See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/334389641

Numerical modelling and design of stainless steel square and rectangular

hollow section column in fire

Conference Paper · July 2019

CITATIONS READS

0 235

2 authors, including:

Asif Mohammed
University of the West of England, Bristol
9 PUBLICATIONS   29 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

HILONG High Strength Long Span Structures View project

Improve robustness of steel end plate connections considering the bolt performance. View project

All content following this page was uploaded by Asif Mohammed on 11 July 2019.

The user has requested enhancement of the downloaded file.

 9th International Conference on
Steel and Aluminium Structures (ICSAS19)
Bradford, UK, 3-5, July, 2019

NUMERICAL MODELLING AND DESIGN OF STAINLESS STEEL

SQUARE AND RECTANGULAR HOLLOW SECTION COLUMNS IN
FIRE

A. MOHAMMED a and S. AFSHAN a

a
Department of Civil and Environmental Engineering, Brunel University London, London, UK
Emails: Asif.Mohammed@brunel.ac.uk, Sheida.Afshan@brunel.ac.uk

Keywords: Axial restraint; Column; Fire; Numerical modelling; Rotational restraint; Stainless steel.

Abstract. This paper presents a numerical study on the buckling behaviour and design of cold-
formed stainless steel square and rectangular hollow section columns in fire. Firstly, finite element
models were developed and validated against a total of nine austenitic and three ferritic stainless
steel fire column tests reported in the literature, which were shown to be capable of safely replicating
the non-linear, large deflection response of the stainless steel columns in fire with a high degree of
predictive accuracy. Following this, a comprehensive numerical parametric study was carried out
where the varied parameters included (1) stainless steel grade (austenitic, duplex and ferritic), (2)
cross-section aspect ratio (h/b = 1 and 1.5), (3) axis of buckling (major and minor) and (4) elevated
temperature member slenderness ( λ θ = 0.1-2.0). The FE generated flexural buckling data were used
to carry out an assessment of the accuracy of the design methods provided in EN 1993-1-2 and the
Design Manual for Stainless Steel Structures. New buckling curves for cold-formed stainless steel
square and rectangular hollow section columns in fire were proposed on the basis of the FE results.
The suitability of the new proposals was confirmed by means of the reliability analysis.

1 INTRODUCTION
The corrosion resistance and durability of stainless steel are well known, offering the
potential for more sustainable construction with increased structural design lives. Stainless
steel is most commonly used in structures in the offshore and onshore industrial sector (e.g.
Oil and Gas, Petrochemical, Pharmaceutical, nuclear, etc.) where fire is a significant hazard.
Other applications where stainless steel has to demonstrate fire resistance are in light interior
structures (e.g. escape routes in airports and office buildings), safety critical structures (e.g.
locations exposed to terrorist attack) and fastening systems. Stainless steel in tunnels, where
the presence of the corrosive environment coupled with the high maintenance costs
necessitates the use of highly durable materials, is also an important environment for fire
resistance. Hence, development of comprehensive and economic guidance on the design of
stainless steel structures in fire is of paramount importance if the use of the material in fire
safety critical applications is to increase. A number of recent studies have indicated that the
fire resistance of stainless steel structural members is greater than that of equivalent carbon
steel members [1]. With superior strength and stiffness retention at elevated temperatures, in
comparison with carbon steels, stainless steels potentially offer substantial improvements in
performance.
The fire resistant design of structural carbon steels is covered in EN 1993-1-2 [2]. EN
1993-1-4 [3] which is the part of Eurocode 3 that provides supplementary design rules for
 A. MOHAMMED and S. AFSHAN

stainless steel structures refers to EN 1993-1-2 [2] for their fire design, where the same
guidelines as those for carbon steels, but in conjunction with the stiffness and strength
reduction factors for stainless steel material, are also adopted for stainless steels. However, the
stress-strain behaviour of stainless steels is of different form to that of carbon steels. Whereas
carbon steels typically exhibit linear elastic behaviour up to the yield strength and a plateau
before strain hardening, stainless steels possess a more rounded response with no well-defined
yield strength. This results in a difference in the structural behaviour between carbon steels
and stainless steels, and consequently different design rules will be needed in certain cases.
This paper focuses on the structural performance and design of stainless steel columns in
fire. A comprehensive numerical modelling study has been conducted to generate structural
performance data for stainless steel columns subjected to fire. The development of the finite
element (FE) models, including their validation against existing test results, as well as the
results of the parametric study, which was performed subsequently, are presented. The
generated FE results are compared with the current design rules in EN 1993-1-2 [2] and the
Design Manual for Structural Stainless Steel [4]. Finally, amendments to the current design
procedures, in line with the obtained results, are proposed, and reliability analysis are carried
out.

2 DEVELOPMENT AND VALIDATION OF NUMERICAL MODELS

A numerical modelling study was performed to investigate the flexural buckling response
of stainless steel square and rectangular hollow section columns in fire. The general purpose
finite element (FE) analysis package ABAQUS [5] was used for conducting the numerical
modelling simulations. The developed finite element models were initially validated against
the fire column tests reported in the literature [6-8], and subsequently used to perform
parametric studies to assess the effect of the variation of key parameters and to generate
further structural performance data. The main features of the developed finite element models
and their validation are presented herein while the conducted parametric studies are presented
in Section 3.

2.1 Summary of literature tests

The results of fire tests on square hollow section (SHS) and rectangular hollow section
(RHS) columns including three grade EN 1.4301 columns reported by Gardner and Baddoo
[6], six grade EN 1.4301 columns reported by Ala Outinen and Oksanen [7], and three grade
EN 1.4003 columns reported by Tondini et al. [8] were utilised for the validation of the
numerical models. A summary of these fire tests is provided in Table 1. The RHS columns
tested in [6] were formed by welding two press-braked channel sections tip-to-tip along the
length of the column. The manufacturing process of the SHS and RHS columns tested in [7]
and [8] involved cold-rolling into a circular tube followed by sizing into the final cross-
section geometry. All fire tests were performed anisothermally, whereby the load was applied
at ambient temperature and was maintained at a constant level while the temperature was
increased until failure at θcrit. The critical temperature θcrit reported in Table 1 refers to the
specimen temperature for the austenitic columns and the furnace temperature for the ferritic
columns. Each model involved a sequentially coupled thermal-stress analysis with three types
of numerical analyses – (1) a linear elastic buckling analysis to determine the buckling mode
shapes, (2) a heat transfer analysis to simulate the temperature development in the columns
and (3) a geometrically and materially nonlinear stress analysis, which incorporated the
buckling mode shapes as initial geometric imperfection from (1) and the temperature
development field from (2).

2
 A. MOHAMMED and S. AFSHAN

Table 1: Summary of literature column tests used for validation of FE models.

Boundary Load θcrit
Specimen reference Reference Grade
condition (kN) (oC)
RHS 150×100×6 268 801
RHS 150×75×6 [6] EN 1.4301 Fixed 140 883
RHS 100×75×6 156 806
SHS 40×40×4-T1 45 872
SHS 40×40×4-T2 129 579
SHS 40×40×4-T3 114 649
[7] EN 1.4301 Pinned
SHS 40×40×4-T4 95 710
SHS 40×40×4-T5 55 832
SHS 40×40×4-T7 75 766
SHS 80×80×3-3000 72 709 (1)
SHS 80×80×3-2500 [8] EN 1.4003 Fixed 78 708 (1)
RHS 120×80×3-2500 100 705 (1)
(1)
Critical furnace temperature

2.2 Heat transfer model

The measured specimen time-temperature data were utilised in the numerical models of the
austenitic stainless steel columns presented in Table 1. The specimen time-temperature was
not measured for the ferritic stainless steel column tests; only the furnace temperature was
measured. Therefore, for these columns, heat transfer analyses were first conducted to obtain
the development of specimen temperature with the fire exposure time, which was required as
input for the nonlinear stress analysis part of the modelling. The thermal analysis of a
structural member may be separated in two parts: the heat transfer from the fire to the exposed
surfaces of the structural member through combined convection and radiation heat transfer
mechanisms, and conductive heat transfer mechanisms within the structural member itself;
these two parts were carefully simulated in the FE models. Radiation was modelled as surface
radiation by means of the *SRADIATE command in ABAQUS with the emissivity coefficient
taken as 0.4 from EN 1993-1-2 [2]. Convection was modelled as a film condition using the
*SFILM command in ABAQUS with the convective heat transfer coefficient taken as 25
W/m²K as specified in EN 1993-1-2 [2]. Other thermal properties including specific heat,
thermal conductivity and thermal expansion provided in the Design Manual for Stainless Steel
Structures (DMSS) [4] were adopted.

2.3 Stress analysis model

The geometrically and materially nonlinear stress analysis was performed in two steps in
order to simulate the anisothermal loading conditions of the tested columns. In the first step, a
static load, equal to the test loads, was applied to the columns at ambient temperature. This
applied load was then kept constant in the second step, during which the increase in specimen
temperature over the fire exposure time was applied from the heat transfer models. For the
austenitic stainless steel columns the measured specimen time-temperature curves from the
tests were directly imported into the FE models.

3
 A. MOHAMMED and S. AFSHAN

2.4 General modelling assumptions

Shell elements were adopted to simulate the stainless steel tubular hollow section columns,
as is customary for the modelling of thin-walled structures [9–11]. The general-purpose four-
noded three-dimensional S4R [5] shell element with reduced integration and hourglass control
was used for the stress analysis models of the SHS/RHS stainless steel columns examined
herein. For the thermal models, the DS4 [5] element, which is compatible with S4R element,
were used. An element mesh size of minimum ten elements across each plate was adopted,
which from a mesh sensitivity analysis was found to give the best combination of accuracy
and computational efficiency. The pinned and fixed end test boundary conditions were
replicated by restraining suitable displacement and the rotation degrees of freedom at the
column ends. The measured geometric dimensions of the tested columns were utilised in the
FE models to replicate the corresponding test specimen.

2.5 Material modelling

In Gardner and Baddoo [6], isothermal material tests at temperatures 20-1000 °C were
performed on the sheet material used to fabricate the press-braked channel sections welded
tip-to-tip to form the hollow sections. Ala-Outinen and Oksanen [7] carried out anisothermal
tensile tests on coupons cut from the flat faces of cold-rolled hollow sections and the results
were converted into stress-strain curves at temperatures 20-900 °C. For these column tests, the
measured elevated temperature stress-strain curves were used to describe the material
behaviour of the flat portions of the sections in the developed FE models. For the ferritic
column tests, the two-stage Ramberg-Osgood material model for elevated temperatures
provided in DMSS [4] given by Equations (1) and (2) was employed to construct full range
stress-strain curves at temperatures 20-800 °C. The reduction factors pertaining to grade EN
1.4003 provided in DMSS [4] and the measured room temperature material properties were
used. These were used to describe the material behaviour of the flat portions of the sections in
the developed FE models.
n
   
     0.002    (1)
E  f 0.2, 

m,2
  f 0.2,  f f    f 
    0.02   0.2,  2, 0.2,    0.2,    0.2, (2)
E 0.2,  E 0.2,   f 2,  f 0.2, 


where σθ and εθ are the stress and strain, respectively at temperature θ, f0.2,θ is the 0.2% proof
stress at temperature θ, Eθ is the Young’s modulus at temperature θ, E0.2,θ is the tangent
modulus at f0.2,θ, ε0.2,θ is the total strain corresponding to f0.2,θ, f2,θ is the stress at 2% total
strain at temperature θ and nθ and mθ,2 are the Ramberg-Osgood model parameters at
temperature θ.
The strength enhancements in the corner regions of the SHS and RHS specimens were also
incorporated in the FE models. For the ferritic column tests, corner material properties were
measured at room temperature. For the austenitic stainless steel tests, where no measured
corner material properties were reported, the predictive equations from Cruise and Gardner
[12] for the strength enhancement in the corner regions of cold-rolled and press-braked
stainless steel sections were used to determine the room temperature 0.2% proof stress of the
corner regions. Equations (1) and (2) together with the reduction factors for EN 1.4301 and
EN 1.4003 from [4] for material with enhanced cold-form strength were used to obtain a

4
 A. MOHAMMED and S. AFSHAN

continuous prediction of the stress-strain response of the cold-worked corner regions at

elevated temperatures.
In ABAQUS, the material behaviour was modelled as elastic-plastic with a Von Mises
yield criterion and isotropic hardening. ABAQUS requires that for shell elements the input
material stress-strain curves are in the form of multi-linear true stress σtrue and logarithmic
plastic strain ε ln responses; these were obtained from the constructed engineering stress-strain
pl

relationships, as defined in Equations (3) and (4), respectively, where σnom and εnom are
engineering stress and strain, respectively and E is the Young’s modulus and incorporated into
the FE models.
true  nom (1  nom ) (3)

 lnpl  ln(1   nom )  true (4)
E

2.6 Geometric imperfections and residual stresses

The lowest local and global buckling mode shapes, determined by means of the elastic
eigenvalue buckling analysis, were assumed for the respective imperfection patterns along the
member length and incorporated into the FE models. For the ferritic stainless steel columns,
the measured global imperfection amplitudes were used. For the austenitic stainless steel
columns, in the absence of measured values, the global imperfection amplitude was taken as
L/2000, where L is the column total length [9,13]. The local imperfection amplitudes
employed for scaling the local buckling mode shapes were taken as the values predicted by the
Dawson and Walker model, as adapted for stainless steel [14], given by Equation (5), where t is
the plate thickness, f0.2 is the material 0.2% proof stress and fcr is the plate elastic buckling stress.
Bending residual stresses were not explicitly included in the FE models since their effect is
inherently present in the measured material properties used, which were obtained from
coupons cut from the finished tubes [15].

f 0.2
o  0.023t (5)
f cr

2.7 Validation results

A total of twelve stainless steel columns were modelled using the sequentially coupled
thermal-stress analysis procedure described. The fire performance criteria set out in EN 1363-
1 [16] for vertically loaded members were used to determine the critical failure temperature of
the columns. It states that a column is deemed to have failed when both the vertical
contraction and the rate of vertical contraction have exceeded L/100 mm and 3L/1000
mm/min, respectively, where L is initial column height in mm.
Figures 1 and 2 compare the test axial deformation versus temperature results with the FE
results for the SHS 80×80×3-2500 and RHS 150×75×6 column, respectively. The axial
deformation versus temperature relationship of the axially loaded compression members may
be described as follows: the column initially shortens as the load is applied at room
temperature; as the temperature is increased, the column then starts to expand. At high steel
temperatures, the rate of increase in the column axial deformation decreases as the column
stiffness reduces and the mechanical shortening, related partially to axial deformation and
partially to out-of-plane deflection as the column buckles, becomes important. Finally, the
mechanical shortening overtakes the thermal expansion of the column, the axial deformation
changes direction and the column contracts until it can no longer sustain the applied load. The

5
 A. MOHAMMED and S. AFSHAN

column mechanical shortening is related to the elevated temperature tangent stiffness, which
reduces rapidly, making the final stage abrupt. As shown in Figures 1 and 2, the FE models
developed herein were capable of accurately replicating the temperature-deflection response
characteristics of tested columns.
A summary of the comparisons between the test and FE results is provided in Table 2. For
the austenitic stainless steel columns, the FE models give a mean FE/test critical temperature
of 0.90 and a coefficient of variation of 0.03, and provide safe-side predictions of the fire
resistance of the test column specimens. For the ferritic stainless steel columns, the FE and
test results are in very good agreement with a mean FE/test critical temperature of 1.00 and a
coefficient of variation of 0.02. Comparison between the test and FE failure modes is depicted
in Figure 3, where comparable results from both are observed. From the comparison of the
test and FE results, it is concluded that the described FE models are capable of safely
replicating the nonlinear, large deflection response of the stainless steel columns in fire.

Table 2: Comparison of critical temperatures between test and FE results.

Critical temperature (°C)
Specimen reference
Test FE FE/Test
RHS 150×100×6 801 757 0.91
RHS 150×75×6 883 814 0.92
RHS 100×75×6 806 744 0.92
SHS 40×40×4-T1 872 750 0.86
SHS 40×40×4-T2 579 502 0.87
SHS 40×40×4-T3 649 608 0.94
SHS 40×40×4-T4 710 646 0.91
SHS 40×40×4-T5 832 722 0.87
SHS 40×40×4-T7 766 681 0.89
SHS 80×80×3-3000 709 726 1.02
SHS 80×80×3-2500 708 718 1.02
RHS 120×80×3-2500 705 709 1.01

40 30
Test
30 FE (Axial displ)
Axial displacement velocity (mm/min)

FE (Axial displ velocity) 20

20
Axial displacement (mm)

10
10

0 0
0 100 200 300 400 500 600 700 800
-10
Axial displacement velocity limit -10
-20
-20
-30 Axial displacement limit

-40 -30
Temperature C
Figure 1: Vertical displacement versus temperature of ferritic SHS 80×80×3-2500 specimen.

6
 A. MOHAMMED and S. AFSHAN

60 30
Test
FE (Axial displ)

Axial displacement velocity (mm/min)

40 FE (Axial displ velocity) 20

Axial displacement (mm)

20 10

0 0
0 100 200 300 400 500 600 700 800 900 1000

-20 Axial displacement velocity limit -10

-40 Axial displacement limit -20

-60 -30
Temperature C

Figure 2: Vertical displacement versus temperature of austenitic RHS 150×75×6 specimen.

Figure 3: Test and FE failure modes for SHS 80×80×3-3000 column tested in [8].

7
 A. MOHAMMED and S. AFSHAN

3 PARAMETRIC STUDY
In this section, a series of parametric studies is presented to examine the flexural behaviour
of stainless steel square and rectangular hollow section columns of different grades at elevated
temperatures. For modelling convenience, all parametric study models presented herein were
performed isothermally, where the material properties for a given temperature θ were
incorporated into the FE models, akin to applying a uniform temperature θ, and the applied
load was increased until failure. A static Riks analysis procedure was used to trace the load-
deformation response of each of the modelled columns and to determine their failure load.
This approach was deemed acceptable, since the influence of time dependent effects, e.g.
creep, was not included in the developed FE models, and therefore both the isothermal and
anisothermal modelling approaches would yield very similar results.
Since the stress-strain response and the elevated temperature properties of stainless steel
vary between the different grades, parametric studies were performed for the three most
common stainless steel grades used in the construction industry – austenitic, duplex and
ferritic. For each stainless steel grade, the varied parameters were: the cross-section aspect
ratio (h/b), the axis of buckling (major/minor) and elevated temperature member slenderness
λ θ . Table 3 provides a summary of the examined parameters. In total 500 cold-formed
stainless steel SHS/RHS columns were modelled. All columns were modelled as pin-ended at
both ends. All cross-sections were classified as Class 1 according to the EN 1993-1-2 [2]
cross-section classification limits. The room temperature plate slenderness of the modelled
cross-sections, taken as the slenderness of the most slender plate element λ p , given by
Equation (6) are also included in Table 3. In Equation (6), fcr is the elastic critical buckling
stress of the plate element, f2 is the stress at 2% total strain, b and t are the flat plate width and
thickness, respectively, E is the Young’s modulus, kσ is the plate buckling coefficient, taken
as 4.0 for internal plate elements in compression and ν is the Poisson's ratio.

0.5
 b   f   12(1-ν ) 
0.5 2
f2
λp = f cr
=   2   2  (6)
 t  E   π k σ 

Table 3: Summary of the parametric study variables.

Grade Section h/b Axis of buckling λp λθ
SHS 100×100×10 1.0 - 0.33 0.1-2.0
Austenitic
RHS 150×100×14 1.5 Major and Minor 0.34 0.1-2.0
SHS 100×100×9 1.0 - 0.38 0.1-2.0
Duplex
RHS 150×100×14 1.5 Major and Minor 0.36 0.1-2.0
SHS 100×100×9 1.0 - 0.32 0.1-2.0
Ferritic
RHS 150×100×11 1.5 Major and Minor 0.39 0.1-2.0

The same modelling assumptions as explained in Section 2 were employed for the
parametric study models with the input parameters taken as those described hereafter. The
room temperature material properties recommended by Afshan et al. [17] for cold-formed
austenitic, duplex and ferritic stainless steel tubes together with the strength and stiffness
reduction factors at elevated temperatures pertaining to Austenitic I, Duplex II and Ferritic II
provided in Table 8.1 of the Design Manual for Stainless Steel Structures [4] were employed.
The two-stage Ramberg-Osgood model, Equations (1) and (2), was then used to develop full-
range stress-strain curves at discrete temperatures of 20-800 °C. The values for nθ were taken

8
 A. MOHAMMED and S. AFSHAN

as the room temperature values for n provided in [4] and values for m2,θ were determined
using Equation (7) as recommended in Clause 8.5 of the Design Manual for Stainless Steel
Structures [4].
f
m2, =1+2.8 0.2, (7)
f u,
The global imperfection amplitude was taken as L/1000, where L is the column length, in
accordance with the permitted out-of-straightness tolerance in EN 1090-2 [18]. The local
imperfection amplitude was taken as b/200, where b is the section width, as recommended in
Annex C of EN 1993-1-5 [19]. Shell element S4R with element size equal to the cross-section
thickness was used to discretise the flat portions of the modelled SHS/RHS cross-sections,
while the corner regions were assigned a finer mesh of four elements to accurately represent
the curved geometry.

4 CURRENT DESIGN METHODS FOR STAINLESS STEEL COLUMNS

4.1 EN 1993-1-2 method

The structural fire design of stainless steel structures is covered in EN 1993-1-2 [2] with
similar treatments as carbon steel structures. Firstly, since in fire situations higher strains than
at room temperature are acceptable, EN 1993-1-2 [2] uses the elevated temperature stress at
2% total stress f2,θ = k2,θfy for the design of columns with Class 1, 2 and 3 cross-sections and
the 0.2% proof stress f0.2,θ = k0.2,θfy for Class 4 cross-sections, where fy is the design yield
strength (taken as the stress at 0.2% proof stress) at room temperature and k2,θ and k0.2,θ are the
reduction factors for f2,θ and f0.2,θ, respectively. These design strength parameters are also
recommended in EN 1993-1-2 [2] for the fire design of stainless steel columns. The column
buckling curve in EN 1993-1-2 [2] is of the same general form as the room temperature
column buckling curves for both carbon steel and stainless steel, with the exception of
exhibiting no plateau i.e. λ0  0 , including a yield strength dependent imperfection factor α =
0.65√235/fy and introducing the elevated temperature member non-dimensional slenderness
λ θ , defined in Equations (8) and (9), where λ is the column slenderness at room temperature
and kE,θ is the reduction factor for Young’s modulus Eθ at temperature θ and all other
parameters are as previously defined. The non-dimensional buckling reduction factor χfi given
by Equation (10) is also recommended for stainless steel columns of all cross-section shapes.

0.5
k 
λ θ = λ  2,θ  for Class 1, 2 and 3 cross-sections (8)
k
 E,θ 
0.5
k 
λ θ =λ  0.2,θ  for Class 4 cross-sections (9)
 k
 E,θ 
1
with φθ =0.5 1+αλθ +λθ 
2
χ fi =
2   (10)
φθ + φ 2θ -λ θ

Finally, the design fire buckling resistance of a compression member Nb,fi,t,Rd at time t and
uniform temperature θ is obtained from Equations (11) and (12), where A is the gross cross-
sectional area, Aeff is the effective cross-sectional area, γM,fi is the partial resistance factor for
member resistance, taken as 1.0 in [2] and all other symbols are as previously defined.

9
 A. MOHAMMED and S. AFSHAN

χ fi Ak 2,θ f y
N b,fi,t,Rd = for Class 1, 2 and 3 cross-sections (11)
γ M,fi
χ fi A eff k 0.2,θ f y
N b,fi,t,Rd = for Class 4 cross-sections (12)
γ M,fi

4.2 Lopes et al. method

Lopes et al. [20] conducted a numerical study on axially loaded stainless steel welded I-
section columns in fire and proposed a modified version of the EN 1993-1-2 [2] buckling
curve. In the proposed buckling curve, (1) the parameter β was introduced in the non-
dimensional buckling reduction factor χfi and φθ formulations, Equation (13) and (2) the
imperfection factor α was defined as a function of temperature, as given by Equations (14) and
(15), resulting in different buckling curves for different temperatures. For welded I-section
columns, β values equal to 1.0 and 1.5 for major and minor axis buckling, respectively for all
stainless steel grades and α values equal to 1.3 for EN 1.4301, 1.4401, 1.4404, 1.4571 and
1.4003 and 0.9 for EN 1.4462 grades were proposed [20]. Using the proposed fire buckling
curves and the same definition of λ θ given in Equations (8) and (9), Nb,fi,t,Rd may similarly be
obtained from Equations (11)-(12).

1
 1.0 with φθ =0.5 1+αλθ +βλθ 
2
χ fi =
2   (13)
φθ + φ 2θ -βλ θ

235 E k E,θ
α=η for Class 1, 2 and 3 cross-sections (14)
f y 210000 k 2,θ

235 E k E,θ
α=η for Class 4 cross-sections (15)
f y 210000 k 0.2,θ

4.3 Design Manual for Structural Stainless Steel method

The design method provided in the Design Manual for Structural Stainless Steel [4] uses
the same buckling curves as for room temperature, with plateau length λ 0 and imperfection
factor α as those recommended in Table 6.1 in [4], for elevated temperature design. The
design fire buckling resistance Nb,fi,t,Rd is obtained from Equation (16) and (17), where A is the
gross cross-sectional area, Aeff is the effective cross-sectional area, χfi is as defined in
Equation (18), where λ θ is defined in Equation (19). For all cross-section classes, the 0.2%
proof stress (k0.2,θfy) is used to determine the flexural buckling resistance of stainless steel
column in fire, where k0.2,θ is the reduction factor for the 0.2% proof strength and fy is the
design yield strength (taken as the stress at 0.2% proof stress) at room temperature.

χ fi Ak 0.2,θ f y
N b,fi,t,Rd = for Class 1, 2 and 3 cross-sections (16)
γ M,fi

χ fi A eff k 0.2,θ f y
N b,fi,t,Rd = for Class 4 cross-sections (17)
γ M,fi

10
 A. MOHAMMED and S. AFSHAN

1
 
 1.0 with φθ =0.5 1+α λθ -λ0 +λθ 
2
χ fi =
2   (18)
φθ + φ θ  λ θ
2

0.5
k 
λ θ =λ  0.2,θ  for all Classes of cross-sections (19)
 k
 E,θ 

5 ANALYSIS OF RESULTS AND DESIGN GUIDANCE PROPOSAL

A comparison of the parametric results obtained in Section 3 with the current design
methods for stainless steel columns in fire presented in Section 4 is provided in this section.
Figure 4 shows the FE ultimate loads normalised by the cross-section elevated temperature
yield loads, defined as the product of the cross-sectional area A and the cross-section elevated
temperature yield strength k2,θfy, plotted against the elevated temperature member slenderness
λ θ , determined using Equation (8), for the cold-formed austenitic, duplex and ferritic
SHS/RHS members with the EN 1993-1-2 [2] buckling curve also depicted. The yield
strength fy at room temperature is taken as the weighted average (by area) 0.2% proof stress
i.e. normalising out the strength enhancements in the corner regions of the cold-formed
sections. Figure 5 compares the FE results with the Design Manual for Structural Stainless
Steel [4] buckling curve, with plateau length λ0  0.3 for the austenitic and duplex grades and
λ0  0.2 for ferritic grade and α = 0.49 for all grades. In data presented in Figure 5, the cross-
section yield strength was taken as k0.2,θfy and the elevated temperature member slenderness
λ θ was determined using Equation (19).
Table 4 presents a summary of the numerical comparisons in terms of the mean, coefficient
of variation (COV), minimum and maximum values of the ratios between the ultimate loads
obtained from the FE models (Nu,FE) and the predicted capacities obtained from the design
methods (Nu,pred), and N is the number of data considered.

Table 4: Comparison between the FE and predicted resistances.

Material Nu,FE/Nu,pred EN 1993-1-2 [2] DMSS [4] Proposed
No. 156 156 156
Mean 0.897 1.049 1.114
Austenitic COV 0.118 0.138 0.062
Max 1.157 1.574 1.256
Min 0.682 0.844 0.965
No. 160 160 160
Mean 0.915 1.036 1.047
Duplex COV 0.136 0.127 0.060
Max 1.212 1.534 1.275
Min 0.649 0.806 0.887
No. 188 188 188
Mean 1.029 1.043 1.069
Ferritic COV 0.091 0.071 0.056
Max 1.166 1.231 1.177
Min 0.801 0.888 0.925

11
 A. MOHAMMED and S. AFSHAN

1.4 1.8
θ = 200 C θ = 200 C
1.2 1.6
θ = 400 C θ = 400 C
1.4
1.0 θ = 600 C θ = 600 C
1.2
θ = 800 C θ = 800 C
Nu,θ/Ak2,θfy

Nu,θ/Ak0.2,θfy
0.8 1.0
EN 1993-1-2 DMSS
0.6 0.8
0.6
0.4
0.4
0.2
0.2
= 0.3 α = 0.49
0.0 0.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
λθ =λ k2,θ kE,θ
λθ = λ k0.2,θ /kE,θ

(a) Austenitic (a) Austenitic

1.4 1.8
θ = 200 C 1.6 θ = 200 C
1.2
θ = 400 C θ = 400 C
1.4
1.0 θ = 600 C θ = 600 C
1.2
θ = 800 C θ = 800 C
Nu,θ/Ak0.2,θfy
Nu,θ/Ak2,θfy

0.8 1.0
EN 1993-1-2 DMSS
0.6 0.8
0.6
0.4
0.4
0.2
0.2 = 0.3 α = 0.49
0.0 0.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
λθ =λ k2,θ kE,θ λθ = λ k0.2,θ /kE,θ

(b) Duplex (b) Duplex

1.4 1.8
θ = 200 C θ = 200 C
1.2 1.6
θ = 400 C θ = 400 C
1.4
1.0 θ = 600 C θ = 600 C
θ = 800 C 1.2
Nu,θ/Ak2,θfy

θ = 800 C
Nu,θ/Ak0.2,θfy

0.8 1.0
EN 1993-1-2 DMSS
0.6 0.8
0.6
0.4
0.4
0.2 0.2 = 0.2 α = 0.49
0.0 0.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0

λθ =λ k2,θ kE,θ λθ = λ k0.2,θ /kE,θ

(c) Ferritic (c) Ferritic

Figure 4: Comparison of FE results with the EN Figure 5: Comparison of FE results with the
1993-1-2 [1] buckling curve (a) austenitic, (b) Design Manual for Structural Stainless Steel
duplex and (c) ferritic SHS/RHS. buckling curves [2] (a) austenitic, (b) duplex and
(c) ferritic SHS/RHS.

The EN 1993-1-2 [2] buckling curve which is common for all temperatures generally over-
predicts the buckling resistance of the austenitic and duplex stainless steel columns for all
temperatures, though it provides a better fit to the ferritic stainless steel columns for

12
 A. MOHAMMED and S. AFSHAN

temperatures below 800 °C. The Design Manual for Structural Stainless Steel provisions,
which use the elevated temperature 0.2% proof stress rather than the elevated temperature
stress at 2% total strain, significantly under-predicts the resistance of the stockier columns
with low elevated temperature member slenderness values as it limits the cross-section
resistance to the squash load based on the 0.2% proof stress. Note that no comparisons have
been made with the Lopes et al. [20] proposed method as it was developed for welded I-
section columns, with distinctly different buckling performance from that of cold-formed box
sections. Buckling curves of the same form as the Lopes et al. [20] formulation developed for
welded I-section columns in fire were fitted to the normalised FE data for the cold-formed
SHS/RHS stainless steel columns generated herein, and are shown in Figure 6. The β and η
parameters were calibrated against the FE data for austenitic, duplex and ferritic stainless steel
columns generated in Section 3. The proposed β and η parameters for the fire design of
stainless steel flexural members is presented in Table 5. The FE results were compared with
the predicted resistances using the proposed new β and η parameters presented in Table 5 and
the numerical comparisons are reported in Table 4. From the comparison results presented in
Table 4, it is shown that the proposed buckling curves provide an improved representation of
the buckling resistance of cold-formed stainless steel SHS/RHS columns in fire and allow a
more accurate and less scattered prediction of their flexural buckling capacity at elevated
temperature.
1.4 1.4
θ = 200 C θ = 200 C
1.2 θ = 400 C 1.2 θ = 400 C
θ = 600 C θ = 600 C
1.0 θ = 800 C 1.0 θ = 800 C
Nu,θ/Ak2,θfy

Nu,θ/Ak2,θfy

Proposed 200 °C Proposed 200 °C

0.8 0.8
Proposed 600 °C Proposed 600 °C
0.6 Proposed 800 °C 0.6 Proposed 800 °C

0.4 0.4

0.2 0.2

0.0 0.0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
λθ =λ k2,θ kE,θ λθ =λ k2,θ kE,θ

(a) Austenitic (b) Duplex

1.4
θ = 200 C
1.2 θ = 400 C
θ = 600 C
1.0 θ = 800 C
Proposed 200 °C
Nu,θ/Ak2,θfy

0.8
Proposed 600 °C
Proposed 800 °C
0.6

0.4

0.2

0.0
0.0 0.5 1.0 1.5 2.0

λθ =λ k2,θ kE,θ

(c) Ferritic
Figure 6: Comparison of FE results with the proposed method (a) austenitic, (b) duplex and (c)
ferritic

13
 A. MOHAMMED and S. AFSHAN

Table 5: Proposed β and η parameters for cold-formed SHS/RHS columns.

Material Section β η
Austenitic SHS/RHS 0.8 1.5
Duplex SHS/RHS 0.8 1.1
Ferritic SHS/RHS 1.0 0.6

6 RELIABILITY ANALYSIS
Safety analysis in accordance with the method recommended by Kruppa [21] were
performed to assess the reliability of the existing and the proposed design methods to predict
the flexural buckling resistance of cold-formed austenitic, duplex and ferritic SHS/RHS
stainless steel columns in fire. The method by Kruppa [21] sets out three distinct reliability
criteria to compare the theoretical resistance values rti, obtained from the design method under
consideration, with the experimental (or numerical) values rei, for each specimen as listed
hereafter and illustrated in Figure 7.
Criterion 1: The percentage of the theoretical resistance values rti on the unsafe side by more
than 15% of the experimental (or numerical) values rei i.e. rti > 1.15rei, which should be zero.
Criterion 2: The percentage of the theoretical resistance values rti on the unsafe side i.e. rti >
1.0 rei, which should be less than 20%.
Criterion 3: The mean value of all percentage difference between the theoretical resistance
values rti and the experimental (or numerical) values rei which should be on the safe side and
less than zero.
Table 6 provides a summary of the safety assessment results for the predicted resistances
from the design methods in EN 1993-1-2 [2], Design Manual for Structural Stainless Steel [4]
and the proposed method, where it is shown that the three reliability criteria are satisfied by
the proposed method.

rti
(1) rti/rei ≤ 1.15
Unsafe

(2) Max number

of unsafe
results

(3) ∑ rti/rei ≤ 1.0

45° Safe
rei

Figure 7: Schematic representation of the reliability criteria set out by Kuppra [21].

14
 A. MOHAMMED and S. AFSHAN

Table 6: Summary of reliability results.

Material Criterion EN 1993-1-2 [2] DMSS [4] Proposed
Criterion 1 41.7% Fail 5.1% Fail 0.0% Pass
Austenitic Criterion 2 76.9% Fail 41.7% Fail 7.7% Pass
Criterion 3 0.131 Fail -0.032 Pass -0.099 Pass
Criterion 1 30.0% Fail 11.9% Fail 0.0% Pass
Duplex Criterion 2 70.0% Fail 35.0% Fail 15.0% Pass
Criterion 3 0.115 Fail -0.020 Pass -0.041 Pass
Criterion 1 9.0% Fail 0.0% Pass 0.0% Pass
Ferritic Criterion 2 31.9% Fail 30.9% Fail 12.2% Pass
Criterion 3 -0.019 Pass -0.036 Pass -0.062 Pass

7 CONCLUSIONS
A numerical modelling study was performed to investigate the flexural buckling response
of stainless steel square and rectangular hollow section columns in fire. The developed finite
element models were initially validated against the fire column tests reported in the literature
and subsequently used to perform parametric studies and to generate structural performance
data. The FE generated flexural buckling data were used to carry out an assessment of the
design methods provided in EN 1993-1-2 [2] and the Design Manual for Stainless Steel
Structures [4], where it was shown that both methods result in inaccurate and unsafe
prediction of the flexural buckling resistance of stainless steel columns in fire. New buckling
curves for cold-formed stainless steel square and rectangular hollow section columns in fire
were proposed on the basis of the FE results, which were shown to give a more accurate and
less scattered prediction of the flexural buckling capacity of cold-formed stainless steel
SHS/RHS columns in fire. The suitability of the proposed buckling curves was confirmed by
means of reliability criteria set out by Kruppa [21].

REFERENCES
[1] Gardner, L. ‘Stainless steel structures in fire’. Proceedings of the ICE - Structures and
Buildings. 160, 129–38, 2007.
[2] EN 1993-1-2. Eurocode 3: Design of steel structures - Part 1-2: General rules - Structural fire
design. Brussels: European Committee for Standardization 2005.
[3] EN 1993-1-4. Eurocode 3: Design of steel structures - Part 1-4: General rules for stainless
steels. Brussels: European Committee for Standardization. 2006.
[4] SCI (Steel Construction Institute). Design Manual For Structural Stainless Steel. Fourth
Edition. 2017.
[5] ABAQUS. ABAQUS 2016, Dassault Systmes Simulia Corp. USA 2016.
[6] Gardner. L, Baddoo. NR. ‘Fire testing and design of stainless steel structures’. Journal of
Constructional Steel Research. 62, 532–43, 2006.
[7] Ala-Outinen, T., Oksanen, T. ‘Stainless steel compression members exposed to fire’. VTT
research notes 1864. Espoo (Finland) 1997.
[8] Tondini, N., Rossi, B., Franssen, JM. ‘Experimental investigation on ferritic stainless steel
columns in fire’. Fire Safety Journal. 62, 238–48, 2013.
[9] Ng, KT., Gardner, L. ‘Buckling of stainless steel columns and beams in fire’. Engineering
Structures. 29, 717–30, 2007.

15
 A. MOHAMMED and S. AFSHAN

[10] Gardner, L., Ng, KT. ‘Temperature development in structural stainless steel sections exposed to
fire’. Fire Safety Journal. 41, 185-203, 2006.
[11] Gardner, L., Nethercot, DA. ‘Numerical Modeling of Stainless Steel Structural Components—A
Consistent Approach’. Journal Structural Engineering. 130, 1586–601, 2004.
[12] Cruise, RB., Gardner, L. ‘Strength enhancements induced during cold forming of stainless steel
sections’. Journal of Constructional Steel Research. 64, 1310–6, 2008.
[13] Afshan, S. Structural Behaviour of Cold-Formed Stainless Steel Tubular Members. Ph.D. thesis,
Imperial College London. 2013.
[14] Dawson, RG., Walker, AC. ‘Post-Buckling of Geometrically Imperfect Plates’. Journal of
Structural Division. 98, 75–94, 1972.
[15] Cruise, RB., Gardner, L. ‘Residual stress analysis of structural stainless steel sections’. Journal
of Constructional Steel Research. 64, 352–66, 2008.
[16] EN 1363-1. Fire resistance tests - Part 1: General requirements. Brussels: European Committee
for Standardization 2012.
[17] Afshan S, Zhao O, Gardner L. ‘Standardised material properties for numerical parametric
studies of stainless steel structures and buckling curves for tubular columns’. Journal of
Constructional Steel Research. 152, 2-11, 2019.
[18] EN 1090-2. Execution of steel structures and aluminium structures - Part 2: Technical
requirements for steel structures. Brussels: European Committee for Standardization. 2008.
[19] EN 1993-1-5. Eurocode 3: Design of steel structures - Part 1-5: Plated structural elements. CEN,
Brussels, 2006.
[20] Lopes, N., Vila Real, P., Silva L, S., Franssen, J-M. ‘Axially Loaded Stainless Steel Columns in
Case of Fire’. Journal of Structural Fire Engineering. 1,43–59, 2010.
[21] Kruppa, J. “Eurocodes-Fire parts: Proposal for a methodology to check the accuracy of
assessment methods,” CEN TC 250, Horizontal Group Fire, Document no: 99/130. 1999.

16

View publication stats