Journal of Constructional Steel Research 143 (2018) 180–195

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Overall buckling behaviour and design of high-strength steel welded

section columns
Huiyong Ban, Gang Shi ⁎
Key Laboratory of Civil Engineering Safety and Durability of China Education Ministry, Department of Civil Engineering, Tsinghua University, Beijing 100084, PR China

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

Article history: Practical use of high-strength (HS) steel in contemporary construction has become one of the most important de-
Received 18 July 2017 sign solutions. Loading capacities of columns may benefit enormously from the HS steel, whilst their overall buck-
Received in revised form 16 December 2017 ling behaves differently compared with conventional mild (CM) steel columns due to varying effects of initial
Accepted 25 December 2017
imperfections and inelastic properties of the HS steel materials. Despite a number of investigations regarding
Available online 8 January 2018
the HS steel columns being undertaken, there is lack of research focused on variation of HS steel grades and
Keywords:
their effects. To deepen understanding of overall buckling behaviour of the HS steel columns, a comprehensive
High-strength steel review of an extensive body of column test data available in the literature is carried out in the present paper,
Variation of steel grade based on which a three-dimensional finite element (FE) model developed herein is validated. Parametric analy-
Column buckling ses are subsequently undertaken with various HS steel grades, welded cross-sectional geometric parameters,
Imperfection slenderness values and initial imperfections being involved. The FE analysis results are also compared with cal-
Parametric analysis culation values in accordance with national standards. It has been demonstrated that with an increase of the
Column curve grade of HS steel, effects of imperfections decrease whilst that of Y/T ratios are rather limited; reduction effects
on the overall buckling strength become less severe, and therefore higher column curves available in current na-
tional standards may be selected and imperfection factors in the alternative column curve equations proposed
herein descend accordingly. In addition, new theoretical column curves based on Perry-Robertson formula are
developed by introducing imperfection parameters independent on the steel strength.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction
High-strength (HS) structural steel has attracted increasing atten-
tion from both practical engineers [1–3] and academic researchers
[4–6], and their usage in contemporary steel construction has become
an important design solution. The most essential attribute of such ad-
vanced steel in terms of higher yield strength (no less than 460 MPa)
compared with conventional mild (CM) steel is particularly beneficial
to steel columns subjected to compression. Loading capacities of the col-
umns may be significantly improved, whilst their overall buckling be-
haviour is quite distinct from CM steel columns. With an increase of
the yield strength of HS steel, compressive residual stresses within
cross-sections, of which ratio to the yield strength becomes markedly
lower [7,8], possess much less severe effects on the overall buckling be-
haviour; whilst initial bending-induced stresses may also result in re-
duced effects [9] because of their descending ratios to the yield
strength of steel. As a consequence, research and development of robust
design methodologies on overall buckling behaviour of HS steel
⁎ Corresponding author.
E-mail address: shigang@tsinghua.edu.cn (G. Shi).
columns are of great significance for providing economical but safe de-
sign solutions.
Despite a number of experimental programmes of HS steel columns
hitherto having been undertaken and reported in the literature, they fo-
cused generally on some individual steel grade without considerations of
its variation in the analyses of buckling mechanism as well as in the de-
velopment of design methods. The very first column tests on HS steel
seem to be that by Usami and Fukumoto [10] in 1982, in which five
welded box section columns excluding stub ones fabricate from HT80
(690 MPa) HS steel were tested. Rasmussen and Hancock [11] undertook
11 welded box and I-shaped column tests for BISALLOY 80 (690 MPa) HS
steel in 1995. After 2010, the authors investigated experimentally and
numerically the overall buckling behaviour of four S690 HS steel col-
umns and four S960 ones with end restraint [12], as well as 12 Q460C
HS steel welded section columns [13] and six Q960 ones [14]. Similarly,
Wang et al. carried out 12 Q460 HS steel welded section column tests
as well as corresponding finite element (FE) analyses [15,16], and Zhou
et al. [17] designed six Q460 HS steel welded H-shaped section columns
with their overall buckling behaviour being investigated experimentally
ad numerically. More recently, Li et al. [18–19] tested 12 Q690 HS steel
welded box and I-section columns, with FE parametric analyses being
conducted. Chung et al. [20] experimentally investigated seven Q690

https://doi.org/10.1016/j.jcsr.2017.12.026
0143-974X/© 2018 Elsevier Ltd. All rights reserved.
 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195 181

Table 1
Detailed information of HS steel column test specimens available in literature.

Specimen Cross-sectional geometry (mm) L0 (mm) e (mm) fy (MPa) Pu·t (kN) λn φ Pu·FEA (kN) P u
FEA
P u
t

S-35-22 [10] B139.00 × 6.00 1880.0 1.18 741.0 2112.0 0.654 0.853 2147.3 1.017
S-50-22 [10] B138.98 × 6.01 2690.0 0.85 741.0 1798.0 0.967 0.740 1658.1 0.922
R-50-22 [10] B139.00 × 6.00 2090.0 0.83 741.0 1622.0 0.930 0.745 1506.3 0.929
R-65-22 [10] B138.98 × 6.01 2720.0 0.66 741.0 1299.0 1.210 0.594 1327.8 1.022
ER-50-22 [10] B138.98 × 6.01 2090.0 11.32 741.0 1220.0 0.932 0.558 1221.5 1.001
B1150C [11] B98.90 × 5.00 1149.0 0.50 705.0 1174.0 0.562 0.911 1161.7 0.990
B1150E [11] B97.50 × 4.95 1150.0 2.10 705.0 1137.0 0.570 0.904 1076.4 0.947
B1950C [11] B98.22 × 4.96 1950.0 0.50 705.0 1078.0 0.960 0.849 962.8 0.893
B1950E [11] B99.34 × 4.97 1950.0 3.20 705.0 926.0 0.948 0.719 864.8 0.934
B3450C [11] B100.14 × 4.97 3451.0 0.40 705.0 469.0 1.664 0.361 478.6 1.020
B3450E [11] B99.78 × 4.94 3451.0 2.90 705.0 438.0 1.670 0.340 442.2 1.010
I1000C [11] H155.40 × 141.50 × 7.70 × 7.70 1000.0 0.70 660.0 2092.0 0.545 0.952 1965.0 0.939
I1000E [11] H157.14 × 141.10 × 7.67 × 7.71 1000.0 1.30 660.0 2192.0 0.550 0.991 1923.1 0.877
I1650C [11] H156.90 × 141.50 × 7.70 × 7.66 1649.0 0.40 660.0 1751.0 0.896 0.800 1739.5 0.993
I1650E [11] H158.42 × 141.50 × 7.71 × 7.75 1649.0 1.00 660.0 1682.0 0.900 0.762 1683.7 1.001
I2950E [11] H157.50 × 140.30 × 7.75 × 7.74 2950.0 2.00 660.0 745.0 1.627 0.337 795.3 1.067
S1-690-1300 [12] I 120.5 × 100.2 × 10.0 × 8.3 1216.8 4.69 799.0 1857.9 0.477 0.821 2028.0 1.092
S2-960-1300 [12] I 89.5 × 79.5 × 7.9 × 6.1 1103.7 8.73 962.5 1368.4 0.625 0.831 1268.6 0.927
S3-690-2700 [12] I 121.1 × 100.0 × 10.0 × 8.0 2385.0 8.08 799.0 1656.5 0.927 0.739 1449.7 0.875
S4-960-2700 [12] I 120.0 × 99.2 × 10.1 × 7.9 2384.9 11.85 996.0 2099.6 1.066 0.756 1487.2 0.708
S5-690-3600 [12] I 79.8 × 70.4 × 6.0 × 6.1 2345.5 34.73 740.3 306.4 1.356 0.328 283.7 0.926
S6-960-3600 [12] I 95.3 × 79.6 × 7.8 × 6.1 2673.3 15.32 962.5 434.6 1.424 0.260 564.9 1.300
S7-690-3600 [12] I 59.7 × 49.9 × 5.1 × 5.2 2015.1 39.44 783.3 136.7 1.613 0.228 148.6 1.087
S8-960-3600 [12] I 59.9 × 59.2 × 6.1 × 6.2 2086.8 17.97 1019.8 210.0 1.919 0.202 226.1 1.077
B1-460 [13] B152.0 × 10.92 1080.2 5.24 531.9 3129.7 0.300 0.955 2944.5 0.941
B2-460 [13] B141.1 × 14.83 1261.1 6.68 492.3 3642.3 0.374 0.988 3641.9 1.000
B3-460 [13] B121.5 × 12.67 1549.4 1.44 492.9 2185.6 0.532 0.804 1961.7 0.898
B4-460 [13] B102.4 × 11.04 1782.4 2.80 531.9 1503.8 0.760 0.701 1546.9 1.029
B5-460 [13] B102.2 × 10.81 2279.8 11.89 531.9 930.6 0.972 0.443 899.1 0.966
I1-460 [13] I 111.7 × 132.1 × 10.96 × 11.37 2571.4 3.13 531.9 1265.4 0.909 0.597 1296.8 1.025
H1-460 [13] H209.4 × 210.0 × 14.80 × 15.02 1089.3 0.07 492.3 4487.2 0.332 1.014 4018.9 0.896
H2-460 [13] H141.6 × 179.7 × 15.16 × 12.96 1312.1 2.11 492.3 2732.0 0.439 0.797 2822.0 1.033
H3-460 [13] H150.2 × 151.5 × 11.08 × 11.35 1535.4 2.51 531.9 1998.6 0.677 0.770 1974.1 0.988
H4-460 [13] H151.1 × 151.2 × 11.02 × 11.07 1815.1 3.90 531.9 1842.4 0.801 0.717 1828.0 0.992
H5-460 [13] H111.2 × 131.9 × 10.76 × 11.34 2026.0 2.28 531.9 1398.8 1.001 0.669 1230.8 0.880
H6-460 [13] H149.4 × 150.3 × 11.02 × 11.09 1315.2 3.06 531.9 2437.8 0.584 0.956 2122.9 0.871
B1-960 [14]a B142.6 × 13.99 1878.6 25.88 973.2 3779.5 0.775 0.540 3711.2 0.982
B2-960 [14] B141.6 × 13.94 2879.8 3.13 973.2 4063.9 1.196 0.587 3704.1 0.911
B3-960 [14] B141.5 × 13.92 4382.3 0.82 973.2 2193.4 1.822 0.317 1981.1 0.903
H1-960 [14]a H211.1 × 209.8 × 13.96 × 13.93 1882.5 18.67 973.2 4682.7 0.813 0.567 4300.3 0.918
H2-960 [14] H209.5 × 210.8 × 13.93 × 13.93 2883.7 4.92 973.2 4282.2 1.238 0.519 4147.1 0.968
H3-960 [14] H209.9 × 211.0 × 13.92 × 13.87 4381.5 4.83 973.2 2322.8 1.879 0.282 2157.9 0.929
H-3-80-1 [15] H171.25 × 154.50 × 20.99 × 11.52 3320.0 2.08 540.9 1913.0 1.365 0.430 2034.2 1.063
H-3-80-2 [15] H171.25 × 154.70 × 20.98 × 11.36 3304.0 1.70 540.9 2107.5 1.354 0.475 2093.4 0.993
H-5-55-1 [15] H245.75 × 227.75 × 21.33 × 11.54 3320.0 0.33 464.0 4357.5 0.858 0.763 4288.0 0.984
H-5-55-2 [15] H245.50 × 229.00 × 21.15 × 11.62 3320.0 3.13 502.5 4290.0 0.889 0.695 4345.7 1.013
H-7-40-1 [15] H317.25 × 308.75 × 21.03 × 11.47 3320.0 3.00 540.9 7596.5 0.682 0.857 7581.7 0.998
H-7-40-2 [15] H318.50 × 308.25 × 21.20 × 11.46 3320.0 1.58 540.9 7534.5 0.683 0.845 7771.5 1.031
B-8-80-1 [16] B110.3 × 11.40 3320.0 3.00 505.8 1122.5 1.288 0.492 1141.6 1.017
B-8-80-2 [16] B112.0 × 11.49 3260.0 0.60 505.8 1473.5 1.245 0.631 1403.2 0.952
B-12-55-1 [16] B156.5 × 11.43 3260.0 4.90 505.8 2591.0 0.866 0.772 2375.1 0.917
B-12-55-2 [16] B156.3 × 11.42 3260.0 3.80 505.8 2436.5 0.867 0.728 2418.6 0.993
B-18-38-1 [16] B220.2 × 11.46 3260.0 2.40 505.8 3774.0 0.602 0.780 4146.6 1.099
B-18-38-2 [16] B220.8 × 11.46 3260.0 3.40 505.8 4010.0 0.601 0.826 4120.5 1.028
L1-H10 [17] H225.2 × 151.6 × 10.82 × 10.82 2120.0 2.12 550.2 1622.5 1.029 0.538 1640.9 1.011
L2-H10 [17] H222.3 × 151.8 × 10.39 × 10.39 2719.0 2.72 550.2 1141.5 1.315 0.395 1211.2 1.061
L3-H10 [17] H221.3 × 151.8 × 11.08 × 11.08 3318.0 3.32 550.2 839.5 1.600 0.274 981.1 1.169
L1-H10 [17] H226.7 × 149.9 × 12.74 × 12.74 2120.0 2.12 515.7 2128.0 1.006 0.646 1814.6 0.853
L2-H10 [17] H225.2 × 150.8 × 12.47 × 12.47 2720.0 2.72 515.7 1298.0 1.281 0.402 1363.7 1.051
L3-H10 [17] H227.5 × 151.6 × 12.65 × 12.65 3321.0 3.32 515.7 1143.0 1.557 0.347 1059.8 0.927
B-30-1 [18] B236.23 × 16.20 2811.0 27.80 624.0 5771.5 0.514 0.649 6447.5 1.117
B-30-2 [18] B236.47 × 16.10 2812.0 4.90 772.0 9751.5 0.571 0.890 9643.7 0.989
B-50-1 [18] B192.37 × 16.02 3610.0 0.90 772.0 6444.5 0.914 0.739 7299.6 1.133
B-50-2 [18] B192.52 × 16.02 3612.0 2.30 772.0 7180.0 0.914 0.822 6934.0 0.966
B-70-1 [18] B140.88 × 16.07 3610.0 0.10 772.0 3258.5 1.286 0.526 3707.3 1.138
B-70-2 [18] B140.48 × 16.08 3609.0 1.50 772.0 2897.0 1.290 0.469 3444.8 1.189
H-30-1 [18] H259.19 × 260.85 × 16.08 × 16.08 2011.0 2.00 772.0 8493.0 0.585 0.914 8486.9 0.999
H-30-2 [18] H260.35 × 260.82 × 16.25 × 16.25 2010.0 0.50 772.0 8994.0 0.585 0.957 8891.5 0.989
H-50-1 [18] H236.30 × 241.75 × 16.03 × 16.03 2912.0 0.50 772.0 7207.0 0.910 0.847 7663.6 1.063
H-50-2 [18] H238.15 × 240.47 × 16.16 × 16.16 2911.0 1.00 772.0 7124.5 0.916 0.832 7456.5 1.047
H-70-1 [18] H204.78 × 209.21 × 16.26 × 16.26 3511.0 2.80 772.0 3039.0 1.263 0.410 4021.3 1.323
H-70-2 [18] H205.24 × 209.38 × 16.24 × 16.24 3512.0 1.50 772.0 3690.0 1.262 0.498 4248.5 1.151
B-120-45 [22]a B120.68 × 12.54 3392.0 45.52 563.0 861.9 1.261 0.282 898.4 1.042

(continued on next page)

182 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195

Table 1 (continued)

Specimen Cross-sectional geometry (mm) L0 (mm) e (mm) fy (MPa) Pu·t (kN) λn φ Pu·FEA (kN) P u
FEA
P u
t

B-120-75 [22]a B121.12 × 12.60 3391.0 80.56 563.0 642.6 1.256 0.209 715.1 1.113
B-168-30 [22]a B168.96 × 12.61 4009.0 27.38 563.0 2004.1 1.034 0.451 2042.5 1.019
B-168-60 [22]a B168.48 × 12.63 4009.0 60.31 563.0 1469.9 1.038 0.332 1576.1 1.072
B-216-45 [22]a B217.23 × 12.57 4072.0 43.46 551.0 2881.5 0.797 0.508 2985.9 1.036
B-216-75 [22]a B216.98 × 12.55 4075.0 70.39 551.0 2240.8 0.798 0.396 2500.3 1.116
B-264-30 [22]a B264.03 × 12.59 3583.0 29.64 551.0 4748.8 0.571 0.681 4903.5 1.033
B-264-60 [22]a B265.10 × 12.63 3582.0 60.47 551.0 3899.9 0.569 0.555 4068.9 1.043
a
Specimens subjected to eccentric loadings.

HS steel welded H-shaped columns and made comparisons with design
results in accordance with Eurocode 3 [21]. Nie et al. [22] completed an
experimental programme of eight Q460GJ HS steel welded box section
columns to investigate their overall buckling behaviour, and FE analyses
were conducted for developing design guidance.
In addition to the limitation of HS steel grade involved in the afore-
mentioned experimental investigations, of which each concerns mainly
one (up to two) individual steel grade, the validation of FE models is
also crude generally with only dependent test results obtained in each ex-
perimental programme being utilised, respectively. Range of parameters
incorporated in the FE analyses is restricted to a less extent, one or two HS
steel grades for instance among 460 MPa, 690 MPa and 960 MPa. The de-
sign guidance developed in the aforementioned research is also specific
to certain steel grade, rather than a unified calculation model with con-
sideration of variation of HS steel grades. Despite some other numerical
research [23,24] addresses statistical buckling strength of HS steel col-
umns as well as column buckling curves for various HS steel grades,
only hot-rolled section columns are involved. With respect to national
standards including Eurocode 3 [21], GB50017 [25], ANSI/AISC 360 [26],
AS 4100 [27], etc. they provide generally overall buckling factor versus
normalised slenderness ratio relationship for designing the column buck-
ling strength, with different specific expressions, most of which are based
on the Perry-Robertson Formula. Eurocode 3 seems to be the only one
that provides design curves for HS steel columns with steel grades up
to S700 [28], which, however, are simply identical to that for S460 HS
steel columns. It implies that either variation of HS steel grades or afore-
mentioned research basis is not considered for determining the curves of
HS steel columns in Eurocode 3. Detailed expressions of the design equa-
tions are given in the following section.
With the aim of overcoming the limitations of existing research and
design approaches and of reasonably designing welded section columns
for various HS steel, a three-dimensional FE model is developed herein
and validated against all the column test data available in the literature.
By utilising this model, parametric analyses incorporating eight HS steel
grades (460 MPa, 500 MPa, 550 MPa, 620 MPa, 690 MPa, 800 MPa,
890 MPa and 960 MPa) are undertaken, and effects of initial geometric
perfections, residual stresses and steel grade as well as yield-to-tensile
strength (Y/T) ratios are clarified. Design methods are developed ac-
cordingly in terms of individually suggested type of column curves for
each HS steel grade column and newly developed alternative column
curves, as well as new theoretical expressions. It is worth noting that re-
sidual stresses incorporated in the FE analyses herein are determined on
a solid basis of the authors' previous research outcomes [8,14,29]. This
study may provide robust and comprehensive design approaches of
overall buckling behaviour of HS steel columns with various steel
grades, and promote their practical application.
for a total of 80 specimens is listed in Table 1, including cross-sectional
geometry (“H” and “I” denote welded I-shaped sections buckling about
minor and major principle axis, respectively, followed by sectional
height h × width b × flange thickness tf × web thickness tw, and “B” rep-
resents welded box sections, followed by sectional width b × plate thick-
ness t), effective length L0, initial geometric imperfection amplitude e,
yield strength of steel fy, ultimate test load Pu·t, as well as calculated slen-
qffiffiffiffiffiffiffiffiffiffi
derness ratio normalised by steel strength λn ¼ ðλ=πÞ f y =E (in which
the slenderness λ is a ratio of effective length l0 to radius of gyration of
cross-section i) and reduction factor φ of overall buckling. It is worth not-
ing that initial imperfections of the six column specimens reported in ref-
erence [17] were not measured, and therefore an identical amplitude of
the initial bending which equals to 1‰ of the length [30] is adopted here-
in for subsequent non-linear FE analyses.
Based on Table 1, it can be seen that the 80 specimens incorporated
an extensive range of geometric parameters, with thickness of compo-
nent plates ranging from 5 to 21 mm, cross-sectional height varying
from 60 to 318 mm, and length of columns differing from 1000 to
4382 mm; whilst the HS steel for fabricating the specimens is limited
to three grades, i.e. 460 MPa, 690 MPa and 960 MPa. All the specimens
failed by flexural buckling, and for welded I-shaped section columns,
only a few of them buckled about the major principal axis with many
others about the minor principal axis.
Fig. 1 plots test data of 70 specimens under axial compression listed
in Table 1 together with column curves in accordance with Eurocode 3
and Euler curve, excluding those subjected to eccentric compression
loadings. It is indicated that most of the test data are above the corre-
sponding design curve, despite some being below due to relatively larger
initial geometric imperfection amplitudes. Besides, column specimens
fabricated from the HS steel with relatively higher grade, e.g. 960 MPa
steel, possess generally higher buckling factors, which is consistent
with previous research outcomes.
Given uncertainties of initial geometric imperfections in column
tests, it is more convincing to apply adequate and validated numerical
model, which may involve standard imperfections in a way ECCS

2. Review of available experimental results

As mentioned in the previous section, a number of experimental

programmes have been carried out to investigate the overall buckling
behaviour of HS steel welded section columns, with steel grades of
460 MPa, 690 MPa and 960 MPa as well as welded sections of box and Fig. 1. Comparison of column test results with column curves in accordance with Eurocode
I shape being involved. Detailed information available in the literature 3.
 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195 183

Fig. 2. Mesh of cross-sections: (a) box section I; (b) box section II; (c) box section III; (d) I-shaped section.

adopted [30], together with column test data for developing design
methodology of column buckling behaviour. As a consequence, all the
80 specimens reviewed in Table 1 were utilised for validation of a
three-dimensional FE model developed subsequently.
3. Finite element modelling and validation
3.1. Finite element model
A three-dimensional FE model incorporating geometric and material
non-linearity was developed through ANSYS. The column was meshed
by utilising BEAM188 beam element, with a number of divisions being
20 along the length. Cross-section was also meshed into cells, in
which 20 divisions were made either within clear width of component
plate between two adjacent plates in box sections or within clear web
depth between welds in I-shaped sections, and 10 divisions were
meshed within the outstanding length of flanges of I-shaped sections,
as shown in Fig. 2. Specific geometric configurations were slightly differ-
ent due to welding types for welded box sections in the afore-reviewed
column tests. Such finely-meshed cross section has been demonstrated
adequate and accurate enough for analysing the column buckling be-
haviour [13], and permits convenient assignment of residual stresses.
Regarding material constitutive models of HS steel, the authors' pre-
vious work indicates that multi-linear stress-strain relationship may be
utilised to simulate the behaviour of HS steel subjected to uniaxial stress
[6]. For HS steel with a nominal yield strength b500 MPa, there is always
a visible yield plateau in the stress-strain curve; whilst for that with
higher grade, the plateau disappears generally. Fig. 3 illustrates the
two typical multi-linear models for HS steel materials. In the case of
modelling the overall buckling behaviour of column test specimens,
the parameters in the models in Fig. 3 should be determined based on
tension coupon test results accordingly; and in the case of parametric
analyses, values of the stresses and elastic modulus may be set as
being nominal ones according to national standard [25,31] as listed in
Table 2, and values of the strains may be determined based on an exten-
sive body of material test results as suggested in Table 2.
Initial imperfections must be incorporated in the non-linear buckling
analyses because they naturally exist in practical compression members.
In the developed FE model, longitudinal residual stresses within cross-
sections were assigned as initial stress at integration points of elements
by using keyword INISTATE (more details can be found in [13,14]), and
initial bending (as slender columns investigated herein all fail by flexural
overall buckling) was involved by updating geometry of the FE model
through keyword UPGEOM, of which geometry is determined by scaling
the first-eigenvalue buckling shape or as being a spline (if possible) fitted
by deviations at the quarter points along the column length. Distribution
patterns of residual stresses within welded I- and box sections are illus-
trated in Fig. 4 developed previously by the authors [8,12,13]; their mag-
nitudes of the residual stresses within welded sections of all the columns
listed in Table 1 were determined in accordance with either correspond-
ing measurement results if available in the literature or a unified model
developed previously by the authors which is valid for steel with various
grades and is determined on the basis of an extensive body of test data
[14], and the latter one was also applied in the parametric analyses here-
in, of which detailed calculation equations for the magnitudes are report-
ed previously by the authors [14]. Amplitude of the initial bending was
determined either by summing the initial out-of-straightness and load-
ing eccentricity measured in the tests for modelling the specimens, or
as being 1/1000 of the column length for parametric analyses, which is
consistent with Eurocode 3 and China's standard GB 50017 [12–14].
Either end of the columns in the FE model was pin-supported except
those reported in [12] with beams rigidly connected or those introduced
in [13] with a non-linear spring element being utilised for simulating
the actual rotation performance of pinned supports. Compression load-
ing was applied at the column end, and arc-length solution method was
applied to obtain the peak load and responses beyond the peak load.
3.2. Validation against experimental results
All the 80 columns listed in Table 1 were modelled through the
aforementioned technique, and non-linear buckling analyses were car-
ried out. Based on the FE simulation results, column buckling capacities
Pu·FEA of all the specimens reviewed herein were obtained and are listed
in Table 1, as well as their ratio Pu·FEA/Pu·t to the test results. Fig. 5 plots a
comparison between the two groups of data, with good agreement
being indicated. It can be seen that a majority of the data points are lo-
cated within a deviation range of ±10%, and only a very limited number

Table 2
Parameters for constitutive model of HS steel with various grades.

HS steel grade (MPa) E (MPa) fy (MPa) fu (MPa) εst (%) εu (%)

460 2.06 × 105 460 550 2.00 14.0

500 2.06 × 105 500 610 0.24 10.0
550 2.06 × 105 550 670 0.27 9.0
620 2.06 × 105 620 710 0.30 9.0
690 2.06 × 105 690 770 0.33 8.0
800 2.06 × 105 800 840 0.39 7.0
890 2.06 × 105 890 940 0.43 6.0
960 2.06 × 105 960 980 0.47 5.5
Fig. 3. Multi-linear constitutive model for HS steel.
184 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195
Fig. 4. Distribution pattern of residual stress: (a) box section; (b) I-shaped section with flame-cut flange edges; (c) I-shaped section with rolled or sheared flange edges.
of data are beyond the range probably due to uncertainties of imperfec-
tion measurements (particularly geometric ones) or boundary condi-
tions. Average value of the ratios Pu·FEA/Pu·t is 1.005 with a standard
deviation of 9.5%. Given such extensive range of HS steel column speci-
mens for simulation and such good consistency between the FE analysis
and test results from both dependent and independent experimental
programmes, it is confident to conclude that the FE model developed
herein is adequate and accurate for predicting the overall buckling ca-
pacity of HS steel columns with various grades.
Fig. 6 shows some typical comparisons of load versus horizontal de-
flection at mid-length responses between FE analysis results and test
curves available in different literature. Good agreement in terms of ini-
tial stiffness and responses beyond the yield or peak point may also be
found. As a consequence, the FE model may be utilised in the following
parametric analyses with a robust validation basis.
4. Parametric analyses
4.1. Parameters considered
To evaluate effects of initial imperfections and material properties
on the overall buckling behaviour of HS steel columns, as well as to de-
velop robust design methodologies, 2700 columns were modelled by
utilising the validated FE model. A variety of parameters relating to
steel grades, column slenderness, cross-sectional geometry, buckling di-
rection, magnitudes of initial bending amplitude and residual stresses,
as well as Y/T ratios, of which range and values are listed in Table 3.
Nine steel grades were involved in the parametric analyses, one of
which was 235 MPa representing CM steel for comparison and the
others were HS steel grades available in national standards for steel
products [32,33]. The steel grades were just the nominal yield strength.
Fig. 5. Comparison of column buckling strength between FE analysis and test results.
Determination of stress-strain relationship for the steel has been intro-
duced in last section.
With respect to geometry of the columns, 20 different slenderness
ratios normalised by the yield strength were adopted, ranging from
0.3 to 2.2 to cover a majority portion of the column curve. 14 various
cross-sections, including six welded square box ones (Sec-B1–B6) and
eight welded I-shaped ones (Sec-H1–H8), were employed for each of
the column slenderness ratios, and their detailed dimensions are listed
in Table 4. For the welded I-shaped columns, behaviour of overall buck-
ling about either minor or major axis was analysed.
For quantifying effects of initial imperfections on the overall buck-
ling performance, except most of the columns applied standard imper-
fection values introduced in last section, some adopted smaller ones, i.e.
1/2500 of the column length as initial bending amplitude or no residual
stresses within the cross-section. The methods of modelling such im-
perfections have been given in Subsection 3.1. In addition, for some col-
umns the upper limit of tensile stress rather than the lower limit one
was utilised in the material constitutive model to produce different Y/
T ratios, of which effects were elucidated accordingly.
4.2. Effects of initial geometric imperfections
To investigate the effect of initial geometric imperfections, 240 pin-
ended columns of all the parametrically modelling ones were selected,
including one typical grade of CM steel (235 MPa) and one typical
grade of HS steel (690 MPa), three section types (welded box-section,
welded I-section buckling about the minor and major principal axes),
20 normalised slenderness ratios λn with values from 0.3 to 2.2, and
two sets of initial geometric imperfections, i.e. 1/1000 and 1/2500 of
the column length, which were designated as the standard one in accor-
dance with national standard [25,30] and as the comparative one, re-
spectively. The cross-sectional dimension of the box section is 270
× 12 (mm, sectional width b × plate thickness t) and that of the I-
shaped section is 360 × 280 × 16 × 12 (mm, height h × width b × flange
thickness tf × web thickness tw).
Figs. 7 and 8 show comparisons of FE analysis results incorporating
different initial bending amplitudes for 235 MPa and 690 MPa steel col-
umns, respectively, in which two series of data are plotted. The standard
one represents the results of columns with an initial bending amplitude
of 1/1000 of the length (i.e. e/L = 1/1000), and the comparative one cor-
responds to e/L = 1/2500. It can be seen that the effects of initial geo-
metric imperfections on the overall buckling behaviour of HS steel
welded section columns are as follows:
The effect is related to the column slenderness ratio, which becomes
more severe generally with an increase of slenderness ratio for 235 MPa
steel columns, and is maximum for 690 MPa HS steel columns with λn
close to 1.2; and
The effect is relatively less severe for 690 MPa HS steel columns com-
pared with the 235 MPa steel ones. A reduction of initial geometric im-
perfection from L/1000 to L/2500 results in an increase of overall
 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195 185

Fig. 6. Comparison of load versus deflection responses between FE analysis and test results. (a) R-65-22 [10]; (b) H2-960 [14]; (c) H-7-40-2 [15]; (d) B-8-80-1 [16]; (e) B-30-2 [18]; (f) B-
264-30 [22].

buckling strength by around 7% on average for 235 MPa steel columns, It can be seen that these effects of residual stresses on the buckling be-
whilst the percentage drops to 5% for 690 MPa HSS columns. This is be- haviour of HS steel welded section columns are as follows:
cause the stress induced by the initial bending and axial loading is inde- The effect is related to the column slenderness, and is more severe
pendent of the steel strength, whilst its ratio to the yield strength with the slenderness λn between 0.8 and 1.3. Beyond this range it be-
becomes lower for HS steel compared with CM steel. Nevertheless, ef- comes less with larger slenderness, and is significantly reduced due to
fects of initial geometric imperfection on the overall buckling behaviour a decrease of the slenderness ratio with lower values;
vary in a limited range for HS steel columns compared with CM steel The effect is less severe for 690 MPa HS steel columns compared
ones. with 235 MPa CM steel columns. The reduction factor (φ) of overall
buckling strength is improved by 18%–25% if ignoring the residual stress
for the 235 MPa CM steel columns, whilst it becomes 6%–10% for the
4.3. Effects of residual stresses 690 MPa HS steel columns, which indicates that the buckling strength
of 690 MPa HS steel columns is markedly improved due to the less se-
240 pin-ended columns were selected to investigate the effect of re- vere effects of residual stresses; and
sidual stresses within cross-sections. Parameters relating to steel The effect of residual stresses must be taken into account for the in-
grades, cross-sectional geometry, slenderness ratios are consistent vestigation of overall buckling behaviour of columns subjected to com-
with those employed in last subsection. An amplitude of 1/1000 of the pression. For the 235 MPa CM steel columns, the maximum effect purely
column length was applied for determining the initial geometric imper- due to existence of residual stresses is close to 42%, and for the 690 MPa
fections. With respect to the residual stresses, two sets of distributions HS steel columns it is still around 15%–25%.
were incorporated, i.e. one is based on the unified model reported in
[14], as designated being the standard one, and the other one with mag-
4.4. Effects of steel grades
nitudes being equal to zero.
Figs. 9 and 10 present the FE analysis results allowing for effects of
2220 pin-ended columns were selected to investigate the effect of
residual stresses for 235 MPa and 690 MPa steel columns, respectively.
steel grades, with standard initial imperfections being incorporated.
The initial geometric imperfection was taken as being 1/1000 of the col-
umn length in terms of the amplitude, and the residual stresses were
Table 3
Parameters and their values.
based on the unified model reported in [14]. All the other parameters
relating to slenderness ratios and cross-sectional geometry are intro-
Parameters Range Values duced previously.
Steel strength fy (MPa) 235–960 235, 460, 500, 550, 620, 690, 800, Fig. 11 presents the FE analysis results of all the 2220 columns
890, 960 allowing for effects of steel grade. It is indicated that the reduction factor
Normalised slenderness 0.3–2.2 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.1,
ratio λn 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0,
2.1, 2.2 or approximately equal Table 4
values Dimension of column sections incorporated in parametric analyses.
Cross-section Total depth H (mm) Sec-B1–B6, Sec-H1–H8, with
Cross-section B t Cross-section H B tf tw
100–1000 detailed geometry given in Table 4
(mm) (mm) (mm) (mm) (mm) (mm)
Buckling direction N/A About minor and major axes for
welded I-shaped section columns Sec-B1 100 10 Sec-H1 150 150 10 10
Initial geometric 1/1000–1/2500 1/1000, 1/2500 Sec-B2 150 10 Sec-H2 360 280 16 12
imperfections (L−1) Sec-B3 270 12 Sec-H3 350 230 12 12
Residual stresses N/A Standard ones in accordance with Sec-B4 392 14 Sec-H4 300 220 12 10
the unified model [14], no residual Sec-B5 320 10 Sec-H5 400 260 14 14
stresses Sec-B6 380 10 Sec-H6 550 350 18 14
Y/T ratio N/A Upper and lower limit of tensile Sec-H7 800 500 26 22
stresses with fixed yield strength Sec-H8 1000 550 30 26
186 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195

Fig. 7. Comparison of FE analysis results allowing for effects of initial geometric imperfections for 235 MPa CM steel columns.

Fig. 8. Comparison of FE analysis results allowing for effects of initial geometric imperfections for 690 MPa HS steel columns.

of overall buckling strength is significantly increased resulting from an
increase of the steel grade. Compared with 235 MPa CM steel columns,
the reduction factor of 460 MPa HS steel columns, for instance, is im-
proved by around 10%–18%, and the percentage goes up to around
15%–25% and 19%–30% for 690 MPa and 960 MPa HS steel columns, re-
spectively. The maximum improvement is even close to 50% in the
range of normalised slenderness ratio λn close to 1.2. This significant im-
provement is produced due to the less-severe effects of initial geometric
imperfections and residual stresses, as demonstrated previously. In ad-
dition, compared with welded box sections (Fig. 11(a)) and I-shaped
sections buckling about major principal axis (Fig. 11(c)), improvement
of the reduction factor of overall buckling strength for welded I-
sections buckling about minor principal axis (Fig. 11(b)) is markedly
higher.
4.5. Effects of Y/T ratios
480 pin-ended columns were selected to investigate the effect of Y/T
ratios, with one typical grade (235 MPa) for CM steel and three typical
ones (460 MPa, 690 MPa and 960 MPa) for HS steel being incorporated.
The Y/T ratio varies by changing the ultimate tensile stress of steel, i.e.
from the lower limit one to the upper limit one in accordance with na-
tional standards for steel products [32,33]; as a result, the Y/T ratio of
these four steel grades was reduced from 0.635, 0.836, 0.896 and
0.980 to 0.470, 0.639, 0.734, and 0.835, respectively. Initial imperfec-
tions were taken as standard ones.
Figs. 12–15 show comparisons of the FE analysis results allowing
for effects of the Y/T ratio for the four various steel grades, respec-
tively. It can be found that the two groups of results corresponding
to different Y/T ratios possess good consistency, with a maximum
difference resulting from the effects of Y/T ratio being only 1.3%,
which implies insignificant effect of the Y/T ratio on the overall buck-
ling behaviour.
To further elucidate the aforementioned phenomenon, development
of strain within cross-section was analysed. Fig. 16 shows maximum
strain at extreme fibre within the critical section obtained from the FE
analysis results when the load reached its peak value. From
Fig. 16(a) it can be seen that the maximum strain for 235 MPa CM
steel columns is basically larger than the yield strain of the steel when
λn b 1.5, whilst less than the yield strain when λn N 1.5; development
of plasticity for I-section columns buckling about the minor principal
axis is the most significant, with the largest strain over 0.5%. The maxi-
mum strain for 460 MPa HS steel columns is basically larger than the
yield strain of the steel when λn b 1.2, whilst less than the yield strain
when λn b 1.2; development of plasticity for I-section columns buckling
around the minor principal axis is also the most significant, with the

Fig. 9. Comparison of FE analysis results allowing for effects of residual stresses for 235 MPa CM steel columns.
 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195 187

Fig. 10. Comparison of FE analysis results allowing for effects of residual stresses for 690 MPa HS steel columns.

Fig. 11. Comparison of FE analysis results allowing for effects of steel grades ranging from 235 MPa to 960 MPa.

largest strain over 0.8%. In consequence, for either 235 MPa or 460 MPa
steel, due to existence of a yield plateau in the stress-strain relationship,
the maximum strain at extreme fibre within the critical section under
peak load may normally not reach the onset of strain hardening,
which is close to 2.0%. Thus, strain hardening does not occur when the
column buckled, which confirms that there are no effects of Y/T
ratio on the buckling strength of such columns when the yield strain
is fixed.
From Fig. 16(b) it can be seen that the maximum strain for 690 MPa
steel columns is basically larger than the yield strain of the steel when
λn b 1.0, whilst less than the yield strain when λn N 1.0; development of
plasticity for I-section columns buckling about the minor principal axis
is the most significant, with the largest strain over 1.1%. The maximum
strain for 960 MPa steel columns is basically larger than the yield strain
of the steel when λn b 0.9, whilst less than the yield strain when λn N
0.9; development of plasticity for I-section columns buckling about the
minor principal axis is also the most significant, with the largest strain
close to 1.1%. Since there is no yield plateau for these two steel grades,
strain hardening occurs once the strain beyond the yield one, and accord-
ingly there is somewhat effect from the Y/T ratio. However, since the
maximum strain observed in the FE analyses is significantly smaller
than the ultimate strain corresponding to the tensile stress, the effects
of Y/T ratio are very limited. In addition, with an increase of the steel
grade, the critical slenderness to distinguish the elastic and inelastic buck-
ling is reduced, which means range of elastic buckling for HS steel col-
umns is larger.

5. Design guidance

5.1. Suggestions for selection of column curve

With the aim of clarifying adequacy of column curves available in national standards for predicting overall buckling strength of HS steel columns,
comparisons of FE analysis results with the column curves in accordance with Eurocode 3 [21] and GB 50017 [25] were carried out.

Fig. 12. Comparison of FE analysis results allowing for effects of Y/T ratio for 235 MPa CM steel columns.
188 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195

Fig. 13. Comparison of FE analysis results allowing for effects of Y/T ratio for 460 MPa HS steel columns.

Fig. 14. Comparison of FE analysis results allowing for effects of Y/T ratio for 690 MPa HS steel columns.

Figs. 17 and 18 show comparisons of FE analysis results with design curves in either Eurocode 3 or GB 50017 for two typical HS steel grade col-
umns, 460 MPa and 690 MPa respectively. It can be seen that the FE analysis results are much larger than their corresponding design strengths in
either Eurocode 3 or GB 50017.
Based on Figs. 17(a) and (b), for all the 460 MPa steel welded box section columns with various width-to-thickness ratios of component plates,
their buckling factors are higher than the curve b in both Eurocode 3 and GB 50017 with average percentages of 1.6% and 1.0% respectively. As a result,
it was suggested that for all welded box section columns made of 460 MPa HS steel, a unique curve b in both Eurocode 3 and GB 50017 may be se-
lected to design them. For welded I-section columns buckling about minor axis as shown in Figs. 17(c) and (d), their buckling strengths are higher
than the curve b in Eurocode 3 and GB 50017 by 2.5% and 1.9% on average respectively, and therefore curve b was suggested in both Eurocode 3 and
GB 50017 for designing such columns which is still conservative. For welded I-section columns buckling about major axis as shown in Figs. 17(e) and
(f), their buckling strengths are averagely higher than the curve b in Eurocode 3 and GB 50017 by 8.1% and 7.4% respectively, thus curve b was also
suggested in both Eurocode 3 and GB 50017 for designing such columns.
Based on Figs. 18(a) and (b), for all the 690 MPa steel welded box section columns with various width-to-thickness ratios of component plates,
their buckling factors are higher than the curve b in both Eurocode 3 and GB 50017 with average percentages of 7.0% and 6.4% respectively. As a result,
it was suggested that for all welded box section columns made of 690 MPa HS steel, a unique curve b in both Eurocode 3 and GB 50017 may be se-
lected to design them. For welded I-section columns buckling about minor axis as shown in Figs. 18(c) and (d), their buckling strengths are higher
than the curve b in Eurocode 3 and GB 50017 by 9.0% and 8.4% on average respectively, and higher than the curve a in Eurocode 3 by 0.6%, and there-
fore curve a in Eurocode 3 and curve b in GB 50017 were suggested for designing such columns. For welded I-section columns buckling about major
axis as shown in Figs. 18(e) and (f), their buckling strengths are averagely higher than the curve b in Eurocode 3 and GB 50017 by 12.7% and 12.1%
respectively, and higher than the curve a by 4.1% and 1.4% respectively, thus curve a was suggested in both Eurocode 3 and GB 50017 for designing
such columns.
Tables 5 and 6 list detailed values of comparison ratios of FE analysis results (φFEA) to design results (φc, φb, φa, φa0) in accordance with different
columns curves from Eurocode 3 and GB 50017, respectively. Suggestions on selection of columns curves for designing the HS steel columns with
various steel grades ranging from 460 MPa to 960 MPa and with welded box and I-sections are also given in these two tables. Such design guidance
completely in accordance with current national standards may provide engineers valuable reference and convenience for practice.

Fig. 15. Comparison of FE analysis results allowing for effects of Y/T ratio for 960 MPa HS steel columns.
 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195 189

Fig. 16. Maximum strain at extreme fibre within the critical section under peak loading. (a) 235 MPa CM steel and 460 MPa HS steel with visible yield plateau; (b) 690 MPa and 960 HS steel
without visible yield plateau.

5.2. Development of alternative column curves by updating imperfection factors

In addition to selection of column curves available in current national standards, alternative column curves were also proposed in this study with
consideration for higher accuracy through a nonlinear fitting based on the FE analysis results described above by using OGININPRO. The expressions
for the column curves in both Eurocode 3 and GB50017-2003 as shown in Eqs. (1) and (2) respectively are derived from the Perry-Robertson formula
[34],

1 h   2
i
χ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; but χ ≤1:0; where Φ ¼ 0:5 1 þ α λ−0:2 þ λ ; ð1Þ
2
Φ þ Φ2 −λ

Fig. 17. Comparison of FE analysis results with design curves for 460 MPa steel columns. (a) Box section, Eurocode 3; (b) box section, GB 50017; (c) I-section about minor axis, Eurocode 3;
(d) I-section about minor axis, GB 50017; (e) I-section about major axis, Eurocode 3; (f) I-section about major axis, GB 50017.
190 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195

1.0 a0 B/t <30: 1.0 B/t ≤20:

a Sec-B1 a Sec-B1
b Euler curve Euler curve

Buckling factor φ
Buckling factor φ
0.8 Sec-B2 0.8 b Sec-B2
c Sec-B3 c B/t >20:
0.6 Sec-B4 0.6 Sec-B3
d d
B/t ≥30: Sec-B4
0.4 Sec-B5 0.4 Sec-B5
Sec-B6 Sec-B6
0.2 0.2

0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
(a) Nondimensional slenderness λn (b) Nondimensional slenderness λn

1.0 a0 Sec-H1 1.0 Sec-H1

a Sec-H2 a Sec-H2
b Euler curve Euler curve
Buckling factor φ

Buckling factor φ
0.8 Sec-H3 0.8 b Sec-H3
c Sec-H4 c Sec-H4
0.6 Sec-H5 0.6 Sec-H5
d d
Sec-H6 Sec-H6
0.4 Sec-H7 0.4 Sec-H7
Sec-H8 Sec-H8
0.2 0.2

0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
(c) Nondimensional slenderness λn (d) Nondimensional slenderness λn

1.0 a0 Sec-I1 1.0 Sec-I1

a Sec-I2 a Sec-I2
b Euler curve Euler curve
Buckling factor φ

Buckling factor φ

0.8 Sec-I3 0.8 b Sec-I3

c Sec-I4 c Sec-I4
0.6 Sec-I5 0.6 Sec-I5
d d
Sec-I6 Sec-I6
0.4 Sec-I7 0.4 Sec-I7
Sec-I8 Sec-I8
0.2 0.2

0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
(e) Nondimensional slenderness λn (f) Nondimensional slenderness λn

Fig. 18. Comparison of FE analysis results with design curves for 690 MPa steel columns. (a) Box section, Eurocode 3; (b) box section, GB 50017; (c) I-section about minor axis, Eurocode 3;
(d) I-section about minor axis, GB 50017; (e) I-section about major axis, Eurocode 3; (f) I-section about major axis, GB 50017.

8 9
>
> 1−α 1 λ2n ; λn ≤ 0:215 >
>
< " rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi# =
λn N0:215
φ¼
> 1  2
  2
>
ð2Þ
>
: 2 α 2 þ α 3 λn þ λn − α 2 þ α 3 λn þ λ2n −4λ2n ; >
;
2λn

and herein are applied to determine the new column design curve, designated as Fitting curves EC and GB, respectively. In Eqs. (1) and (2), both χ and
φ are reduction factors of overall buckling, and λ and λn denote the normalised slenderness ratio by steel strength.
Coefficients in the expression (i.e. the imperfection factor α in Eq. (1) and α1, α2, α3, in Eq. (2)) were obtained through a non-linear fitting of the FE
analysis results for each HS steel grade and for each type of section. In addition, given no significant difference in the suggested column curve for var-
ious section types being indicated in Tables 5 and 6, another unique new column curve applicable to both welded box and I-sections was also

Table 5
Comparison of FE analysis results with column curves in Eurocode 3 and suggested selection of curve.

Steel grade (MPa) Welded box sections Welded I-sections, buckling about minor Welded I-sections, buckling about major
axis axis
φFEA φFEA φFEA φFEA φFEA φFEA φFEA φFEA φFEA
φc φb φa
Sel. φc φb φa
Sel. φb φa φa0
Sel.

460 1.099 1.016 0.939 b 1.109 1.025 0.946 b 1.081 0.998 0.943 b
500 1.124 1.040 0.961 b 1.128 1.043 0.963 b 1.095 1.011 0.956 a
550 1.148 1.061 0.981 b 1.146 1.059 0.979 b 1.109 1.024 0.968 a
620 1.172 1.083 1.001 b 1.162 1.073 0.991 b 1.113 1.028 0.971 a
690 1.158 1.070 0.988 b 1.180 1.090 1.006 a 1.127 1.041 0.983 a
800 1.191 1.100 1.016 a 1.199 1.108 1.023 a 1.144 1.056 0.997 a
890 1.208 1.117 1.032 a 1.223 1.130 1.043 a 1.151 1.063 1.004 a
960 1.222 1.129 1.042 a 1.234 1.139 1.051 a 1.158 1.069 1.010 a0
 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195 191

Table 6
Comparison of FE analysis results with column curves in GB 50017 and suggested selection of curve.

Steel grade (MPa) Welded box sections Welded I-sections, buckling about minor Welded I-sections, buckling about major
axis axis
φFEA φFEA φFEA φFEA φFEA φFEA φFEA φFEA φFEA
φc φb φa
Sel. φc φb φa
Sel. φc φb φa
Sel.

460 1.136 1.010 0.915 b 1.147 1.019 0.921 b 1.210 1.074 0.972 b
500 1.163 1.034 0.937 b 1.167 1.037 0.939 b 1.226 1.089 0.985 b
550 1.188 1.056 0.956 b 1.186 1.054 0.954 b 1.242 1.103 0.998 b
620 1.212 1.077 0.975 b 1.202 1.067 0.965 b 1.246 1.107 1.001 a
690 1.198 1.064 0.962 b 1.221 1.084 0.980 b 1.262 1.121 1.014 a
800 1.232 1.094 0.990 b 1.241 1.102 0.997 b 1.283 1.138 1.029 a
890 1.250 1.111 1.005 a 1.266 1.124 1.016 a 1.290 1.145 1.036 a
960 1.264 1.122 1.015 a 1.278 1.133 1.023 a 1.298 1.151 1.041 a

developed for each steel grade of the HS steel columns. The fitting process was based on the average level of all parametric analysis results, which was
adopted in current design codes.
Fitting results of the imperfection factors based on the expression in Eurocode 3 and GB 50017 are given in Table 7, and it can be seen that the values
decrease with an increase of the steel grade, which implies improvement of the buckling behaviour and is consistent with aforementioned discussion
results. The new column curves are plotted with the results of the parametric analysis in Fig. 19, from which satisfactory agreement can be found.
In addition to updating the imperfection factors with individual values listed in Table 7 for each HS steel grade, the only factor α in expression
(Eq. (1)) of Eurocode 3 column curves may be replaced by a HS steel grade-dependent one αs satisfying a power function αs = α·(235/fy)n [9]
where fy is in units of MPa. Taking the column curve c (α = 0.49) in Eurocode 3 as the original and basic one, the HS steel grade-dependent imper-
fection factor αs may be determined through a non-linear fitting of the 1940 HS steel columns' FE analysis results, and be expressed as Eq. (3) with a
R-Square(COD) value of 0.985. Fig. 20 illustrates comparisons of the calculation results of buckling factors with design values in accordance with
Eurocode 3 by adopting the factor αs, from which good agreement can be seen. Average of the ratios between the FE analysis results to that calculated
through the factor αs is 1.024 with a standard deviation of 0.052, implying conservation on an average basis and good consistency for the prediction
by utilising the proposed factor αs.

 0:75
as ¼ 0:49 235=f y ð3Þ

5.3. Adaptation of column curve expression

The aforementioned two design approaches of overall buckling behaviour of HS steel columns, either selection of higher columns curves available
in national standards or development of alternative curves by updating magnitudes of the imperfection factors, does not change the analytical ex-
pressions of the columns curves, which are formulated based on the Perry-Robertson equation [34,35]. This section aims to clarify the reasonability
of the analytical expressions of columns curves in Eurocode 3 and GB 50017 in terms of their dependence of various HS steel grades.
The Perry-Robertson formula may be expressed as [35,36]
" rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi#
1  2
φ¼ 1 þ ε0 þ λ2n − 1 þ ε0 þ λ2n −4λ2n ; ð4Þ
2λ2n

where φ is reduction factor of overall buckling, λn denotes the slenderness ratio λ normalised by the yield strength fy of steel as Eq. (5), and ε0 rep-
resents equivalent imperfection parameter and may be determined based on test results in case of practical usage.

sffiffiffiffiffi
λ fy
λn ¼
; ð5Þ
π E

Table 7
Values of imperfection factors for proposed column curves.

Steel grade (MPa) Welded box sections Welded I-sections, buckling about Welded I-sections, buckling about Welded box and I-sections
minor axis major axis

α α1 α2 α3 α α1 α2 α3 α α1 α2 α3 α α1 α2 α3

460 0.342 0.00 0.903 0.377 0.354 1.66 1.029 0.235 0.248 0.88 1.000 0.189 0.310 0.84 0.984 0.255
500 0.308 0.00 0.917 0.334 0.319 1.44 1.022 0.216 0.226 0.83 1.001 0.172 0.280 0.79 0.987 0.229
550 0.276 0.00 0.936 0.286 0.289 1.38 1.024 0.191 0.205 0.77 1.002 0.155 0.253 0.79 0.993 0.202
620 0.245 0.16 0.956 0.238 0.270 1.82 1.058 0.139 0.206 1.18 1.028 0.126 0.239 1.13 1.018 0.162
690 0.272 0.91 0.996 0.213 0.240 1.58 1.048 0.128 0.185 1.14 1.030 0.108 0.227 1.25 1.028 0.142
800 0.230 1.09 1.017 0.157 0.208 1.52 1.050 0.103 0.160 1.08 1.031 0.089 0.195 1.25 1.035 0.110
890 0.205 1.03 1.019 0.134 0.170 0.84 1.014 0.115 0.151 0.94 1.024 0.090 0.172 0.94 1.020 0.110
960 0.189 1.11 1.027 0.114 0.158 0.81 1.014 0.108 0.142 0.92 1.025 0.082 0.160 0.92 1.022 0.099
192 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195

Fig. 19. Comparison of FE analysis results with proposed columns curves for HS steel welded box and I-section columns. (a) 460 MPa; (b) 500 MPa; (c) 550 MPa; (d) 620 MPa; (e) 690 MPa;
(f) 800 MPa; (g) 890 MPa; (h) 960 MPa.

For convenience of comparisons, the analytical expression of columns curves in Eurocode 3 as shown in Eq. (1) may be transformed into Eq. (6):
" rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi#
1 h i2
χ¼ 1 þ α ðλn −0:2Þ þ λ2n − 1 þ α ðλn −0:2Þ þ λ2n −4λ2n ; but χ ≤1; ð6Þ
2λ2n

where χ is also the reduction factor of overall buckling.

Based on comparison of the expression of Perry-Robertson formula Eq. (4) with that of analytical column curves in Eurocode 3 Eq. (6), it can be
found that Eurocode 3 employs a linear expression of λn to determine ε0 for formulating its curves,

ε 0
EC ¼ α ðλn −0:2Þ: ð7Þ

Similarly, by comparing the expression of Perry-Robertson formula Eq. (4) with that of analytical column curves in GB 50017 Eq. (2), it can be
found that GB 50017 also employs a linear expression of λn to determine ε0 for formulating its curves,

ε 0
GB ¼ α 2 þ α 3
λn −1: ð8Þ

Eqs. (7) and (8) indicate identical assumption of both Eurocode 3 and GB 50017 for the equivalent imperfection parameter ε0 in the Perry-
Robertson formula, which is dependent on the slenderness ratio λn and therefore is dependent on the steel grade as λn is normalised by the yield
 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195 193

Fig. 20. Comparisons of FE analysis results of buckling factor with design values in accordance with Eurocode 3 by adopting the proposed imperfection factor αs.

strength fy of steel as Eq. (5). However, initial geometric imperfections relating to fabrication quality are barely dependent on the steel strength, and
residual stresses within cross-sections have been demonstrated unrelated to the steel strength in terms of their magnitudes. Thus, the equivalent im-
perfection parameter ε0 has actually quite minimal effects from the steel grade. To eliminate this relation with fy for determining the parameter ε0,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Eqs. (7) and (8) may be modified by introducing a factor 235=f y as being

 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 
ε 0
EC ¼ α λn
235=f y −0:2 ð9Þ

and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ε0
GB ¼ α 2 þ α 3
λn
235=f y −1; ð10Þ

respectively. By substituting either Eq. (9) or (10) into the Perry-Robertson formula Eq. (4), new theoretical expressions of column curves in
Eurocode 3 and GB 50017 may be obtained accordingly, as given in Eqs. (11) and (12), respectively. The modification to that of Eurocode 3 is similar
to suggestion by Jönsson and Stan [24] with different derivation approaches, however.

1 h  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  2
i
χ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; but χ ≤1:0; where Φ ¼ 0:5 1 þ α λ
235=f y −0:2 þ λ ; ð11Þ
2
Φ þ Φ2 −λ
8 9
>
> 1−α 1 λ2n ; λn ≤ 0:215 >
>
< " rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi# =
φ¼ 1  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 ð12Þ
>
> 2
α 2 þ α 3 λn
235= f y þ λ2n −4λ2n ; λn N 0:215 > >
: 2 α 2 þ α 3 λn
235=f y þ λn − ;
2λn

but φ ≤ 1.0
In Eqs. (11) and (12), the imperfection factors α and α2, α3 remain the original values in accordance with Eurocode 3 and GB 50017, respectively;
whilst the factor α1 may be determined based on continuity of the function at λn = 0.215 and therefore possess different values from the original
ones. It should be noted that in accordance with GB 50017, original values of the factors α2 and α3 for curves c and d vary beyond λn = 1.05, whilst
they remain constant in the range of λn N 0.215 in Eq. (12).
Fig. 21 illustrates a series of columns curves corresponding to different steel grades, in which that of 235 MPa are the original ones. It can be seen
that with an increase of the steel grade, the column curves become higher, which is consistent with the previous discussion results herein. In addition,
it is also indicated that with an increase of the steel grade, the plateau with χ or φ equals to 1.0 becomes longer; for 235 MPa CM steel the plateau
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
extends until λ ¼ 0:2, and for 960 MPa HS steel it extends until λ ¼ 0:2
960=235 ¼ 0:40.

Fig. 21. Column curves with modified theoretical expressions. (a) Eurocode 3; (b) GB 50017.
194 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195

Fig. 22. Comparisons of the FE analysis results with corresponding design values in accordance with modified theoretical column curves. (a) Eurocode 3; (b) GB 50017.

By utilising the updated theoretical expressions of columns curves (i.e. Eqs. (11) and (12)) for designing the HS steel columns, curve c in either
Eurocode 3 or GB 50017 is suggested for the welded box section and I-sections buckling about the minor axis, and curve b in these two standards is
suggested for the welded I-sections buckling about the major axis. Fig. 22 plots comparisons of the FE analysis results with corresponding design
values in accordance with Eqs. (11) and (12), respectively, and good agreement can be found. Average of the ratios between the FE analysis results
to the predictions is 1.016 and 1.052, with a standard deviation of 0.058 and 0.062 for Eurocode 3 and GB 50017 ones, respectively, implying
conservativism on an average basis and good consistency for the prediction by utilising the updated theoretical column curves.
In a word, three design approaches have been developed for predicting the overall buckling behaviour of HS steel columns in the present work
based on different concepts, i.e. selection of higher column curves available in current national standards, development of alternative column curves
by proposing new imperfection factors in terms of individual suggested magnitudes or fitting equation dependent on steel grades, derivation of new
theoretical column curves with imperfection parameters independent on the steel grade. Each of the three approaches may be applied by either ac-
ademic researchers or practical engineers for different design scenarios and requirements.

6. Conclusions appropriate approach for engineering practice, as it is the simplest

As many as 80 independent and dependent HS steel column tests
available in the literature were reviewed, against which a three-
dimensional FE model was validated thoroughly. To elucidate effects
of imperfections, steel grades and Y/T ratios, an extensive range of para-
metric analyses involving 2700 pin-ended columns subjected to axial
compression was carried out. Based on the research outcomes, follow-
ing conclusions were made.
• The FE model developed herein is adequate for evaluating the overall
buckling behaviour of HS steel columns with various grades ranging
from 460 MPa to 960 MPa in terms of buckling strengths and load-
deflection responses.
• Effects of initial geometric imperfections on the overall buckling be-
haviour of HS steel columns are slightly lower compared with that
of CM steel ones, however, are generally b 10% in percentage. Effects
of residual stresses are significantly less severe with an increase of
steel grade, varying from 18%–25% to 6%–10% basically, which must
be considered in the overall buckling analyses. Effects of Y/T ratios
are negligible due to rather limited or even no development of strain
hardening at the peak load.
• Resulting from less severe effects from initial imperfections, reduction
factor of overall buckling strength of the HS steel columns rises on a
non-dimensional basis with an increase of steel grade, and higher col-
umn curves available in Eurocode 3 and GB 50017 may be selected for
their design.
• Alternative column curves with individually updated values of imper-
fection factors are adequate for designing the HS steel columns with
various grades, and a fitting function of the imperfection factor depen-
dent on the steel strength may also be utilised for such design.
• New theoretical column curves are derived based on the Perry-
Robertson formulae by introducing modified imperfection parame-
ters independent on the steel strength for Eurocode 3 and GB 50017,
respectively, which have been demonstrated conservative and accu-
rate for designing the overall buckling behaviour of HS steel columns
with various grades; this method is recommended as the most
one, which is consistent with current design method for CM steel col-
umns and has the strongest theoretical basis.
The research findings and proposed design methods are of great sig-
nificance for robustly and accurately evaluating the overall buckling be-
haviour of HS steel columns, and therefore for promoting their practical
application; more importantly, the new theoretical column curve ex-
pressions may reasonably and scientifically reflect the effects of steel
grades, and are suggested to be applied in future national standards.
Acknowledgements
This work was financially supported by the National Natural Science
Foundation of China (Nos. 51608300, 51478244) and the Excellent
Young Scientist Programme of the National Natural Science Foundation
of China (No. 51522806) awarded to the second author. All the sources
of funding are gratefully acknowledged.
References
[1] R. Bjorhovde, Development and use of high performance steel, J. Constr. Steel Res.
60 (3) (2004) 393–400.
[2] G. Pocock, High strength steel use in Australia, Japan and the US, The Structural En-
gineer 84 (21) (2006) 27–30.
[3] Y.J. Shi, Recent developments on high performance steel for buildings, Adv. Struct.
Eng. 15 (9) (2012) 1617–1622.
[4] G. Shi, Hu FX, Y.J. Shi, Recent research advances of high strength steel structures and
codification of design specification in China, Int. J. Steel Struct. 14 (4) (2014)
873–887.
[5] G. Shi, H.Y. Ban, Y.J. Shi, Y.Q. Wang, Progress in application and research of high
strength steel structures, Proceedings of the Twelfth International Symposium on
Structural Engineering, Vol. 1, 2012, pp. 583–586 (Wuhan, China).
[6] H.Y. Ban, G. Shi, A review of research on high-strength steel structures, Proceedings
of the Institution of Civil Engineers — Structures and Buildings, 2017, https://doi.
org/10.1680/jstbu.16.00197.
[7] H.Y. Ban, G. Shi, Y.J. Shi, Y.Q. Wang, Residual stress tests of high-strength steel equal
angles, J. Struct. Eng. 138 (12) (2012) 1446–1454.
[8] H.Y. Ban, G. Shi, Y. Bai, Y.J. Shi, Y.Q. Wang, Residual stress of 460 MPa high strength
steel welded I section: experimental investigation and modeling, Int. J. Steel Struct.
13 (4) (2013) 691–705.
 H. Ban, G. Shi / Journal of Constructional Steel Research 143 (2018) 180–195
[9] International Association for Bridge and Structural Engineering, Use and Application
of High-performance Steels for Steel Structures, IABSE, Zurich, 2005.
[10] T. Usami, Y. Fukumoto, Local and overall buckling of welded box columns, J. Struct.
Div. 108 (ST3) (1982) 525–542.
[11] K.J.R. Rasmussen, G.J. Hancock, Tests of high strength steel columns, J. Constr. Steel
Res. 34 (1) (1995) 27–52.
[12] G. Shi, H.Y. Ban, F.S.K. Bijlaard, Tests and numerical study of ultra-high strength steel
columns with end restraints, J. Constr. Steel Res. 70 (2012) 236–247.
[13] 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.
Steel Res. 74 (2012) 140–150.
[14] H.Y. Ban, G. Shi, Y.J. Shi, M.A. Bradford, Experimental investigation of the overall
buckling behaviour of 960 MPa high strength steel columns, J. Constr. Steel Res.
88 (2013) 256–266.
[15] Y.B. Wang, G.Q. Li, S.W. Chen, F.F. Sun, Experimental and numerical study on the be-
havior of axially compressed high strength steel columns with H-section, Eng.
Struct. 43 (2012) 149–159.
[16] Y.B. Wang, G.Q. Li, S.W. Chen, F.F. Sun, Experimental and numerical study on the be-
havior of axially compressed high strength steel box-columns, Eng. Struct. 58
(2014) 79–91.
[17] F. Zhou, L.W. Tong, Y.Y. Chen, Experimental and numerical investigations of high
strength steel welded H-section columns, Int. J. Steel Struct. 13 (2) (2013) 209–218.
[18] T.J. Li, G.Q. Li, S.L. Chan, Y.B. Wang, Behavior of Q690 high-strength steel columns:
part 1: experimental investigation, J. Constr. Steel Res. 123 (2016) 18–30.
[19] T.J. Li, S.W. Liu, G.Q. Li, S.L. Chan, Y.B. Wang, Behavior of Q690 high-strength steel
columns: part 2: parametric study and design recommendations, J. Constr. Steel
Res. 122 (2016) 379–394.
[20] K.F. Chung, G.Q. Li, K. Wang, T.Y. Ma, X. Liu, Experimental investigation into high
strength steel columns of Q690 welded H-sections, Proceedings of the 11th Pacific
Structural Steel Conference, Shanghai, China 2016, pp. 936–942.
[21] British Standards Institution (BSI), BS EN 1993-1-1: 2005 Eurocode 3: Design of
Steel Structures Part 1-1: General Rules and Rules for Buildings, BSI, London, 2005.
[22] S.D. Nie, S.B. Kang, L. Shen, B. Yang, Experimental and numerical study on global
buckling of Q460GJ steel box columns under eccentric compression, Eng. Struct.
142 (2017) 211–222.
195
[23] R.C. Spoorenberg, H.H. Snijder, L.G. Cajot, N. Popa, Buckling curves for heavy wide
flange QST columns based on statistical evaluation, J. Constr. Steel Res. 101 (2014)
280–289.
[24] J. Jönsson, T.C. Stan, European column buckling curves and finite element modelling
including high strength steels, J. Constr. Steel Res. 128 (2017) 136–151.
[25] GB 50017-2003, Code for Design of Steel Structures, China Architecture & Building
Press, Beijing, 2006.
[26] AISC, ANSI/AISC 360-16, Specification for Structural Steel Buildings, AISC, Chicago,
2016.
[27] Australian Standard (AS), AS 4100-1998, Steel Structures, Standards Australia Limit-
ed, Sydney, 2012.
[28] British Standards Institution (BSI), BS EN 1993-1-12: 2007 Eurocode 3: Design of
Steel Structures Part 1–12: Additional Rules for the Extension of EN 1993 up to
Steel Grades S700, BSI, London, 2007.
[29] H.Y. Ban, G. Shi, Y.J. Shi, Y.Q. Wang, Residual stress of 460 MPa high strength steel
welded box section: experimental investigation and modeling, Thin-Walled Struct.
64 (2013) 73–82.
[30] European Convention for Constructional Steelworks, Manual on Stability of Steel
Structures Part 3.1 Compression Member Design Based on Flexural Buckling Failure
Modes, Second Edition ECCS Publ, Bruxelles, 1976.
[31] H.Y. Ban, M.A. Bradford, B. Uy, X.P. Liu, Available rotation capacity of composite
beams with high-strength materials under sagging moment, J. Constr. Steel Res.
118 (2016) 156–168.
[32] GB/T1591-2008, High Strength Low Alloy Structural Steel, China Standards Press,
Beijing, 2009 (in Chinese).
[33] GB/T 16270-2009, High Strength Structural Steel Plates in the Quenched and Tem-
pered Condition, China Standards Press, Beijing, 2007 (in Chinese).
[34] N.S. Trahair, M.A. Bradford, D.A. Nethercot, L. Gardner, The Behaviour and Design of
Steel Structures to EC3, 4th edn Spon Press, London, 2008.
[35] R.D. Ziemian, Guide to Stability Design Criteria for Metal Structures, 6th edn John
Wiley & Sons, Inc., New Jersey, 2010.
[36] A. Robertson, The strength of struts, Selected Engineering Papers 1 (28) (1925)
1–55.