Structures 27 (2020) 2082–2095

Contents lists available at ScienceDirect

Structures
journal homepage: www.elsevier.com/locate/structures

Analysis and design of cold-formed steel storage rack uprights under T

localised fires
Chong Rena, Peng Zhanga, Shen Yanb, Liusi Daia,
⁎

a
Department of Civil Engineering, Shanghai University, Shanghai 200444, China
b
School of Civil Engineering, The University of Sydney, NSW 2008, Australia

ARTICLE INFO ABSTRACT

Keywords: A fire in a storage warehouse creates a scenario that the load bearing members, notably storage rack uprights,
Cold-formed steel are exposed to localised fires, generating the temperature field whose distribution is not uniform along the
Storage rack uprights member length. This non-uniformity in temperature distribution is different from the assumption implied in the
Localised fires current structural fire design guidelines. This paper investigates the resistance of the uprights subjected to lo­
Temperature gradient
calised fires, in particular, fires at the bottom of the uprights, by accounting for the temperature gradients along
Buckling interactions
Finite element analysis
the member length. Finite Element (FE) models of cold-formed steel storage rack uprights are developed, and are
validated against experimental results from aspects of both heat transfer and buckling behaviour. The validated
FE models are employed in a parametric study to investigate the resistance and buckling behaviour of the
uprights under localised fires. The parametric study results show that the failure modes of the uprights are very
sensitive to the localised fires. Moreover, it is observed that the Direct Strength Method (DSM), in the current
form, does not produce accurate predictions for the cold-formed steel storage rack uprights under localised fires.
Hence, modified DSM global curves are proposed in the paper to allow for the effects of localised fires, and prove
to be sufficiently accurate through reliability analysis.

1. Introduction Since the inception of the new century, a few investigations have
studied the behaviour of cold-formed steel structures when subjected to
Cold-formed steel is increasingly becoming a favourable option in uniform elevated temperature caused by fires. The major findings of the
lightweight steel constructions such as storage rack systems, by virtue relevant studies include: (1) the design methods in specifications need
of its high strength to weight ratio, environmentally friendly features to be modified to be applied into cold-formed steel members at elevated
and convenience in manufacture. The current design codes including temperatures [4,5]. For instance, for cold-formed steel channel beams,
RMI [1] in North America, EN 15512 [2] in Europe, and AS/NZS4084 Eurocode 3 Part 1.2 [6] gives over-conservative predictions on the
[3] in Australia and New Zealand have been successfully applied in the lateral-torsional buckling resistance, while Eurocode 3 Part 1.3 [7] was
design of cold-formed steel storage racks. However, the influences of found to be unsafe for some temperatures and over-conservative for
elevated temperatures on cold-formed steel members are not con­ others [5]; (2) for cold-formed steel members such as channel columns,
sidered in the aforementioned design codes. Due to the extreme lack of the Effective Width Method (EWM) and Direct Strength Method (DSM)
fire resistance evaluation during the rack structural design, cold-formed produce conservative but acceptable predictions when certain mod­
steel members in the rack system have been facing serious challenges. ifications are adopted, notably by accounting for the reduction of ma­

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

https://doi.org/10.1016/j.istruc.2020.08.022
Received 13 April 2020; Received in revised form 3 July 2020; Accepted 6 August 2020
Available online 14 August 2020
2352-0124/ © 2020 Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.
C. Ren, et al. Structures 27 (2020) 2082–2095

Table 1 temperature but it needs to be modified to evaluate the behaviour of

Summary of models. steel members under non-uniform temperatures, and the relevant in­
Specimens b(mm) h(mm) t(mm) A(mm2) Anet(mm2) L(mm) T( °C) vestigation is proposed in the paper.
It can be seen that there are knowledge gaps in the behaviour of
75-55-1.8 75 55 1.8 369.4 254.2 1500, 20, cold-formed steel storage rack uprights under localised fires, and in the
90-56-1.8 90 56 1.8 411.8 296.6 2000, 200,
corresponding DSM modifications. Therefore, this study presents a
90-66-1.8 90 66 1.8 496.5 379.3 2400, 400,
90-66-2.0 90 66 2.0 548.2 420.2 3000, 600,
numerical investigation into the buckling behaviour of cold-formed
100-60-1.8 100 60 1.8 477.7 362.5 3500, 800 steel storage rack uprights under localised fires. Having validated
100-60-2.3 100 60 2.3 610.2 463.0 4000, against available experimental results, a total of 280 finite element (FE)
100-90-2.0 100 90 2.0 653.4 525.4 4500 models are developed, and the results of heat transfer analysis, linear
120-80-2.0 120 80 2.0 658.7 530.7
analysis and nonlinear analysis are presented and discussed. The in­
fluences of temperature gradients along the upright length on the
buckling behaviour are explicitly explained. Finally, by comparing the
terial properties at elevated temperatures [8]; (3) at the critical tem­ ultimate loads derived from the FE analysis and the current DSM,
peratures, the support condition and the stiffness of the surrounding modifications on the DSM design curves are proposed for storage rack
structures have a significant influence on the structural behaviour of uprights at various elevated temperatures under localised fires.
cold-formed steel members, such as lipped channel sections [9].
However, when a fire outbreaks in a large compartment such as a 2. Development and validation of finite element analysis models
storage warehouse, the load bearing members are usually exposed to
localised fires. Therefore, recent investigations started to focus on the FE models are developed in the commercial FE package ABAQUS to
behavioural differences of cold-formed steel members under uniform investigate the structural behaviour of steel storage rack uprights ex­
and non-uniform elevated temperatures [10–18]. The major conclu­ posed to localised fires. Each FE model is subjected to three analyses,
sions include: (1) the gradient temperature generated by localised fires i.e., heat transfer analysis, linear analysis and nonlinear analysis. A heat
has a significant influence on the failure mode [15,16]; For instance, for transfer analysis is performed first to determine the temperature dis­
the cold-formed thin-walled steel channel columns investigated in [16], tribution along the member length under a localised fire, which are
the failure mode changes from local buckling at ambient temperature to then input into a mechanical analysis to study the behaviour of the
interactions of buckling modes under localised fires; (2) the location of uprights subjected to the localised fire, either being a linear analysis for
localised fires impacts greatly on the behaviour of cold-formed steel determining the buckling modes of the upright and the associated cri­
members, notably the buckling temperatures of columns [17]; (3) the tical buckling load, or being a material and geometric non-linear ana­
current design codes for steel structures in fire [6,7] do not allow for the lysis considering geometrical imperfections (MGNIA) for obtaining the
temperature gradient and the resulted failure mode change, and ultimate load and the failure mode. The Section 2.7 provides that the
thereby are not accurate for the design of cold-formed steel members modelling techniques used in each analysis are validated against the
under localised fires, normally underestimating their performance [18]. experiments reported in [9,23,29].
Although many researches have been mentioned above, very limited
investigations reported on the buckling behaviour of cold-formed sto­ 2.1. Model geometries
rage rack uprights under localised fires, in which the influence of the
temperature distribution along the member length is of great sig­ A total of 280 FE models were developed in this study, comprising
nificance. eight types of cross-sections, seven lengths of uprights and five target
The DSM is an efficient design method to predict the strengths of temperatures, as shown in Table 1. The models are designated as “web
cold-formed steel columns or beams undergoing local buckling (LB), length (b) – flange height (h) – thickness (t) – upright length (L) –
distortional buckling (DB), and/or global buckling (GB). Therefore, it is temperature at the upright bottom end (T)”. For example, Model 90–56-
potentially applicable for predicting the strength of cold-formed storage 1.8-1500-200 denotes an upright with b = 90 mm, h = 56 mm,
rack uprights under localised fires. Investigations of DSM for the single t = 1.8 mm and L = 1500 mm, and it is subjected to a target tem­
buckling mode [19–21] or the interaction of buckling modes [22–25] at perature T = 200 °C. Fig. 1 shows the cross-sectional geometries and
ambient temperature were reported in the literature, as were the more perforation features of the uprights used in this study.
recent studies and modifications of DSM considering elevated tem­
peratures. Based on the experimental and numerical studies, Gunalan 2.2. Material properties
et al. [26] improved the accuracy of the current DSM design method for
cold-formed steel members featuring a local buckling mode, by con­ The uprights are manufactured using cold-formed Q235 steel, which
sidering the effects of nonlinear stress-strain characteristics at elevated is widely used in the storage rack industry of China. Material non­
temperatures. Shahbazian and Wang [27,28] developed and validated linearities are included in the FE analysis to achieve accurate simula­
numerical models to assess the suitability of the current global and tions. The true stress-strain curves are employed in the material defi­
distortional DSM curves for the cold-formed steel compression members nition. It is noted that the cold-forming process significantly reduces the
under elevated temperatures. The studies pointed out that the current ultimate-to-yield stress ratio of the material, and thus an elastic-per­
global and distortional DSM curves are appropriate for uniform fectly plastic material model with the yield stress σy = 235 MPa and the

2083
C. Ren, et al. Structures 27 (2020) 2082–2095

Fig. 1. (a) Cross-sectional geometries (The relative position of the centre of holes is indicated). (b) Perforation features.

2084
C. Ren, et al. Structures 27 (2020) 2082–2095

Table 2
Comparison of numerical results with different mesh sizes.
Specimens Ultimate Load Ultimate Load Ultimate Load (kN)
(75-55-1.8-L-20) (kN) (kN) (Mesh size: 10 mm)
(Mesh size: 2 mm) (Mesh size: 5 mm)

L = 1500 mm 46.45 46.70 47.35

L = 2000 mm 35.65 35.76 36.10
L = 2400 mm 27.64 27.73 27.87
L = 3000 mm 19.06 19.13 19.26

freedom are constrained, i.e. U1 = U2 = UR3 = 0.

2.4. Mesh

Due to the small wall thickness, cold-formed thin-walled steel

members are typically modelled using shell elements. In Abaqus, two
types of shell elements, the four-node general-purpose shell element
Fig. 2. Input elastic-plastic stress-strain relationships of steel at different tem­ (S4) and the four-node general-purpose shell element with reduced
peratures base on EN 1993-1-2. integration and hourglass control (S4R), are commonly adopted.
According to Shahbazian and Wang [28], both S4 and S4R element
Young's modulus E = 210 GPa is assumed at room temperature. The types can be used in FE analysis of cold-formed steel channel sections
deterioration of the yield stress and elastic modulus under elevated under elevated temperatures, showing comparable accuracy. Con­
temperatures are determined according to EN 1993-1-2 [6], with the sidering its advantage in computing efficiency, S4 element type is em­
deteriorated strain-stress curves under considered temperatures shown ployed in this study. Mesh convergence analysis is performed. The re­
in Fig. 2. It should be noted that high temperatures make stress-strain sult shows that a mesh size of 10 mm can be used in this study to predict
curves of cold-formed steels become rounded. It has been proved that the ultimate load of the storage rack uprights, as shown in Table 2.
the buckling occurred in the elastic range which is mainly governed by
Young’s modulus, the influences of different stress-strain curves are 2.5. Geometrical imperfections
insignificant [30]. Therefore, the elastic-perfectly plastic material
model in high temperature, as shown in Fig. 2, are employed in FE It is well known that the ultimate load of the cold-formed steel
models. The Poisson's ratio is 0.3 and the density is 7850 kg/m3. columns is significantly affected by geometrical imperfections, whether
at ambient temperature or elevated temperatures [4]. Seeing that the
geometrical imperfection is generated during the process of production
2.3. Boundary and loading conditions and transportation, i.e., prior to the occurrence of fires, in this study the
initial geometrical imperfection of each storage rack upright is defined
Storage rack uprights in practice are simply supported at the ends by its first critical buckling mode at room temperature. A series of in­
and are subjected to axial compression. In the FE model, the cross- vestigations are conducted for the magnitude of initial global im­
section at each end is tied through coupling constraints to a reference perfections, i.e., L/2000, L/1500, L/1250, L/1000 and L/500, respec­
point defined at the centroid of the cross-section, and a concentrated tively. Benchmarked against the experimental results reported in [23]
force is set on the reference point at the top of the member, as shown in which different lengths and sections of uprights are considered, and it
Fig. 3. At the base end, all translational and the longitudinal-axis ro­ can be found that uprights with a maximum imperfection equal to L/
tational degrees of freedom are constrained, i.e. 1000 provides the ultimate loads closest to the experimental results, as
U1 = U2 = U3 = UR3 = 0; at the loading end, the lateral and shown in Table 3. Therefore, the value of geometric imperfection is
transverse translational and the longitudinal-axis rotational degrees of defined as L/1000 for all FE models.

Fig. 3. Boundary conditions and loading position.

2085
C. Ren, et al. Structures 27 (2020) 2082–2095

Table 3
Sensitivity study of geometrical imperfections based on the experimental results reported in [23].
Specimen label Ptest PL/2000 PL/1500 PL/1250 PL/1000 PL/500 PL/2000/ PL/1500/ PL/1250/ PL/1000/ PL/500/
(kN) (kN) (kN) (kN) (kN) (kN) Ptest Ptest Ptest Ptest Ptest

75-55-1.8-1500-20 59.79 64.77 63.79 63.06 62.01 57.61 1.084 1.068 1.056 1.038 0.964
90-66-1.8-2000-20 73.92 82.15 80.26 78.88 77.06 70.22 1.111 1.086 1.067 1.042 0.950
90-66-1.8-2400-20 58.70 64.02 63.48 61.39 59.91 54.47 1.090 1.081 1.046 1.021 0.928
90-66-2.0-2000-20 81.07 88.71 86.79 85.40 83.51 76.25 1.094 1.071 1.053 1.03 0.941
Average – – – – – – 1.095 1.076 1.056 1.033 0.945

Fig. 5. Comparison of heat transfer between the FEA and the experiment
conducted by Craveiro et al. [9].

Table 4
Comparisons of buckling modes and buckling stress of simple lipped channel
between the experiment from Young et al. [29] and the FEA.
Specimens Buckling Mode Buckling Stress (MPa)

Test FEA Test FEA

CH1-5-800 D(2) D(2) 45.0 44.5

CH1-6-800 D(2) D(2) 55.5 58.7
CH1-7-400 D(1) D(1) 88.8 89.7
CH1-7-600 D(1) D(1) 75.8 86.3
CH1-7-800 D(2) D(2) 68.6 84.0

Note: D = Distortional buckling mode; () = Numbers of buckle half-wave­

length.

(warehouse) and a shopping centre in terms of the large space, the

Fig. 4. Localised fire model. maximum rate of heat release (RHRf) of the warehouse is estimated as
250 kW/m2, similar to the RHRf of a shopping centre in heat transfer
analysis. The formulas for calculating gas temperature along the ver­
2.6. Localised fire models
tical flame axis are obtained from EN 1991-1-2 [31]. In heat transfer
analysis, the convection coefficient c = 25W / m2K , and radiation
As shown in Fig. 4, the localised fire model is divided into the lower
coefficients f = 0.8, m = 0.8 where f is the emissivity of flames and m
combustion (flame) and upper non-combustion (plume) regions. In
is the surface emissivity of members [31]. In practice, the distance
view of the similarity between a storage rack upright environment

2086
C. Ren, et al. Structures 27 (2020) 2082–2095

Fig. 6. Comparison of buckling mode (DB) of section CH1-7-800 between the

FEA and the experiment [29].

Fig. 7. Comparison of failure mode (FTB) of section 90-66-2.0-H-2000 between

between the adjoining uprights is from 1 m to 3 m, which is the length
the FEA and the experiment [23].
of beams. Moreover, the study focuses on single uprights without
considering the whole structure under fires. Therefore, the diameter of
the fire D is defined as 0.5 m-1.5 m equalling to the half-length of beam
with no effect to the adjoining uprights. According to EN 1991-1-2 [31], validate the heat transfer analysis modelling technique, the FE model of
the flame lengths Lf have relevance to the diameter of the fire D and the the specimen C_PP_30LL_K1-3 reported in the paper [9] is developed
rate of heat release of the fire Q. Based on EN 1991-1-2 [31], the cal­ following the above described modelling procedure. In the validation of
culation result of the flame lengths Lf is 0.5 m-1.5 m equalling to a third heat transfer analysis, due to a slender column maintain certain dis­
length of uprights. Particularly, the height of upright is from 1.5 m to tance from fire source in chamber (2.5 m × 1.5 m × 1.5 m), heating
4.5 m. Because the flame of a localised fire will not cover the entire process not consider the emissivity of fire but thermal convection. It is
height of an upright, the flame length of a localised fire is assumed to be observed that the FE heat transferring analysis and experimental results
one-third of the upright height. are in good agreement, as shown in Fig. 5. In regard to the linear
analysis, the FE model of the simple lipped channel sections (CH1)
2.7. Validations subjected to compression load tested in the paper [29] is developed and
the FE results are compared with the experimental data in Table 4 and
In this section, the modelling techniques used in the heat transfer Fig. 6. Satisfactory agreement is observed in terms of the buckling mode
analysis, linear analysis and nonlinear analysis are validated against the and the associated buckling stresses, demonstrating that the above
experiments reported in the literature. It should be noted that the described FE technique is capable of obtaining the buckling behaviour
material model and geometric imperfection of the specimen in FEM are of the cold-formed steel uprights. With respect of the nonlinear analysis,
based on the assumed values described in Section 2.2 and 2.5. To FE models of the experimental investigations with hinged support and

Table 5
Comparisons of test results from Zhao et al. [23] and FEA ultimate loads.
Specimens FEA column dimensions (mm) fy (MPa) E (GPa) Ultimate load (kN) Pu,FEA/Pu,test

b d t L Pu,test Pu,FEA

75-55-1.8-1500 75 55 1.8 1500 371.3 206 59.79 62.05 1.04

90-66-1.8-1500 90 66 1.8 1500 373.7 206 91.63 94.50 1.03
90-66-1.8-2000 90 66 1.8 2000 373.7 206 73.92 77.06 1.04
90-66-1.8-2400 90 66 1.8 2400 373.7 206 58.70 59.91 1.02
90-66-2.0-2000 90 66 2.0 2000 341.7 206 81.07 83.50 1.03
120-80-2.0-1500 120 80 2.0 1500 299.3 206 119.43 112.8 0.94
Mean 1.02
Standard Deviation 0.035

2087
C. Ren, et al. Structures 27 (2020) 2082–2095

along the length of upright simulated from the present FEA model is
adopted in this study, and parametric analyses are conducted to explain
the influences of different temperature distributions.

3.2. Buckling behaviour analysis

Based on the Section 3.1, the buckling behaviour of the cold-formed

steel uprights under localised fires is investigated in this section. The
first buckling mode and the corresponding critical buckling load (Pcr,1)
are obtained using a liner analysis, while the ultimate loads (Pu) and the
failure mode are obtained using a nonlinear analysis. The detailed re­
sults of all FE models can be found in the Appendix, where LB, DB and
GB represent local buckling, distortional buckling and global buckling
respectively.

3.2.1. Linear analysis

The critical buckling load is an important indicator in assessing the
stability of compression members as well as an essential parameter used
in the DSM. Due to the temperature gradient along the length, material
properties of the cold-formed steel storage rack uprights, including the
Young’s modulus and yield stress, are weakened with varying tem­
peratures, which needs to be accounted for in the analysis [32,33]. The
explicit linear analysis results are provided in the Appendix, including
the first buckling modes and the related critical buckling loads (Pcr,1) at
room temperature (Pcr,1,20) and at elevated temperature (Pcr,1,T) where
the subscript T denotes the temperature (T = 200 °C, 400 °C, 600 °C,up
to 800 °C).
Table 6 summaries the reductions in the first critical buckling loads
when subjected to localised fires, through defining a reduction factor as
Pcr,1,T/Pcr,1,20. It is observed that the reduction factor is affected by both
the elevated temperature and the upright length. Elevated temperature
has a significant negative impact on the critical buckling loads for
compression uprights shorter than 3000 mm (i.e., 1500 mm, 2000 mm
Fig. 8. The results of heat transfer analysis on 75-55-1.8-1500 under a localised
and 2400 mm). It can be found through this study that that the re­
fire.
duction factor increases with the increased upright length.
Fig. 9 shows the mean values and standard deviations of the re­
axial load presented in [23] are developed and analysed via the Riks duction factors. Each point in the Fig. 9 represents the mean value of
method, which determine the ultimate loads and failure mode of the the reduction factors of all sections with the same length and subjects to
uprights in line with experimental results, as shown in Table 5 and the same temperature elevation, and the vertical coloured bands re­
Fig. 7. present the standard deviation of these sections. It can be seen from the
Fig. 9 that when the temperature is elevated to 200 °C, the critical
3. Results buckling loads of the uprights deteriorate by the minimum amount with
the reduction factors greater than 0.98 for all upright lengths. When the
3.1. Heat transfer analysis temperature reaches 400 °C and 600 °C, the effect of elevated tem­
perature on the critical buckling loads starts to appear, particularly for
The validated FE method is employed in the heat transfer analysis of those members shorter than 3000 mm. When the temperature keeps
the 280 FE models of cold-formed steel storage rack uprights under increasing to 800 °C, the uprights are significantly undermined by the
localised fires. Results are explained by taking Upright 75-55-1.8-1500 elevated temperature, as seen in the Fig. 9 that the reduction factors are
as an example. Fig. 8 (a) presents the temperature field of the upright well below 0.80 for all upright length. Moreover, the members shorter
when the localised fire generates a temperature of 800 °C at the bottom than 3000 mm can develop a notable interaction of buckling modes, as
end. Fig. 8 (b) shows the development of the temperature distribution demonstrated by the much deeper vertical coloured band. The existence
along the upright length as the temperature at upright bottom increases of buckling mode interaction makes it more difficult for the prediction
from 20 °C to 200 °C, 400 °C, 600 °C, up to 800 °C. In Fig. 8 (c), the of the critical buckling loads of the short members, comparing with the
temperature time-history curves obtained at the top, middle and bottom longer members undergoing single buckling mode.
of the upright are compared to the ISO 834 curve, showing very dif­
ferent temperature gradient distributions along the length from that 3.2.2. Nonlinear analysis
specified in ISO 834. Therefore, the temperature gradient distribution The ultimate axial loads and failure modes of the cold-formed steel

2088
C. Ren, et al. Structures 27 (2020) 2082–2095

Table 6
Reduction factors (Pcr,1,T/Pcr,1,20) of uprights at elevated temperatures with different lengths and sections.
Length/mm Temperature/ °C Pcr,1,T/Pcr,1,20 St. Dev.

75-55-1.8 90-56-1.8 90-66-1.8 90-66-2.0 100-60-1.8 100-60-2.3 100-90-2.0 120-80-2.0

1500 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 0.986 0.986 0.977 0.976 0.990 0.984 0.977 0.979 0.00493
400 0.949 0.948 0.915 0.909 0.954 0.940 0.920 0.920 0.01657
600 0.870 0.867 0.806 0.800 0.875 0.857 0.796 0.757 0.04133
800 0.630 0.598 0.606 0.640 0.610 0.688 0.557 0.497 0.05333
2000 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 0.996 0.995 0.985 0.985 0.996 0.993 0.989 0.994 0.00436
400 0.976 0.975 0.938 0.931 0.976 0.965 0.949 0.973 0.01723
600 0.903 0.899 0.829 0.823 0.901 0.879 0.836 0.792 0.04015
800 0.713 0.704 0.666 0.667 0.708 0.708 0.537 0.486 0.08201
2400 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 0.997 0.998 0.988 0.987 0.997 0.998 0.992 0.997 0.00424
400 0.986 0.986 0.948 0.943 0.984 0.980 0.958 0.984 0.01712
600 0.926 0.927 0.845 0.839 0.924 0.899 0.870 0.910 0.03412
800 0.728 0.731 0.680 0.681 0.723 0.720 0.603 0.583 0.05429
3000 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 1.000 1.000 0.991 0.992 0.998 1.000 0.993 0.999 0.00367
400 0.996 0.996 0.958 0.954 0.993 0.990 0.966 0.993 0.01696
600 0.960 0.963 0.872 0.863 0.956 0.929 0.894 0.946 0.03818
800 0.753 0.680 0.696 0.690 0.737 0.733 0.698 0.702 0.02449
3500 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 1.000 1.000 0.994 0.995 1.000 1.000 0.995 0.997 0.00249
400 1.000 0.995 0.963 0.960 0.997 0.995 0.968 0.996 0.01613
600 0.970 0.975 0.885 0.879 0.973 0.950 0.905 0.964 0.03843
800 0.772 0.750 0.703 0.698 0.753 0.744 0.718 0.736 0.02418
4000 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 1.000 1.000 0.996 0.997 1.000 1.000 0.997 1.000 0.00163
400 1.000 1.000 0.967 0.965 0.996 0.997 0.972 0.998 0.01478
600 0.984 0.987 0.895 0.893 0.978 0.963 0.915 0.976 0.03814
800 0.773 0.758 0.707 0.698 0.761 0.752 0.735 0.753 0.02502
4500 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 1.000 1.000 0.996 1.000 1.000 1.000 0.996 1.000 0.00173
400 1.000 1.000 0.973 1.000 0.995 1.000 0.974 0.998 0.01107
600 0.990 0.992 0.907 0.944 0.984 0.974 0.921 0.983 0.03113
800 0.822 0.810 0.743 0.765 0.813 0.802 0.778 0.803 0.02544

storage rack uprights subjected to localised fires are investigated using

the GMNIA, with the results summarised in Appendix. Fig. 10 (a) shows
the effect of temperature on the load-displacement response of Upright
75-55-1.8-1500, indicating a significant softening in the upright re­
sponse and a significant reduction in the ultimate resistance caused by
the elevated temperature. Fig. 10 (b) illustrates the load-displacement
curves of Upright 75-55-1.8 with different lengths at room temperature,
demonstrating that the upright responses are reduced with the increase
of upright length. Fig. 11 shows the typical deformed shapes of uprights
at the ultimate load and the corresponding buckling mode, including
distortional buckling, local-global buckling interaction, distortional-
global buckling interaction and local-distortional-global buckling in­
teraction.
Fig. 12 shows the dependence of the failure mode interaction on
temperature, length and flange width-thickness ratio h/t. In the Fig. 12,
points marked by “GB” represent global buckling failure mode, while
points marked by “Others” have other buckling modes, including local
buckling (LB), distortional buckling (DB), local-distortional buckling
interaction (LB + DB), local-global buckling interaction (LB + GB),
distortional-global buckling interaction (DB + GB) and local-distor­
tional-global buckling interaction (LB + DB + GB). A modest elevation
Fig. 9. Mean values and standard deviations of reduction factors of critical in temperature, i.e. 200 °C, does not change the failure mode compared
buckling loads (Pcr,1,T/Pcr,1,20) at elevated temperature with different lengths.
with room temperature, as shown in Fig. 12 (a). However, when the

2089
C. Ren, et al. Structures 27 (2020) 2082–2095

temperature is further elevated, notable changes in the failure mode of

uprights can be observed for shorter uprights with lengths no larger
than 2400 mm, as shown in Fig. 12 (b).
The reduction factor of the ultimate load, Pu,T/Pu,20, is defined as the
ratio of the ultimate loads at elevated and room temperatures. Table 7
shows the calculation results of the reduction factors, and Fig. 13 shows
the mean values and the standard deviations of the reduction factors,
along with the dependence of the statistical characteristics on the
temperature and upright length. Noted that similar to Fig. 9 each point
represents the mean value of reduction factors for all studied sections
with a specified length and temperature, and the vertical coloured
bands reflect the standard deviation of the reduction factors. It can be
seen that the ultimate loads are barely reduced to 200 °C, i.e., reduction
factors are close to unity at 200 °C, but are significantly undermined by
the ever-elevating temperature, especially for the relatively shorter
uprights. For a steel upright with a length of 1500 mm, the reduction in
ultimate load can be up to 60%.

4. Design approach

Since the propose of the Direct Strength Method (DSM) by Schafer

and Peköz [34], and Schafer [35], the DSM has been used in the design
for cold-formed steel components in North American Specification [36]
and Australian/New Zealand Standard [37]. However, there is no ap­
plicable DSM for cold-formed steel storage rack uprights at elevated
temperatures, especially under localised fires. In this section, the
modified DSM curves for axial compression cold-formed steel rack up­
rights under localised fires are presented and evaluated.
Normally, upright racks are comparatively long, so global buckling
mode is the dominant buckling mode. In the DSM, the axial compres­
sive strength of columns can be calculated using the global equation for
global buckling:
2
(0.658 c ) Py if c 1.5
Pu =
( )P
0.877
2
c
y if c > 1.5
(1)

where c = Py / Pcre , Py = Anet Fy . Anet is the net cross-sectional area, Fy is

the yield stress at room temperature and Pcre is the critical flexural

Fig. 10. (a) Load-displacement curves of 75-55-1.8-1500 with different tem­

peratures. (b) Load-displacement curves of section 75-55-1.8 with different
lengths at room temperature.

Fig. 11. Deformed shapes of perforated uprights at failure point.

2090
C. Ren, et al. Structures 27 (2020) 2082–2095

calculate the critical buckling load is not accurate for calculation of an

individual critical flexural buckling load. Hence, for convenience pur­
pose, in calculating the slenderness considering localised fires (λc,T), the
first critical buckling load considering the localised fire (Pcr,1,T) is
adopted, replacing the critical flexural buckling load (Pcre):

c, T = Py / Pcr ,1, T , Py = Anet Fy (2)

where Py is the yield strength of uprights at room temperature and Anet
is the net cross-sectional area.
Fig. 14 shows the five groups of results obtained from the FE ana­
lysis, each corresponding to a temperature T = 20, 200, 400, 600,
800 °C. Fig. 14 (a) (T = 20 °C) and Fig. 14 (b) (T = 200 °C) show that
the current DSM global curve is able to predict the ultimate load of
compressive uprights at room temperature and at 200 °Cwith accep­
table accuracy. It is to be noted that because Py in Eq. (1) is the yield
strength at room temperature, the reduction in material properties due
to elevated temperatures must be accounted for. When the tempera­
tures reach 400 °C, 600 °C and 800 °C, as shown in Fig. 14 (c) (d) (e),
the current DSM curve significantly over-estimates the ultimate loads of
the axially compressive rack uprights. For more accurate prediction on
the strength of the uprights under elevated temperatures, the modified
DSM global equations are presented for 400 °C as Eq. (3), 600 °C as Eq.
(4) and 800 °C as Eq. (5):
2
(0.77 c ) Py × 0.77 if c 1.3
Pu,400 =
( )P 0.877
2
c
y if c > 1.3
(3)
2
(0.87 c ) Py × 0.5 if c 1.5
Pu,600 =
( ) P × 0.5
1.7
2
c
y if c > 1.5
(4)
2
(0.85 c ) Py × 0.4 if c 1.95
Pu,800 =
( ) P × 0.4
2
2
c
y if c > 1.95
(5)
Reliability analyses are performed for the modified DSM global
curves at different temperatures, as shown in Fig. 15, in which 20 °C
and 200 °C are evaluated using the current DSM global equation, while
400 °C, 600 °C and 800 °C are evaluated using Eqs. (3)–(5) respectively.
It is observed that the modified DSM global curves provide accurate
predictions of the ultimate strength for the cold-formed steel uprights
dominated by the global buckling under localised fires.

5. Conclusions

Fig. 12. Three-dimensional parametric distributions of failure modes with h/t, A series of numerical investigations have been presented in this
temperatures and lengths. (a)T = 20 °C, 200 °C, (b)T = 400 °C, 600 °C, 800 °C. paper to study the temperature distribution and buckling behaviour of
the axially compressive cold-formed steel storage rack uprights under
localised fires. The results of FEA presented and discussed in this paper
consist of three parts, i.e., the heat transfer analysis, the linear analysis
buckling load at room temperature. and the nonlinear analysis. Some significant conclusions drawn from
Due to the continuous perforations and non-uniform temperature the present numerical results are summarised as follow:
distribution, the critical flexural buckling load (Pcre) cannot be calcu­
lated by the finite strip software CUFSM. Therefore, in this study, the
finite element software ABAQUS is used instead. Due to the existence of
• Fires at the bottom of the uprights generate nonlinear temperature
gradients along the length of the uprights.
temperature gradient, the linear perturbation analysis which is used to • The buckling behaviour of the uprights under localised fires is

2091
C. Ren, et al. Structures 27 (2020) 2082–2095

Table 7
Reduction factors (Pu,T/Pu,20) of upright at elevated temperature with different lengths and sections.
Length/mm Temperature/°C Pu,T/Pu,20 St. Dev.

75-55-1.8 90-56-1.8 90-66-1.8 90-66-2.0 100-60-1.8 100-60-2.3 100-90-2.0 120-80-2.0

1500 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 0.988 0.986 0.975 0.973 0.990 0.984 0.980 0.967 0.00757
400 0.857 0.812 0.782 0.765 0.763 0.744 0.744 0.717 0.04129
600 0.552 0.520 0.500 0.507 0.490 0.530 0.503 0.481 0.02134
800 0.419 0.401 0.384 0.403 0.373 0.439 0.392 0.379 0.02057
2000 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 0.995 1.000 0.987 0.987 0.995 0.995 0.988 0.995 0.00462
400 0.976 0.974 0.878 0.866 0.919 0.885 0.806 0.740 0.07483
600 0.712 0.662 0.599 0.610 0.590 0.634 0.542 0.497 0.06259
800 0.499 0.484 0.451 0.471 0.438 0.492 0.418 0.386 0.03668
2400 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 0.997 0.997 0.990 0.989 0.998 0.998 0.993 0.998 0.00347
400 0.986 0.986 0.921 0.916 0.986 0.981 0.834 0.838 0.06109
600 0.868 0.841 0.696 0.707 0.720 0.759 0.555 0.557 0.10734
800 0.578 0.569 0.515 0.532 0.506 0.554 0.435 0.418 0.05559
3000 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 1.000 1.000 0.990 0.993 1.000 1.000 0.995 0.999 0.00380
400 0.995 0.996 0.948 0.944 0.994 0.993 0.891 0.994 0.03592
600 0.965 0.963 0.801 0.803 0.940 0.905 0.675 0.687 0.11084
800 0.662 0.537 0.593 0.597 0.607 0.636 0.501 0.493 0.05775
3500 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 1.000 1.000 0.993 0.986 1.000 1.000 0.995 1.000 0.00481
400 1.000 0.995 0.957 0.954 0.996 0.994 0.934 0.998 0.02424
600 0.979 0.973 0.844 0.842 0.978 0.953 0.775 0.840 0.07595
800 0.697 0.689 0.639 0.640 0.663 0.678 0.563 0.566 0.04873
4000 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 1.000 1.000 0.954 1.000 1.000 1.000 0.994 1.000 0.01494
400 0.991 1.000 0.938 0.996 1.000 0.996 0.954 0.998 0.02265
600 0.983 0.979 0.880 0.975 0.981 0.966 0.826 0.968 0.05507
800 0.735 0.718 0.681 0.674 0.708 0.707 0.621 0.632 0.03822
4500 20 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.00000
200 1.000 1.000 1.000 1.000 1.000 1.000 0.995 1.000 0.00154
400 1.000 1.000 1.000 1.000 1.000 0.995 0.965 1.000 0.01145
600 0.989 0.991 0.985 0.982 0.982 0.972 0.867 0.985 0.03899
800 0.839 0.779 0.749 0.742 0.768 0.766 0.692 0.709 0.04199

affected by the temperature gradient along the upright length.

• Although global buckling dominates the failure mode at ambient
temperature, the uprights tend to develop local-global interaction
when subjected to localised fires with temperature greater than
400 °C.
• Temperatures below 200 °C have an insignificant effect on the ul­
timate loads of the uprights, and the current DSM is still reliable for
design of cold-formed steel uprights receiving temperature raise less
than 200 °C. However, for the rack uprights subjected to localised
fires generating temperature greater than 400 °C, the current DSM
global equations at ambient temperatures is not safe.
• Modified DSM curves are proposed for the axially compressive cold-
formed steel storage rack uprights under localised fires, showing
reasonably good accuracy when compared with the results of the FE
parametric study.

Declaration of Competing Interest

The authors declare that they have no known competing financial

interests or personal relationships that could have appeared to influ­
Fig. 13. Mean values and standard deviation of reduction factors of ultimate ence the work reported in this paper.
load at elevated temperature versus lengths.

2092
C. Ren, et al. Structures 27 (2020) 2082–2095

Fig.14. Comparison of the current DSM global curve and modified curves.

2093
C. Ren, et al. Structures 27 (2020) 2082–2095

Fig. 15. Results of reliability analysis for modified global DSM curves.

2094
C. Ren, et al. Structures 27 (2020) 2082–2095

Acknowledgement bare steel columns subjected to localized fire. J Struct Eng 2015;141.
[18] Zhang C, Choe L, Seif M, Zhang Z. Behavior of axially loaded steel short columns
subjected to a localized fire. J Constr Steel Res 2015;111:103–11.
The work presented in the paper was supported by the National [19] Hancock GJ. Cold-formed steel structures. J Constr Steel Res 2003;59:473–87.
Natural Science Foundation of China (Grant No. 51908347). [20] Casafont M, Pastor MM, Roure F, Peköz T. An experimental investigation of dis­
tortional buckling of steel storage rack columns. Thin-Walled Structures
2011;49:933–46.
Appendix A. Supplementary data [21] Macdonald M, Heiyantuduwa MA, Rhodes J. Recent developments in the design of
cold-formed steel members and structures. Thin-Walled Struct 2008;46:1047–53.
Supplementary data to this article can be found online at https:// [22] Kwon YB, Kim BS, Hancock GJ. Compression tests of high strength cold-formed
steel channels with buckling interaction. J Constr Steel Res 2009;65:278–89.
doi.org/10.1016/j.istruc.2020.08.022. [23] Zhao X, Ren C, Qin R. An experimental investigation into perforated and non-per­
forated steel storage rack uprights. Thin-Walled Structures 2017;112:159–72.
References [24] Ren C, Wang BH, Zhao XZ. Numerical predictions of distortional-global buckling
interaction of perforated rack uprights in compression. Thin-Walled Struct
2019;136:292–301.
[1] RMI, Specification for the design, testing and utilization of industrial steel storage [25] Ren C, Zhao X, Chen Y. Buckling behaviour of partially restrained cold-formed steel
racks, ANSI MH16.1-2012', Rack Manufacturers Institute, Technical report, 2012. zed purlins subjected to transverse distributed uplift loading. Eng Struct
[2] EN 15512, Steel static storage systems, European Technical Committee CEN/TC 2016;114:14–24.
344, European Specifications, 2009. [26] Gunalan S, Heva YB, Mahendran M. Local buckling studies of cold-formed steel
[3] AS/NZS 4084, Steel storage racking, Standards Australia/Standards New Zealand, compression members at elevated temperatures. J Constr Steel Res
Sydney, Australia, 2012. 2015;108:31–45.
[4] Feng M, Wang YC, Davies JM. A numerical imperfection sensitivity study of cold- [27] Shahbazian A, Wang YC. Calculating the global buckling resistance of thin-walled
formed thin-walled tubular steel columns at uniform elevated temperatures. Thin- steel members with uniform and non-uniform elevated temperatures under axial
Walled Struct 2004;42(4):533–55. compression. Thin-Walled Struct 2011;49(11):1415–28.
[5] Kankanamge ND, Mahendran M. Behaviour and design of cold-formed steel beams [28] Shahbazian A, Wang YC. Direct Strength Method for calculating distortional
subject to lateral–torsional buckling at elevated temperatures. Thin-Walled Struct buckling capacity of cold-formed thin-walled steel columns with uniform and non-
2012;61:213–28. uniform elevated temperatures. Thin-Walled Struct 2012;53:188–99.
[6] EN.1993-1-2.2005: “General rules—Structural fire design”. Brussels, Belgium: [29] Young B, Kwon B, Hancock GJ. Tests of cold-formed channels with local and dis­
European Committee for Standardization (CEN). tortional buckling. J Struct Eng 1992;117(7):1786–803.
[7] EN.1993-1-3. 2006: “General rules — Supplementary rules for cold-formed mem­ [30] Shahbazian A, Wang YC. Application of the Direct Strength Method to local buck­
bers and sheeting”. Brussels, Belgium: European Committee for Standardization ling resistance of thin-walled steel members with non-uniform elevated tempera­
(CEN). tures under axial compression. Thin-Walled Struct 2011;49:1573–83.
[8] Chen J, Young B. Cold-formed steel lipped channel columns at elevated tempera­ [31] EN.1991-1-2. 2002: “General actions—Actions on structures exposed to fire”.
tures. Eng Struct 2007;29:2445–56. Brussels, Belgium: European Committee for Standardization (CEN).
[9] Craveiro HD, Rodrigues JPC, Laím L. Cold-formed steel columns made with open [32] Chen J, Young B. Experimental investigation of cold-formed steel material at ele­
cross-sections subjected to fire. Thin-Walled Struct 2014;85:1–14. vated temperatures. Thin-Walled Struct 2007;45:96–110.
[10] Tan KH, Yuan WF. Buckling of elastically restrained steel columns under long­ [33] Kankanamge ND, Mahendran M. Mechanical properties of cold-formed steels at
itudinal non-uniform temperature distribution. J Constr Steel Res 2008;64:51–61. elevated temperatures. Thin-Walled Struct 2011;49(1):26–44.
[11] Byström A, Sjöström J, Wickström U, Lange D, Veljkovic M. Large scale test on a [34] Schafer BW, Peköz T. Direct strength prediction of cold-formed steel members using
steel column exposed to localized fire. J Struct Fire Eng 2014;5(2):147–60. numerical elastic buckling solutions. Proceedings of the 14th International
[12] Ferraz G, Santiago A, Rodrigues JP, Barata P. Thermal analysis of hollow steel Specialty Conference on Cold-Formed Steel Structures, St. Louis, University of
columns exposed to localised fires. Fire Technol 2016;52:663–81. Missouri-Rolla, MO, USA. 69-76, 1998.
[13] Zhang C, Gross JL, McAllister TP. Lateral torsional buckling of steel W-beams [35] Schafer BW. Local, distortional, and Euler buckling of thin-walled columns. J Struct
subjected to localized fires. J Constr Steel Res 2013;88:330–8. Eng ASCE 2002;128(3):289–99.
[14] Ren C, Dai L, Huang Y, He W. Experimental investigation into post-fire mechanical [36] AISI. North American Specification for the design of cold-formed steel structural
properties of Q235 cold-formed steel, Thin-Walled Structures. (In press). members, North American Cold-Formed Steel Specification, American Iron and
[15] Feng M, Wang YC, Davies JM. Axial strength of cold-formed thin-walled steel Steel Institute, AISI S100−16, Washington, D.C, 2016.
channels under non-uniform temperatures in fire. Fire Saf J 2003;38:679–707. [37] AS/NZS. Cold-formed steel structures, Australian/New Zealand Standard, AS/NZS
[16] Zhang C, Li G, Usmani A. Simulating the behavior of restrained steel beams to flame 4600:2005, Standards Australia, Sydney, Australia, 2005.
impingement from localized-fires. J Constr Steel Res 2013;83:156–65.
[17] Zhang C, Gross JL, McAllister TP, Li GQ. Behavior of unrestrained and restrained

2095