Journal of Constructional Steel Research 140 (2018) 153–162

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Global buckling behaviour of welded Q460GJ steel box columns under

axial compression
Shao-Bo Kang, Bo Yang, Xiong Zhou, Shi-Dong Nie ⁎
a
Key Laboratory of New Technology for Construction of Cities in Mountain Area (Chongqing University), Ministry of Education, Chongqing 400045, China
b
School of Civil Engineering, Chongqing University, Chongqing 400045, China

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

Article history: This paper describes an experimental and numerical study on the global buckling behaviour of welded Q460GJ
Received 18 October 2016 steel box columns. In the experimental programme, seven steel columns with different cross sections and wall
Received in revised form 27 September 2017 thicknesses were tested under axial compression. The load capacity of steel columns was quantified. Compari-
Accepted 14 October 2017
sons were made between experimental results and design values calculated in accordance with national stan-
Available online 5 November 2017
dards. Furthermore, numerical models were established in which initial geometric imperfections and residual
Keywords:
stress distributions were considered. The model was validated against test data with reasonably good accuracy.
Q460GJ steel A parametric study was conducted on the effects of initial geometric imperfections and normalised slenderness
Box columns on the load capacity of box columns. Experimental and numerical results indicated that Q460GJ steel box
Axial compression columns could develop higher global buckling resistances than the values calculated from GB50017-2003 and
Global buckling Eurocode 3, but ANSI/AISC360-10 might not be safe for welded box columns with small width-thickness ratios.
Numerical simulations Therefore, the design approaches for conventional steel columns were modified so that the buckling behaviour
of box columns fabricated of Q460GJ steel could be accurately evaluated.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction that current design approaches for conventional steel members pro-
vided rather conservative results when used for high-strength steel
As a typical high-performance structural steel, GJ steel has been columns. Likewise, residual stresses in H and box sections fabricated
widely used for structural members in construction practices in China. of Q460 steel were reported and similar conclusions were drawn
Compared with high strength steel, it has relatively lower yield-to- from the experimental and numerical studies on overall buckling of
ultimate strength ratio, better impact toughness, lamellar tearing resis- high-strength steel columns conducted by Wang et al. [6–8]. As for
tances and weldability [1,2]. However, due to different processing tech- Q690 steel, design recommendations were provided based on exper-
niques and metallurgical structures, the mechanical behaviour of GJ imental and numerical results [9,10]. Besides global buckling, Q460
steel components may be different from those of high strength steel. high-strength steel column stubs were also tested to gain insight
To date, there is still a lack of test data on structural behaviour of GJ into local buckling behaviour [11]. Comparisons between numerical
steel members under various loading conditions. Thus, experimental results and calculated values from current design methods suggested
and numerical investigations are necessary in order to incorporate this that the methods were fairly conservative for I-shaped columns but
type of steel in current design codes and guidelines. unsafe for box columns.
Several experimental tests have been conducted on the buckling As another typical high-performance steel in China, GJ steel has sim-
behaviour of high-strength steel members. As for columns, Ban et al. ilar material properties to its counterpart in the United States, Japan and
measured the residual stress distribution in 460 MPa high-strength Europe. Available experimental studies mainly focused on the lateral-
steel columns [3] and studied the overall buckling of the columns torsional buckling resistance of steel beams. Yang et al. [12] investigated
through experimental tests [4]. It was concluded that the effects of the lateral-torsional buckling behaviour of I-shaped GJ steel beams
initial geometric imperfections and residual stresses on buckling resis- under a concentrated point load. Test results showed that the design
tance became insignificant with increasing yield strength of steel. methods in GB 50017-2003 [13] and ANSI/AISC360-10 [14] might not
In the meantime, 960 MPa strength steel columns were also investi- be conservative for global stability design of welded GJ steel beams.
gated by Ban et al. [5]. Experimental and numerical results indicated Furthermore, Yang et al. [15] tested the residual stress distribution in
welded Q460GJ steel I-beams and the results were later used for numer-
⁎ Corresponding author at: School of Civil Engineering, Chongqing University,
ical analyses on the lateral-torsional buckling of steel beams under a
Chongqing 400045, China. concentrated point load [16]. However, the global buckling resistance
E-mail address: [email protected] (S.-D. Nie). of GJ steel columns has not drawn sufficient attention.

https://doi.org/10.1016/j.jcsr.2017.10.013
0143-974X/© 2017 Elsevier Ltd. All rights reserved.
154 S.-B. Kang et al. / Journal of Constructional Steel Research 140 (2018) 153–162

B Q460GJ steel and with different width-thickness ratios and normalised

slendernesses, were tested under axial compression. Fig. 1 shows the
definition of symbols for a column section. Table 1 summarises the
dimensions of steel columns. In the designation, “B” represents box
tf columns; the first numeral denotes the nominal height of column sec-
TF tions, and the last is the nominal thickness of column web and flange.
It is noteworthy that column B-120-20a, with nearly the same dimen-
sions as B-120-12, was fabricated and tested to verify the reliability of
tw y
test setup.
Steel plates with thicknesses of 12 and 25 mm were used for box
columns. The width of columns with 12 mm thick plates ranged
from 120 mm to 264 mm. The corresponding width-thickness ratio
o x was in the range of 7.8 and 19.6. As for columns with 25 mm thick

H
LW RW
plates, the width-thickness ratio was from 4.8 to 7.9 due to the con-
straints of the testing machine. Steel columns with a wall thickness
of 12 mm were fabricated by using carbon dioxide welding, whereas
b' submerge arch welding was employed when the thickness was
increased to 25 mm. The column height was adjusted so that the
slenderness was in the range of 50 and 120. All steel columns were de-
signed to fail in global buckling prior to local buckling by controlling
BF the width-thickness ratio. The maximum width-thickness ratio of
column walls was less than 20, and thus local buckling of steel columns
was averted in accordance with GB 50017-2003 [13], as expressed in
Fig. 1. Definition of symbols for a column section. Eq. (1). It also satisfied the requirements in Eurocode 3 [17] and
ANSI/AISC 360-10 [14]. All the columns were classified as “c” in GB
50017–2003 and Eurocode 3.
This paper presents an experimental and numerical study on the
global buckling behaviour of Q460GJ steel box columns subject to sffiffiffiffiffiffiffiffiffi
0 235
axial compression. In the experimental programme, seven steel b =t ≤40 ð1Þ
fy
columns with various slendernesses and wall thicknesses were tested
to buckling. Corresponding axial load and horizontal displacement
were measured. Numerical models were developed and validated where b' and t are the net width and thickness of column walls, respec-
against experimental results. Based on the validated numerical model, tively, and fy is the yield strength of steel plates.
parametric study was carried out to investigate the effects of normalised
slenderness on buckling factor. Comparisons between numerical results 2.2. Test setup and instrumentation
and design curves suggest that design approaches in GB50017-2003
[13] and Eurocode 3 [17] are relatively conservative, whereas ANSI/ Fig. 2 shows the test set-up and instrumentation. To prevent prema-
AISC360-10 [14] may be inaccurate when used for box columns with ture failure, two strengthening plates with a thickness of 30 mm were
relatively small width-thickness ratios. Finally, design recommenda- welded to the ends of each column. Steel blocks with a groove were
tions were proposed for global stability design of Q460GJ steel box bolted to the column, as shown in Fig. 2(a). Besides, rigid arc supports
columns in accordance with experimental and numerical results. were designed and placed in the grooves to provide pin supports.
Thus, the column could rotate freely about the y-axis. A displacement-
2. Experimental programme controlled hydraulic testing machine was employed in the tests.
During testing, the vertical load applied to the column was measured
2.1. Specimen design by means of a load sensor. In addition, displacement transducers were
utilised to measure horizontal and vertical displacements, as shown in
The load capacity of a compression member is determined by differ- Figs. 2(a and b). Transducer DH1 to DH5 was erected along the column
ent column curves, in which the width-thickness ratio and slenderness height to measure its horizontal deflections. The axial deformation of
play an important role. The width-thickness ratio mainly affects the re- the column was recorded by DV1 and DV2. Moreover, inclinometer Q1
sidual stress distribution and buckling mode of steel plates. By varying was mounted at the testing machine to monitor its rotation. Another
the slenderness ratio, different buckling factors can be obtained. In the two inclinometers Q2 and Q3 were placed at both ends of the column
experimental programme, the effects of width-thickness ratio and nor- to measure the angles of rotation. Besides axial and horizontal deforma-
malised slenderness on buckling behaviour of box columns were taken tions, the strains of steel columns at the mid-height were also captured
into consideration. Seven box columns, fabricated of high-performance through strain gauges, as shown in Fig. 3.

Table 1
Measured dimensions of steel columns.

Specimen H (mm) tw (mm) B (mm) tf (mm) b'/t iy* (mm) L (mm) Le (mm) λ

B-120-12a 120.30 12.38 119.59 12.30 7.7 44.13 3114.00 3492.00 110
B-120-12 120.37 12.37 120.21 12.23 7.8 44.39 3115.00 3493.00 110
B-168-12 168.51 12.43 168.04 12.45 11.5 63.74 3733.50 4111.50 90
B-216-12 216.62 12.32 215.94 12.28 15.4 83.35 3801.50 4179.50 70
B-264-12 264.43 12.27 263.91 12.21 19.6 102.94 3306.00 3684.00 50
B-175-25 176.18 25.87 174.42 25.53 4.8 61.75 4945.60 5323.60 120
B-200-25 201.96 25.62 197.97 25.42 6.8 71.41 4776.50 5154.50 100

*: iy denotes the radius of gyration about y-axis.

 S.-B. Kang et al. / Journal of Constructional Steel Research 140 (2018) 153–162 155

Q1 Q1

DV1

DV2

DV1
DV2
DH5
DH5

Le/4

Le/4
Q2 Q2

DH4W

DH4 DH4W DH4

Le/4

Le/4
DH6

Le
Le
DH3 DH3W DH3 DH6
DH3W
Le/4

Le/4
DH2

DH2 DH2W DH2W

Le/4

Le/4
Q3 Q3

Hinge support DH1

DH1

(a) Front view (b) Side view

Fig. 2. Test setup and instrumentation for steel columns under axial compression.

2.3. Initial imperfections as shown in Fig. 4. v1, v2 and v3 represent the deviations from the col-
umn centreline. As for the principal initial bending induced by inhomo-
Prior to installation, initial geometric imperfections of steel columns geneous cooling after welding, the initial bending curve was continuous
were measured. Three points were arranged along the column height, and smooth. Thus, deformations at the three points could represent the
deformation profile of steel columns. Each specimen was measured five
times to minimise system error. Table 2 lists the average values of defor-
TF-1 TF-2 TF-3 TF-4 TF-5 mations at the measured points.
As for loading eccentricities, a laser demarcation device was used
to determine the horizontal displacements between the column
centreline and the centre of rotation at the top and bottom supports.
Table 2 summarises the loading eccentricity. Accordingly, the geometric
imperfection was calculated from Eq. (2). It was approximately equal to
LW-1 RW-1 L/1000, consistent with the value considered in design codes [13,17].
y

e ¼ ðet þ eb Þ=2 þ vmax ð2Þ

LW-2 o x RW-2
in which e is the geometric imperfection, et and eb are the loading eccen-
tricities measured at the top and bottom ends of each column and vmax
LW-3 RW-3 is the maximum initial bending along the column height.

Bottom Top

v1 v2 v3
BF-1 BF-2 BF-3 BF-4 BF-5 L/4 L/4 L/4 L/4

Fig. 3. Layout of strain gauges. Fig. 4. Measurement of geometric imperfections.

156 S.-B. Kang et al. / Journal of Constructional Steel Research 140 (2018) 153–162

Table 2 750
Initial geometric imperfection in buckling plane. T12
Specimen Initial bending (mm) Loading e (mm) e/L (%)
T25-1
eccentricity 600 T25-2
(mm)

v1 v2 v3 vmax eb et

Stress (MPa)
B-120-12a 1.74 2.18 2.13 2.18 1.00 8.00 6.68 0.194 450
B-120-12 2.34 2.20 2.16 2.34 0.50 1.75 3.45 0.100
B-168-12 0.42 1.08 1.19 1.19 2.00 5.25 4.82 0.119
B-216-12 2.00 1.42 1.29 2.00 2.50 1.50 4.00 0.097
B-264-12 0.70 0.88 0.61 0.88 2.00 2.00 2.88 0.079 300
B-175-25 2.74 2.33 2.74 2.74 3.00 2.25 5.37 0.102
B-200-25 3.03 3.42 2.57 3.42 0.00 3.75 5.23 0.103

150
Longitudinal residual stresses in each type of box sections were mea-
sured by using sectioning method. Similar to I-sections [15], a short
column segment with a gauge length of 150 mm was extracted from
0
the middle of each box column to eliminate end effect and sectioned 0 5 10 15 20 25
to 12 mm wide steel strips to determine its axial deformations. Axial
strain and stress could be calculated correspondingly. The measured Strain (%)
residual stresses would be used for numerical simulations of steel box
columns under axial compression. Fig. 5. Stress-strain curves of steel plates.

3. Test results and discussions

In addition to the load-displacement curves, the load capacity and
3.1. Material properties associated horizontal displacement at the mid-height of columns were
determined from test results, as included in Table 4. It is apparent that
Material properties of 12 and 25 mm thick steel plates were obtained by increasing the cross section of box columns, the load capacity was in-
by tension tests. For each plate, three tension coupons were tested and creased, whereas corresponding horizontal displacement was signifi-
their average value was calculated. Table 3 includes the yield and ulti- cantly reduced. Comparisons between load capacity and calculated
mate strengths of steel plates. Besides yield and ultimate strengths, yield strength indicate the buckling resistance of steel columns was sub-
elongations were also obtained by strain gauges and an extensometer. stantially smaller than their yield strength. Besides, the buckling factor
Yield and ultimate strains were determined correspondingly. Fig. 5 of steel columns was also calculated from Eq. (3). It ranged from 0.49
shows the stress-strain relationships of steel plates. to 0.83 for steel columns with 12 mm thick walls.

3.2. Buckling resistance of steel columns φ ¼ P u =Af y ð3Þ

Through axial compression tests, the axial load-horizontal displace-

ment curve of steel columns was obtained, as shown in Fig. 6. Note that in which A is the cross-sectional area of box columns and fy is the yield
the horizontal displacement represents the displacement measured at strength of columns obtained from experimental tests.
the mid-height of steel columns. It can be observed in Fig. 6(a) that Comparisons were also made between load capacities obtained from
the applied load increased almost linearly with increasing horizontal experimental tests and calculated in accordance with GB50017-2003
displacement at the initial stage. Before attaining the load capacity, non- [13]. The ratio of experimental to calculated load capacity varied from
linear behaviour developed in the columns. Thereafter, global buckling 1.08 to 1.31, depending on the slenderness. Thus, it can be concluded
of columns decreased the load gradually, whereas rapid increases in that conservative results could be obtained when the design approach
horizontal displacement were recorded at the mid-height. Columns B- in GB50017-2003 was used for global buckling design of welded
120-12a and B-120-12 with the same dimensions exhibited similar Q460GJ steel box columns.
load-displacement curves, despite differences between load capacities.
The load capacity of B-120-12a was only 91.4% of that B-120-12. Rela- 3.3. Typical failure mode
tively lower load capacity of B-120-12a was mainly induced by larger
initial imperfections, as included in Table 2. By increasing the cross sec- Under axial compression, all steel columns failed in global buck-
tion from 120 mm to 264 mm, the load capacity of steel columns was in- ling about y-axis. Significant horizontal deformations could be ob-
creased by nearly three times. When the thickness of column wall was served in B-120-12 and B-175-25 after the vertical load reached its
increased to 25 mm, similar load-displacement curves were obtained, maximum value, as shown in Figs. 7(a and b). Other specimens de-
as shown in Fig. 6(b). veloped similar failure modes. As for columns B-216-12 and B-264-
12 with comparatively large width-thickness ratios, insignificant
Table 3 local buckling took place after the post-peak stage of vertical load,
Material properties of steel plates. but it did not affect the global buckling capacities. Fig. 8 shows the
deformed profile at different stages. When the axial load was 50% of
Plate Thickness Elastic Yield Yield Ultimate Strain at Remark
number (mm) modulus strength strain strength ultimate the load capacity, the maximum horizontal displacement was only
(GPa) (MPa) strength 2.7 mm at the mid-height. The displacement was slowly increased
T12 12 209 546 0.019 678 0.094 For all walls to 18.7 mm at the load capacity. However, after buckling of the column
T25-1 25 211 473 0.013 628 0.114 For LW and occurred, the displacement at the mid-height increased rapidly to
RW only 89.4 mm. It is noteworthy that the deformed profile remained symmet-
T25-2 208 401 0.023 563 0.140 For TF and rical about the mid-height at each load and could be represented by a
BF only
half sine curve.
 S.-B. Kang et al. / Journal of Constructional Steel Research 140 (2018) 153–162 157

6000 5000
B-120-12a B-175-25
B-120-12 B-200-25
5000
B-168-12 4000
B-216-12
4000 B-264-12

Axial load (kN)

Axial load (kN)

3000
3000
2000
2000

1000
1000

0 0
0 15 30 45 60 75 90 0 30 60 90 120 150
Horizontal displacement (mm) Horizontal displacement (mm)

(a) Columns with 12 mm plate (b) Columns with 25 mm plate

Fig. 6. Axial load-horizontal displacement curves of steel columns.

Fig. 9 shows the variation of strains in column B-120-12. Prior to condition, type = stress”. Fig. 10 shows the residual stress distribution
buckling of the column, only compressive strains were measured on box sections B-120-12 and B-175-25. As for the material model, a tri-
on all four walls. When the load capacity was attained, the maximum linear stress-strain model was utilised based on coupon tests. The model
compressive strain of the column was substantially less than the yield comprised an elastic stage, yield plateau and hardening stage of steel
strain. Thus, the yield strength had not been achieved. Beyond the plates.
load capacity, the compressive strain of RW-2 increased rapidly. On
the contrary, LW-2 on the opposite wall was shifted to tensile strain 4.2. Model validation
as a result of considerable horizontal displacements.
The numerical model was validated against experimental results
4. Numerical modeling and parametric study of steel column under axial compression. Comparisons between exper-
imental and numerical results suggest that the numerical model is ca-
4.1. Development of numerical model pable of estimating the load-displacement curve of columns with good
accuracy, as shown in Fig. 11. In addition to load-displacement curves,
In addition to experimental tests, numerical simulations were also comparisons were also made between load capacities of steel columns
conducted on welded Q460GJ steel box columns to study global buck- obtained from experimental tests and numerical simulations. Table 5
ling behaviour. By using ABAQUS [18], numerical models were devel- includes the ratio of load capacities. The average ratio of numerical
oped for steel columns under compression loads. In the model, pin and experimental results is 0.980, with a coefficient of variation
supports were defined at both ends of steel columns. Vertical displace- of 3.6%. Therefore, the numerical model yields reasonably accurate
ment was allowed so that axial load could be applied to the column. predictions of the load capacity of box columns under axial compres-
C3D8R element with reduced integration was employed for steel sion and it can be used for a parametric study on the effects of initial
columns. The mesh size of the columns was determined elaborately so geometric imperfections and normalised slenderness on buckling
that the residual stress pattern in the columns could be accurately incor- factor.
porated. The columns with 12 mm thick wall were divided into two
layers in the thickness direction, as shown in Fig. 10. Three layers
were meshed for 25 mm thick plates. The mesh size along the column
width remained the same as that of steel strips used for residual stress
measurement.
Loading eccentricities at the top and bottom supports were taken
into consideration in the numerical model. In order to consider initial
bending, eigenvalue buckling analyses were performed for steel
columns. The first Eigenmode could be used to approximately repre-
sent the geometric imperfections [19]. A few sets of elements were
defined for each column section and residual stresses measured
through sectioning method were applied to these sets by using “initial

Table 4
Load capacity and horizontal displacement of steel columns.

Specimen Load Yield Buckling Load capacity PuG in Pu/PuG

capacity strength factor φ GB50017–2003 (kN)
Pu (kN) Py (kN)

B-120-12a 1495.3 3032.5 0.49 1250.8 1.20

B-120-12 1635.5 3033.8 0.54 1251.3 1.31
B-168-12 2739.9 4428.1 0.62 2260.0 1.21
B-216-12 4493.6 5730.9 0.78 3656.7 1.23
B-264-12 5852.3 7042.8 0.83 5402.0 1.08 (a) B-120-12 (b) B-175-25
B-175-25 3452.9 7077.0 0.49 2785.0 1.24
B-200-25 4511.4 8181.9 0.55 3917.6 1.15
Fig. 7. Deformed profile of steel columns.
158 S.-B. Kang et al. / Journal of Constructional Steel Research 140 (2018) 153–162

1.0 4.3. Parametric study

0.5Pu
1.0Pu Based on the validated numerical model, parametric studies were
0.8 conducted on the effects of initial geometric imperfection and slender-
Point along column height (xL)

Failure
ness. In the study, box columns with the same cross sections as those
listed in Table 1 but different heights were simulated under axial com-
0.6 pression. Material properties of steel columns remained identical to
those in Table 3.

4.3.1. Effect of initial geometric imperfection

0.4 To study the effect of initial geometric imperfection on the buckling
resistance of steel columns, two different values, namely L/1000 and
L/500, were selected in this study. Table 6 summarises the load capaci-
0.2 ties of box columns with different geometric imperfections. It can be
concluded that by increasing the value from L/1000 to L/500, the load
capacity is decreased by around 7.2% on average. It agrees well with
0.0 the differences between load capacities of B-120-12a and B-120-12.
0 15 30 45 60 75 90 Therefore, the initial geometric imperfection plays an important role
Horizontal displacement (mm) in the global buckling resistance of steel box columns.

Fig. 8. Horizontal displacements of steel column B-120-12.

4.3.2. Effect of slenderness
In addition to initial geometric imperfection, the effect of slender-
ness on global buckling resistance is also investigated through the
numerical model. Besides the column sections in the experimental pro-
2000 gramme, two additional box columns B-225-25 and B-250-25 were
RW-2 simulated through the model. The normalised slenderness is less than
LW-2 3.0, as calculated from Eqs. (4) and (5).
1600 TF-3
BF-3 λy qffiffiffiffiffiffiffiffiffiffi
λn ¼ f y =E ð4Þ
π
Axial load (kN)

1200
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
λy ¼ λ 235=f y ð5Þ
800
in which E is the modulus of elasticity and λ is the slenderness.
Comparisons are made between numerical results and design
400 values provided by different codes [13,14,17]. In GB50017-2003 [13],
four curves are classified in accordance with column section and
manufacturing method. In this study, the maximum width-thickness
0
-6000 -4000 -2000 0 2000 4000 ratio of column web (b'/t) was less than 20, and thus the box column
Strain was classified as ‘c’ in accordance with GB 50017-2003. However, com-
parisons between numerical results and design curves in GB 50017-
Fig. 9. Load-strain curves of steel column B-120-12. 2003 suggest that all numerical values are above the recommended
curve by the code, as shown in Figs. 13(a and b). Therefore, current de-
sign curves for conventional steel can be fairly conservative when used
for high-performance structural steel. As for box columns with a
Fig. 12 shows the deformed profile and longitudinal stress in B-120- wall thickness of 12 mm, curve ‘a’ appears more appropriate for global
12 at failure. The column develops significant horizontal deformations buckling design, whereas curve ‘b’ agrees reasonably well with numer-
when the buckling resistance is reached. The overall profile is in good ical results of columns with a thickness of 25 mm. Fig. 14 shows
agreement with that measured during testing (see Fig. 7(a)). Due to the comparison between numerical results and design curves in
the presence of longitudinal residual stresses, the maximum longitudi- Eurocode 3 [17]. The buckling curve for column sections is classified as
nal stress is greater than the measured values in Fig. 9. ‘c’ based on Eurocode 3, as the thicknesses of walls are 12 and 25 mm,

(a) B-120-12 (b) B-175-25

Fig. 10. Residual stresses in steel columns.

 S.-B. Kang et al. / Journal of Constructional Steel Research 140 (2018) 153–162 159

1800 1800
Test results Test results
Numerical results Numerical results
1500 1500

1200 1200

Axial load (kN)

Axial load (kN)

900 900

600 600

300 300

0 0
0 15 30 45 60 75 0 15 30 45 60 75
Horizontal displacement (mm) Horizontal displacement (mm)

(a) B-120-12a (b) B-120-12

3000 5000
Test results Test results
Numerical results Numerical results
2500
4000

2000
Axial load (kN)

Axial load (kN)

3000
1500
2000
1000

1000
500

0 0
0 15 30 45 60 75 0 15 30 45 60 75
Horizontal displacement (mm) Horizontal displacement (mm)

(c) B-168-12 (d) B-216-12

7500 4500
Test results Test results
Numerical results Numerical results
6000 3600
Axial load (kN)

Axial load (kN)

4500 2700

3000 1800

1500 900

0 0
0 10 20 30 40 50 0 30 60 90 120 150
Horizontal displacement (mm) Horizontal displacement (mm)

(e) B-264-12 (f) B-175-25

5000
Test results
Numerical results
4000
Axial load (kN)

3000

2000

1000

0
0 30 60 90 120 150
Horizontal displacement (mm)

(g) B-200-25

Fig. 11. Comparisons between experimental and numerical results.

160 S.-B. Kang et al. / Journal of Constructional Steel Research 140 (2018) 153–162

Table 5 Table 6
Comparisons between load capacities of steel columns. Effect of initial geometric imperfection on load capacity.

Specimen Test value Pu (kN) Numerical result Pe (kN) Pe Specimen Load capacity (kN) P2
Pu P1
P1 when e/L = 1/1000 P2 when e/L = 1/500
B-120-12a 1495.3 1414.3 0.946
B-120-12 1487.2 1353.1 0.910
B-120-12 1635.5 1603.1 0.980
B-168-12 2837.0 2570.8 0.906
B-168-12 2739.9 2779.8 1.015
B-216-12 4541.8 4300.1 0.947
B-216-12 4493.6 4532.7 1.009
B-264-12 6182.7 5908.4 0.956
B-264-12 5852.3 5561.4 0.950
B-175-25 3048.2 2839.5 0.932
B-175-25 3452.9 3241.6 0.939
B-200-25 5054.7 4632.4 0.916
B-200-25 4511.4 4601.8 1.020
Mean ratio 0.928
Mean ratio 0.980
Coefficient of variation 2.2%
Coefficient of variation 3.6%

tends to be unsafe when the normalised slenderness ranges from 0.5

respectively, more than 0.5tf, and B/tf and H/tw are smaller than 30. to 2.0, as shown in Fig. 15(b).
Similar to GB50017-2003, design curve ‘c’ in Eurocode 3 considerably Comparisons between numerical and design buckling factors also
underestimates the buckling factor, in particular for columns with a indicate that the buckling factor of welded box columns decreases
thickness of 12 mm, as shown in Fig. 14(a). The numerical results with decreasing width-thickness ratio of the column, as shown in
are in good agreement with curve ‘a’. However, curve ‘b’ in Eurocode 3 Figs. 13(b), 14(b) and 15(b). It mainly results from the effect of residual
can be used for the design of box columns of 25 mm thickness stress. By reducing the width-thickness ratio, residual stresses in the
(see Fig. 14(b)). Compared to GB50017-2003 and Eurocode 3, ANSI/ box section can be significantly increased [3], thereby decreasing the
AISC360-10 [14] provides relatively accurate results for the global buck- buckling resistance of the column [20]. This effect becomes more signif-
ling design of welded box columns under axial compression when the icant for steel columns with a wall thickness of 25 mm, as the associated
wall thickness is 12 mm (see Fig. 15(a)). Nevertheless, when the wall width-thickness ratio is considerably smaller than that of columns with
thickness is increased to 25 mm, the design curve in ANSI/AISC360-10 a wall thickness of 12 mm, as included in Table 1.

5. Design recommendations

Comparisons between numerical results and design values from

GB50017-2003 [13] and Eurocode 3 [17] demonstrate that current de-
sign approaches for conventional steel columns yield fairly conservative
estimations of buckling resistances of Q460GJ steel columns. Thus,
when used for the design of Q460GJ steel columns, the advantages of
GJ steel over conventional steel cannot be fully taken. Nevertheless,
ANSI/AISC360-10 [14] provides greater buckling factors than numerical
results, especially when the width-thickness ratio is small. Therefore,
design approaches need to be revised in accordance with experimental
and numerical results. In this section, design recommendations are pro-
posed to take account of the enhancement of GJ steel to global buckling
resistances of columns.
Based on numerical results of box columns under axial compression,
curve ‘b’ in GB500172003 [13] and Eurocode 3 [17] is recommended
for global buckling design of Q460GJ steel box columns under axial com-
pression. Figs. 16(a and b) show the ratio of the numerical result to
design value at each normalised slenderness. It can be observed that
the ratio of numerical to design buckling factor is greater than 1 for
welded box columns with a wall thickness of 12 mm. The ratio attains
its maximum value of 1.15 at a normalised slenderness of 1.25, thereaf-
ter decreasing with increasing normalised slenderness. Nonetheless, the
ratio can be less than 1 for columns with a wall thickness of 25 mm. It is
due to higher residual stresses in the column when the width-thickness
ratio is small. By increasing the width-thickness ratio, the ratio of nu-
merical to design buckling factor can be increased significantly for
steel column of 25 mm thick walls. As for ANSI/AISC360-10 [14], when
a reduction factor of 0.9 is used for welded box columns, the design
curve provides reasonably conservative results for columns with a
wall thickness of 12 mm, as shown in Fig. 16(c). However, when the
normalised slenderness is in the vicinity of 1.25, the ratio of numerical
to design buckling factor is still less than 1 due to the influence of higher
residual stresses.

6. Conclusions

This paper describes a study on global buckling of Q460GJ steel

columns subject to axial compression. Buckling resistance of steel
Fig. 12. Deformed profile and stress contour of B-120-12 at failure. columns was quantified through experimental tests. Besides, numerical
 S.-B. Kang et al. / Journal of Constructional Steel Research 140 (2018) 153–162 161

Fig. 13. Comparisons between numerical results and design curves in GB50017-2003 [13].

Fig. 14. Comparisons between numerical results and design curves in Eurocode 3 [17].

models were established and validated by experimental results. A para- the mid-height of columns. The deformed shape at each loading stage
metric study was carried out to investigate the effects of initial geomet- could be simplified as a half sine curve.
ric imperfection and normalised slenderness on buckling factors of (2) Numerical models were developed by using solid elements. In
Q460GJ steel columns. The following conclusions were drawn from ex- the model, initial geometric imperfections and residual stresses were
perimental and numerical results. considered. The model was validated by experimental results and is
(1) Q460GJ steel box columns exhibited global buckling under axial capable of predicting the load capacity of steel columns with reasonably
compression, with their load capacities significantly lower than calcu- good accuracy. Thus, it can be used for evaluation of buckling resistance
lated yield strengths. Significant lateral deflections were measured at of steel columns under axial compression.

Fig. 15. Comparisons between numerical results and design curves in ANSI/AISC360-10 [14].
162 S.-B. Kang et al. / Journal of Constructional Steel Research 140 (2018) 153–162

Fig. 16. Ratio of numerical to design buckling factor from different codes.

(3) Parametric studies on the effects of slenderness suggest that cur- [4] H.Y. Ban, G. Shi, Y.J. Shi, Y.Q. Wang, Overall buckling behavior of 460 MPa high
strength steel columns: experimental investigation and design method, J. Constr.
rent design approaches for conventional steel columns can be conserva- Steel Res. 74 (2012) 140–150.
tive when used for GJ steel box columns. Thus, design recommendations [5] H.Y. Ban, G. Shi, Y.J. Shi, M.A. Bradford, Experimental investigation of the overall
are made based on experimental and numerical results to accurately buckling behaviour of 960 MPa high strength steel columns, J. Constr. Steel Res.
88 (2013) 256–266.
assess the buckling resistance of GJ steel columns. [6] Y.B. Wang, G.Q. Li, S.W. Chen, The assessment of residual stresses in welded high
Present study only focused on global buckling of Q460GJ steel strength steel box sections, J. Constr. Steel Res. 76 (2012) 93–99.
columns subject to axial compression. When bending moments are [7] Y.B. Wang, G.Q. Li, S.W. Chen, Residual stresses in welded flame-cut high strength
steel H-sections, J. Constr. Steel Res. 79 (2012) 159–165.
applied to the column, different behaviour can be expected. Therefore,
[8] Y.B. Wang, G.Q. Li, S.W. Chen, F.F. Sun, Experimental and numerical study on the
further experimental tests and numerical simulations on steel column behavior of axially compressed high strength steel box-columns, Eng. Struct. 58
under combined axial load and bending moment are still necessary in (79–91) (2014).
[9] T.-J. Li, S.-W. Liu, G.-Q. Li, S.-L. Chan, Y.-B. Wang, Behavior of Q690 high-strength
order to fully quantify the advantages of GJ steel columns over conven-
steel columns: Part 2: parametric study and design recommendations, J. Constr.
tional members in buckling behaviour. Steel Res. 122 (2016) 379–394.
[10] T.-J. Li, G.-Q. Li, S.-L. Chan, Y.-B. Wang, Behavior of Q690 high-strength steel
columns: Part 1: experimental investigations, J. Constr. Steel Res. 123 (2016) 18–30.
[11] G. Shi, W. Zhou, Y. Bai, C. Lin, Local buckling of 460 MPa high strength steel welded sec-
Acknowledgements tion stub columns under axial compression, J. Constr. Steel Res. 100 (2014) 60–70.
[12] B. Yang, G. Xiong, K. Ding, S.D. Nie, W. Zhang, Y. Hu, G.X. Dai, Experimental and nu-
The authors gratefully acknowledge the financial support provid- merical studies on lateral-torsional buckling of GJ structural steel beams under a
concentrated loading condition, Int. J. Struct. Stab. Dyn. 16 (1) (2016) 1640004.
ed by the National Natural Science Foundation of China (51578089), [13] Ministry of Construction, Code for Design of Steel Structures, GB50017-2003, China
the Fundamental and Frontier Research Project of Chongqing Planning Press, Beijing, 2003.
(cstc2015jcyjBX0024), Chongqing Postdoctoral Science Foundation [14] American Institute of Steel Construction, Specification for Structural Steel Buildings,
ANSI/AISC360-10, Chicago, Illinois, 2010.
(Xm2016001) and the Fundamental Research Funds for the Central [15] B. Yang, S.D. Nie, G. Xiong, Y. Hu, J. Bai, W. Zhang, G.X. Dai, Residual stresses in
Universities (No. 106112017CDJQJ208849). welded I-shaped sections fabricated from Q460GJ structural steel plates, J. Constr.
Steel Res. 122 (2016) 261–273.
[16] G. Xiong, S.-B. Kang, B. Yang, S. Wang, J. Bai, S.D. Nie, Y. Hu, G.X. Dai, Experimental
References and numerical studies on lateral torsional buckling of welded Q460GJ structural
steel beams, Eng. Struct. 126 (2016) 1–14.
[1] R. Bjorhovde, Development and use of high performance steel, J. Constr. Steel Res. [17] BSI, Eurocode 3: Design of Steel Structures—Part 1-1: General Rules and Rules for
60 (3) (2004) 393–400. Buildings, BS EN 1993-1-1:2005, British Standards Institution, London, 2005.
[2] Y. Fukumoto, New constructional steels and structural stability, Eng. Struct. 18 (10) [18] ABAQUS, Analysis User's Manual Version 6.13, ABAQUS Inc, 2013.
(1996) 786–791. [19] W.-F. Chen, T. Atsuta, Theory of Beam-Columns Volume 1: In-Plane Behavior and
[3] H.Y. Ban, G. Shi, Y.J. Shi, Y.Q. Wang, Residual stress of 460 MPa high strength steel DesignMcGraw-Hill Inc 1976.
welded box section: experimental investigation and modeling, Thin-Walled Struct. [20] N.E. Shanmugan, S.P. Chiew, S.L. Lee, Strength of thin-walled square steel box
64 (2013) 73–82. columns, J. Struct. Eng. 113 (4) (1987) 818–831.