liang M, Wu ZM. Size effect of axial compressive strength of CFRP
confined concrete cylinders. Int J Concr Struct M 2011;5(1):49–55.
[32] Elkadi AS, Van Mier JGM. Experimental investigation of size effect in concrete
fracture under multiaxial compression. Int J Fract 2006;140:55–71.
[33] Luksha LK, Nesterovich AP. Strength testing of large-diameter concrete filled
steel tubular members. In: Proceedings of the third international conference
on steel-concrete composite structures; 1991. p. 67–72.
[34] Yamamoto T, Kawaguchi J, Morino S. Experimental study of the size effect on
the behavior of concrete filled circular steel tube columns under axial
compression. J Struct Constr Eng, AIJ 2002;561:237–44 [in Japanese].
[35] Chen YJ, Li Y, Yan WM, Pan D. Bearing capacity test of large size concrete filled
steel tubular column. Chin J Highw Transp 2011;24(4):33–8 [in Chinese].
[36] Caner FC, Bazant ZP. Lateral confinement needed to suppress softening of
concrete in compression. J Eng Mech 2002;128(12):1304–13.
[37] Wang WJ, Tang ZY, Li ZB, Ma H. Bearing capacities of different-diameter
concrete-filled steel tubes under axial compression. Adv Mater Sci Eng
2016:1–10. Article ID 7982365.
[38] Ministry of Construction of the People’s Republic of China. GB/T 50081–2002,
Standard for test method of mechanical properties on ordinary concrete.
Beijing: China Architecture & Building Press; 2003 [in Chinese].
[39] General Administration of Quality Supervision, Inspection and Quarantine of
the People’s Republic of China. GB/T 228.1-2010, Metallic materials-Tensile
testing-Part 1: Method of test at room temperature. Beijing: Standards Press of
China; 2010 [in Chinese].
[40] Gu WP. Stress analysis of concrete filled steel tubular columns. Build Sci
1987;2:14–21 [in Chinese].
[41] Liang QQ, Fragomeni S. Nonlinear analysis of circular concrete-filled steel
tubular short columns under axial loading. J Constr Steel Res
2009;65:2186–96.
[42] Zhou JP, Niu XQ, Jia JS. Design of concrete gravity dams – A 20 year
perspective. Beijing: China Water & Power Press; 2008. p. 591 [in Chinese].
[43] Richart FE, Brandtzage A, Brown RL. A study of the failure of concrete under
combined compressive stresses. Bulletin No. 185. Urbana (Illinois):
Engineering Experimental Station, University of Illinois; 1928.
[44] Zeng SJ, Ren XD, Li J. Triaxial behavior of concrete subjected to dynamic
compression. J Struct Eng 2013;139(9):1582–92.
[45] Cai SH, Jiao ZS. Behavior and ultimate strength of short concrete-filled steel
tubular columns. J Build Struct 1984;6:13–29 [in Chinese].