Journal of Constructional Steel Research 129 (2017) 54–74

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Design and optimization of 420 MPa large-section HSSY columns

Xianglin Yu a, Hongzhou Deng a,⁎, Lei Cui b, Donghong Zhang b
a
Department of Structural Engineering, Tongji University, Shanghai 200092, China
b
Guangdong Electric Power Design Institute of China Energy Engineering Group Co., Ltd., Guangzhou 510663, Guangdong, China

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

Article history: A new type of large-section high strength steel Y-shaped (HSSY) column was presented. 24 specimens in total
Received 19 January 2016 were designed and subject to experiment in order to investigate the overall buckling behavior. The columns in-
Accepted 30 September 2016 cluding two commonly-used slenderness ratios of 35 and 40 were divided into two groups for axial and eccentric
Available online xxxx
compression, respectively. Prior to the buckling test, the material properties, initial geometric imperfections and
longitudinal residual stresses were measured. The buckling deformations and ultimate strengths were obtained
Keywords:
HSSY column
by test and compared with numerical results, which were based on validated finite element models taking the
Axial and eccentric loadings real initial imperfections and boundary conditions into account. A good agreement was reached between the
Initial imperfections test and numerical approaches. Buckling factors and column curves of axial compression specimens were obtain-
Column curves ed and compared with existing code design curves. To accurately predict the overall buckling strengths of HSSY
Section optimal selection columns, recommended column curves and their formulae corresponding to different codes were proposed
through nonlinear regression method. In addition, the influence of loading eccentricity on ultimate buckling ca-
pacity was analyzed. Furthermore, the cross section efficiency was formulated and criteria for section optimal se-
lection were concluded for making the best use of the column. Finally, a variety of large-section and medium-
section columns with potentially high utilization ratios were recommended for engineering application.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction
Due to its superiority over normal strength steel (NSS) in material
saving, structural safety and economic benefits, high strength steel
(HSS) has been increasingly used in transmission towers in recent
years. During the design of steel members, simple cross section (i.e.
steel angles or circular tube) and built-up cross section (i.e. cruciform
angle or quad angle section connected with fill plates) [1–3] were devel-
oped and applied in high-rise structures. On the one hand, latticed
transmission towers with multi-circuit, high voltage, light weight and
super height are developing rapidly in China for the huge demands for
electric power; on the other hand, main members do not satisfy the re-
quirement of sufficient bearing capacity if designed with simple cross
section, nor economic and reasonable with conventional built-up sec-
tion due to its drawbacks of connective material consuming, non-uni-
form stress distribution and relatively low section utilization. Under
these circumstances, a new type of Y-section column featuring better
stress performance, fabrication simplicity and material saving was
invented and has been safely used in practical latticed transmission
towers, which had been subject to full-scale field test by us as shown
in Fig. 1. The test result showed that one diagonal bracing member at
front tilt slope was buckled firstly, which caused the final collapse of
the whole tower. For those HSSY main members, they did not fail
⁎ Corresponding author.
E-mail address: denghz@tongji.edu.cn (H. Deng).
until the external load reached 130% of the designed value, at which
moment the maximum compressive stress of main members reached
115% of the designed full stress. Therefore, the HSSY column was con-
siderably reliable and safe to be used as the main member in transmis-
sion tower. Meanwhile, the full-scale test of an individual HSSY column
was essential as well to investigate its overall buckling behavior, be-
cause no provisions at present were stipulated in current design codes
regarding to such a novel column [3–11].
Research on diversified bucking curves of different cross-sectional
columns is an important objective. Some studies have been focusing
on curves acquisition and pertinent formulae derivation through exper-
imental and numerical approaches. Deterministic and probabilistic so-
lutions to developing a total of 112 maximum strength column curves
were elaborated in detail by Bjorhovde et al. [12,13]. It was concluded
that the probabilistic multiple curves represented a scientifically
sound solution and was regarded as more adequate than those obtained
by the deterministic approach, which provided theoretical evidence for
the American column curves. For HSSY columns, the mono-symmetric
cross section shares similar geometric and mechanical properties to
equal-leg angle section column or tee section column. 66 and 90 col-
umns both fabricated from 420 MPa HSS equal-leg angles were subject
to axial compression tests by Ban et al. [14] and Cao et al. [15], respec-
tively. Cardoso and Rasmussen [16] investigated the buckling behavior
of concentrically loaded T-section steel columns with measured yield
strength of 300 MPa, which concluded that the Australian standard
AS4100 provided the most accurate predictions for the buckling

http://dx.doi.org/10.1016/j.jcsr.2016.09.020
0143-974X/© 2016 Elsevier Ltd. All rights reserved.
 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74
(a)
Fig. 1. Field test of full-scale latticed transmission tower with HSSY main members: (a) Tower member assembly field; (b) full-scale tower test.
strength, but Eurocode 3 and ANSI/AISC 360-10 were more conservative
and modifications for improving the design strength predictions were
recommended.
The assessment of initial imperfections (i.e. initial out-of-straight-
ness and residual stresses) of HSS columns is another research issue.
Some studies [17–21] highlighted that the influence of initial bending
and residual stresses on overall buckling capacities was less significant
when compared with NSS columns. The overall buckling behavior of
HSS columns was investigated by experiments [22–24], and it revealed
that the buckling curves specified in the selected codes (GB 50017-
2003, Eurocode 3 and ANSI/AISC 360-10) were conservative to predict
the buckling strengths of these HSS columns. Eccentric and concentric
axial loading tests of 13 box- and I-section HSS columns with nominal
yield strength of 690 MPa were carried out by Rasmussen and Hancock
[17]. The residual stresses in three box columns and three welded
flame-cut H-section columns fabricated from Q460 steel plates were
measured and assessed by Wang et al. [25,26]. The result showed that
the residual stresses ratios of HSS welded box sections and H-sections
tended to be less detrimental to the buckling strength than the ordinary
steel sections. Similar measurements of residual stresses in three box-
and H-sections fabricated from Q690 steel plates were conducted by Li
Fig. 2. Geometric parameter definition of Y-section.
55
Firstly buckled
member
Y-section
column
(b)
et al. [27] and lower ratios of residual compressive stress were obtained
when compared with that of Q460 test specimens by Wang et al. [25,
26]. Summary can be concluded from the aforementioned studies that
the buckling strengths corresponding to code design curves were
underestimated and the influence of initial imperfections on buckling
strengths were overestimated. However, it remains to be testified to de-
termine whether these conclusions are equally applicable to the HSSY
column or not in view of the differences of manufacturing process and
residual stress distribution between a new and conventional section.
In addition, material properties and boundary conditions are also im-
portant factors influencing the overall buckling capacities of steel col-
umns, which were carefully studied in [19,28–30].
The present study focused on the test and numerical investigations
of 420 MPa large-section HSSY columns subject to axial and eccentric
compression with 12 specimens for each. It aimed to explore the overall
buckling behavior of a new-style column. Before the buckling tests, ini-
tial geometric imperfections, longitudinal residual stresses and material
properties were measured and described in detail. The buckling defor-
mations and buckling capacities were obtained from the tests and com-
pared with numerical results. The finite element (FE) models of all
specimens were established and validated by test results. Sensitivity of
buckling capacities to loading eccentricities was evaluated regarding
to eccentrically compressed specimens. Based on extended numerical
data and various code formulae, corresponding column curves derived
through parametric analyses were recommended to predict the ulti-
mate buckling capacities of the columns. Finally, criteria for section op-
timal selection and potentially high efficient sections were proposed for
experimental and practical uses.
2. Test program
2.1. Test specimen details
The test specimens comprised a total of 24 large-section HSSY col-
umns with different sectional width-to-thickness ratios (for steel an-
gles: 7.8 and 9.1; for steel plate: 7.9 and 9.1). 12 columns were
designed for axial compression and the other 12 columns were subject
to eccentric loadings. The equal-leg steel angles of all columns were fab-
ricated from hot-rolled Q420 steel and the steel plates were flame cut
also from Q420 steel. The steel angles were welded at the limb back
56 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74

Table 1
Nominal geometric parameters of large-section HSSY specimens.

Specimen Steel angles Steel plate A (mm2) λx λy rx ry Loading Number

(mm) (mm) eccentricity (mm)
ba (mm) ta (mm) ba0/ta bb (mm) tb (mm) bb0/tb

Z135 250 26 7.8 180 18 9.1 15,655.4 33 35 92.0 85.7 0 3

Z135P 250 26 7.8 180 18 9.1 15,655.4 33 35 92.0 85.7 50 3
Z235 250 26 7.8 200 20 9.1 16,415.4 29 35 100.4 83.7 0 3
Z235P 250 26 7.8 200 20 9.1 16,415.4 29 35 100.4 83.7 50 3
Z340 220 20 9.1 140 16 7.9 10,715.6 39 40 77.0 75.8 0 3
Z340P 220 20 9.1 140 16 7.9 10,715.6 39 40 77.0 75.8 15 3
Z440 220 20 9.1 160 16 9.1 11,035.6 36 40 83.3 74.7 0 3
Z440P 220 20 9.1 160 16 9.1 11,035.6 36 40 83.3 74.7 20 3
In total 24

with the steel plate to form a Y-shaped column by THY-51B weld rod as
illustrated in Fig. 2. The nominal and measured geometric parameters of
the specimens were listed in Tables 1 and 2, respectively. All test speci-
mens were hot zinc galvanized. The 12 specimens for axial compression
were designed to study the overall buckling behavior of large-section
columns used as main members in latticed transmission towers. In
order to investigate the influence of loading eccentricities on overall
buckling capacities of Y-section columns, another group of 12 speci-
mens were designed to be eccentrically compressed and the eccentrici-
ties were designed along the symmetric axis. Thus, the slenderness
ratios of all specimens and the width-to-thickness ratios of steel angles
and steel plates were properly designed in prevention of the occurrence
of local buckling prior to overall buckling.
Both the industry standard [1] and the Chinese national standard [5]
specified the limits of width-to-thickness ratios in correlation with slen-
derness ratios as follows:
For axial compression members according to both standards:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b=t ≤ ðb=t Þlim ¼ ð10 þ 0:1λÞ 235=f y
For beam-columns according to industry standard:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b=t ≤ ðb=t Þlim ¼ 15 235=f y ; 30 ≤λ≤100 ð2Þ
For beam-columns according to Chinese national standard:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b=t ≤ ðb=t Þlim ¼ 13 235=f y ; 30 ≤λ≤100 ð3Þ
where λ was the larger value of the column slenderness ratios in two di-
rections (buckling around axis-x and around axis-y), taken as 30 when
λ b 30, and as 100 when λ N 100. fy was the nominal yield strength
taken as 420 MPa. b was the free outstand length taken as bb0 (see Fig.
2, for steel angles: b0 = ba − ta − r; for steel plate: b0 = bb − h. r was
the internal bend radius of rolled members taken as 22 mm for
Z1(P)and Z2(P), 18 mm for Z3(P) and Z4(P). h was the leg size of fillet
weld taken as 16 mm for Z1(P), 18 mm for Z2(P), and 14 mm for
Z3(P) and Z4(P)). The lower and upper limits of width-to-thickness ra-
ð1Þ tios were 9.7 (when λ = 30) and 15.0 (when λ = 100) with respect to
axial compression members. Besides, the limits (b/t)lim were taken as
10.1 and 10.5, which corresponded to λ = 35 and λ = 40, respectively.

Table 2
Measured geometric parameters and initial geometric imperfections of large-section HSSY specimens.

Specimen Measured geometric parameters Initial geometric imperfections

L L0 (mm) ba (mm) ta (mm) bb (mm) tb (mm) Ix (×108 mm4) Iy (×108 mm4) λx λy |u|max |v|max(mm) |θ|max
(mm) (mm) (10−3 rad)

Z135-1 2542 3002 250.2 26.78 180.4 18.38 1.37 1.18 32.4 34.8 0.99 0.76 2.47
Z135-2 2526 2986 250.5 26.08 180.6 18.40 1.36 1.16 32.3 34.9 1.37 0.76 3.17
Z135-3 2530 2990 250.4 26.23 181.9 18.45 1.37 1.16 32.3 35.1 1.11 0.82 1.10
Z135P-1⁎ 2526 2986 250.3 26.21 180.8 18.03 1.34 1.16 32.4 34.8 1.33 0.61 2.78
Z135P-2 2538 2998 250.2 26.11 180.4 18.26 1.34 1.16 32.5 35.0 1.13 0.79 1.61
Z135P-3 2540 3000 250.2 26.18 180.1 18.20 1.34 1.16 32.5 34.9 0.65 0.60 1.67
Z235-1 2456 2916 250.8 26.56 201.1 20.37 1.71 1.18 29.6 35.7 0.89 0.71 2.00
Z235-2 2470 2930 250.7 26.23 200.5 20.01 1.68 1.17 29.0 34.7 0.69 0.55 1.60
Z235-3 2470 2930 251.0 26.20 201.0 20.17 1.69 1.17 29.0 34.9 1.17 0.26 2.81
Z235P-1 2460 2920 250.1 26.22 200.4 20.70 1.70 1.16 29.0 35.1 1.05 0.72 2.43
Z235P-2 2458 2918 250.8 26.58 200.4 20.22 1.70 1.18 29.0 34.8 1.29 0.75 2.72
Z235P-3 2460 2920 251.1 26.44 200.3 20.27 1.70 1.18 29.0 34.7 1.25 0.48 3.37
Z340-1 2522 2982 220.2 20.60 140.4 16.07 0.65 0.63 38.0 38.4 1.35 0.96 3.56
Z340-2 2532 2992 220.2 20.23 140.4 16.07 0.64 0.62 37.9 38.5 0.83 0.84 1.37
Z340-3 2530 2990 220.5 20.26 140.2 16.60 0.66 0.63 37.8 38.7 1.01 0.66 2.62
Z340P-1 2524 2984 220.1 20.24 140.2 16.35 0.64 0.62 38.7 39.3 0.86 0.80 1.45
Z340P-2 2534 2994 220.8 20.36 140.1 16.20 0.64 0.63 38.9 39.3 0.70 0.64 3.86
Z340P-3 2534 2994 220.1 20.13 140.5 16.12 0.64 0.62 38.8 39.4 1.34 0.58 2.77
Z440-1 2534 2994 220.5 20.01 161.3 16.11 0.72 0.62 36.4 39.2 1.37 0.61 3.65
Z440-2 2532 2992 220.2 20.19 160.1 16.03 0.71 0.62 36.8 39.4 1.05 0.44 2.54
Z440-3 2525 2985 220.4 20.32 160.3 16.01 0.72 0.63 36.8 39.4 1.48 1.02 3.02
Z440P-1 2532 2992 220.8 20.02 160.2 16.01 0.77 0.62 35.9 39.9 0.67 0.84 1.06
Z440P-2 2532 2992 221.2 20.01 160.3 16.31 0.77 0.63 35.8 39.8 1.24 0.71 2.78
Z440P-3 2530 2990 220.1 20.08 160.4 16.25 0.77 0.62 35.8 39.9 0.98 0.58 3.51

Note: L was the measured column length; L0 was the effective length between two hinged supports and was taken as the measured length L plus 460 mm (distance between the rotation
center and the end of the specimen, i.e. 230 mm for each end).
Z135(P)-1⁎: Z represents “large-section HSSY column”, the first “1” represents the group No. of the cross section, 35 represents the maximum slenderness ratio λmax around axis-y, −1
represents the first test specimen of the three columns with the same nominal geometric dimension and length.
 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74
Table 3
Limits of width-to-thickness ratios in axial compression.
Codes/standards
ASCE 10-97
BS 5950-1
ANSI/AISC 360-10
AS4100-1998
CSA S16-09
AIJ 2010
For eccentric compression members, (b/t)lim were taken as 11.2 (ac-
cording to industry standard) and 9.4 (according to national standard).
The limits of width-to-thickness ratios of compression parts accord-
ing to the British standard Eurocode 3 [4] were given as:
For outstand flange subject to compression:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b=t ≤ ðb=t Þlim ¼ 14 235=f y
For outstand flange subject to bending and compression:
pffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b=t ≤ ðb=t Þlim ¼ 21 K σ 235=f y
The limit of width-to-thickness ratio for outstand flange subject to
axial compression was 10.5 when the cross-section was defined as
Class 3. In the case of bending and compression members, Kσ in Eq.
(5) was the buckling factor corresponding to the stress ratio ψ
(1N ψ ≥0 herein) and boundary conditions [31]. Thus, the buckling factor
Kσ ranged from 0.43 to 0.57 for outstand flanges and the limits of (b/t)lim
in Eq. (5) accordingly ranged from 10.3 to 11.9.
All the width-to-thickness ratios of axial and eccentric compression
members as shown in Table 1 satisfied the requirement of their corre-
sponding limits in accordance with the Chinese standards [1,5] and
the British standard [4] and furthermore, within the relevant limits as
specified in other 6 codes [3,7–11] (see Table 3). Therefore, the local
buckling was not expected to occur prior to the overall buckling, and
the latter became the decisive failure mode for predicting the ultimate
buckling capacities of the test specimens.
The most commonly-used slenderness ratios of large sectional main
members under compression in latticed transmission towers were be-
tween 30 and 40, thus two slenderness ratios of 35 and 40 were de-
signed for buckling tests. To ensure the sufficient utilization of the
cross sectional strengths either buckling around the major axis (axis-
x) or around the minor axis (axis-y), each section was designed to
Fig. 3. Tension coupon sampling location and dimension: (a) steel angles; (b) steel plate; (c) dimension.
57
Formulae Limits
pffiffiffiffiffiffi
ðw=tÞlim ¼ 80Ψ= F y , where Ψ = 2.62 9.9
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(b/T)lim = 15ε, where ε ¼ 275= f y 12.1
qffiffiffiffiffiffiffiffiffiffi
ðb=tÞlim ¼ 0:45 E=f y 10.0
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðb=tÞlim ¼ 14 250= f y 10.8
pffiffiffiffiffiffi
ðbel =tÞlim ¼ 200= F y 9.8
qffiffiffiffiffiffiffiffiffiffi
ðb=t f Þlim ≤0:82 E= f y 18.3
make the radius of gyration around both axes approximately equal.
Considering that the actual end restraints of the testing system were
unidirectional cylindrical supports (rotation around axis-x was restrict-
ed, see Fig. 2 and Fig. 9), the radius of gyration around axis-x (rx) was
slightly larger than that around axis-y (ry, see Table 1) so that the col-
umns were expected to buckle around the symmetric axis (axis-y),
ð4Þ thus the boundary conditions (BC) were similar to the case of fully
pin-ended supports. 18 standard tension coupons (three for each sec-
tion) were cut from the parent steel angles and plates to determine
the stress-strain relationship, as elaborated in Section 2.2. The longitudi-
nal residual stresses were quantified and modelled by blind hole drilling
ð5Þ method and a simplified distribution pattern was put forward, as
expounded in Sections 2.3 and 4.3.
2.2. Material properties
Tension coupon tests were carried out to determine the stress-strain
properties of Q420 steel by the 1000kN tensile and compressive univer-
sal testing machine. A total of 18 tension coupons with 3 for each section
were cut from their parent plates and subject to uniaxial tensile loads
according to the Chinese standards GB/T 2975-1998 [32] and GB/T
228-2002 [33]. The longitudinal axes of the tension coupons were par-
allel with the main processing direction of the parent steels, and the
sampling location and dimension were shown in Fig. 3 [32]. The test
setup, before fracture and fracture moment of tension coupons were
shown in Fig. 4. Two bidirectional strain gauges were attached at the
center of both sides to measure the tensile strains. The measured me-
chanical parameters of every three coupons cut from the same parent
plate were averaged and summarized in Table 4. Because noticeable
yield plateau was hardly observed during the test, the averaged tensile
stress corresponding to 0.2% plastic strain was defined as the yield
strength ReL. Table 4 indicated that the yield strength of each tension
coupon was about 445 MPa, which was higher than the nominal yield
strength of 420 MPa. In addition, the yield strengths of all steel plates
58 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74

Fig. 4. Tension coupon test: (a) test setup; (b) before fracture; (c) fracture moment.

were higher than those of steel angles and the averaged ultimate
strengths were around 595 MPa. The strains after fracture were unable
to be read from strain gauges and the ultimate deformations were much
larger than what the computer recorded.
2.3. Initial geometric imperfections
Initial geometric imperfections were represented by initial out-of-
straightness and loading eccentricity. Their influence on the buckling
capacity of steel column was sometimes phenomenal. Initial out-of-
straightness was inevitably arising from the manufacturing process,
thus it should not be neglected. In the case of Y-section column, the ini-
tial out-of-straightness was measured as follows:
The deviations u and v of initial bending at the centroid (see Fig. 5)
were calculated by [15]
8 were e0max = L/1405 (for Z440-3), e0min = L/2871 (for Z135P-3) and
>
> ba2 u1 þ ba1 u2
<u ¼
ba
> v ¼ ba2 v1 þ ba1 v2
>
: center and the column center at both ends could be mitigated by using
ba
where u, v were the deviations (in mm) along the two limbs of steel an-
gles; ba1 ,ba2 were the distances from the centroid to the steel angle limb
back and the limb tip, respectively (see Fig. 2); u1 , v1 and u2 , v2 were the
deviations of the steel angle limb back and limb tip, and the positive or
negative sign denoted the deviating direction of the measured section.
The initial rotating angle (in rad) of measured section was derived
from
θ ¼ ½ðu1 −u2 Þ=ba þ ðv2 −v1 Þ=ba þ Δ=bb =3
where θ was the average rotation angle (in rad) of the three limbs of Y-
section; Δ was the transverse deviation perpendicular to the plate width
as shown in Fig. 5(b) and
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 pffiffiffi 2 
Δ¼ bb = 2−u1 −u3 þ v1 þ v3 −bb = 2
where u3 , v3 were the horizontal and vertical distances (in mm) be-
tween the plate limb tip and the welding conjunction. When the Y-sec-
tion made an initial rotation in a counter-clockwise direction, Δ took the
positive value, otherwise the negative value was adopted.
Each column was assumed to have initial bending in a half sine
wave, thus the maximum deformation occurred at the mid-length sec-
tion as illustrated in Fig. 5(c). The deviations u1,v1; u2,v2andu3,v3
correpsonding to the mid-length section of the column were measured
3 times and the whole sectional deviations u, v and θ for each measure-
ment were calculated and obtained by Eqs. (6)–(8). Table 2 listed the
maximum deviations of each column. The maximum initial rotation
among all test columns was max(θmax) = 3.86× 10−3 rad, which was a
small angle that could be ignored. The centroidal deviation of the mea-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sured section was calculated by e0 ¼ u2max þ v2max . The maximum,
minimum and averaged centroidal deviations of all test specimens
e0ave = L/2014, respectively.
ð6Þ The influence of loading eccentricities between the loading end plate
horizontal bubble gradienter, lead plummet and manual adjustment as
well as intelligent modulation.
2.4. Longitudinal residual stresses
The longitudinal residual stresses were measured by using the blind
hole drilling method. Each three standard specimens cut from the same
parent Y-shaped columns with the same nominal cross sections were
ð7Þ prepared, thus 12 specimens in total were fabricated for tests. It was as-
sumed herein that the longitudinal residual stresses did not vary along
the column length direction nor along the thickness direction. Thus
the specimens were representative of arbitrary sections along the col-
umn lengths. Through measurements of the 4 different large sectional
Y-shaped specimens (RS-Z-1-RS-Z-4), the average residual stress ratios
pffiffiffi2 corresponding to the pivotal locations were summarized in Table 5. It
ð8Þ
was found that the maximum compressive residual stresses occurring
at the plate tips and angle tips reached − 241 MPa and − 147 MPa,

Table 4
Measured mechanical properties of tension coupons.

Tension coupons MP-A MP-B

L250 × 26 L220 × 20 B200 × 20 B180 × 18 B160 × 16 B140 × 16

5 5 5 5 5
Modulus of elasticity E (MPa) 2.13 × 10 2.11 × 10 2.12 × 10 2.11 × 10 2.08 × 10 2.10 × 105
Yield strength ReL (MPa) 441 449 454 468 465 459
Ultimate strength Rm (MPa) 591 615 595 593 582 607
ReL/Rm 0.75 0.73 0.76 0.79 0.80 0.76
Ultimate strain ɛu 0.084 0.082 0.080 0.081 0.078 0.077
Elongation ratio δ (%) 24.4 27.0 23.5 24.3 24.7 29.2
 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74 59

Fig. 5. Measurement of initial geometric imperfections of HSSY specimens.

respectively. In addition, the maximum tensile residual stresses occur-
ring at the welding conjunctions of steel angles and plates were taken
as 411 MPa and 400 MPa, respectively. It was also found that the resid-
ual stress distribution of all specimens generally conformed to a qua-
dratic pattern, thus a more generalized and simplified residual stress
model was proposed in Fig. 6. The measured residual stresses were in-
corporated into the validated models in Section 4.3, and the simplified
residual stress model was applied for parametric analyses in Section 5.1.
2.5. Test setup and procedures
Tests on the overall buckling behavior of all specimens were classi-
fied into two different groups: concentric loading and eccentric loading.
A 10,000 kN universal testing machine in addition to computer-aided
control system was introduced to carry out the tests at the Building
Structure Laboratory of Tongji University. The diagram and photograph
of test configuration were illustrated in Fig. 7. The unidirectional cylin-
drical supports at both the top and bottom ends were designed as
hinged restraints except that the rotation around axis-x (major axis)
was restrained as shown in Fig. 8(a) and (b). The column ends and
the cylindrical supports were connected with slots and end plates as
shown in Fig. 8(c) and (d).
Three critical cross sections (the top, bottom and intermediate cross
sections) were chosen to arrange displacement transducers and strain
gauges as shown in Fig. 9. The most unfavorable lateral deflections
along the column lengths were measured by displacement transducers
D1–D3, which were attached perpendicularly to the three limb edges at
the intermediate section. Transducers D4–D7 were deployed at both ba-
sically freely rotational middle edges of the top and bottom end plates
and vertical to the plates in order to measure the in-plane rotating an-
gles (around axis-y; D4, D6 for the top end and D5, D7 for the bottom
end), which was calculated by

θtop1 ¼ arctan½ðRD4 −RD6 Þ=D01  180=π
ð9Þ
θbottom1 ¼ arctan½ðRD5 −RD7 Þ=D01  180=π
where θtop1, θbottom1 were the in-plane rotating angles in degree at the
top and bottom ends; RD4–RD7 were the readings from displacement
transducers D4–D7; D01 was the horizontal distance between D4 and
D6 (or between D5 and D7); and for Z1(P) ~ Z4(P), D01 = 500 mm.
Transducers D8–D9 were placed at both almost non-rotational mid-
dle edges of the end plates and vertical to them, so that the end shorten-
ings of the columns were obtained through averaging the readings of D8
and D9. In addition, the out-of-plane rotating angles can be calculated
by

θtop2 ¼ arctan½ðRD8 −RD9 Þ=D02  180=π
ð10Þ
θbottom2 ¼ 0
where θtop2, θbottom2 were the out-of-plane rotating angles in degree at
the top and bottom ends; RD8 and RD9 were the readings from displace-
ment transducers D8 and D9; D02 was the horizontal distance between
D8 and D9; D02 = 700 mm (for Z1(P) and Z2(P)) and D02 = 500 mm
(for Z3(P) and Z4(P)).
The stress and strain states of the three critical sections (top, bottom
and intermediate sections) were monitored by strain gauges distributed
at the edges of three limbs. 8 strain gauges (S4–S6, S4a–S6b) were uni-
formly arranged at the mid-length section to measure the vertical
strains accurately in addition to validating the geometric imperfections.
The vertical strains at the edges of both end sections were measured by
gauges S1–S3 and S7–S9, respectively. During the whole real-time load-
ing process, the load-end shortening curves, load-deflection curves and

Table 5
Ratios of residual stresses.

Specimen Width-to-thickness ratio Steel angles Steel plates

ba0/ta bb0/tb a1 a2 a3 b1 b2 b3

RS-Z-1 7.8 9.1 0.893 −0.129 −0.281 0.855 −0.302 −0.515

RS-Z-2 7.8 9.1 0.884 −0.134 −0.303 0.847 −0.314 −0.486
RS-Z-3 9.1 7.9 0.915 −0.115 −0.296 0.813 −0.369 −0.497
RS-Z-4 9.1 9.1 0.907 −0.121 −0.327 0.860 −0.338 −0.473

Note: ba0/ta and bb0/tb were the width-to-thickness ratios of the steel angles and the steel plates, respectively. The ratios of residual stresses over the yield strengths of steel angles and steel
plates were denoted as ai =σra/fya and bi =σrb/fyb, i = 1, 2, 3; the subscript i indicated the concerned locations.
60 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74
Fig. 6. Simplified residual stress model.
load-strain curves were drawn by the monitoring system, which guar-
anteed the test stabilities and precisions.
The loading process included two steps: pre-loading and formal
loading. Before the formal loading step, a pre-load with 20% of the pre-
dicted buckling strength was applied to examine the normal operation
of the test apparatus and the compressive strains in the columns and
then unloaded it. For concentric loaded specimens, formal loading was
not conducted until strains at the intermediate cross section almost uni-
formly distributed among the three limbs. For eccentric loaded speci-
mens, strains in the two limbs of steel angles were ensured to be
roughly identical during the pre-loading step. Based on the displace-
ment loading control method, the external force was imposed on the
specimens at a rate of 1 mm/min until the maximum load was attained.
After that, the loading rate was speeded up until the external force
dropped to 70% of the maximum load. Finally, the specimen was
unloaded and disassembled to finish an individual column test. The
strains, displacements, deformations of the specimens and the imposed
forces were rigorously monitored and recorded by computer during the
Fig. 7. Test configuration: (a) diagram; (b) photograph of real test setup.
whole loading process. It was observed that almost all the failure modes
of the axial loaded and eccentric loaded columns demonstrated similar
overall buckling deformations. The local buckling was not observed in
evidence and the out-of-plane lateral deflections around the major
axis (axis-x) were so small that the columns were preferably considered
as overall in-plane buckling as were designed and expected.
3. Test results and discussion
3.1. Overall buckling deformations
As expected, all test specimens were observed to fail in the form of
overall buckling around the minor axis (axis-y) and the local-overall in-
teraction buckling did not occur on any specimen. The typical buckling
deformations and failure modes with some similarities between Z340-
1 and Z340P-1 were obtained from the test results and from the numer-
ical approach as shown in Fig. 10. It can be seen that one limb of the steel
angles started to yield at the earliest and then entered into the plastic
stage. With the increase of external loads, the plastic zone continued
to develop around the intermediate section and followed by the other
two limbs' buckling at the mid-section. As a result, the maximum com-
pressive stresses occurred at the middle of each limb.
The typical load versus deflection curves of Z340-1 and Z340P-1
were shown in Fig. 11, in which the horizontal ordinate represented
the mid-length lateral deflections of three limbs read from D1–D3
(see Fig. 9), the positive and negative values were the deflections
away from and towards the displacement transducers, and the longitu-
dinal ordinate represented the axial or eccentric loads imposed at the
top ends of columns Z340-1 and Z340P-1. It was concluded from the
curve directions and trends that almost no torsional deformations oc-
curring on the columns. Furthermore, the load-deflection curves com-
prised of almost linearly ascending part and mildly descending part,
which indicated that the columns had very sound ductility and isotropy.
This could be testified by the load-vertical displacement curves and
load-strain curves as presented in Figs. 12 and 13. Moreover, the bend-
ing directions and stress states were easily distinguished from the fig-
ures. The maximum tensile strains and transverse deflections were
found to occur at the mid-length sections of almost all test specimens.
As seen from Fig. 13, the compressive stain (recorded by S4) of one
 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74 61

Fig. 8. End restraint conditions: (a) top; (b) bottom; (c) slot without eccentricity; (d) slot with eccentricity.

limb fringe for steel angles at the intermediate section did not start to
decrease from the maximum value to the tensile strain until the peak
load was reached, while the other limb fringe (recorded by S6) was con-
stantly in compressive status and the compressive strain had been in-
creasing to the maximum value until the failure of the specimen. The
strains around the centroid at the intermediate section (recorded by
S4b, S5a and S6b) represented the average strain state of the entire sec-
tion. For Z340-1 and Z340P-1, the average values of the recorded strains
corresponding to the peak loads were 194.2% and 56.3% of their yield
strains ReL/E, respectively. The load-rotation curves of both end supports
for Z340-1 and Z340P-1 were shown in Fig. 14, in which the longitudinal
ordinate represented the axial or eccentric load, and the horizontal ordi-
nate represented the in-plane rotation around the minor axis or the out-
of-plane rotation around the major axis.
3.2. Overall buckling capacities
The overall buckling capacities of all specimens subject to axial or ec-
centric loadings were obtained both from the test results and FE analy-
ses as summarized in Table 6. The buckling factor was calculated by
where fya, fyb were the measured yield strengths of steel angles and steel
plate, Aa and Ab were the corresponding measured cross sectional areas.
The slenderness ratio λmax = λy =L0y/iy and the non-dimensional slen-
qffiffiffiffiffiffiffiffiffiffi
derness λn ¼ ðλ=πÞ f y =E were summarized in Table 6 as well.
3.3. Test results compared with design codes
The Chinese Standard ⟨⟨Code for design of steel structures (GB
50017-2003)⟩⟩ [5] and the British Standard ⟨⟨Eurocode 3: Design of
steel structures: Part 1-1: General rules and rules for buildings (BS EN
1993-1-1)⟩⟩ [6] were selected to predict the overall buckling capacities
of HSSY columns. In addition, 6 other codes and standards including
American specifications ASCE 10-97 [3] and ANSI/AISC 360-10 [8], Cana-
dian standard CSA S16-09 [10], Australian standard AS4100 [9], British
standard BS 5950-1 [7] and Japanese code AIJ 2010 [11] were also intro-
duced. The predicted results based on the measured initial imperfec-
tions and material properties were compared with the test results for
assessing the reasonability and accuracies of the experimental design
and process.

 
φt ¼ Nu = f ya Aa þ f yb Ab ð11Þ

Fig. 9. Layout of displacement transducers and strain gauges.

62 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74

D4 D6

D5 D7
D9

Before test After buckling Top view outlines Deformed shape Element contour plot
(a)

D4 D6

D5 D7
D9

Before test After buckling Top view outlines Deformed shape Element contour plot
(b)

Fig. 10. Buckling deformations and failure modes: (a) Z340-1; (b) Z340P-1.

5000 4000
Eccentric load (kN)

4000
Axial load (kN)

3000
3000
2000
2000 D1 D1
D2 1000 D2
1000
D3 D3
0 0
-50 -40 -30 -20 -10 0 10 20 30 40 50 -60 -30 0 30 60 90
Mid-length lateral deflection (mm) Mid-length lateral deflection (mm)
(a) (b)

Fig. 11. Load-lateral deflection curves: (a) Z340-1; (b) Z340P-1.

 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74 63

(a)
5000 5000

4000 4000

Axial load (kN)

Axial load (kN)
3000 3000
D4
2000 D5 2000 D8
D6 D9
1000 1000
D7 End shortening
0 0
-30 -20 -10 0 10 -15 -12 -9 -6 -3 0
Displacement (mm) Displacement (mm)
(b)
4000 4000

Eccentric load (kN)

Eccentric load (kN)

3000 3000

2000 D4 2000
D5 D8
1000 D6 1000 D9
D7 End shortening
0 0
-60 -40 -20 0 20 -40 -30 -20 -10 0
Displacement (mm) Displacement (mm)

Fig. 12. Load-vertical displacement curves: (a) Z340-1; (b) Z340P-1.

3.3.1. Compared with GB 50017-2003 [10], Australian standard AS4100 [9], British standard BS 5950-1 [7]
The buckling capacity of a steel column according to the Chinese Na- and Japanese code AIJ 2010 [11] were presented herein for comparison
tional Standard GB 50017-2003 [5] was calculated by as well. The corresponding formulae for buckling factors and curve fit-
tings were summarized in Table 7. The buckling curves obtained from
N ¼ φA f y =γR ð12Þ these codes and standards were made comparison with the test buck-
ling factors of axial compression columns as shown in Fig. 15. The test
where γR was the resistance partial factor, for steel Q420, γR = 1.111. φ results were higher than the buckling curve a and curve b of GB
was the stability factor of axial compression members calculated from 50017-2003 by 4.1% and by 11.2% on average, respectively. Although
the following formulae there was one experimental data point lower than curve a by 4.0%, it
8 qffiffiffiffiffiffiffiffiffiffi was safe enough in general to predict the buckling capacities by using
> 2
< 1−α"1 λn ; for λn ¼ ðλ=πÞ r
> f y =E ≤0:215
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi#
curve a. For Eurocode 3, the test buckling factors were higher than
φ¼ 1    2 curve a0 and curve a by 1.1% and 4.5% on average. However, it was
>
> 2 2 2
: 2 α 2 þ α 3 λn þ λn − α 2 þ α 3 λn þ λn −4λn ; forλn N0:215 more justifiable to use curve a to predict the buckling strengths of the
2λn
columns, since only one data point was lower than curve a, while
ð13Þ
three out of 12 points were lower than curve a0. Besides, curve b was
too conservative to predict the strengths with a deviation of 9.8% on av-
where λn was the non-dimensional slenderness ratio, α1, α2, α3 were
erage. For 6 other codes and standards, the test results were higher than
the imperfection factors determined by the cross-section classification
curve of ASCE10-97 by 2.7%, curve of ANSI/AISC 360-10 by 7.5%, curve of
(a, b, c and d) in Table 5.1.2 of this code [5].
CSA S16-09 by 8.4%, curves of AS4100 by 0.6% (for αb = − 1) and 4.8%
(for αb = − 0.5), curves of BS 5950-1 by 1.5% (for curve a) and 5.9%
3.3.2. Compared with Eurocode 3
(for curve b), and curve of AIJ 2010 by 14.6%. It should be noted that
According to Eurocode 3 [6], the buckling resistance of an axial com-
the test slenderness ratios were 35 and 40, which only represented a
pression member was taken as
very limited range of the most commonly-used large-section HSS Y-
Nb;Rd ¼ χAf y =γ M1 ð14Þ shaped columns. Therefore, the test data were not sufficient to assess
the ultimate buckling capacities of the columns, and more extended
where γM1 was the partial factor for resistance of members to instability data corresponding to broader slenderness ratios were necessary by
assessed by member checks, χ was the reduction factor for the relevant means of numerical approach.
buckling mode and was calculated according to
8 qffiffiffiffiffiffiffiffiffiffi
>
>  ¼ λ f =E ≤0:2
< 1; for λ
4. Finite element model and validation
y
 π
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
χ¼ ð15Þ
>
> 2  2 ; for λN0:2
 4.1. Simplified model of steel
: 1= Φ þ Φ −λ

The material mechanical properties of steel angles and steel plates

2
where Φ ¼ 0:5½1 þ αðλ−0:2Þ þ λ , α was an imperfection factor corre-
sponding to the appropriate buckling curve (a0, a, b, c and d) and was
obtained from Tables 6.1 and 6.2 of this code [6].
3.3.3. Compared with other codes
Six other codes and standards such as American specifications ASCE
10-97 [3] and ANSI/AISC 360-10 [8], Canadian standard CSA S16-09
based on the tension coupon test were employed in the finite element
analysis (FEA). The multi-linear kinematic hardening plasticity models
(MKIN) complying with von Mises yield criterion were applied to the
stress-strain curves as proposed in Fig. 16. The elastic modulus and
hardening modulus corresponding to Z series and ZP series were obtain-
ed from the tension coupon test, and the Poisson's ratio was taken as
0.3.
64 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74

(a)
5000 5000

4000 4000

Axial load (kN)

Axial load (kN)
S4
3000 3000 S4a
S4b
S1 S5
2000 2000 S5a
S2 S6
1000 S3 1000 S6a
S6b
0 0
-8000 -6000 -4000 -2000 0 2000 -12000 -9000 -6000 -3000 0 3000
Strain (µε) Strain (µε)
5000 5000

4000 4000
Axial load (kN)

Axial load (kN)

S4b
3000 3000
S7 S5a
2000 S8 2000 S6b
S9 Averaged
1000 1000
Yield strain
0 0
-3000 -2000 -1000 0 1000 2000 -12000 -9000 -6000 -3000 0 3000
Strain (µε) Strain (µε)
(b)
4000 4000
Eccentric load (kN)

Eccentric load (kN)

3000 3000
S4
S4a
2000 2000 S4b
S1 S5
S2 S5a
1000 1000 S6
S3 S6a
S6b
0 0
-15000 -10000 -5000 0 5000 -25000-20000-15000-10000 -5000 0 5000
Strain (µε) Strain (µε)
4000 4000
Eccentric load (kN)
Eccentric load (kN)

3000 3000
S4b
2000 2000 S5a
S7
S6b
S8
1000 1000 Yield strain
S9
Averaged
0 0
-4000 -3000 -2000 -1000 0 1000 -6000 -4000 -2000 0 2000
Strain (µε) Strain (µε)

Fig. 13. Load-strain curves: (a) Z340-1; (b) Z340P-1.

4.2. Meshes and boundary conditions
All test specimens were modelled by SHELL181 in the finite ele-
ment (FE) software ANSYS. SHELL181 was well-suited for analyzing
linear, large rotation and large strain nonlinear problems of thin-
walled structures. It was a four-node element with six degrees of
freedom (DOFs) at each node: three for translations and three for ro-
tations. Any user-defined cross section can be meshed into an appro-
priate number of elements. Each specimen herein was meshed into
10 elements along the limb width and 100 elements along the axial
direction, respectively. The typical element mesh was shown in Fig.
17.

Fig. 14. Load-rotation curves: (a) Z340-1; (b) Z340P-1.

 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74 65

Table 6 4.3. Model validation

Comparison of test buckling strengths and FEA results.

Specimen λmax λn Nu,t (kN) Nu,FEA (kN) φt φFEA Nu,FEA/Nu,t The measured initial bending and simplified residual stress
(a) Z series
model were taken into account in the validated FE models. Based
Z135-1 35.0 0.511 6616 6720 0.946 0.961 1.016 on the equilibrium principle of sectional resultant force and resul-
Z135-2 34.8 0.508 6736 6723 0.963 0.962 0.998 tant moment, the residual stress existing in each element was con-
Z135-3 34.9 0.509 6639 6723 0.950 0.962 1.013 verted from the simplified model (see Fig. 8) into the validated
Z235-1 34.8 0.506 7276 7062 0.998 0.969 0.971
model as shown in Fig. 19.
Z235-2 35.0 0.509 6967 7059 0.956 0.968 1.013
Z235-3 35.0 0.509 6941 7059 0.952 0.968 1.017 The predicted overall buckling strengths and the test results of all
Z340-1 39.3 0.579 4608 4512 0.953 0.933 0.979 specimens were compared as shown in Table 6. Fig. 20 provides the
Z340-2 39.5 0.582 4220 4509 0.873 0.933 1.068 fitting effect of the numerical results and the test results. It was
Z340-3 39.5 0.582 4827 4509 0.999 0.933 0.934 seen that the numerical solutions showed a good agreement with
Z440-1 40.1 0.592 4733 4670 0.947 0.935 0.987
Z440-2 40.1 0.592 4739 4670 0.949 0.935 0.985
the experimental results. For axial compression members, the
Z440-3 40.0 0.591 4840 4673 0.969 0.935 0.965 mean ratio of numerical result to experimental result was 0.996
Mean value 0.996 with a standard deviation of 0.034 and for eccentric compression
Standard deviation 0.034 members, the mean value was 0.994 with a standard deviation of
(b) ZP series 0.057. The differences between the test and the numerical buckling
Z135P-1 34.8 0.508 5086 4992 0.982 strengths of columns were attributed to several reasons. For in-
Z135P-2 35.0 0.511 4972 4989 1.003 stance, when the influence of the residual stresses made the column
Z135P-3 35.0 0.511 5004 4989 0.997
bending towards the direction of the initial out-of-straightness, a
Z235P-1 34.9 0.508 5322 5658 1.063
Z235P-2 34.9 0.508 5974 5658 0.947 lower buckling strength will be reached and vice versa. In addition,
Z235P-3 34.9 0.508 5628 5658 1.005 the different combinations of residual stresses and initial geometric
Z340P-1 39.4 0.580 3759 3971 1.056 imperfections will result in an opposite bending direction as seen
Z340P-2 39.5 0.582 3978 3969 0.998 in Fig. 21. The comparison between test result and numerical result
Z340P-3 39.5 0.582 3650 3969 1.087
of load-end shortening curves of Z340-1 and Z340P-1 were shown
Z440P-1 40.1 0.592 4304 4146 0.963
Z440P-2 40.1 0.592 4440 4146 0.934 in Fig. 22. Their load-average lateral deflection curves at the interme-
Z440P-3 40.0 0.591 4670 4146 0.888 diate sections were shown in Fig. 23, in which the horizontal ordi-
Mean value 0.994 nate represented the average lateral deflections of three limbs
Standard deviation 0.057
around the minor axis. Except for the initial loading stage of the col-
umns, the numerical results agreed well with the test results from
To keep consistent with the real test setup, the boundary conditions of Figs. 22 and 23. It can be seen that the strain hardening effect was
both ends were simulated by prescribing the applied translations, rota- distinct after the attainment of the ultimate strength. The disparity
tions and forces as illustrated in Figs. 17 and 18. Both ends of the columns of the initial loading deformation might be caused by the instability
were set as fixed around axis-x and axis-z, but freely rotating around axis- of the loading eccentricities and the unevenness of both ends. In gen-
y. Besides, all the node translations along axis-y at both ends were fixed eral, the FE models can well predict the overall buckling capacities
other than the z-displacement of the top nodes. The concentrated axial and deformations of the specimens under both axial and eccentric
or eccentric load was imposed at the center of the top end plate, through compressive loadings. Therefore, the numerical approach and the
which the external force was well-distributed along the three limbs. FE models were accurate and reliable.

Table 7
Comparison of design formulae of buckling factor.

Code Calculation formula of buckling factor

(
ASCE 10-97 1−0:5ðλ=C C Þ2 ; λ≤C C
φ ¼ π2 E
λ2 F
; λNC C
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
y

whereλ = KL/r was the slenderness ratio, C C ¼ π 2E= F y , Fy was the minimum guaranteed yield stress.
(
ANSI/AISC 2
0:658λ ; λ≤1:5
360-10 φ¼ 2
0:877=λ ; λN1:5
pffiffiffiffiffiffiffiffiffiffiffi
whereλ ¼ ðλ=πÞ F y =E was the non-dimensional slenderness.
CSA S16-09 2n −1=n
φ ¼ ð1 þ λ Þ
pffiffiffiffiffiffiffiffiffiffiffi
where λ ¼ λπ F y =E was the non-dimensional slenderness parameter, n was the parameter for compressive resistance.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
AS4100 2
α c ¼ ξf1− 1−½90=ðξλÞ g
where αc was the compression member slenderness reduction factor, ξ= [(λ/90)2 + 1 +η]/[2(λ/90)2] was the compression member factor, λ = λn + αaαbwas
qffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the slenderness ratio, η = 0.00326(λ − 13.5) ≥ 0 was the compression member imperfection factor, λn ¼ ðle =rÞ k f f y =250 was the modified compression
member slenderness, αa = [2100(λn − 13.5)]/(λ2n − 15.3λn + 2050) was the compression member factor.
8
BS 5950-1 < 1:0; λ≤0:2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
φ¼ 2 2 2 2
: 12 ½ðλ þ η þ 1Þ− ðλ þ η þ 1Þ −4λ ; λN0:2
2λ
pffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffi
απ
Whereλ ¼ ðλ=πÞ F y =E was the non-dimensional slenderness, η ¼ 1000 ðλ−0:2Þ E=f y N0 was the Perry factor denoting an equivalent imperfection, α was the
imperfection factor.
AIJ 2010
8
>
< 1:0; λbλp
φ¼ 1−0:5ðλ−λp Þ=ðλe −λp Þ; λp bλ ≤λe
>
: 2
1=ð1:2λ Þ; λNλe
qffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffi
Whereλ ¼ ðλ=πÞ f y =Ewas the non-dimensional slenderness ratio, λp ¼ 0:15was the plastic limitation, λe ¼ 1= 0:6 ¼ 1:291was the elastic limitation.
66 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74

Fig. 15. Comparison of test buckling factors with code buckling curves according to: (a) Eurocode 3; (b) GB 50017-2003; (c) ASCE 10-97; (d) ANSI/AISC 360-10; (e) CSA S16-09; (f)
AS4100; (g) BS 5950-1; (h) AIJ 2010.

5. Parametric analyses and buckling curves
5.1. Parametric analyses
The initial geometric imperfections and the residual stresses were
two primary factors influencing the ultimate buckling capacities of
HSSY columns. The validated FE models were established to perform
parametric analyses of the influence of the two factors on the overall
buckling strengths and failure modes. Numerical solutions based on
the validated FE models were carried out by ANSYS. These specimens
had the same width-to-thickness ratios as those of Z1–Z4 series (see
Table 1) and contained 27 slenderness ratios ranging from 20 to 150.
 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74 67

Fig. 16. Stress-strain curves of multi-linear model for HSSY columns: (a) Steel angles; (b) steel plates.

Thus a total of 108 columns under axial and eccentric compression spec-
imens were investigated and the buckling factors were calculated ac-
cording to the FE results.
Since the initial out-of-straightness was assumed to take the maxi-
mum tolerances of material, fabrication and erection specified in the
current codes [3–5,9], the allowable bending value taken as L/1000
was adopted in the FE models, where L was the column length between
the brace or framing point. The sectional residual stresses were incorpo-
rated by inputting each average value to its corresponding divided
SHELL181 element in Figs. 6, 17 and 19. The stress-strain relation
models of steel angles and steel plates were employed as shown in
Fig. 16, in which E, εy, εu, ReL and Rm were taken from the results of the
BC on top (x=0, z=L0):
BC at bottom (x=0, z=0):
Ux=Uy =0
Ux=Uy =Uz=0
ROTX=ROTZ =0
ROTX=ROTZ =0
Imposed load F
tension coupon tests (Table 4). The Poisson's ratio was taken as 0.3.
Both ends of the columns were partially hinged as illustrated in Figs.
17 and 18. An axial or eccentric compression force after appropriate
magnification of the buckling load obtained from elastic buckling anal-
ysis was imposed at the top of the end plate. The failure modes and
the ultimate bucking strengths were obtained to evaluate the overall
buckling behavior of HSSY columns.
The influence of loading eccentricities on overall buckling capacities
of ZP series was analyzed by FE method as shown in Fig. 24. The hori-
zontal ordinate was represented by loading eccentricity ratioβ, which
was defined as β = eL/ba (eL was the loading eccentricity and ba was
the limb width of steel angles), and the longitudinal ordinate in Fig.
24(b) was represented by the decline percentage of overall buckling ca-
pacity Nu. It was found that the overall buckling capacities of columns
Z135P – Z440P declined by 16.7%, 12.5%, 15.5% and 11.9% on average
when the loading eccentricity ratio reached 0.1, and declined by
25.8%, 19.8%, 23.0% and 18.2% on average when β was 0.2. The decline
percentage of Nu did not exceed 10% when β was no more than 0.05,
which indicated that majority buckling factors of axial compression col-
umns were still above curve b specified in GB 50017-2003 and in
Eurocode 3 even if 10% of Nu was lost due to the occurrence of connec-
tion eccentricity at the joint between main members of latticed trans-
mission towers. Therefore, an allowance of 0.05ba for connection
eccentricity was tolerable to design HSSY columns in accordance with
curve b of GB 50017-2003 and of Eurocode 3.

Fig. 17. FE model and boundary conditions of Y-section column.

Rigid platen surface

BC imposed on axis-y
Centroid

Shell element in flange

Y-section column

Fig. 18. Boundary conditions illustration. Fig. 19. Residual stress model applied in FEA.
68 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74

Fig. 20. Comparison of numerical results and test results: (a) Z series; (b) ZP series.

5.2. Comparison with existing codes
The ultimate buckling capacities of the 108 columns obtained from
the parametric analyses were made comparison with the predicted
buckling strengths according to the current design codes such as GB
50017-2003 [5], Eurocode 3 [6] and other codes [3,7–11] on non-di-
mensional basis. The buckling curves considering the initial imperfec-
tions and the real boundary conditions in the FE models were
compared with the code design curves as shown in Fig. 25. It was
found that the buckling factors corresponding to the cross sections stud-
ied by tests were in compact and reversed “S” shape. The exceedance
percentage of the average FEA buckling factors over the code results
((φFEA / φcode − 1) × 100 (%)) were summarized in Fig. 25(i). For
Eurocode 3, the average buckling factor of Z1–Z4 from FE models was
0.2% higher than curve a0 with a standard deviation of 0.022, and 5.7%
higher than curve a with a standard deviation of 0.016 on average. Con-
sequently, curve a0 was suitable for predicting the buckling strengths of
HSSY columns, while curves a, b, c and d were too conservative. For GB
50017–2003, the FE value was on average 1.2% higher than curve a, thus
it was suitable to use curve a to predict the buckling strengths. As for BS
5950-1, the FE value was on average 1.0% and 8.1% higher than curve a
and curve b with a standard deviation of 0.020 and 0.022, respectively.
Thus curve a was more accurate for buckling strength prediction. The
average FE values were higher than the buckling curves of the other
four codes (i.e. higher than ANSI/AISC 360-10 by 6.0%, CSA S16-09 (for
n = 1.34) by 10.8%, AS4100 (forαb = − 0.5) by 6.0%, AIJ 2010 by
11.1%), while lower than the design values of ASCE 10–97 by 4.9%,
AS4100 (forαb = − 1) by − 0.9% and CSA S16-09 (for n = 2.24) by
3.3%. Therefore, it was conservative and safe enough to design the col-
umns around the minor axis (axis-y) according to the former, but not
safe according to the latter.
5.3. Proposed buckling curves
The overall buckling behavior of 420 MPa large section HSSY columns
had never been investigated before, thus their buckling curves were not
categorized in any current design codes or specifications. In this study,
six new buckling curves corresponding to six codes other than ASCE
10-97 and AIJ 2010 were proposed for the design of 420 MPa large-
section HSSY columns by nonlinear regression analysis on the basis of

Fig. 21. Different buckling directions: (a) Z235P-1 (left); (b) Z235P-2 (right); Z235P-3 (right).
 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74 69

5000 4000

Eccentric load (kN)

4000
3000

Axial load (kN)

3000
2000
2000 Test result Test result
1000 1000
Numerical result Numerical result
0 0
-12 -9 -6 -3 0 -20 -15 -10 -5 0
End shortening (mm) End shortening (mm)
(a) (b)

Fig. 22. Load-end shortening curves: (a) Z340-1; (b) Z340P-1.

5000 4000

Eccentric load (kN)

4000
3000
Axial load (kN)

3000
2000
2000 Test result Test result
1000
1000 Numerical result Numerical result
0 0
0 20 40 60 0 20 40 60
Average lateral deflection (mm) Average lateral deflection (mm)
(a) (b)

Fig. 23. Load-average lateral deflection curves: (a) Z340-1; (b) Z340P-1.

the FE results of 108 columns, as shown in Fig. 26. The regression models
were derived from the following Perry-Robertson's formula
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
h 2
i h i
2 2
φ ¼ 1 þ ð1 þ ε 0 Þ=λ =2− 1 þ ð1 þ ε0 Þ=λ =4−1=λ
where φ was the non-dimensional buckling factor,λ was the non-dimen-
sional slenderness and ε0 was the equivalent imperfection which incor-
porated various initial imperfections (i.e. initial bending, initial rotation,
initial loading eccentricity, residual stresses and material heterogeneity).
For GB 50017-2003, the regression results of the imperfection fac-
tors were: α1 = 0.030, α2 = 0.956, α3 = 0.161, thus the equivalent im-
perfection ε0 ¼ α 3 λ þ α 2 −1 ¼ 0:161λn −0:044 . Its corresponding
2 proposed curve was higher than curve a (see Fig. 26(b)). For Eurocode
ð16Þ
3, the regressed imperfection factor α = 0.144, thusε0 ¼ αðλ−0:2Þ ¼ 0:
144λ−0:029 , which was between curve a0 (α = 0.13) and curve a
(α = 0.21). Therefore, the proposed buckling curve was most approxi-
mate to curve a0 as shown in Fig. 26(a). For BS 5950-1, the regressed im-
perfection factor α = 2.124 N 2.0(curve a), thus the proposed buckling
curve was very close to curve a as shown in Fig. 26(f). With respect to

Fig. 24. Influence of loading eccentricity on buckling capacity.

70 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74

Fig. 25. Comparison of FEA results with current code design curves according to: (a) Eurocode 3; (b) GB 50017-2003; (c) ASCE 10-97; (d) ANSI/AISC 360-10; (e) CSA S16-09; (f) AS4100;
(g) BS 5950-1; (h) AIJ 2010; (i) summary.
 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74 71

Fig. 26. Comparison of proposed buckling curves with current code design curves and FEA results according to: (a) Eurocode 3; (b) GB 50017-2003; (c) ANSI/AISC 360-10; (d) CSA S16-09;
(e) AS4100; (f) BS 5950-1.

the other three codes, the regression results (including imperfection
factors, equivalent imperfections and goodness of fit) of the buckling
factors were summarized in Table 8, in which the notations were the
same as those in Table 7. The proposed curves were above their corre-
sponding column curves of ANSI/AISC 360-10 (see Fig. 26(c)), CSA
S16-09 (see Fig. 26(d)) and AS4100 (for αb = − 0.5, see Fig. 26(e)). Ac-
tually, the proposed curves corresponding to different codes (except for
ANSI/AISC 360-10) resembled each other, because the expressions of
equivalent imperfection ε0 in Perry-Robertson's formula were very
similar.

Table 8
Regression results for buckling curves by using FEA data.

Code/standard Regression results for buckling curves by using FEA data

Eurocode 3 α = 0.144 (0.139, 0.150), R2 = 99.78%, ε 0 ¼ αðλ−0:2Þ ¼ 0:144λ−0:029

GB 50017-2003 α2 = 0.956 (0.948, 0.965), α3 = 0.161 (0.150, 0.172), R2 = 99.79%,α1 = 0.030,
ε0 ¼ 0:161λn −0:044
ANSI/AISC 360-10 a = 0.685 (0.668, 0.690), b = 0.962 (0.948, 0.975), R2 = 98.08%
CSA S16-09 n = 1.926 (1.899, 1.953), R2 = 99.84%
AS4100 αb = −0.861 (−0.867, −0.855), R2 = 99.74%
BS 5950-1 α = 2.124 (2.046, 2.203), R2 = 99.78%,
pffiffiffiffiffiffiffiffi
απ E= f
ε0 ¼ 1000 y ðλ−0:2Þ ¼ 0:149λ−0:030
72 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74

Table 9
Comparison of section efficiency between test results and numerical results.

Specimen Nu,t (MPa) γave ,t (MPa) Nu,FEA (MPa) γave , FEA (MPa) Specimen Nu,t (MPa) γave ,t (MPa) Nu,FEA (MPa) γave , FEA (MPa)

Z135-1 423 426 429 429 M1160-1 391 379 356 356
Z135-2 430 429 M1160-2 362 356
Z135-3 424 429 M1160-3 384 356
Z135P-1 325 321 319 319 M8830-1 444 451 435 435
Z135P-2 318 319 M8830-2 459 435
Z135P-3 320 319 M8830-3 450 435
Z235-1 443 430 430 430 M8840-1 447 418 419 419
Z235-2 424 430 M8840-2 387 419
Z235-3 423 430 M8840-3 419 419
Z235P-1 324 344 345 345 M8850-1 408 391 395 395
Z235P-2 364 345 M8850-2 400 395
Z235P-3 343 345 M8850-3 366 395
Z340-1 430 425 421 421 M8860-1 404 379 357 357
Z340-2 394 421 M8860-2 340 357
Z340-3 450 421 M8860-3 392 357
Z340P-1 351 354 371 370 M8870-1 365 367 311 311
Z340P-2 371 370 M8870-2 363 311
Z340P-3 341 370 M8870-3 374 311
Z440-1 429 432 423 423 M8880-1 356 332 265 265
Z440-2 429 423 M8880-2 355 265
Z440-3 439 423 M8880-3 287 265
Z440P-1 390 405 376 376 M6630-1 412 408 436 436
Z440P-2 402 376 M6630-2 406 436
Z440P-3 423 376 M6630-3 406 436
M1130-1 408 414 425 425 M6640-1 391 399 424 424
M1130-2 397 425 M6640-2 407 424
M1130-3 436 425 M6640-3 399 424
M1135-1 409 410 419 419 M6650-1 402 387 405 405
M1135-2 418 419 M6650-2 387 405
M1135-3 402 419 M6650-3 373 405
M1140-1 398 403 412 412 M6660-1 312 370 375 375
M1140-2 389 412 M6660-2 413 375
M1140-3 421 412 M6660-3 386 375
M1145-1 401 393 402 402 M6670-1 368 365 315 315
M1145-2 385 402 M6670-2 325 315
M1145-3 393 402 M6670-3 402 315
M1150-1 402 387 390 390 M6680-1 272 278 263 263
M1150-2 386 390 M6680-2 301 263
M1150-3 372 390 M6680-3 263 263
M1155-1 380 380 375 375
M1155-2 372 375
M1155-3 387 375

6. Section efficiency and optimization
6.1. Section efficiency
Section efficiency herein was defined as the ultimate buckling
strength carried by unit cross-sectional area of the column, that was,
γ = Nu/A. It was a comprehensive index relating with many factors
(such as steel grade, section type, initial imperfection, boundary
condition and buckling factor) that evaluated the section utilization
ratio. Comparison of the section efficiencies between test results and
numerical results of medium-section (M series, M11: 140 × 10–
100 × 10, M88: 125 × 10–80 × 8, M66: 125 × 10–60 × 6) [34] and
large-section (Z and ZP series) HSSY columns was summarized in
Table 9. The average section efficiencies with the same cross section
and the same slenderness ratio based on the test results and numerical
results was shown in Figs. 27 and 28.

Fig. 27. Average section efficiencies of medium-section and large-section HSSY columns.
 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74 73

Fig. 28. Average section efficiencies based on numerical results.

It was found from Fig. 27(a) that the average section efficiencies of
Z2(P), Z4(P) were higher than those of Z1(P), Z3(P) by 0.9%(7.2%),
1.6%(14.4%) respectively, which indicated that under the premise of the
same cross-sectional steel angles, the section utilization ratio could be ef-
fectively heightened with the appropriate increase of the length of steel
plate (so that the centroid of Y-section and its shear center could get as
close as possible). Fig. 27(b) revealed that M88 had the highest average
section efficiencies among all M-series columns [34], from which we
can conclude that when the moments of inertia around axis-x and around
axis-y were approximately equal (Ix/Iy =1.016), the cross section reached
the maximum utilization ratio. Fig. 28 presented the average section
efficiencies based on numerical results. All curves resembled their cor-
responding column curves, because the section efficiency γ = Nu/A
and the buckling factor φ =Nu/(fyA) had similar graphical shape.
6.2. Section optimal selection
Section optimization and selection was of great significance from
economic and practical perspectives with respect to design of HSSY col-
umns. Based on the optimization principles concluded from Section 6.1
Table 10
Optimized and selected columns with potentially high section efficiency.
and within the limits of width-to-thickness ratios, a certain number of
cross sections having potentially high section efficiencies were selected
from thousands of sectional combinations as summarized in Table 10.
The width-to-thickness ratios of steel angles (ba0/ta) and steel plates
(bb0/tb) were limited between 7.0 and 9.4. The moment of inertia around
axis-x (Ix) to that around axis-y (Iy) was limited between 1.0 and 1.2 for
large-section columns, and between 0.9 and 1.1 for medium-section col-
umns. These columns provided reference for the design and application
of practical main members in latticed transmission towers.
7. Conclusions
A package of full-scale tests on 420 MPa large-section HSSY columns
under axial and eccentric compression was conducted to investigate
their overall buckling behavior in-depth. The failure modes and buck-
ling strengths were acquired and compared with the numerical results
based on validated FE models. The buckling factors corresponding to
the test data and numerical results were calculated and made compari-
son with the current code design curves. The buckling curves and their
design formulae were obtained through nonlinear regression method.
Section efficiency and optimization were formulated and criteria for
section selection were put forward. Conclusions from the study were
summarized as:

Steel angles Steel plate A Ix Iy Ix/Iy 1) All test specimens failed in overall buckling mode around the minor
(mm2) (×108 (×108 axis and local buckling did not appear, because all designed width-
ba ta ba0/ta bb tb bb0/tb
mm4) mm4) to-thickness ratios were within the code limits. Rotation compo-
(mm) (mm) (mm) (mm)
nents were so insignificant that can be neglected.
280 28 8.1 200 20 9.1 19,003.7 2.02 1.75 1.15
280 28 8.1 180 22 7.3 18,963.7 1.85 1.75 1.06 2) The buckling deformations and ultimate strengths from numerical
280 26 8.8 180 20 8.1 17,575.4 1.69 1.64 1.03 results showed a good agreement with those from the test results,
250 28 7.0 180 18 9.1 16,563.7 1.37 1.22 1.12 thus the FE model and analysis procedure were reliable to predict
250 28 7.0 220 22 9.1 15,563.7 1.38 1.22 1.12 the test results satisfactorily.
250 26 7.7 180 18 9.1 15,655.4 1.32 1.15 1.15
250 26 7.7 220 24 8.3 14,655.4 1.36 1.15 1.18
3) Residual stresses existing in cross section can be simulated as qua-
250 24 8.5 160 18 8.0 14,380.4 1.09 1.07 1.01 dratic distribution. The peak compressive stresses were found occur-
240 26 7.3 160 18 8.0 14,775.4 1.06 1.01 1.04 ring at each limb tip and reached about 0.3fya for steel angles and
240 24 8.1 160 16 9.1 13,580.4 0.95 0.94 1.00 0.5fyb for steel plates, respectively. The peak tensile stresses arising
240 24 8.1 160 18 8.0 13,900.4 1.02 0.94 1.08
in the welding conjunction of Y-section reached about 0.9fya for
240 22 9.0 160 16 9.1 12,698.8 0.91 0.88 1.04
220 22 8.1 160 16 9.1 11,818.8 0.80 0.67 1.19 steel angles and 0.85fyb for steel plates, respectively.
200 20 8.1 125 14 8.1 9400.5 0.43 0.46 0.95 4) The test buckling strengths of most columns on non-dimensional
200 20 8.1 130 14 8.4 9470.5 0.45 0.46 1.00 basis were higher than curve a of GB 50017-2003 and of Eurocode
200 18 9.2 120 14 7.7 8595.5 0.39 0.42 0.94 3, as well as higher than curve b of BS 5950-1, curves of ASCE10-
200 18 9.2 125 14 8.1 8665.5 0.41 0.42 0.99
200 18 9.2 130 14 8.4 8735.5 0.43 0.42 1.04
97, ANSI/AISC 360-1, CSA S16-09, AS4100 (for αb = − 0.5) and AIJ
180 18 8.1 110 12 8.3 7515.5 0.27 0.30 0.90 2010. Buckling factors and column curves derived from a large num-
180 18 8.1 120 12 9.2 7635.5 0.30 0.30 1.00 ber of FE data proved to be higher than these curves. Therefore, it
180 16 9.4 110 12 8.3 6853.9 0.25 0.27 0.94 was reasonable to use them to predict HSSY column strengths
180 16 9.4 120 12 9.2 6973.9 0.28 0.27 1.05
from a statistical standpoint.
160 16 8.1 100 12 7.5 6093.9 0.18 0.19 0.99
140 14 8.1 85 10 7.7 4595.6 0.10 0.11 0.93 5) Design buckling formulae obtained on the basis of nonlinear regres-
140 14 8.1 90 10 8.2 4645.6 0.11 0.11 1.00 sion models in various codes were recommended to be incorporated
140 14 8.1 95 10 8.7 4695.6 0.12 0.11 1.07 into these codes for a new section design, because it will be too con-
140 14 8.1 90 9 9.3 4555.6 0.10 0.11 0.94 servative to design the 420 MPa large-section HSSY column as a NSS
125 12 8.6 80 8 9.3 3510.6 0.06 0.07 0.94
column with T-section or equal-leg angles.
74 X. Yu et al. / Journal of Constructional Steel Research 129 (2017) 54–74

6) Section efficiency and section optimal selection were clarified. Under Acknowledgements
the condition of the same cross-sectional steel angles, a higher sec-
tion efficiency can be attained with the appropriate increase of the This study was funded by the Extra High Voltage Power Transmis-
length of steel plate (so that the centroid of Y-section and its shear sion Company of China Southern Power Grid, which is gratefully ac-
center could be as close as possible). The cross section reached the knowledged. All authors would like to appreciate the Building
maximum utilization ratio when moments of inertia around axis-x Structure Laboratory of Tongji University for the test facilities and con-
and around axis-y were approximately equal in the event of the ditions provided. Anhui Hongyuan Tower Co., Ltd. is also acknowledged
width-to-thickness ratio within the code limits. for processing all test specimens and slots.
7) For eccentric compression members, the sensitivity of ultimate buck-
ling capacities to loading eccentricities was significant. No more than References
0.1Nu was lost when the loading eccentricity ratio β≤ 0.05, thus the
maximum connection eccentricity should be limited within 0.05ba [1] DL/T 5154-2012, Technical Code for the Design of Tower and Pole Structures of
Overhead Transmission Line, China Planning Press, Beijing, 2012.
in prevention of a substantial loss of buckling strength. [2] BS EN 50341-1:2012, Overhead Electrical Lines Exceeding AC 1 kV: Part 1: General
Requirements-common Specifications, BSI, London, 2013.
[3] ASCE 10-97, Design of Latticed Steel Transmission Structures, ASCE, Washington DC,
Major symbols 1997.
[4] BS EN 1993-1-1, Eurocode 3: Design of Steel Structures: Part 1–1: General Rules and
Rules for Buildings, BSI, London, 2005.
L measured length of a column [5] GB 50017-2003, Code for Design of Steel Structures, China Architecture & Building
L0 effective length of a column Press, Beijing, 2006.
[6] BS EN 1993-1-12, Eurocode 3: Design of Steel Structures: Part 1–12: Additional
A sectional area of a column Rules for the Extension of EN 1993 up to Steel Grades S700, BSI, London, 2007.
Aa sectional area of steel angles [7] BS 5950-1, British Standard: Structural Use of Steelwork in Building: Part 1: Code of
Ab sectional area of steel plate Practice for Design - Rolled and Welded Sections, BSI, London, 2000.
[8] ANSI/AISC 360-10, Specification for Structural Steel Buildings, AISC, Chicago, 2010.
ba width of equal-leg angles [9] AS4100-1998 (incorporating Amendment 1-2012), Steel Structures, Standards Aus-
bb width of steel plate tralia, Sydney, 2012.
ta thickness of equal-leg angles [10] CSA S16-09, Design of Steel Structures, CSA, Mississauga, 2009.
[11] AIJ, Recommendation for Limit State Design of Steel Structures, AIJ, Tokyo, 2010
tb thickness of steel plate
2010 (in Japanese).
umax, vmax maximum translation of Y-section [12] R. Bjorhovde, L. Tall, Maximum Column Strength and the Multiple Column Curve
Ix moment of inertia around the x-axis Concept, Fritz Laboratory Report, Lehigh University, 1971.
[13] R. Bjorhovde, Deterministic and Probabilistic Approaches to the Strength of Steel
a1, a2, a3 residual stress factor of steel angles
Columns, (Ph.D. dissertation), Lehigh University, 1972.
β loading eccentricity ratio [14] H.Y. Ban, G. Shi, Y.J. Shi, Y.Q. Wang, Column buckling tests of 420 MPa high strength
Iy moment of inertia around the y-axis steel single equal angles, Int. J. Struct. Stab. Dyn. 13 (2) (2013) 1250069-1–1250069-23.
θmax maximum rotation of Y-section [15] K. Cao, Y.J. Guo, D.W. Zeng, Buckling behavior of large-section and 420 MPa high-
strength angle steel columns, J. Constr. Steel Res. 111 (2015) 11–20.
θtop, θbottom rotation of top end or bottom end [16] F.S. Cardoso, K.J.R. Rasmussen, Behavior and design of concentrically loaded T-sec-
RD reading of displacement transducer tion steel columns, J. Struct. Eng. 140 (2014) 04014039-1–04014039-17.
E elastic modulus of steel [17] K.J.R. Rasmussen, G.J. Hancock, Tests of high strength steel columns, J. Constr. Steel
Res. 34 (1) (1995) 27–52.
Kσ buckling factor [18] H.Y. Ban, G. Shi, Y.J. Shi, Y.Q. Wang, Residual stress tests of high-strength steel equal
δ elongation ratio after fracture angles, J. Struct. Eng. 138 (2012) 1446–1454.
σarc, σbrc compressive residual stress [19] 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.
σart, σbrt tensile residual stress [20] Y.B. Wang, G.Q. Li, S.W. Chen, F.F. Sun, Experimental and numerical study on the be-
Nu buckling capacity or buckling strength havior of axially compressed high strength steel columns with H-section, Eng.
Nu,t column test buckling capacity Struct. 43 (2012) 149–159.
[21] Y.B. Wang, G.Q. Li, S.W. Chen, F.F. Sun, Experimental and numerical study on the be-
Ne first-order eigenvalue buckling load in FE model
havior of axially compressed high strength steel box-columns, Eng. Struct. 58
fy, Fy, ReL yield strength (2014) 79–91.
fu, Rm ultimate tensile strength [22] 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.
fya yield strength of equal-leg angles
Steel Res. 74 (2012) 140–150.
fyb yield strength of steel plate [23] H.Y. Ban, G. Shi, Y.J. Shi, M.A. Bradford, Experimental investigation of the overall
e0 deviation of centroid buckling behavior of 960 MPa high strength steel columns, J. Constr. Steel Res. 88
ɛu ultimate strain (2013) 256–266.
[24] G. Shi, X. Jiang, W.J. Zhou, T.M. Chan, Y. Zhang, Experimental study on column buck-
λn, λ non-dimensional slenderness ratio ling of 420 MPa high strength steel welded circular tubes, J. Constr. Steel Res. 100
λx slenderness ratio around the x-axis (2014) 71–81.
λy slenderness ratio around the y-axis [25] Y.B. Wang, G.Q. Li, S.W. Chen, The assessment of residual stresses in welded high
strength steel box sections, J. Constr. Steel Res. 76 (2012) 93–99.
λmax the larger value of λx and λy [26] Y.B. Wang, G.Q. Li, S.W. Chen, Residual stresses in welded flame-cut high strength
α, α1, α2, α3 imperfection factor steel H-sections, J. Constr. Steel Res. 79 (2012) 159–165.
b1, b2, b3 residual stress factor of steel plates [27] T.J. Li, G.Q. Li, Y.B. Wang, Residual stress tests of welded Q690 high-strength steel
box- and H-sections, J. Constr. Steel Res. 115 (2015) 283–289.
eL loading eccentricity [28] Y. Huang, B. Young, The art of coupon tests, J. Constr. Steel Res. 96 (2014) 159–175.
ɛ0 equivalent imperfection [29] R. Adman, M. Saidani, Elastic buckling of columns with end restraint effects, J.
γ section efficiency Constr. Steel Res. 87 (2013) 1–5.
[30] K.J.R. Rasmussen, N.S. Trahair, Exact and approximate solutions for the flexural
γt, γFEA section efficiency from test and FEA buckling of columns with oblique rotational end restraints, Thin-Walled Struct. 43
γR, γM1 resistance partial factor (2005) 411–426.
ψ stress ratio [31] BS EN 1993-1-5, Eurocode 3: Design of Steel Structures: Part 1–5: Plated structural
Elements, BSI, London, 2006.
rx, ry radius of gyration
[32] GB/T 2975-1998, Steel and Steel Products-Location and Preparation of Test Pieces
χ, φ, αc reduction/buckling/stability factor for Mechanical Testing, China Standards Press, Beijing, 1998 (in Chinese).
φt buckling factor of column test [33] GB/T 228-2002, Metallic materials: tensile testing at ambient temperature, China
φFEA FEA buckling factor Standards Press, Beijing, 2002 (in Chinese).
[34] Yu XL, Deng HZ, Zhang DH, Cui L. Buckling behavior of 420 MPa high strength steel
φcode code buckling factor Y-section columns: test investigation and design approach. Thin-Walled Struct.
Nu,FEA FEA buckling capacity (Under review)