Engineering Structures 157 (2018) 179–191

Contents lists available at ScienceDirect

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

The continuous strength method for the design of hot-rolled steel cross- T
sections
⁎
X. Yuna, L. Gardnera, , N. Boissonnadeb
a
Department of Civil and Environmental Engineering, Imperial College London, London, UK
b
Department of Civil and Water Engineering, Laval University, Quebec, Canada

A R T I C L E I N F O A B S T R A C T

Keywords: The continuous strength method (CSM) is a deformation-based structural design approach that enables material
Continuous strength method strain hardening properties to be exploited, thus resulting in more accurate and consistent capacity predictions.
Finite element modelling To date, the CSM has featured an elastic, linear hardening material model and has been applied to cold-formed
Hot-rolled carbon steel steel, stainless steel and aluminium. However, owing to the existence of a yield plateau in its stress-strain re-
Reliability analysis
sponse, this model is not well suited to hot-rolled carbon steel. Thus, a tri-linear material model, which can
Strain hardening
closely represent the stress-strain response of hot-rolled carbon steel, is introduced and incorporated into the
Yield plateau
CSM design framework. Maintaining the basic design philosophy of the existing CSM, new cross-section re-
sistance expressions are derived for a range of hot-rolled steel structural section types subjected to compression
and bending. The design provisions of EN 1993-1-1 and the proposed CSM are compared with experimental
results collected from the literature and numerical simulations performed in this paper. Overall, the CSM is
found to offer more accurate and consistent predictions than the current design provisions of EN 1993-1-1.
Finally, statistical analyses are carried out to assess the reliability level of the two different design methods
according to EN 1990 (2002).

1. Introduction resistance to the plastic moment capacity Mpl for Class 1 or Class 2
cross-sections and the elastic moment capacity Mel for Class 3 cross-
The section capacities of hot-rolled structural steel members are sections, which results in a step (or recently proposed linear or para-
traditionally determined following the process of cross-section classi- bolic transition [2–4]) from Mpl to Mel at a particular slenderness limit.
fication, which is based on the assumption of elastic, perfectly plastic Experimental results have shown that the current design rules in EN
material behaviour. Cross-section classification is a key feature of many 1993-1-1 [1] are often conservative in estimating the resistance of
modern steel design codes, such as EN 1993-1-1 [1], and determines the stocky hot-rolled steel cross-sections in both compression and bending
extent to which the resistance and deformation capacities of a cross- due primarily to the omission of strain hardening [5–7], and the failure
section are limited by the effects of local buckling. The classification of to account accurately for the spread of plasticity in Class 3 cross-sec-
a cross-section is assessed by comparing the width-to-thickness ratios of tions [8,9].
its constituent plate elements to corresponding slenderness limits, The importance of plate element interaction effects on the ultimate
which take account of the edge support conditions and applied loading. resistance of structural cross-sections has been examined by a number
The plate elements are typically treated individually, thus neglecting of researchers. Studies have been conducted on square and rectangular
the interaction effects between adjacent elements, i.e. the ability of the hollow sections (SHS/RHS) [10–13] and I-sections [13–18], while ex-
less slender elements to provide some assistance in resisting local plicit allowance for element interaction through the use of cross-section
buckling to the more slender elements, and the classification of the rather than element elastic buckling stresses in the determination of
most slender element defines that of the overall cross-section. Fur- local and distortional slendernesses is a key feature of the Direct
thermore, the maximum stress in the cross-section is generally limited Strength Method (DSM) [19–21]. The beneficial influence of strain
to the material yield stress fy, neglecting the beneficial effects of strain hardening has also been observed by a number of researchers
hardening. Hence, for non-slender cross-sections the compression re- [7,22–24], and the exploitation of strain hardening is a key feature of
sistance is limited to the yield load Ny and the cross-section bending the Continuous Strength Method [25,26].

⁎
Corresponding author.
E-mail address: [email protected] (L. Gardner).

https://doi.org/10.1016/j.engstruct.2017.12.009
Received 12 April 2017; Received in revised form 29 October 2017; Accepted 5 December 2017
Available online 22 December 2017
0141-0296/ © 2017 Elsevier Ltd. All rights reserved.
X. Yun et al. Engineering Structures 157 (2018) 179–191

The CSM is a deformation-based design approach that offers more behaviour. In cases where the full stress-strain curves were not reported
consistent and continuous resistance predictions than traditional ap- in the literature, the nonlinear material model for hot-rolled carbon
proaches based on cross-section classification, and allows system- steels proposed in [35] was used. Since the analyses may involve large
atically for the beneficial influence of strain hardening. To date, this inelastic strains, the engineering (nominal) stress-strain curves were
method has been established for stainless steel [27–29], structural converted to true stress-logarithmic plastic strain curves as required in
carbon steel [25,30–32] and aluminium alloy design [33,34] utilizing a ABAQUS for the chosen element type.
simple elastic, linear hardening material model, with a strain hardening Initial geometric imperfections were incorporated into the FE
modulus Esh varying with material grade. This model has been shown to models using the mode shapes obtained from a prior elastic buckling
offer a suitably close representation of the stress-strain responses of analysis. The lowest buckling mode under the considered loading
metallic materials with nonlinear rounded stress-strain behaviour, such conditions with an odd number buckling half-waves was used as the
as stainless steel, cold-formed carbon steel and aluminium alloys, to imperfection shape, which generally represents the most unfavourable
enable efficient and accurate designs. However, due to the existence of imperfection pattern [43], as shown in Fig. 2. Four different im-
a yield plateau, this CSM bi-linear material model is less suitable for perfection amplitudes were examined: a/400, a/200 and a/100, where
hot-rolled carbon steels. Thus, a revised CSM material model has been a is the flat width of the most slender constituent plate element in the
proposed for hot-rolled carbon steels [35] that exhibits a yield point, cross-section (i.e. that with the highest value of fy / σcr under the ap-
followed by a yield plateau and a strain hardening region. In this paper, plied loading conditions), and a value obtained from the modified
the application of the CSM to hot-rolled steel structural elements, i.e. Dawson and Walker predictive model [5,44], as given by Eq. (1), where
SHS/RHS and I-sections, focusing primarily on cross-sections in com- ω0 is the magnitude of the local imperfection, t is the plate thickness, fy
pression and bending, including the development of the material model is the material yield stress, and σcr is the elastic buckling stress of the
and the corresponding resistance functions, is outlined. The accuracy of most slender cross-section plate element assuming simply supported
the current design methods in EN 1993-1-1 and the CSM are then as- conditions between adjacent plates, taking due account of the stress
sessed based on the results of experiments collected from the literature distribution through the buckling factor kσ [45].
and numerical simulations performed herein.
fy
ω0 = 0.064 ⎛⎜ ⎟⎞ t
2. Numerical modelling ⎝ σcr ⎠ (1)

Numerical analyses were carried out using the finite element pro- The residual stresses introduced into hot-rolled steel sections are
gramme ABAQUS, version 6.13 [36] to simulate the cross-sectional primarily associated with non-uniform cooling rates during the fabri-
response of hot-rolled steel SHS/RHS and I-sections in both compres- cation process, with the more rapidly cooling regions of the sections
sion and bending. The numerical models were initially validated being left in residual compression and the slower cooling regions in
against existing experimental results on stub columns [5,37–39] and residual tension. The residual stress patterns recommended by the
simply-supported beams [7,40], and were subsequently used for para- European Convention for Constructional Steelwork (ECCS) [46], as
metric studies to expand the numerical data over a wider range of cross- shown in Fig. 3, where compressive residual stresses are designated as
section slendernesses, section geometries and loading conditions. Note positive and tensile residual stresses as negative, were applied to the FE
that three grade S460 beam tests reported in [40] were discontinued models for the I-sections. The magnitude of the initial stresses depends
before the maximum bending moment was reached and the results are on whether the height-to-width ratio of the cross-section is less than or
therefore excluded from the validation study and subsequent assess- equal to 1.2 or greater than 1.2 and is independent of the yield stress,
ment. Details of the finite element (FE) modelling approach and the with the nominal stress fy∗ = 235 MPa taken as the reference value. For
parametric studies performed in the current study are presented in the SHS and RHS, a typical pattern of residual stresses, based on DIN re-
following subsections. commendations [47], is shown in Fig. 4(a). To simplify the distribution
for ease of application in FE modelling, a modified residual stress pat-
2.1. Description of the FE models tern was proposed by Nseir [37] based on the analysis of measured
residual stress magnitudes and distributions, as represented in Fig. 4(b).
The reduced integration four-noded doubly curved shell element A constant value of 0.5fy∗ was assumed for the compressive residual
S4R was employed in all FE models; this element type has been suc- stresses in the corner regions while the amplitudes of the tensile re-
cessfully utilised by others in similar applications [10,40,41]. Mesh sidual stresses in the flat regions of the web and flanges were de-
sensitivity studies were conducted to determine an appropriate mesh termined to achieve self-equilibrium. The simplified residual stress
density to provide accurate results while minimizing computational pattern was employed herein for both the validation of the models and
time. Hence, an average element size equal to one twentieth of the the parametric analyses.
largest plate width that makes up the cross-section was adopted for the The boundary conditions were carefully selected to simulate the
flat parts of the modelled cross-sections, while four elements were experimental setups. For the stub column FE models, each end section
employed to model the curved geometry of the corner regions for the was connected to a concentric reference point through rigid body
SHS/RHS and two elements for each fillet zone of the I-sections. For the constraints such that the degrees of freedom of all nodes at each end
I-sections, particular attention was given to ensure that the properties of were constrained to the degrees of freedom of its corresponding re-
the fillet zones could be accurately represented. The nodes at each end ference point. All degrees of freedom were then restrained at each end
of the web were shifted by a distance of half the flange thickness to apart from the longitudinal translation at the loaded end to allow for
avoid overlapping of the elements at the web-to-flange junction [42], vertical displacement. For the FE models of the simply-supported
and these nodes were tied to their corresponding nodes at the mid- beams, the reference points were positioned on the lower flange of the
thickness of the flanges using the “General multi-point constraints section and connected to the nodes within the corresponding end plate
(∗MPC)”; this ensured that the translational and rotational degrees of region, with boundary conditions applied at each reference point to
freedom were equal for this pair of nodes. The additional area in the allow appropriate movement and rotation to simulate simple support
fillet zones was allowed for by increasing the thickness of the adjacent conditions. The point loads were applied to the lower part of the web at
web elements (see Fig. 1). the web-to-flange junction to prevent web crippling under concentrated
During the validation of the models, measured cross-section di- loading. Also, for modelling convenience, beam models under pure
mensions and material properties from the existing tests were in- bending moment were developed and validated against 4-point bending
corporated into the FE models to replicate the observed experimental tests and then used in the subsequent parametric studies. The length of

180
X. Yun et al. Engineering Structures 157 (2018) 179–191

Nodes of flange Nodes of flange Fig. 1. Representation of the fillet zones of I-sec-
tions in FE models.

Fillet zone Increased web thickness

Nodes of web Nodes of web

Fig. 2. Typical elastic buckling mode shapes adopted

as initial local imperfections in FE models.

(a) SHS/RHS stub column. (b) I-section stub column.

(c) SHS/RHS in 4-point bending.

(d) I-section in 4-point bending.

the pure bending models was equal to the length of the constant mo- that, for both compression and bending resistances, the most consistent
ment region in the corresponding 4-point bending tests, and similar predictions were obtained using the local imperfection amplitude of a/
boundary conditions as those used for the stub column models were 200, which is also in accordance with the recommendations of EN
applied to the pure bending models, the only difference being that the 1993-1-5 [45]. Note however that the results of the FE simulations were
reference points were allowed to rotate in the direction of bending, as relatively insensitive to the magnitude of the imperfections, especially
shown in Fig. 5. The modified Riks method was employed to perform for the more stocky cross-sections. Generally, the ultimate capacities
the geometrically and materially nonlinear analyses of the FE models to predicted by the FE models are slightly lower than those obtained from
capture the full load-deformation response of the modelled specimens, the tests, indicating conservative predictions of the numerical results.
including the post-ultimate path. The average ratios of FE-to-test ultimate loads are 0.96 and 0.94, with
corresponding coefficients of variation (COV) of 0.069 and 0.080, for
2.2. Validation of the FE models stub column and beam tests, respectively. The initial test stiffnesses
were also constantly predicted to within 2%. The pure bending models
The FE models were validated by comparing the numerically ob- were found to provide accurate and consistent predictions compared
tained ultimate capacities, load-deformation responses and the de- with the corresponding tests and 4-point bending models and were thus
formed shapes against those reported in a series of experiments. The considered to be appropriate to perform parametric studies for bending.
normalized results obtained from the FE models for varying imperfec- The load-deformation curves obtained from the FE models with the
tion magnitudes for stub columns [5,37–39] and beams [7,40] are local imperfection amplitude of a/200 are compared against the cor-
presented in Tables 1 and 2, respectively. It was generally observed responding tests results for typical stub columns and beams in Figs. 6–9,

181
X. Yun et al. Engineering Structures 157 (2018) 179–191

0.5fy* 0.5fy* 0.5fy*

0.2fy*
0.5fy*
0.5fy* 0.5fy*
0.5fy* ri

ri tf
t 0.2fy*

H
0.5fy*
H

tw
0.5fy*
B

B 0.5fy* (a) Residual stress pattern presented in DIN [47].

(a) H/B . 0.5fy* 0.5fy*

*
f=2fy (ri+t/2)sin( /8)/(B-2ri-2t)
0.3fy* f
0.5fy*

ri
0.3fy* 0.3fy*
0.3fy*
t w
*
w=2fy (ri+t/2)sin( /8)/(H-2ri-2t)
H

tf
ri

0.3fy*
H

0.5fy*
tw B

(b) Simplified residual stress pattern.

Fig. 4. Residual stress patterns applied to FE models of hot-finished SHS/RHS.

B 0.3fy*
The parametric study included a range of SHS and RHS, with the
cross-section aspect ratios of the latter ranging from 1.3 to 2.0. The
(b) H/B > 1.2.
height of the SHS was varied from 40 to 140 mm, whereas for the RHS,
Fig. 3. Residual stress patterns applied to FE models of hot-finished I-sections. the height ranged from 80 to 160 mm and the width ranged from 60 to
80 mm. For both types of cross-section (SHS and RHS), the thickness
where δ is the end shortening of the stub columns, θ is the end rotation was varied between 1.3 and 5 mm and internal corner radii were taken
of the beams and θpl is the elastic portion of the total rotation corre- equal to the thickness of the section. For I-sections, the height, width
sponding to the plastic moment Mpl. Overall, initial stiffnesses, ultimate and corner radii were kept constant at 300 mm, 200 mm and 15 mm,
capacities and general shape of the load-deformation histories from the respectively. The flange thickness was varied from 8 to 28 mm and the
FE models matched closely with those obtained from the tests, and the web thickness from 3 to 28 mm. The modelled specimens were chosen
failure modes of the FE models mirrored those observed in the physical to cover the full range of local slenderness from very compact to slender
experiments. It is therefore concluded that the FE models are able to cross-sections. The obtained results are discussed in detail in the fol-
produce accurate and consistent predictions of the test response and are lowing section.
suitable for performing parametric studies.
3. Extension of the CSM to hot-rolled steel cross-sections
2.3. Parametric study
3.1. General concept
Based on the developed FE models, an extensive parametric study
was carried out to expand the available results over a wider range of The CSM is a deformation (strain) based design method with two
geometries and cross-section slenderness in order to investigate the main components: (1) a base curve defining the maximum strain εcsm
applicability and accuracy of the codified provisions given in EN 1993- that a cross-section can tolerate as a function of the cross-section
1-1 [1] and the revised CSM for hot-rolled steel cross-sections. The slenderness λ p and (2) a material model that allows for the influence of
parametric study included stub column models and pure major and strain hardening. The CSM base curve relates the cross-section de-
minor axis bending models following the basic modelling assumptions formation capacity, which is defined in a normalised form as εcsm di-
described previously. The length of all FE models was set equal to three vided by the yield strain εy, to the cross-section slenderness λ p . Within
times the average cross-section dimensions. The average measured the CSM, the cross-section slenderness λ p is defined in non-dimensional
material properties from the tensile coupon tests performed in [38] on form as the square root of the ratio of the yield stress fy to the elastic
grade S355 steel were adopted in the parametric study. buckling stress σcr, as given by Eq. (2). The elastic buckling stress σcr

182
X. Yun et al. Engineering Structures 157 (2018) 179–191

Fig. 5. Boundary conditions in the 4-point bending

and pure bending FE models.

UY = UZ = 0
RX = RY = 0
Reference point
P
P P
U = UY = UZ = 0 P
Reference point X
RX = RY = 0
4-point bending model

M
M
Reference point
Reference point UY = UZ = 0
UX = UY = UZ = 0 RX = RY = 0
RX = RY = 0
Pure bending model

should be determined for the full cross-section either using numerical sections in [31]. This was confirmed in [49], where the relationship
methods, such as the finite strip software CUFSM [48], or approximate between local slenderness and rotation capacity was also studied. Two
analytical methods [13], both of which take into account the effects of upper bounds have been placed on the predicted cross-section de-
plate element interaction. Conservatively, the elastic buckling stress of formation capacity εcsm/εy; the first limit of 15 corresponds to the
the most slender individual plate element can be used to define the material ductility requirement expressed in EN 1993-1-1 [1] and pre-
slenderness of the cross-section. In the present study, CUFSM was vents excessive deformations and the second limit of C1εu/εy, where εu
adopted for the determination of σcr for the full cross-section. is the ultimate strain of the material in tension and C1 is a coefficient
corresponding to the adopted CSM material model as described in the
λp = fy / σcr (2) subsequent section, defines a ‘cut-off’ strain to prevent over-predictions
of material strength. Note that the CSM is focussed primarily on non-
The CSM design base curve (Eq. (3)) and the transition point
slender sections where local buckling occurs after yielding, though
(λ p = 0.68) between non-slender and slender sections used for stainless
extension of the method to cover slender sections has recently been
steel sections were shown to be applicable to hot-rolled carbon steel

Table 1
Comparison of stub column test results with FE results for varying imperfection amplitudes.

Reference Steel grade Specimen Cross-section Nu,FE/Nu,test

Imperfection amplitude

a/400 a/200 a/100 D-W model

Gardner et al. [5] S355 SHS 100 × 100 × 4-HR1 SHS 1.07 1.06 1.06 1.08
S355 SHS 100 × 100 × 4-HR2 SHS 1.05 1.04 0.99 1.08
S355 SHS 60 × 60 × 3-HR1 SHS 0.95 0.94 0.93 0.97
S355 SHS 60 × 60 × 3-HR2 SHS 0.93 0.92 0.91 0.95
S355 RHS 60 × 40 × 4-HR1 RHS 0.97 0.95 0.93 1.04
S355 RHS 60 × 40 × 4-HR2 RHS 0.97 0.94 0.92 1.03
S355 SHS 40 × 40 × 4-HR1 SHS 0.94 0.89 0.82 0.98
S355 SHS 40 × 40 × 4-HR2 SHS 0.93 0.89 0.82 0.97
S355 SHS 40 × 40 × 3-HR1 SHS 0.87 0.85 0.84 0.93
S355 SHS 40 × 40 × 3-HR2 SHS 0.89 0.87 0.85 0.94

Nseir [37] S355 LC1_HF_250 × 150 × 5 RHS 1.05 1.01 0.95 1.11
S355 LC1_HF_200 × 100 × 5 RHS 1.06 1.04 1.01 1.06
S355 LC1_HF_200 × 200 × 5 SHS 1.04 0.98 0.89 1.12
S355 LC1_HF_200 × 200 × 6.3 SHS 1.04 1.04 1.03 1.04
S355 Stub_HF_250 × 150 × 5 RHS 1.08 1.04 0.99 1.16
S355 Stub_HF_200 × 100 × 5 RHS 1.06 1.02 0.97 1.07
S355 Stub_HF_200 × 200 × 5 SHS 1.09 1.04 0.98 1.11
S355 Stub_HF_200 × 200 × 6.3 SHS 0.97 0.94 0.86 1.02

Foster et al. [39] – 305 × 127 × 48-SC1 I-section 0.97 0.96 0.93 0.98
– 305 × 127 × 48-SC2 I-section 0.99 0.98 0.95 1.00
– 305 × 165 × 40-SC1 I-section 1.02 1.01 0.98 1.06
– 305 × 165 × 40-SC2 I-section 0.99 0.97 0.94 1.02

Yun et al. [38] S235 P7 HEB160 I-section 0.89 0.87 0.85 0.94
S355 C7 HEB160 I-section 0.90 0.87 0.86 0.91

Mean 0.99 0.96 0.93 1.02

COV 0.068 0.069 0.073 0.066

183
X. Yun et al. Engineering Structures 157 (2018) 179–191

Table 2
Comparison of bending test results with FE results for varying imperfection amplitudes.

Reference Steel grade Specimen Cross-section Bending configuration Mu,FE/Mu,test

Imperfection amplitude

a/400 a/200 a/100 D-W model a/200 (Pure bending model)

Byfield and Nethercot [7] FE430A Y1 203 × 102 × 23UB I-section 4-point 0.91 0.88 0.84 0.92 0.89
FE430A Y2 203 × 102 × 23UB I-section 4-point 0.91 0.88 0.85 0.92 0.90
FE430A Y3 203 × 102 × 23UB I-section 4-point 0.95 0.93 0.90 0.96 0.95
FE430A Y4 203 × 102 × 23UB I-section 4-point 0.90 0.87 0.84 0.92 0.89
FE430A Y5 203 × 102 × 23UB I-section 4-point 0.94 0.90 0.87 0.95 0.92
FE430A Y6 203 × 102 × 23UB I-section 4-point 0.95 0.92 0.89 0.96 0.94

Wang et al. [40] S460 SHS 50 × 50 × 5 SHS 3-point 1.00 1.00 1.00 1.00 –
S460 SHS 50 × 50 × 4 SHS 3-point 1.06 1.06 1.06 1.06 –
S460 SHS 100 × 100 × 5 SHS 3-point 0.98 0.98 0.98 0.98 –
S460 SHS 90 × 90 × 3.6 SHS 3-point 1.01 1.01 1.01 1.01 –
S460 RHS 100 × 50 × 4.5 RHS 3-point 1.11 1.11 1.11 1.11 –
S460 SHS 100 × 100 × 5 SHS 4-point 0.88 0.88 0.86 0.88 0.88
S460 SHS 90 × 90 × 3.6 SHS 4-point 0.92 0.91 0.88 0.92 0.92
S460 RHS 100 × 50 × 6.3 RHS 4-point 0.87 0.87 0.87 0.87 0.87
S460 RHS 100 × 50 × 4.5 RHS 4-point 0.91 0.90 0.90 0.91 0.91

0.95 0.94 0.92 0.96 0.91

0.072 0.080 0.092 0.068 0.028

3000 90

80
2500
70

2000 60
M (kNm)

50
N (kN)

1500
40

1000 30

20
500 Test Test
10
FE FE
0 0
0 5 10 15 20 25 0 1 2 3 4 5 6 7 8 9

(mm) (degree)
Fig. 8. Comparison of experimental and numerical moment-end rotation curves for Y6
Fig. 6. Comparison of experimental and numerical load-end shortening curves for C7
203 × 102 × 23UB [7].
HEB160 [38].

1.2
2500

1.0
2000
0.8

1500
M/Mpl
N (kN)

0.6

1000 0.4

0.2
500 Test
Test FE
FE 0.0
0 0 1 2 3 4 5 6 7
0 1 2 3 4 5 6 7 8 / pl
(mm)
Fig. 9. Comparison of experimental and numerical normalized moment-rotation curves
Fig. 7. Comparison of experimental and numerical load-end shortening curves for for SHS 100 × 100 × 5 [40].
Stub_HF_200 × 200 × 5 [37].

presented [50,51]. Note also that the maximum strain limit of 15εy
εcsm
⎧ 0.25 , but εcsm ⩽ min 15, C1 εu
⎪ λp3.6 εy εy ( ) for λ p ⩽ 0.68
could be set by the designer on a project by project basis, depending on =
εy ⎨ ⎛ 0.222 ⎞ 1
the accepted degree of plasticity at the ultimate limit state. ⎪ 1− λ 1.05 λ 1.05
⎜ ⎟ for λ p > 0.68
⎩⎝ p ⎠ p (3)

Owing to the existence of a yield plateau in hot-rolled steels, the

184
X. Yun et al. Engineering Structures 157 (2018) 179–191

fy
εu = 0.6 ⎜⎛1− ⎟⎞, but εu ⩾ 0.06 for hot-rolled steels
⎝ fu⎠ (6)

fy
εsh = 0.1 −0.055, but 0.015 ⩽ εsh ⩽ 0.03
fu (7)

εsh + 0.25(εu−εsh )
C1 =
εu (8)

εsh + 0.4(εu−εsh )
C2 =
εu (9)

Fig. 10. Typical measured stress-strain curve for hot-rolled carbon steel and proposed 3.3. CSM resistance functions for hot-rolled steels
quad-linear material model.
The CSM analytical and simplified design resistance expressions for
elastic-linear hardening material model employed to date in the CSM hot-rolled steel cross-sections under compression Ncsm and bending
for cold-formed steel, stainless steel and aluminium is less appropriate. Mcsm are derived in this section, based on the material model described
The process of extending the CSM to cover the design of hot-rolled steel in Section 3.2. Note that only the first three stages of the quad-linear
cross-sections requires the selection of a suitable material model and material model are used in the CSM, since progression into the fourth
the derivation of simple yet accurate resistance expressions based on stage would correspond to very high strains and lead to overly com-
the adopted material model; these two aspects are addressed in the plicated resistance functions, particularly in bending.
following sub-sections.
3.3.1. Compression resistance
3.2. Material model The design CSM compression resistance for a hot-rolled steel cross-
section Ncsm,Rd is calculated as the product of the gross cross-section
Unlike stainless steel, aluminium and cold-formed carbon steel area A and the CSM limiting stress fcsm, divided by the partial factor γM0
which exhibit predominantly nonlinear rounded stress-strain curves, with a recommended value of unity for hot-rolled steel, as given by Eq.
hot-rolled carbon steels have an elastic response, with a distinctively (10), in which the fcsm may be calculated from Eq. (11) based on the
defined yield point, followed by a yield plateau and a moderate degree first three stages of the quad-linear material model.
of strain hardening. A material model to reflect this response is required Afcsm
for the development and application of the CSM to hot-rolled carbon Ncsm,Rd =
γM0 (10)
steel cross-sections. The first three stages of the quad-linear material
model proposed by Yun and Gardner [35] for hot-rolled carbon steels, for εcsm ⩽ ε y
⎧ Eεcsm
as illustrated in Fig. 10 and described in Eq. (4), is adopted herein as the ⎪f for ε y < εcsm ⩽ εsh
fcsm = y
CSM material model that takes account of both the yield plateau and ⎨
f
⎪ y + E (ε
sh csm−ε sh ) for εsh < εcsm ⩽ C1 εu
the strain hardening of hot-rolled steels. ⎩ (11)

⎧ Eε for ε ⩽ ε y The post-yield strength benefit due to strain hardening can be

⎪ fy for ε y < ε ⩽ εsh achieved once the CSM limiting stress εcsm obtained from the base curve
⎪
f (ε ) = fy + Esh (ε−εsh ) for εsh < ε ⩽ C1 εu Eq. (3) exceeds the strain hardening strain εsh; thereafter the corre-
⎨ sponding CSM limiting stress fcsm will be larger than the yield stress fy.
⎪ fu − fC1 εu
⎪ fC1 εu + εu − C1 εu (ε−C1 εu) for C1 εu < ε ⩽ εu
⎩ (4)
3.3.2. Bending resistance
In Fig. 10, E is the Young’s modulus, fy and fu are the yield and For slender cross-sections, the design CSM bending resistance is
ultimate tensile stress, εy and εu are the strains at the yield and ultimate given simply by the product of the CSM limiting stress and the elastic
stresses, respectively, εsh is the strain hardening strain where the yield section modulus, as given by Eq. (12).
plateau ends and subsequently the strain hardening initiates, C1εu re- Wel fy
Wel ε
presents the strain at the intersection point of the third stage of the Mcsm,Rd = fcsm = csm for εcsm ⩽ ε y
γM0 ε y γM0 (12)
model and the actual stress-strain curve, and fC1 εu is the corresponding
stress. Two material coefficients, C1 and C2, are used in the material The analytical CSM bending resistance expression for non-slender
model; C1 represents the interaction point discussed previously and hot-rolled steel I-sections and SHS/RHS is derived herein based on the
effectively defines a ‘cut-off’ strain to avoid over-predictions of material assumption that plane sections remain plane and normal to neutral axis
strength and is also included in the base curve (Eq. (3)); C2 is used in in bending, and that the cross-section shape does not significantly de-
Eq. (5) to define the strain hardening slope Esh. Note that the quad- form before the strain at the extreme outer-fibre εcsm is attained. The
linear material model adopts the same definitions for the material linear strain and tri-linear stress distributions (arising from the CSM
coefficients (C1 and C2) as used in the previous CSM elastic, linear material model for hot-rolled carbon steels described in Section 3.2) for
hardening material model [29,31], and can indeed be seen as a superset half of a symmetric cross-section are shown in Fig. 11. The bending
of the previous model, with the principal difference being that the capacity at a strain ratio (εcsm/εy) of unity is equal to the elastic mo-
strain hardening strain εsh is equal to the yield strain εy in the CSM ment Mel. With an increase in deformation capacity (i.e. strain ratio
elastic, linear hardening model. εcsm/εy), the bending capacity is asymptotic to the plastic moment Mpl
fu −fy due to the spread of plasticity until the strain ratio reaches εsh/εy, after
Esh = which benefit from strain hardening can be exploited.
C2 εu−εsh (5)
The CSM cross-section resistance in bending Mcsm,Rd depends upon
The predictive expressions for the required parameters of the model, whether or not strain hardening is experienced (i.e. whether or not
εu, εsh, C1 and C2, are given by Eqs. (6)–(9), respectively. εcsm > εsh). If εcsm ≤ εsh, the bending resistance expressions for hot-

185
X. Yun et al. Engineering Structures 157 (2018) 179–191

csm fy fcsm-fy B B

z-z z-z
sh Ww*
H/2 (B/2)

y-y

hw
H

H
hw
Y*
y

y tw tw/2
Y

Ww tf
Neutral axis tf
Fig. 11. Strain and stress distributions for half of an I-section or SHS/RHS in bending.
Fig. 12. Cross-section geometry for the CSM analytical bending expressions (I-section and
rolled steel cross-sections are the same as those derived previously SHS/RHS).

using the bi-linear CSM material model [31] with Esh taken equal to
zero to represent the yield plateau; the design resistance expressions are moment Msh resulting from the material strain hardening properties,
given by Eqs. (13) and (14) for major and minor axis bending, re- which can be calculated by integration of the triangular shaped stress
spectively, block associated with the stress (fcsm − fy), for Y ∗ ⩽ |y| ⩽ H /2 (major
α axis) as shown in Fig. 11, where H is the height of the cross-section and
Wpl,y fy ⎡ ⎛ Wel,y ⎞
My,csm,Rd = 1−⎜1− ⎟
⎛ εcsm ⎞ ⎤, for ε < ε
⎜ ⎟ y csm ⩽ εsh
Y∗ is the distance from the neutral axis to the point of first strain
γM0 ⎢ ⎝ Wpl,y ⎠ ⎝ εy ⎠ ⎦
⎥
(13) hardening, as given by Eq. (18).
⎣
H
Wpl,z fy ⎡ ⎛ Wel,z ⎞ α Y∗ =
Mz,csm,Rd = 1−⎜1− ⎛ εcsm ⎞ ⎤, for ε < ε 2
εcsm
⎟ ⎜ ⎟ y csm ⩽ εsh (18)
γM0 ⎢ ⎝ Wpl,z ⎠ ⎥ εsh
⎣ ⎝ εy ⎠ ⎦ (14)
Considering an I-section in bending about the major axis, the deri-
where Wpl is the plastic section modulus, Wel is the elastic section ∗
vation process for W w,y is described below. The same approach can be
modulus, y and z refer to the major and minor axes, respectively, and α applied to a box section (SHS or RHS), simply by setting the wall
is a dimensionless coefficient that depends on the type of cross-section thickness as half of the web thickness of an I-section, as shown in
and axis of bending, as defined in Table 3. Fig. 12.
For the more stocky cross-sections, where εcsm > εsh, the outer fibre The strain hardening moment Msh,y is defined by Eq. (19), from
strain will enter into the strain hardening regime and the CSM bending which the introduced modulus W w,y ∗
can be determined using Eq. (20),
resistance can achieve Mcsm > Mpl. With reference to Fig. 11, the where the function g(y) represents the triangular stress distribution
bending capacity of a cross-section can now be expressed by Eq. (15), normalised by (fcsm − fy). The integral is evaluated from the first strain
∗
where the appropriate section moduli Wpl,y, Ww,y and W w,y or Wpl,z, hardening point |y| = Y ∗ to |y| = H /2 , and the associated area to in-
∗
Ww,z and W w,z are used for bending about the major and minor axes, tegrate over is AY ∗.
respectively. The Ww and W w∗ terms represent the section moduli as-
sociated with the two shaded triangular stress blocks shown in Fig. 11.
∗
Msh,y = W w,y (fcsm −fy ) = ∫A Y∗
fy dAY ∗ = ∫A Y∗
(fcsm −fy ) g (y ) y dAY ∗
(19)
For strain ratios greater than unity, the introduced modulus Ww can be
approximated by Eq. (16), the derivation of which is given in [52]. This
section of the present paper is, therefore, mainly focused on the deri-
∗
W w,y = ∫A Y∗
g (y ) y dAY ∗
(20)
vation of W w∗ . In Fig. 13(a), the first strain hardening point lies within the web i.e.
Mcsm = Wpl fy −Ww fy + W w∗ (fcsm −fy ) (15) Y∗ ≤ hw/2. This scenario may be alternatively expressed as:
εcsm H
α ≥
Ww = (Wpl−Wel ) ⎛ εcsm ⎞ εsh hw (21)
⎜ ⎟
⎝ εy ⎠ (16) ∗
In this case, the modulus W w,y has contributions from two parts: a
Substituting Eq. (16) and the CSM strength trapezoidal part acting within the flanges and a triangular part acting
fcsm = fy + Esh(εcsm − εsh) into Eq. (15) and normalizing the equation within the web. The modulus is calculated as:
by the plastic moment capacity Mpl = Wplfy = WplEεy gives Eq. (17). BH 2 ⎛ ε ε
∗
W w,y = 1− sh ⎞ ⎛2 + sh ⎞
⎜ ⎟⎜ ⎟
α 12 ⎝ εcsm ⎠ ⎝ εcsm ⎠
Mcsm W ⎛ εcsm ⎞ + W w∗
Esh ⎛ εcsm−εsh ⎞
= 1−⎛⎜1− el ⎞⎟ ⎜ ⎟ ⎜ ⎟ 2 −1
Mpl W ε W E ⎝ εy ⎠ (17) (B−tw ) H 2 ⎛ h w εsh ⎞ ⎛ ε ε h
⎝ pl ⎠ ⎝ y ⎠ pl
− −⎜ 1− sh ⎞ ⎛ sh + 2 w ⎞
⎟ ⎜ ⎟ ⎜ ⎟

12 ⎝ H εcsm ⎠ ⎝ εcsm ⎠ ⎝ εcsm H⎠ (22)

The final term W w∗ (fcsm −fy ) in Eq. (15) represents a strain hardening
∗
When Y lies outside the web, as shown in Fig. 13(b), the modulus
∗
Table 3 W w,y is calculated as:
CSM coefficients α and β for use in bending resistance functions.
BH 2 ⎛ ε ε
∗
W w,y = 1− sh ⎞ ⎛2 + sh ⎞
⎜ ⎟⎜ ⎟

Axis of bending α β 12 ⎝ εcsm ⎠ ⎝ εcsm ⎠ (23)

Major Minor Major Minor which applies when the strain ratio is bounded by:
I-section 2 1.2 0.1 0.05 εcsm H
1⩽ ⩽
SHS/RHS 2 2 0.1 0.1 εsh hw (24)

186
X. Yun et al. Engineering Structures 157 (2018) 179–191

B fy fcsm-fy ∗
forW w,z in which the first strain hardening point lies within the flange is
considered herein, and is given by:

z-z t f B2 ⎛ ε ε ε B
∗
W w,z = 1− sh ⎞ ⎛2 + sh ⎞, for csm ⩽
Ww,* y
⎜ ⎟⎜ ⎟

6 ⎝ εcsm ⎠ ⎝ εcsm ⎠ εsh tw (27)

Y*
∗
Substituting the expressions for the introduced modulus W w,z into
Eq. (17) leads to the exact analytical CSM bending equations, but these
are lengthy for practical design due to the W w∗ term. A simplified yet
hw

y-y
H

Neutral axis
accurate approximation is therefore sought for design purposes. The
simpler design expression for W w∗ is obtained based on a regression
tw analysis of the geometric properties of commercially available struc-
tural I-sections and SHS/RHS in [53] to determine the dimensionless
coefficient β to be used in Eq. (28); β is presented as a function of the
tf type of cross-section and axis of bending in Table 3.

(a) Y* within the web. ε −ε

W w∗ = β ⎜⎛ csm sh ⎟⎞ Wpl
⎝ εy ⎠ (28)

B fcsm-fy
The simpler design expression of W w∗ (Eq. (28)) can then be sub-
stituted into Eq. (17) to give the simplified Mcsm equations provided as
z-z fy Eq. (29) and Eq. (30) for bending about the major and minor axis, re-
Ww,* y spectively.
Y*

α 2
Wpl,y fy ⎡ ⎛ Wel,y ⎞ ⎛ εcsm ⎞ + β ⎛ εcsm−εsh ⎞ Esh ⎤,
My,csm,Rd = 1−⎜1−
γM0 ⎢ ⎝ Wpl,y ⎠ ⎥
⎟ ⎜ ⎟ ⎜ ⎟

⎣ ⎝ εy ⎠ ⎝ εy ⎠ E ⎦
H

hw

Neutral axis y-y for εcsm > εsh (29)

α 2
Wpl,z fy ⎡ ⎛ Wel,z ⎞ ⎛ εcsm ⎞ + β ⎛ εcsm−εsh ⎞ Esh ⎤,
tw Mz,csm,Rd = 1−⎜1−
γM0 ⎢ ⎝ Wpl,z ⎠ ⎥
⎟ ⎜ ⎟ ⎜ ⎟

⎣ ⎝ εy ⎠ ⎝ εy ⎠ E ⎦

tf for εcsm > εsh (30)

The predicted CSM bending resistances, using the exact analytical

(b) Y* within the flange.
expression (Eq. (15)) and the simplified design expression (Eq. (29)),
∗
Fig. 13. Geometry for the derivation of W w,y for I-sections. are normalised by the plastic bending moment Mpl and plotted against
the strain ratio for a typical SHS bending about one of its principal axes,
In summary, for box and I-sections bending about their major axis, as shown in Fig. 14. The design equation can be seen to tend towards
∗
the W w,y modulus is defined by Eq. (25). the analytical solution for the higher strain ratios, indicating a high
level of accuracy in approximation of the W w∗ term (Eq. (28)), while for
2 lower strain ratios where Ww term has a greater influence, the design
⎪ 12 ( εcsm )(
⎧ BH 1− εsh 2 + εsh
εcsm ) for
εcsm
εsh
⩾
H
hw expression (Eq. (16)) may be seen to provide a safe-sided approxima-
⎪ (B − tw ) H2 hw 2 εsh −1 εsh
∗
W w,y =
⎪ − 12 (H
ε
− ε sh
csm ) (1− ) (εcsm εcsm
tion to the analytical solution [52]. A close match to the exact analy-
tical solution could of course also be obtained from a numerical fibre
⎨
⎪ +2H
⎪ 2
hw
) element model, and indeed guidelines for the use of the CSM in such a
context are currently under development.
⎩ 12
(
⎪ BH 1− εsh
εcsm )(2 + ) εsh
εcsm
for 1 ⩽
εcsm
εsh
⩽
H
hw
1.10
(25)
1.05
CSM/plastic moment Mcsm/Mpl

For box sections bending about the minor axis, the expression for
∗ 1.00
W w,zis similar to Eq. (25), as given in Eq. (26).

2 0.95
⎪ 12 ( εcsm )(
⎧ HB 1− εsh 2 + εsh
εcsm ) for
εcsm
εsh
⩾
B
B − 2tw
0.90
⎪ hw B 2 B − 2tw εsh 2 −1
⎪ − 12
⎪
( B
−ε
csm )(
ε
1− ε sh
csm ) ( εsh
εcsm 0.85
∗
W w,z =
⎨ + 2
B − 2tw
B ) 0.80
⎪ HB2 Design Eq. (29)
⎪ 12 1− ε
⎪
( εsh
csm )( ε
2 + ε sh
csm ) for 1 ⩽
εcsm
εsh
⩽ 0.75
Analytical Eq. (15)
⎪ B
0.70
⎩ B − 2tw
1 3 5 7 9 11 13 15
(26) Strain ratio csm/ y

For I-sections bent about their minor axis, the first strain hardening Fig. 14. Comparison between exact analytical and simplified design CSM normalised
bending resistances Mcsm/Mpl with varying strain ratios εcsm/εy.
point is unlikely to lie within the web, hence only the expression

187
X. Yun et al. Engineering Structures 157 (2018) 179–191

1.5
Test
1.4 FE
CSM strain ratio limit to 15
1.3 CSM strain ratio limit to 20
EN 1993-1-1
1.2
Nu,FE(test)/Ny

1.1 csm/ y=20

csm/ y=15
1.0

0.9

0.8

0.7

0.6
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Cross-section slenderness p
Fig. 17. Comparison of CSM and EN 1993-1-1 resistance predictions with FE/test data for
Fig. 15. Comparison of CSM and EN 1993-1-1 resistance predictions with FE/test data for SHS/RHS in bending.
SHS/RHS in compression.

1.6
Test
4. Assessment of proposals and reliability analysis FE
1.5
CSM strain ratio limit to 15
1.4 CSM strain ratio limit to 20
Ultimate capacities from the numerical parametric studies, together EN 1993-1-1
with the experimental results summarised in Tables 1 and 2, on stub 1.3

Mu,FE(test) /Mpl
columns (Nu,FE(test)) and beams (Mu,FE(test)), respectively, have been 1.2
used to assess the accuracy of the resistance predictions of the CSM
(Nu,csm and Mu,csm). For comparison purposes, the codified design 1.1 csm/ y=20

csm/ y=15
provisions of EN 1993-1-1 [1] (Nu,EC3 and Mu,EC3) are also examined. 1.0
Note that the comparisons were performed using the measured geo- 0.9
metric and material properties of the tested and modelled cross-sec-
0.8 Class 2 limit based on EN 1993-1-1 ( p
tions, with all partial safety factors set to unity. Cross-sections with
0.15 ⩽ λ p ⩽ 0.68, which cover most of the commercially available non- 0.7 Class 2 limit utilising CUFSM ( p

slender hot-finished I-sections and SHS/RHS, were the focus of the 0.6
present study. The comparisons are presented in Figs. 15–19, in which 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
the experimentally and numerically obtained ultimate capacities have Cross-section slenderness p
been normalised by the yield load Ny and the plastic moment capacity
Fig. 18. Comparison of CSM and EN 1993-1-1 resistance predictions with FE/test data for
Mpl for cross-sections in compression and in bending, respectively, and I-sections under major axis bending.
are plotted against the cross-section slenderness λ p . Curves indicating
design capacities from EN 1993-1-1 [1] and the CSM have also been
1.6
plotted together with the experimental and numerical data in FE
Figs. 15–19. Since the CSM curve depends on the shape of the cross- CSM strain ratio limit to 15
CSM strain ratio limit to 20
sections and the material properties, a curve based on the average 1.4 CSM strain ratio limit to 25
geometric and material properties used in the parametric studies has EN 1993-1-1
been depicted for illustration purposes. The EN 1993-1-1 [1] resistance
1.2
Mu,FE(test)/Mpl

curves have been positioned based on the slenderness limits corre-

sponding to the Class 2 limiting width-to-thickness ratio given in csm/ y=25

csm/ y=20
1.0
csm/ y=15

Class 2 limit based on EN 1993-1-1 ( p

0.8
Class 2 limit utilising CUFSM ( p

0.6
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Cross-section slenderness p

Fig. 19. Comparison of CSM and EN 1993-1-1 resistance predictions with FE/test data for
I-sections under minor axis bending.

Table 5.2 of EN 1993-1-1, applied on an element-by-element basis [45]

i.e. for internal compression elements, the Class 2 width-to-thickness
ratio of 38ε (ε is the material factor ε = 235/ f y ) corresponds to
λ p ≈ 0.67 (see Fig. 17), and for outstand elements in compression, the
Class 2 width-to-thickness ratio of 10ε corresponds to λ p ≈ 0.54 (see
Figs. 18 and 19). The mean slenderness limits corresponding to the
Fig. 16. Comparison of CSM and EN 1993-1-1 resistance predictions with FE/test data for Class 2 width-to-thickness ratio in EN 1993-1-1 based on the elastic
I-sections in compression. buckling stress of the cross-sections considered in this study,

188
X. Yun et al. Engineering Structures 157 (2018) 179–191

Table 4 the I-sections and 0.089 and 0.078 for the SHS/RHS, respectively. The
Comparison of CSM and EN 1993-1-1 resistance predictions with FE/test results for cross- CSM predictions remain conservative for the stocky cross-sections due
sections in compression.
primarily to the imposed strain limit of εcsm = 15εy. This limit could be
Cross- λp ⩽ 0.35 All (λp ⩽ 0.68 )
adjusted if a designer wished to impose higher or lower strain limits on
section a given project. Ten of the collected stub column test specimens had
type No. of Nu,FE(test)/ Nu,FE(test)/ No. of Nu,FE(test)/ Nu,FE(test)/ slender cross-sections (i.e. λ p > 0.68). For these specimens, the CSM
FE Nu,EC3 Nu,csm FE Nu,EC3 Nu,csm yielded good mean capacity predictions (Nu,test/Nu,csm = 1.06), but
(tests) Mean Mean (tests) Mean Mean with higher scatter than those obtained from EC3. Further investigation
(COV) (COV) (COV) (COV)
is required on this topic.
I-sections 82 (0) 1.28 1.22 140 1.17 1.14
(0.055) (0.049) (4) (0.117) (0.098)
SHS/RHS 26 (6) 1.20 1.16 159 1.03 1.02 4.2. Assessment of CSM for hot-rolled steel cross-sections in bending
(0.089) (0.078) (10) (0.097) (0.080)

All 108 1.26 1.21 299 1.09 1.07 Comparisons between the test and FE capacities of cross-sections in
(6) (0.070) (0.062) (14) (0.127) (0.107)
bending and the CSM and EN 1993-1-1 [1] resistance predictions are
shown in Figs. 17–19, while the key results of the comparisons are re-
ported in Table 5. Based on the cross-section types and axis of bending,
determined using CUFSM [48], are also depicted in Figs. 17–19. In
the data have been classified into 3 groups: I-sections bent about their
Figs. 17 and 18, the slenderness limits determined on an element-by-
major axis, I-sections bent about their minor axis and SHS/RHS bent
element basis (i.e. based on EN 1993-1-1) are similar to the values
about either axis. For the more stocky sections with λ p ⩽ 0.35, the EN
determined based on buckling of the full cross-section (i.e. using σcr
1993-1-1 [1] predictions are unduly conservative, especially for I-sec-
from CUFSM, with some influence from element interaction). However,
tions, with mean values of Mu,FE(test)/Mu,EC3 equal to 1.29 and 1.30 for I-
for I-sections in bending about the minor axis, the web, which is at a
sections bent about their major and minor axis, respectively. The con-
low level of stress, and the outstand flanges on the tension side, which
servatism is less pronounced in the CSM predictions, except for I-sec-
experience stabilising tensile stresses, remain unbuckled and thus pro-
tions bent about their minor axis, where high strains are needed to
vide significant assistance to the flange plates in compression. In this
reach the plastic bending moment Mpl due to the high cross-section
instance, the boundary conditions on the unloaded edges of the com-
shape factor (i.e. the ratio of plastic-to-elastic section modulus Wpl/
pressed outstands are approaching fixed-free. However, in EN 1993-1-1,
Wel); here, the limited strain hardening is largely compensating for the
the outstand flanges of I-sections are treated, ignoring element inter-
more gradual plastification of the section to enable the attainment of
action, as having pinned-free boundary conditions, resulting in sig-
Mpl, rather than bringing resistance benefits beyond Mpl. This point
nificantly higher slenderness values when compared to the mean
highlights one of the additional benefits of the CSM, which is knowl-
CUFSM value, as highlighted in Fig. 19. The key results from the
edge of the strains required to attain a given capacity; this information
comparisons are provided in Table 4 for the cross-sections in com-
is not available in traditional strength-based design. Note that if the
pression and Table 5 for the cross-sections in bending.
maximum CSM strain ratio limit were to be relaxed to 20 or 25, further
exploitation of strain hardening would be possible and a closer match
4.1. Assessment of CSM for hot-rolled steel cross-sections in compression with the ultimate test and FE capacities is achieved, as indicated in
Figs. 15–19. Discrepancies between the CSM resistance predictions and
The compressive strengths predicted by EN 1993-1-1 [1] and the the test/FE data also stem from scatter in deformation capacity relative
CSM have been compared with experimental and numerical results for to the base curve and the safe-sided approximations introduced in the
both SHS/RHS and I-sections, as shown in Table 4 and Figs. 15 and 16. derivation of the simplified design expression for bending resistance.
The comparisons demonstrate that the very stocky hot-rolled steel For slenderness values λ p > 0.35, the CSM is shown to offer more
cross-sections, i.e. λ p ⩽ 0.35, have the potential to exceed their yield accurate ultimate capacity predictions and far fewer on the unsafe side
load substantially. It can be seen from Figs. 15 and 16 that the CSM towards the slender end of the Class 2 range by allowing systematically
provides the same predictions for cross-section resistance in the local for the increasing bending resistance with increasing deformation ca-
slenderness range 0.35 < λ p ⩽ 0.68 due to the existence of the yield pacity (i.e. strain ratio εcsm/εy) due to the spread of plasticity.
plateau in the adopted CSM material model, while for stockier cross- Specifically, for Class 3 cross-sections, the CSM yields significantly
sections (λ p ⩽ 0.35), where the predicted failure strain εcsm extends into more accurate than EN 1993-1-1 due to the allowance for the partial
the strain hardening region, the CSM yields improved resistance pre- plastification, with mean values of Mu,FE(test)/Mu,csm in this slenderness
dictions over EN 1993-1-1 [1]. As shown in Table 4, for the more stocky range of 1.03, 1.09 and 1.03 for I-sections in major axis bending, I-
sections with λ p ⩽ 0.35, the mean values of FE (or test)-to-predicted sections in minor axis bending and SHS/RHS in bending about either
ultimate loads for the EN 1993-1-1 [1] (Nu,FE(test)/Nu,EC3) and the CSM axis, respectively, all of which are lower (i.e. closer to unity and hence
(Nu,FE(test)/Nu,csm) are 1.28 and 1.22 for the I-sections and 1.20 and 1.16 more accurate) than the corresponding values of Mu,FE(test)/Mu,EC3.
for the SHS/RHS, with the corresponding COV of 0.055 and 0.049 for Moreover, the nature of the cross-section classification system in EN

Table 5
Comparison of CSM and EN 1993-1-1 resistance predictions with FE/test results for cross-sections in bending.

Cross-section-axis of λp ⩽ 0.35 Class 3 All ( λp ⩽ 0.68 )

bending
No. of FE Mu,FE(test)/ Mu,FE(test)/ No. of FE Mu,FE(test)/ Mu,FE(test)/ No. of FE Mu,FE(test)/ Mu,FE(test)/Mu,csm
(tests) Mu,EC3 Mu,csm (tests) Mu,EC3 Mu,csm (tests) Mu,EC3
Mean (COV) Mean (COV) Mean (COV) Mean (COV) Mean (COV) Mean (COV)

I-sections – major 180 (6) 1.29 (0.033) 1.26 (0.032) 27 (0) 1.10 (0.014) 1.03 (0.016) 265 (6) 1.22 (0.088) 1.20 (0.088)
I-sections – minor 98 (0) 1.30 (0.054) 1.30 (0.053) 41 (0) 1.48 (0.043) 1.09 (0.049) 147 (0) 1.34 (0.090) 1.23 (0.096)
SHS/RHS – both axes 82 (6) 1.15 (0.069) 1.13 (0.064) 35 (0) 1.07 (0.012) 1.03 (0.023) 309 (9) 1.03 (0.091) 1.03 (0.071)

All 360 (12) 1.26 (0.068) 1.24 (0.070) 103 (0) 1.24 (0.163) 1.05 (0.044) 721 (15) 1.16 (0.140) 1.13 (0.116)

189
X. Yun et al. Engineering Structures 157 (2018) 179–191

n
Table 6 1 re,i
Summary of reliability analysis results for cross-sections in compression. b=
n
∑ rt,i
i= 1 (32)
Design Cross- No. of tests and kd,n b Vδ Vr γM0
approach section simulations The results of the reliability analysis based on the EN 1993-1-1 [1]
predictions are also reported in Tables 6 and 7 for comparison purposes.
EN 1993-1- I-sections 144 3.159 1.171 0.119 0.132 1.10 For I-sections in compression and in bending, the obtained partial safety
1 SHS/RHS 169 3.149 1.025 0.088 0.106 1.16 factors γM0 were found to be less than or very close to unity, which is
CSM I-sections 144 3.159 1.141 0.099 0.115 1.07 the target value in EN 1993-1-1, for both EN 1993-1-1 and the CSM, as
SHS/RHS 169 3.149 1.017 0.074 0.094 1.13 shown in Tables 6 and 7. The CSM is therefore deemed to be reliable for
the design of hot-rolled steel I-sections. For SHS/RHS, the obtained
partial safety factors γM0 for the CSM are lower than the corresponding
1993-1-1 [1] results in a sharp step in the predicted resistance at a
values for EN 1993-1-1 [1], with CSM partial safety factors γM0 of 1.13
discrete slenderness limit; this step is more pronounced for the I-sec-
for compression and 1.10 for bending and EN 1993-1-1 partial safety
tions bent about their minor axis, as shown in Fig. 19, due to the higher
factors γM0 of 1.16 for both compression and for bending. Therefore,
ratio of Wpl/Wel, while this discontinuity is eliminated in the CSM ap-
although the CSM partial safety factors γM0 are slightly higher than the
proach.
target value of 1.0, the CSM design expressions are deemed to be ac-
ceptable on the basis of historical precedent.
4.3. Reliability analysis
5. Conclusions
In this section, the reliability level and required partial safety factor
γM0 for the proposed CSM resistance expressions for hot-rolled steel The continuous strength method (CSM) has been extended herein to
cross-sections are assessed using the first order reliability method cover the design of hot-rolled steel cross-sections, including I-sections
(FORM) set out in Annex D of EN 1990 [54]. Within the adopted FORM, and SHS/RHS. The developments of the method and the derivation of
the probability of failure Pf, i.e. the probability that the resistance (R) is the resistance functions has been described. Numerical models were
less than or equal to the action effect (E), is expressed in terms of the created and thoroughly validated against available test results for hot-
“safety” or “reliability” index β, as given in Eq. (31), where ϕ is the rolled steel cross-sections under compression and bending. Upon vali-
cumulative distribution function of the standardised normal distribu- dation of the FE models, a comprehensive parametric study was con-
tion. The FORM is based on the assumptions that there is no statistical ducted to generate further numerical data on hot-rolled steel cross-
dependence between the variables in the resistance function and that all sections under different loading cases. The experimental and numerical
the variables follow either a normal or log-normal distribution; a target data were then used to evaluate the accuracy of the provisions of EN
reliability index of 3.8 is selected for ultimate limit state design [54]. A 1993-1-1 [1] and the extended CSM for the structural design of hot-
detailed description of the theoretical background and calculation rolled steel cross-sections.
procedures of the reliability method adopted in EN 1990 can be found A refined quad-linear material model considering both the yield
in [54,55]. plateau and the strain hardening of hot-rolled steels was adopted as the
CSM material model, and the corresponding cross-section resistance
Pf = P (R ⩽ E ) = ϕ (−β ) (31)
expressions for axial compression and bending were presented.
Comparisons with experimental and numerical data revealed that the
In the reliability analysis, the material over-strength, i.e. the mean-
CSM generally provides more accurate and consistent predictions than
to-nominal yield strength ratio fy,mean/fy,norm, was taken as 1.16 and the
the existing design provisions, especially for very stocky cross-sections
COVs of material strength and geometric properties were taken as 0.05
and for Class 3 sections in bending. Reliability analysis of the EN 1993-
and 0.03, respectively, in accordance with the values recommended by
1-1 and the CSM design provisions was conducted against test and FE
Byfield and Nethercot [56]. The key statistical parameters for the
data according to EN 1990 [54]; it was found that the CSM satisfies the
proposed CSM expressions are summarized in Tables 6 and 7, including
reliability requirements for I-sections (i.e. the required value of γM0 is
the number of tests and FE simulations, the design (ultimate limit state)
less than unity) and requires lower partial safety factors γM0 than the
fractile factor kd,n, the mean value of the correction factor b, the coef-
corresponding values of EN 1993-1-1 11] for SHS/RHS. The scope of
ficient of variation of the tests and FE simulations relative to the re-
the foundational work described in the present paper is limited to the
sistance model Vδ, the combined coefficient of variation incorporating
determination of cross-section resistances; further work is needed to
both model and basic variable uncertainties Vr and the required partial
extend the CSM to consider combined loading, member buckling and
safety factor γM0. Note that the parameter b is taken as the average of
the response of indeterminate frames.
the ratios of the test and FE resistances to theoretical (predicted) values,
as given in Eq. (32), in which n is the total number of the tests and FE
simulations, re is the experimental or FE value and rt is the theoretical Acknowledgements
(predicted) value. The definition (Eq. (32)) does not bias the value of b
towards the test or FE results with higher failure resistance, which is The financial support provided by the China Scholarship Council
unlike the least squares method adopted in Annex D of EN 1990 [54]. (CSC) for the first author’s PhD study at Imperial College London is

Table 7
Summary of reliability analysis results for cross-sections in bending.

Design approach Cross-section – axis of bending No. of tests and simulations kd,n b Vδ Vr γM0

EN 1993-1-1 I-sections – major 271 3.126 1.220 0.092 0.109 0.99

I-sections – minor 147 3.158 1.338 0.091 0.108 0.90
SHS/RHS – both axes 318 3.121 1.027 0.087 0.105 1.16

CSM I-sections – major 271 3.126 1.200 0.092 0.109 1.00

I-sections – minor 147 3.158 1.234 0.115 0.099 1.00
SHS/RHS – both axes 318 3.121 1.031 0.068 0.090 1.10

190
X. Yun et al. Engineering Structures 157 (2018) 179–191

gratefully acknowledged. Eng 2004;82(21):21–8.

[28] Ashraf M, Gardner L, Nethercot DA. Structural stainless steel design: resistance
based on deformation capacity. J Struct Eng ASCE 2008;134(3):402–11.
References [29] Afshan S, Gardner L. The continuous strength method for structural stainless steel
design. Thin-Walled Struct 2013;68:42–9.
[1] EN 1993-1-1. Eurocode 3: Design of steel structures — Part 1–1: General rules and [30] Gardner L, Wang FC, Liew A. Influence of strain hardening on the behaviour and
rules for buildings. Brussels: European Committee for Standardization (CEN); 2005. design of steel structures. Int J Struct Stab Dyn 2011;11(5):855–75.
[2] Greiner R, Kettler M, Lechner A, Freytag B, Linder J, Jaspart JP, et al. Plastic [31] Liew A, Gardner L. Ultimate capacity of structural steel cross-sections under com-
member capacity of semi-compact steel sections – a more economic design. Eur Res pression, bending and combined loading. Structures 2015;1:2–11.
Fund Coal Steel; 2008. [32] Gardner L, Yun X, Macorini L, Kucukler M. Hot-rolled steel and steel-concrete
[3] Shifferaw Y, Schafer BW. Inelastic bending capacity of cold-formed steel members. J composite design incorporating strain hardening. Structures 2017;9:21–8.
Struct Eng ASCE 2012;138(4):468–80. [33] Su MN, Young B, Gardner L. Testing and design of aluminium alloy cross-sections in
[4] Lechner A, Kettler M, Greiner R, Unterweger H. Verbesserte Bemessungsregeln für compression. J Struct Eng ASCE 2014;140(9):04014047.
Stäbe mit Klasse 3-Querschnitt. Stahlbau 2012;81(4):265–81. [34] Su MN, Young B, Gardner L. The continuous strength method for the design of
[5] Gardner L, Saari N, Wang F. Comparative experimental study of hot-rolled and cold- aluminium alloy structural elements. Eng Struct 2016;122:338–48.
formed rectangular hollow sections. Thin-Walled Struct 2010;48(7):495–507. [35] Yun X, Gardner L. Stress-strain curves for hot-rolled steels. J Constr Steel Res
[6] Liew A, Boissonnade N, Gardner L, Nseir J. Experimental study of hot-rolled rec- 2017;133:36–46.
tangular hollow sections. In: Proceedings of the Annual Stability Conference, [36] Abaqus. Reference manual, version 6.13. Simulia, Dassault Systèmes, France; 2013.
Structural Stability Research Council, 25–28 March 2014, Toronto, Canada. [37] Nseir J. Development of a new design method for the cross-section capacity of steel
[7] Byfield MP, Nethercot DA. An analysis of the true bending strength of steel beams. hollow sections. PhD thesis, Liège, Belgium: Université de Liège; 2015.
Proc Inst Civil Eng Struct Build 1998;128:188–97. [38] Yun X, Nseir J, Gardner L, Boissonnade N. Experimental and numerical investiga-
[8] Kettler M. Elastic-plastic cross-sectional resistance of semi-compact H-and hollow tion into the local imperfection sensitivity of hot-rolled steel I-sections. In:
sections. PhD thesis, TU Graz, 2008. Proceedings of the 7th international conference on coupled instabilities in metal
[9] Boissonnade N, Jaspart JP, Oerder R, Weynand K. A new design model for the re- structures, Baltimore, Maryland, November 7–8; 2016.
sistance of steel semi-compact cross-sections. In: Proceedings of the 5th European [39] Foster ASJ, Gardner L, Wang Y. Practical strain-hardening material properties for
conference on steel structures, 3–5 September 2008, Graz. use in deformation-based structural steel design. Thin-Walled Struct
[10] Zhou F, Chen Y, Young B. Cold-formed high strength stainless steel cross-sections in 2015;92:115–29.
compression considering interaction effects of constituent plate elements. J Constr [40] Wang J, Afshan S, Gkantou M, Theofanous M, Baniotopoulos C, Gardner L. Flexural
Steel Res 2013;80:32–41. behaviour of hot-finished high strength steel square and rectangular hollow sec-
[11] Zhou F, Long G. Element interaction of cold-formed stainless steel cross-sections tions. J Constr Steel Res 2016;121:97–109.
subjected to major axis bending. J Constr Steel Res 2016;118:22–40. [41] Theofanous M, Chan TM, Gardner L. Flexural behaviour of stainless steel oval
[12] Bock M, Real E. Effective width equations accounting for element interaction for hollow sections. Thin-Walled Struct 2009;47(6):776–87.
cold-formed stainless steel square and rectangular hollow sections. Structures [42] Kucukler M, Gardner L, Macorini L. Lateral–torsional buckling assessment of steel
2015;2:81–90. beams through a stiffness reduction method. J Constr Steel Res 2015;109:87–100.
[13] Seif M, Schafer BW. Local buckling of structural steel shapes. J Constr Steel Res [43] Silvestre N, Gardner L. Elastic local post-buckling of elliptical tubes. J Constr Steel
2010;66(10):1232–47. Res 2011;67(3):281–92.
[14] Kato B. Deformation capacity of steel structures. J Constr Steel Res [44] Dawson RG, Walker AC. Post-buckling of geometrically imperfect plates. J Struct
1990;17(1):33–94. Div ASCE 1972;98(1):75–94.
[15] Beg D, Hladnik L. Slenderness limit of class 3 I cross-sections made of high strength [45] EN 1993-1-5. Eurocode 3: Design of steel structures. Part 1–5: Plated structural
steel. J Constr Steel Res 1996;38(3):201–17. elements. Brussels: European Committee for Standardization (CEN); 2006.
[16] Ren WX, Zeng QY. Interactive buckling behaviour and ultimate load of I-section [46] ECCS. Ultimate limit state calculation of sway framed with rigid joints. Technical
steel columns. J Struct Eng 1997;123(9):1210–7. Committee 8 (TC 8) of European Convention for Constructional Steelwork (ECCS)
[17] Davids AJ, Hancock GJ. Compression tests of short welded I-sections. J Struct Eng 1984, Tech. Rep.; No. 33.
1986;112(5):960–76. [47] DIN 18800 1–4. Stahlbauten. Beuth Verlag GmbH, Berlin; 1990.
[18] Kettler M, Ein Greiner R. “Overall-Konzept” für die Querschnittstragfähigkeit im [48] Schafer BW, Ádány S. Buckling analysis of cold-formed steel members using
elasto-plastischen Bereich. Stahlbau 2009;78(10):742–9. CUFSM: conventional and constrained finite strip methods. In: Proceedings of the
[19] Schafer BW. Review: the direct strength method of cold-formed steel member de- 18th international specialty conference on cold-formed steel structures, 26–27
sign. J Constr Steel Res 2008;64(7):766–78. October 2006, Orlando, Florida, USA. p. 39–54.
[20] Yu C, Schafer BW. Simulation of cold-formed steel beams in local and distortional [49] Torabian S, Schafer BW. Role of local slenderness in the rotation capacity of
buckling with applications to the direct strength method. J Constr Steel Res structural steel members. J Constr Steel Res 2014;95:32–43.
2007;63(5):581–90. [50] Zhao O, Afshan S, Gardner L. Structural response and continuous strength method
[21] AISI S100. North American Specification for the design of cold-formed steel struc- design of slender stainless steel cross-sections. Eng Struct 2017;140:14–25.
tural members. Washington, DC: American Iron and Steel Institute (AISI); 2012. [51] Ahmed S, Ashraf M, Anwar-Us-Saadat M. The continuous strength method for
[22] Kemp AR, Byfield MP, Nethercot DA. Effect of strain hardening on flexural prop- slender stainless steel cross-sections. Thin-Walled Struct 2016;107:362–76.
erties of steel beams. Struct Eng 2002;80(8):188–97. [52] Liew A. Design of structural steel elements with the Continuous Strength Method.
[23] Gardner L, Nethercot DA. Experiments on stainless steel hollow sections—Part 1: PhD thesis, Department of Civil and Environmental Engineering, Imperial College
Material and cross-sectional behaviour. J Constr Steel Res 2004;60(9):1291–318. London, UK; 2014.
[24] Chan TM, Gardner L. Compressive resistance of hot-rolled elliptical hollow sections. [53] SCI, Steel building design: Design data, The Steel Construction Institute; 2013.
Eng Struct 2008;30(2):522–32. [54] EN 1990. Eurocode: Basis of structural design. Brussels: European Committee for
[25] Gardner L. The continuous strength method. Proc Inst Civil Eng Struct Build Standardization (CEN); 2002.
2008;161(3):127–33. [55] Afshan S, Francis P, Baddoo NR, Gardner L. Reliability analysis of structural
[26] AISC Design Guide 27: Structural Stainless Steel. American Institute of Steel stainless steel design provisions. J Constr Steel Res 2015;114:293–304.
Construction (AISC); 2013. [56] Byfield MP, Nethercot DA. Material and geometric properties of structural steel for
[27] Gardner L, Nethercot DA. Stainless steel structural design: a new approach. Struct use in design. Struct Eng 1997;75(21). 393-67.

191